Structural integrity of additive manufactured materials and parts 9780803177086, 0803177089

Citation preview

Shamsaei | Seifi

Structural Integrity of Additive Manufactured Materials and Parts STP1631

ASTM INTERNATIONAL Helping our world work better

ASTM INTERNATIONAL Selected Technical Papers

Structural Integrity of Additive Manufactured Materials and Parts STP 1631 Editors: Nima Shamsaei Mohsen Seifi

ISBN: 978-0-8031-7708-6 Stock #: STP1631 www.astm.org .

SELECTED TECHNICAL PAPERS STP1631

Editors: Nima Shamsaei and Mohsen Seifi

Structural Integrity of Additive Manufactured Materials and Parts ASTM STOCK #STP1631 DOI: 10.1520/STP1631-EB

ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959 Printed in the U.S.A.

.

Library of Congress Cataloging-in-Publication Data Names: Shamsaei, Nima, editor. | Seifi, Mohsen, editor. Title: Structural integrity of additive manufactured materials and parts / editors, Nima Shamsaei, Mohsen Seifi. Description: West Conshohocken, PA, USA : ASTM International, [2020] | Series: Selected technical papers ; STP 1631 | Includes bibliographical references. | Summary: “THIS COMPILATION OF Selected Technical Papers, STP1631, Structural Integrity of Additive Manufactured Materials & Parts, contains peer-reviewed papers that were presented at a symposium held October 7–10, 2019, in National Harbor, MD, USA. The symposium was sponsored by ASTM International Additive Manufacturing Center of Excellence and additional ASTM Technical Committees”– Provided by publisher. Identifiers: LCCN 2020032340 (print) | LCCN 2020032341 (ebook) | ISBN 9780803177086 | ISBN 9780803177093 (pdf) Subjects: LCSH: Strength of materials. | Machine parts–Testing. | Structural analysis (Engineering) | Additive manufacturing. Classification: LCC TA405 .S767 2020 (print) | LCC TA405 (ebook) | DDC 620.1/12–dc23 LC record available at https://lccn.loc.gov/2020032340 LC ebook record available at https://lccn.loc.gov/2020032341 ISBN: 978-0-8031-7708-6 C 2020 ASTM INTERNATIONAL, West Conshohocken, PA. All rights reserved. This material Copyright V may not be reproduced or copied, in whole or in part, in any printed, mechanical, electronic, film, or other distribution and storage media, without the written consent of the publisher.

Photocopy Rights Authorization to photocopy items for internal, personal, or educational classroom use, or the internal, personal, or educational classroom use of specific clients, is granted by ASTM International provided that the appropriate fee is paid to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, Tel: (978) 646-2600; http://www.copyright.com/ ASTM International is not responsible, as a body, for the statements and opinions expressed in this publication. ASTM International does not endorse any products represented in this publication. Peer Review Policy Each paper published in this volume was evaluated by two peer reviewers and at least one editor. The authors addressed all of the reviewers’ comments to the satisfaction of both the technical editor(s) and the ASTM International Committee on Publications. The quality of the papers in this publication reflects not only the obvious efforts of the authors and the technical editor(s), but also the work of the peer reviewers. In keeping with long-standing publication practices, ASTM International maintains the anonymity of the peer reviewers. The ASTM International Committee on Publications acknowledges with appreciation their dedication and contribution of time and effort on behalf of ASTM International. Citation of Papers When citing papers from this publication, the appropriate citation includes the paper authors, “paper title,” in STP title, book editor(s) (West Conshohocken, PA: ASTM International, year), page range, paper doi, listed in the footnote of the paper. A citation is provided on page one of each paper. .

Printed in Hanover, PA September, 2020

Foreword THIS COMPILATION OF Selected Technical Papers, STP1631, Structural Integrity of Additive Manufactured Materials and Parts, contains peer-reviewed papers that were presented at a symposium held October 7–10, 2019, in National Harbor, MD, USA. The symposium was sponsored by ASTM International Additive Manufacturing Center of Excellence and additional ASTM Technical Committees. Symposium Chairs and STP Editors: Nima Shamsaei Auburn University Auburn, AL, USA Mohsen Seifi ASTM International Washington, DC, USA

.

iii

.

Contents

Overview

ix

Demonstration of Closed-Loop Control for Laser Powder Bed Fusion (LPBF) David Maass On the Challenges toward Realization of Functionally Graded Structures by Electron Beam Melting—Fe-Base Shape Memory Alloy and Stainless Steel ´ Bauer, Malte Vollmer, and Thomas Niendorf Christof Torrent, Andre

20

Residual Stress Formation in Laser-Based Powder Bed Fusion (PBF-LB) of Invar 36 Mostafa Yakout and Mohamed A. Elbestawi

34

Origin of Oxides and Oxide-Related Pores in Laser Powder Bed Fusion Parts Tomio Ohtsuki, Lonnie Smith, Ming Tang, and P. Chris Pistorius

45

Toward Understanding the Role of Surface Texture for Additively Manufactured Metal Parts Adam J. Brooks, Arushi Dhakad, Agustin Diaz, and Daniel Kowalik Optimizing X-Ray Computed Tomography Settings for Dimensional Metrology Using 2D Image Analysis Younes Chahid, Andrew Townsend, Alexander Liu, Paul Bills, Philip Sperling, and Radu Racasan Challenges in Inspecting Internal Features for SLM Additive Manufactured Build Artifacts Ahmed Tawfik, Radu Racasan, Desi Bacheva, Liam Blunt, Andre´ Beerlink, and Paul Bills A Critical Discussion on the Diffraction-Based Experimental Determination of Residual Stress in AM Parts ˜oz, Tobias Fritsch, Alexander Ulbricht, Tatiana Mishurova, Itziar Serrano-Mun Maximilian Sprengel, Alexander Evans, Arne Kromm, Mauro Madia, and Giovanni Bruno

.

1

v

61

88

102

122

X-Ray Computed Tomography Inspection in Metal Additive Manufacturing: The Role of Witness Specimens Anton du Plessis, Jess M. Waller, Stephan G. le Roux, Ina Yadroitsava, Igor Yadroitsev, Johan Els, and Jacobus Prinsloo Perspective on Nondestructive Evaluation of Additive Manufactured Components Eric Lindgren and Bryce Jolley Prediction of Residual Stress Evolution for End-To-End Process Chain of Laser Powder Bed Fusion Process and Determination of Fatigue S-N Curves Shukri Afazov, Jamie Frame, Utkarsha Ankalkhope, Prveen Bidare, Yijun Liu, Wilson Vesga, and Ben Dutton Design of Coupons and Test Methodology for Orthotropic Characterization of FFF-Processed Ultem 9085 Tommy Hyatt, Richard Martin, and Rich Fields Intrinsic Threshold Stress Intensity of Additive Manufactured Metals R. Sunder, Ramesh Koraddi, and Andrei Gorunov A Multiscale Material Modeling Approach to Predict the Mechanical Properties of Powder Bed Fusion (PBF) Metal Yang Li, Hongyi Xu, Wei-Jen Lai, Ziang Li, and Xuming Su Alternate Method for Determining Yield Strength of Polymer Additive Manufacturing Chul Y. Park, Keith E. Rupel, Chelsey E. Henry, Kevin F. Malik, Sayata Ghose, and Upul R. Palliyaguru Effects of Surface Roughness and Porosity on Fatigue Behavior of AlSi10Mg Produced by Laser Powder Bed Fusion Process Wei-Jen Lai, Ziang Li, Avinesh Ojha, Yang Li, Joy Forsmark, Carlos Engler, and Xuming Su

139

157

165

176

188

203

214

229

Characterization of Functionally Graded Materials Based on Inconel 718 and Stainless Steel 316L Manufactured by DED Process 247 ´, Jaroslav Vavrˇ´k, Jan Dzˇugan, Daniel Melzer, Martina Koukolı´kova ı and Mohsen Seifi Fretting Fatigue Characterization in Press-Fit Joints of AM Parts by X-Ray Tomography and Digital Image Correlation Inigo Bacaicoa, Sascha Horn, Angelika Brueckner-Foit, Julia Richter, and Thomas Niendorf

.

257

Effect of Microstructure and Internal Defects on the Mechanical Properties of Ti6Al4V Gyroid Lattice Structures for Biomedical Implants Dalia Mahmoud, Mohamed A. Elbestawi, and Kassim S. Al-Rubaie

271

Full-Scale High-Load, Thermal, and Fatigue Testing of Additive Manufactured Powder Bed Fusion Component for Oil Field Applications Matthew Wayne Sanders, Adam Rowe, and Suresh Divi

289

vi

Dynamic Compression Response of Porous Zirconium-Based Bulk Metallic Glass (Zr41Ti14Cu12.5Ni10Be22.5) Honeycomb: A Numerical Study Nand Kishore Singh, Shashi Kant Kumar, Satish K. S. N. Idury, K. K. Singh, and Ratneshwar Jha Preclinical Testing of a Novel, Additive-Manufactured, Three-Dimensional Porous Titanium Structure Erik Woodard, Zach Post, and Mark Morrison

322

Investigation of Microstructure and Mechanical Properties of SLM-Produced Inconel 718 and Hastelloy-X Alloys Guney Mert Bilgin, Cansinem Tuzemen, Cemre Tigli, and Yesim Nur Gulcan

340

Analysis of Data Streams for Qualification and Certification of Inconel 738LC Airfoils Processed through Electron Beam Melting Michael Kirka, Derek Rose, William Halsey, Amirkoushyar Ziabari, Vincent Paquit, Daniel Ryan, and Paul Brackman

.

308

vii

352

.

Overview To ensure the structural integrity of additively manufactured (AM) parts, there is a need for establishing feedstock-process-structure-property-performance relationships, specifically where these components and structures are being used in safety critical applications. Therefore, the Fourth ASTM International Symposium on the Structural Integrity of Additive Manufactured Materials and Parts took place from October 7–10, 2019, in the Washington, DC, area in coordination with partners from government, industry, and academia, to provide a forum for the exchange of ideas regarding the structural integrity of parts fabricated using additive manufacturing with a focus on the lack of industry standards, design principles, as well as qualification and certification challenges. The first event of this kind was sponsored by ASTM International Committee E08 as a workshop in San Antonio in May 2016. The second event was sponsored by Committees E08 and F42 as a symposium in Atlanta in November 2017. The third event was held in Washington, DC and sponsored by Committees F42, E08, and E07 in November 2018. After the creation of ASTM International Additive Manufacturing Center of Excellence (AM CoE) in early 2018 and the growth of the AM industry, the effort has been centralized and is now being led and sponsored by the AM CoE, involving several co-sponsoring ASTM technical committees, including: B09 on Metal Powders and Metal Powder Products; D20 on Plastics; D30 on Composite Materials; E07 on Nondestructive Testing; E08 on Fatigue and Fracture; F04 on Medical and Surgical Materials and Devices; and F42 on Additive Manufacturing Technologies. In addition, the fourth event was supported by the National Aeronautics and Space Administration (NASA), National Institute of Standards and Technology (NIST), European Structural Integrity Society (ESIS) TC15, and the Food and Drug Administration. This event continues to receive interest from the international community and has been growing drastically, becoming the main gathering around standardization, qualification, and certification for additively manufactured materials and parts. This year we are pleased to report that we have 133 talks, including 44 invited speakers, 10 posters, 25 student competition talks, 3 panel discussions, and over 300 attendees, representing industry, academia, and government from more than 30 countries. There are 24 selected technical papers (STPs) in this collection that will provide an opportunity to delve more deeply into the wide variety of topics covered during the symposium. While individual papers often touch on multiple topics, the

.

ix

following list presents the major areas covered in the symposium and the papers that address each topic as their primary focus. A significant emphasis of this symposium was on nondestructive evaluation in situ monitoring and process control of AM parts, and the below articles are in this category:      

Optimizing X-Ray Computed Tomography Settings for Dimensional Metrology Using 2D Image Analysis Challenges in Inspecting Internal Features for SLM Additive Manufactured Build Artifacts A Critical Discussion on the Diffraction-Based Experimental Determination of Residual Stress in AM Parts X-Ray Computed Tomography Inspection in Metal Additive Manufacturing: The Role of Witness Specimens Perspective on Nondestructive Evaluation of Additive Manufactured Components Demonstration of Closed-Loop Control for Laser Powder Bed Fusion (LPBF)

As the lack of understanding of the factors affecting the structural integrity of AM parts and the related standards is still a major roadblock against the adoption of AM in load-bearing, safety critical applications, many papers related to fatigue, fracture, tensile, and corrosion behavior of materials and parts fabricated using AM were presented. Effects of design, process, and post-process parameters on fatigue and fracture properties as well as process optimization to improve structural integrity of AM parts were also discussed extensively. Applicability of existing test methods to AM materials and parts and innovative approaches to standardization were specifically emphasized in this symposium. The following papers focus on related topics for this category:        

.

Prediction of Residual Stress Evolution for End-To-End Process Chain of Laser Powder Bed Fusion Process and Determination of Fatigue S-N Curves Design of Coupons and Test Methodology for Orthotropic Characterization of FFF-Processed Ultem 9085 Intrinsic Threshold Stress Intensity of Additive Manufactured Metals A Multiscale Material Modeling Approach to Predict the Mechanical Properties of Powder Bed Fusion (PBF) Metal Alternate Method for Determining Yield Strength of Polymer Additive Manufacturing Effects of Surface Roughness and Porosity on Fatigue Behaviors of AlSi10Mg Produced by Laser Power Bed Fusion Process Characterization of Functionally Graded Materials Based on Inconel 718 and Stainless Steel 316L Manufactured by DED Process Fretting Fatigue Characterization in Press-Fit Joints of AM Parts by X-Ray Tomography and Digital Image Correlation

x

        

Effect of Microstructure and Internal Defects on the Mechanical Properties of Ti6Al4V Gyroid Lattice Structures for Biomedical Implants Full-Scale High-Load, Thermal, and Fatigue Testing of Additive Manufactured Powder Bed Fusion Component for Oil Field Applications Dynamic Compression Response of Porous Zirconium-Based Bulk Metallic Glass (Zr41Ti14Cu12.5Ni10Be22.5) Honeycomb: A Numerical Study Preclinical Testing of a Novel, Additive-Manufactured, Three-Dimensional Porous Titanium Structure Investigation of Microstructure and Mechanical Properties of SLM-Produced Inconel 718 and Hastelloy-X Alloys Residual Stress Formation in Laser-Based Powder Bed Fusion (PBF-LB) of Invar 36 On the Challenges toward Realization of Functionally Graded Structures by Electron Beam Melting—Fe-Base Shape Memory Alloy and Stainless Steel Origin of Oxides and Oxide-Related Pores in Laser Powder Bed Fusion Parts Toward Understanding the Role of Surface Texture for Additively Manufactured Metal Parts

Finally, recent advances in standardization, qualification, and certification were discussed. The following paper provides a snapshot of such discussions: 

Analysis of Data Streams for Qualification and Certification of Inconel 738LC Airfoils Processed through Electron Beam Melting

Following the successes of these events, we are excited to announce that starting in 2020, the event will be organized as the ASTM International Conference of Additive Manufacturing (ASTM ICAM) with a wider scope, while still focusing on standardization, qualification, and certification. This will be a major event involving additional ASTM committees and external stakeholders from the international community, such as NIST, FAA, NASA, FDA, America Makes, ESIS, ESA, EASA, CECIMO, and more, setting the stage to bring experts from all around the world to exchange and share the latest development in the field of additive manufacturing. We would like to extend our gratitude to everyone who made this symposium possible. The hard work of the authors, co-authors, our symposium co-chairs, peer reviewers, ASTM personnel, ASTM Committees and ASTM AM CoE partners, as well as the support of NASA, NIST, FAA, FDA, and ESIS TC15, all played a major role in the success of the symposium and this publication. Nima Shamsaei Mohsen Seifi STP Editors

.

xi

.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190130

David Maass1

Demonstration of Closed-Loop Control for Laser Powder Bed Fusion (LPBF) Citation D. Maass, “Demonstration of Closed-Loop Control for Laser Powder Bed Fusion (LPBF),” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 1–19. http://doi.org/10.1520/ STP1631201901302

ABSTRACT

Currently, AM processes such as LPBF are performed open loop, using a fixed, preprogrammed definition of material deposition (path, speed, laser power, and so on). Actual layer and part formation details, even when measured, are not fed back to the print controller to account for actual, as-made layer conditions. Unanticipated layer and part deviations occur frequently that, in the worst case, can result in print failure, part rejection, higher scrap rate, lower yield, and more expensive parts. Process anomalies are sometimes detected manually by the operator. In-process inspection methods such as melt pool monitoring typically do not provide accept/reject guidance. When anomalies are noted, no instructions are provided to the operator or the machine to repair or compensate for the flaw and to salvage the build, in cases where this is possible. We demonstrate development of a closed-loop control capability using a nonthermal in-process inspection method on every layer. Layer Topographic Mapping (LTM) is an in-process inspection method using an optical profilometer to generate a dense, precise map of layer surface height. Algorithms process this data to detect melt flaws with excellent performance. Demonstrated detection of lack of fusion flaws in more than 1,800 Inconel 625 layers is 98.2% probability of detection (POD) and 1.0% probability of false detection (POFD). Optimum

Manuscript received October 31, 2019; accepted for publication January 30, 2020. 1 Flightware, Inc., 829 Podunk Rd., Guilford, CT 06437, USA 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

1

2

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

repair/rework processes were developed for lack of fusion flaw regions one to three layers thick. LTM software was modified to not only detect flaws but also to define the optimum repair process to employ upon detection based on the number of flaw layers present. Intentionally created (or seeded) lack of fusion flaws were restored to less than 0.1% porosity for one- to three-layer flaws. Porosity in the flaw regions was reduced by up to 98% as verified by CT scans. Keywords closed-loop control, defect repair, flaws, defects, lack of fusion, laser powder based fusion (LPBF), in situ inspection, in-process inspection, surface profilometry, additive manufacturing

Introduction Laser powder bed fusion (LPBF) is the most widely used additive manufacturing (AM) process to produce high-quality metal parts from raw powder.1 The quality of the formed part is typically only determined after final inspection, using methods such as computed tomography (CT) to locate small internal defects such as lack of fusion, porosity, crack, inclusion, and other flaws. This approach is both expensive and inefficient. LPBF is a relatively slow and expensive process compared with more traditional fabrication methods. However, if flaws produced early in the build process are only discovered after the entire part is completed and inspected, significant machine time and expense can be expended (i.e., wasted) producing a part that ultimately must be rejected. A more efficient inspection approach is to perform intermediate inspections, called in situ or in-process inspections, to detect, identify, and disposition flaws in the melt layer as soon as they are created. If the flaw cannot be repaired or reworked, the part build is terminated immediately, thereby eliminating the additional costs of completing the part and subsequent final inspection. However, some types of LPBF flaws can be repaired, such as lack of fusion. This condition occurs when insufficient laser energy is applied to the powder, resulting in less than 100% melting and fusing, thereby creating melt regions that contain minute porosity between particles. The reapplication of sufficient laser energy to the lack of fusion region can remelt and reflow the material, eliminating the lack of fusion defect. Parts that contain repairable flaws can be salvaged (i.e., restored to an acceptable condition) provided the flaw is detected when it is created (i.e., at the layer level) and the repair is performed at such time, when the flaw is accessible. This can be considered a form of closed-loop control applied at the layer level for LPBF. The work described herein consists of two elements required to implement this approach: 1. Development and demonstration of an in situ inspection method performed after the creation of every melt layer. This method must be able to demonstrate a high level of sensitivity and reliability for flaw detection. .

MAASS, DOI: 10.1520/STP163120190130

2. Development and demonstration of a closed-loop control method to restore melt layers with lack of fusion flaws to an acceptable quality condition.

Surface Profilometry Data Acquisition AS-FORMED QUALITY METRIC

We use the surface profile (height) of the as-formed melt layer to determine melt layer quality because (a) it is easily measured, and (b) variation of the surface profile, also called surface texture or roughness, can provide useful insight into the quality of the metal material. For example, figure 1 depicts the surface texture in camera images for three different process conditions of an LPBF Inconel 625 melt layer: (a) high laser fluence conducive to keyhole formation, (b) nominal process laser fluence that produces low porosity, and (c) low laser fluence that can lead to lack of fusion. The texture variation is apparent to the naked eye, which suggests that more sensitive data processing algorithms may more reliably detect flaws and associated subtleties. SENSOR

The surface height of the melt layer is measured using a commercial off-the-shelf (COTS) laser line profilometer such as shown in figure 2. Sensor characteristics of the Keyence LJ-7060K device are presented in table 1. BED SURFACE PROFILOMETRY

The sensor is hard mounted to the powder recoater arm in a purpose-built LPBF printer developed by EWI, an independent engineering consultancy with advanced manufacturing technology resources dedicated specifically to metal additive manufacturing, called the open architecture test bed2 as seen in figure 3. While the recoater spreads a fresh layer of powder for a new layer, the sensor acquires profile line data of the melt layer just created, resulting in a melt surface scan with high point density. Measurement point spacing in the direction of the sensor line is fixed

FIG. 1 Surface texture variations in melt layers processed under several process conditions: (A) high laser fluence, (B) nominal laser fluence, and (C) low laser fluence.

.

3

4

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 Line profilometer captures the height of 800 closely spaced surface points along the projected laser line profile.

TABLE 1 Line profilometer sensor parameters

Specification

Units

Line length, nominal

mm

14

–

800

Points measured

Value

Point spacing

:m

20

Z repeatability

:m

0.4

X repeatability

:m

5

Wavelength

nm

405 (blue)

Communication interface

–

1000BASET/100BASETX

Transmission rate (max)

Mb/s

1,000

at 20 lm while the line-to-line spacing parallel to the recoater motion depends upon the profile acquisition frequency and the recoater velocity. Data were acquired at both 7.7 and 20 lm line-to-line spacing, corresponding to point densities of 2,500 to about 6,500 points per square millimeter. One of the benefits of this approach is that measurement occurs simultaneously with powder recoating, thereby adding no inspection time delay to the normal LPBF process. .

MAASS, DOI: 10.1520/STP163120190130

FIG. 3 The profilometer sensor is mounted to the powder recoater such that a profile scan is created due to recoater motion simultaneous with powder spreading.

FIG. 4 Texture aberration algorithm processes surface profile data (in the form of an image) to detect regions of texture variation.3

LAYER TOPOGRAPHIC MAPPING (LTM) IN SITU INSPECTION ALGORITHM

Profilometry data are processed using a technique we call Layer Topographic Mapping (LTM) using a texture aberration method developed by Timm and Barth,3 as shown conceptually in figure 4. The procedure includes: • Segmentation of a layer profile image into many smaller subimages or “windows” representing small regions of the layer .

5

6

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Plot of all window Weibull parameters for a single layer.

Calculation of the Gaussian image gradient values within each window Determination of Weibull probability distribution of Gaussian image gradient values for each window, defined by two probability density function (PDF) values: shape and scale, which are sometimes referred to as k and k4 • Plotting the shape and scale PDF values for each window on a common layer plot, as shown in figure 5 The LTM plot shown in figure 5 includes overlays of many nominal quality layers, and it is apparent that most of the X,Y values tend to cluster in an elliptical region in the center of the plot. Using statistical methods to determine inliers and outliers, inlier points are marked dark gray while outliers are marked light gray. The dashed ellipses indicate the outlier boundary. The points (layer regions) of any newly created layer are similarly processed and each point is marked black if within the boundary (i.e., a nominal quality region) or red if outside the boundary. In the example shown in figure 6A, the majority of the current layer regions are nominal (black points), with only a few possible flaws or red points. However, other layer LTMs can be different, such as shown in figure 6B. Here, the majority of the current layer regions points are red and likely are flaws. In fact, this scan was derived from data for a layer that had been intentionally produced with process conditions known to create lack of fusion flaws. • •

Test Coupon Fabrication Test coupons were fabricated containing nominal and seeded flaw layers (a) to determine how reliably flaws could be detected using the LTM algorithm, (b) to .

MAASS, DOI: 10.1520/STP163120190130

FIG. 6 LTM plot from a layer made with known lack of fusion process parameters: (A) LTM map of a nominal layer, and (B) LTM map of a flawed layer.

develop optimum repair processes capable of healing or repairing intentionally created flaws, and (c) to validate how effective the automated detect and repair processes performed, as described in the following sections. MATERIAL

The material used was a nickel-based superalloy, Inconel 625, often used in gas turbines. The material was virgin, with spherical particles from 15 to 45 lm in diameter with a Gaussian size distribution. PROCESS PARAMETERS

LPBF process conditions used are defined on table 2. This included nominal parameters, known from prior experience to generate high-quality material on this printer, lack of fusion or LOF parameters known to generate lack of fusion and porosity, and three repair processes. The repair process conditions included a range of laser energy flux designed to find what conditions were most effective at healing or repairing LOF in an exposed melt layer.

TABLE 2 LPBF process parameters used

Condition Process Parameter

Nominal

LOF

Repair Low

Repair Moderate

Repair High

Laser power

watts

280

100

200

270

300

Scan speed

mm/s

960

1,000

900

1,000

1,000

Spot size

:m

75

75

75

75

75

Hatch

:m

100

100

100

100

90

distance .

Units

7

8

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

A heated build plate held at 80 C was used as well as an argon atmosphere, with gas monitoring to assure that oxygen content remained below 500 ppm. GENERATION OF SEEDED FLAWS

We focused on LOF flaws only because these flaws offered good promise of effective in situ repair. Each seeded flaw consisted of the entire layer (i.e., 100% of the coupon layer area) made with process conditions known to create LOF for simplicity. It is understood that to be practical, ultimately specific layer regions containing flaws such as LOF need to be detected and repaired. Initially flaws were defined and created containing either one, three, five, seven, or nine consecutive layers with lack-of-fusion conditions to bracket the level of defect we might detect and repair. We later scaled back to one, two, or three successive LOF layers because more than three layers created excessive porosity that was unlikely to be repairable. The purpose of using one, two, or three successive LOF layers was to create varying levels of LOF porosity using a single LOF process parameter set. As discussed later in this paper, the closed-loop repair activity was only performed on the top layer of the LOF flaw region (i.e., on the third LOF layer if a three-layer flaw region) to determine if the defined repair process could remove or repair the level of LOF porosity known to exist in this condition. FLAW REPAIR PROCESS DEFINITION

Six process conditions were evaluated for LOF repair to determine which process provided the most LOF porosity reduction for one, two, or three successive LOF layer flaws. Combinations of two repair process variables were evaluated including: • Laser power level (three): Lower (L), moderate (M), and higher (H) than nominal laser power settings • Powder recoat (two): Intermediary (I) process, with no additional powder recoat step prior to repair and subsequent (S), where an additional powder recoat step was performed as part of the repair process FLAW COUPON CONFIGURATION

Test coupons were produced as seen in figure 7 and described in table 3. LOF flaws were intentionally seeded throughout the thickness of each coupon, with at least 50 layers between such flaw regions in each coupon to ensure that process conditions used to create one flaw were thermally isolated from flaw regions above or below the current region. Depending upon coupon height, each coupon contained five or six flaw regions. As previously noted, a flaw region contained either one, two, or three successive LOF layers. A fiducial marking system was used to allow correlation of CT scan results with known locations of the seeded flaws. Coupon layers were defined with a tapering width in the flaw region such that the first layer containing an LOF flaw .

MAASS, DOI: 10.1520/STP163120190130

FIG. 7 Test coupons were produced with nominal and lack-of-fusion flaw regions.

TABLE 3 Test coupon parameters

Item

Material Layer thickness Powder particle diameter

Value

Inconel 625 40 :m 20 :m nominal

Planform dimensions

20 mm  20 mm

Coupon height

Up to 464 layers

Flaw-to-flaw separation

30 layers

occurred on the minimum width layer, as shown in figure 8. This later allowed determination of the build layer number (from geometry information present in the CT data) independent of porosity information in the CT data.

Validation of Flaws via CT Scan CT scans were performed for all coupons as described in table 4. CT POROSITY DATA PROCESSING

Raw CT data were processed using Thermo Fisher Aviso software. Two parameters were extracted for every layer (i.e., every 16.5 lm in the Z direction) of every coupon, the layer area and porosity as a percentage of the total layer area. A plot of layer area in the Z direction enabled determination of the fiducial mark where the first LOF flaw layer was created within that flaw group. Porosity was determined by thresholding voxel attenuation to segregate Inconel material from no material present. However, these data were first filtered using adaptive thresholding to eliminate CT noise and ring artifacts. .

9

10

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 8 The minimum width layer was a fiducial mark to define the first LOF layer in a flaw region for later CT data analysis (not to scale): (A) View of AA one-layer LOF flaw, and (B) View of AA three-layer LOF flaw.

TABLE 4 CT scan parameters

Item

CT machine

Description

Nikon XTH 225ST

X-ray source

225 kV

Voxel resolution

16.5 :m

Automated Flaw Detection and Repair Demonstration (Closed-Loop Control) Today, most LPBF is primarily performed open loop, meaning there is no active measurement and feedback of the melt quality state used to alter process conditions. Establishing closed-loop control is an active area of research.5–7 In this program, a closed-loop control architecture was implemented, as shown in figure 9. After layer melt was performed, in situ inspection was performed and a determination was made for every coupon layer in the build as to whether that layer was acceptable (i.e., of nominal quality) or flawed based solely on the in situ inspection algorithm. If the layer was nominal, the programmed build plan was incremented to produce the next layer. However, if one or more coupon layers were determined .

MAASS, DOI: 10.1520/STP163120190130

FIG. 9 Closed-loop control architecture was implemented and demonstrated.

as a flaw, the software determined if it was a one-, two-, or three-layer flaw using the known build plan utilized to create these flaws. The software then defined the appropriate repair process to perform based on earlier trials in which the optimum repair for one-, two-, and three-layer flaws had been determined. As noted previously in the section on the generation of seeded flaws, although LOF flaw regions were created with as many as three separate, successive layers, our goal was to create increasing levels of LOF porosity, which succeeded as noted here. The control architecture was written to recognize if a detected flaw layer was part of a one-, two-, or three-layer flaw region, based on a priori knowledge of where the LOF flaw layers had been programmed. We programmed the control architecture to only repair the topmost layer, to determine if that level of LOF porosity could be reduced or eliminated by the repair. For example, consider the case of three-layer flaw. In situ inspection detected the first LOF layer (i.e., flaw detection was not based on the build plan). This was compared with the build plan, which indicated that this was the first layer of an intentional three-layer flaw. Therefore, the repair process was deliberately not performed at this time. This sequence was repeated for the second layer of the three-layer flaw. On the third layer, the control system recognized that this was the topmost layer of the flaw region (with the highest level of porosity) and at this point the repair procedure was commanded. In this manner, we could evaluate effectiveness of the repair procedures used for a range of LOF porosity. In an actual operating environment with closed-loop control, repairs would be performed as soon as a single flaw layer was detected, to prevent the formation of higher levels of porosity. In a fully automated system, these determinations would be communicated directly to the printer controller. For simplicity in this proof-of-concept demonstration, the software defined the proper repair process to apply for the flawed coupons in that layer on a screen. The operator then selected this process to be used only on .

11

12

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

the flawed coupons through a manual entry to the controller, where the controller software had been modified to enable this out-of-sequence operation.

Results and Discussion LOF FLAW CREATION

An isometric view of the porosity present in a test coupon CT scan is shown in figure 10. It can be seen from the density of individual pores that the greater the number of successive pore layers, the greater the porosity, as expected. An exception, however, occurs for the nine-layer LOF flow, where porosity was so high the coupon literally delaminated at this location, providing an inaccurate porosity reading. In fact, porosity levels in the five- and seven-layer LOF flaws were also very high, so we restricted further work to one-, two-, and three-layer LOF flaws only. As seen in figure 11, porosity in the control coupon (no seeded flaws, nominal melt conditions only) was 0.16%, which is typical of aerospace-grade LPBF alloys. When LOF conditions were used, porosity conditions ranged from 0.76% for a single LOF layer to 14.21% for three successive LOF layers. These values are averages from two to three replicate flaw regions.

FIG. 10 CT scan of coupon with LOF flaws comprised of one to nine successive LOF layers.

.

MAASS, DOI: 10.1520/STP163120190130

FIG. 11 Porosity as a function of number of flaw layers.

Porosity was not limited to the LOF layer only, as seen in figure 12. Porosity extended both below and above the intentional flaw layer(s), as would be expected due to layer remelting that occurs below each newly formed layer. LTM FLAW DETECTION ACCURACY

Initial LTM Algorithm

In the demonstration, four coupons were produced, comprised of 1,831 layers total (average of 458 layers each, or about 18.3-mm-tall each). The distribution of layers made with nominal conditions (no flaw) and LOF flaw conditions is shown in table 5. Results of the LTM detection algorithm were then determined based on the build plan location of the seeded flaws. This is conservative in the sense that some flaw indications were classified as false simply because they had not been intentionally programmed. Accuracy of the LTM algorithm was determined using the metrics in equations (1)–(3): POD ¼ POMD ¼

TP ðTP þ FN Þ

FN ¼ 1  POD ðTP þ FN Þ

(2)

FP ðTP þ FN Þ

(3)

POFD ¼

.

(1)

13

14

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 12 Porosity extends below and above the seeded LOF layers.

TABLE 5 Layers produced in four demonstration test coupons.

Detection Status

No Flaw (Negative)

LOF Flaw (Positive)

Total

Detection opportunities

1,776

55

1,831

True (correct)

1,758

54

1,812

1

18

19

False (incorrect)

where: POD ¼ probability of detection, POMD ¼ probability of missed detection (“missed”), POFD ¼ probability of false detection, TP ¼ true positive, FN ¼ false negative, and FP ¼ false positive. Based on the results in table 5, LTM detection performance was calculated as shown in figure 13. Aerospace industry standards require that nondestructive inspection (NDI) procedures used to detect flaws of a certain type and size in primary structural elements must demonstrate a minimum 90% probability with 95% confidence.8,9 The number of flaw layers is relatively small and significantly smaller than the number of nominal layers. While the statistical basis to estimate confidence bands was not performed here, initial results are considered promising. .

MAASS, DOI: 10.1520/STP163120190130

FIG. 13 LTM algorithm detection performance.

Preliminary Machine Learning Approach

We also performed a preliminary study using machine learning to leverage the profile data collected. LTM results presented here consisted only of whole layer LOF flaws (55 layers total) and nominal layers (1,758 total layers). For the machine-learning task, we subdivided profile maps into much smaller regions, each 1.6 mm by 1.6 mm. This was performed both for nominal layers and LOF flaw layers, generating 1,000 data sets (or profile images) of each type. These labeled images (nominal or flawed) were used to train a neural network using Google’s AutoML cloud-based service. A total of 1,358 independent labeled test images were then used to assess the performance of the neural network, with results presented in table 6. The probability of detection (POD) for the LOF flaw was 93% while the probability of false detection (POFD) was 4%. While these results are slightly less accurate than shown for the LTM algorithm in the previous section, these results are nevertheless very encouraging. This was achieved with a relatively small investment of time and CPU resources, and there are many ways in which this approach can be refined and improved. .

15

16

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 6 Confusion matrix and LOF flaw prediction accuracy

Predicted Label True Label

Nominal

LOF Flaw

Nominal

98%

2%

LOF flaw

4%

93%

LOF FLAW REPAIR PROCESSES DEVELOPED

Six different repair processes were performed on LOF flaws regions comprised of one, two, and three successive LOF layers. CT-derived porosity results for the postrepaired coupons containing one- and three-layer LOF flaws are shown in figures 14 and 15, respectively. Porosity levels in repaired flaws comprised of more than three successive LOF layers were very high; therefore, this level of LOF was determined not repairable. In these plots, the bright green bar represents the optimum of the six repair processes, with the largest reduction in residual porosity after repair, while the red bar indicates the poorest repair (i.e., least porosity reduction). In fact, for the singlelayer flaw repair, one of these processes actually increased porosity after repair.

FIG. 14 Porosity reduction for one-layer LOF flaw after repair.

.

MAASS, DOI: 10.1520/STP163120190130

FIG. 15 Porosity reduction for three-layer LOF flaw after repair.

Note that the best-performing repair process was different for a one-layer and three-layer flaw. The “SL” repair (subsequent powder recoat performed at low power) performed best for the one-layer LOF flaw, while the “IH” repair (no subsequent powder recoat performed at high power) performed best for the three-layer LOF flaw. CLOSED-LOOP CONTROL DEMONSTRATION RESULTS

Four coupons were produced in the closed-loop control demonstration, and each contained six LOF flaw regions comprising one-, two-, and three-layer LOF flaws, for a total of 24 flaws. These flaws were detected using only the LTM algorithm, and the software defined the repair process to use when a flaw was detected. Repairs were made in situ; the coupons were completed and then CT scanned and analyzed for residual porosity. Residual porosity after repair is presented in figure 16 (blue bars) as well as the as-formed porosity (red bars) from a control coupon as previously shown in figure 11. Note that the reported porosity levels are the average of two to three replicates for each flaw condition. For all flaw levels (one-, two-, and three-layer flaws), the repairs almost completely eliminated the as-formed porosity. The residual porosity after repair was essentially equivalent to the nominal or virgin material. .

17

18

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 16 Residual porosity before and after in situ repair using closed-loop control.

Conclusions This work has demonstrated: 1. An in situ inspection procedure was developed called LTM that relies upon a direct measurement of the as-formed melt layer quality, as compared with other methods that measure process parameters during melt formation. 2. The LTM method exhibits very good reliability detecting LOF melt flaws, with high probability of detection and low probability of false detection. 3. A neural network was also developed to assess the accuracy of using profile data to detect LOF flaws. While not quite as accurate as the LTM method, these results were still quite good and very encouraging. 4. Lack of fusion for three successive layers created porosity of more than 14%. 5. Specific repair processes were developed and optimized to minimize or eliminate LOF porosity for one to three successive layers. 6. Different LOF flaws (i.e., degree of porosity) have different optimum repair processes. 7. In situ repair of LOF porosity can be accomplished by closed-loop control. While this demonstration was semiautomated, fully automated operation can easily be envisioned. 8. LOF porosity of up to 14% was almost completely eliminated using closedloop control. .

MAASS, DOI: 10.1520/STP163120190130

9. Closed-loop control has the potential to salvage parts and builds where LOF flaws have occurred. As such, use of closed-loop control can reduce scrap rate, improve yield, reduce unit part cost, and improve LPBF throughput. ACKNOWLEDGMENTS

This work was funded under Defense Logistics Agency contract SP4701-18-P-0107 under the direction of Denise Price, program manager for the Small Business Innovation Program. Paul Boulware and Heimdall Mendoza of EWI contributed to LPBF coupon design and fabrication, while Adam Brooks at EWI’s Buffalo Manufacturing Works performed CT scans. Trevor Lacon of Thermo Fisher assisted in the reduction of CT data, while Eric Anstadt of Quartus Engineering assisted in the implementation of the LTM and closed-loop control software, as well as the machine learning activity.

References 1. 2.

3.

4. 5.

6.

7. 8.

9.

.

L. Cherido, “Metal 3D Printer Market in 2019,” https://perma.cc/UA5N-RZG6 S. M. Kelly, P. C. Boulware, L. Cronley, G. Firestone, M. Jamshidinia, J. Marchal, T. Stempky, and C. Reichert, “In-Process Sensing of Laser Powder Bed Fusion Additive Manufacturing,” (paper presentation, Workshop on Predictive Theoretical and Computational Approaches for Additive Manufacturing, Washington, DC, October 7–9, 2015). F. Timm and E. Barth, “Non-Parametric Texture Defect Detection Using Weibull Features,” in Proceedings Volume 7877, Image Processing: Machine Vision Applications IV (Bellingham, WA: SPIE, 2011). “Weibull Distribution,” Wikpedia, https://perma.cc/N4WU-LEFJ A. R. Nassar, J. S. Keist, E. W. Reutzel, and T. J. Spurgeon, “Intra-Layer Closed-Loop Control of Build Plan during Directed Energy Additive Manufacturing of Ti–6Al–4V,” Additive Manufacturing 6 (2015): 39–52. J. Mireles, S. Ridwan, P. A. Morton, A. Hinojos, and R. B. Wicker, “Analysis and Correction of Defects within Parts Fabricated Using Powder Bed Fusion Technology,” Surface Topography: Metrology and Properties 3, no. 3 (2015), https://doi.org/10.1088/2051672X/3/3/034002 V. Renken, L. Lu ¨bbert, H. Blom, A. von Freyberg, and A. Fischer, “Model Assisted ClosedLoop Control Strategy for Selective Laser Melting,” Procedia CIRP 74 (2018): 659–663. Federal Aviation Administration, A Quantitative Assessment of Conventional Nondestructive Inspection Techniques for Detecting Flaws in Composite Laminate Aircraft Structures, DOT/FAA/TC-15/6 (Washington, DC: FAA, 2016). U.S. Department of Defense, Nondestructive Evaluation System Reliability Assessment, MIL-HDBK-1823A (Washington, DC: U.S. Department of Defense, 2009).

19

20

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190128

Christof Torrent,1 Andre´ Bauer,1 Malte Vollmer,1 and Thomas Niendorf1

On the Challenges toward Realization of Functionally Graded Structures by Electron Beam Melting—Fe-Base Shape Memory Alloy and Stainless Steel Citation C. Torrent, A. Bauer, M. Vollmer, and T. Niendorf, “On the Challenges toward Realization of Functionally Graded Structures by Electron Beam Melting—Fe-Base Shape Memory Alloy and Stainless Steel,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 20–33. http:// doi.org/10.1520/STP1631201901282

ABSTRACT

In the present study, an iron-manganese-aluminum-nickel (Fe-Mn-Al-Ni) shape memory alloy was processed on an austenitic steel (AISI 304) build platform by electron beam melting in order to study the feasibility of realizing functionally graded structures consisting of two different materials (i.e., a functional and a structural material). Compression specimens consisting of the processed shape memory alloy and the austenitic build platform in equal parts were investigated. The microstructure was analyzed in the as-built state and after different heat treatments, focusing on the interface between both materials. Scanning electron microscopy and electron backscatter diffraction measurements were conducted to reveal the relation between processing steps and the microstructural evolution. It is shown that the microstructure after the electron beam melting process is characterized by a preferred h001i orientation with respect to the build direction and that a suitable microstructure for good pseudoelastic Manuscript received October 30, 2019; accepted for publication February 25, 2020. 1 Universita¨t Kassel, Institute of Materials Engineering, Mo ¨nchebergstraße 3, 34125 Kassel, Germany https://orcid.org/0000-0002-9487-0853, A. B. https://orcid.org/0000-0002-9392-803X, C. T. M. V. https://orcid.org/0000-0002-8098-8498, T. N. https://orcid.org/0000-0003-2622-5817 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

TORRENT ET AL., DOI: 10.1520/STP163120190128

performance can be realized by post-processing heat treatments. Finally, incremental strain tests up to 12% compressive strain were conducted to analyze the overall mechanical performance of the specimens. Keywords additive manufacturing, shape memory alloy, austenitic steel, functionalized structures, secondary recrystallization

Introduction Apart from the unprecedented design freedom, additive manufacturing (AM) provides the possibility of influencing the microstructure of specimens and components as well as pathways toward processing and obtaining new materials. As an example, silver– iron compounds and other nanoparticle modified materials, which usually do not form an alloy in common melting processes, can be processed by AM.1,2 Moreover, it is shown that materials with tailored microstructural and mechanical properties over arbitrary component directions can be obtained by using different parameters during processing.3,4 Thus, it is possible to build a specimen and simultaneously tailor its microstructure. An advanced approach is the combined processing of different materials (e.g., by applying powder layers of various materials). In this study, it is demonstrated that specimens consisting of two different materials, that is, a functional shape memory alloy (SMA) and a structural material, can be realized by AM. The specimens in the electron beam melting (EBM) process are built layer-bylayer from material powder. The first layer is built up directly on the platform used. This is followed by moving the platform downward, spreading new powder and selective melting. Eventually, the parts grow along the build axis. In most cases, the platform and the specimen consist of different materials, and the interface between both is characterized by diffusion and compositional changes.5 Moreover, the thermal conditions change during the process and are influenced by preheating, melting, and spreading new powder. This thermal history is often referred to as in situ annealing and intrinsic heat treatment, respectively.6 While the influence of the thermal history cannot be completely suppressed, the influence of diffusion and compositional changes in vicinity of the platform is usually minimized by cutting the specimens well above the affected zone between specimens and platform. Therefore, the behavior in the specimen–platform interface has been barely investigated so far. However, for multi-material AM, the interface between the different materials is of high interest. In the present study, the platform is used as a part of the specimens in order to gain insights into the microstructural evolution of the interface and the mechanical behavior of specimens consisting of two different materials (i.e., an iron-base SMA and an austenitic steel). Due to their characteristic thermoelastic martensitic transformation, SMAs are a promising class of materials exhibiting functional properties such as pseudoelasticity, one-way effect, and two-way effect.7 These effects can be used for actuation, .

21

22

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

damping, and sensing applications, respectively. In recent years, processing of SMAs by AM attracted a lot of attention.8–10 Unique microstructural features upon AM (e.g., induced by directional solidification) can be beneficial for the functional performance of SMAs.11 One of the alloy systems of particular interest is the recently developed iron-manganese-aluminum-nickel (Fe-Mn-Al-Ni) alloy. It shows a low Clausius-Clapeyron slope of 0.53 MPa/K and a wide temperature range for pseudoelastic application (196 C to 240 C).12 Good pseudoelastic properties have been observed in single13–18 as well as in oligocrystalline12,19–24 states. However, the material is characterized by a highly anisotropic behavior leading to constraints, most importantly resulting in severe stress concentrations at grain boundary triple junctions. Therefore, it is of crucial importance to obtain highly textured microstructures through processing. Moreover, it was shown that oligocrystalline structures consisting of grains exceeding the cross section of the specimens improve the pseudoelastic behavior in this SMA (referred to as bamboo structures). Recently, a cyclic heat treatment procedure was developed for a copper (Cu)-based SMA leading to oligocrystalline and single crystalline structures in the order of several centimeters.25 The same kind of procedure was successfully applied to Fe-Mn-Al-Ni.23,26 In 2015, Niendorf et al.8 demonstrated the feasibility of processing Fe-Mn-Al-Ni by AM, i.e., by selective laser melting (SLM). Good pseudoelastic properties in compression were revealed. However, the high cooling rates as well as the evolution of residual stresses, both being characteristic for SLM, were shown to lead to processinduced cracking preferentially along grain boundaries.23 To account for these shortcomings, EBM was used in the present study in order to minimize crack formation. EBM is a hot powder bed process (i.e., being conducted at elevated process temperatures and lower cooling rates), eventually leading to decreasing residual stresses and the formation of the ductile c phase at the grain boundaries.23 Advantages of using Fe-Mn-Al-Ni as the SMA in focus are the expected good compatibility to AISI 304 and the relatively low dependency of the functional properties on the chemical composition as compared to other SMAs, such as nickel-titanium (Ni-Ti). To evaluate the microstructure within the additively manufactured SMA as well as in the diffusion zone between the SMA and the austenitic steel, different characterization methods were used, i.e., optical microscopy (OM), scanning electron microscopy (SEM), electron backscatter diffraction (EBSD), and energy dispersive X-ray spectroscopy (EDS). Moreover, different post-processing heat treatments were conducted to obtain a coarse-grained microstructure, which is crucial for a good pseudoelastic performance. Afterward, the mechanical behavior was investigated by incremental strain tests (ISTs) accompanied by in situ OM in order to correlate the overall performance with local microstructural features.

Materials and Experimental Methods Fe-36Mn-8Al-8.5Ni (wt%) with a powder size ranging from 60 lm to 100 lm was processed using an Arcam A2X EBM (Arcam, Mo¨lndal, Sweden) machine on an .

TORRENT ET AL., DOI: 10.1520/STP163120190128

TABLE 1 Parameter sets employed for processing of specimens and relative density measured by contrast image analysis based on OM micrographs

Set

Current (mA)

Voltage (kV)

Power (W)

Speed (mm/s)

Hatch (mm)

Layer (mm)

Volume Energy (J/mm3)

Rel. Density (%)

#1

3

60

180

3,000

0.05

0.05

24

92.1

#2

3.5

60

210

3,000

0.05

0.05

28

98.4

#3

4

60

240

3,000

0.05

0.05

32

98.5

AISI 304 stainless steel platform. Three parameter sets, differing in terms of volume energy, were used (table 1). Preheating to 950 C was conducted for all conditions. Compression specimens with dimensions of 3 by 3 by 6 mm3 were wire-cut by electrical discharge machining in a way that the loading direction was parallel to the build direction (BD). Care was taken to ensure that the interface between AISI 304 and Fe-Mn-Al-Ni was roughly in the middle of the specimens. For further investigations, selected specimens of each parameter set were subjected to subsequent thermal treatments, that is, solution annealing at 1,225 C for 1 h followed by water quenching in 80 C warm water and alternatively a cyclic heat treatment according to the procedure shown in figure 1 in order to promote abnormal grain growth (AGG). The cyclic heat treatment consists of four cycles between 1,225 C and 900 C with dwell times of 30 min and 15 min, respectively, and four cycles between 1,150 C and 900 C with dwell times of 15 min each. The heating and cooling rates

FIG. 1 Cyclic heat treatment procedure used in this study to promote abnormal grain growth in the Fe-Mn-Al-Ni SMA.

.

23

24

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

were 1 K min1. Afterward, the specimens were solution annealed at 1,225 C for 3 h and finally quenched in 80 C warm water. As it was shown by Omori et al.,25 the driving force for AGG is provided by subgrain structures introduced during formation and dissolution of a second phase. In further studies, it was demonstrated that the misorientation of the subgrain structures and, thus, the driving force for AGG, increased when cycles with upper temperatures only slightly above the solvus temperature were used.27,28 For further details on the cyclic heat treatment procedure, readers are referred to a recent study by Vollmer et al.28 Finally, specimens were immediately aged for 3 h at 200 C in order to form b precipitates, which are essential for the thermoelastic martensitic transformation in the Fe-Mn-Al-Ni SMA,17 and to avoid any room-temperature aging effects.15,21 Microstructure analysis was carried out on planes parallel to the build direction in order to investigate the interface between the materials as well as the porosity distribution over the build height. For that purpose, specimens were ground to 5 lm grit size and vibropolished to 0.02 lm using colloidal suspension. Microstructural characterization was conducted using OM as well as an SEM system operated at a nominal voltage of 20 kV. The relative density of the specimens was evaluated by pore analysis based on OM images. Polished surfaces with an area of about 1.5 by 1.5 mm2 were analyzed in as-built condition using the software ImageJ. The SEM system was equipped with EBSD and EDS units for grain orientation analysis and determination of the chemical composition, respectively. Compression tests on a cyclic heat-treated Fe-Mn-Al-Ni/AISI 304 specimen as well as on an AISI 304 reference specimen subjected to the same heat treatment were performed using a servohydraulic testing machine operated in displacement control at a rate of 5 lm s1. Strain was measured by an extensometer having a gauge length of 12 mm being directly attached to the compression grips, which were considered to be absolutely rigid. Strain values were recalculated according to the actual size of the specimens. Incremental strain tests up to 12% strain with a step size of 1% were performed. For in situ characterization, a Keyence microscope with a VH Z100 objective was mounted directly in front of the servohydraulic testing machine. Surface images in the loaded conditions as well as in the unloaded conditions of each increment were recorded.

Results and Discussion MICROSTRUCTURAL CHARACTERIZATION OF THE AS-BUILT STATE

In order to study the impact of the applied volume energy, three different beam currents were used. The micrographs (OM) in figure 2 show characteristic microstructures of the different sets of specimens in the as-built state. Obviously, Specimen #1 is characterized by a large number of pores. This can be related to the low volume energy and lack of fusion defects, respectively.29 In contrast, Specimens #2 and #3 show nearly the same relative density of about 98.5% (table 1). At higher volume energy densities, a tendency toward columnar solidification of the a phase can .

TORRENT ET AL., DOI: 10.1520/STP163120190128

FIG. 2 Optical micrographs of the additively manufactured Fe-Mn-Al-Ni in as-built state (build directions are highlighted by white arrows). Conditions shown differ by the used volume energy: (A) 24 J/mm3; (B) 28 J/mm3; and (C) 32 J/mm3.

be seen. In line with the results and conclusions detailed in Kurz, Bezenc¸on, and Ga¨umann,30 it is very likely that lower solidification and cooling rates as well as changes in the melt pool shape and depths caused by the higher beam currents and energy density, respectively, are responsible for the columnar solidification in these specimens. Eventually, microstructure evolution can be rationalized by the strong relationships between the solidification and cooling rate, the thermal gradient, and the resulting grain morphology.30 In all three conditions, a large fraction of serrated c-phase is formed, being sufficient to effectively suppress the formation of cracks alongside the grain boundaries during cooling.23 The high-volume fraction of c phase can be related to the high temperature prevailing within the chamber during processing. However, since the c phase is detrimental to the pseudoelastic behavior,31 additional heat treatments are necessary in order to obtain adequate functional properties. The AISI 304/Fe-Mn-Al-Ni interfaces of the as-built conditions are shown in figure 3. In all conditions, small pores are visible directly at the interface (dashed arrows in fig. 3C) as well as a segregated, dark spotted area with a thickness of about 30 to 50 lm next to the interface (dashed circles in fig. 3). Unfortunately, resolution of the EDS system is not high enough to analyze the composition of these spots. In contrast to the microstructure shown in figure 2, the ductile c phase can hardly be observed in the diffusion zone near the interface between the SMA and the platform. Changes in local chemical composition are, thus, assumed to be the reason for the occurrence of individual transcrystalline cracks in the vicinity of the interface as marked by the dashed arrows in figure 3B. It is very likely that the lack of ductile c-phase is linked to chromium diffusion from the AISI 304 to the Fe-MnAl-Ni, as it is well known that chromium stabilizes the a phase in Fe-Mn-Al-Ni.28 Moreover, a loss of the c-stabilizing manganese was observed for each parameter set in comparison to the original powder composition as is highlighted in figure 4. Manganese is known to stabilize the c phase; however, the element is also known to evaporate due to its very low vapor pressure in EBM processing; still, the loss of .

25

26

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 3 Backscattered electron (BSE) micrographs showing the interface area of Fe-Mn-Al-Ni and AISI 304 built with different volume energies: (A) 24 J/mm3; (B) 28 J/mm3; and (C) 32 J/mm3. Porosity along the interface is highlighted (dashed arrows in C) as well as transcrystalline cracks (dashed arrows in B); furthermore, a dark spotted area can be seen within the first 50 μm of every build job (white ovals). In each condition, a white solid arrow highlights the BD.

FIG. 4 Chemical analysis of the as-received powder and the material processed by the three investigated parameter sets as obtained by EDS analysis.

manganese can be controlled by the energy density applied for melting as has been shown very recently for a medium-manganese steel.32 MICROSTRUCTURAL CHARACTERIZATION AFTER DIFFERENT HEAT TREATMENTS

Microstructures present in the vicinity of the interface of the SMA and the austenitic steel after solution annealing are shown in figure 5. It should be emphasized that AGG occurred randomly in the specimens, independent from the volume energy .

TORRENT ET AL., DOI: 10.1520/STP163120190128

FIG. 5 BSE micrographs of the interface after solution annealing at 1,225 C for 1 h. Grain coarsening can be clearly seen in (A) and hardly in (B) (24 J/mm3 and 28 J/mm3). In both conditions, the microstructure differs in direct vicinity of the build plate. There is no coarsening; however, differing microstructure up to about 1 mm from the build plate can be seen in (C) (32 J/mm3; see white marker). The dark segregations in direct vicinity of the interface have become larger in comparison to the as-built state. BD is highlighted by white arrows.

applied in the EBM process. A characteristic microstructure is shown for Specimen #1 (fig. 5A). Niendorf et al.8 already reported on AGG after solution annealing of SLM-processed Fe-Mn-Al-Ni. They assumed that residual stresses originating from the high-temperature gradients and rapid solidification in the SLM process (conducted at relatively low platform temperature) provided the necessary energy for some grains to grow abnormally.33 However, the residual stresses in the EBMprocessed Fe-Mn-Al-Ni should be comparatively low and, thus, it is more likely that the different microstructures in the as-built condition, being characterized by grain morphology and second phase distribution, affect AGG kinetics. It was shown that a change in the chemical composition (i.e., the addition of titanium and chromium to Fe-Mn-Al-Ni) has a significant effect on the morphology of the c-phase of conventionally processed material, finally leading to pronounced changes in AGG kinetics and final grain sizes.28 In-depth analysis of the AGG kinetics in AM Fe-Mn-Al-Ni after solution treatments is, however, beyond the scope of the present study and, thus, will be the subject of future work. Besides random AGG in parts of the specimens, all sets of specimens reveal an area with a thickness of about 1 mm being characterized by a fundamentally different microstructural appearance. This is probably due to pronounced diffusion of chromium upon processing and during the annealing process. Furthermore, the segregations already mentioned in the case of the as-built condition grew during annealing. EDS analysis (not shown) demonstrated local enrichment in aluminum and nickel. In order to investigate AGG promoted by cyclic heat treatment in AMprocessed Fe-Mn-Al-Ni, specimens were subjected to the heat treatment procedure previously applied by Vollmer et al. (fig. 1).28 Since Specimen #1 showed a too high density of defects, the procedure was only applied to Specimens #2 and #3. After cyclic heat treatment, Specimen #2 does not show pronounced grain coarsening, .

27

28

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 6 EBSD analysis (inverse pole figure maps) of (A) Parameter Set #2 and (B) Parameter Set #3 after the cyclic heat treatment and subsequent aging. Color coding is with respect to BD. Corresponding pole figures of the areas highlighted by the dashed boxes in (A) and (B) are shown in (C) and (D).

whereas in Specimen #3, a large-grained, assumedly single crystalline structure with a high density of subgrain structures was obtained, as can be deduced from the pole figures (fig. 6C and 6D) of the areas highlighted by the dashed boxes in figure 6A and 6B. Similar to the results after the solution annealing, AGG cannot be observed in all specimens after cyclic heat treatment. In Specimen #3, it is very likely that grain growth was already completed before the final solution annealing step because subgrain structures are only present in areas where the grain growth was already completed before the last solution annealing step.28 The preferred orientation upon AGG (i.e., near h001i) indicates that the EBM induced texture of the as-built condition can be maintained in the condition after AGG. However, a final evaluation of effects of initial texture on the grain orientation upon AGG here is difficult due to the limited number of specimens. Relationships will be further analyzed on a statistical basis in future work. .

TORRENT ET AL., DOI: 10.1520/STP163120190128

FIG. 7 SEM EDS mappings of the interface region (built plate below the superimposed red dashed line) for Parameter Set #3 after cyclic heat treatment.

Again, a zone of about 1 mm, with a different microstructural appearance characterized by the absence of substantial grain growth, can be seen between the AISI 304 and the Fe-Mn-Al-Ni. The EDS mapping shown in figure 7 clearly reveals that the zone, where grain growth is hampered, coincides with the diffusion zone of chromium. Recently, Vollmer et al.28 revealed that the addition of chromium strongly inhibits AGG in Fe-Mn-Al-Ni. In light of these results, it is very likely that the growth of the single crystalline part stopped in specimen regions where a critical amount of chromium is exceeded. INCREMENTAL STRAIN TEST

In order to study the overall performance of the AISI 304/Fe-Mn-Al-Ni compound, incremental strain tests (ISTs) were conducted on the specimen shown in figure 6B as well as on an AISI 304 reference specimen. The critical stress for plastic deformation of the latter is quite low (fig. 8A), and it should be noted that the AISI 304 reference specimen reaches a stress of 400 MPa at approximately 8% applied strain. Taking into account that the share of AISI 304 is about half of the compound specimen and that no martensitic transformation has been observed in the very first cycles, it is obvious that, up to about 4% strain, the plastic deformation is mainly accommodated by the AISI 304, and the Fe-Mn-Al-Ni initially only deforms elastically. At about 4% applied strain, the stress level increased to about 400 MPa, and a martensitic transformation is seen for the first time in the Fe-Mn-Al-Ni alloy (not shown). This is in good agreement with results of Tseng et al.,17 showing critical stresses of about 400 MPa for single-crystal Fe-Mn-Al-Ni with h001i orientation in compression after an aging treatment of 3 h at 200 C. In the course of further cycling, clear traces of pseudoelasticity can be seen. Fairly good reversibility of martensitic transformation can be deduced from the stress-strain curves indicated by the opening between the unloading and loading path. .

29

30

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

In order to provide for deeper insights into the mechanical behavior, the stressstrain curve of the eighth cycle is highlighted in figure 8B. It is obvious that the mechanical response is partially characterized by plastic deformation of the AISI 304 alloy; however, a significant contribution of the pseudoelastic deformation of the Fe-Mn-Al-Ni alloy can be seen as well. In situ micrographs of the loaded and unloaded condition obtained within the eighth cycle are shown in figure 8C and 8D, respectively. In the loaded condition (figure 8C), pronounced topography changes, which can be linked to the deformation of the AISI 304, can be observed at the bottom of the micrograph. In contrast, martensitic structures can be seen in the FeMn-Al-Ni single crystal in the upper part. These structures partly disappear upon unloading (figure 8D), providing a proof of reversibility and, thus, pseudoelastic behavior. In addition, the grain boundaries within the diffusion zone are visible. It is assumed that the partly incomplete reversibility of the Fe-Mn-Al-Ni is linked to the interaction of the martensite with the subgrain structures,28 martensite variant interaction,16 as well as to effects related to the experimental setup (i.e., friction between the specimen and the grips). A certain amount of pores could also have a detrimental effect on the pseudoelastic behavior due to an interaction between

FIG. 8 Incremental strain test of an AISI 304 reference specimen and for an AISI 304/ Fe-Mn-Al-Ni compound specimen (Parameter Set #3) after cyclic heat treatment. In (A), the responses recorded during the complete ISTs up to 12% are plotted; (B) detail highlighting only the eighth cycle of the AISI 304/Fe-Mn-Al-Ni compound specimen; (C) corresponding in situ OM images of the AISI 304/FeMn-Al-Ni compound specimen in the loaded and (D) unloaded condition (see “x” in B).

.

TORRENT ET AL., DOI: 10.1520/STP163120190128

pores and martensite plates. Based on present results, quantitative analysis of the pseudoelastic performance is difficult. The plastic deformation of the AISI 304 can hardly be separated from the irrecoverable pseudoelastic strain of the Fe-Mn-Al-Ni. In-depth analysis of mechanical behavior using local strain evaluation by means of digital image correlation as well as an in-depth analysis of the microstructural features leading to the functional degradation will be the subject of follow-up studies.

Conclusions The present study reveals the possibility of directly producing functionally graded Fe-Mn-Al-Ni/AISI 304 specimens with good pseudoelastic properties by using EBM. Due to the high process temperature and the lower cooling rates as compared to SLM, a crack-free microstructure was observed in the Fe-Mn-Al-Ni SMA. However, cracks occurred in the diffusion zone between the build plate and the SMA. This assumedly can be attributed to an insufficient amount of ductile c phase at the grain boundaries in the deteriorated region, which is attributed to the diffusion of chromium from the substrate to the SMA. Subsequent heat treatments (i.e., solution annealing treatments and cyclic heat treatments) are adequate means to promote AGG in the Fe-Mn-Al-Ni alloy. Up to now, it remains an open question whether the strong texture after AM can be robustly maintained upon the AGG process. The in situ IST revealed a superimposed material response of the Fe-Mn-Al-Ni SMA and the austenitic stainless steel AISI 304. Primarily plastic deformation of the AISI 304 austenitic steel was observed at the beginning of the test followed by a partly reversible martensitic transformation of the SMA in later stages of the test. The polycrystalline diffusion zone of the Fe-Mn-Al-Ni showed hardly any martensitic transformation. To open up the full potential of multimaterial structures using Fe-Mn-Al-Ni as a functional alloy, future studies should focus on the application of high-strength structural materials. ACKNOWLEDGMENTS

Financial support by Deutsche Forschungsgemeinschaft (Project No. 250216343; NI1327/7-3) within the Emmy Noether-Program is gratefully acknowledged.

References 1. 2.

3.

.

J. T. Sehrt, S. Kleszczynski, and C. Notthoff, “Nanoparticle Improved Metal Materials for Additive Manufacturing,” Progress in Additive Manufacturing 2 (2017): 179–191. T. Niendorf, F. Brenne, P. Hoyer, D. Schwarze, M. Schaper, R. Grothe, M. Wiesener, G. Grundmeier, and H. J. Maier, “Processing of New Materials by Additive Manufacturing: Iron-Based Alloys Containing Silver for Biomedical Applications,” Metallurgical and Materials Transactions A 46 (2015): 2829–2833. T. Niendorf, F. Brenne, M. Schaper, A. Riemer, S. Leuders, W. Reimche, D. Schwarze, and H. J. Maier, “Labelling Additively Manufactured Parts by Microstructural Gradation— Advanced Copy-Proof Design,” Rapid Prototyping Journal 22 (2016): 630–635.

31

32

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

4.

5.

6.

7. 8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

.

H. Helmer, A. Bauereiß, R. F. Singer, and C. Ko ¨rner, “Grain Structure Evolution in Inconel 718 during Selective Electron Beam Melting,” Materials Science and Engineering: A 668 (2016): 180–187. A. Hinojos, J. Mireles, A. Reichardt, P. Frigola, P. Hosemann, L. E. Murr, and R. B. Wicker, “Joining of Inconel 718 and 316 Stainless Steel Using Electron Beam Melting Additive Manufacturing Technology,” Materials & Design 94 (2016): 17–27. D. Deng, R. L. Peng, H. So ¨derberg, and J. Moverare, “On the Formation of Microstructural Gradients in a Nickel-Base Superalloy during Electron Beam Melting,” Materials & Design 160 (2018): 251–261. K. Otsuka and C. M. Wayman, Shape Memory Materials (Cambridge, UK: Cambridge University Press, 2002). T. Niendorf, F. Brenne, P. Krooß, M. Vollmer, J. Gu ¨nther, D. Schwarze, and H. Biermann, “Microstructural Evolution and Functional Properties of Fe-Mn-Al-Ni Shape Memory Alloy Processed by Selective Laser Melting,” Metallurgical and Materials Transactions A 47 (2016): 2569–2573. B. E. Franco, J. Ma, B. Loveall, G. A. Tapia, K. Karayagiz, J. Liu, A. Elwany, R. Arroyave, and I. Karaman, “A Sensory Material Approach for Reducing Variability in Additively Manufactured Metal Parts,” Scientific Reports 7 (2017): 3604. J. Ma, B. Franco, G. Tapia, K. Karayagiz, L. Johnson, J. Liu, R. Arroyave, I. Karaman, and A. Elwany, “Spatial Control of Functional Response in 4D-Printed Active Metallic Structures,” Scientific Reports 7 (2017): 46707. Q. Zhou, M. D. Hayat, G. Chen, S. Cai, X. Qu, H. Tang, and P. Cao, “Selective Electron Beam Melting of NiTi: Microstructure, Phase Transformation and Mechanical Properties,” Materials Science and Engineering: A 744 (2019): 290–298. T. Omori, K. Ando, M. Okano, X. Xu, Y. Tanaka, I. Ohnuma, R. Kainuma, and K. Ishida, “Superelastic Effect in Polycrystalline Ferrous Alloys,” Science 333, no. 6038 (2011): 68–71. L. W. Tseng, J. Ma, S. J. Wang, I. Karaman, M. Kaya, Z. P. Luo, and Y. I. Chumlyakov, “Superelastic Response of a Single Crystalline FeMnAlNi Shape Memory Alloy under Tension and Compression,” Acta Materialia 89 (2015): 374–383. L. W. Tseng, J. Ma, S. J. Wang, I. Karaman, and Y. I. Chumlyakov, “Effects of Crystallographic Orientation on the Superelastic Response of FeMnAlNi Single Crystals,” Scripta Materialia 116 (2016): 147–151. M. Vollmer, M. J. Kriegel, A. Walnsch, V. Klemm, A. Leineweber, and T. Niendorf, “On the Microstructural and Functional Stability of Fe-Mn-Al-Ni at Ambient and Elevated Temperatures,” Scripta Materialia 162 (2019): 442–446. M. Vollmer, M. J. Kriegel, P. Krooß, S. Martin, V. Klemm, A. Weidner, Y. Chumlyakov, H. Biermann, D. Rafaja, and T. Niendorf, “Cyclic Degradation Behavior of h001i-Oriented Fe–Mn–Al–Ni Single Crystals in Tension,” Shape Memory and Superelasticity 4 (2017) 335–346. L. W. Tseng, J. Ma, B. C. Hornbuckle, I. Karaman, G. B. Thompson, Z. P. Luo, and Y. I. Chumlyakov, “The Effect of Precipitates on the Superelastic Response of [1 0 0] Oriented FeMnAlNi Single Crystals under Compression,” Acta Materialia 97 (2015): 234–244. L. W. Tseng, J. Ma, Y. I. Chumlyakov, and I. Karaman, “Orientation Dependence of Superelasticity in FeMnAlNi Single Crystals under Compression,” Scripta Materialia 166 (2019): 48–52. L. W. Tseng, J. Ma, M. Vollmer, P. Krooß, T. Niendorf, and I. Karaman, “Effect of Grain Size on the Superelastic Response of a FeMnAlNi Polycrystalline Shape Memory Alloy,” Scripta Materialia 125 (2016): 68–72.

TORRENT ET AL., DOI: 10.1520/STP163120190128

20.

21.

22. 23.

24.

25.

26. 27. 28.

29.

30. 31.

32.

33.

.

H. Ozcan, J. Ma, S. J. Wang, I. Karaman, Y. Chumlyakov, J. Brown, and R. D. Noebe, “Effects of Cyclic Heat Treatment and Aging on Superelasticity in Oligocrystalline FeMn-Al-Ni Shape Memory Alloy Wires, Scripta Materialia 134 (2017): 66–70. H. Ozcan, J. Ma, I. Karaman, Y. I. Chumlyakov, R. Santamarta, J. Brown, and R. D. Noebe, “Microstructural Design Considerations in Fe-Mn-Al-Ni Shape Memory Alloy Wires: Effects of Natural Aging,” Scripta Materialia 142 (2018): 153–157. T. Omori, M. Okano, and R. Kainuma, “Effect of Grain Size on Superelasticity in Fe-MnAl-Ni Shape Memory Alloy Wire,” APL Materials 1 (2013): 32103. M. Vollmer, C. Segel, P. Krooß, J. Gu ¨nther, L. W. Tseng, I. Karaman, A. Weidner, H. Biermann, and T. Niendorf, “On the Effect of Gamma Phase Formation on the Pseudoelastic Performance of Polycrystalline Fe–Mn–Al–Ni Shape Memory Alloys,” Scripta Materialia 108 (2015): 23–26. M. Vollmer, P. Krooß, M. J. Kriegel, V. Klemm, C. Somsen, H. Ozcan, I. Karaman, A. Weidner, D. Rafaja, H. Biermann, and T. Niendorf, “Cyclic Degradation in Bamboo-Like Fe– Mn–Al–Ni Shape Memory Alloys—The Role of Grain Orientation,” Scripta Materialia 114 (2016): 156–160. T. Omori, T. Kusama, S. Kawata, I. Ohnuma, Y. Sutou, Y. Araki, K. Ishida, and R. Kainuma, “Abnormal Grain Growth Induced by Cyclic Heat Treatment,” Science 341 (2013): 1500–1502. T. Omori, H. Iwaizako, and R. Kainuma, “Abnormal Grain Growth Induced by Cyclic Heat Treatment in Fe-Mn-Al-Ni Superelastic Alloy,” Materials & Design 101 (2016): 263–269. T. Kusama, T. Omori, T. Saito, S. Kise, T. Tanaka, Y. Araki, and R. Kainuma, “Ultra-Large Single Crystals by Abnormal Grain Growth,” Nature Communications 8 (2017): 354. M. Vollmer, T. Arold, M. J. Kriegel, V. Klemm, S. Degener, J. Freudenberger, and T. Niendorf, “Promoting Abnormal Grain Growth in Fe-Based Shape Memory Alloys through Compositional Adjustments,” Nature Communications 10 (2019): 2337. S. Tammas-Williams, H. Zhao, F. Le ´onard, F. Derguti, I. Todd, and P. B. Prangnell, “XCT Analysis of the Influence of Melt Strategies on Defect Population in Ti–6Al–4V Components Manufactured by Selective Electron Beam Melting,” Materials Characterization 102 (2015): 47–61. W. Kurz, C. Bezenc¸on, and M. Ga¨umann, “Columnar to Equiaxed Transition in Solidification Processing,” Science and Technology of Advanced Materials 2 (2001): 185–191. M. Vollmer, P. Krooß, I. Karaman, and T. Niendorf, “On the Effect of Titanium on Quenching Sensitivity and Pseudoelastic Response in Fe-Mn-Al-Ni-Base Shape Memory Alloy,” Scripta Materialia 126 (2017): 20–23. J. Gu ¨nther, F. Brenne, M. Droste, M. Wendler, O. Volkova, H. Biermann, and T. Niendorf, “Design of Novel Materials for Additive Manufacturing—Isotropic Microstructure and High Defect Tolerance,” Scientific Reports 8 (2018): 1298. L. Thijs, F. Verhaeghe, T. Craeghs, J. van Humbeeck, and J.-P. Kruth, “A Study of the Microstructural Evolution during Selective Laser Melting of Ti–6Al–4V,” Acta Materialia 58 (2010): 3303–3312.

33

34

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190149

Mostafa Yakout1 and Mohamed A. Elbestawi1

Residual Stress Formation in Laser-Based Powder Bed Fusion (PBF-LB) of Invar 36 Citation M. Yakout and M. A. Elbestawi, “Residual Stress Formation in Laser-Based Powder Bed Fusion (PBF-LB) of Invar 36,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 34–44. http:// doi.org/10.1520/STP1631201901492

ABSTRACT

Laser-based powder bed fusion (PBF-LB), also known as selective laser melting (SLM), is a metal additive manufacturing process associated with residual stress formation due to rapid heating and cooling. This paper aims at investigating residual stresses and deflections of Invar 36 parts produced using a selective laser melting machine that is equipped with a 400-W ytterbium fiber laser source. Invar 36 has been used in the aerospace industry for the past decade because it is known for its ferromagnetic property, high strength, and improved toughness. Invar 36 has a very low coefficient of thermal expansion below its Curie temperature (2798C); therefore, it is a good candidate for the PBF-LB process because it shows low thermal stresses and small deflections. Parts manufactured by the PBF-LB process usually experience void formation, internal cracks, metallurgical changes, vaporization of alloying elements, and residual stress formation. In this paper, evolution of residual stresses and deflections of Invar 36 parts is analyzed using a coupled thermal-mechanical finite element model in ANSYS Additive 19.2 software. The numerical results are validated experimentally. Residual stresses are measured using an X-ray diffraction (XRD) instrument, and part deflections are measured using a coordinate measuring machine (CMM). Parts are produced at the optimum process parameters for Invar 36 in order to eliminate the formation of excessive residual stresses during Manuscript received November 26, 2019; accepted for publication March 30, 2020. 1 Dept. of Mechanical Engineering, McMaster University, 1280 Main St. W., Hamilton, ON L8S 4L7, Canada, M. Y. http://orcid.org/0000-0002-0887-0217, M. A. E. http://orcid.org/0000-0003-0982-6127 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

YAKOUT AND ELBESTAWI, DOI: 10.1520/STP163120190149

the PBF-LB process. These optimum process parameters that give the smallest deflections and lowest residual stresses are tested using the finite element model. In addition, the relationships between the thermal properties of Invar 36 and the process-induced residual stresses and deflections are discussed. Keywords additive manufacturing, selective laser melting, powder bed fusion, Invar 36, residual stress, part deflection, thermal expansion, thermal properties, finite element, ANSYS additive

Introduction Residual stress formation is one of the main challenges in laser additive manufacturing (AM) as it affects part quality and process productivity.1–4 ISO/ASTM 52900, Standard Terminology for Additive Manufacturing—General Principles—Terminology, defines powder bed fusion (PBF) as an AM process in which thermal energy selectively fuses regions of a powder bed.5 In the laser-based powder bed fusion (PBF-LB) process, a laser source is used to fully melt metal powder for producing metal objects layer-by-layer.6,7 The PBF-LB process is associated with high-temperature gradients, which in turn induce high residual stresses in parts produced.8–10 The PBF-LB parts can also have a range of internal defects such as lack-of-fusion defects, gas pores, layered voids, keyholes, cracks between subsequent layers, spatters, and soot.11–13 Residual stresses and internal defects alter the mechanical properties and quality of parts produced via the PBF-LB process. These defects and stress risers depend on the material properties, processing conditions, and part design.14–16 Invar 36, a nickel-iron alloy, is relatively difficult to cut but weldable. Invar 36 has a very low coefficient of thermal expansion, making it a good candidate for laser AM.12,17 Invar 36 has been used for several applications in the electronics and aerospace industries including, but not limited to, ring laser gyroscopes, electronic tubes for radios, precision instruments, electronic devices, orbiting satellites, laser components, microwave devices, waveguide tubes, and timing devices.18–22 This paper presents residual stress formation in PBFLB of Invar 36 and its dependency on material properties and processing conditions. Residual stresses can be measured using either relaxation destructive methods or diffraction nondestructive methods.23 The X-ray diffraction (XRD) method is a nondestructive technique that measures residual stresses in metal parts using the spacing of atomic planes of the crystalline lattice.24 In this study, the XRD method is used to measure the evolution of residual stresses in Invar 36 parts produced using the PBFLB process.

Materials and Methods FEEDSTOCK MATERIAL AND PROCESS CONDITIONS

Gas-atomized spherical Invar 36 powder, provided by Sandvik Osprey LTD in the size range of 15 to 45 lm, was used. Table 1 presents the chemical composition of .

35

36

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 1 Chemical composition of Invar 36 powder (wt.%)

Provided by Sandvik Osprey

25

EDS results of powder samples

Fe

Ni

Mn

Si

Balance

35.5–36.5

< 0.5

< 0.25

63.10

36.32

0.41

0.17

Note: Fe ¼ iron; Ni ¼ nickel; Mn ¼ manganese; Si ¼ silicon.

Invar 36 powder tested using energy dispersive X-ray spectroscopy (EDS) compared to that provided by the manufacturer. The powder was sieved using a 200-mesh sieve (75 lm) to ensure good printability. Invar 36 cantilevers were produced using an EOSINT M280 machine (EOS Electro Optical Systems Ltd.) for deflection measurements, and cubes were produced using an OmniSint-160 machine (OmniTek Tecnologia) for residual stress analysis. Both machines are selective laser melting machines equipped with a 400W ytterbium fiber laser and an Argon gas flow for the PBF-LB process. The OmniSint machine produces parts on a circular build plate with a diameter of 160 mm, while the EOSINT machine produces parts on a square build plate with an edge length of 250 mm. Figure 1A shows the cantilever dimensions and the coordinate measuring machine (CMM) measurement locations, while figure 1B shows the cube dimensions and the directional residual stress components. The locations of the XRD measurements are shown in figure 1C. These cantilevers and cubes were introduced in another publication by the authors.26 All test samples were produced at a laser power (P) of 250 W, scanning speed (v) of 600 mm/s, and hatch spacing (h) of 0.12 mm. The layer thickness, t, stripe width, w, stripe overlap, c, and scanning rotation between subsequent layers were maintained constant at 0.04 mm, 10 mm,

FIG. 1 (A) Cantilever dimensions and the CMM measurement locations, (B) cube dimensions and the directional residual stress components, and (C) the residual stress measurement locations.

.

YAKOUT AND ELBESTAWI, DOI: 10.1520/STP163120190149

0.08 mm, and 67 , respectively. These samples were built directly on the build plate without supports and were cut from the build plate using wire electrical discharge machining (EDM). The volumetric laser energy density (Ev) at these process parameters equals 86.8 J/mm3 and is defined as the critical laser energy density (EC) of Invar 36. Parts produced at EC have the highest density and mechanical toughness as reported in the open literature.27,28 FINITE ELEMENT ANALYSIS

ANSYS Additive 19.2, a coupled thermal-mechanical finite element (FE) model, was used for predicting residual stresses and deflections in PBF-LB of Invar 36. The model uses an orthogonal Cartesian mesh in which each mesh layer is an average of multiple scan layers. It consists of (i) transient thermal analysis that predicts the transient temperature field using a three-dimensional (3D) Fourier conduction approach and correction coefficients for convection and radiation as well as (ii) static structural analysis that predicts the stress field using an inherent strain method. Mesh sensitivity analysis was performed, and the temperature-dependent properties of Invar 36 were imported from the open literature. All boundary conditions described by Yakout et al.26 were applied in the current study, and then the FE model was used for both cubes and cantilevers. The analysis of the cubes was used for obtaining the residual stress profiles, while the analysis of the cantilevers was used for predicting the part deflections. EXPERIMENTAL WORK

Residual Stress Analysis

A high-speed XRD laboratory system (Proto Manufacturing Ltd., Canada) was used to measure in-depth residual stress profiles for Invar 36 cubes produced using the PBF-LB process. These residual stress profiles were determined by measuring the biaxial residual stresses at nine different depth locations from the surface, as shown in figure 1C. The top surface of each sample was cleaned via air blasting to eliminate the effect of the surface roughness. Each depth measurement location was reached by removing monolayers from the surface using electropolishing in order to avoid introducing additional residual stresses. The lattice deformation at each location ˚ , a Bragg angle was obtained using a Mn Ka target with a wavelength of 1.79026 A of 152.88 for the crystallographic plane (3 1 1), and 11 different tilting angles from 258 to þ258. The in-depth residual stress profiles were determined at both top and side surfaces of the cubes. The top surface profiles represent the evolution of longitudinal and transverse horizontal residual stresses in the vertical (build) direction, rx and ry, respectively. The side surface profiles represent the evolution of vertical and horizontal residual stresses in the layer direction, rz and r½yx , respectively. Part Deflection

After the build plate removal, the cantilever deflection was measured using a CRYSTA-Apex S544 CMM (Mitutoyo) with a scale value of 0.1 lm. A total of .

37

38

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

40 measurement points, shown in figure 1A, was taken along the 50-mm length of the cantilever in which each measurement point was an average of three measurements. The cantilever deflection measured via CMM was compared to that predicted using the FE model.

Results and Discussion RESIDUAL STRESS OF INVAR 36

The FE model was used to predict residual stress profiles of Invar 36 cubes in all directions. Figure 2A–C shows the 3D numerical profiles of longitudinal horizontal residual stress (rx), lateral horizontal residual stress (ry), and vertical residual stress (rz), respectively. High-tensile horizontal stresses were observed at the top and bottom surfaces. During the cooling process, the temperature of the top layer decreases more rapidly than the temperature of the subsequent layer, leading to volume contraction at the top layer. Thus, tensile thermal stresses were induced in the top layers and compressive thermal stresses were induced in the core. Similarly, hightensile vertical stresses were observed at the side surfaces because of the rapid contraction of the side surfaces of each layer during the cooling of that layer. It was also observed that the vertical stresses were higher than the horizontal stresses at the side surfaces. This was attributed to the high-temperature gradients between layers when feeding a new layer of powder. High compressive residual stresses were observed at the center of the part in all directions, as shown in figure 2. The longitudinal and lateral horizontal stresses reached to 130 MPa at the center, while the vertical residual stress reached to 207 MPa at the center. Compressive residual stresses are usually found at the center of the AM parts when there are tensile residual stresses at the surface.29 If the part has internal cracks, the tensile residual stresses may cause an extensive propagation of this crack, while the compressive residual stresses may help in closing the crack. The FE numerical results agreed with the XRD experimental results except near surfaces as shown in figure 3A and B. The residual stresses measured using XRD

FIG. 2 Profiles of (A) longitudinal horizontal, (B) transverse horizontal, and (C) vertical residual stresses predicted using the FE model.

.

YAKOUT AND ELBESTAWI, DOI: 10.1520/STP163120190149

FIG. 3 In-depth profiles of (A) longitudinal and transverse horizontal residual stresses in the build direction and (B) horizontal and vertical residual stresses in the layer direction.

near the surface (up to 0.2 mm from the surface) were lower than those obtained via simulation. This is likely because of some errors in the XRD results near the surface due to the surface roughness of the as-built AM parts.30 The horizontal residual stress profiles showed high tensile stresses at the top surface, transitioning to compressive stresses at a depth of approximately 1.5 mm. Similarly, tensile horizontal stresses were observed at the side surface, leading to compressive stresses at a 1-mm depth in the metal. The highest tensile residual stresses (348 6 30 MPa) were observed in the vertical direction within the top 0.5 mm from the side surface. The vertical residual stresses started with tensile stresses at the side surface, leading to compressive stresses at a depth of approximately 1.2 mm from the side surface. It was concluded that: 1. Invar 36 parts produced by PBF-LB had high tensile residual stresses at the outer surfaces and compressive residual stresses at the center of the parts. 2. Tensile residual stresses were induced at the outer surfaces along the surface plane, which meant that the top and bottom surfaces had tensile stresses in the horizontal direction and the side surfaces had tensile stresses in the vertical direction. 3. Invar 36 parts showed vertical residual stresses much higher than the horizontal residual stresses. This was attributed to the high temperature gradients in the vertical direction during the deposition of a new layer. .

39

40

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

DEFLECTION OF INVAR 36 CANTILEVERS

The FE model was used to predict the deflection of Invar 36 cantilevers. In addition, cantilevers were produced using the PBF-LB process at the aforementioned process parameters for experimental validation. Figure 4A shows the distribution of the cantilever deformation in the vertical direction (z) as determined by the FE model. The heights of 28 points along the top surface of the cantilever were determined and plotted against the horizontal positions of these points. The best fitting curve of these located points was parabola, which represented the final shape of the cantilever after the build plate removal. The cantilever deflection determined by the FE model was compared to that obtained experimentally using the CMM (fig. 4B). The rapid heating and cooling phases during the PBF-LB process induced internal residual stresses. If these stresses were relieved prior to the removal of the build plate, the parts produced would not warp and deflect.31 The cantilevers did not undergo any stress-relieving treatment; therefore, they were expected to wrap and deflect. It was observed that the cantilevers manifested convex deflection after wire EDM, as shown in figure 4A. It was concluded that: 1. The maximum cantilever deflection was 0.1 mm. The overall deflection of the cantilever represented the residual stress relief that occurred during the wire EDM process. 2. The numerical results agreed with the experimental results. The maximum difference between the experimental deflection and numerical deflection was 0.03 mm, which was likely due to the roughness of the cantilever surface.

Conclusions This paper used an FE model to predict deflections and residual stresses of Invar 36 parts produced using the PBF-LB process. The parts were produced at the optimum

FIG. 4 (A) Deformation in the z-direction obtained from the FE model and (B) a comparison between the experimental and numerical cantilever deflections.

.

YAKOUT AND ELBESTAWI, DOI: 10.1520/STP163120190149

process conditions that developed a critical laser energy density (EC) of 86.8 J/mm3. The part deflection was experimentally examined using a CMM, while residual stresses were measured using the XRD method. Because the model did not explicitly include the effect of pore formation and spatter formation, it could not be used for any other processing conditions where pores and spatters were formed. The numerical results agreed with the experimental results at the critical laser energy density. Due to its nature, the PBF-LB process would always be associated with residual stress formation. Rather than avoiding residual stress formation during the PBF-LB process, which is practically impossible, this paper used the optimum process parameters that minimized the amount of residual stresses induced during the process. In the PBF-LB process, residual stress formation depends on the thermal properties of the material, process parameters, and postprocessing procedures. Because Invar 36 has a very low coefficient of thermal expansion, it was the process parameters that most likely caused residual stress formation during the process. Invar 36 is a good candidate for the PBF-LB process because it does not induce high-tensile residual stresses during the process compared to other aerospace alloys such as Ti-6Al-4V32 and stainless steel 316L.26 Ti-6Al-4V parts had tensile residual stresses in the vertical direction up to 920 MPa, while stainless steel 316L parts showed residual stresses up to 600 MPa in the vertical direction. The highest tensile residual stress observed in the vertical direction of Invar 36 parts was 350 MPa, which is approximately half of what was observed in stainless steel 316L parts and a third of what was observed in Ti-6Al-4V parts. This is likely because Invar 36 has a low coefficient of thermal expansion compared to the other two alloys. Studying the influence of powder preheating,33 build plate preheating,34,35 heat treatment,36 and layer remelting37 on residual stress formation in Invar 36 is considered for future work. ACKNOWLEDGMENTS

The use of the wire EDM facility at Liburdi Engineering in Ontario, Canada, is gratefully acknowledged. Experiments in this work were funded by the Natural Sciences and Engineering Research Council of Canada (NSERC), Grant: RGPIN-2016-06268.

References 1.

2. 3.

.

Y. Yang, M. Allen, T. London, and V. Oancea, “Residual Strain Predictions for a Powder Bed Fusion Inconel 625 Single Cantilever Part,” Integrating Materials and Manufacturing Innovation 8, no. 3 (2019): 294–304, https://doi.org/10.1007/s40192-019-00144-5 C. Li, Z. Y. Liu, X. Y. Fang, and Y. B. Guo, “Residual Stress in Metal Additive Manufacturing,” Procedia CIRP 71 (2018): 348–353, https://doi.org/10.1016/j.procir.2018.05.039 L. Zumofen, C. Beck, A. Kirchheim, and H.-J. Dennig, “Quality Related Effects of the Preheating Temperature on Laser Melted High Carbon Content Steels,” in Industrializing Additive Manufacturing—Proceedings of Additive Manufacturing in Products and Applications—Ampa2017, ed. M. Meboldt and C. Klahn (Cham, Switzerland: Springer, 2018), 210–219, https://doi.org/10.1007/978-3-319-66866-6_21

41

42

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14. 15.

16.

17.

18. 19. 20.

.

¨ chsner, “On the Anisotropic L. Hitzler, J. Hirsch, B. Heine, M. Merkel, W. Hall, and A. O Mechanical Properties of Selective Laser-Melted Stainless Steel,” Materials 10, no. 10 (2017): 1136, https://doi.org/10.3390/ma10101136 Standard Terminology for Additive Manufacturing—General Principles—Terminology, ISO/ASTM52900-15(2015) (West Conshohocken, PA: ASTM International, approved December 1, 2015), https://doi.org/10.1520/ISOASTM52900-15 M. Yakout, M. A. Elbestawi, and S. C. Veldhuis, “A Review of Metal Additive Manufacturing Technologies,” Solid State Phenomena 278 (2018): 1–14, https://doi.org/10.4028 /www.scientific.net/SSP.278.1 Additive Manufacturing—Design—Part 1: Laser-Based Powder Bed Fusion of Metals, ISO/ ASTM52911-1-19(2019) (West Conshohocken, PA: ASTM International, approved July 1, 2019), https://doi.org/10.1520/F3280-19 L. Mugwagwa, D. Dimitrov, S. Matope, and I. Yadroitsev, “Influence of Process Parameters on Residual Stress Related Distortions in Selective Laser Melting,” Procedia Manufacturing 21 (2018): 92–99, https://doi.org/10.1016/j.promfg.2018.02.099 P. K. Gokuldoss, S. Kolla, and J. Eckert, “Additive Manufacturing Processes: Selective Laser Melting, Electron Beam Melting and Binder Jetting-Selection Guidelines,” Materials (Basel, Switzerland) 10, no. 6 (2017): 672, https://doi.org/10.3390/ma10060672 ¨ . Poyraz and M. C. Kus O han, “Residual Stress-Induced Distortions in Laser Powder Bed Additive Manufacturing of Nickel-Based Superalloys,” Strojnisˇki Vestnik—Journal of Mechanical Engineering (2019), https://doi.org/10.5545/sv-jme.2019.6004 T. DebRoy, H. L. Wei, J. S. Zuback, T. Mukherjee, J. W. Elmer, J. O. Milewski, A. M. Beese, A. Wilson-Heid, A. De, and W. Zhang, “Additive Manufacturing of Metallic Components—Process, Structure and Properties,” Progress in Materials Science 92 (2018): 112– 224, https://doi.org/10.1016/j.pmatsci.2017.10.001 M. Yakout, A. Cadamuro, M. A. Elbestawi, and S. C. Veldhuis, “The Selection of Process Parameters in Additive Manufacturing for Aerospace Alloys,” The International Journal of Advanced Manufacturing Technology 92, no. 5 (2017): 2081–2098, https://doi.org /10.1007/s00170-017-0280-7 M. Yakout, M. A. Elbestawi, and S. C. Veldhuis, “On the Characterization of Stainless Steel 316L Parts Produced by Selective Laser Melting,” The International Journal of Advanced Manufacturing Technology 95, no. 5 (2018): 1953–1974, https://doi.org/10.1007/s00170017-1303-0 J. L. Bartlett and X. Li, “An Overview of Residual Stresses in Metal Powder Bed Fusion,” Additive Manufacturing 27 (2019): 131–149, https://doi.org/10.1016/j.addma.2019.02.020 D. Bourell, J. P. Kruth, M. Leu, G. Levy, D. Rosen, A. M. Beese, and A. Clare, “Materials for Additive Manufacturing,” CIRP Annals 66, no. 2 (2017): 659–681, https://doi.org/10.1016 /j.cirp.2017.05.009 Y. P. Yang, M. Jamshidinia, P. Boulware, and S. M. Kelly, “Prediction of Microstructure, Residual Stress, and Deformation in Laser Powder Bed Fusion Process,” Computational Mechanics 61, no. 5 (2018): 599–615, https://doi.org/10.1007/s00466-017-1528-7 C. Qiu, N. J. E. Adkins, and M. M. Attallah, “Selective Laser Melting of Invar 36: Microstructure and Properties,” Acta Materialia 103 (2016): 382–395, https://doi.org/10.1016 /j.actamat.2015.10.020 Y. Nakamura, “The Invar Problem,” IEEE Transactions on Magnetics 12, no. 4 (1976): 278–291, https://doi.org/10.1109/TMAG.1976.1059049 M. Shiga, “Invar Alloys,” Current Opinion in Solid State and Materials Science 1, no. 3 (1996): 340–348, https://doi.org/10.1016/S1359-0286(96)80023-4 T. Ogawa and T. Koseki, “Weldability of Invar (Fe-36%Ni) and Its Large-Diameter Pipe,” Welding Journal 65, no. 28 (1986): 35–44.

YAKOUT AND ELBESTAWI, DOI: 10.1520/STP163120190149

21. 22.

23.

24.

25. 26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

.

C. E. Guillaume, “Invar and Its Applications,” Nature 71, no. 1832 (1904): 134–139, https:// doi.org/10.1038/071134a0 L. L. Harner, “After 100 Years, the Uses for Invar Continue to Multiply” (Reading, PA: Carpenter Technology Corp., 2020), http://web.archive.org/web/20200330155755 /https://www.carpentertechnology.com/en/alloy-techzone/technical-information /technical-articles/after-100-years-the-uses-for-invar-continue-to-multiply G. S. Schajer and C. O. Ruud, “Overview of Residual Stresses and Their Measurement,” in Practical Residual Stress Measurement Methods, ed. G. S. Schajer (West Sussex, UK: John Wiley & Sons, 2013), 1–27, https://doi.org/10.1002/9781118402832.ch1 C. E. Murray and I. C. Noyan, “Applied and Residual Stress Determination Using X-Ray Diffraction,” in Practical Residual Stress Measurement Methods, ed. G. S. Schajer (Sussex, UK: John Wiley & Sons, 2013), 139–161, https://doi.org/10.1002/9781118402832.ch6 “Nickel Alloy Metal Powder,” Sandvik Osprey, http://web.archive.org/web/20200330155514 /https://www.metalpowder.sandvik/en/products/metal-powder-alloys/nickel-alloys/ M. Yakout, M. A. Elbestawi, S. C. Veldhuis, and S. Nangle-Smith, “Influence of Thermal Properties on Residual Stresses in SLM of Aerospace Alloys,” Rapid Prototyping Journal (2019), https://doi.org/10.1108/rpj-03-2019-0065 M. Yakout, M. A. Elbestawi, and S. C. Veldhuis, “A Study of Thermal Expansion Coefficients and Microstructure during Selective Laser Melting of Invar 36 and Stainless Steel 316L,” Additive Manufacturing 24 (2018): 405–418, https://doi.org/10.1016/j.addma.2018.09.035 M. Yakout, M. A. Elbestawi, and S. C. Veldhuis, “Density and Mechanical Properties in Selective Laser Melting of Invar 36 and Stainless Steel 316L,” Journal of Materials Processing Technology 266 (2019): 397–420, https://doi.org/10.1016/j.jmatprotec.2018.11.006 A. K. Syed, B. Ahmad, H. Guo, T. Machry, D. Eatock, J. Meyer, M. E. Fitzpatrick, and X. Zhang, “An Experimental Study of Residual Stress and Direction-Dependence of Fatigue Crack Growth Behaviour in As-Built and Stress-Relieved Selective-Laser-Melted Ti6Al4V,” Materials Science and Engineering: A 755 (2019): 246–257, https://doi.org /10.1016/j.msea.2019.04.023 P. Mercelis and J. P. Kruth, “Residual Stresses in Selective Laser Sintering and Selective Laser Melting,” Rapid Prototyping Journal 12, no. 5 (2006): 254–265, https://doi.org /10.1108/13552540610707013 F. Calignano, M. Galati, and L. Iuliano, “A Metal Powder Bed Fusion Process in Industry: Qualification Considerations,” Machines 7, no. 4 (2019), https://doi.org/10.3390 /machines7040072 M. Yakout, M. A. Elbestawi, and S. C. Veldhuis, “A Study of the Relationship between Thermal Expansion and Residual Stresses in Selective Laser Melting of Ti-6Al-4V,” Journal of Manufacturing Processes 52 (2020): 181–192, https://doi.org/10.1016 /j.jmapro.2020.01.039 L. Papadakis, D. Chantzis, and K. Salonitis, “On the Energy Efficiency of Pre-Heating Methods in SLM/SLS Processes,” The International Journal of Advanced Manufacturing Technology 95, no. 1 (2018): 1325–1338, https://doi.org/10.1007/s00170-017-1287-9 J. D. Roehling, W. L. Smith, T. T. Roehling, B. Vrancken, G. M. Guss, J. T. McKeown, M. R. Hill, and M. J. Matthews, “Reducing Residual Stress by Selective Large-Area Diode Surface Heating during Laser Powder Bed Fusion Additive Manufacturing,” Additive Manufacturing 28 (2019): 228–235, https://doi.org/10.1016/j.addma.2019.05.009 R. Mertens, S. Dadbakhsh, J. V. Humbeeck, and J. P. Kruth, “Application of Base Plate Preheating during Selective Laser Melting,” Procedia CIRP 74 (2018): 5–11, https:// doi.org/10.1016/j.procir.2018.08.002 M. S. I. N. Kamariah, W. S. W. Harun, N. Z. Khalil, F. Ahmad, M. H. Ismail, and S. Sharif, “Effect of Heat Treatment on Mechanical Properties and Microstructure of Selective

43

44

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

37.

.

Laser Melting 316L Stainless Steel,” IOP Conference Series: Materials Science and Engineering 257 (2017): 012021, https://doi.org/10.1088/1757-899x/257/1/012021 J. Vaithilingam, R. D. Goodridge, R. J. M. Hague, S. D. R. Christie, and S. Edmondson, “The Effect of Laser Remelting on the Surface Chemistry of Ti6Al4V Components Fabricated by Selective Laser Melting,” Journal of Materials Processing Technology 232 (2016): 1–8, https://doi.org/10.1016/j.jmatprotec.2016.01.02

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190137

Tomio Ohtsuki,1 Lonnie Smith,1 Ming Tang,1 and P. Chris Pistorius1

Origin of Oxides and OxideRelated Pores in Laser Powder Bed Fusion Parts Citation T. Ohtsuki, L. Smith, M. Tang, and P. C. Pistorius, “Origin of Oxides and Oxide-Related Pores in Laser Powder Bed Fusion Parts,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 45–60. http://doi.org/10.1520/STP1631201901372

ABSTRACT

Fatigue cracks grow from pores at the surface of components that were produced by laser powder bed fusion (LPBF). In AlSi10Mg components produced by LPBF, large oxides apparently interfere with consolidation of powder into the melt pool, contributing to part porosity; the oxides may also nucleate hydrogen porosity. In previous work, it was found that the effect of such porosity on fatigue life could be predicted by measuring pores found on a sample size of a few square millimeters and extrapolating to the much larger surface of a fatigue test specimen. The aim of this work is to understand the fundamental origin of oxides in LPBF as a basis for controlling the defects. The sources considered here are the native oxide on the surface of metal powder and oxidation of hot spatter in the build chamber for the case of LPBF of UNS N07718 samples. Kinetic analysis indicates that the rate of oxidation of a spatter droplet would be controlled by the oxygen concentration in the build chamber. From measurement of the surface coverage of deposited oxide particles (apparently oxidized spatter) on the build surface, and estimating the thickness of these deposits, it is concluded that about twice as much oxidized spatter is deposited on the part surface (during building of each layer) than the amount of oxygen incorporated into the part from this source. A possible reason for Manuscript received November 1, 2019; accepted for publication February 24, 2020. 1 Dept. of Materials Science and Engineering, Carnegie Mellon University, 5000 Forbes Ave., Pittsburgh, PA http://orcid.org/0000-0002-7111-1627, P. C. P. http://orcid.org/0000-0002-296615213, USA, M. T. 1879 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

45

46

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

this difference is that spatter oxides might be partially removed from the part surface during recoating. Keywords oxide inclusions, spatter, oxidation, fatigue, porosity

Introduction This work concerns oxides in components produced by laser powder bed fusion (LPBF): how the oxides form and an example of how defects at oxides cause fatigue failure in AlSi10Mg parts. The focus is on the origins of the oxides, with the aim of lowering the oxide content to improve fatigue resistance. In previous work, the presence of oxides in AlSi10Mg and their effect on fatigue were reported; here, this previous work is summarized briefly and some previously unpublished micrographs of oxides in this AlSi10Mg are shown. Formation of oxides in UNS N07718 has not been reported in detail. In the current work, new information on oxides in UNS N07718 is presented, with detailed consideration of mechanisms leading to oxide formation. The oxygen content of LPBF components is high compared with wrought material. For example, analysis (by inert-gas fusion) of UNS N07718 (Inconel 718) components produced by LPBF (in the current work, under standard conditions listed later in this paper) showed that the components contain around 200 parts per million (ppm, by mass) of oxygen—much higher than the total oxygen content of modern bearing steels,1 which is around 5 ppm. Because of the low solubility of oxygen in solid alloys that contain strong oxide formers such as aluminum (Al), titanium (Ti), and magnesium (Mg), all the oxygen would be bound as oxides after solidification. An example of typical oxides found in UNS N07718 components is shown in figure 1. Microanalysis showed the major elements in these oxides to be aluminum and oxygen (O). Oxides can be detrimental to fatigue resistance in at least two ways: (1) Oxides can initiate fatigue cracks; the effects of oxide particles on fatigue resistance is much larger than that of pores of the same size.2 (2) In AlSi10Mg, pores were found to be associated with—and likely caused by—oxides; pores associated with oxides, at the sample surface, were found to be the nucleation sites for all fatigue failures in samples tested in uniaxial tension.3 Examples of pores associated with oxides in AlSi10Mg parts are shown in figure 2. In these backscattered electron images, the pores appear black, the matrix is light gray (with the cellular silicon structure the lightest), and the oxides are darker gray. The major elements in the oxides were found to be aluminum and oxygen, sometimes with a minor amount of magnesium. In addition to approximately equiaxed pores associated with the oxides, other defects are similar to “bifilms” found in aluminum castings.4 Bifilms form when aluminum oxide on the surface of the liquid alloy is mixed into the alloy during turbulent mold filling; because aluminum oxide is not wetted by the metal, the enfolded “dry surface film” constitutes a crack in the cast part.5 In LPBF of AlSi10Mg, porosity may form at oxides because .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

FIG. 1 Examples of oxides in a UNS N07718 part built by LPBF, as seen in backscattered electron micrographs of polished sections.

FIG. 2 Examples of pores (black regions) associated with oxides in AlSi10Mg built by LPBF, as seen in backscattered electron micrographs of polished sections. The white arrows indicate oxides, and the black arrows mark “bifilms.”

nonwetting of the oxides by the metal retards melting and consolidation of powder into the melt pool; oxides have also been shown to nucleate hydrogen porosity in aluminum castings.4 An example of a fatigue crack that grew from such a pore in AlSi10Mg is shown in figure 3. The sample was tested under cyclic axial loading; the test conditions .

47

48

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 3 Example of (A) a fatigue crack that grew from a pore (outlined with a broken line) at the surface of an AlSi10Mg sample tested in axial tension (secondary electron micrograph). The enlarged view in (B) is a backscattered electron micrograph showing the oxides associated with the pore.

have been reported previously.3 Figure 3B shows oxides associated with the pore. Preliminary results indicated that the fatigue life of such samples could be predicted with reasonable accuracy by using an extreme-value distribution to extrapolate the size of pores found on polished sections to the largest pores expected on the surface of fatigue samples.6 Pores can affect both crack formation and crack propagation. However, measurements on UNS N06625 builds showed little or no effect of pores on crack propagation for the typical porosity (around 1%) in parts produced by LPBF.7 In contrast, the effect of porosity on crack formation has been clearly demonstrated. For example, Sheridan et al.8 and Sheridan, Gockel, and Scott-Emuakpor9 demonstrated that the fatigue life of UNS N07718 specimens could be predicted reasonably by considering near-surface porosity. Such porosity was found to have a stronger effect on fatigue life than surface roughness.10 While the relationship among oxides, pores, and fatigue has been demonstrated for AlSi10Mg, the origin of the micron-sized oxides associated with pores has not been established. The work presented here evaluated spatter oxidation as a possible oxide source. First, the contribution of the native oxide film on the metal powder is considered briefly.

Possible Oxide Sources OXIDE FILM ON METAL POWDER

The metal powder used in LPBF is covered by an oxide film, which is typically several nanometers thick. For aluminum powder produced by inert-gas atomization, .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

the oxide film thickness is approximately 4 nm,11 which is similar to the thickness of the film formed on aluminum under ambient conditions.12 When the powder is melted, this surface oxide will not dissolve in the metal because the solubility of oxygen in liquid aluminum is very low—approximately 0.2 parts per billion (by mass).13 Rather, there are indications from the appearance of partially melted powder particles that the surface oxide de-wets from the molten metal, forming submicron aluminum oxide particles; an example is given in figure 4. Because of their small size, these oxide particles are not readily detected on polished sections but contribute significantly to the total oxygen concentration in the as-built parts, as evaluated here for UNS N07718. In the system consisting of the alloy powder and passive film, the mass fraction of oxygen is given by equation (1): powder

wO

¼ woxide qoxide Loxide Spowder O

(1)

powder

In this expression, wO is the mass fraction of oxygen in the component is the mass fraction of oxygen in that originated from the passive film, and woxide O the passive film—as given by stoichiometry. For example, for Cr2O3, ¼ ð3  16Þ=ð2  52 þ 3  16Þ, qoxide is the density of the oxide (values taken woxide O from Rumble14), Loxide is the thickness of the passive film, and Spowder is the specific area of the powder calculated using the size distribution shown in figure 5. For UNS NO7718, the contribution of the oxide on powder to total oxygen was assessed by taking the air-formed passive film on the powder particles to be 2.4 nm thick, which is the thickness on the similar alloy, Inconel 625.16 For that alloy, the

FIG. 4 Partially molten AlSi10Mg powder particle covered by fine surface oxides, possibly formed from the native oxide film that was present on the atomized powder before melting. The large dome-shaped oxide appears to be oxidized spatter (secondary electron micrograph).

.

49

50

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Number distribution of particle sizes in UNS NO7718 powder used to estimate the contribution of the passive film to the oxygen concentration in built parts (redrawn from a size distribution reported by Sudbrack et al.15).

composition (in molar percentages) of the passive film was measured as 60.5% NiO, 30.9% Cr2O3, and 8.6% MoO3.16 The resulting oxygen concentration in parts produced from UNS NO7718 powder was estimated by taking the oxide to be an ideal mixture of the pure oxides (i.e., the partial molar volumes are the same as that of pure oxides), with oxidation states as shown in the stated composition. The calculation used one of the alloy powder size distributions reported by Sudbrack et al.15 (for a powder used in LPBF) and redrawn in figure 5. In this figure, the particle sizes are shown as a number distribution.17 In the calculation, the powder particles were assumed to be spherical. Using these inputs, the contribution of the passive film to the oxygen content is estimated to be approximately 100 ppm. This estimate would be affected by the size distribution of the powder. However, the details of the composition of the film are less important—taking the passive film to be pure Cr2O3 (instead of the mixture detailed previously) changes the estimated oxygen content by less than 5%. Based on this estimate, approximately half of the oxygen in the completed parts is contributed by the passive film on the metal powder. SPATTER OXIDATION

A significant fraction of the top surfaces of as-built parts is covered by deposited oxide. Examples for UNS NO7718 are shown in figure 6. As for the other examples shown in this paper, these samples were produced in an EOS M290 instrument under standard building conditions: laser power of 285 W, scanning speed 0.96 m/s, beam diameter approximately 100 lm, and 40-lm layer thickness for UNS NO7718; the chamber atmosphere was argon containing 0.1% of oxygen. Previous work showed that spatter particles are covered by oxide.18 Here, the possibility is tested that the deposited oxide on the build surface originated from hot spatter droplets (molten metal) that were ejected from the melt pool, oxidized .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

FIG. 6 Examples of oxide deposited on the top surface of as-melted UNS NO7718 parts for (A) remelted plate (no powder) and (B) part built with powder. (In backscattered electron micrographs oxides appear darker than the underlying metal surface; note difference in magnification.)

during their trajectory through the chamber atmosphere, and deposited on the part surface. For this to be a valid mechanism, sufficient hot spatter droplets need to be produced to yield the observed oxide, the rate of oxidation must be high enough, .

51

52

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

and oxidation must be complete before the spatter cools. In addition, the amount of surface oxide would determine whether this oxide could be a significant source of oxide within the parts. Here, the latter three points are addressed in a preliminary way. We do not attempt to quantify whether sufficient spatter droplets are produced. However, hot spatter ejected from the melt pool is a well-known phenomenon,19 and it seems reasonable to propose that hot spatter could be the source of oxidized particles. Spatter formation depends on the melt-pool conditions: highspeed photography showed that the rate of spatter production is higher under keyholing conditions.20 The direction of ejection is also affected: under keyholing conditions, spatter was found to be directed backward along the scan path, whereas ejection was vertical or forward if keyholing did not occur.21 The UNS NO7718 alloy typically contains around 0.5% aluminum, which would be the first element to be oxidized because of the high affinity of aluminum for oxygen (titanium, typically 1% of the alloy, would be oxidized after most of the aluminum has been oxidized). The appearance of deposited oxide (fig. 6) indicates that the oxidized droplets would be several microns in diameter. For such small droplets, traveling at a speed of a few meters per second through the chamber atmosphere, mass transfer through the gas phase can be approximated by static diffusion, for which the mass transfer coefficient is given by:22 mgas ¼ Dgas =rdroplet

(2)

where: mgas is the mass transfer coefficient, Dgas is the binary diffusivity in the O2-Ar gas mixture, and rdroplet is the droplet diameter. This approximation is valid because of the small Reynolds number for metal droplets moving through the gas phase and the resulting weak effect on the mass transfer coefficient. In general, the mass transfer coefficient between a sphere and a relatively moving fluid is given by equation (3):23 Sh ¼ 2 þ 0:552 Re1=2 Sc1=3

(3)

where: Sh ¼ 2mgasrdroplet/Dgas is the Sherwood number, Re ¼ 2urdroplet/ gas is the Reynolds number (in which u is the speed of the gas relative to the sphere and  gas the kinematic viscosity of the gas), and Sc ¼  gas/Dgas is the Schmidt number of the gas. Because the typical speed is of the order of 1 m/s, and the typical spatter diameter is around 10 lm, the Reynolds number is small, giving Sh  2, which is the basis of equation (2). It is reasonable to assume that the interfacial reaction between oxygen in the gas phase and the aluminum in the alloy would be close to equilibrium, given the high temperatures involved. The calculation approach is similar to that used by other .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

workers to model the oxidation of dissolved aluminum from liquid steel at the steelgas interface.24 As also assumed in previous work, here the partial pressure of oxygen at the metal-gas interface is taken to be much lower than the oxygen pressure in the bulk gas because dissolved aluminum in the molten metal is a very effective getter of oxygen. In addition to gas-phase mass transfer as a rate-limiting step, transport of dissolved aluminum within the spatter droplet was also considered in this work. However, as shown by the results, aluminum transport was found not to be limiting. The oxygen activity at the droplet surface would be much lower than in the bulk of the gas, and the molar flux of oxygen to the droplet surface would be given by equation (4): JO2 ¼

mgas pO2 Dgas pO2 ¼ RTfilm rdroplet RTfilm

(4)

where: pO2 is the partial pressure of oxygen in build chamber, R is the ideal gas constant, and T is the absolute film temperature. The rate of mass transfer of aluminum within the droplet to the droplet surface (to react with oxygen from the gas phase) is similarly given by equation (5):25 JAl ¼

 DAl qalloy  ½%Albulk  ½%Alsurface , 20rdroplet MAl

(5)

where: DAl is the diffusivity of Al in the liquid metal, qalloy is the alloy density (7,388 kg/m3, value from Mills26), MAl is the molar mass of Al (0.027 kg/mol), and [%Al] is the mass percentage of Al in the alloy. If mass transfer of Al is rate-determining, its surface concentration would approach zero. At a film temperature of approximately 690 C (the average of the alloy melting point of 1,350 C and chamber temperature, taken to be 30 C), the diffusivity in the gas phase is calculated to be approximately 1.5 cm2/s (using the correlations for binary gas diffusivity27). The diffusivity of Al in the alloy is approximately 3  105 cm2/s (utilizing Thermo-Calc Software TCS Ni-alloys Mobility Database28). Using these values, together with the stoichiometric requirement that the molar flux of O2 is 0.75 times the molar flux of Al, it is concluded that the rate of oxidation of droplets would be limited by transfer of oxygen to the droplets, until the aluminum concentration drops below 0.06% (at a chamber pressure of 1 atm and oxygen concentration of 0.1%). Given that the initial aluminum concentration in the alloy is 0.5%, this means that spatter oxidation is largely controlled by the oxygen concentration in the chamber: a higher oxygen concentration would cause more rapid oxidation. More detailed analysis (not presented here) indicates that oxidation is rapid at the 0.1% oxygen concentration, with 90% of the aluminum oxidized in 4 ms, from a UNS NO7718 droplet with a diameter of .

53

54

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

10 lm. Smaller droplets would be oxidized more quickly—in general, for reactions under mass transfer control, the oxidation time is proportional to the square of the droplet diameter. A second test of the feasibility of spatter oxidation is whether the droplets would remain molten long enough for extensive oxidation to occur. Smaller spatter droplets would tend to cool rapidly by radiative heat transfer. However, the exothermicity of the oxidation reaction would counteract cooling. Because the oxidation rate is controlled by gaseous diffusion, the ratio of heat generation by the oxidation reaction to heat loss by radiation can be written as follows: qreaction =qrad ¼ 

4Dgas pO2 DHreaction   3reRTfilm Tp4  Ts4 rdroplet

(6)

where: qreaction is the heat generation rate by oxidation, qrad is the heat loss rate by radiation, DHreaction is the heat of reaction (approximately 700 kJ per mol Al, taking into account the heat of solution of Al in the alloy), r is the Stefan-Boltzmann constant, e is the emissivity of the liquid droplet surface (approximately 0.36 as reported by Pottlacher et al.29), Tp is the surface temperature of the oxidizing droplet (taken to be at least 1,350 C to remain liquid), and Ts is the temperature of the surroundings (taken to be 25 C). Substitution of these values in equation (6) shows that the rate of heat generation by oxidation would exceed the rate of heat loss by radiation for all droplets with diameters smaller than approximately 24 lm: droplets smaller than this would heat up during oxidation. Equation (6) also shows that, at higher oxygen concentrations in the chamber, larger droplets can be oxidized: the droplet size below which net heating occurs during oxidation is proportional to the oxygen concentration in the chamber gas. This is in line with the observation that much more oxidation of spatter from laser melting of a titanium alloy was observed when the oxygen concentration was higher, even though the amount of spatter produced depended only weakly on the oxygen concentration.30 Based on the analysis presented here, a higher oxygen concentration in the chamber would cause oxidation of a larger portion of the size distribution of spatter droplets, thereby causing more oxidation overall. The other factor controlling the amount of oxidation would be the rate of generation of spatter droplets. In line with this suggestion, it was found in the current work that the surface concentration of oxides (of the kind shown in fig. 6) is approximately three times as large for samples melted under keyholing conditions (laser power of 285 W and speed of 0.48 m/s) than without keyholing; the stronger stirring and deeper melt pool associated with keyholing are expected to produce more spatter. .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

SURFACE CONCENTRATION OF OXIDES

As discussed so far, the surfaces of as-built parts show considerable coverage by oxides that appear to originate from hot spatter that was ejected from the melt pool. If these oxides on the surface of each layer were to be incorporated during deposition of the next layer, this would cause oxygen pickup in the build. The purpose of the analysis presented here is to assess whether this would contribute significantly to the oxygen concentration in parts. The extent of such oxygen pickup can be estimated from the measured surface coverage and the estimated thickness of the oxides. In scanning electron micrographs (backscattered electron [BSE] mode), the oxides vary in brightness, indicating a range of thicknesses (thicker oxide layers would appear darker). The thickness was estimated by measuring the energy-dispersive X-ray (EDX) spectrum when placing the electron beam on representative oxides—oxide regions that were the darkest, the brightest (but still darker than the surrounding metal surface), and intermediate in brightness. The ratio of the Ka peak heights of Al and nickel (Ni) was measured and converted to an estimated thickness. The thickness estimate relied on Monte Carlo simulations (performed with DTSA-II Lorentz31) of a uniform Al2O3 layer (density: 3.99 kg/dm3) on a flat UNS NO7718 substrate (density and composition as listed by Mills26). The surfaces of the metal and oxide were taken to be perpendicular to the beam. It was found that an accelerating voltage of 20 kV gives good sensitivity over the oxide thickness range encountered on these samples. EDX spectra were measured at this voltage, using a silicon drift detector (window: Moxtek AP3.3) at an elevation of 35 .

FIG. 7 Ratio of Al Ka to Ni Ka counts for simulated EDX spectra from a dense alumina layer on a UNS NO7718 substrate at an accelerating voltage of 20 kV. The dotted line is the fitted relationship used to analyze the measured spectra.

.

55

56

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 8 Examples of (A) measured and (B) simulated EDX spectra (20 kV accelerating voltage) for Al2O3 layers with different thicknesses on a UNS NO7718 substrate.

TABLE 1 Estimated thicknesses (in μm) of oxide deposited on build surface for oxides judged to be thin, intermediate in thickness, or thick (based on BSE images)

Thin

0.15; 0.17; 0.2

Intermediate

Thick

0.28; 0.33; 0.67

0.87; 0.99; 1.28

Results from the simulations are summarized in figure 7, showing good sensitivity of the Al/Ni count ratio to the thickness of alumina, over the thickness range encountered in this work. The examples of measured and simulated spectra in figure 8 are for thicknesses at the extremes of the estimated range: 0.2 to 1.0 lm. The figure illustrates the similarity between the measured and simulated spectra; the main differences are that the oxide did contain some titanium (whereas the simulated oxide was pure Al2O3), and that the ubiquitous carbon peak is present in the measurements. .

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

TABLE 2 Calculated oxygen pickup in built part if all the spatter oxide on the build surface is incorporated

Fraction of build surface covered by oxide

¼

10%

Average thickness of oxide

¼

0.5 :m

Density of oxide (Al2O3)

¼

3,990 kg/m3

Density of alloy (UNS NO7718)

¼

8,190 kg/m3

Layer thickness (of built part)

¼

40 :m

Mass fraction of oxygen in build

¼

0.000287

¼

287 ppm

The estimated oxide film thickness for regions covered by thin, thick, and intermediate oxide (based on BSE brightness) are summarized in table 1. Three representative regions (on three different samples) were measured. The results show that the thicknesses range from approximately 0.2 lm (thin regions) to around 1 lm (thick oxide). Based on these measurements, the average oxide thickness can be taken to be in the vicinity of 0.5 lm. This thickness, together with the surface coverage, determines the mass of oxide deposited on each build layer. The surface coverage of oxides would depend on several factors: the spatter production rate (higher for keyholing conditions), the extent of oxidation of spatter (greater if the oxygen concentration in the chamber is higher), and possible removal of oxides from the build surface during recoating. In this work, the surface coverage was typically 10%. Using this coverage, together with the estimated average thickness, the amount of oxygen imparted to the build by the deposited spatter oxide was estimated. The estimate—shown in table 2—is around 300 ppm. This is more than the expected extent of oxygen pickup from the oxidized spatter; as mentioned earlier, the total oxygen in UNS NO7718 components (built under the conditions considered here) is around 200 ppm, of which 100 ppm is estimated to originate from the oxide on the metal powder. The mismatch is between the estimated amount of oxygen that can be picked up from the oxidized spatter on the sample surface (200 to 300 ppm) and the amount from this source as estimated from the total oxygen in the components and oxygen from powder (around 100 ppm). The mismatch indicates that at least some of the oxidized spatter is removed from the surface before the next layer is deposited; it is speculated that interactions among the recoater blade, powder, and the sample surface may contribute to removing some of the oxidized spatter.32

Conclusion Oxides are prominent and potentially important defects in LPBF parts. The work presented here shows that both the oxide film on the metal powder (the raw material for LPBF) and spatter oxidation during part building contribute significantly to the total oxygen content in LPBF parts. The main conclusions are as follows: .

57

58

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

•

•

•

•

Pores in AlSi10Mg builds nucleate fatigue cracks. The pores are associated with oxides, some which have a similar “bifilm” morphology to that found in conventional aluminum castings. The rate of oxidation of UNS N07718 spatter droplets is controlled by mass transfer of oxygen in the gas surrounding spatter. The exothermicity of oxidation prevents cooling of UNS N07718 spatter for spatter droplets smaller than a critical size. This size is larger if the oxygen concentration in the build chamber is higher, leading to more oxidation. From these results, two ways to limit the oxygen content in builds are to use coarser powder (with a smaller specific area and, hence, less surface oxide relative to the mass) and to decrease the oxygen concentration in the build chamber.

ACKNOWLEDGMENTS

The authors would like to acknowledge use of the Materials Characterization Facility at Carnegie Mellon University under Grant No. MCF-677785. This work was supported by an Early Stage Innovations grant from NASA’s Space Technology Research Grants Program (Grant No. NNX17AD03G).

References 1.

2.

3.

4.

5.

6.

7.

.

¨ lund, “A Review of Swedish Bearing Steel T. I. Johansson, T. B. Lund, and P. L. J. O Manufacturing and Quality Assurance of Steel Products,” Journal of ASTM International 3, no. 10 (2006): 1–13, https://doi.org/10.1520/JAI14022 G. Le ´opold, Y. Nadot, T. Billaudeau, and J. Mendez, “Influence of Artificial and Casting Defects on Fatigue Strength of Moulded Components in Ti-6Al-4V Alloy,” Fatigue & Fracture of Engineering Materials & Structures 38 (2015): 1026–1041, https://doi.org /10.1111/ffe.12326 M. Tang and P. C. Pistorius, “Oxides, Porosity and Fatigue Performance of AlSi10Mg Parts Produced by Selective Laser Melting,” International Journal of Fatigue 94 (2017): 192–201, https://doi.org/10.1016/j.ijfatigue.2016.06.002 T. Ludwig, M. Di Sabatino, L. Arnberg, and D. Dispinar, “Influence of Oxide Additions on the Porosity Development and Mechanical Properties of A356 Aluminium Alloy Castings,” International Journal of Metalcasting 6, no. 2 (April 2012): 41–50, https:// doi.org/10.1007/BF03355526 J. Campbell, “An Overview of the Effects of Bifilms on the Structure and Properties of Cast Alloys,” Metallurgical and Materials Transactions B 37 (2006): 857–863, https:// doi.org/10.1007/BF02735006 M. Tang and P. C. Pistorius, “Fatigue Life Prediction for AlSi10Mg Components Produced by Selective Laser Melting,” International Journal of Fatigue 125 (2019): 479–490, https://doi.org/10.1016/j.ijfatigue.2019.04.015 J.-R. Poulin, A. Kreitcberg, P. Terriault, and V. Brailovski, “Long Fatigue Crack Propagation Behavior of Laser Powder Bed-Fused Inconel 625 with Intentionally Seeded Porosity,” International Journal of Fatigue 127 (2019): 144–156, https://doi.org/10.1016 /j.ijfatigue.2019.06.008

OHTSUKI ET AL., DOI: 10.1520/STP163120190137

8.

9.

10.

11.

12.

13. 14. 15.

16.

17.

18.

19.

20.

21. 22.

.

L. Sheridan, O. E. Scott-Emuakpor, T. George, and J. E. Gockel, “Relating Porosity to Fatigue Failure in Additively Manufactured Alloy 718,” Materials Science and Engineering: A 727 (2018): 170–176, https://doi.org/10.1016/j.msea.2018.04.075 L. Sheridan, J. E. Gockel, and O. E. Scott-Emuakpor, “Primary Processing Parameters, Porosity Production, and Fatigue Prediction for Additively Manufactured Alloy 718,” Journal of Materials Engineering and Performance 28 (2019): 5387–5397, https:// doi.org/10.1007/s11665-019-04305-7 J. Gockel, L. Sheridan, B. Koerper, and B. Whip, “The Influence of Additive Manufacturing Processing Parameters on Surface Roughness and Fatigue Life,” International Journal of Fatigue 124 (2019): 380–388, https://doi.org/10.1016/j.ijfatigue.2019.03.025 J. Joys, “Production of Aluminum and Aluminum-Alloy Powder,” in ASM Handbook, Volume 7: Powder Metallurgy, ed. P. Samal and J. Newkirk (Metals Park, OH: ASM International, 2015), 569–580, https://doi.org/10.31399/asm.hb.v07.a0006065 M. R. Alexander, G. E. Thompson, and G. Beamson, “Characterization of the Oxide/Hydroxide Surface of Aluminium Using X-Ray Photoelectron Spectroscopy: A Procedure for Curve Fitting the O 1s Core Level,” Surface and Interface Analysis 29 (2000): 468–477, https:// doi.org/10.1002/1096-9918(200007)29:73.0.CO;2-V H. A. Wriedt, “The (Al-O) (Aluminum-Oxygen) System,” Bulletin of Alloy Phase Diagrams 6, no. 6 (1985): 548–553, https://doi.org/10.1007/BF02887157 J. R. Rumble, Jr., ed., CRC Handbook of Chemistry and Physics, 100th Ed. (Boca Raton, FL: CRC Press, 2019). C. K. Sudbrack, B. A. Lerch, T. A. Smith, I. E. Locci, D. L. Ellis, A. C. Thompson, and B. Richards, “Impact of Powder Variability on the Microstructure and Mechanical Behavior of Selective Laser Melted Alloy 718,” in Proceedings of the 9th International Symposium on Superalloy 718 & Derivatives: Energy, Aerospace, and Industrial Applications, ed. E. Ott, X. Liu, J. Andersson, Z. Bi, K. Bockenstedt, I. Dempster, J Groh, K. Heck, P. Jablonski, M. Kaplan, D. Nagahama, and C. Sudbrack (Cham, Switzerland: The Minerals, Metals & Materials Series, Springer, 2018), 89–113, https://doi.org/10.1007/978-3-319-89480-5_5 A. C. Lloyd, J. J. Noe ¨l, S. McIntyre, and D. W. Shoesmith, “Cr, Mo and W Alloying Additions in Ni and Their Effect on Passivity,” Electrochimica Acta 49 (2004): 3015–3027, https://doi.org/10.1016/j.electacta.2004.01.061 Standard Practice for Calculation of Mean Sizes/Diameters and Standard Deviations of Particle Size Distributions, ASTM E257807(2018) (West Conshohocken, PA: ASTM International, approved January 1, 2018), https://doi.org/10.1520/E2578-07R18 M. Simonelli, C. Tuck, N. T. Aboulkhair, I. Maskery, I. Ashcroft, R. D. Wildman, and R. Hague, “A Study on the Laser Spatter and the Oxidation Reactions during Selective Laser Melting of 316L Stainless Steel, Al-Si10-Mg, and Ti-6Al-4V,” Metallurgical and Materials Transactions A 46 (2015): 3842–3851, https://doi.org/10.1007/s11661-015-2882-8 S. A. Khairallah, A. T. Anderson, A. Rubenchik, and W. E. King, “Laser Powder-Bed Fusion Additive Manufacturing: Physics of Complex Melt Flow and Formation Mechanisms of Pores, Spatter, and Denudation Zones,” Acta Materialia 108 (2016): 36–45, https:// doi.org/10.1016/j.actamat.2016.02.014 H. Zheng, H. Li, L. Lang, S. Gong, and Y. Ge, “Effects of Scan Speed on Vapor Plume Behavior and Spatter Generation in Laser Powder Bed Fusion Additive Manufacturing,” Journal of Manufacturing Processes 36 (2018): 60–67, https://doi.org/10.1016 /j.jmapro.2018.09.011 P. Bidare, I. Bitharas, R. M. Ward, M. M. Attallah, and A. J. Moore, “Fluid and Particle Dynamics in Laser Powder Bed Fusion,” Acta Materialia 142 (2018): 107–120. J. Szekely, J. W. Evans, and H. Y. Sohn, Solid-Gas Reactions (New York: Academic Press, 1976).

59

60

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

23.

24. 25. 26. 27. 28.

29.

30.

31.

32.

.

R. L. Steinberger and R. E. Treybal, “Mass Transfer from a Solid Soluble Sphere to a Flowing Liquid Stream,” AIChE Journal 6 (1960): 227–232, https://doi.org/10.1002 /aic.690060213 K. Sasai and Y. Muzikama, “Reoxidation Behavior of Molten Steel in Tundish,” ISIJ International 40 (2000): 40–47, https://doi.org/10.2355/isijinternational.40.40 E. Glueckauf, “Formulae for Diffusion into Spheres and Their Application to Chromatography,” Transactions of the Faraday Society 51 (1955): 1540–1551. K. N. Mills, Recommended Values of Thermophysical Properties for Selected Commercial Alloys (Cambridge, UK: Woodhead Publishing, 2002). R. B. Bird, E. N. Lightfoot, and W. E. Stewart, Transport Phenomena, Revised 2nd Ed. (New York: Wiley, 2007). J.-O. Andersson, T. Helander, L. Ho ¨glund, P. Shi, and B. Sundman, “Thermo-Calc & DICTRA, Computational Tools for Materials Science,” Calphad 26 (2002): 273–312, https:// doi.org/10.1016/S0364-5916(02)00037-8 G. Pottlacher, H. Hosaeus, E. Kaschnitz, and A. Seifter, “Thermophysical Properties of Solid and Liquid Inconel 718 Alloy,” Scandinavian Journal of Metallurgy 31 (2002): 161–168, https://doi.org/10.1034/j.1600-0692.2002.310301.x H. Amano, Y. Yamaguchi, T. Sasaki, T. Sato, T. Ishimoto, and T. Nakano, “Effect of Oxygen Concentration on the Generation of Spatter during Fabrication via Selective Laser Melting,” Journal of Smart Processing 18 (2019): 102–105, https://doi.org/10.7791 /jspmee.8.102 National Institute of Standards and Technology, “DeskTop Spectrum Analyzer-II (DTSA-II): Introduction,” http://web.archive.org/web/20191030175906/https://cstl.nist.gov/div837/ 837.02/epq/dtsa2/ H. Gong, K. Rafi, H. Gu, T. Starr, and B. Stucker, “Analysis of Defect Generation in Ti–6Al– 4V Parts Made Using Powder Bed Fusion Additive Manufacturing Processes,” Additive Manufacturing 1 (2014): 87–98, http://dx.doi.org/10.1016/j.addma.2014.08.002

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190160

Adam J. Brooks,1 Arushi Dhakad,1 Agustin Diaz,2 and Daniel Kowalik1

Toward Understanding the Role of Surface Texture for Additively Manufactured Metal Parts Citation A. J. Brooks, A. Dhakad, A. Diaz, and D. Kowalik, “Toward Understanding the Role of Surface Texture for Additively Manufactured Metal Parts,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 61–87. http://doi.org/10.1520/STP1631201901603

ABSTRACT

In the interest of evaluating as-built and finished surfaces of additively manufactured (AM) parts, the surfaces of AM parts must be better understood. The variability in surface quality of an AM produced part, both before and after surface finishing, has made it difficult to standardize reporting methodologies for measurement and characterization. While previous standards provide a starting point for measurement, the surfaces of metal powder bed fusion (MPBF) AM parts vary greatly from conventionally machined and formed parts. Recent work at EWI through the ASTM Additive Manufacturing Center of Excellence is concentrating on developing a standardized guide for measuring surface texture, part characterization, and metrics of AM parts. While this guide is currently under development, here we review the current state of surface texture with respect to AM, its analysis, and we describe the associated challenges. Areas of focus include measurement, analysis, and application-based issues such as comparing metrics across measurement techniques (contact stylus profilometry, laser confocal, focus variation, coherence scanning interferometry), knowing the appropriate bandpass filters to use for analysis, and how to incorporate inspection into AM part design. Mainly, there is a large disconnect among the AM, surface metrology, materials science, and application realms in regards to Manuscript received December 6, 2019; accepted for publication February 20, 2020. 1 EWI, 683 Northland Ave., Buffalo, NY 14211, USA 2 REM Surface Engineering, 2107 Longwood Dr., Brenham, TX 77833, USA 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

61

62

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

optimal AM surface specifications, indicating a high need for collaborative efforts among these groups. Keywords additive manufacturing, metal powder bed fusion, surface roughness, surface texture, surface measurement parameters, contact stylus profilometry, focus variation, laser confocal microscopy, coherence scanning interferometry, surface finishing

Introduction When it comes to surface measurement, new materials, processes, and geometric designs can disrupt the previously understood methodologies and standards. Traditional manufacturing relies on previously developed standards for all aspects of the process, including, but not limited to, component design, surface measurement, and final inspection. Metal additive manufacturing (AM) technology has revolutionized metal component manufacturing via the reduction of operation steps, scrap material, component lead time, and component cost.1–3 This indicates that many of the current AM research efforts may need to be altered in the future as qualification and certification requirements become established. AM has a profound impact on part design, allowing for optimization of component characteristics, such as shape and weight, that would not be possible or viable with traditional subtractive manufacturing techniques. Starting with an intended sample geometry, feedstock material (e.g., powder) is fed into an AM system on a layer-by-layer basis for part production. There are seven AM processes for generating parts: (1) vat photopolymerization, (2) powder bed fusion (PBF), (3) binder jetting, (4) material jetting, (5) sheet lamination, (6) material extrusion, and (7) directed energy deposition. With each of these seven processes, the surface topography of the as-built part will vary from process to process depending upon a multitude of factors including feedstock, equipment, and parameters. When describing surfaces, we must first understand what we are looking at. Several recent manuscripts and book chapters have provided background on surface metrology, topography, and surface texture.4,5 Surface metrology describes the measurement and characterization of surface topography, where topography incorporates all of the information from a surface’s shape (form) and features.6 In the measurement world, surface metrology plays a key role in knowledge of detecting features on AM parts. With the industry-altering potential of AM comes inherent complications: extremely high surface textures, surface defects, and near-surface defects, which are typical of AM manufacturing techniques.4,7–9 These surface texture characteristics are exhibited in both powder and wire-based AM, as well as in laser and electron beam-based building styles. Additional processing is typically needed to improve the as-built surface quality via finishing methods such as electrochemical polishing, .

BROOKS ET AL., DOI: 10.1520/STP163120190160

chemical polishing, and mechanical abrasive finishing. The end goal of optimizing surface finishing is to significantly improve the surface and, consequently, the mechanical performance of AM components. When looking at the mechanical performance needs of an application, the behavior of the material in question needs to be fully understood. Mechanical properties such as hardness, fatigue, and yield strength depend upon the material properties of the part. In the case of structural integrity, the structure property relationship between a part’s microstructure and its fatigue strength can be correlated to one another.10 While alloys have been developed for AM parts that result in superior mechanical properties, the “layer-by-layer” building of AM allows for defects to be incorporated into the part.11 This can result in surface defects that are detected using surface measurement techniques, internal defects near the surface that show up as pits during surface finishing processes, and a reduction of the final part mechanical properties from porosity.12 Given that there are AM, materials science, and surface metrology realms, these areas should be geared toward an end-use application. If the requirements for an application are more stringent than reported out in each of the methods, there will be no value in performing research efforts in those individual fields. For example, AM allows fabrication of complex parts, and conformal internal channels become easier to produce, but inspecting the quality of these internal surfaces becomes more difficult. If the channels need to have a smooth finish for fatigue application, the longevity and mechanical performance will influence the surface measurement, surface finish, and AM design requirements. Figure 1 shows how the AM, material science, surface metrology, and end-use application worlds collide. Currently, there are zero standards or guidelines on how to verify an AM part with regard to surface measurement and analysis. Many ISO and ASME standards exist for manufactured parts; however, these standards are geared for traditional manufacturing processes and not AM.13–15 While there are several research efforts currently in progress for AM, these are focused in the metrology realm, application realm, materials science realm, or AM realm. This is introducing new challenges that have not yet been standardized, thus creating a wide range of measurement, parameter, and process variability from machine to machine, part to part, and application to application. Relevant work attempting to bridge these gaps is underway in the ASTM AM CoE and ASME PROJECT TEAM 53, where standards (ASME B46.1)16 and guidelines (ASTM) are being developed for surface finish and measurement techniques geared toward additive manufacturing. In this manuscript, we provide a look at: 1. Terminology for describing surface texture 2. Contributions of AM and surface finishing to applications 3. Current challenges connecting AM, metrology measurement techniques, materials science, and end-use applications

.

63

64

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 Venn diagram showing how AM, material science, surface metrology, and enduse application ties together. The end use of the application along with the manufacturing method determines the part geometry. The manufacturing method and the aspects of material science influence the structural capabilities of the part. The inspection system (surface metrology) is influenced by the chemistry of the material and the method. Lastly, the parameters reported for surface metrology will influence and be influenced by the end-use application.

Surface Texture To understand how surfaces are generated in an AM process, surface texture must first be understood. When looking at a surface, that surface’s texture is the “scalelimited surface remaining after a series of operations are applied to the primary extracted surface.”13 In this regard, scale can be thought of on the macro or micro level. On the macro level, scale can refer to form or waviness; whereas on the micro level, scale can refer to roughness. In most simplistic terms, the surface texture can be defined as the combination of three main components: form, waviness, and roughness. .

BROOKS ET AL., DOI: 10.1520/STP163120190160

FIG. 2 How a surface profile is generated and can vary across a material surface. The surface profile is made up of form, roughness, and waviness features. Copyright ISO. This material is reproduced from ISO 4287:1999, figure 2, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.14

For a given surface, a surface profile including form, waviness, and roughness can be extracted from a region of interest on the sample using either twodimensional (2D) (contact) or three-dimensional (3D) (areal) techniques. Figure 2 shows an example of a profile extraction from an areal surface measurement. Given that a profile will include a sample’s roughness, waviness, and form, a filtering operation should be performed to separate the roughness and waviness components. Figure 3 shows how roughness and waviness can vary depending upon the wavelength of the spatial frequency filters applied. Small-scale filters, large-scale filters, and form filters all play a key role in knowing the true nature of a surface. With many different measurement techniques and characterization paths to acquire and analyze surface data, for a given set of raw data, the underlying form can be filtered out for a statistically significant sample size.13,14 After form removal, the resulting data can be further analyzed using a set of scalar parameters to measure roughness or waviness. In the case of filtering, there are requirements for both areal and contact methods as defined in ISO 25178-2 and ISO 25178-3 (table 1).13,15 The maximum sampling distance is calculated as a 5:1 ratio, although sometimes .

65

66

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 3 Separation of surface features using spatial frequency filters; ks refers to a small-scale filter, kc refers to a large-scale filter, and kf refers to a form filter. Copyright ISO. This material is reproduced from ISO 4287:1999, figure 1, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.14

TABLE 1 The numerical relationship between the roughness cutoff wavelength kc, the tip radius, and roughness cutoff ratio

kc (lm)

ks (lm)

kc/ks

rtip (lm)

Maximum Sampling Spacing (lm)

0.08

2.5

30

2

0.5

0.25

2.5

100

2

0.5

0.8

2.5

300

2a

0.5

2.5

8

300

5b

1.5

8

25

300

10b

5

a

For surfaces with Ra > 0.5 lm, rtip ¼ 5 lm can usually be used without significant differences in the measurement result. b For cutoff wavelengths ks of 2.5 lm and 8 lm, it is almost certain that the attenuation characteristic resulting from the mechanical filtering of a stylus with the recommended tip radius will lie outside the defined transmission band. Since this is the case, a small variation in stylus radius or shape will have negligible effect on the values of parameters calculated from the measured profile. If another cutoff ration is deemed necessary to satisfy the application, this ration must be specified. Copyright ISO: This material is reproduced from ISO 3274:1998-table 1, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.17

these distances may not be able to be achieved given certain geometries, orientations, or features. ISO 25178-2 (fig. 4) provides a figure representing the relationship between small-scale S filters, large-scale L filters, and form F filters.13 Relationships between S-filter nesting index value, sampling distance, and the .

BROOKS ET AL., DOI: 10.1520/STP163120190160

FIG. 4 Figure showing the relationship between the S-filter, L-filter, F-operation, and S-F and S-L surface. Copyright ISO. This material is reproduced from ISO 251782:2012, figure 1, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.13

lateral period limit for optical surface can also be found in ISO 25178-3 for areal techniques.15 Once the results have been filtered according to the existing standards, the appropriate parameters need to be determined for reporting out results. There are many options to choose from when considering surface texture parameters. ISO specification standards 4287 and 25178-2 define most of the commonly used parameters for both profile measurements14 and areal measurements.13 The most commonly used parameter is Ra, which is the arithmetic mean deviation of a profile.4,18–21 Equation (1) refers to how to calculate Ra for a given surface. ðL 1 Ra ¼ jZðxÞjdx L

(1)

0

In equation (1), Z(x) is the height of the assessed profile at any position x, and L is the sampling length. The sampling length refers to the measurement length of the probe in the x-axis direction. Other common parameters include Rq, the root mean square deviation of a profile; Rz, the maximum height of the profile; Rv, the distance between the deepest valley and the mean line within the evaluation length; Rp, the peaks from the profile; and Rt, the total height of the profile. Figure 5 shows how to calculate the parameters Ra, Rz, and Rt.22,23 .

67

68

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Sample line profile exhibiting how Ra, Rz, and Rt are determined.

Two-dimensional measurements and parameters do not represent a highly variable AM surface, as only one line of the surface is represented at a time. This is exemplified in figure 2, where only one profile is extracted from a large surface area. In 3D measurements, a larger area of the surface can be measured and analyzed to gain more representation of the true form, roughness, and waviness of the surface. To mathematically represent these areas, areal parameters have been established (ISO 25178-3), and several good practice guides for areal techniques have also been written.15,24–26 Sa is the most commonly used areal parameter, defined by equation (2): Sa ¼

1 A

ðð

jZ ðx; yÞjdxdy

(2)

A

where A is the area of measurement and Z(x,y) is the height of the peaks/valleys in the area. Similar to 2D R parameters, there are analogous areal S parameters such as Sq, Sz, Sv, and Sp that provide surface details. Skewness (Rsk, Ssk) refers to a proportion of material above or below the mean line/plane (fig. 6). Kurtosis (Rku, Sku) is another parameter that describes the “spikiness” of a 2D or 3D surface. In the ISO 25178–2 document, there are several field parameters to characterize spatial information, functional information, and feature parameters.13 Other efforts by Leach27 and others have focused on parameters defined from a subset of predefined topographic features from the scale-limited surface. The most common technique for measuring surfaces is the contact stylus instrument (fig. 7). A recent review determined that roughly 40% of examined literature regarding surface measurement used a stylus-based contact instrument.4 Given their relatively low cost, stylus instruments are common in the industrial workplace. A contact stylus instrument works on the premise that the vertical motion of a probe as it moves along a surface being measured is converted into an electrical .

BROOKS ET AL., DOI: 10.1520/STP163120190160

FIG. 6 Skewness curve measuring the asymmetry of the measured profile. ADF represents the probability density function that displays the probability of a certain height Z (x) for a given evaluation length.5

FIG. 7 Schematic representation of a contact stylus measurement system. The stylus moves a spherical probe across the surface, capturing the topography of the surface.

signal via a transducer. The contact probe is moved at a constant speed and measures the height of the probe tip at regular intervals. The probe tip can be as small as 2 lm and is generally made of diamond. The limitations of a contact technique are due to the stylus probe tip size, where a large probe tip will not be able to resolve .

69

70

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

small features (“stylus flight”). Another limitation is that the probe tip can cause damage or can be damaged by the surface. There are multiple publications regarding the theory, measurement, and analysis behind contact instruments.19,28,29 National Physical Library (NPL) Good Practice Guide 37 provides an in-depth overview of how to take surface measurements using a contact stylus instrument; however, this is not geared toward AM surfaces.30 Another common technique is confocal microscopy (CM). During this technique, a beam of light is focused at an objective lens onto a small area of a sample. The reflected light returns through the system’s optics and is imaged onto a second pinhole where the intensity of the light is captured.31 The objective lens is then stepped along the z axis to build up a complete height map of the surface. Figure 8 shows a typical confocal microscope.

FIG. 8 Setup showing confocal imaging. A light source travels through a pinhole to a focus spot on a surface. The spot is then imaged onto the discrimination pinhole. A confocal system is characterized by two conjugate pinholes where light passes through the objective twice in opposite directions, and the setup is coaxial. Copyright ISO. This material is reproduced from ISO 25178-602:2010, figure B.1, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.32

.

BROOKS ET AL., DOI: 10.1520/STP163120190160

The coherence scanning interferometer (CSI) is another 3D surface topology method that uses a light source and measures the reflected light from the surface it is measuring. Two beams of light are generated from one light source where an interference pattern is detected by combining a reference path with a sample path of light at the detector. The instrument is scanned in the Z axis, where interference fringes are detected and measured as a sample is moved into focus for each pixel. Once the sample is out of focus, the interference patterns are no longer present and thus cannot be resolved. Figure 9 shows a CSI setup including the interference fringes when a sample is in and out of focus.9,26,33,34 Another technique, focus variation (FV), combines the small depth of focus of a microscopic optical system with vertical scanning to provide topographical and color information from the variation of focus.36,37 The main component of the FV instrument is a precision optical unit containing various lenses. FV operates by stacking images to detect in-focus points and generate a surface height corresponding to each pixel. This technique can be affected by how light is reflected back into

FIG. 9 Conceptual drawing of the operation of a CSI instrument showing the interference fringe pattern on a curved object for three different successive scan positions (A), (B), and (C). Copyright ISO. This material is reproduced from ISO 25178-604:2013, figure B.1, with permission of the American National Standards Institute (ANSI) on behalf of the International Organization for Standardization. All rights reserved.35

.

71

72

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

the detector. For example, a smooth surface may run into issues due to a lack of required local roughness to create a dominant peak in the focus curve, resulting in faulty height maps. Further explanations describing the focus variation technique can be seen in ISO 25178-602:2015.32 Figure 10 shows the technical workings behind an FV instrument.38 Recent efforts into another technique have been established to use X-ray computed tomography (XCT) for areal surface texture measurement. XCT works by measuring the attenuation across a sample and is dependent upon the atomic elements of the material. Equispaced projections are acquired at increments around the sample and, by combining the 2D projections acquired at each angle, the sample can then be reconstructed into 3D volume using various reconstruction and visualization software. While XCT has shown advantages for detecting surface roughness, the technique is still in its infancy due to the necessary high resolution for roughness scans. Several other techniques not described here previously have also been adopted for surface measurement. Included in this are chromatic confocal microscopy,39 conoscopic holography,40 atomic force microscopy (AFM),41 elastomeric sensor,42–44 optical microscopy,19,41 scanning electron microscopy (SEM),18 and Raman spectroscopy.45 Optical techniques, SEM, and Raman measurements do not give topographical information. Regarding the elastomeric sensor, this may have applicability; however, this is in the initial stage of research and needs further studies performed to use it as a valuable tool in AM surface measurement. Chromatic confocal does not

FIG. 10 Schematic diagram of the focus variation technology.

.

BROOKS ET AL., DOI: 10.1520/STP163120190160

provide as much value as other systems for measuring surfaces properly.25 These techniques are further described in the referenced publications.

Additive Manufacturing and Surface Finishing Of all metal-based AM processes, power bed fusion is the most widely accepted across industries.1 Detailed comparisons and contrasts of different AM processes can be found in other AM-centric reviews.46–48 Because the main focus in the AM community is with metal power beam (electron or laser) powder bed fusion, surfaces from these processes will be the initial focus of this document. When observing the metal laser/electron beam powder bed fusion (L/E-PBF) process, powder is fused on a layer-by-layer basis until the layer is complete. Upon completion of a layer, the build platform is lowered by one layer thickness, and this process is repeated until the sample layering is finished. During this process, a wide variety of features such as spatter, leftover powder particles, and porosity can vastly influence the parameters reported out. One study performed by Senin observed that the surface topography of as-built AM surfaces revealed laser and electron beam melt paths and spatter.34 In the case of reusing powder for AM builds, poor powder quality can be associated with a bad surface.9,42,49–52 Other factors influencing the surface of L/E-PBF processes include the inclination angle (up skin/down skin),21 scan speed and hatching spacing,53 and small bonded particles.9 These aspects of the process signature present several considerations for surface measurement. For an L/E-PBF process, it was discovered that the heat on the edge borders was not adequate to fuse the powder particles, resulting in unmelted and partially sintered particles bonded to the part at the edges.21 During the layer-bylayer process, as this inclination angle increases, the step edges are located closer to one another, where bonded particles become more commonplace.9 In attempts to correlate top and side surface roughness of an AM laser-PBF part, Mumtaz and Hopkinson found that due to different surface tension forces from thermal variations in the melt pool, as the arithmetic mean surface roughness (Ra) on the top decreases, the side surface Ra increases and vice versa.19 There are many other common features included on an L/E-PBF surface. Weld tracks are large surface features that result from the formation and solidification of the melt pool as the laser moves across the powder bed.54 When trying to characterize a weld track, this can be quite difficult due to a variety of factors including heat exchange, interactions between the material and the laser, and also the topography of the layers beneath.34 Another common feature is spatter, where the beam source is passed over the melt pool and particles are partially melted, form agglomerates, or are the result of droplets of molten material.54 In the case of L/E-PBF, the shape, size, and quantity of the spatter on the AM part can influence the surface texture. Surface and near-surface defects are generally found in the form of incomplete particle fusion (lack of fusion), v-notch surface defects (“valleys”), and surface/ near-surface pores or voids (such as the keyhole effect). These surface-related .

73

74

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

defects (SRDs) reduce the mechanical strength and dynamic performance of AM components and impact other performance characteristics, such as cleanliness and fluid drag, depending on the type of component and its applications.55–58 The extent of these SRDs has been identified to depths up to 1 mm from the surface, and some of these defects can be correlated with the overlapping between the inner contour line and the hatching pass; thus, significant surface metal removal through a surface finishing process is needed to produce a defect-free surface, fit for many potential applications that can maximize AM’s potential in all industries.59–61 Secondary subtractive finishing processes such as grinding, milling, and so on may be used to remove some of these defects. Nevertheless, the efficacy of these subtractive technologies to finish AM components is extremely limited due to the geometrical complexity, size, and line-of-sight inaccessibility characteristic of AM components. As a result, AM requires alternative surface finishing techniques to reach its full potential.4,8,56,62,63 In order to be a viable surface finishing technology, the surface finishing process should be capable of (1) reducing the surface roughness to a 0.8-lm Rq surface finish or lower; (2) uniformly removing 0.5 to 1 mm of material from the internal or external surface (or both) of geometrically complex AM components; (3) eliminating surface defects that would otherwise result in reduced tensile or fatigue properties (or both); (4) surface finishing of hundreds or thousands of components without an excessive footprint requirement; and (5) scalability, automation, and minimal operator interaction. The only practical technologies that can potentially meet all of these properties are electropolishing, chemical polishing, mechanical abrasive finishing, and chemically accelerated mechanical finishing. Electrochemical polishing (EP) has been widely used as a surface finishing technique for over a century. In general, EP consists of using the workpiece as an anode immersed on an electrolyte to selectively oxidize (dissolve) the peaks on the surface, achieving remarkably smooth and planar surfaces. However, when used for AM parts, EP has not been effective in targeting the valleys on the surface (i.e., the v-notches); rather, it preferentially targets the peaks and leaves most of the near-surface defects untouched. In order to remove the deepest valleys from the surface, the process needs to run for extended cycles and, due to its preferential surface removal on the surface’s peaks, extreme rounding on sharp features and edges takes place. To overcome this challenge, specialized counter electrodes need to be built to perfectly match the complex geometries of AM components and control the electron transfer on areas of high preferential accumulation for power density. This is needed because it is the complex geometry typical of AM components that causes the differences in power density and, thus, the heterogeneous/nonuniform surface metal removal. It is important to mention that further optimization of the EP process for AM components can improve the uniformity on the material removal ratio on the parts and produce a remarkably smooth surface,64–66 but this will require a considerable amount of time and money, which might be a limitation. .

BROOKS ET AL., DOI: 10.1520/STP163120190160

Another technique with great potential to improve the surface texture and eliminate the SRDs from AM surfaces is chemical polishing (CP). CP is also a common process in the industry and yields highly uniform metal removal and efficacy on complex geometries or internal surfaces (or both). The CP process consists of using specific chemicals to dissolve the surface of a component, which will preferentially dissolve the higher peaks first while limiting the material removal on the lowest valleys. The mechanism of action for CP relies on the gradient difference between the active chemistry available at the peaks, which is exposed completely to the working solution, and the chemistry at the valleys, which is shielded from the working solution. It is rapidly consumed and depleted, limiting the reaction kinetics to effective diffusion into the cavity. A shortcoming that CP has on AM components is that the process cannot planarize the strong waviness that is characteristic of AM-built workpieces because the spatial wavelength is so big and no effective chemistry gradient can be established between the peaks and the much accessible valleys on the surface. Existing CM technologies can eliminate some of the SRDs, but they suffer from an inability to reduce or eliminate the waviness and the macro surface defects of AM workpieces (the so-called layering surface feature) or eliminate the exposed pores and regions with lack of fusion at the subsurface. In addition, there are some environmental aspects and safety engineering aspects that need to be met to employ CP because most of the chemicals used in the process and the by-products can be very dangerous and toxic. The other two technologies with great potential to improve the surface texture and eliminate the SRD characteristics of AM are mechanical abrasive finishing (MAF) and chemically accelerated mechanical finishing (CAMF). Both technologies share certain basic principles, and for this reason, they are going to be discussed together. Both technologies depend on certain rubbing action imparted by masses that can vary in size, geometry, density, and composition. Those masses are called media (in some parts of Europe they are called “chips”), and they are put in motion usually by a vibratory machine or a tumbler. In addition, these technologies can also be employed with high-energy-type equipment such as centrifugal barrels, centrifugal disks, drag finishers, and so on. In the case of mechanical abrasive finishing, metal removal will depend on the effective rubbing action of abrasive media against the workpiece to cause erosion of the surface, eliminating the higher peaks on the surface, and hence, the surface imperfections. Nevertheless, this mechanism of action is also one of the biggest disadvantages of this technology because areas of the workpiece’s surface that are accessible to the rubbing action will experience a higher metal removal rate than less accessible surfaces, such as internal channels and complex shapes. In addition, this effect will also cause significant rounding of edges and corners on the workpiece. On the other hand, the chemically accelerated mechanical finishing is based on chemically activating the surface of the component with a formulation specifically designed (unique per alloy) to produce a self-assembled layer on the surface, known as a conversion coating. This conversion coating is easily removed by any .

75

76

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

mechanical rubbing action against the surface. In a vibratory bowl, where constant contact with moving media and dosing chemistry is achieved, the conversion coating at the surface is only removed at the point of contact between the media and the higher points of the surface roughness. This reduces the peaks first, followed by the valleys, when planarization is achieved. This provides an advantage over abrasive mass finishing without the chemical assistance, in which the media has to reach the surface with a certain force and frequency to create an effective abrasive action to eliminate the peaks. Thus, in cases where the surfaces are inaccessible, the media cannot exert the necessary force; therefore, no effective action occurs. CAMF overcomes this limitation because the chemistry is able to homogeneously reach all surfaces, and the media only needs to contact the surface with a gentle rubbing action to remove the chemically activated surface. This simple, yet elegant, approach considerably reduces the rounding, edging, and heterogeneous surface removal problems faced by the abrasive mass finishing via vibratory and tumbling processes. Furthermore, the CAMF has shown a dramatic effect on the mechanical performance of AM components by eliminating surface defects and v-notches from the surface.67–69

Current State Challenges When comparing previous AM research efforts, guidelines need to be established to help clarify what is being observed when taking a surface measurement of an AM sample. Several research efforts have provided input on how to look at AM surfaces, whereas a recent article by Leach et al. describes the infrastructure under development for specification standards in AM, research efforts regarding geometrical dimensioning and tolerancing for AM, postprocess finishing, and postprocess metrology, including the measurement of surface form, texture, and internal features.69 In the AM realm, those industries that are using AM may run into several challenges when ensuring the validity of their AM surface measurements. Several of these gaps are identified in figure 11 in the realm of AM measurement, analysis, and application. APPLICATION CHALLENGES

The AM surface realm is beginning to transition toward application-based measurement and characterization. However, the requirements for the measurements are not well defined by the application. Each industry will have their own qualification and certification requirements for an AM surface; however, these guidelines or standards have not been established for each industry yet. For example, in aerospace, many AM applications care about a long fatigue life. However, the research that has been performed correlating surface features or parameters to fatigue life has been limited by a wide variety of surface measurement techniques with limited characterization of the surfaces.70 Many more studies need to be performed for truly understanding the process-structure-property functional relationship. .

BROOKS ET AL., DOI: 10.1520/STP163120190160

FIG. 11 Flow chart showing the current challenges facing the AM community regarding application, measurement, and analysis.

When looking at surfaces, the community needs to understand that there are differences within each AM process. Currently, there are seven unique AM processes with four metal processes: powder bed fusion, binder jetting, fused filament fabrication (FFF), and directed energy deposition (DED).71 For example, when talking about a powder bed system, there are multiple options for a melting source (electron beam or laser). The AM parts that come from an electron beam source will have different surface roughness and texture than AM parts from a laser source. For an as-built electron beam melted (EBM) system, the surfaces (e.g., Ra, Sa) are rougher than an as-built L-PBF system. While both technically fall within the same AM category of powder bed fusion, the surface topography varies drastically, and each process needs its own research efforts to understand the melt pool interactions with the energy density. This would also apply to the FFF and DED processes with differences between filament/wire-fed and blown powder DED. Another challenge is the location of the measurement on an AM part. We know that AM allows for more complex geometries to be fabricated, but if you are unable to measure surfaces in these parts, then it makes it near impossible to correlate to the end-use goal. The CAD designer in the beginning of the AM process needs to keep in mind the inspection requirements for the end-use goal; however, this disconnect between design and inspection leads to delays. In cases such as a .

77

78

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

downskin versus upskin region of an AM built part, several research efforts have shown that the process parameters play a key role in the variability of AM surface texture.72,73 Given the results of these studies, there is a high need of research for surface measurements correlating geometrical features to the measurement parameters. When focusing on parameters, the relationship between Ra and Sa is still in exploratory stages. Out of habit, the industry is still currently using Ra generated from contact stylus measurement techniques. However, with AM trending toward part surface to function correlation, larger areas of surfaces are needed to be measured with more applicable parameter reporting. For example, for a turbine blade, due to an end goal of optimized flow, the surface needs to be finished for as little peaks as possible. This indicates that the commonly used surface parameters like Ra or Sa may not reflect what we are looking for in an AM surface. Other functional parameters such as Rp (peaks) and Rv (valleys) would be more indicative and valuable for quality control. Research efforts are ongoing in this field as there are many unique parameters that can be reported out for individual applications. As of this writing, we know that the size of the powder particles is directly proportional to the surface texture. The powder size can vary within each AM technique and from material to material. In the case of Al2O3 powder, the powder size distribution is 20 to 60 lm, while for Inconel718, the powder is 45 to 105 lm.74 Although both powders can be used on the same Laser-PBF system, the resulting AM part will have different-sized features on the surface. Another factor could be powder quality, where porosity in the feedstock can cause additional surface defects. In the case of loosely attached particles, this can greatly affect the reported-out surface parameters if some of the smaller- or larger-sized particles are measured and influence the average mean line. Essentially, if powder-related features can be identified and understood for a surface, a baseline can be established for surface measurements. A standardized measurement process will assist in this realm. When the material in question is not a metal and the process is not PBF, the end user does not know what this looks like on an AM surface. For example, for a polymer AM sample, there are five different AM techniques: VAT photopolymerization, material jetting, sheet lamination, material extrusion, and binder jetting. If we try to compare these polymer surfaces to one another, do they look the same on the surface? The short answer is no because with each method, there is a different source, feedstock material, and so on. Also, the end use of these products will be different from one another. This is leading toward an application-based approach for surface measurement, which is discussed further in the next section. MEASUREMENT CHALLENGES

Within each measurement technique discussed earlier in the section on additive manufacturing and surface finishing, there are several advantages and disadvantages associated in the measurement of AM parts. For confocal microscopy, it is difficult for the instrument to correctly reconstruct recess topography due to a lower .

BROOKS ET AL., DOI: 10.1520/STP163120190160

amount of reflected light.26,75–77 This limitation can also be observed in the case of using CSI, where CSI can have difficulties in measuring highly irregular surfaces or recessed valleys.78 However, CSI can have a unique advantage of measuring high slope angles or small-scale topography of AM surfaces.75,79 In the case of the focus variation technique, several works have found that the repeatability error in height determination is highly dependent on input variables such as illumination, vertical resolution, and lateral resolution.28,36,80,81 By knowing the constraints for each of these techniques, it is possible for an AM user to make a more informed decision when choosing how to measure and analyze their AM surface. Aside from the several established techniques mentioned here, research efforts are ongoing into the use of new measurement applications such as XCT for surface roughness. The advantages of XCT include a nondestructive technique to extract surface information and allow filter and parameter generation82 and also the ability to reconstruct reentrant features (undercuts).83 Recent efforts by Turner and Brierley have validated XCT for feasibility of surface metrology with XCT systems.84 Challenges in this realm involve a long process chain including but not limited to surface determination, generating an STL/PLY file, trimming data, converting STL to PLY, aligning surfaces, performing deviation analysis, cropping, cleaning meshes, converting to height maps, further cropping, filtering per ISO 25178-3, and generating parameter data per ISO 25178-2.13,15 One example includes a successful roundrobin study using XCT for measuring the surfaces of five Ti6Al4V AM samples.85 While further work will need to be performed in this new measurement application realm, this is highly indicative of the value of XCT for surface measurement. While several established and newer measurement techniques were discussed previously with respect to AM surfaces, one of the largest challenges is that there is limited knowledge of cross-technique correlation. This falls back to the fact that there are no standards for AM areal measurements. When trying to compare techniques such as laser confocal microscopy, focus variation, and coherence scanning interferometry, the everyday operator of the equipment may not have the option of selecting a measurement technique. If this is the case, the relationship between measurements from one technique to another needs to be identified. Similar to how AM machines have variability in part fabrication, contact and areal surface measurement techniques have variability in the reported-out measurement parameters. Part of the current challenge is that areal techniques such as laser confocal and focus variation are making their way to the forefront of AM surface measurement without guidelines as to how they should be used for AM surfaces. There is a large gap for research and development (R&D) in each of these measurement techniques in regard to variability of the techniques for AM surfaces. This measurement variability can also add to unmeasurable operator error. While there are multiple contact and areal techniques to measure an AM surface, there is a need for repeatability and reproducibility limits. One of the current challenges is that there are no range requirements for the filtering parameters. Several questions arise from this issue, such as: (1) Does switching from one PBF .

79

80

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

system to another PBF system influence the reported-out surface parameter? (2) How reproducible are the results from one EOS M290 system to another EOS M290 system? (3) What is the quality of equipment and measurement, where do error bars, uncertainty, and bias factor into the results? There is a great need for a good practice guide in regard to knowing how to report out repeatability and reproducibility within AM measurement. ANALYSIS CHALLENGES

AM surfaces are unpredictable. From changing processes to measurement techniques, there are many ways to introduce variety in fabricating an AM part. While this may be the case, recent efforts into analyzing AM surfaces has shown that it is possible to separate out common features on AM surfaces for analysis. Work performed by Senin, Thompson, and Leach34 revealed that a feature-based analysis approach to separate out common AM surface features such as spatter is possible. Transitioning into a feature-based analysis approach would simplify and speed up the analysis process, allowing the operator and end user to know the critical waviness and roughness of their AM surface.34 While some work has been performed to date, there is a strong need for understanding the science behind feature-based surface analysis for AM parts. ASME Project Team 53 is working on this challenge through the identification of common and special surface characteristics of AM parts. After measurement, leveling can be performed where a reference plane is defined to the raw data. This is the basis for how all surface parameters are calculated, meaning that any error here will introduce a much larger error into final reporting of parameters. Currently, it is unknown how several AM areal surface system original equipment manufacturers (OEMs) define the leveling plane. This is an area of research needed to compare different OEM measurement equipment with one another in regard to filtering and leveling and to provide guidelines on the value of each type of measurement equipment. There are many considerations that go into how the raw data are further filtered to generate a list of relevant parameters. The typical treatment of data involves outlier removal, leveling, and filtering. Filtering involves eliminating unwanted vibrations (low pass filter) and form (high pass filter). Based on how the raw data are filtered, outliers can be removed from the data by playing around with the low pass, high pass, and form filters. Further research is needed in the realm of knowing the appropriate filtering steps for an AM surface. This is likely to transition into an application-based approach where knowing what a typical surface roughness Ra value is (e.g., 1 to 5 lm) would indicate a need for an L (kc) filter of 0.8 mm. When discussing filtering, the commonly used filters of kc, ks, and kf can be correlated across measurement techniques. However, few operators or technicians understand this correlation and the bandwidths they are measuring, where the bandwidth is the ratio between kc/ks. Based on the profile measurements, it is known that the kc and ks filters vastly influence the amount of roughness and .

BROOKS ET AL., DOI: 10.1520/STP163120190160

waviness data filtered. When compared to areal techniques, little is known as to whether or not the same filters are applicable from contact to areal measurements. This has created a need for R&D in bandwidth matching. ISO 25178-600, on metrological characteristics for areal surface measurements, is one such document under review to incorporate this information; however, further work needs to be performed.86

Conclusions and Future Direction The current state of surface measurement and characterization in relation to AM and applications has been discussed. Most knowledge and standards regarding surface measurement is in regard to traditional manufacturing, limiting the applicability of previous standards to AM. In the realm of AM, there are many polymer and metal processes that can result in unique surface features. When trying to measure these features, the AM community has hit a roadblock in finding appropriate measurement techniques and analysis methods to validate surface quality (refer to the section “Additive Manufacturing and Surface Finishing”). As the field of AM grows larger, this will create a big challenge in qualification and certification of parts. In this study, we tried to define the current state and address some of the biggest challenges in AM surface measurement. Areas of focus for future work can revolve around applications, measurement, and analysis as summarized in figure 11. A more collaborative effort will be required among the AM process designer, measurement operator, and end-application user. By having the application user define the end-use requirements, design limitations can be incorporated into a measurement plan and vice versa. Based on the application, guidelines can be established for the reporting parameter, whether a mathematical average such as Ra or Sa or a functional parameter such as Sku or Ssk. With many factors involved in surface measurement, guidelines need to be established for how to take measurements, how to know what is being measured, and how to report parameters. Some of the challenges shown here are being worked on by several research institutes around the world; however, this knowledge is limited in exposure to its own specific discipline, industry, and department. Interdisciplinary research efforts such as those funded by the ASTM Additive Manufacturing Center of Excellence and ASME Project Team 53 can provide value to spread the knowledge of advances in each other’s realm. One such example of this is a current ASTM work item that will function as a guide for surface measurement. While this guide will not address every problem mentioned in this study, it will serve as a starting point for future AM surface measurement needs across the materials science, metrology, AM, and their end-use applications. ACKNOWLEDGMENTS

The authors would like to acknowledge ASTM for their support of the ASTM Additive Manufacturing Center of Excellence. As part of the ASTM Additive .

81

82

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Manufacturing Center of Excellence, EWI is working in collaboration with world experts in the field such as Nottingham University and REM Surface Engineering to develop a useful guide for surface measurement. Without funding from ASTM and the assistance and in-kind support of these experts, the research efforts to advance this challenging task would not be possible.

References 1. 2. 3. 4. 5.

6. 7. 8.

9.

10.

11.

12.

13.

14.

15.

.

T. Wohlers, Wohlers Report 2016: 3D Printing and Additive Manufacturing State of the Industry (Fort Collins, CO: Wohlers Associates, Inc., 2016). M. Adams, “The ‘Third Industrial Revolution’: 3D Printing Technology and Australian Designs Law,” Journal of Law, Information and Science 24, no. 1 (2016): 1–27. W. E. Frazier, “Metal Additive Manufacturing: A Review,” Journal of Materials Engineering and Performance 23 (2014): 1917–1928. A. Townsend, N. Senin, L. Blunt, R. K. Leach, and J. S. Taylor, “Surface Texture Metrology for Metal Additive Manufacturing: A Review,” Precision Engineering 46 (2016): 34–47. C. A. Brown, H. N. Hansen, X. J. Jiang, F. Blateyron, J. Berglund, N. Senin, T. Bartkowiak et al., “Multiscale Analyses and Characterizations of Surface Topographies,” CIRP Annals 67, no. 2 (2018): 839–862. D. J. Whitehouse, Surfaces and Their Measurement (London: Hermes Penton, 2002). T. M. Mower and M. J. Long, “Mechanical Behavior of Additive Manufactured, PowderBed Laser-Fused Materials,” Materials Science Engineering A 651 (2015): 198–213. A. Diaz, L. Winkelmann, J. Michaud, and C. Terrazas, “Surface Finishing and Characterization of Titanium Additive Manufacturing Components: From Rich to Smooth Surface,” in World PM2016 Proceedings (Shrewsbury, UK: European Powder Metallurgy Association, 2016), 6. T. Grimm, G. Wiora, and G. Witt, “Characterization of Typical Surface Effects in Additive Manufacturing with Confocal Microscopy,” Surface Topography: Metrology and Properties 3, no. 1 (2015): 014001, https://doi.org/10.1088/2051-672x/3/1/014001 J. Pegues, N. Shamsaei, M. Roach, and R. S. Williamson, “Fatigue Life Estimation of Additive Manufactured Parts in the As-Built Surface Condition,” Material Design & Processing Communications 1, no. 3 (2019): E36. M. Qian, W. Xu, M. Brandt, and H. P. Tang, “Additive Manufacturing and Postprocessing of Ti-6Al-4V for Superior Mechanical Properties,” MRS Bulletin 41, no. 10 (2016): 775–784. M. Kahlin, H. Ansell, D. Basu, A. Kerwin, L. Newton, B. Smith, and J. J. Moverare, “Improved Fatigue Strength of Additively Manufactured Ti6Al4V by Surface Post Processing,” International Journal of Fatigue 134 (2020): 105497. Geometrical Product Specifications (GPS)—Surface Texture: Areal—Part 2: Terms, Definitions and Surface Texture Parameters, ISO 25178-2:2012 (London: British Standards Institute, 2012). Geometrical Product Specifications (GPS)—Surface Texture: Profile Method: Terms, Definitions and Surface Texture Parameters, ISO 4287 (London: British Standards Institute, 2000). Geometrical Product Specifications (GPS)—Surface Texture: Areal—Part 3: Specification Operators, ISO 25178-3:2012 (London: British Standards Institute, 2012).

BROOKS ET AL., DOI: 10.1520/STP163120190160

16. 17.

18.

19. 20.

21.

22.

23. 24. 25.

R. K. Leach, Calibration of Imaging Confocal Microscopes for Areal Surface Texture Measurement (Middlesex, UK: National Physical Laboratory, 2013).

26.

J. Petzing, J. M. Coupland, and R. K. Leach, The Measurement of Rough Surface Topography Using Coherence Scanning Interferometry (Middlesex, UK: National Physical Laboratory, 2010). R. Leach, “Surface Topography Characterisation,” in Fundamental Principles of Engineering Nanometrology (Amsterdam, The Netherlands: Elsevier, 2014), 241–294.

27. 28.

29.

30. 31.

32.

33. 34.

.

Surface Texture (Surface Roughness, Waviness and Lay), ASME B46.1:2009 (New York: American Society of Mechanical Engineers, 2009). Geometrical Product Specifications (GPS)—Surface Texture: Profile Method—Nominal Characteristics of Contact (Stylus) Instruments, ISO 3274:1998 (Geneva, Switzerland: International Organization for Standardization, 1998). A. Safdar, H. Z. He, L. Y. Wei, A. Snis, and L. E. Chavez De Paz, “Effect of Process Parameters Settings and Thickness on Surface Roughness of EBM Produced Ti-6Al-4V,” Rapid Prototyping Journal 18, no. 5 (2012): 401–408. K. Mumtaz and N. Hopkinson, “Top Surface and Side Roughness of Inconel 625 Parts Processed Using Selective Laser Melting,” Rapid Prototyping Journal 15, no. 2 (2009): 96–103. F. Calignano, D. Manfredi, E. P. Ambrosio, L. Iuliano, and P. Fino, “Influence of Process Parameters on Surface Roughness of Aluminum Parts Produced by DMLS,” The International Journal of Advanced Manufacturing Technology 67, nos. 9–12 (2013): 2743–2751. G. Strano, L. Hao, R. M. Everson, and K. E. Evans, “Surface Roughness Analysis, Modelling and Prediction in Selective Laser Melting,” Journal of Materials Processing Technology 213, no. 4 (2013): 589–597. D. Ahn, J. H. Kweon, S. Kwon, J. Song, and S. Lee, “Representation of Surface Roughness in Fused Deposition Modeling,” Journal of Materials Processing Technology 209, nos. 15– 16 (2009): 5593–5600. R. I. Campbell, M. Martorelli, and H. S. Lee, “Surface Roughness Visualisation for Rapid Prototyping Models,” Computer Aided Design 34, no. 10 (2002): 717–725. R. K. Leach and C. L. Giusca, Calibration of Stylus Instruments for Areal Surface Texture Measurement (Middlesex, UK: National Physical Laboratory, 2011).

A. Triantaphyllou, C. L. Giusca, G. D. Macaulay, F. Roerig, M. Hoebel, R. K. Leach, B. Tomita, and K. A. Milne, “Surface Texture Measurement for Additive Manufacturing,” Surface Topography: Metrology and Properties 3, no. 2 (2015): 024002, https://doi.org/ 10.1088/2051-672x/3/2/024002 M. Kro ´l, L. A. Dobrzan ´ski, L. Reimann, and I. Czaja, “Surface Quality in Selective Laser Melting of Metal Powders,” Archives of Materials Science and Engineering 60, no. 2 (2013): 87–92. R. K. Leach, The Measurement of Surface Texture Using Stylus Instruments (Middlesex, UK: National Physical Laboratory, 2001). Geometrical Product Specifications (GPS)—Surface Texture: Areal—Part 607: Nominal Characteristics of Non-Contact (Confocal Microscopy) Instruments, ISO 25178–607 (Geneva, Switzerland: International Organization for Standardization, 2019). Geometrical Product Specifications (GPS)—Surface Texture: Areal—Part 602: Nominal Characteristics of Non-Contact (Confocal Chromatic Probe) Instruments, ISO 25178602:2010 (London: British Standards Institute, 2010). R. Artigas, “Imaging Confocal Microscopy,” in Optical Measurement of Surface Topography, ed. R. Leach (Berlin, Germany: Spring-Verlag, 2011), 237–286. N. Senin, A. Thompson, and R. Leach, “Feature-Based Characterisation of Signature Topography in Laser Powder Bed Fusion of Metals,” Measurement Science and Technology 29, no. 4 (2018), https://doi.org/10.1088/1361-6501/aa9e19

83

84

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

35.

36.

37.

38.

39.

40.

41.

42.

43.

44. 45.

46. 47. 48.

49.

50.

51.

.

Geometrical Product Specification (GPS)—Surface Texture: Areal—Part 604: Nominal Characteristics of Non-Contact (Coherent Scanning Interferometry) Instruments, ISO 25178-604:2013 (London: British Standards Institute, 2013). L. Newton, N. Senin, C. Gomez, R. Danzl, F. Helmli, L. Blunt, and R. Leach, “Areal Topography Measurement of Metal Additive Surfaces Using Focus Variation Microscopy,” Additive Manufacturing 25 (2019): 365–389. N. Duboust, H. Ghadbeigi, C. Pinna, S. Ayvar-Soberanis, A. Collis, R. Scaife, and K. Kerrigan, “An Optical Method for Measuring Surface Roughness of Machined Carbon FibreReinforced Plastic Composites,” Journal of Composite Materials 51, no. 3 (2017): 289–302. L. Newton, N. Senin, and R. Leach, “Focus Variation Measurement of Metal Additively Manufactured Surfaces,” in Proceedings of the 2018 ASPE euspen Summer Topical Meeting Advancing Precision in Additive Manufacturing (Albany, NY: American Society for Precision Engineering, 2018), 232–237. M. Jamshidinia and R. Kovacevic, “The Influence of Heat Accumulation on the Surface Roughness in Powder-Bed Additive Manufacturing,” Surface Topography: Metrology and Properties 3, no. 1 (2015), https://doi.org/10.1088/2051-672X/3/1/014003 L. M. Galantucci, F. Lavecchia, and G. Percoco, “Experimental Study Aiming to Enhance the Surface Finish of Fused Deposition Modeled Parts,” CIRP Annals Manuacturing Technology 58, no. 1 (2009): 189–192. T. Ko ¨pplmayr, L. Ha¨usler, I. Bergmair, and M. Mu ¨hlberger, “Nanoimprint Lithography on Curved Surfaces Prepared by Fused Deposition Modelling,” Surface Topography: Metrology and Properties 3, no. 2 (2015): 024003. M. Vetterli, M. Schmid, and K. Wegener, “Comprehensive Investigation of Surface Characterization Methods for Laser Sintered Parts,” in Proceedings of the Fraunhofer Direct Digital Manufacturing Conference (Stuttgart, Germany: Fraunhofer Verlag, 2014), 1–6. M. K. Johnson and E. H. Adelson, “Retrographic Sensing for the Measurement of Surface Texture and Shape,” in Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition (Piscataway, NJ: Institute of Electrical and Electronics Engineers, 2009), 1070–1077. M. K. Johnson, F. Coley, A. Raj, and E. H. Adelson, “Microgeometry Capture Using an Elastomeric Sensor,” ACM Transactions on Graphics 30, no. 4 (2011): 1–8. M. A. Beard, O. R. Ghita, and K. E. Evans, “Using Raman Spectroscopy to Monitor Surface Finish and Roughness of Components Manufactured by Selective Laser Sintering,” Journal of Raman Spectroscopy 42, no. 4 (2011): 744–748. W. E. Frazier, “Metal Additive Manufacturing: A Review,” Journal of Materials Engineering and Performance 23, no. 6 (2014): 1917–1928. B. Dutta and F. H. Froes, Additive Manufacturing of Titanium Alloys (Amsterdam, The Netherlands: Elsevier, 2014). J. J. Lewandowski and M. Seifi, Metal Additive Manufacturing: A Review of Mechanical Properties (Postprint) Interim Report (Wright-Patterson Air Force Base, OH: Air Force Research Laboratory Materials and Manufacturing Directorate, U.S. Air Force, 2016). T. J. Gornet, K. R. Davis, T. L. Starr, and K. M. Mulloy, “Characterization of Selective Laser Sintering Materials to Determine Process Stability,” Solid Freeform Fabrication Symposium (Austin: University of Texas at Austin, 2002), 546–553. S. Grießbach, R. Lach, and W. Grellmann, “Einfluss der Materialzusammensetzung und Herstellungsbedingungen auf das Eigenschaftsprofil lasergesinterter [Influence of Material Composition and Manufacturing Conditions on the Property Profile of Laser Sintered],” Polymerwerkstoffe (2006): 272–275. J. Kruth, G. Levy, R. Schindel, T. Craeghs, and E. Yasa, “Consolidation of Polymer Powders by Selective Laser Sintering,” in Proceedings of the Third International Conference

BROOKS ET AL., DOI: 10.1520/STP163120190160

52.

53.

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

.

on Polymers and Moulds Innovations, ed. L. Cardon (Ghent, Belgium: University College Ghent, 2008), 15–30. B. Fulcher and D. K. Leigh, “Effect of Segregated First and Second Melt Point on Laser Sintered Part Quality and Processing,” in Proceedings of the 23rd Annual International Solid Freeform Fabrication Symposium, SFF 2012 (Austin, TX: University of Texas at Austin, 2012), 556–564. M. Rombouts, L. Froyen, D. Bourell, and J. P. Kruth, “Roughness after Laser Melting of Iron Based Powders,” in Virtual Modeling and Rapid Manufacturing (London: Taylor & Francis, 2005), 329–335. N. Senin, A. Thompson, and R. K. Leach, “Characterisation of the Topography of Metal Additive Surface Features with Different Measurement Technologies,” Measurement Science and Technology 28, no. 9 (2017), https://doi.org/10.1088/1361-6501/aa7ce2 K. S. Chan, “Characterization and Analysis of Surface Notches on Ti-Alloy Plates Fabricated by Additive Manufacturing Techniques,” Surface Topography: Metrology and Properties 3, no. 4 (2015): 044006. D. Witkin, D. Patel, H. Helvajian, and A. Diaz, “Surface Treatment of Powder-Bed Fusion Additive Manufactured Metals for Improved Fatigue Life,” Journal of Materials Engineering and Performance 28 (2019): 681–692. Y. Yamashita, T. Murakami, R. Mihara, M. Okada, and Y. Murakami, “Defect Analysis and Fatigue Design Basis for Ni-based Superalloy 718 Manufactured by Additive Manufacturing,” Procedia Structural Integrity 7 (2017): 11–18. H. Masuo, Y. Tanaka, S. Morokoshi, H. Yagura, T. Uchida, Y. Yamamoto, and Y. Murakami, “Effects of Defects, Surface Roughness and HIP on Fatigue Strength of Ti-6Al-4V Manufactured by Additive Manufacturing,” Procedia Structural Integrity 7 (2017): 19–26. D. Witkin, H. Helvajian, L. Steffeney, and W. Hansen, “Laser Post-Processing of Inconel 625 Made by Selective Laser Melting,” in Proceedings Volume 9738: Laser 3D Manufacturing III, ed. G. Gu, H. Helvajian, and A. Pigue ´ (Bellingham, WA: Society of Photo-Optical Instrumentation Engineers (SPIE), 2016), 97380W. B. Wysocki, J. Idaszek, K. Szla ˛zak, K. Strzelczyk, T. Brynk, K. J. Kurzydlowski, and W. S ´wie ˛szkowski, “Post Processing and Biological Evaluation of the Titanium Scaffolds for Bone Tissue Engineering,” Materials (Basel) 9, no. 3 (2016): 197. C. de Formanoir, M. Suard, R. Dendievel, G. Martin, and S. Godet, “Improving the Mechanical Efficiency of Electron Beam Melted Titanium Lattice Structures by Chemical Etching,” Additive Manufacturing 11 (2016): 71–76. Y. Y. Sun, S. Gulizia, C. H. Oh, D. Fraser, M. Leary, Y. F. Yang, and M. Qian, “The Influence of As-Built Surface Conditions on Mechanical Properties of Ti-6Al-4V Additively Manufactured by Selective Electron Beam Melting,” JOM 68, no. 3 (2016): 791–798. S. Bagehorn, J. Wehr, and H. J. J. Maier, “Application of Mechanical Surface Finishing Processes for Roughness Reduction and Fatigue Improvement of Additively Manufactured Ti-6Al-4V Parts,” International Journal of Fatigue 102 (2017): 135–142. L. Yang, Y. Wu, A. Lassell, and B. Zhou, “Electropolishing of Ti6Al4V Parts Fabricated by Electron Beam Melting,” in Proceedings of the 26th Annual International Solid Freeform Fabrication Symposium—An Additive Manufacturing Conference, ed. D. L. Bourell (Austin, TX: University of Texas at Austin, 2016), 1333–1344. Y.-C. Wu, C.-N. Kuo, Y.-C. Chung, C.-H. Ng, and J. C. Huang, “Effects of Electropolishing on Mechanical Properties and Bio-Corrosion of Ti6Al4V Fabricated by Electron Beam Melting Additive Manufacturing,” Materials (Basel) 12, no. 9 (2019): 1466. Y. Zhang, J. Li, and S. Che, “Electrochemical Science Electropolishing Mechanism of Ti6Al-4V Alloy Fabricated by Selective Laser Melting,” International Journal of Electrochemical Science 13 (2018): 4792–4807.

85

86

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

67.

68.

69. 70.

71. 72.

73.

74.

75.

76.

77.

78. 79.

80.

81.

82.

83.

.

D. B. Witkin, D. N. Patel, H. Helvajian, L. Steffeney, and A. Diaz, “Surface Treatment of Powder-Bed Fusion Additive Manufactured Metals for Improved Fatigue Life,” Journal of Materials Engineering and Performance 28, no. 2 (2019): 681–692. A. Diaz, “Surface Texture Characterization and Optimization of Metal Additive Manufacturing-Produced Components for Aerospace Applications,” Additive Manufacturing for the Aerospace Industry (2019): 341–374. R. K. Leach, D. Bourell, S. Carmignato, A. Donmez, N. Senin, and W. Dewulf, “Geometrical Metrology for Metal Additive Manufacturing,” CIRP Annals 68, no. 2 (2019): 677–700. J. Tong, C. R. Bowen, J. Persson, and A. Plummer, “Mechanical Properties of TitaniumBased Ti–6Al–4V Alloys Manufactured by Powder Bed Additive Manufacture,” Materials Science and Technology (United Kingdom) 33, no. 2 (2017): 138–148. Additive Manufacturing—General Principles—Part 2: Overview of Process Categories and Feedstock, ISO 17296-2:2015 (London: British Standards Institute, 2015). J. C. Fox, S. P. Moylan, and B. M. Lane, “Effect of Process Parameters on the Surface Roughness of Overhanging Structures in Laser Powder Bed Fusion Additive Manufacturing,” Procedia CIRP 45 (2016): 131–134. R. Udroiu, I. C. Baraga, and A. Nedelcu, “Evaluating the Quality Surface Performance of Additive Manufacturing Systems: Methodology and a Material Jetting Case Study,” Materials (Basel) 12, no. 6 (2019): 995. J. A. Slotwinski, E. J. Garboczi, P. E. Stutzman, C. F. Ferraris, S. S. Watson, and M. A. Peltz, “Characterization of Metal Powders Used for Additive Manufacturing,” Journal of Research of the National Institute of Standards and Technology 119 (2014): 460–493. A. Thompson, “Surface Texture Measurement of Metal Additively Manufactured Parts by X-Ray Computed Tomography” (PhD thesis, University of Nottingham, 2018), https:// doi.org/10.13140/RG.2.2.26204.03207 R. K. Leach and C. L. Giusca, Calibration of Phase Stepping Interferometers and Coherence Scanning Interferometers for Areal Surface Texture Measurement (London: National Physical Laboratory, 2013). A. Temmler, E. Willenborg, and K. Wissenbach, “Laser Polishing,” in SPIE Proceedings, Vol. 8243: Laser Applications in Microelectronic Optoelectronic Manufacturing (LAMOM) XVII (Bellingham, WA: SPIE, 2012), https://doi.org/10.1117/12.906001 P. de Groot, “Coherence Scanning Interferometry,” Optical Measurement of Surface Topography (2011): 187–208. M. F. Fay, V. G. Badami, and X. C. De Lega, “Characterizing Additive Manufacturing Parts Using Coherence Scanning Interferometry,” in Proceedings of ASPE 2014 Spring Topical Meeting on Dimensional Accuracy and Surface Finish in Additive Manufacturing (Raleigh, NC: American Society for Precision Engineering, 2014), 135–138. G. Krolczyk, P. Raos, and S. Legutko, “Experimental Analysis of Surface Roughness and Surface Texture of Machined and Fused Deposition Modelled Parts,” Tehnicˇki Vjesnik 21, no. 1 (2014): 217–221. R. Danz, F. Helmli, and S. Scherer, “Focus Variation—A Robust Technology for High Resolution Optical 3D Surface Metrology,” Journal of Mechanical Engineering 57, no. 3 (2011): 245–256. A. Townsend, L. Blunt, and P. Bills, “Investigating the Capability of Microfocus X-Ray Computed Tomography for Areal Surface Analysis of Additively Manufactured Parts,” in Proceedings of the 2016 ASPE Euspen Summer Topical Meeting Advancing Precision in Additive Manufacturing (Albany, NY: American Society for Precision Engineering, 2016), 206–210. L. Pagani and P. J. Scott, “Areal Surface Texture Parameters on Surface,” in Proceedings of the 2016 ASPE Euspen Summer Topical Meeting Advancing Precision in Additive Manufacturing (Albany, NY: American Society for Precision Engineering, 2016), 162–167.

BROOKS ET AL., DOI: 10.1520/STP163120190160

84.

85.

86.

.

A. T. N. Turner and N. Brierley, “3-in-1 X-Ray Computed Tomography” (paper presentation, Ninth Conference on Industrial Computed Tomography, Padova, Italy, February 13– 15, 2019). A. Townsend, R. Racasan, R. Leach, N. Senin, A. Thompson, A. Ramsey, D. Bate, P. Wolliams, S. Brown, and L. Blunt, “An Interlaboratory Comparison of X-Ray Computed Tomography Measurement for Texture and Dimensional Characterisation of Additively Manufactured Parts,” Additive Manufacturing 23 (2018): 422–432. Geometrical Product Specifications (GPS)–Surface Texture: Areal—Part 600: Metrological Characteristics for Areal Topography Measuring Methods, ISO 25178-600:2019 (London: British Standards Institute, 2019).

87

88

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190141

Younes Chahid,1 Andrew Townsend,2 Alexander Liu,3 Paul Bills,1 Philip Sperling,4 and Radu Racasan1

Optimizing X-Ray Computed Tomography Settings for Dimensional Metrology Using 2D Image Analysis Citation Y. Chahid, A. Townsend, A. Liu, P. Bills, P. Sperling, and R. Racasan, “Optimizing X-Ray Computed Tomography Settings for Dimensional Metrology Using 2D Image Analysis,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 88–101. http://doi.org/10.1520/ STP1631201901415

ABSTRACT

The current way to choose X-ray computed tomography (XCT) scanning settings is usually manual and prone to operator errors. This paper presents an effective semiautomatic protocol that proves a high correlation between the local contrast-to-noise (CNR) of XCT two-dimensional (2D) projection image (prior to reconstruction) quality and the resulting XCT 3D volume scan quality. This high correlation allowed the comparison of four XCT settings to determine the one with the smallest error, solely by locally using the CNR equation on one 2D projection (prior to reconstruction) of an additive manufactured lattice structure. Verification of the protocol was done by using a workpiece and comparing the chosen XCT setting reconstructed workpiece dimensions to the ones measured

Manuscript received November 6, 2019; accepted for publication May 27, 2020. 1 EPSRC Future Metrology Hub, School of Computing and Engineering, University of Huddersfield, http://orcid.org/0000-0002-1816-4869, P. B. http:// Queensgate, HD1 3DH Huddersfield, UK, Y. C. orcid.org/0000-0002-5468-6330, R. R. http://orcid.org/0000-0001-5067-9494 2 Nondestructive Evaluation Group, Nondestructive Characterization Institute, Lawrence Livermore National Laboratory, 7000 East Ave., Livermore, CA 94550, USA http://orcid.org/0000-0002-1215-3018 3 NTU Innovation Centre, 71 Nanyang Dr. #04-01, Singapore 638075 4 Product Management Additive Manufacturing, Volume Graphics GmbH, Speyerer Str. 4-6, 69115 Heidelberg, Germany http://orcid.org/0000-0001-9944-264X 5 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

CHAHID ET AL., DOI: 10.1520/STP163120190141

using a coordinate-measuring machine (CMM). This new method can reduce the operator error and time needed to compare different XCT setting combinations. The proposed protocol is a step closer to an automated XCT parameter selection procedure, limiting user dependency and error while increasing accuracy and fidelity. Keywords quality control, metrology, X-ray computed tomography (XCT), additive manufacturing

Introduction X-ray computed tomography (XCT) is one of the few tools that can assist in measuring and assessing both external and internal features. With the advancements in design techniques and additive manufacturing methods, the industry has become capable of producing parts such as custom lattice structures with extreme complexity that cannot be measured with common metrology tools. Lattice structures are usually topologically ordered and repeated unit cells;1 their unique geometry can, for example, be beneficial for patient-specific orthopedic implants2 or in aerospace.3,4 However, while XCT is usually the tool used to measure these geometries in a nondestructive way, there is still no clear model correlating the effects of different XCT parameters on the measurement uncertainty.5 This means that: • Current users cannot usually provide an uncertainty statement.6 • No holistic model can currently aid the user in choosing settings for a part where the material and internal features are unknown. • Most XCT settings are being chosen by the user, leading to a great influence and large variation in the measurements.7 • A considerable amount of time can be spent choosing the XCT settings optimized for scanning a given part. CURRENT PROCEDURES AND CHALLENGES WHEN CHOOSING XCT SETTINGS

When faced with a new part, of which the material and internal features are unknown, an experienced user can usually limit the range of the different combinations of settings that should be used, following common initial user guidelines provided in the machine manufacturer manuals. These guidelines suggest, for example, that X-rays penetrate the part from all angles without saturating any of the projections used for reconstruction or that the X-ray spot stays smaller than the effective pixel size.* After limiting the choice of settings to be used, a user is usually not capable of accurately deciding from this range of settings which one results in the highest contrast and minimum noise—therefore leading to less artifacts and measurement

*Nikon Metrology, “Inspect-X: X-Ray and CT Acquisition and Processing Software,” 2020, http://web.archive.org/web/20200519163325/https://www.nikonmetrology.com/en-us/product/inspect-x .

89

90

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

error. The reason is that it can be challenging for the user to detect subtle changes in contrast and noise from the histogram alone or the two dimensional (2D) image displayed on the machine screen without relying on image analysis tools before starting the XCT scan. Currently, to choose the XCT machine settings, different approaches can be considered. Alongside the manufacturer’s XCT guidelines, the ISO 15708-2:20198 descriptive table can be used as a starting point. The latter links the approximate adequate voltages, screen scintillators, and filtration values to different materials. The materials, however, are described as pure elements, which makes it hard to choose the ideal parameter for an alloy composed of a combination of different elements. Another approach is to use simulation techniques and software such as aRTist (BAM, Germany) to estimate or optimize the scan quality. Usually, based on the attenuation law for the primary radiation and Monte Carlo models for scattered radiation,9 different part positions and scan parameters can be simulated to effectively reduce beam hardening and ring artifacts and to increase scan quality.10–12 However, this method is only feasible if the part outer, internal features, and material with its element percentages are known prior to scanning. A semiempirical approach has also been suggested in the literature. Called a knowledge-based system (KBS), the method assists in choosing the XCT settings of a part aided by the use of available knowledge of the XCT device to be used and previously scanned parts of a similar shape. The aforementioned paper also showed that the method provided more accurate results compared to relying on user experience only.7 Another solution is, of course, user experience. Experienced users yield more accurate results.7 EVALUATING 2D IMAGE QUALITY OF XCT PROJECTIONS OR RECONSTRUCTIONS

Using 2D image quality to discern the scan quality has been used numerous times in the literature. The contrast-to-noise ratio (CNR) equation is a common tool in this field.9,10,13 The two parameters in the CNR equation (1) are l1–2, mean gray value, and r1–2, standard deviation of both the part and the background values extracted locally from the image, as used in the medical field14 or when using a scanning electron microscope (SEM).15 An experienced user is often needed to avoid choosing a region of interest (ROI) with a local artifact leading to it not representing the quality of the whole image. To minimize the user input error when choosing an ROI, the CNR equation parameters can be extracted directly from the histogram peaks of a reconstructed 2D image.16 CNR ¼

Contrast jl1  l2 j ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Noise r21 þ r22

(1)

CNR and signal to noise ratio (SNR) alongside other 2D image quality quantifiers have been used to analyze the effect of different XCT settings on each image .

CHAHID ET AL., DOI: 10.1520/STP163120190141

quality parameter.17 The research showed that the image quality parameters correlated with XCT scanning settings, but the accuracy was not enough to rank the ideal XCT settings purely from evaluating the 2D image projections. The method proposed in this paper not only assists the user in choosing from a range of XCT settings based on the image quality of the projections (prior to reconstruction) but also represents a crucial step closer to a closed-loop XCT, where the settings are automatically chosen, by comparing a range of XCT settings and choosing the ideal one, in an optimized procedure.

Methodology INTRODUCTION

The proposed protocol, as shown in figure 1, starts by taking 2D XCT images of the part at different settings; CNR is then calculated for each one using nonreconstructed 2D XCT images. The setting with the highest CNR is considered the ideal XCT setting. To validate the proposed protocol, the obtained results were compared to the conventional protocol where 3D XCT reconstructed scans of a dimensional workpiece (previously measured using a coordinate-measuring machine [CMM]) and the lattice are taken at the same time with different settings. The 3D reconstructed and thresholded volume XCT scans are then analyzed, and the one with the least difference from the CMM measurement is considered the ideal one. THE DIMENSIONAL WORKPIECE

The dimensional workpiece (fig. 2) was computer numerical control (CNC) machined from aluminum bar stock to produce a consistent gray value during XCT scans and

FIG. 1 Suggested and conventional protocol used to compare four XCT setting combinations to find the one with the least error. Total time is calculated if each of the four scans is only done once.

.

91

92

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 (A) The dimensional workpiece dimensions (in mm) used in the experiment. (B) Manufactured dimensional workpiece with pencil tip.19

to have less chance of porosity. The part has a 1.5-mm outer diameter (OD), the same as the lattice truss diameter, with the aim of achieving almost the same gray value when XCT scanning both parts. This allows an edge detection that is consistent for both parts. Designing and scanning a dimensional workpiece from the same material as the main scanned part (in this case, the lattice structure) is a common method used in the literature, especially for XCT porosity measurements.18 The CMM, a Zeiss Prismo Access, was used to measure the OD. The CMM measured mean OD after five repeat measurements was 1.5117 mm (0.0001 mm standard deviation). To determine the OD, as shown in figure 3, four circles were measured (five times each) to then extract the cylinder they formed.

FIG. 3 Measurement strategy for CMM to extract the outer cylinder diameter (OD).

.

CHAHID ET AL., DOI: 10.1520/STP163120190141

LATTICE STRUCTURE AND FIXTURING

The cube vertex centroid lattice structure, as shown in figure 4, was designed in Ntop (Ntopology, New York) with a dimension of 15 by 15 by 25 mm and a truss diameter of 1.5 mm. The lattice was additive manufactured using an EOS M290 in AlSi10Mg material (fig. 4B). The lattice structure and the dimensional workpiece, as seen in figure 5, were held using an additive manufactured polyactide (PLA) plastic fixture using a fused

FIG. 4 (A) Rendering and (B) additive manufactured lattice structure with 15-by-15-by25-mm box dimension and 1.5-mm truss diameter.

FIG. 5 (A) The FDM 3D-printed holder used to grip both the lattice and dimensional workpiece. (B) Rendering of the fixed position.

.

93

94

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

deposition modeling (FDM) process. The fixture minimized the movement of the part without being in direct contact with the measured features, allowing for better surface determination. XCT MEASUREMENT PROTOCOL

The XCT machine used was a Nikon XT H 225 (Nikon Metrology), and the different XCT settings used in this study are shown in table 1. The voxel size, exposure time, and number of projections were fixed to 0.023 mm, 2,829 ms, and 1,583 projections, respectively, with no filter. Each setting combination was done three times. This resulted in a total of 12 scans that were all performed in a random order. The magnification was fixed in order to avoid the need to scale the ROIs analyzed. More settings will be simultaneously changed in future research, especially exposure time and number of projections, two factors usually decisive in scan times. 2D XCT MEASUREMENTS

Using the Nikon XT H 225, three 2D XCT projections (after shading correction) were taken of each XCT setting combination mentioned in table 1. The analysis of the 2D projections was done using Fiji software.* The projection images used were 16-bit TIFF images, each with a size of 1008 by 1008 pixels. The gray value of each pixel ranged from 0 to 65,536. The CNR equation used is from ISO 15708-3:2019.20 The higher the CNR, the higher the image quality. In this case, and as seen in equation (2), an absolute value is calculated between the mean gray value of the part and mean gray value of the background and is divided by the noise of the background. The goal is to have an image with a high gray value difference between the part and the background but with a small noise value. CNR ¼

jlf  lb j rb

(2)

To extract these values from the 2D nonreconstructed XCT projection image, four different ROIs were defined, as shown in figure 6: TABLE 1 Setting combinations used in the study

Setting Combinations

Voltage (kV)

Power (W)

100_6 (3 meas.)

100

6

120_7 (3 meas.)

120

7

140_8 (3 meas.)

140

8

160_9 (3 meas.)

160

9

*Fiji (ImageJ 2.0), https://web.archive.org/web/20200729223035/https://fiji.sc/ .

CHAHID ET AL., DOI: 10.1520/STP163120190141

FIG. 6 Regions of interest used in CNR formula shown in two different 2D projections: (A) one for dimensional and (B) one for lattice structure.

ROI_1 and ROI_3: Used to extract the local mean gray value (lf) of the workpiece (ROI_1) and of the lattice part (ROI_3). • ROI_2 and ROI_4: Used to extract both the local gray value of the background (lb) and the noise of the background (rb) for the workpiece (ROI_2) or the lattice (ROI_4). As mentioned previously, the dimensional workpiece diameter and material were chosen to have almost the same gray value of the lattice strut. To limit the impact of the chosen projection on the gray value, it was made sure that the chosen projection angle for the lattice did not have superimposed struts (making the pixels darker). The chosen projection isolated the strut so that its gray value was closer to the dimensional workpiece. The individual effects of different XCT parameters on the 2D projection image quality has not been discussed in this study and has been previously addressed.21 This study does not individually look at the cause and effect of each parameter and their influence on the CT scan; the purpose was to rather assess if there is a strong correlation between the 2D image quality (before reconstruction) using CNR and CT settings (voltage and power, in this case) to assist in swiftly selecting parameters before each scan. The proposed method is semiautomated in the sense that the user chooses the XCT parameters to be compared and defines the 2D image ROIs, while the XCT scanning and CNR extraction are done as a batch. The batch process was done using Inspect-X software (Nikon Metrology) for automating XCT scans and Fiji software for automating the extraction of CNR from 2D images. •

3D VOLUME XCT SCANS AND ANALYSIS

Using Nikon XT H 225, three 3D volume XCT scans were taken of each XCT setting combination mentioned in table 1. Reconstruction was performed using .

95

96

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Nikon XCT Pro 3D (Nikon Metrology), and no beam hardening or noise reduction algorithm was applied during the reconstruction. Since the XCT machine used is not a metrology one, and to reduce reproducibility errors, the manipulator and the part were not moved during measurements and all 12 XCT scans were done as a batch. For surface determination, local iterative surface determination was chosen using VGSTUDIO MAX3.1 (Volume Graphics, Heidelberg, Germany). This type of surface determination delivers the most accurate results during reconstruction.22 A global surface determination was also applied, and both methods were compared.

Results 2D XCT PROJECTION ANALYSIS

Using equation (2) on the ROIs described in figure 6, the CNR of each 2D projection for each setting was determined. Because three 2D projections were taken per XCT setting, table 2 shows the average CNR value as well as the standard deviation. It can be seen from table 2 that the setting with 160 kV and 9 W had the highest CNR for both the dimensional workpiece and the lattice. The raw data used to calculate CNR can be seen in table A.1 and table A.2 in the Appendix. 3D XCT SCAN RESULTS AND MEASUREMENTS USED FOR VALIDATION

To verify the results of the 2D XCT projection analysis, the extracted OD from 3D volume analysis was compared to the CMM measured OD. The calculated difference (using two types of surface determinations) for each XCT setting combination can be seen in figure 7. Because each setting was taken three times, the standard deviation intervals are also plotted, showing the repeatability of the method. The standard deviation of all settings was 0.0001 mm except in automatic surface determination where two settings (100_6 and 120_7) have a standard deviation of 0.0003 mm and 0.0005 mm. Based on the results in figure 7, the XCT setting combination of 160 kV and 9 W has the smallest error—in this case, meaning the least difference from CMM measurement. It can also be seen from figure 7 that the mean difference value to the CMM using automatic global surface determination was higher compared to using local

TABLE 2 Average CNR value per XCT setting for dimensional workpiece (left) and lattice (right)

XCT Settings

.

Dimensional Workpiece CNR [Std. Dev.] (Gray Value)

Lattice CNR [Std. Dev.] (Gray Value)

100_6 (3 meas.)

16.3 [0.3]

19.3 [0.5]

120_7 (3 meas.)

16.3 [0.6]

18.4 [0.6]

140_8 (3 meas.)

20.2 [0.9]

19.5 [1.2]

160_9 (3 meas.)

21.6 [0.2]

20.1 [0.2]

CHAHID ET AL., DOI: 10.1520/STP163120190141

FIG. 7 Mean OD difference between XCT volume data and CMM data with standard deviation.

iterative surface determination, which was expected and conforms to previous findings published in the literature.22

Discussion Based on the aforementioned results, the XCT setting combination determined by the 2D projection analysis (table 2) matches the XCT setting combination determined using the 3D volume analysis with both surface determinations (fig. 2). This proves a high correlation between the 2D projections’ image quality (prior to reconstruction) and the reconstructed 3D volume quality. These results also prove that before a XCT scan, comparing the local CNR of one 2D projection (prior to reconstruction) of each XCT setting combination (voltage and power in this case) can assist in finding the ideal XCT scan setting. The X-ray voltage affects the contrast between low density materials and the background noise level.23 As expected, the ideal setting was the one with the highest power (W). Higher power leads to the broadening of the histogram peaks, leading to more contrast in the 2D projections, an easier edge detection, and smaller surface determination error after reconstruction and thresholding. However, measuring the contrast alone can be misleading because the noise can induce artificial high differences between pixels, hence the importance of using the CNR equation because it takes the noise value into consideration. .

97

98

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

To minimize noise, one of the user guidelines from ISO 15708-2:2019 is to have a minimal X-ray transmission of 10%.8 Minimal X-ray transmission is calculated by comparing the screen gray value when the X-ray is off with the scanned part’s lowest gray value. While the suggested protocol assists in comparing different XCT settings and deciphering the ideal one, it does not suggest which XCT settings should initially be considered for comparison. This means that when getting a new part, it is expected that the user selects the range of XCT setting combinations needed. The number and selected values for these combinations will depend on the user’s experience as well as on the maximum allowed time, cost of the experiment, and allowable error. Due to the XCT repeatability error when scanning after long intervals5 (impacted by filament life and other factors), it is expected, when using this protocol, that the XCT scan is performed shortly after performing the suggested protocol. As seen in figure 1 and in this study, the XCT scanning and 3D volume comparison of four XCT scans (if each scan is only done once) took around 7 h. Alternatively, the time needed for the 2D projection analysis (prior to reconstruction) of four XCT scans took only 35 min. While a low standard deviation was achieved when analyzing three samples of each scan (in both 3D volume and 2D analysis), the verification method relied solely on the difference between the CMM measured OD and the XCT 3D volume extracted OD. This means that the CNR method produces accurate results when the goal is to have XCT scans with the least dimensional measurement error.

Conclusion This paper presents and verifies a new protocol for evaluating XCT setting combinations in order to find the optimum voltage (kV) and power (W) (prior to reconstruction). While the results were expected, this study proved the high correlation between the 2D nonreconstructed image CNR and reconstructed 3D volume. While an experienced user can choose better XCT settings than a nonexperienced one, the suggested protocol of batch scanning and batch analyzing 2D projections prior to reconstruction can be used on a range of settings set by the experienced user, taking advantage of the user’s experience, saving even more time, and increasing the machine accuracy while limiting user input and error. This paper compared setting combinations in which only two XCT parameters were altered in the combinations. More parameters need to be considered in the future alongside their impact on the CNR equation. It is expected, for example, that if the magnification is changed, the ROI needs to be scaled accordingly as well. The protocol also does not take in consideration artifacts that can only be visible after reconstruction (e.g., beam hardening, ring artifacts, and so on). The proposed protocol needs to be assessed for other scenarios such as having minimal error when calculating porosity or extracting areal surface roughness. Implementing this .

CHAHID ET AL., DOI: 10.1520/STP163120190141

protocol will provide a closer step to a semiautomated and, ideally in the future, a fully automated XCT setup process, limited in terms of user input and error, with a high accuracy and minimal noise. ACKNOWLEDGEMENTS

The authors gratefully acknowledge the UK’s Engineering and Physical Sciences Research Council (EPSRC) funding of the Future Metrology Hub (Grant Ref: EP/ P006930/1). This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC5207NA27344. LLNL Document number LLNL-BOOK-811638. The authors also gratefully acknowledge Ntopology for providing the license of Ntop.

Appendix TABLE A.1 Dimensional workpiece ROI_1, ROI_2, and CNR raw data

.

XCT Setting

Mean Part (ROI_1)

Mean Back Part (ROI_2)

St. Dev. Back Part (ROI_2)

CNR Part

100_6

43546.8

52576.1

563.7

16.0

100_6

43520.3

52554.0

545.6

16.6

100_6

43552.6

52504.1

545.7

16.4 16.1

120_7

45313.6

52885.2

471.2

120_7

45253.0

52740.0

439.0

17.1

120_7

45300.3

52876.1

478.2

15.8

140_8

45614.6

51854.9

311.8

20.0

140_8

45481.7

51757.6

297.2

21.1

140_8

45572.2

51801.6

320.8

19.4

160_9

46099.1

51782.4

264.0

21.5

160_9

45762.6

51383.4

263.3

21.3

160_9

46172.9

51893.1

262.6

21.8

99

100

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE A.2 Lattice ROI_3, ROI_4, and CNR raw data

XCT Setting

Mean Lattice (ROI_3)

Mean Lattice Background (ROI_4)

Std. Dev. (Lattice Back) ROI_4

CNR

100_6

42178.6

50507.6

422.8

19.7

100_6

42216.6

50540.3

426.7

19.5

100_6

42277.7

50512.3

439.8

18.7

120_7

44206.5

51245.8

369.5

19.1

120_7

44213.8

51212.3

381.7

18.3

120_7

44278.0

51272.1

391.2

17.9

140_8

44947.3

50853.2

292.1

20.2

140_8

44890.1

50901.3

298.6

20.1

140_8

45119.4

51016.3

327.4

18.0

160_9

45880.2

51410.8

277.6

19.9

160_9

46119.5

51633.9

271.4

20.3

160_9

47880.0

53590.3

285.1

20.0

References 1.

2. 3.

4.

5. 6.

7.

8.

9.

.

T. Maconachie, M. Leary, B. Lozanovski, X. Zhang, M. Qian, O. Faruque, and M. Brandt, “SLM Lattice Structures: Properties, Performance, Applications and Challenges, Materials & Design 183 (2019): 1–18. E. Alabort, D. Barba, and R. C. Reed, “Design of Metallic Bone by Additive Manufacturing,” Scripta Materialia 164 (2019): 110–114. M. Bici, S. Brischetto, F. Campana, C. G. Ferro, C. Secli, S. Varetti, P. Maggiore, and A. Mazza, “Development of a Multifunctional Panel for Aerospace Use through SLM Additive Manufacturing,” Procedia CIRP 67 (2018): 215–220. H. Zhou, X. Zhang, H. Zeng, H. Yang, H. Lei, X. Li, and Y. Wang, “Lightweight Structure of a Phase-Change Thermal Controller Based on Lattice Cells Manufactured by SLM,” Chinese Journal of Aeronautics 32 (2019): 1727–1732. J. P. Kruth, M. Bartscher, S. Carmignato, R. Schmitt, L. De Chiffre, and A. Weckenmann, “Computed Tomography for Dimensional Metrology,” CIRP Annals 60 (2011): 821–842. S. Carmignato, A. Pierobon, and E. Savio, “First International Intercomparison of Computed Tomography Systems for Dimensional Metrology,” in Proceedings of the Eleventh International Conference of the European Society for Precision Engineering and Nanotechnology, EUSPEN 2011, ed. H. Spaan (Bedfordshire, UK: EUSPEN, 2011), 84–87. R. Schmitt, C. Isenberg, and C. Niggemann, “Knowledge-Based System to Improve Dimensional CT Measurements,” in Proceedings of the Fourth Conference on Industrial Computed Tomography, (Du ¨ren, Germany: Shaker Verlag GmbH, 2012), 363–372. Non-Destructive Testing. Radiation Methods for Computed Tomography. Principles, Equipment and Samples, BS EN ISO 15708-2:2019 (London: British Standards Institution, 2019). G.-R. Jaenisch, C. Bellon, and U. Samadurau, “McRay—A Monte Carlo Model Coupled to CAD for Radiation Techniques,” in Proceedings of the 9th European Conference on NonDestructive Testing ECNDT (Berlin, Germany: ECNDT, 2006), 1–8.

CHAHID ET AL., DOI: 10.1520/STP163120190141

10.

11.

12.

13.

14.

15. 16.

17.

18.

19.

20. 21.

22.

23.

.

U. Hilpert, M. Bartscher, M. Neugebauer, J. Goebbels, G. Weidemann, and C. Bellon, “Simulation-Aided Computed Tomography (CT) for Dimensional Measurements” (paper presentation, International Symposium on Digital Industrial Radiology and Computed Tomography, Lyon, France, June 25–27, 2007). J. Tabary, “Sindbad—A Realistic Multi-Purpose and Scalable X-Ray Simulation Tool for NDT Applications” (paper presentation, International Symposium on Digital Industrial Radiology and Computed Tomography, Lyon, France, June 25–27, 2007). M. Reiter, B. Harrer, C. Heinzl, D. Salaberger, C. Gusenbauer, C. Kuhn, and J. Kastner, “Simulation Aided Study for Optimising Industrial X-Ray CT Scan Parameters for NonDestructive Testing and Materials Characterisation” (poster, International Symposium on Digital Industrial Radiology and Computed Tomography, Berlin, Germany, June 20–22, 2011). A. du Plessis, M. Tshibalanganda, and S. G. le Roux, “Not All Scans Are Equal: X-Ray Tomography Image Quality Evaluation,” Materials Today Communications 22 (2020): 1–10. B. Bechara, C. A. McMahan, W. S. Moore, M. Noujeim, H. Geha, and F. B. Teixeira, “Contrast-to-Noise Ratio Difference in Small Field of View Cone Beam Computed Tomography Machines,” Journal of Oral Science 54 (2012): 227–232. F. Timischl, “The Contrast-To-Noise Ratio for Image Quality Evaluation in Scanning Electron Microscopy,” The Journal of Scanning Microscopies 37 (2015): 54–62. M. Reiter, D. Weiss, C. Gusenbauer, M. Erler, C. Kuhn, S. Kasperl, and J. Kastner, “Evaluation of a Histogram-Based Image Quality Measure for X-Ray Computed Tomography” (paper presentation, Fifth Conference on Industrial Computed Tomography, Wels, Austria, February 25–28, 2014). A. Kraemer, E. Kovacheva, and G. Lanza, “Projection Based Evaluation of CT Image Quality in Dimensional Metrology” (paper presentation, International Symposium on Digital Industrial Radiology and Computed Tomography, Lyon, France, June 25–27, 2007). P. Hermanek, F. Zanini, and S. Carmignato, “Traceable Porosity Measurements in Industrial Components Using X-Ray Computed Tomography,” Journal of Manufacturing Science and Engineering 141 (2019): 1–8. A. Townsend, R. Racasan, R. Leach, N. Senin, A. Thompson, A. Ramsey, D. Bate, P. Woolliams, S. Brown, and L. Blunt, “An Interlaboratory Comparison of X-Ray Computed Tomography Measurement for Texture and Dimensional Characterisation of Additively Manufactured Parts,” Additive Manufacturing 23 (2018): 422–432. Non-Destructive Testing. Radiation Methods for Computed Tomography. Operation and Interpretation, BS EN ISO 15708-3:2019 (London: British Standards Institution, 2019). A. Buratti, N. Grozmani, C. Voigtmann, L. V. Sartori, and R. H. Schmitt, “Determination of the Optimal Imaging Parameters in Industrial Computed Tomography for Dimensional Measurements on Monomaterial Workpieces,” Measurement Science and Technology 29 (2018): 1–13. A. Townsend, L. Pagani, L. Blunt, P. J. Scott, and X. Jiang, “Factors Affecting the Accuracy of Areal Surface Texture Data Extraction from X-Ray CT,” CIRP Annals 66 (2017): 547–550. S. Carmignato, W. Dewulf, and R. Leach, Industrial X-Ray Computed Tomography (Cham, Switzerland: Springer International, 2018).

101

102

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120200003

Ahmed Tawfik,1 Radu Racasan,1 Desi Bacheva,2 Liam Blunt,1 Andre´ Beerlink,3 and Paul Bills1

Challenges in Inspecting Internal Features for SLM Additive Manufactured Build Artifacts Citation A. Tawfik, R. Racasan, D. Bacheva, L. Blunt, A. Beerlink, and P. Bills, “Challenges in Inspecting Internal Features for SLM Additive Manufactured Build Artifacts,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 102–121. http://doi.org/10.1520/STP1631202000034

ABSTRACT

Additive manufacturing (AM) is a process where the component is built layer by layer using powder or wire precursors. AM is a new and developing technology offering advantages over conventional subtractive machining in terms of design optimization and weight reduction and enabling the creation of complex internal and external features that are impossible to achieve with conventional subtractive machining. AM technologies continue to be the subject of rapid development and, consequently, the geometrical repeatability and mechanical properties of AM parts are still the subject of research. X-ray computed tomography (XCT) is a nondestructive inspection method that can be utilized in characterizing and measuring the internal defects/features of metallic AM components and is becoming the go-to tool for AM metrology. This paper presents several challenges associated with the inspection of the internal features and defects. The parts utilized in the present study were a 10-mm aluminum (AlSi10Mg) AM artifact/sample manufactured using a Renishaw

Manuscript received January 19, 2020; accepted for publication May 13, 2020. 1 EPSRC Future Advanced Metrology Hub, University of Huddersfield, Queensgate Huddersfield, HD1 3DH, UK, A. T. https://orcid.org/0000-0002-4871-5369 2 HiETA Technologies Ltd, Bristol & Bath Science Park, Dirac Crescent, Emersons Green, Bristol, BS16 7FR, UK 3 YXLON International GmbH, Essener Bogen 15, 22419 Hamburg, Germany 4 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

AM250 (Renishaw, UK) selective laser melting (SLM) AM system. The sample contains several “designed-in” internal features, varying in size from 50 :m to 1 mm, and located between 50 :m and 5 mm from the outer surfaces of the component. The features were designed as geometric features (spheres, cylinders, prisms, and helical prisms). A Nikon XTH 225 (Nikon Tring, UK) industrial XCT was used to analyze the internal features’ location, form, and volume. The results from the XCT were compared to the prebuild slicing software to attempt to identify the cause of the variation from design. The sample was then physically sectioned to confirm the actual variation of the features from the design intent. After sectioning, the defects were characterized/verified using an Alicona G4 (Alicona, Graz) focus variation instrument. Data processing, surface determination processes, and defect analysis were carried out using VG Studio Max 3.1 (Volume Graphics, Heidelberg). The focus of this study is on identifying the limitations in designing, building, and characterizing micro internal features in AM SLM components. Keywords additive manufacturing, powder bed fusion, selective laser melting, internal features, nondestructive inspection, porosity analysis, X-ray computed tomography, focus variation

Introduction In recent years, manufacturing technologies have evolved rapidly, enabling manufacturers to acquire additive manufacturing (AM) machines at costs comparable to multiaxis computer numerically controlled (CNC) machines. There are many advantages of AM over conventional subtractive machining, and these are essentially based around the geometrical design freedom; for example, by using AM, now it is possible to print components with complex internal and external features for functions such as internal cooling channels or optimized weight reduction.1 The selective laser melting (SLM) technology was first developed by Pierre Ciraud, who filed a patent application in 1971 for manufacturing components by fusing powdered material onto a surface using a laser prior to solidification (also using a laser).2–4 This technology incorporates the use of advanced techniques such as topology optimization and the incorporation of trabecular structures into components.5–8 This has given SLM significant potential in aerospace and orthopedic implant applications in particular. Unfortunately, however, several disadvantages prevent manufacturers from more widely exploiting AM technologies. A significant barrier in the perception of AM component end users is a lack of understanding of the structural integrity of AM components. Currently, there are several AM technologies using powder or wire for direct deposition; however, a more common range of technologies utilize powder bed fusion (PBF), where the energy source for powder melting is provided by lasers or .

103

104

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

electron beams. A further group of techniques uses a binder to achieve powder adhesion prior to a sintering process. A metrology technology that has advanced rapidly in recent years is X-ray computed tomography (XCT). XCT is a nondestructive inspection method that has the capability to measure geometry—both external and internal. Crucially, XCT also has the capability to detect internal porosity, providing position, location, distribution, and volume of defects.9 Generally, in XCT, the component is mounted to a rotating stage and a penetrating X-ray beam is projected through the component. The attenuated beam is then collected at a detector. The obtained data are filtered by back projection algorithms and can be reconstructed as a three-dimensional (3D) model of the measured component. There are several proprietary software analysis systems that threshold and differentiate between the gray values of the object and the background based on data collected from the detector. The reconstructed 3D model consists of several million 3D pixels (voxels). Each voxel has a unique gray value; the lighter gray values are always assigned to denser materials, and the darker gray values are assigned to less dense materials or air.10,11 There are several limitations when detecting porosity with XCT, for example: part size, spatial resolution and its impact on detectability, and repeatability and reproducibility.12–14 Several researchers attempted to explore the overall accuracy of XCT scanning. Shah et al.15 compared the accuracy of CT measurement to the coordinate measurement machine (CMM; the traceable CMM was considered to provide benchmark values). The authors used polymer artifacts with several external geometric features. The artifact features ranged in size from 2 mm to 8 mm. The authors faced difficulties of surface determination of the XCT data; some features such as cubes were found to have a 19% percentage dimensional error compared to the CMM, and other features such as cylindricity and parallelism showed a 86% error compared to CMM results.15 More encouragingly, Capel et al.16 investigated the printing resolution of a miniaturized reactor with diameters varying from 1 mm to 3 mm using an SLM process.16 The authors used a Renishaw AM250, and the printing material used was titanium Ti6AL4V. The part geometrical resolution was “acceptable,” but digitally removing the semifused powder proved to be a difficult challenge. Further data in metal AM printing were provided by Zachary et al.17 They designed and printed a Nickel 625 alloy artifact with internal defects ranging in size from 0.5 mm to 2.5 mm, as shown in figure 1. An EOS M290 Direct Metal Laser Sintering machine was used. The print layer thickness was 40 lm; this investigation highlighted the relation between subsurface defects and chevron patterning on the upward facing surface. Furthermore, this experiment showed that the defect takes up to three layers (120 lm) to “fill over” and up to eight layers (320 lm) for the defect traces to disappear.17 Looking more deeply into the AM process, du Plessis et al.18 investigated the correlation between voids/pore morphology and improper process parameters. The authors used 10-mm Ti6AL4V cube-shaped coupons for the study, which investigated two different types of pores, subsurface lack of fusion due to improper .

TAWFIK ET AL., DOI: 10.1520/STP163120200003

FIG. 1 Defect propagation through build (0–8) layers from left to right: (A) open hole, (B) keyhole, and (C) lack of fusion.17

contour scanning, and pores growing in the build direction. In the experiment, the authors used a micro CT, and the sample was scanned with a 15-lm voxel size. The data were analyzed with VGStudioMAX3.1 software. The authors discussed the challenges in the process data containing unfused/semifused powder. Surface determination protocols and the impact of noise on the accuracy of the measurement process were also discussed. In PBF processes, identifying the melt pool dimension is crucial, not only due to its effect on geometrical features accuracy but also on contour chevron patterning, which affects the final surface finish. Furthermore, the metallurgical grain morphology/microstructure can be predicted from the melt pool dimensions.19,20 Additionally, it has been found that the average residual stresses increased with the number of melted powder layers, which are primarily a function of the melt pool size.19,20 Several researchers have investigated the simulation of the melt pool in the AM process.21–23 Zinoviev et al.21 successfully adapted the Goldak heat source model to predict the evolution of grain structure, but the authors recommended taking the 3D effects into account for accurate micro structure prediction such effects could be measured using XCT methods. The present study focuses on the challenges of using XCT as a metrology tool to establish validity of build parameters for a given material when producing internal features. The work investigates establishing the limitations of manufacturing and XCT inspection of an aluminum test artifact produced by the SLM AM process.

Study Rationale In an attempt to develop technologies capable of assessing AM part integrity, several studies have analyzed the accuracy of measuring internal features or simulated defects in components. Such “defects” were machined accurately with conventional .

105

106

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

subtractive machining and then measured with optical measurement systems such as a focus variation system. These results were then compared to the XCT results.9,24 The problem with these methods was their poor representation of the various AM defects existing in real AM parts due to the lack of or limited powder fusion. Identifying the proper XCT settings for accurately detecting unfused and semifused powder (shown in fig. 2) in the AM process and understanding the effects of semifused and unfused powder on part metrology were the key drivers for this reported research. SLM operators widely believe that the printing resolution is directly related to the laser spot size and that it is not possible to print features smaller than this. This, however, is not entirely true; the melt pool can effectively melt several layers beneath the surface. Consequently, a secondary driver for this study is identifying the limitation of manufacturing capability of geometrical features/internal features as they approach laser spot or powder particle size. While tensile strength of AM components can be compared to their wrought alternative, fatigue strength of AM components can be significantly reduced due to the subsurface porosity.25–29 Several studies investigating the fatigue life of AM mechanical components confirmed that the fatigue cracks often initiate from subsurface pores/defects.29–33 Consequently, a further objective of this study is to assess whether proximity to the surface affects the SLM internal feature’s manufacturability. In this study, different geometries with various sizes have been built as close as 50 lm from the surface to duplicate subsurface pores. The final objective of the present work is to investigate the correlation of the slicing software images with the actual built sample using XCT verified by physical sectioning.

Methodology The designed and manufactured test artifact shown in figure 3 contains 64 internal features ranging from 50 lm up to 1 mm. The geometrical figures include centerlocated cylinders and prisms, surface cylinders, spheres, edge-truncated prisms, two 350-lm internal cylindrical channels, and two truncated prism helix/spirals. FIG. 2 High-resolution (4-μm voxel size) XCT image of unfused and semifused powder in an enclosed internal feature.

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

FIG. 3 (A) Artifact 3D model and (B) XCT (26 μm) image of the artifact with the internal features highlighted.

The artifact diameter was 11 mm and the length was 24 mm; on the outside, two circumferential marks were printed. These were used as physical sectioning marks for location verification. The artifact was built using a Renishaw AM 250, and the powder used was aluminum AlSi10Mg recycled eight times. The used powder was analyzed by scanning electron microscopy (SEM) prior to machine filling. The powder was 20 to 45 lm in diameter in the virgin state prior to recycling. The used artifact was placed on the far right corner of the build plate. The built settings used on AM250 are shown in table 1. The scanning pattern utilized is shown in figure 4—the outer contour with hatches in arrays of parallel stripes. In this study, the primary XCT investigation was carried out by scanning the full part at low magnification (voxel size: 26 lm); then, key zones of the part were scanned at higher magnification (4-lm voxel size). Finally, the artifact was physically sectioned, and the center rectangles were characterized using a focus variation microscope (Alicona G4) to determine reference geometey values and ensure the reliability of the results.34 Sample spacing was 10 lm, the data were not filtered, in addition the Alicona was calibrated per the manufacturer’s recommendations. A Nikon XTH225 industrial CT was used for the low-magnification (26-lm voxel size) scan; the settings used for the low-magnification scan are shown in table 2.

TABLE 1 Build parameters

.

Laser Power

Laser Focus

Laser Speed

Point Distance

Exposure Time

Point Jump Delay

Jump Speed

Jump Delay

200 W

0 mm

0.55 m/s

80 :m

140 :s

NULL

NULL

NULL

107

108

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 4 Features close to the edge: (A) CT image of the area of interest and (B) QuantAM software slice.

TABLE 2 Low-magnification CT settings

Filter

0.25 mm

Exposure

Filament Current

Acceleration Voltage

Voxel Size

4,000 ms

58 :m

135 kV

26 :m

The high-magnification scan (4-lm voxel size) was carried out using a YXLON FF20 CT scanner (YXLON International GMBH) utilizing a YXLON FXE 190.61 X-ray tube and Varex Imaging Detector, Model 4343CT. The settings used are shown in table 3. The results obtained are presented in the form of a histogram, shown in figure 5, where the voxel gray value is plotted against the number of voxels. A typical histogram of a part contains two sets of peaks—one representing the material and one representing the background air. The automatic threshold (ISO) value is 50%.35,36 This is set exactly between these two peaks and delineates part materials from the background air. This standard method has been shown by the present authors to induce errors. Consequently, the present study uses a more iterative method.37

TABLE 3 High-magnification CT settings

Filter

0.25 mm .

Exposure

Filament Current

Acceleration Voltage

Voxel Size

4,000 ms

58 :m

135 kV

4 :m

TAWFIK ET AL., DOI: 10.1520/STP163120200003

FIG. 5 Surface determination histogram.

The surface determination methodology used presently is uniquely optimized to detect and measure internal features/defects.38 The method shown in figure 6A is applied by identifying the gray value of the trapped air between the unfused powder within the denser section of the work piece (longest X-ray path) shown in figure 6B.

FIG. 6 Surface determination optimization methods.38

.

109

110

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

This exact gray value is used as the threshold for the defect analysis. This method enhances subvoxel detection by taking into consideration the voxels that contain air and material. Following the application of the method shown in figure 6, this study adopted the surface determination threshold of 62%.

Results The results from the present study are divided into three different sections: (i) highand low-magnification XCT results, (ii) focus variation (Alicona) verification, and (iii) comparison of the XCT model to QuantAM slicing software. The low-resolution (26 lm) results were mainly used for assessing the proximity to edge/surface effects, location verification, and methodology qualification. The high-resolution (4 lm) results were used for feature dimension comparison. In the designed artifact, several cylinders and prisms were designed close to the outer surface/edge of the part. The feature distance from surface varied from 50 lm to 350 lm; any features closer than 100 lm from the surface were not designed to be enclosed. Figure 4A shows an XCT image of the edge cylinders; it is clear that the closest cylinder to the surface is open and not enclosed. This was also highlighted in figure 4B, which shows the slicing software; here, the feature was not enclosed. As described previously, there were two circumferential location marks printed in the middle of the artifacts; the indication marks were located in the central zones of the five cylinders. Figure 7 shows the location indication marks and illustrates the location verification process. The center of the cylinders was designed to be 14.40 mm from the base; the location was verified by the XCT and found to be approximately 14 lm from the base, 18 lm lower than the designed location. Another method used for location verification used the lower and upper helix features and compared the angular position to the computer aided design (CAD) drawing. The verification proved that the angular location was identical to CAD, with the exact start and end location within the XCT voxel resolution. The artifact was sectioned in the middle in between the indication marks using a CNC machine. The sectioned slice is shown in figure 4. The length and width of the five cylinders were verified using an Alicona G4 Focus Variation Microscope (Alicona Austria). The five cylinders are numbered one to five from left to right; the length is named

FIG. 7 Location verification and physical section slice.

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

the major axis and the width as the minor axis. The major and minor axes for the five cylinders shown in figure 8 were used for the methodology qualification by comparing the design dimension to the Alicona-based measurement, then to the XCT ISO 50% threshold, and then, finally, to the custom threshold method.38 The dimensional comparison is shown in figure 9. The minor (width) values were lower than the designed values; the only exception was Cylinder 5, with 1% more than the designed width. The difference between the reference Alicona results and the custom threshold (surface determination) for Cylinder 1 is 1.8%, Cylinder 2 is 0.5%, Cylinder 3 is 4.5%, and Cylinders 4 and 5 are 0.9% and 1%, respectively. While the difference between the Alicona reference values and ISO 50% threshold for Cylinders 1 and 2 are 6% and 4.8%, respectively, Cylinders 3 and 4 are 11.4% and 1.6%, respectively, and for Cylinder 5, 5.4%.39 In the case of major (length) comparison, the custom threshold values were closer to the Alicona reference results than those obtained from the ISO 50% threshold. The custom threshold values’ differences from Alicona for Cylinders 1 and 2 are 0.5% and 0.9%, respectively, and for Cylinders 3 and 4 are 0.3% and 0.4%, respectively. In the case of the ISO 50% threshold, the difference for Cylinder 1 is 9% and for Cylinder 2 is 7%. For Cylinders 3 and 4, the differences are 7.8% and

FIG. 8 Minor and major axis of printed cylinders.38

FIG. 9 (A) Minor axis comparison; (B) major axis comparison.

.

111

112

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

5.1%, respectively. The length for Cylinder 5 was not compared due to the presence of semifused powder. The results show that across all measurements, the optimized surface determination method gives the least error compared to the verified values. The next section of the results investigates the effect of feature geometry on SLM manufacturability and the minimum build resolution. The results for each feature are shown in table 4. The sphere minimum resolution was 100 lm; the 50-lm spheres were not seen to be present within the build. The smallest truncated prisms were designed with a 100-lm base and a 50-lm top face. These prisms were not present in the build, but the second smallest truncated prism, with a 150-lm base and a 100-lm top face, were detected as present. The next feature was the vertical cylinder; the vertical cylinders with a diameter less than 150 lm were not present. In the case of the horizontal cylinders, any cylinder with a diameter less than 100 lm was not present. Finally, the helix and spiral features were present, and high-resolution scanning revealed that the start and end angular position were very accurate, matching the designed location.

Discussion The characterization of semifused and unfused powder is one of the biggest challenges in the metrology of AM components. This study highlighted the effect of semifused and unfused powder presence on the part metrology. Generally, powder particle size plays a very important role in the AM process; it was clear in this study that the presence of large powder particles, as shown in figure 10, had a negative impact on the feature resolution, as shown in figure 11. This is mainly because bigger particles require higher energy to melt.40 This lack of required energy will result in smaller melted powder particles being fused to the bigger unmelted particles, creating semi- or unfused powder zones within the build. This phenomenon will result in poor printing resolution for small (under 500 lm) geometric features as well as poor overall structural integrity due to the presence of internal pores/defects. The SEM images highlighted the presence of nonspherical particles, as shown in

TABLE 4 Minimum printed feature resolution

Feature

Truncated Prism

Designed Size

Build Resolution Limit

Base size 100 :m–500 :m

Base size 150 :m

Top face 50 :m–250 :m

Top face 100 :m

Vertical Cylinders

50 :m–1,000 :m dia. 

150 :m dia.  500 :m length

Horizontal Cylinders

50 :m–1,000 :m dia. 

500–1,000 :m length 100 :m dia.  500 :m length

1,000 :m length Spheres Helix/Spiral

50 :m–1,000 :m dia.

100 :m dia.

Base size 350 um

Not applicable

Top face 100 :m

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

FIG. 10 SEM image of the powder used: (A) 300-μm magnification, (B) 100-μm magnification, and (C) 50-μm magnification.

FIG. 11 XCT images of the internal features containing semifused and unfused powder.

figure 10A and 10C. The irregular-shaped powder can cause packed pores between

the powder particles. Figure 10B shows a powder particle with surface humidity. The presence of humidity can cause splatter. In PBF, splatter is one of the causes of porosity and also compromises the powder quality and spreadability by introducing shape irregularity,41 reducing its service life. The existence of humidity indicates the presence of high oxygen content; the high oxygen levels will lead to a decrease in ductility and impact toughness of printed parts. Typical recycled powder, as used in this study, can contain more than 55% more oxygen than virgin powder.42,43 That virgin powder would not suffer such issues and feature resolution could be improved is under consideration. The presence of humidity allegedly has affected this experiment by introducing spherical shape irregularity, which may have compromised the printing resolution. In light of the aforementioned statement, further investigation is needed. .

113

114

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

In the present study, the presence of the semifused powder illustrated the minimal difference in gray values between semifused powder and fully dense materials; similar gray values directly affect the internal features’/defects’ detectability. A large difference in gray values will result in better contrast,44 resulting in better surface determination and more accurate defect analysis. The surface determination study proved that an ISO 50% threshold is not adequate for detecting semifused powder zones. Consequently, a more bespoke approach is required. This approach encompasses acquiring the gray value of the pores in the denser section of the part (longest X-ray path). The threshold can vary among scans due to different scanning parameters, part shapes, and dimensions. Furthermore, the gray value of the unfused powder and semifused powder will always be closer to the full dense material. Even the air between the powder particles enclosed inside the part will have a higher gray value than the background air found outside the part.37 Thus, these issues justify the use of the optimized surface determination method. All the internal features are affected directly by the melt pool dimensions. Any feature less than 150 lm did not conform to the designed shape, and any feature closer than 100 lm from surface was not enclosed. This is mainly due to melt pool dimensions, which are a 150-lm length, 90-lm width, and a 180-lm depth for the settings used.19,45 One of the interesting points found in the present study concerns the slicing software accuracy because the software used did not consider melt pool or the solidification process. Figure 12 shows a comparison of (A) the slicing software, (B) CT image, and (C) sectioned part at the exact same location. What is clear from the figure is that the distance from the cylinder end face and the outer surface of the part is less in the actual builds than in the slicing image. Additionally, the length and width of the cylinders were less than designed. Consequently, the spacing of the end faces was larger than designed. It is clear that slicing software also does not take into consideration the powder particle size and degradation from recycling, which in this case

FIG. 12 (A) Slicing software image, (B) CT image, and (C) sectioned part image.38

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

shows as internal feature geometry errors. For the presently used slicing software, there appear to be shortcomings in determining the manufacturability of the parts containing small intricate features near the surface. In this instance, care/experience is needed in order to realize such features.

Conclusions This study discussed a recently developed artifact manufactured to understand AM build feature resolution using a given set of manufacturing conditions. This artifact established process capability of the SLM AM for manufacturing small internal features for this particular powder condition. The study highlighted the limitation of the manufacturing capability for specific geometrical features (i.e., internal features approaching the size of laser spot/powder size and how this appears to be degraded with repeated recycling of powder). The artifact enabled the assessment of the capability of the XCT metrology process and proved that optimization of surface determination can enhance the capability of nondestructive inspection to detect unfused or semifused powder and the difference in gray value between them. From this study, it can be concluded that features with sharp corners are poorly resolved at small dimensions (e.g., spheres at 100 lm, prism at 150 lm). The internal feature locations appear to be accurate within the mechanical sectioning tolerance (20–40 lm). The proximity to the surface does have an effect on manufacturability of features close to the surface; however, if the feature is closer than 100 lm, it will not print as a closed feature. In future studies, samples can be designed with subsurface pores to accurately evaluate components’ fatigue life. The powder size and quality directly affect the internal feature resolution and the overall AM part’s structural integrity. Further study is needed on assessing melt pool dimensions and how they affect feature manufacturability using bespoke artifacts. In further studies, samples can be designed with subsurface pores to accurately evaluate the effect on component fatigue life. The methodology described here, concerning manufactured test artifacts, can be adapted to other AM processes such as electron beam melting or binder jet and other build materials such as titanium, Inconel, and so on.

.

115

116

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Appendix DESIGN Feature

Edge spheres

Size

Location

50 :m to 1 mm 500 :m apart 80 :m to 500 :m from edge

Edge cylinders

50 :m–1 mm diameter  500 :m length 100 :m to 500 :m from edge

Center cylinders and prisms

Cylinders: 50 :m to 1 mm diameter 500 :m apart Truncated prisms: Base 500 :m top face 250 :m 500 :m apart

(continued)

.

TAWFIK ET AL., DOI: 10.1520/STP163120200003

Feature

Size

Edge prisms

Location

Base: 100 :m to 500 :m Top face: 50 to 250 :m 500 :m from edge

Internal channel

150 :m from surface 200 :m diameter 60 and 115 sweep angles

Helix/spiral

Base: 350 :m Top: 100 :m Start point: part center End: 1.5 mm from edge The feature was mirrored.

AM BUILD

.

Laser Power

Laser Focus

Laser Speed

Point Distance

Exposure Time

Point Jump Delay

Jump Speed

Jump Delay

200 W

0 mm

0.55 m/s

80 :m

140 :s

NULL

NULL

NULL

117

118

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

CT Settings LOW MAGNIFICATION Filter

0.25 mm

Exposure

Filament Current

Acceleration Voltage

Voxel Size

4,000 ms

58 :m

135 kV

26 :m

Exposure

Filament Current

Acceleration Voltage

Voxel Size

4,000 ms

58 :m

135 kV

4 :m

HIGH MAGNIFICATION Filter

0.25 mm

References 1. 2.

3.

4.

5.

6.

7.

8.

9.

10.

.

N. Hopkinson, R. Hague, and P. Dickens, Rapid Manufacturing: An Industrial Revolution for the Digital Age (West Sussex, UK: Wiley, 2006). P. Ciraud. Verfahren und Vorrichtung zur Herstellung beliebiger Gegensta ¨nde aus beliebigem schmelzbarem Material [Method and device for producing any object from any meltable material]. German Patent DE 2263777, filed December 28, 1971, and issued July 5, 1973. M. Shellabear and O. Nyrhila ¨, “DMLS—Development History and State of the Art,” in Proceedings of the Fourth Laser Assisted Net Shape Engineering, LANE, Vol. 1 (Bamberg, Germany: Meisenbach, 2004), 393–404. M. Baumers, C. Tuck, R. Wildman, I. Ashcroft, and R. Hague, “Shape Complexity and Process Energy Consumption in Electron Beam Melting,” Journal of Industrial Ecology 21, no. S1 (2017): S157–S167. A. Aremu, I. Ashcroft, R. Wildman, R. Hague, C. Tuck, and D. Brackett, “The Effects of Bidirectional Evolutionary Structural Optimization Parameters on an Industrial Designed Component for Additive Manufacture,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture 227, no. 6 (2013): 794–807. A. Aremu, I. Ashcroft, R. Wildman, R. Hague, C. Tuck, and D. Brackett, “The Effects of Bidirectional Evolutionary Structural Optimization Parameters on an Industrial Designed Component for Additive Manufacture,” Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture 227, no. 6 (2013): 794–807. L. Parry, I. A. Ashcroft, and R. D. Wildman, “Understanding the Effect of Laser Scan Strategy on Residual Stress in Selective Laser Melting through Thermo-Mechanical Simulation,” Additive Manufacturing 12, pt. A (2016): 1–15. I. Maskery, A. O. Aremu, M. Simonelli, C. Tuck, R. D. Wildman, I. A. Ashcroft, and R. J. M. Hague, “Mechanical Properties of Ti-6Al-4V Selectively Laser Melted Parts with Body-Centred-Cubic Lattices of Varying Cell Size,” Experimental Mechanics 55 (2015): 1261–1272. A. Tawfik, P. Bills, and L. Blunt, “Characterisation of Powder-Filled Defects in Additive Manufactured Surfaces Using X-Ray CT” (poster presentation at Eighth Conference on Industrial Computed Tomography, Wels, Austria, February 6–9, 2018). A. Tawfik, S. Nicholson, R. Racasan, L. Blunt, and P. Bills, “Utilizing Detector Filters for Noise Reduction in X-Ray Computer Tomography Scanning for the Inspection of the Structural Integrity of Additive Manufactured Metal Parts,” Smart and Sustainable Manufacturing Systems 3, no. 1 (2019): 18–30, https://doi.org/10.1520/SSMS20180042

TAWFIK ET AL., DOI: 10.1520/STP163120200003

11. 12.

13. 14.

15.

16.

17.

18.

19.

20.

21.

22. 23.

24.

25.

26.

.

A. Townsend, N. Senin, L. Blunt, R. K. Leach, and J. S. Taylor, “Surface Texture Metrology for Metal Additive Manufacturing: A Review,” Precision Engineering 46 (2016): 34–47. D. Wildenschild, C. M. P Vaz, M. L. Rivers, D. Rikard, and B. S. B. Christensen, “Using X-Ray Computed Tomography in Hydrology: Systems, Resolutions, and Limitations,” Journal of Hydrology 267, nos. 3–4 (2002): 285–297. O. J. Tretiak, “Noise Limitations in X-Ray Computed Tomography,” Journal of ComputerAssisted Tomography 2, no. 4 (1978): 477–480. T. B. Kim, T. Bo, S. Yue, Z. Zhang, E. Jones, J. R. Jones, and P. D. Lee, “Additive Manufactured Porous Titanium Structures: Through-Process Quantification of Pore and Strut Networks,” Journal of Materials Processing Technology 214, no. 11 (2014): 2706–2715. P. Shah, R. Racasan, and P. Bills, “Comparison of Different Additive Manufacturing Methods Using Computed Tomography, Case Studies in Nondestructive Testing and Evaluation 6, pt. B (2016): 69–78. A. J. Capel, S. Edmondson, S. D. R. Christie, R. D. Goodridge, R. J. Bibb, and M. Thurstans, “Design and Additive Manufacture for Flow Chemistry,” Lab on a Chip 13, no. 23 (2013): 4583–4590. Z. Reese, J. Fox, F. Kim, J. Taylor, and C. Evans, “Effect of Subsurface Defects on the Surface Topography of Additive Manufactured Components,” National Institute of Standards and Technology (NIST), 2018, https://web.archive.org/web/20200508130315/ https://www.nist.gov/publications/effect-subsurface-defects-surface-topography-additive-manufactured-components A. du Plessis, P. Sperling, A. Beerlink, L. Tshabalala, S. Hoosain, , N. Mathe, and S. G le Roux, “Standard Method for MicroCT-Based Additive Manufacturing Quality Control 2: Density Measurement,” MethodsX 5 (2018): 1117–1123, https://doi.org/10.1016/j.mex.2018.09.006 Z. Reese, J. Fox, J. Taylor, and C. Evans, “Evolution of Cooling Length in Parts Created through Laser Powder Bed Fusion Additive Manufacturing,” National Institute of Standards and Technology (NIST), 2018, https://web.archive.org/web/20200508130412 /https://www.nist.gov/publications/evolution-cooling-length-parts-created-through-laserpowder-bed-fusion-additive I. Roberts, “Investigation of Residual Stresses in the Laser Melting of Metal Powders in Additive Layer Manufacturing” (doctoral thesis, Indiana University of Pennsylvania, University of Wolverhampton, Wolverhampton, 2019), https://web.archive.org/web /20200508130438/https://wlv.openrepository.com/handle/2436/254913 A. Zinoviev, O. Zinovieva, V. Ploshikhin, V. Romanova, and R. Balokhonov, “Evolution of Grain Structure during Laser Additive Manufacturing. Simulation by a Cellular Automata Method,” Materials & Design 106 (2016): 321–329. J. A. Goldak and M. Akhlaghi, Computational Welding Mechanics (New York: Springer Science, 2005). A. Foroozmehr, M. Badrossamay, E. Foroozmehr, and S. Golabi, “Finite Element Simulation of Selective Laser Melting Process Considering Optical Penetration Depth of Laser in Powder Bed, Materials & Design 89 (2016): 255–263. P. Hermanek and S. Carmignato, “Reference Object for Evaluating the Accuracy of Porosity Measurements by X-Ray Computed Tomography,” Case Studies in Nondestructive Testing and Evaluation 6 (2016): 122–127. J. M. Hyzak and I. M. Bernstein, “The Effect of Defects on the Fatigue Crack Initiation Process in Two P/M Superalloys: Part II. Surface-Subsurface Transition,” MTA 13 (1982), 45–52, https://doi.org/10.1007/BF02642414 E. Santos, F. Abe, and Y. Kitamura, “Mechanical Properties of Pure Titanium Models Processed by Selective Laser Melting,” Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science 218, no. 7 (2004): 711–719.

119

120

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

27. 28.

29.

30.

31. 32.

W. Fricke, Guideline for the Fatigue Assessment by Notch Stress Analysis for Welded Structures (Paris: International Institute of Welding, 2008).

33.

B. Zhang, D. R. Poirier, and W. Chen, “Microstructural Effects on High-Cycle FatigueCrack Initiation in A356.2 Casting Alloy,” Metallurgical and Materials Transactions A 30 (1999): 2659–2666, https://doi.org/10.1007/s11661-999-0306-3 T. Kowaluk and A. Wozniak, “A New Threshold Selection Method for X-Ray Computed Tomography for Dimensional Metrology,” Precision Engineering 50 (2017): 449–454, https://doi.org/10.1016/j.precisioneng.2017.06.020 F. Borges de Oliveira, A. Stolfi, M. Bartscher, L. De Chiffre, and U. Neuschaefer-Rube, “Experimental Investigation of Surface Determination Process on Multi-Material Components for Dimensional Computed Tomography,” Case Studies in Nondestructive Testing and Evaluation 6, pt. B (2016): 93–103. J. Yagu ¨e-Fabra, S. Ontiveros, R. Jime ´nez, S. Chitchian, G. Tosello, and S. Carmignato, “A 3D Edge Detection Technique for Surface Extraction in Computed Tomography for Dimensional Metrology Applications,” CIRP Annals 62, no. 1 (2013): 531–534. J. P. Kruth, M. Bartscher, S. Carmignato, R. Schmitt, L. De Chiffre, and A. Weckenmann, “Computed Tomography for Dimensional Metrology,” CIRP Annals 60, no. 2 (2011): 821–842, https://doi.org/10.1016/j.cirp.2011.05.006 A. Tawfik, P. Bills, L. Blunt, and R. Racasan, “Optimization of Surface Determination Strategies to Enhance Detection of Unfused Powder in Metal Additive Manufactured Components,” in Proceedings of the 2018 ASPE and EUSPEN Summer Topical Meeting: Advancing Precision in Additive Manufacturing (Raleigh, NC: American Society for Precision Engineering, 2018), 281–285, https://web.archive.org/web/20200508131737/https:// www.scopus.com/record/display.uri?eid¼2-s2.0-85060392349&origin¼inward&txGid¼ 9227030e97f695ffcd1a23285dab8340 A. Tawfik, L. Blunt, C. Dawson, R. Racasan, D. Bacheva, and P. Bills, “Optimisation of Surface Determination to Improve the Accuracy and Repeatability of Detecting Unfused Powder in AM Aluminum Component” (paper presentation, 9th Conference on Industrial Computed Tomography, Padova, Italy, February 13–15, 2019). A. B. Spiering, N. Herres , and G. Levy, “Influence of the Particle Size Distribution on Surface Quality and Mechanical Properties in AM Steel Parts,” Rapid Prototyping Journal 17, no. 3 (2011), http://doi:10.1108/13552541111124770 A. B. Anwar and Q.-C. Pham, “Study of the Spatter Distribution on the Powder Bed during Selective Laser Melting,” Additive Manufacturing 22 (2018): 86–97, https://doi.org /10.1016/j.addma.2018.04.036

34.

35.

36.

37.

38.

39.

40.

41.

.

P. Edwards and M. Ramulu, “Fatigue Performance Evaluation of Selective Laser Melted Ti-6Al-4V,” Materials Science Engineering A 598 (2014): 327–337. E. Wycisk, A. Solbach, S. Siddique, D. Herzog, F. Walther, and C. Emmelmann, “Effects of Defects in Laser Additive Manufactured Ti-6Al-4V on Fatigue Properties,” Physics Procedia 56 (2014): 371–378. S. J. Li, L. E. Murr, X. Y. Cheng, Z. B. Zhang, Y. L. Hao, R. Yang, F. Medina, and R. B. Wicker, “Compression Fatigue Behaviour of Ti-6Al-4V Mesh Arrays Fabricated by Electron Beam Melting,” Acta Materialia 60, no. 3 (2012): 793–802. H. Zhang, R. Chiang, H. Qin, Z. Ren, X. Hou, D. Lin, G. L. Doll, V. K. Vasudevan, Y. Dong, and C. Ye, “The Effects of Ultrasonic Nanocrystal Surface Modification on the Fatigue Performance of 3D-Printed Ti64,” International Journal of Fatigue 103 (2017): 136–146, https://doi.org/10.1016/j.ijfatigue.2017.05.019 A. Hobbacher, ed., Recommendations for Fatigue Design of Welded Joints and Components (Cham, Switzerland: Springer, 2009).

TAWFIK ET AL., DOI: 10.1520/STP163120200003

42.

43.

44.

45.

.

A. Strondl, O. Lyckfeldt, H. Brodin, and U. Ackelid, “Characterization and Control of Powder Properties for Additive Manufacturing,” JOM 67, no. 3 (2015): 549–554, https:// doi.org/10.1007/s11837-015-1304-0 V. V. Popov, A. Katz-Demyanetz, A. Garkun, and M. Bamberger, “The Effect of Powder Recycling on the Mechanical Properties and Microstructure of Electron Beam Melted Ti6Al-4 V Specimens,” Additive Manufacturing 22 (2018): 834–843, https://doi.org /10.1016/j.addma.2018.06.003 A. du Plessis, P. Sperling, A. Beerlink, W. B. du Preez, and S. G. le Roux, “Standard Method for Micro-CT-Based Additive Manufacturing Quality Control 4: Metal Powder Analysis,” MethodsX 5 (2018): 1336–1345, https://doi.org/10.1016/j.mex.2018.10.021 D. Abolhasani, S. M. Seyedkashi, N. Kang, Y. J. Kim, Y. Y. Woo, and Y. H. Moon, “Analysis of Melt-Pool Behaviors during Selective Laser Melting of AISI 304 Stainless-Steel Composites,” Metals 9, no. 8 (2019): https://doi.org/10.3390/met9080876

121

122

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190148

Tatiana Mishurova,1 Itziar Serrano-Mun ˜oz,1 Tobias Fritsch,1 1 Alexander Ulbricht, Maximilian Sprengel,1 Alexander Evans,1 Arne Kromm,1 Mauro Madia,1 and Giovanni Bruno1,2

A Critical Discussion on the Diffraction-Based Experimental Determination of Residual Stress in AM Parts Citation T. Mishurova, I. Serrano-Mun ˜oz, T. Fritsch, A. Ulbricht, M. Sprengel, A. Evans, A. Kromm, M. Madia, and G. Bruno, “A Critical Discussion on the Diffraction-Based Experimental Determination of Residual Stress in AM Parts,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 122–138. http://doi.org/10.1520/STP1631201901483

ABSTRACT

As opposed to reviewing results on experimental determination of residual stress by diffraction, this paper discusses the open issues when dealing with residual stress determination in additive manufactured parts, in particular those manufactured with laser powder bed fusion techniques. Three points are addressed in detail: (a) the proper determination of the strain-free reference d0, (b) the problem of the determination of the principal axes, and (c) the use of the correct diffraction elastic constants. It is shown that all methods to determine the strain-free reference d0 suffer from caveats, and care must be taken in evaluating the most suitable for the problem being tackled. In addition, it is shown that, in some systems, the principal axes do correspond to the

Manuscript received November 21, 2019; accepted for publication January 10, 2020. 1 Bundesanstalt fu ¨r Materialforschung und –pru ¨fung (BAM), Unter den Eichen 87, 12205 Berlin, Germany T. M. https://orcid.org/0000-0002-9349-0453, I. S.-M. https://orcid.org/0000-0002-5585-6637, A. K. https://orcid.org/0000-0003-0340-9656, G. B. http://orcid.org/0000-0001-9632-3960 2 University of Potsdam, Institute of Physics and Astronomy, Karl-Liebknecht-Straße 24/25, 14476 Potsdam, Germany 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

geometrical axes of the specimen, but this needs to be systematically checked, especially in the case of uni- or bidirectional hatching strategies. Finally, the need to experimentally determine the proper diffraction elastic constants is underlined, especially in the case of strongly textured specimens, which again depends on the deposition strategy. Keywords neutron and X-ray diffraction, diffraction elastic constants, strain-free reference, principal directions, nondestructive testing, residual stress

Introduction Additive manufacturing (AM) technologies have boomed in the last few years. They are very promising for applications where freedom of (structural) design is concerned, and even cost issues are being successfully addressed in the industry.1 Belated, but relevant, questions about the microstructural stability, the mechanical performance, and the structural integrity have recently become a much larger subject for debate than at the introduction of AM technologies. In particular, residual stress (RS) is now acknowledged as one of the main issues of AM parts, especially when manufactured by laser powder bed fusion (LPBF) technologies.2,3 There are a couple of aspects to take into account: (1) Internal stresses can cause rupture even during production. Their origin is being thoroughly studied both by means of experimental and (even more so) by simulations.2–5 (2) Metallic materials are often exposed to postbuild heat treatments, which aim to reduce residual stresses, close porosity, and achieve the desired microstructure and mechanical properties. Such heat treatments rest, however, on the materials knowledge of conventional manufactured products; they are therefore seldom tailored and optimized for the AM material and the component in question. (3) Properly simulating the RS in AM parts is a formidable task because a detailed knowledge of the thermal history is needed.6 Such detailed knowledge is extremely difficult to gain because the cooling rates in LPBF are enormous, and the microstructure remains extremely unstable in as-built conditions.7 Finally, experimental determination of RS, especially in complex geometries associated with the freeform designs made possible by AM, is challenging for the majority of the established techniques. The experimental determination of RS seems a valid path to acquiring more knowledge at least about the effect of different printing parameter sets, scanning strategies, and geometries. Indeed, it represents the reference for any simulation work and adds (obviously together with a detailed knowledge of the microstructure) important data on this new class of materials. Because it is always desirable to gain knowledge of the RS state in a nondestructive fashion, diffraction methods play the most eminent role in tackling this challenge. Among them, neutron and synchrotron X-ray diffraction (SXRD)8 are the most appealing because they promise determination of bulk RS, but laboratory X-ray techniques are also relevant because the manufacturing possess induces very high (tensile) surface RS (balanced by compressive bulk RS). .

123

124

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

It is well known that RS (defined as those stresses that remain inside a material/ component in absence of external loads) can be classified into three types:8 Type I: Homogeneous to a length scale comparable to the sample size (macroscopic stresses), they balance over the whole component. Their moments also balance. They induce, if relaxed, macroscopic size changes. Type II: Homogeneous to a length scale comparable to a few grains (intergranular stresses), they balance over a few grains. Their moments also balance on the same scale. They can induce macroscopic size changes, if relaxed. Type III: Homogeneous to a length scale comparable to a grain (intragranular stresses), they balance over a grain or a part thereof. They cannot induce macroscopic size changes. Such classification is relevant for two reasons. The first is that often, even in the absence of macroscopic stress, the microstructure accommodates micro-RS, and such stresses can be as deleterious as macro-RS, so that it is important to detect Type II and III RS; the second is that the diffraction method can detect all of them, unlike other methods—destructive and nondestructive. This implies that diffraction methods could yield a wider palette of information than conventional methods (e.g., contour, hole-drilling, ultrasound, or magnetic methods) when dealing with safety-relevant components. Such components are the ultimate goal of many scientific (e.g., fundamental understanding, alloy development) and technological (design optimization and so on) actions around AM. In recent times, a lot of scientific literature has appeared on the diffraction-based experimental determination of RS in AM parts.9–17 However, the critical points about diffraction techniques are not addressed in all such publications (and often also not even mentioned). We will see that the classic issues to deal with in diffraction-based stress analysis of components are exacerbated by the complex microstructure and the thermal history of AM parts (especially if manufactured by LPBF). Such issues are: (a) the proper determination of the strain-free reference d0, (b) the problem of the determination of the principal axes, and (c) the use of the correct diffraction elastic constants. In the following sections, we discuss these issues based on recent studies, rather than showcasing results on different materials and components.

Materials and Methods DIFFRACTION-BASED DETERMINATION OF RESIDUAL STRESS

.

There are three main techniques we briefly describe in the following text: neutron, synchrotron radiation, and laboratory X-ray diffraction. This description is by no means exhaustive and is only meant to make this paper self-consistent. The reader is referred to the books of Hauk,8 Hutchings et al.,18 Noyan and Cohen,19 and Spieß et al.20 for comprehensive treatments of many of the theoretical and experimental methods referred to in this work. All diffraction techniques are based on Bragg’s law k ¼ 2dhkl sinhhkl, where k is the wavelength of the probing radiation, dhkl the interplanar distance of the lattice plane (with Miller indices hkl) under investigation, and hhkl the Bragg’s angle of

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

scattering. If we have a reference material with grains possessing interplanar distance dhkl,0 (corresponding to a diffraction angle hhkl,0) a compression strain will cause a decrease of d (increase of hhkl) and a tensile strain an increase of d (decrease of hhkl). By the comparison to the reference material, one can evaluate the lattice 0 strain e ¼ dd d0 , where obviously such strain depends on the measured lattice plane (hkl). Such strain is purely elastic: It is a dilation of the grain interplanar distance, and a slip of the planes—plastic deformation—would not be visible as a peak shift in diffraction. However, plastic deformation does manifest itself in a broadening of the diffraction peak as accumulation of intergranular stresses (Type II). From the strain, we can calculate the associated stress, if we assume that no free deformation occurs (i.e., all strain converts into stress), as r 5 C  e, where C is the stiffness tensor (fourth rank) and r the stress tensor (second rank). Later, we will discuss the approximations linked to this basic treatment. Neutron Diffraction

Neutron diffraction (ND) methods allow the measurement of the bulk lattice spacing (d) within a crystalline material because neutron penetration depth is considerably high (some centimeters) in most materials. Nevertheless, ND has the limitation of relatively low intensities leading to high acquisition times. By using an xyzx-table (fig. 1), measurements can be performed at any desired location by

FIG. 1 Schematic illustration of the main components of the E3 beamline at the BER II reactor (now closed). A collimator or slits can be used in front of the detector.

.

125

126

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

suitable positioning of the sample in the beam. The sensitivity of the ND method depends on the angular definition of the system and the volume of material sampled (i.e., the gauge volume; the smallest gauge sizes used are generally of the order of 1 mm3). Strains in the x, y, and z directions are obtained by orientating the scattering vector to correspond with the coordinate axes of the sample. The ND experiments reported in this article were carried out on the beamline E3 at BER II reactor (Helmholtz Zentrum Berlin, Germany).21 In this beamline (beam ˚ ), the gauge volume is controlled by the primary slit syswavelength set at k ¼ 1.476 A tem aperture, as well as by the secondary collimator using a focal width of 2 mm. Also, the acquisition time typically varies from 5 min up to 90 min, depending on the measured stress component and the neutron beam path length through the material. Synchrotron X-Ray Diffraction

In synchrotron radiation sources, high-energy electrons are accelerated in a circular ring to speeds close to the speed of light. The centripetal acceleration induced by the curvature of the ring causes electrons to emit electromagnetic radiation. Modern X-ray synchrotron sources can provide very intense narrow beams of highly collimated and highly penetrating (energetic) X-ray photons (between 40 and ˚ ]), which offer micron-scale spatial resolu200 keV [approximately 0.4 and 0.06 A tion combined with very fast rates of data acquisition. At this energy range, X-ray path lengths of a couple of centimeters are possible, even in steel. Nevertheless, this high-energy range inevitably leads to low scattering angles that cause gauge volumes to be elongated along the beam direction. The energy dispersive SXRD near-surface experiments reported in this work were performed at the energy dispersive diffraction (EDDI) beamline,22 synchrotron BESSY II (Helmholtz Zentrum Berlin, Germany, fig. 2). Analogue to the lab XRD, the sin2w method in reflection setup is used for assessment of near-surface RS. The gauge volume is determined by primary and secondary slit systems. The probed region (i.e., the information depth) is usually smaller than the immersion depth of the gauge volume. Note that both beam attenuation (according to Lambert-Beer’s law) and gauge volume shape determine the effective information depth. FIG. 2 Schematic illustration of the main components of the EDDI beamline at the BESSY II synchrotron.

.

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

Laboratory X-Ray Diffraction

Laboratory X-ray diffraction (Lab XRD) utilizes sealed tubes either with static or rotating anodes to produce X-rays by accelerating electrons (originated by heating a tungsten filament in vacuum) through a high potential field and directing them into a target (the characteristic wavelength k of the tube depends therefore on the target material), which then emits X-rays. The maximum voltage usually is 50 or 60 kV, which limits the probed region to the surface. Automated X-ray sources equipped with diffractometers are usually in operation for 24 h a day in many laboratories. The lab XRD experiments conducted at BAM were performed using a Xstress G3 diffractometer (fig. 3) from StressTech, applying the sin2w-method.23 The measurements were carried out using an MnKa radiation source with a collimated beam with a diameter of 2 mm. The sample was aligned at every measurement point by a tactile probe in order to avoid the influence of roughness. Peak fitting was performed with the Xtronic software using a Pearson VII function.

Results and Discussion We discuss the three issues previously mentioned on the base of some worked examples. CHOICE OF THE APPROPRIATE ELASTIC CONSTANTS

The stiffness tensor converting strains into stresses must correspond to the ensemble of the measured grains: it does not coincide with the single crystal stiffness tensor. Instead, such a tensor is often identified with a macroscopic isotropic tensor (i.e., with two elastic constants, Young’s modulus E and Poisson’s ratio ), but it is clear that such an approximation is only valid under strong assumptions. A further

FIG. 3 (A) Image of the laboratory X-ray diffractometer available at BAM; (B) schematic illustration of the setup.

.

127

128

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

complication arises; the stiffness tensor C connects measured microstrains (i.e., specific to a family of grains within the material, which satisfy the Bragg’s condition during the experiment) and calculated macrostresses (i.e., not depending on the family of grain but belonging to the specimen itself). This issue has been tackled to a great extent with the use of homogenization methods, such as the Voigt and Reuß averages24,25 or the Kro¨ner scheme.26 In such ways, the so-called diffraction elastic constants (DECs) have been introduced. The aforementioned hypothesis of quasiisotropy has been frequently used because the tensor r represents the engineering stress. However, in the case of strongly textured materials (such as cold rolled plates, extruded wires, and AM components), the use of such a quasi-isotropic hypothesis is incorrect. A few works thoroughly describe more or less practical ways to correct the average stiffness tensor entries to take into account texture.27–30 The most commonly used is the replacement of C with the so-called stress factor F (very often erroneously not associated with a tensor, though; see Gna¨upel-Herold, Creuziger, and Iadicola28). This gives a good starting point for a correct diffraction-based stress analysis of AM parts, but still the single crystal elastic constants (SCECs) of AM materials remain largely unknown. In fact, tabulated SCECs are still used in the models, even if in practice the chemical composition and the microstructure of AM parts strongly differ from their analogous conventionally produced materials (we mainly mean metals and alloys here), for which such constants have been tabulated. Some efforts are being devoted to the determination of SCECs of AM materials from uniaxial tensile tests on polycrystalline materials (see Heldmann et al.31), but figure 4 (also discussed in Mishurova et al.32) shows that the inaccuracy of stress determination in LPBF Ti-6Al-4V can reach 10% because of the uncertainty on the proper DEC.

FIG. 4 DECs for LPBF Ti-6Al-4V obtained experimentally (for as-built, AB, and heattreated, HT, points) and their modeled values (lines) as a function of the orientation parameter H2. The solid cyan lines show the linear fit of experimental data and 6 5% scatters around the mean value (dashed lines) (see Mishurova et al.32).

.

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

We therefore need to practice extreme care to use the appropriate SCEC or DEC, and for AM materials, it is strongly recommended to determine them experimentally for the same material as the part under consideration. While this seems to constrain diffractionists to repeatedly measure basic quantities, it also corresponds to the level of ignorance we have about such quantities for AM materials. As for many other fields of science, we will only be able to tabulate quantities when those will have been validated by repeated measurements. By nature, the same AM material presents a large variety of microstructures, according to all the process parameters and to the subsequent posttreatments it undergoes. Additionally, it has been found that even by using the same process parameters, neither the microstructure nor the RS distribution are perfectly reproducible.9 In fact, increasing numbers of process parameters, formerly considered irrelevant, are being observed to influence the RS state. Such parameters need to be included in simulations. It is therefore unrealistic to assume that the variety of AM materials stemming from the same nominal composition would possess the same DEC (if not within a large error bar). It is necessary to collect enough experimental data and improve the current models if we want to be able to use such models in a reliably predictive fashion (for a recent attempt at suggesting a solution, see Mishurova et al.33). CHOICE OF THE APPROPRIATE STRAIN-FREE REFERENCE d0

The quantitative determination of stresses in a component requires the reliable determination of the reference d0, according to the definition of strain given earlier in this paper. This metrological problem has been tackled for decades—since the very beginning of the use of diffraction techniques for stress analysis. For laboratory XRD, the treatment of experimental data may include the determination of d0, if simplifying assumptions can be made on the stress state (e.g., plane stress) or many independent measurements are made (or both). For bulk measurements, where truly three-dimensional (3D) stress states have to be assumed a priori, some strategies have been established to tackle the problem of the d0 determination. The determination of d0 is discussed in a few papers referenced later in this paper. The possibilities offered to the diffractions are: a. Use of a powder. In the case of a bulk component original, raw powders (for all phases in the case of a composite) or mechanical filings extracted from the sample have been used (see, e.g., Thiede et al.13). They provide a global value of d0. b. Small cubes (coupons) extracted from the sample/part. This strategy has been used for quite some time16,17,34,35 and was initiated by the acknowledgment that d0 can locally vary in components where compositional or microstructural gradients (at the scale of the gauge volume) are present, such as in the case of welds. As a derivation of this strategy, combs are also widely used, especially when the stress state is biaxial (see Ganguly, Edwards, and Fitzpatrick36). In such a way, a position-dependent d0 can be determined. .

129

130

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

This has been reported in literature for the blown powder AM techniques in IN625 walls (Wang et al.37). c. Stress balance/boundary condition (e.g., plane stress for thin walls structures). Because RS needs to balance over the whole sample or over a cross section of it, such analytical conditions could greatly help in determining a global value of d0 (see Albertini et al.38). In the case of AM materials and components, analogous possibilities are available, with some (important) modifications: 1. Use of the initial powder. Raw powders have the same nominal composition of the printed material but do not undergo the same thermal history, which can lead to differences in local segregation of elements that can alter the lattice parameter. Therefore, the conditions under which they could represent a valid reference are limited for complex alloy systems. 2. Mechanical filings extracted from the sample do not suffer from the problem mentioned in Point 1 but may exhibit significant plastic deformation. As mentioned before, powders only represent a global strain-free reference, unless an array of filings are produced from a positionally distinct cross section of material. 3. Small cubes (coupons) extracted from the sample/part. This strategy ably tackles the problem of d0 variation that is common in AM components. AM materials can be strongly textured and anisotropic, which can give rise to high intergranular stresses. These microstresses may be present in mechanically macrostress-relieved cubes; the cubes must also be sufficiently small to relax the long-range macro-RS. 4. Stress balance condition. As mentioned before, this strategy is good at determining a global value of d0. An important fact to take into account in AM parts (especially LPBF-made) is that the outer surface of components is often printed with different parameters, the so-called contour-scan mentioned previously). Such an outer region usually has a different microstructure (and also a very different RS, typically highly tensile in the as-built conditions). In figure 5, the case of the d0 determination in an IN718 prism is shown (see Thiede et al.13): The cubes show extremely anisotropic values (in three perpendicular directions—N-normal, T-transversal, and L-longitudinal) and possibly only an average could be considered (see Altenkirch et al.39). The filings from the prism (SP) show consistent values among different measurements (and agree with the average of the coupons within the error bars). Indeed, they show a large plastic deformation, as the peak full width at half maximum values show. The use of filings seems to be a good choice in the case where a global d0 value can be applied. If surface RS is of importance (e.g., for contour scans), the use of laboratory XRD can yield d0 under the assumption of vanishing normal stress at the surface. In general, a global d0 is acceptable for simple-shaped components (such as the case in fig. 5) in the absence of strong microstructural variation within the part. .

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

FIG. 5 Determination of d0 by means of different reference methods: raw powder (RP), filings (SP), and cubes (three principal components of the cube L,T,N).13 Note that while the RP could seem to be a valid reference, its use would be incorrect.

The latter strategy also holds in the case where stress balance conditions are utilized. Figure 6A shows a planar distribution of RS that was mapped at half the build height of a prism of AISI 316L, as measured by ND. A certain value of d0 for the 311-Fe reflection is calculated upon the application of the stress balance ˚ ) correcondition. The value of d0 calculated from ND data only (d0 ¼ 1.075 A sponds to the average d0 value over three orthogonal directions measured in a ˚ ). However, in this case, the contour scan small sectioned coupon (d0 ¼ 1.075 A FIG. 6 ND normal RS at the mid-height of an AISI 316L prism (d0 was determined using the stress balance condition for the normal component, which corresponds to the build direction): (A) as measured by ND without surface stresses and (B) stress-balanced results after taking into account additional XRD surface stresses. The dots within the figures represent the interpolation grid.

.

131

132

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

needs to be taken into account because high tensile stresses near the surface are inherent to AM metallic parts produced by LPBF.2 Therefore, the stress balance must include the surface regions (measured by XRD). The stress balance condition, including (sub)surface tensile stresses, resulted in higher compressive ˚ . The differences stresses in the bulk as well as in an increased d0 value of 1.075 A between the d0 value determined using solely the ND and that using both data sets can be observed when comparing figure 6A and 6B. PROPER DETERMINATION OF THE PRINCIPAL AXES

However trivial it may sound, different scanning strategies yield different RS states. More importantly, it cannot be assumed that the principal axes follow the geometrical ones, if the scanning strategy does not follow them too. This is analogous to the case of a V-shaped weld (see, e.g., Bruno40); it is well-known that the principal axes rotate at the border between the weld pool and the parent material (i.e., in the heat-affected zone). In figure 7, the RS along building direction in

FIG. 7 ND RS LPBF IN718 prism printed with different scanning strategies: (A) T-scan, (B) L-T-scan, and (C) rotation-scan. Note that rN corresponds to the building direction.

.

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

IN718 prisms with three different scanning strategies (along the T direction of the sample, T-scan; alternated along the L and T directions at each second layer, L-T-scan; and inclined 67 at each layer, rotation-scan) is shown, as determined by ND. The difference among stress maps is driven by the orientation of the scanning vectors. Thus, the T-scan sample presents a stress gradient along the T direction, while the two other samples show more homogeneous RS distribution in the plane perpendicular to the build direction (the hatch length is half of the sample width in the T direction). In the case of the L-scan sample, additional directions to the three perpendicular geometrical directions (N, T, L) were investigated, using the sin2w method. In that case, it was found that the d versus sin2w plots were basically linear, as displayed in figure 8A. This confirms that the principal directions corresponded to the geometrical ones. For the rotation-scan sample, more directions (þ180 from principal directions) were measured by SXRD to estimate the presence of shear RS at the top surface of the sample. Also in this case, the sin2w plots presented no splitting for both transversal and longitudinal component scans, indicating absence of shear stresses at the top surface. Obviously, this is not the case for every scanning strategy and every component geometry. In particular, truly anisotropic scanning strategies or geometries (such as lattice structures) need to be properly evaluated; d versus sin2w plots are recommended or, in any case, more than six independent directions should be measured (see the thesis of T. Thiede41). SAMPLE ALIGNMENT

One more issue needs to be mentioned: sample alignment. Several strategies are used to properly align samples and components. Typically, lasers and theodolites are used at both neutron and synchrotron radiation sources; they provide a

FIG. 8 (A) ND d versus sin2w plots in the LPBF IN718 prism with L-scanning strategy. (B) Synchrotron XRD d versus sin2w at the surface of 67 rotation LPBF IN718 prism. The labels T þ 180 and L þ 180 correspond to the w angles.

.

133

134

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

positioning accuracy sometimes better than 50 lm. Such strategies are usually followed by the so-called entrance-scans, that is, by positioning the specimen with respect to the beam (as opposed to positioning with respect to the center of rotation). Needless to say, these procedures are beamtime consuming. This has been partially addressed at certain neutron instruments using the SScanSS software and metrology approach (see the works of James et al.42). Dial gauges are frequently used on laboratory X-ray diffractometers. They allow a precision of the order of a few micrometers. In the case of LPBF parts, the surface roughness is typically of the order of the powder particle size (i.e., 20–50 lm). While for ND this may not be critical, such a roughness implies that when using synchrotron radiation and laboratory XRD, extreme care must be taken in determining the sample surface. In fact, in the case of SXRD, the gauge volume is often of the order of 50 lm (at least in one direction), and in the case of XRD, the penetration depth of the radiation lies even below such roughness. In the case of curved surfaces (e.g., lattice structures), sample positioning can be an even bigger issue. It is recommended to properly take into account the geometry of the surface (i.e., to measure the roughness and adjust the “true” position of the surface by a correction). Figure 9 shows different stress scans at the surface (using synchrotron radiation) on a Ti-6Al-4V prism produced without contour scan (higher roughness) and with contour scan (decreased roughness).10 Normally, when roughness is low, the scan starts from the surface, which is measured by a laser pointer. In this case, for a sample without contour, the scan showed negligible RS at the very surface; therefore, two additional scans starting from 150 and 250 lm below were

FIG. 9 White beam (energy-dispersive) synchrotron radiation residual stress scanning on a Ti-6Al-4V prism—comparison of scans starting from the measured surface and from 150 and 250 μm below it (see Mishurova et al.10).

.

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

performed (fig. 9). In fact, after penetrating deeper into the sample (250 lm), well beyond the region influenced by the surface roughness, the stress profiles of the samples with and without contour almost coincide. This fact proves that surface roughness can lead to erroneous determination of surface RS and should always be considered as an influencing factor.

Concluding Remarks In the case of AM materials and components, and especially those produced by LPBF, the microstructure often contains elongated grains with a distinct texture. Such a microstructure is also sometimes very unstable, as demonstrated by its spectacular evolution under heat treatments. (For instance, from as-built to annealed IN718, the microstructure goes from elongated to nearly equiaxed grains.) Moreover, a mesostructure is created by the laser passes (the scanning strategy), which is particularly visible by the alignment of internal defects or of surface features (or both). On top of that, the surface is often very rough (some authors even talk about waviness rather than roughness). All these facts impart particular features to the RS fields and require particular care when determining them by means of diffraction methods. Some of the classic assumptions of RS analysis need to be verified nearly case by case. We focused on three main aspects in detail: 1. Use of proper DECs. The classic problems created by texture and coarse grains become ordinary in the case of AM materials. They cannot be ignored (as is done in many cases), and all models need to be adjusted to include the effect of texture. Moreover, the chemical composition of the AM alloys may not be the same as that of the conventional ones for which single-crystal elastic constants have been tabulated, so that some work is needed to determine such constants for AM materials. It is therefore recommended to experimentally determine the DEC in the case where the microstructure is far from that of conventionally produced materials. 2. Determination of the strain-free reference d0. Due to the complex thermal history of AM parts, one cannot always expect a constant d0 over the whole sample; therefore, cubes are recommended in some cases to obtain d0 in a spatially resolved fashion. In the case where at least the bulk of the sample can be considered homogeneous, cubes may not be suited to evaluate d0 because of the complex microstructure. In such cases, filings from the sample should be used, even if they are plastically deformed. 3. Determination of the principal stresses. Because the mesostructure mentioned previously determines the heat flow during the build job, the RS also should be expected to align to such a microstructure rather than to the geometrical axes of the sample. Therefore, verifying that the geometrical directions align to the principal stress by using the sin2w technique or measuring more than six independent strain directions is recommended. .

135

136

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

ACKNOWLEDGMENTS

Mirko Boin, Robert Wimpory, Christoph Genzel, and Manuela Klaus are kindly acknowledged for their support at BESSY II and BER II, HZB, Berlin, Germany.

References 1.

2.

3. 4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

.

A. Gisario, M. Kazarian, F. Martina, and M. Mehrpouya, “Metal Additive Manufacturing in the Commercial Aviation Industry: A Review,” Journal of Manufacturing Systems 53 (2019): 124–149, https://doi.org/10.1016/j.jmsy.2019.08.005 P. Mercelis and J. P. Kruth, “Residual Stresses in Selective Laser Sintering and Selective Laser Melting,” Rapid Prototyping Journal 12, no. 5 (2006): 254–265, https://doi.org/ 10.1108/13552540610707013 J. L. Bartlett and X. Li, “An Overview of Residual Stresses in Metal Powder Bed Fusion,” Additive Manufacturing 27 (2019): 131–149, https://doi.org/10.1016/j.addma.2019.02.020 Y. Lu, S. Wu, Y. Gan, T. Huang, C. Yang, L. Junjie, and J. Lin, “Study on the Microstructure, Mechanical Property and Residual Stress of Slm Inconel-718 Alloy Manufactured by Differing Island Scanning Strategy,” Optics & Laser Technology 75 (2015): 197–206, https:// doi.org/10.1016/j.optlastec.2015.07.009 B. Cheng, S. Shrestha, and K. Chou, “Stress and Deformation Evaluations of Scanning Strategy Effect in Selective Laser Melting,” Additive Manufacturing 12, Part B (2016): 240–251, https://doi.org/10.1016/j.addma.2016.05.007 T. Mukherjee, W. Zhang, and T. DebRoy, “An Improved Prediction of Residual Stresses and Distortion in Additive Manufacturing,” Computational Materials Science 126 (2017): 360–372, https://doi.org/10.1016/j.commatsci.2016.10.003 L. Thijs, F. Verhaeghe, T. Craeghs, J. V. Humbeeck, and J.-P. Kruth, “A Study of the Microstructural Evolution during Selective Laser Melting of Ti–6al–4v,” Acta Materialia 58, no. 9 (2010): 3303–3312, https://doi.org/10.1016/j.actamat.2010.02.004 V. Hauk, “Structural and Residual Stress Analysis by Nondestructive Methods A2,” in Structural and Residual Stress Analysis by Nondestructive Methods (Amsterdam, The Netherlands: Elsevier Science B.V., 1997). T. Mishurova, K. Artzt, J. Haubrich, G. Requena, and G. Bruno, “New Aspects about the Search for the Most Relevant Parameters Optimizing Slm Materials,” Additive Manufacturing 25 (2019): 325–334, https://doi.org/10.1016/j.addma.2018.11.023 T. Mishurova, K. Artzt, J. Haubrich, G. Requena, and G. Bruno, “Exploring the Correlation between Subsurface Residual Stresses and Manufacturing Parameters in Laser Powder Bed Fused Ti-6al-4v,” Metals 9, no. 2 (2019): 261, https://doi.org/10.3390/met9020261 T. Mishurova, S. Cabeza, K. Artzt, J. Haubrich, M. Klaus, C. Genzel, G. Requena, and G. Bruno, “An Assessment of Subsurface Residual Stress Analysis in SLM Ti-6al-4v,” Materials 10, no. 4 (2017): 348–362, https://doi.org/10.3390/ma10040348 T. Mishurova, S. Cabeza, T. Thiede, N. Nadammal, A. Kromm, M. Klaus, C. Genzel, C. Haberland, and G. Bruno, “The Influence of the Support Structure on Residual Stress and Distortion in SLM Inconel 718 Parts,” Metallurgical and Materials Transactions A 49, no. 7 (2018): 3038–3046, https://doi.org/10.1007/s11661-018-4653-9 T. Thiede, S. Cabeza, T. Mishurova, N. Nadammal, A. Kromm, J. Bode, C. Haberland, and G. Bruno, “Residual Stress in Selective Laser Melted Inconel 718: Influence of the Removal from Base Plate and Deposition Hatch Length,” Materials Performance and Characterization 4, no. 7 (2018): 717–735, https://doi.org/10.1520/MPC20170119

MISHUROVA ET AL., DOI: 10.1520/STP163120190148

14.

15.

16.

17.

18. 19.

I. C. Noyan and J. B. Cohen, Residual Stress. Measurement by Diffraction and Interpretation (New York: Springer-Verlag, 1987), https://doi.org/10.1007/978-1-4613-9570-6

20.

L. Spieß, G. Teichert, R. Schwarzer, H. Behnken, and C. Genzel, Moderne Ro ¨ntgenbeugung (Wiesbaden, Germany: ViewegþTeubner, 2009), https://doi.org/10.1007/978-38349-9434-9 M. Boin and R. C. Wimpory, “E3: Residual Stress Neutron Diffractometer at Ber II,” Journal of Large-Scale Research Facilities 2, no. A100 (2016), https://doi.org/10.17815/jlsrf-2-126

21. 22.

23.

24.

25.

26.

27.

28. .

M. Ghasri-Khouzani, H. Peng, R. Rogge, R. Attardo, P. Ostiguy, J. Neidig, R. Billo, D. Hoelzle, and M. R. Shankar, “Experimental Measurement of Residual Stress and Distortion in Additively Manufactured Stainless Steel Components with Various Dimensions,” Materials Science and Engineering: A 707 (2017): 689–700, https://doi.org/10.1016/j.msea.2017.09.108 D. W. Brown, J. D. Bernardin, J. S. Carpenter, B. Clausen, D. Spernjak, and J. M. Thompson, “Neutron Diffraction Measurements of Residual Stress in Additively Manufactured Stainless Steel,” Materials Science and Engineering: A 678 (2016): 291–298, https:// doi.org/10.1016/j.msea.2016.09.086 F. Bayerlein, F. Bodensteiner, C. Zeller, M. Hofmann, and M. F. Zaeh, “Transient Development of Residual Stresses in Laser Beam Melting—A Neutron Diffraction Study,” Additive Manufacturing 24 (2018): 587–594, https://doi.org/https://doi.org/10.1016/ j.addma.2018.10.024 L. Kolbus, E. Payzant, P. Cornwell, T. Watkins, S. Babu, R. Dehoff, M. Lorenz, O. Ovchinnikova, and C. Duty, Comparison of Residual Stresses in Inconel 718 Simple Parts Made by Electron Beam Melting and Direct Laser Metal Sintering 46 (2015): 1419–1432, https:// doi.org/10.1007/s11661-014-2722-2 M. T. Hutchings, P. J. Withers, T. M. Holden, and T. Lorentzen, Introduction to the Characterization of Residual Stress by Neutron Diffraction (Boca Raton, FL: CRC Press, 2005).

C. Genzel, I. A. Denks, J. Gibmeier, M. Klaus, and G. Wagener, “The Materials Science Synchrotron Beamline EDDI for Energy-Dispersive Diffraction Analysis,” Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 578, no. 1 (2007): 23–33, https://doi.org/10.1016/ j.nima.2007.05.209 E. Macherauch and P. Mu ¨ller, “Das Sin2w—Verfahren Der Ro ¨ntgenographischen Span2 nungsmessung [The Sin w Method of X-ray Voltage Measurement],” Journal of Applied Mathematics and Physics 13 (1961): 305–312. W. Voigt, “Uber Die Beziehung Zwischen Den Beiden Elastizitatskonstanten Isotroper Korper [About the Relationship between the Two Elasticity Constants of Isotropic Bodies],” Wiedemanns Annalen 38 (1889): 573–587. A. Reuss, “Berechnung Der Fliessgrenze Von Mischkristallen Auf Grund Der Plastizitatsbedingung Fur Einkristalle [Calculation of the Yield Point of Mixed Crystals Based on the Plasticity Condition for Single Crystals],” Journal of Applied Mathematics and Mechanics 9 (1929): 49–58. E. Kro ¨ner, “Berechnung Der Elastischen Konstanten Des Vielkristalls Aus Den Konstanten Des Einkristalls [Calculation of the Elastic Constants of the Multi-Crystal from the Constants of the Single Crystal],” Zeitschrift fu ¨r Physik 151, no. 4 (1958): 504–518, https://doi.org/10.1007/BF01337948 U. Welzel and E. J. Mittemeijer, “Diffraction Stress Analysis of Macroscopically Elastically Anisotropic Specimens: On the Concepts of Diffraction Elastic Constants and Stress Factors,” Journal of Applied Physics 93, no. 11 (2003): 9001–9011, https://doi.org/ 10.1063/1.1569662 T. Gna¨upel-Herold, A. A. Creuziger, and M. Iadicola, “A Model for Calculating Diffraction Elastic Constants,” Journal of Applied Crystallography 45, no. 2 (2012): 197–206, https:// doi.org/10.1107/s0021889812002221

137

138

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41. 42.

.

U. Welzel, J. Ligot, P. Lamparter, A. C. Vermeulen, and E. J. Mittemeijer, “Stress Analysis of Polycrystalline Thin Films and Surface Regions by X-Ray Diffraction,” Journal of Applied Crystallography 38, no. 1 (2005): 1–29, https://doi.org/10.1107/s0021889804029516 H. Do ¨lle, “The Influence of Multiaxial Stress States, Stress Gradients and Elastic Anisotropy on the Evalution of (Residual) Stress by X-Rays,” Journal of Applied Crystallography 12 (1979): 489–501. A. Heldmann, M. Hoelzel, M. Hofmann, W. Gan, W. W. Schmahl, E. Griesshaber, T. Hansen, N. Schell, and W. Petry, “Diffraction-Based Determination of Single-Crystal Elastic Constants of Polycrystalline Titanium Alloys,” Journal of Applied Crystallography 52, pt. 5 (2019): 1144–1156, https://doi.org/10.1107/S1600576719010720 T. Mishurova, K. Artzt, J. Haubrich, S. Evsevleev, A. Evans, M. Meixner, I. Serrano Munoz, I. Sevostianov, G. Requena, and G. Bruno, “Connecting Diffraction-Based Strain with Macroscopic Stresses in Laser Powder Bed Fused Ti-6al-4v,” Metallurgical and Materials Transactions A 51 (2020): 3194–3204, https://doi.org/10.1007/s11661-020-05711-6 T. Mishurova, G. Bruno, S. Evsevleev, and I. Sevostianov, “Determination of Macroscopic Stress from Diffraction Experiments: A Critical Discussion,” Journal of Applied Physics 128 (2020): 025103, https://doi.org/10.1063/5.0009101 R. Stegemann, S. Cabeza, V. Lyamkin, G. Bruno, A. Pittner, R. Wimpory, M. Boin, and M. Kreutzbruck, “Residual Stress Characterization of Steel TIG Welds by Neutron Diffraction and by Residual Magnetic Stray Field Mappings,” Journal of Magnetism and Magnetic Materials 426 (2017): 580–587. L. Edwards, G. Bruno, M. Dutta, P. J. Bouchard, K. L. Abbott, and R. L. Peng, “Validation of Residual Stress Predictions for a 19 MM Thick, J-Preparation Stainless Steel Pipe Girth Weld Using Neutron Diffraction” in Proceedings of the Sixth International Conference on Residual Stress (Oxford, UK: Oxford University Institute of Materials, 2000), 1523–1526. S. Ganguly, L. Edwards, and M. E. Fitzpatrick, “Problems in Using a Comb Sample as a Stress-Free Reference for the Determination of Welding Residual Stress by Diffraction,” Materials Science and Engineering: A 528 (2011): 1226–1232. Z. Wang, A. D. Stoica, D. Ma, and A. M. Beese, “Stress Relaxation in a Nickel-Base Superalloy at Elevated Temperatures with in Situ Neutron Diffraction Characterization: Application to Additive Manufacturing,” Materials Science and Engineering: A 714 (2018): 75–83. G. Albertini, G. Bruno, B. D. Dunn, F. Fiori, W. Reimers, and J. S. Wright, “Comparative Neutron and X-Ray Residual Stress Measurements on Al-2219 Welded Plate,” Materials Science and Engineering: A 224, no. 1 (1997): 157–165, https://doi.org/10.1016/S09215093(96)10546-3 J. Altenkirch, M. J. Peel, A. Steuwer, and P. J. Withers, “Comparison of Methods to Determine Variations in Unstrained Unit Cell Parameter across Welds,” Journal of Strain Analysis for Engineering Design 46 (2011): 651–662. G. Bruno, “Relaxation of Residual Stress in As-Welded and Aged AISI 347 Pipe by Means of Time-of-Flight Neutron Diffraction,” International Journal of Materials Research 93 (2002): 33–41. T. Thiede, “A Multiscale Analysis of Additively Manufactured Lattice Structures” (PhD diss., Potsdam University, Germany, 2019), https://doi.org/10.25932/publishup-47041 J. A. James, J. R. Santisteban, L. Edwards, and M. R. Daymond, “A Virtual Laboratory for Neutron and Synchrotron Strain Scanning.” Physica B: Condensed Matter 350, Suppl. 1–3 (2004): E743–E746.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190109

Anton du Plessis,1 Jess M. Waller,2 Stephan G. le Roux,1 Ina Yadroitsava,3 Igor Yadroitsev,3 Johan Els,3 and Jacobus Prinsloo4

X-Ray Computed Tomography Inspection in Metal Additive Manufacturing: The Role of Witness Specimens Citation A. du Plessis, J. M. Waller, S. G. le Roux, I. Yadroitsava, I. Yadroitsev, J. Els, and J. Prinsloo, “X-Ray Computed Tomography Inspection in Metal Additive Manufacturing: The Role of Witness Specimens,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 139–156. http://doi.org/10.1520/STP1631201901095

ABSTRACT

This work highlights the capabilities for high-resolution X-ray computed tomography (CT) inspection of witness specimens, built alongside a complex part, in metal additive manufacturing. Such witness specimens, which can be standardized in their dimensions (fixed diameter of 15 mm, with cylindrical shape built in a vertical orientation), allow X-ray CT inspections with fixed and reproducible workflows. The detection of improper process parameters of the additive manufacturing system is possible, as is demonstrated in this paper. It is also demonstrated how the presence of inclusions/contamination in the powder feedstock can be detected in the witness specimen. A series of Ti6Al4V witness Manuscript received September 17, 2019; accepted for publication November 25, 2019. 1 Research Group 3D Innovation, Stellenbosch University, Stellenbosch 7602, South Africa A. D. http:// http://orcid.org/0000-0002-5617-8137 orcid.org/0000-0002-4370-8661, S. G. L. 2 NASA-JSC White Sands Test Facility, 12600 NASA Rd., Las Cruces, NM, 88012 USA http://orcid.org/ 0000-0002-7847-8376 3 Dept. of Mechanical Engineering, Central University of Technology, 20 President Brand St., Bloemfontein 9301, South Africa I. Y. http://orcid.org/0000-0003-3132-5724, I. Y. http://orcid.org/0000-0002http://orcid.org/0000-0002-4837-8892 7556-8675, J. E. 4 ADC Aeroswift, CSIR Main Campus, Meiring Naude Rd., Brummeria, Pretoria 0081, South Africa 5 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

139

140

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

specimens with varying porosity distributions are presented, which were part of a previous study of builds of the same set of parts on different laser powder bed fusion systems. This demonstrates how various process parameter errors are highlighted and proven to be detectable in witness specimens using standardized CT procedures. More importantly, it also allows the potential to detect layered flaws, which can occur horizontally in the build plane. Such layered flaws may originate from reduced laser power, improper powder spreading, or due to complete shutdown and restart of a build. A complex bracket and witness specimen cylinder was built and a layered flaw was artificially induced by shutting down the system and restarting it. The positive detection of the flaw by CT in the witness rod is demonstrated. This witness rod was recently part of a round-robin test, and the layered flaw was successfully identified by all ten participants in the round-robin test. The witness rod and complex part were subsequently sectioned and optical microscopy reported here. This approach is especially useful for inspection of larger parts, which cannot be inspected using X-ray CT at the highest possible resolution due to part size and associated CT scanning time limits. Keywords X-ray tomography, microCT, nondestructive testing, witness specimen, metal additive manufacturing, laser powder bed fusion, stop-start flaw, porosity

Introduction Additive manufacturing (AM) is an emerging technique used to manufacture custom and complex parts for a variety of commercial applications.1–3 One major industrial interest is the production of metal parts, which is possible for a variety of alloys with excellent mechanical properties. One popular alloy is Ti6Al4V, which is used in biomedical and aerospace applications.4,5 Laser-powder bed fusion (L-PBF) is the most widely adopted metal AM technique, which allows the manufacturing of relatively large parts with intricate, complex designs by melting layer by layer in a powder bed, using a laser beam. For parts built using Ti6Al4V alloy by L-PBF, the mechanical performance can be superior to conventionally manufactured cast and wrought parts.6 However, despite the huge potential of AM, various manufacturing imperfections can occur that lead to compromised mechanical properties. Of the many types of AM imperfections possible in L-PBF parts, the most technologically important is the presence of porosity. Different forms of porosity can originate from improper process parameters,7,8 changes in the powder morphology (e.g., due to changing from virgin to used powder9), the separation of the part from the support structures during processing, and redistribution of loose powder in the form of funnels,10 as well as other causes that cannot always be predicted. Stop-start flaws are of particular interest in this work. These flaws can be created when the system stops and restarts, for example, due to power failure. The formation of these flaws in the build plane (horizontal in plane of powder bed) is due .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

to the shrinkage during cooling of the solidified part below the powder level during the “off-time” of the laser, creating a thicker powder layer than previous layers that is then not entirely melted on the next layer when the laser restarts. There is also a thermal mismatch, which could contribute to the observed porosity formation. A similar effect can occur if the laser power unexpectedly drops, creating one or more layers that are imperfectly melted; in this case, imperfect melting occurs over a large area, creating a similar flaw type. These horizontal flaws are particularly important as they can potentially extend across the entire part. Even when the extent is not large, the layered flat shape makes this kind of flaw a strong stress concentrator at its (side) edges when subjected to loading conditions. The grain evolution during solidification depends on heat flow, so stop-start flaws and other types of porosity can influence the microstructural grain growth in the vicinity of the flaw. The interaction of the pore morphology and microstructural features results in different stress distributions during loading, thus leading to unexpected damage behavior with different types of pore shapes.11 One of the best-suited methods to analyze AM parts for porosity or other flaw types and to optimize AM processes for porosity minimization is X-ray microcomputed tomography (microCT). A recent comprehensive review of the capabilities of present-day microCT for the analysis of additively manufactured parts highlights the importance of this type of nondestructive testing for process optimization and final product inspection.12 The use of microCT is not new in the field of materials science in general,13 and in additive manufacturing in particular.14–18 However, its wider acceptance and adoption has been limited in the AM community, mainly due to the high costs and complexity of analysis, which varies for each part. Although the capabilities of microCT are now starting to be appreciated more widely in the AM community, there is a need for standardization of microCT inspections. This is particularly true for measurement of AM part porosity and dimensional metrology of AM parts, as mentioned in Thompson, Maskery, and Leach,14 in order to improve the interpretation and ultimately the proper usage of the technique, as discussed in Seifi et al.19 To this end, we have developed a number of simplified and standardized methods for characterization of porosity, density, and surface roughness of small coupons of 1 cm3 cubes20–22 and for characterization of powder feedstock.23 These methods include prescribed scanning parameters and subsequent image analysis steps, in order to enhance reproducibility of these analyses across different microCT systems and users. Ultimately, the hope is that these methods will be adopted by industry and formally promulgated in voluntary consensus standards published by standards development organizations such as ASTM and the International Organization for Standardization (ISO).24 These methods can be used to optimize processing conditions prior to building critical parts. In this paper, we demonstrate a similar proposed method of analyzing cylindrical witness specimens and highlight the potential for standardization—fixed cylinder sizes allow recipes for CT scanning and image analysis, improving the reliability .

141

142

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

of flaw detection. We demonstrate how the process-specific porosity from different pore formation mechanisms is present in both the witness specimen and the complex part built alongside it for a series of different sets of samples. This confirms the ability to detect these types of flaws in witness specimens with the proposed scan and image analysis steps. The idea is that the witness specimen analysis will always take place with the same resolution and other scan settings, despite having a potentially larger complex part. Additionally, a witness specimen with an artificially induced stop-start flaw is analyzed here in detail, including subsequent physical cross sectioning and imaging by optical microscopy. In this example, the machine was stopped and restarted 12 h later, to artificially induce a stop-start flaw. This type of layer defect has been previously detected by microCT scans of a complex part, as reported in du Plessis et al.25 The concept of a witness specimen is not new, and their characterization by microCT was reported previously in Seifi et al.26 Witness specimens are now specified for all Class A and B metal parts fabricated using PBF and directed energy deposition (DED).27 These parts are used in critical and semicritical applications, whose failure would cause significant danger to personnel, loss of control, loss of a system, loss of a major component, an operating penalty, or loss of intended function. The aim of this present work is to demonstrate the suggested fixed scan parameters and a step-by-step workflow to improve the reproducibility of microCT inspections of such witness specimens. The ultimate aim of this is to allow easier usage of the microCT technique for routine quality inspections, thus improving the quality and reliability of AM parts.

Methods For the series of witness specimens (rods and cubes) and complex parts (brackets) having different process porosity types, the data were taken from previously reported round-robin testing encompassing a variety of L-PBF systems and different pore distributions.28 These sets of samples were produced on a variety of different L-PBF systems with the optimal process parameters of each system. All samples were nearly fully dense at greater than 99.87% density, but the (unexpected) porosity distributions were different and are further described in du Plessis and le Roux.28 More recent work on 5-mm cubes of Ti6Al4V studied the effects of varying process parameters on one system, creating a variety of porosity distributions artificially.8 In the present work, the fabrication of a witness rod and corresponding bracket with a stop-start flaw was accomplished with an EOS-M280 L-PBF system located at the Centre for Rapid Prototyping and Manufacturing (CRPM) at the Central University of Technology, Free State, South Africa. The powder used consisted of gas-atomized Ti6Al4V extra-low interstitials from TLS Technic with a mean spherical particle size of 45 lm. Standard process parameters for Ti6Al4V were used as recommended by the L-PBF system manufacturer for a layer thickness of 30 lm. Argon was used as a protective atmosphere with oxygen content controlled to stay below 0.12%. .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

The parts were all scanned in a microCT system at the Stellenbosch CT facility,29 similar to the complex part scans reported in du Plessis et al.30 The parts include a complex part (a bracket), its witness specimen, and a 1-cm3 cubic coupon produced during the same build. The witness specimen is a 15-mm-diameter cylinder built vertically up to the total height of the complex part height (in this case, roughly 40 mm high). The microCT scans of the witness specimen can be done at a resolution of up to 10 lm with typical microCT systems, but this requires reasonably long scan times and does not allow for mounting the sample at an angle. The selected voxel size is 25 lm, which allows a larger field of view and faster scan times, with sufficient quality and resolution to allow detection of important porosity distributions as shown in this work. In the case of this work, the X-ray tomography parameters were: 200 kV, 100 lA with a 0.5-mm beam filter, 250 ms acquisition time per image, and no averaging of images to allow fast scan time of 20 min per sample. Because the bracket was designed to be used in a load-bearing application, topology optimization was performed to ensure optimal load-bearing capacity relative to weight; this is reported elsewhere.31 The bracket was scanned at 46 lm initially and then close-up sections were scanned at 23 lm, with similar X-ray settings as previously noted.

Results and Discussion The results are split into two sections. The first section demonstrates how the porosity distributions in the cubes, witness rods, and brackets correlate with one another. This was partly reported in du Plessis, le Roux, and Guelpa,29 where the focus was on differences in cubes and the witness specimens were not analyzed yet. The second section focuses on the detection of an artificially created stop-start flaw present in a single witness rod. PROCESS POROSITY DISTRIBUTIONS

Several types of process-induced porosity, each with a unique mechanism responsible for its formation, are clearly distinguishable in the microCT scans of the samples made by different commercial L-PBF systems, as described in du Plessis and le Roux.28 In that previous work, contour pores, lack of fusion pores, and keyhole pores (subsurface at the top surface only) were detected in the 1-cm3 cubes, for example. These were unexpected at the time, which highlights the potential for microCT to add value in detection of process errors and for process optimization or refinement. The presence of these same pore types in the associated complex brackets made by the same L-PBF system was also confirmed by microCT scans, though with less clarity in places due to poorer resolution of scans of the brackets. In this paper, microCT scans of witness rods confirm the presence of these same pore types and pore distributions and show that each distribution type can be positively detected in this type of sample. Figures 1, 2, and 3 show the examples of contour pores, lack of .

143

144

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 Process-induced porosity detected in both witness specimen and complex part, demonstrated here for contour porosity at end of scan tracks. Views shown of witness specimen (A) surface view, (B) transparent angled view, (C) transparent side view, and (D) transparent top view. View shown of bracket (E) surface view, (F) transparent angled view, (G) transparent side view, and (H) transparent top view.

fusion pores, and keyhole pores (at the top surface only), respectively. Different three-dimensional (3D) views are used to illustrate the presence of the same signature porosity in each set of samples. The importance here is that the witness specimens can be used to check when slight power drops might create lack of fusion pores in a series of layers. Figure 1 shows contour porosity, which is just below the surface at all vertical walls and is clearly seen in the top view of the witness specimen in figure 1D. Figure 2 shows lack of fusion porosity and, in the witness specimen as seen in figure 2C in a side view, the lack of fusion porosity is not uniformly spread across the build height. More porosity is present near the bottom of the witness specimen in this case. In the associated bracket, this difference is not clear (see fig. 2F for .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

FIG. 2 Process-induced porosity detected in both witness specimen and complex part, demonstrated here for lack of fusion porosity. Views shown of witness specimen (A) surface view, (B) transparent angled view, and (C) transparent side view. View shown of bracket (D) surface view, (E) transparent angled view, (F) transparent side view, and (G) transparent top view.

side view). This is presumably due to additional pore formation mechanisms at work in the complex part. The build strategy, the presence or absence of supports, powder delivery, and part orientation on the plate have to be analyzed as well as process parameters. Figure 3 shows an example of keyhole mode pores at the top surface only; this is best seen in the witness specimen in a side view, such as in figure 3C. It is also seen in figure 3F that the same type of pores are present in the bracket at horizontal topfacing surfaces. A different set of samples with less porosity is shown in figure 4, but the witness specimen clearly shows layered porosity—a particularly worrying form of porosity. When inspecting the associated bracket closely at a height corresponding to the .

145

146

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 3 Process-induced porosity detected in both witness specimen and complex part, demonstrated here for keyhole mode porosity under top surface. Views shown of witness specimen (A) surface view, (B) transparent angled view, and (C) transparent side view. Views shown of bracket (D) surface view, (E) transparent angled view, (F) transparent side view, and (G) transparent top view.

appearance of the lower layer of porosity seen in the witness rod, this layered porosity is also detected in the in-plane CT slice data (fig. 4B) and was missed in out-ofplane CT slice data (fig. 4C, bottom). This porosity distribution might occur due to imperfect powder spreading on this particular layer and confirms the utility of the witness specimen for detecting layered flaws of this type. In addition to pores, contamination can occur in L-PBF systems such as from previous builds with different powder.32 These inclusions can influence the melting process and act as stress concentrators in final parts under loading conditions. Such inclusions have a different density, size, and shape than the rest of the powder. Powder contamination also plays a potential role in the formation of porosity in LPBF parts because the inclusion particles have a different melting temperature. It .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

FIG. 4 Layered lack of fusion porosity, presumably due to imperfect powder spreading—shown in (A) the witness specimen in 3D and in the bracket; by using carefully aligned slice images, it is possible to image the pores (B) in the flaw plane and (C) out-of-plane in one arm of the bracket.

.

147

148

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Inclusions (powder contamination) detected in (A) the witness specimen and also in (B) the associated bracket—white dots seen in slice images are denser particles.

has been shown, for example, that during in situ alloying with controlled amounts of different powders, the ideal process parameters change.33 One set of parts in this series contained such inclusions as shown in figure 5. This figure shows the detection of high-density inclusions, which appear as white dots in both the witness specimen and bracket. Thus, the manufacturing of witness samples is useful for identifying negative features such as powder contamination, which is unacceptable in the manufacture of critical parts. In this case, it was confirmed that contamination from a previous build was likely. STOP-START FLAW

Besides inherent process porosity, some errors can occur that create localized porosity or flaws, which can extend across the entire build plane or large parts of it. Of these, one of the most important is the stop-start flaw, which occurs when there is a shutdown of the system and a restart later. This type of flaw is caused by shrinkage of the solidified part, which creates a larger layer height of powder in the next layer upon restart that does not fully melt, thus leading to a specialized form of lack of fusion. A similar effect of imperfect melting on a single layer can occur when the laser power drops temporarily or when powder spreading is uneven due to part warping or recoater damage, for example. Such effects may potentially be spread across the entire build, which means they may be detectable by using witness specimens. This is shown for an artificially induced stop-start flaw in a witness rod in figure 6. Detecting a start-stop flaw or other type of layer defect in a production part at the same height as the flaw that was observed in the corresponding witness specimen (fig. 6) can be problematic for X-ray tomography. In this case, despite a stopstart interval of 12 h (the machine was stopped during the build and restarted the next day), no layer defects were found inside the complex part (bracket) at the 46-lm resolution of the bracket scan. A higher resolution “zoom scan” of the .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

FIG. 6 Stop-start flaw detected in witness specimen—different views of (A) 3D surface, (B) transparent angled view, (C) transparent side view, (D) slice top view in plane of flaw, (E) with the associated slice plane indicated, (F) slice side view, and (G) associated slice plane indicated.

bracket at 23-lm voxel size of the potentially problematic area corresponding to the known build height and build orientation where the build was stopped and restarted is shown in figure 7. However, no layer defects were found in this area or elsewhere in the build plane corresponding to the location of the stop-start flaw. The bracket was also sectioned, and no flaws were found under nominal magnification with an optical microscope. Stop-start flaws may not be present in all locations in the build plane, as also seen in the witness specimen (the defect does not cover the entire area of the cylinder); in this case, they did not extend into the bracket. .

149

150

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 7 High-resolution “zoom scan” of a potentially problematic area corresponding to stop-start layer defect in a witness rod at a known Z-height in the build. No layer defects were found in this case (23-μm resolution). Shown here are (A) the cross-sectional (side) view and (B) top view in plane of expected location of layered defects (13 mm from top, as measured from witness specimen).

The detection of planar 2D flaws in AM parts perpendicular to the build (Z) direction is inherently challenging for any nondestructive evaluation technique, including microCT. As pointed out in the NASA standard,34 AM processes tend to prohibit volumetric defects with significant height in the Z direction. The major concern, therefore, is for planar defects, such as aligned or chained porosity or even laminar cracks, or the stop-start defects, as examined in this paper, that form along the build plane. The implications of this are: (1) planar defects are well suited for growth; (2) planar defects generally have low contained volume; (3) the orientation of defects of concern must be known before inspection, especially when detection sensitivity depends on the defect orientation relative to the inspection direction; and (4) the Z-height of planar defects can be demanding on incremental step inspection methods such as CT. Therefore, it is important to manually assess slice images in microCT data from at least two orthogonal orientations, and it is critical that the part is scanned at an angle relative to its original build direction. Regardless, when a larger production part cannot be inspected at sufficiently high resolution due to size limits, and layer or planar defects are positively identified in a smaller matching witness coupon, it would be safer to assume the presence of undetected layer defects in the production part and reject the part. Microscopic analysis of physical cross sections of the cylinder shows that the stop-start layered defect is comprised of a chain of pores with irregular shapes. Large, .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

irregular pores were found with vertical sizes ranging from 120 to 180 lm with narrow (up to 20 lm) shrinkage cavities. Usually, big, irregular pores correlate with low energy input, when laser power density was not enough to fully melt the powder layer and previously melted material (also known as lack of fusion). Taking into account the 30-lm powder layer thickness used in this experiment, and the optimal process parameters needed to produce a fully dense part, the reason large stop-start defects occurred in the witness rod in this case can be attributed to shrinkage of the whole system during cooling, including the powder delivering system, baseplate, and as-built part that had been previously melted in the first cycle prior to machine stoppage. The redistribution of residual stresses detaching from the substrate or the warping of parts during cooling for several hours can lead to uneven layer thickness when the next powder layer is delivered. So, interaction of all these factors can lead to random porosity in L-PBF parts after a stop-start cycle, as was found in this experiment and as revealed by the presence of horizontally aligned pores in the witness rod (fig. 6 and fig. 8). The same defects were expected in the bracket but were absent (fig. 7 and fig. 8B at the arrows). In this stop-start L-PBF process, defects occurred along several layers, taking into account their size (fig. 8A); but despite this, the extent was not across the entire build plane and did not extend into the bracket. As previously stated, the interaction of the microstructure and different types of porosity can be critical for the performance properties of the L-PBF part.35 The interruption of the microstructural grain growth (in Ti6Al4V, prior beta grains typically grow vertically along the build direction) makes for possible new locations of crack initiation and growth along the inner (top and bottom) edges of the flaw along the vertical grain boundaries ending at the flaw. Sharp edges of pores interrupted by prior beta grains and notches coinciding with the direction of acicular martensitic a’ phase (fig. 8A) can influence not only the crack initiation under loading but can also deteriorate the fatigue performance of as-built and stress-relieved L-PBF components. Textured microstructure related to anisotropic structural properties usually remains even after heat treatment, for example, in Ti6Al4V.5,36 The advantages of the microCT inspection of witness specimens using standardized workflows has been clearly demonstrated in this work. The inherent disadvantage is that some layered flaws or irregular porosity distributions may occur in a complex part but not in the witness specimen. This means that the microCT inspection should be complemented by other inline process monitoring and postprocess quality control tools. The shape and size of witness specimens, their position near the complex L-PBF component, and the extent of layered flaws across the build plane justify a separate study in the future. For denser metals, the method will need some modification compared to that presented here. The work in this paper was presented for Ti6Al4V and will be suitable for less radiodense materials (e.g., aluminum alloys, plastics, and so on). For denser materials, a narrower/finer witness specimen may be required to allow penetration of typical laboratory microCT X-ray beams. Also, it is imperative that the .

151

152

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 8 Cross section and microstructure of cylinder with (A) layered defects and (B) part of the bracket where layered flaw was expected but was not found at location indicated by white arrows.

orientation of defects of concern is known before inspection to maximize CT detectability of known or suspected planar flaws. This means the build/print direction must be known and part angled relative to this to ensure proper detection of the layered flaws in the build plane. For larger L-PBF parts, the Z-height of planar defects, such as the stop-start flaw examined here, can be demanding on incremental step inspection methods such as CT. Nevertheless, this method should be useful for routine analysis, with the only modification being the resolution of the scan of the complex part, which in turn depends on part size. The standardization of witness specimen geometry (e.g., uniform coupon diameter) allows a fixed methodology for all identically shaped additively manufactured witness specimens, regardless of the AM platform used, machine-to-machine variation, or variation within a single AM machine. The only limitation is that larger parts will require longer witness specimens, which will require longer scan times to identify flaws with the requisite .

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

resolution (less than or equal to 20-lm voxel size in this paper). The advantage is that only one witness specimen may be adequate for an entire batch of parts to ensure the absence of unwanted layer, cross-layer, or other volumetric defect types (e.g., inclusions, trapped powder, cracks, and so on).

Conclusions The advantages of using witness specimens and microCT scanning thereof according to fixed workflows was investigated. It was shown how this approach can accurately identify process porosity signatures, which can act as “witness” to in-process changes in the parameters over a single layer or multiple layers. The presence of contamination of metal powders was demonstrated in one case, and this was accurately detected in both a witness specimen and its corresponding complex part. Lack of fusion porosity detected in a witness specimen was found to occur across multiple build layers in one case, and analogous lack of fusion porosity was confirmed in the complex bracket associated with this witness specimen. Finally, an artificially induced stop-start flaw was investigated and its detection in witness specimen confirmed and analyzed in detail using microCT and optical microscopy of cross sections. This stop-start flaw was found to extend widely but not completely over the entire part and, in this case, did not extend to the complex part built alongside it (also investigated by microCT and optical microscopy). This points to the possibility that unexpected flaws, including layered flaws, may occur in complex parts despite passing a witness specimen microCT test. Similarly, there may be situations where localized power fluctuations occur or where build quality varies with location in the powder bed. This means that additional complementary tools are needed for 100% quality control, and understanding the limits of the microCT technique to detect planar defects is therefore important. This work is expected to contribute to the wider understanding and better utility of microCT as an inspection tool, especially with standardized workflows using witness specimens. ACKNOWLEDGMENTS

The Collaborative Program for Additive Manufacturing (CPAM), funded by the South African Department of Science and Technology, is acknowledged for financial support. This work is also based on the research supported by the South African Research Chairs Initiative of the Department of Science and Technology and National Research Foundation of South Africa (Grant No. 97994). Samples were built in CRPM at Central University of Technology, Free State.

References 1.

.

M. Schmidt, M. Merklein, D. Bourell, D. Dimitrov, T. Hausotte, K. Wegener, L. Overmeyer, F. Vollertsen, and G. N. Levy, “Laser-Based Additive Manufacturing in Industry and Academia,” CIRP Annals 66, no. 2 (2017): 561–583, https://doi.org/10.1016/J.CIRP.2017.05.011

153

154

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13. 14.

15.

16.

17. .

T. DebRoy, T. Mukherjee, J. O. Milewski, J. W. Elmer, B. Ribic, J. J. Blecher, and W. Zhang, “Scientific, Technological and Economic Issues in Metal Printing and Their Solutions, Nature Materials 18 (2019): 1026–1032, https://doi.org/10.1038/s41563-019-0408-2 T. Wohlers, T. Caffrey, R. I. Campbell, O. Diegel, and J. Kowen, Wohlers Report 2018: 3D Printing and Additive Manufacturing State of the Industry (Fort Collins, CO: Wohlers Associates, Inc., 2018). T. Debroy, H. L. Wei, J. S. Zuback, T. Mukherjee, J. W. Elmer, J. O. Milewski, A. M. Beese, A. Wilson-Heid, A. De, and W. Zhang, “Additive Manufacturing of Metallic Components—Process, Structure and Properties,” Progress in Materials Science 92 (2018): 112–224, https://doi.org/10.1016/j.pmatsci.2017.10.001 I. Yadroitsev, P. Krakhmalev, I. Yadroitsava, and A. du Plessis, “Qualification of Ti6Al4V ELI Alloy Produced by Laser Powder Bed Fusion for Biomedical Applications,” JOM 70 (2018): 372–377, https://doi.org/10.1007/s11837-017-2655-5 M. Qian, W. Xu, M. Brandt, and H. P. Tang, “Additive Manufacturing and Postprocessing of Ti-6Al-4V for Superior Mechanical Properties,” MRS Bulletin 41 (2016): 775–784, https://doi.org/10.1557/mrs.2016.215 H. Gong, K. Rafi, H. Gu, T. Starr, and B. Stucker, “Analysis of Defect Generation in Ti-6Al4V Parts Made Using Powder Bed Fusion Additive Manufacturing Processes,” Additive Manufacturing 1–4 (2014): 87–98, https://doi.org/10.1016/j.addma.2014.08.002 A. du Plessis, “Effects of Process Parameters on Porosity in Laser Powder Bed Fusion Revealed by X-Ray Tomography,” Additive Manufacturing 30 (2019): 100871, https:// doi.org/10.1016/j.addma.2019.100871 L. Cordova, M. Campos, and T. Tinga, “Revealing the Effects of Powder Reuse for Selective Laser Melting by Powder Characterization,” JOM 71 (2019): 1062–1072 https:// doi.org/10.1007/s11837-018-3305-2 I. Yadroitsavaa and I. Yadroitsev, “Residual Stress in Metal Specimens Produced by Direct Metal Laser Sintering,” in Proceedings of the Solid Freeform Fabrication Symposium (Austin, TX: University of Texas, 2015), 614–625, http://sffsymposium.engr.utexas.edu/ sites/default/files/2015/2015-49-Yadroitsev.pdf M. Kabir and H. Richter, “Modeling of Processing-Induced Pore Morphology in an Additively-Manufactured Ti-6Al-4V Alloy,” Materials 10, no. 2 (2017): 145, https:// doi.org/10.3390/ma10020145 A. du Plessis, I. Yadroitsev, I. Yadroitsava, and S. G. le Roux, “X-Ray Microcomputed Tomography in Additive Manufacturing: A Review of the Current Technology and Applications,” 3D Printing and Additive Manufacturing 5, no. 3 (2018): 3dp.2018.0060, https://doi.org/10.1089/3dp.2018.0060 E. Maire and P. J. Withers, “Quantitative X-Ray Tomography,” International Materials Review 59 (2014): 1–43, https://doi.org/10.1179/1743280413Y.0000000023 A. Thompson, I. Maskery, and R. K. Leach, “X-Ray Computed Tomography for Additive Manufacturing: A Review,” Measurement Science and Technology 27, no. 7 (2016): 072001, https://doi.org/10.1088/0957-0233/27/7/072001 F. H. Kim, S. P. Moylan, E. J. Garboczi, and J. A. Slotwinski, “Investigation of Pore Structure in Cobalt Chrome Additively Manufactured Parts Using X-Ray Computed Tomography and Three-Dimensional Image Analysis,” Additive Manufacturing 17 (2017): 23–38, https://doi.org/10.1016/J.ADDMA.2017.06.011 A. du Plessis, S. G. le Roux, G. Booysen, and J. Els, “Directionality of Cavities and Porosity Formation in Powder-Bed Laser Additive Manufacturing of Metal Components Investigated Using X-Ray Tomography,” 3D Printing and Additive Manufacturing 3, no. 1 (2016): 48–55, https://doi.org/10.1089/3dp.2015.0034 H. Villarraga-Go ´mez, C. Lee, and S. T. Smith, “Dimensional Metrology with X-Ray CT: A Comparison with CMM Measurements on Internal Features and Compliant Structures,” Precision Engineering 51 (2018): 291–307, https://doi.org/10.1016/J.PRECISIONENG.2017.08.021

DU PLESSIS ET AL., DOI: 10.1520/STP163120190109

18.

19.

20.

21.

22.

23.

24.

25.

26.

27. 28.

29.

30.

31.

32.

.

H. Villarraga-Go ´mez, E. L. Herazo, and S. T. Smith, “Progression of X-Ray Computed Tomography from Medical Imaging to Current Status in Dimensional Metrology,” Precision Engineering 51 (2019): 291–307, https://doi.org/10.1016/J.PRECISIONENG.2019.06.007 M. Seifi, A. Salem, J. Beuth, O. Harrysson, and J. J. Lewandowski, “Overview of Materials Qualification Needs for Metal Additive Manufacturing,” JOM 68 (2016): 747–764, https://doi.org/10.1007/s11837-015-1810-0 A. du Plessis, P. Sperling, A. Beerlink, L. Tshabalala, S. Hoosain, N. Mathe, and G. Stephan, “MethodsX Standard Method for MicroCT-Based Additive Manufacturing Quality Control 1: Porosity Analysis,” MethodsX 5 (2018): 1102–1110, https://doi.org/10.1016/j.mex.2018.09.005 A. du Plessis, P. Sperling, A. Beerlink, L. Tshabalala, S. Hoosain, N. Mathe, and G. Stephan, “MethodsX Standard Method for MicroCT-Based Additive Manufacturing Quality Control 2: Density Measurement,” MethodsX 5 (2018): 1117–1123, https://doi.org/10.1016/ j.mex.2018.09.006 A. du Plessis, P. Sperling, A. Beerlink, O. Kruger, L. Tshabalala, S. Hoosain, and S.G. le Roux, “Standard Method for MicroCT-Based Additive Manufacturing Quality Control 3: Surface Roughness,” MethodsX 5 (2018): 1111–1116, https://doi.org/10.1016/j.mex.2018.09.006 A. du Plessis, P. Sperling, A. Beerlink, W. du Preez, and S. G. le Roux, “Standard Method for MicroCT-Based Additive Manufacturing Quality Control 4: Metal Powder Analysis,” MethodsX 5 (2018): 1336–1345, https://doi.org/10.1016/j.mex.2018.09.006 Additive Manufacturing—General Principles—Non-Destructive Testing of Additive Manufactured Products, ISO/ASTM DTR 52905 (ASTM F42) (Geneva, Switzerland: ISO Technical Committee 261 2019), http://web.archive.org/web/20190115092649/https:// www.iso.org/standard/71988.html A. du Plessis, S. G. Le Roux, J. Els, G. Booysen, and D. C. Blaine, “Application of MicroCT to the Non-Destructive Testing of an Additive Manufactured Titanium Component,” Case Studies in Nondestructive Testing and Evaluation 4 (2015): 1–7, https://doi.org/ 10.1016/j.csndt.2015.09.001 M. Seifi, M. Gorelik, J. Waller, N. Hrabe, N. Shamsaei, S. Daniewicz, and J. J. Lewandowski, “Progress towards Metal Additive Manufacturing Standardization to Support Qualification and Certification,” JOM 69 (2017): 439–455, https://doi.org/10.1007/s11837-0172265-2 Specification for Fabrication of Metal Components Using Additive Manufacturing, AWS D20.1/D20.1M:2019 (Miami, FL: American Welding Society, 2019). A. du Plessis and S. G. le Roux, “Standardized X-Ray Tomography Testing of Additively Manufactured Parts: A Round Robin Test,” Additive Manufacturing 24 (2018): 125–136, https://doi.org/10.1016/j.addma.2018.09.014 A. du Plessis, S. G. le Roux, and A. Guelpa, “The CT Scanner Facility at Stellenbosch University: An Open Access X-Ray Computed Tomography Laboratory,” Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms 384 (2016): 42–49, https://doi.org/10.1016/J.NIMB.2016.08.005 A. du Plessis, S. G. Le Roux, G. Booysen, and J. Els, “Quality Control of a Laser Additive Manufactured Medical Implant by X-Ray Tomography,” 3D Printing and Additive Manufacturing 3 (2016): 175–182, https://doi.org/10.1089/3dp.2016.0012 C. H. Hands, A. du Plessis, N. Minnaar, B. A. Blakey-Milner, and E. Burger, “Can Additive Manufacturing Help Win the Race?” Preprints (2018), https://doi.org/10.20944/ PREPRINTS201811.0040.V1 A. D. Branda ˜o, R. Gerard, J. Gumpinger, S. Beretta, A. Makaya, L. Pambaguian, and T. Ghidini, “Challenges in Additive Manufacturing of Space Parts: Powder Feedstock CrossContamination and Its Impact on End Products,” Materials 10, no. 5 (2017): 522, https:// doi.org/10.3390/ma10050522

155

156

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

33.

34.

35. 36.

.

P. Krakhmalev and I. Yadroitsev, “Microstructure and Properties of Intermetallic Composite Coatings Fabricated by Selective Laser Melting of Ti–SiC Powder Mixtures,” Intermetallics 46 (2014): 147–155, https://doi.org/10.1016/J.INTERMET.2013.11.012 Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals, MSFC-STD-3716 (Huntsville, AL: NASA Marshall Spaceflight Center, 2017). S. Liu and Y. C. Shin, “Additive Manufacturing of Ti6Al4V Alloy: A Review,” Materials & Design 164 (2019): 107552, https://doi.org/10.1016/J.MATDES.2018.107552 P. Barriobero-Vila, J. Gussone, A. Stark, N. Schell, J. Haubrich, and G. Requena, “Peritectic Titanium Alloys for 3D Printing,” Nature Communications 9 (2018): 3426, https://doi.org/10.1038/s41467-018-05819-9

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190152

Eric Lindgren1 and Bryce Jolley1

Perspective on Nondestructive Evaluation of Additive Manufactured Components Citation E. Lindgren and B. Jolley, “Perspective on Nondestructive Evaluation of Additive Manufactured Components,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 157–164. http://doi.org/10.1520/STP1631201901522

ABSTRACT

Nondestructive evaluation (NDE) has been used for materials and component manufacturing, typically for process quality control or at multiple steps in manufacturing processes. In addition, NDE is used in sustainment of aircraft components to detect defects with a validated capability quantified by probability of detection. U.S. Air Force (USAF) Structures Bulletin EZ-SB-19-01, “Durability and Damage Tolerance Certification for Additive Manufacturing of Aircraft Structural Metallic Parts,” emphasizes the use of NDE in manufacturing by defining five factors to be evaluated to specify a new material, process, joining method, and/or structural concepts. The factors include stability, producibility, characterization of mechanical or physical properties, predictability of structural performance, and supportability. Included is Section 3.4, “Nondestructive Inspection Development, Validation, Verification, and Implementation,” which provides instructions to determine the validated capability for all damage types, orientations, and locations for specific regions of interest. Factors that affect NDE capability include where in the manufacturing process the assessment occurs, the nature of the raw feedstock, and the diagnostic information required to ensure the quality of the produced part. Once manufactured, additional factors must be considered, including the nature/ orientation of defects, access, surface condition, geometric complexity, internal

Manuscript received November 29, 2019; accepted for publication May 4, 2020. 1 U.S. Air Force Research Laboratory, 2230 10th St., B655, Ste 1, Wright-Patterson AFB, OH 45324, USA 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. This work is not subject to copyright law. ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

.

157

158

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

microstructural characteristics, and potential residual stresses. Representative defects within additive manufactured (AM) parts/materials include micro-cracks, voids, delaminations, lack of fusion, porosity, and inclusions. Systematic approaches are needed that leverage existing methods for other manufacturing processes. Methods to assist in determining which NDE method(s) have the highest potential of success are discussed. However, the intent is not to provide an overview of all NDE methods and their applicability to assessing AM components. In addition, methods to simplify or accelerate the assessment of the probability of detection of the defect of interest are addressed. These items should provide insight into how to ensure AM parts can be qualified and sustained for aerospace applications. Keywords nondestructive evaluation, nondestructive testing, damage tolerance

Introduction Nondestructive inspection (NDI) is defined in the U.S. Air Force (USAF) Structures Bulletin EZ-SB-19-01, “Durability and Damage Tolerance Certification for Additive Manufacturing of Aircraft Structural Metallic Parts,” as “an inspection process or technique designed to reveal the damage at or beneath the external surface of a part or material without adversely affecting the material or part being inspected. NDI generally refers to inspections that are conducted using equipment that is not part of or permanently affixed to the part being inspected. Inspections that do involve such equipment are generally referred to as in-situ NDI or structural health monitoring.”1 The USAF incorporates NDI and nondestructive evaluation (NDE) for the life management of USAF aircraft structures and propulsion systems.2,3 NDE, in contrast, is the capability to characterize the damage/defect (e.g., size, location, and orientation). NDI focuses on detecting damage greater than a certain size with a known probability. As the accomplishment of NDI-based procedures requires certified technicians in USAF Depots, or maintenance and overhaul facilities, the management of NDI is centralized across each Depot in addition to the field-based inspection teams, the vast majority of which are USAF uniformed service members. The USAF NDI Executive Working Group was established to ensure common practice of NDI across all locations where inspections are accomplished. This includes procedures, training, verification/validation of new inspection capability, and vetting of new technology before it is implemented. In addition to life management of structures, nondestructive methods (for the remainder of this paper referred to as NDE/I) are extensively used in manufacturing as a quality assurance tool. For example, all aerospace composites that are intended for a structural application, such as a wing or fuselage, are 100% inspected using ultrasound to detect unacceptable levels of porosity or delaminations (or both) greater than a specified size. The criteria for exact size and distribution of rejectable indications are determined by the intended use of the component being inspected. .

LINDGREN AND JOLLEY, DOI: 10.1520/STP163120190152

Thus, there is a significant reliance on NDE/I in aerospace, and this is directly related to the intent for the use of additive manufactured (AM) components in aerospace. The desired potential for additive manufacturing in aerospace is to manufacture any component for any application at any time. However, for fracture critical components, this can become a tremendous challenge as a function of the certification process for the component. If the part is not critical to ensuring the safety of the aircraft, then there is greater flexibility. MIL STD 1530D provides the definition of the four general classes of components that are used on USAF aircraft.2 They are normal controls, durability critical, nontraceable fracture critical, and traceable fracture critical. This list is in order of increased requirements for qualification and certification, with the last one having the most stringent processes. The USAF has success in implementing AM parts that are managed by normal controls. Several use cases that have been highlighted in publications include plastic components that do not carry load but have realized cost savings for the USAF.4 Typically, these components have addressed parts that are no longer being manufactured or that function as support components to other integrated systems (or both), including such applications as knobs and handles. However, as AM parts are being considered for fracture critical applications, they must be evaluated in the general scheme of durability and damage tolerance (DADT), which is the method used to ensure the USAF meets its structural risk metrics.

Overview of Durability and Damage Tolerance DADT can be defined as the ability of a structure to resist failure due to the presence of flaws, cracks, or other damage for a specified period of usage. The details of the DADT approach are given in MIL STD 1530D. The following is a brief implementation introduction and is unique to the USAF. Civil aviation uses a similar, but not identical, approach, and the U.S. Navy uses a different method to ensure structural integrity, or safety, as their usage spectrum is quite different from that of the USAF. Figure 1 shows a typical fatigue crack growth curve for a metallic component assuming the nucleation of a fatigue crack has occurred. The USAF makes the assumption that a rogue flaw is present based on experiences of having this occur and having this result in fatal mishaps.5 The recurring inspection interval is determined by the validated capability of the NDE/I method being used. In this case, the inspection being used during manufacturing is assumed to be slightly better than that being used once the system is placed in service. As can be seen from the graph, the intent is to have two opportunities to inspect a component before the fatigue crack can grow to critical size. If no crack is detected, the crack growth curve is reset to the size of the capability of the inspection procedure. The method to determine this capability, called probability of detection (POD), will be introduced later in this paper. .

159

160

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 Validated NDE/I capability sets’ inspection intervals as a function of fatigue crack growth.

Critical

Crack Size

Crack Propagation Margin (aka Flaw Growth Interval, aka Residual Life): Must be 2X Inspection Interval

Safety Limit (SL) Minimum Failure

½ SL Initial (A1) Inspect 1

Inspect 2 Inspect 3 Cycles or Time

Inspect 4

In addition to the NDE/I capability, the interval is determined by knowing the predicted fracture growth in the component as a function of factors such as material type, applied stresses, and geometry. Without understanding the fracture behavior of this material, this approach would not be possible. Therefore, there is a significant effort in the additive manufacturing research and development community to understand the fracture behavior of AM components and the effect of defects on the fracture behavior to integrate that information into this analysis method.1 The integration of NDE/I capability, materials properties, and intended use to determine the risk of structural failure is shown in figure 2. These inputs can be tailored as a function of material system and known fracture behavior. As shown, the critical safety metric for the USAF is a single flight probability of failure, and this metric is tracked for all weapon systems. For material systems where the fracture mechanics are not well understood and cannot be readily predicted, such as polymer matrix composites, this approach cannot be used. As an alternative to this approach, the material can be qualified and certified for a limited time and, once this time limit is reached, the component needs to be replaced or the system should be retired.

Structures Bulletin EZ-SB-19-01 To facilitate the implementation of AM components on USAF aircraft, the USAF structures community has published Structures Bulletin (SB) EZ-SB-19-01, titled “Durability and Damage Tolerance Certification for Additive Manufacturing of Aircraft Structural Metallic Parts,” which is available to the general public.1 The reader is encouraged to obtain a copy of this SB and read it carefully to fully understand the processes used by the USAF structures community. Several highlights from this SB include the five factors for evaluation of new material capability to ensure safety .

LINDGREN AND JOLLEY, DOI: 10.1520/STP163120190152

FIG. 2 Typical factors used to calculate risk of failure for USAF structural components.

and discussion of damage types and the need to understand their attributes on life management. Factors that can influence processing of materials include stability, producibility, characterization of mechanical or physical properties, prediction of structural performance, and supportability. While each one of these characteristics will be influenced by the AM method being used, NDE/I plays an important role in both producibility and supportability to enable prediction of fracture behavior possible as a function of the criticality of the component. As an example, for a normal controls component, the need for predicting fracture behavior may not be needed. This type of component could be a part that does not carry load. In this simplified case, the inspection requirement could be a quality assurance (QA) assessment to ensure the component meets the geometric tolerances and can be readily used as a replacement for an existing part. However, if the component is considered fracture critical, then methods to ensure the quality of the part plus a method that enables support over its expected service life need to be identified. This USAF SB provides additional guidance in this situation and refers to damage as “any flaw, defect, crack, corrosion, disbond, delamination, discontinuity, or other type that degrades, or has the potential to degrade, the performance of the affected component.”1 All attributes of the damage need to be understood, including type, size, orientation, and location, as these factors can affect the ability to detect damage. As an analogy, consider that unique damage modes can emerge in welding and casting processes. The typical defects that occur in these processes have been studied and are understood, resulting in detailed standards or requirements (or both) for the inspection of components made by these processes to ensure they meet the performance requirements of the application. .

161

162

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Considerations for NDE/I The selection of the proper NDE/I technique is dependent on the type of damage present and the effect of this damage on the performance of the component for its intended application. For this reason, a number of “effects of defects” studies are underway6 to determine the attributes of damage that impact the performance of the component. Most of these studies are focused on fatigue behavior of test samples prepared by laser powder bed fusion methods. When this approach is considered, ASTM 52900, Standard Terminology for Additive Manufacturing—General Principles—Terminology,7 provides the following list of possible defects: localized porosity, residual stress, microcracks, delamination, inclusion, geometry tolerance, uncontrolled microstructure, balling, lack of fusion, keyholing, and undesired surface finishes. Each one of these defects can affect fatigue behavior, and their impact on the performance of a specific part needs to be understood to determine the proper NDE/I approach to detect these defects. In addition, the size and location of the defect to be detected needs to specified. Ideally this would include the defect orientation if this information is available. This information provides a metric to determine if the NDE/I technique has the capability to detect the defect(s) of interest. This is a parameter that needs to be specified for a fracture critical component by the structural engineer as a result of the effects of defects studies. A common request is to find the “smallest defect possible,” but this can lead to extensive and unnecessary investigations to refine a capability that may not be needed. By properly specifying the type and size of defect that has a negative impact on performance, the most effective NDE/I method can be determined. As a representative example, computed tomography (CT) is frequently used as an NDE/I technique for AM parts. However, if the defect in question has a very low aspect ratio (e.g., more like a delamination than a void), then CT capability could be compromised because this method is challenged to detect this type of defect geometry when compared to alternative NDE/I methods, such as ultrasound. Attributes of the defects of interest can have a significant impact on the selection of the NDE/I techniques that have the greatest potential to detect the defects of interest.

Validation of NDE/I Capability: Probability of Detection Once the defect type, location, and size to be detected is defined and the NDE/I technique to detect the defects is determined, the capability of the NDE/I technique needs to be validated. For engineering applications, validation is a rigorous assessment using many test samples with defects of differing sizes. The guidance for performing these assessments, called POD studies in USAF terminology, is given in

.

LINDGREN AND JOLLEY, DOI: 10.1520/STP163120190152

MIL HDBK 1823A.8 Attributes of a properly executed POD study include the use of representative flaws, not artificial defects, that are independent (e.g., the same flaw cannot be sampled multiple times) and that are well characterized, meaning their size and location are known before the POD study is initiated. This process can be burdensome, but a POD curve is required for fracture critical components to enable performing risk calculations that determine the probability of failure. The rigor of the POD study is intended to determine the largest flaw that can be regularly missed, not the smallest flaw that can be found. This large undetected flaw is used as the feature that will determine the fatigue behavior in a fracture critical location. POD assessments typically have been applied to postprocess assessments and not for in-process monitoring. The challenge for the latter application is the requirement of a distribution of defects representative of their state at the time they are detected, adding time as a factor in the assessment. Current research is addressing various capabilities for in-process monitoring,6 but the output of these embedded monitoring systems needs to provide a parameter that can be used to manage the quality of the build. While several techniques show promise, this area is still under investigation.

Concluding Comments The promise of additive manufacturing as a cost-effective approach to make aerospace components has been extensively documented. Several normal controls (i.e., nonfracture critical components) have been implemented in the USAF and have realized considerable cost savings when compared to traditional procurement processes. However, the challenges for fracture critical components are much greater as their fatigue performance needs to be determined. This includes an understanding of the effect of differing defects on this performance and determining the preferred approach to detecting them with NDE/I. The severity of these challenges depends on multiple factors, such as intended application, geometry, access, location of assessment, and repeated assessment requirements. Because defining the requirements for NDE/I when additively manufacturing remains in the research and development stage, it is premature to consider additional standards development for these applications. However, as AM capabilities mature and as improvements in the understanding of NDE/I requirements and performance parameters are realized, specific standards addressing NDE/I of additive manufacturing for aerospace components need to be considered for potential development. ACKNOWLEDGMENTS

The views expressed in this paper are those of the authors and do not reflect the official policy or position of the United States Air Force, Department of Defense, or the United States Government.

.

163

164

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

References 1. 2. 3. 4.

5.

6. 7.

8.

.

USAF, “Durability and Damage Tolerance Certification for Additive Manufacturing of Aircraft Structural Metallic Parts,” EZ-SB-019-01 (Wright Patterson AFB, OH: USAF, 2019). Department of Defense Standard Practice—Aircraft Structural Integrity Program (ASIP), MIL STD 1530Dc1 (Washington, DC: U.S. Department of Defense, 2016). Department of Defense Standard Practice—Propulsion System Integrity Program (PSIP), MIL STD 3024 (Washington, DC: U.S. Department of Defense, 2008). D. Naguy, “Leveraging Additive Manufacturing to Enhance Mission Generation,” in Proceedings of the 2018 Additive Manufacturing for Maintenance Operations Workshop, ed. P. Hurt (Ann Arbor, MI: Department of Defense Additive Manufacturing for Maintenance Operations, 2018). Threats to Aircraft Structural Safety, Including a Compendium of Selected Structural Accidents/Incidents, ASC-TR-2010-5002 (Wright Patterson AFB, OH: USAF Aeronautical Systems Center, 2010). N. Shamsaei and M. Seifi, Structural Integrity of Additive Manufactured Materials and Parts (West Conshohocken, PA: ASTM International, 2020). Standard Terminology for Additive Manufacturing—General Principles—Terminology, EN ISO/ASTM 52900 (West Conshohocken, PA: ASTM International, approved December 1, 2015), https://doi.org/10.1520/F2792-15 U.S. Department of Defense, Nondestructive Evaluation System Reliability Assessment, MIL HDBK 1823A (Washington, DC: U.S. Department of Defense, 2009).

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190124

Shukri Afazov,1,2 Jamie Frame,1 Utkarsha Ankalkhope,1 Prveen Bidare,1 Yijun Liu,1 Wilson Vesga,1 and Ben Dutton1

Prediction of Residual Stress Evolution for End-To-End Process Chain of Laser Powder Bed Fusion Process and Determination of Fatigue S-N Curves Citation S. Afazov, J. Frame, U. Ankalkhope, P. Bidare, Y. Liu, W. Vesga, and B. Dutton, “Prediction of Residual Stress Evolution for End-To-End Process Chain of Laser Powder Bed Fusion Process and Determination of Fatigue S-N Curves,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 165–175. http://doi.org/10.1520/STP1631201901243

ABSTRACT

This paper presents prediction of residual stress evolution for end-to-end process chain of laser powder bed fusion (L-PBF) for an aero-casing component made of IN718. The end-to-end process chain includes the simulation of the L-PBF build process, removal of support structures, heat treatment cycle for stress relief, and application of surface-hardening processes such as shot peening and laser shock peening. The simulation of the end-to-end process chain was performed using validated process models for IN718. Validation of the L-PBF process model was carried out for the aero-casing component, where predicted and measured distortions were found to be in good agreement. The predicted residual stresses after each process of the chain were used to develop Manuscript received October 29, 2019; accepted for publication February 26, 2020. 1 Manufacturing Technology Centre, Pilot Way, Ansty Park, Coventry, CV7 9JU, UK S. A. http://orcid.org/ 0000-0001-5346-1933 2 Dept. of Engineering, Nottingham Trent University, Clifton St., Clifton Campus, Nottingham, NG11 8NS, UK 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

165

166

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

theoretical fatigue S-N curves using the endurance limit approach. This approach requires knowledge for the surface roughness, as well as the ultimate strength and relative density of the material (defect level). Understanding the evolution of stresses in manufacturing process chains as well as prediction of material properties for functional performance is essential to reduce iterations during the process and product development of L-PBF parts. The advantage of this approach is that S-N curves can be determined very rapidly at every location of the aero-casing or any other component without conducting any fatigue tests physically, which saves considerable cost and time. The S-N curves after the L-PBF build process, heat treatment, and surface hardening were determined and discussed. It was concluded that the proposed S-N curve predictive methodology can be employed in design workflows for estimation of high cycle fatigue in L-PBF process chains very early in the design stage. This would enable designers to mitigate fatigue problems by designing parts more proactively for L-PBF process constraints such as surface roughness, ultimate strength, residual stresses, and relative density or defect level in the given material. Keywords residual stress, powder bed fusion, fatigue, S-N curve

Introduction

.

Powder bed additive technologies are enablers for light-weighting, consolidating, and increasing the functionalities of parts by the ability to print complex threedimensional (3D) geometries. Many parts, including safety critical components, are subject to cycling loads that could affect the structural integrity of the components. To prevent parts from experiencing high cycle fatigue (HCF) in service, the HCF should be tackled at the design stage. The conventional way of conducting HCF analysis is to relate static or dynamic stresses to S-N curves that are obtained experimentally at the specimen level. This approach assumes that the specimen could capture the specifics of the manufacturing process route and its performance can be assumed at the component level. This approach assumes that all locations of the part experience the same fatigue performance. For instance, the surface roughness in L-PBF is a function of the build angle,1 which would result in different fatigue performance. The fatigue strength is also dependent on the manufacturing process, including the thermal and mechanical effect on the grains’ evolution.2 The idea of this paper is to present an alternative way of determining the S-N curves that would allow the HCF calculation at the design stage. This includes taking into account the surface roughness, ultimate strength, induced residual stresses, and additive manufacturing (AM) defects. This approach assumes that the effect of microstructure is captured by the change of ultimate strength during the manufacturing process chain. The surface roughness can be either measured or predicted at the entire surface in L-PBF. It can be also measured and predicted using empirical methods after applying postprocessing surface engineering technologies, such as sand blasting, linishing, laser polishing, machining, and mass finishing.3 The ultimate strength can be obtained by uniaxial testing of specimens, measuring the hardness

AFAZOV ET AL., DOI: 10.1520/STP163120190124

and relating it to ultimate strength or using physics-based predictive models (or both).4 The residual stresses for the whole component are typically obtained by validated numerical models using finite element (FE) analysis, which are compared to experimentally obtained results.5 The effect of defects on the HCF performance is known, but how to take into account the defects in the lifing calculations is not well understood. There are researchers working on the inclusion of the defects in the geometry from computed tomography scans and conducting stress analysis.6 The alternative is to incorporate them in the S-N curve as presented in this paper. The main contribution of this paper is to demonstrate how numerical predictive models can be used for the prediction of S-N curves that can enable designers and analysts to conduct HCF calculations and ensure the structural fatigue performance of components that are manufactured by L-PBF. The paper first describes the theory of the endurance limit approach and how theoretical S-N curves could be obtained. The paper then presents a process chain simulation of L-PBF followed by heat treatment (HT) and surface-hardening processes of an aero-casing component. The simulation results present the validation of the L-PBF process model by comparing predicted and measured distortion, as well as the prediction of the residual stress evolution in two manufacturing process chains. Finally, the evolution of the S-N curves in the process chains is determined at a different location of the aero-casing following a validation example.

Endurance Limit Approach The endurance limit approach has been used in industry to obtain fatigue stress limits in metals.7 The approach determines the fatigue stress limits at 1,000 and 1,000,000 cycles using equations (1) and (2): rf ¼ 0:9ru at 1,000 cycles

(1)

rf ¼ 0:5ru fd at 1,000,000 cycles

(2)

where: rf is the fatigue strength, ru is the ultimate strength, f is the surface factor, and d is a defect factor introduced in this study. The ultimate strength can be obtained from a uniaxial test. The surface factor is a function of the average surface roughness value Ra and the ultimate strength. The relationship is described in figure 1 where the experimentally obtained Ra value and the ultimate strength are used to determine the surface factor. As known, the L-PBF can introduce defects in the parts such as porosity and microcracks. Those defects affect the fatigue performance and are represented by the parameter d. Considering the effect of the mean stresses using the Goodman approach, equation (2) can be expanded to equation (3): rf ¼ 0:5ru fdð1  rm =ru Þ at 1,000,000 cycles .

(3)

167

168

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 Fatigue surface factor for metal parts plotted against material tensile strength and average surface roughness.8 Note that units are in ksi and microinch(μin).

where rm is the mean stresses due to applied load to the component and induced residual stresses. In most metals, the S-N curve is described by equation (4): rf ¼ AN n

(4)

where: N is the number of cycles, and A and n are material constants. In this study, equations (1) and (3) are used to obtain two points from the S-N curves. The two points are then used to obtain the material constants A and n by fitting.

Prediction of Residual Stresses in L-PBF Process Chains The modeling techniques for L-PBF process chain simulation presented by Yaghi et al.9 were applied to the aero-casing. The FE method was applied to simulate the build process of the aero-casing in ABAQUS using the inherent strain method. The inherent strain method consists of directly applying strain values in the FE model instead of using the traditional method of predicting temperature and using a thermal expansion model to compute the thermal strains. The fundamental question when using the inherent strain method is what strain value needs to be applied in the FE model. In this paper, a constant compressive inherent strain is applied based .

AFAZOV ET AL., DOI: 10.1520/STP163120190124

on the ratio between the yield stress and modulus of elasticity. The idea behind this approach is that the applied strain will be sufficient to reach plasticity in the material. Once the plasticity is reached based on the von Mises criterion, the plasticity calculations are executed, and the direct and shear stresses are calculated based on the applied plasticity model. An elastic perfectly plastic material model is applied to conduct the plasticity calculations in this study. The main challenge in this approach is to select the yield stress and the modulus of elasticity for the specific process parameters and L-PBF machine. Therefore, a calibration step is introduced to gain confidence in the applied yield stress and modulus of elasticity. A cantilever beam component is used to obtain the values for the yield stress and modulus of elasticity. Based on the study by Yaghi et al.5 on In718, the inherent strain value prescribed in the model for IN718 in the three Cartesian directions is 0.0042353, equivalent to the negative value of dividing the yield stress of the material (720 MPa) by the elastic modulus (170 GPa). The inherent strain values have been applied per activated layer using the UEXPAN subroutine in ABAQUS. The activation of layers has been done using the element birth technique, where all elements in the layer are activated in the stiffness matrix. Once the elements have been activated, the inherent strain values are applied by the UEXPAN subroutine in the form of a strain increment per layer. The strain increment is applied to the activated elements just once. The same modeling approach and assumptions have been applied to the aero-casing component including the same slicing of 1-mm layers and meshing with 1-mm, eight-node, linear, hexahedral, 3D solid elements. The support structures were modeled with spring elements where stiffness was prescribed in all three directions of the Cartesian system.9 After the simulation of the build process was completed, the spring elements representing the support structures were deactivated to simulate the support removal process. The final predicted shape of the process model was aligned to the nominal computer-aided design (CAD) geometry and the surface deviations were obtained. The same approach was applied to the experimentally measured shape using blue light optical scanning technology. Both surface deviations are compared in figure 2. It can be seen that predicted and experimentally obtained surface deviations agree considering the complex shape of the aero-casing; hence, the L-PBF process model can be considered as valid. The residual stresses and distortion from the L-PBF process were used as an input for the stress relief model. The stress relief thermal cycle simulated heating up to 980 C, keeping at that temperature for 2.5 h, and rapid cooling down to room temperature. The rapid cooling down was assumed in the simulation to avoid the extra residual stress generated due to precipitation of the secondary phases. Temperature-dependent modulus of elasticity and yield stress were applied for the heating and the cooling steps as given in table 1. The holding step was modeled by applying the Norton creep law e_ ¼ Arn , where e_ is the strain rate, r is the applied stress, and A and n are creep constants. The applied stress is represented by the residual stresses from the L-PBF. The temperature-dependent material properties .

169

170

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 Comparison of measured and predicted surface deviations.

are provided in table 1. The Norton creep constants used in this study are A ¼ 2.05  1015 and n ¼ 5.53 (units in MPa and seconds) adopted from Thomas et al.10 The constants corresponding to 980 C have been prescribed in the mechanical FE model.

TABLE 1 Temperature-dependent material properties for IN718

Temperature, 8C

.

Modulus of Elasticity, GPa

Poisson’s Ratio

Yield Stress, MPa

Thermal Expansion, lm/m8C

20

170

0.284

720

13.7

100

170

0.284

685

13.8

200

170

0.284

643

13.9

300

167

0.284

618

14.4

400

163

0.284

595

15.2

500

154

0.284

579

16.2

600

149

0.284

562

17.4

700

142

0.284

546

18.7

800

138

0.284

530

19.9

900

111

0.284

268

21.0

1,000

79

0.284

81

21.6

AFAZOV ET AL., DOI: 10.1520/STP163120190124

The simulation of shot peening (SP) and laser shock peening (LSP) was performed at microscale, where the stress profiles were predicted and compared against experimentally measured residual stresses.11 The stress profiles, including the stress profiles for the three direct and three shear stresses, were mapped onto the entire surface of the aero-casing. The FE mesh for the two surface-hardening processes was generated using inflation layers, where five layers of elements were created for a 1-mm depth from the surface. The distortion and residual stresses from the heat treatment model were mapped onto the surface-hardening process as initial conditions using published methods.12 The evolution of the residual stresses is shown in figure 3. It can be seen that compressive and tensile stresses are generated during the L-PBF. The stresses in the heat treatment were reduced to less than 50 MPa. The application of SP and LSP after heat treatment-induced compressive residual stresses at the surfaces. The predicted residual stresses in the L-PBF process chain were used for the prediction of the S-N curves using the endurance limit approach.

Prediction of Fatigue S-N Curves The endurance limit approach was first validated against test data published by Yahollahi and Shamsaei13 where the fatigue specimens were produced using L-PBF. The specimens were first stress relieved followed by hot isostatic pressing and aging. An average surface roughness value Ra of 20 mm has been measured for the tested specimens. In this study, an ultimate strength of 1,360 MPa is used, which is a typical value for this type of heat treatment process and material. The fatigue surface factor of 0.55 was obtained based on an Ra of 20 mm and an ultimate strength of 1,360 MPa (note that the units in fig. 1 are in inches and ksi). A defect factor of 0.9 was obtained empirically based on the best representation of modeled and experimental

FIG. 3 Predicted principal residual stresses in the L-PBF process chain.

.

171

172

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

results from Yahollahi and Shamsaei.13 It needs to be noted that this method is used based on one set of experimental data where the specimens might not have had many defects affecting the fatigue performance. Based on the simulation results after heat treatment, tensile residual stress of þ50 MPa was assumed to represent the mean value of the Goodman approach. It needs to be noted that the hot isostatic pressing and aging stages were not modeled, and the stresses could be lower due to precipitation of the gamma prime phase that has the ability to relive residual stresses, too. The fatigue strength at 1,000 cycles was calculated to be 1,224 MPa, while at 1,000,000 cycles, it was calculated to be 324 MPa. Based on these two predicted points of the S-N curve and the use of fitting to equation (4), constants A and n were obtained to be 4,620 and 0.192, respectively. Figure 4 compares the predicted and the experimentally obtained S-N curves. It can be seen that the proposed methodology for S-N curve prediction agrees with the experimental data. Following the same principles, the S-N curves can be predicted for each location of the part by knowing the residual stresses, surface roughness, ultimate strength, and classification of defects at each location of the component. As an example, the S-N curves for two nodes, depicted as Node A and Node B in figure 3, are predicted to demonstrate how the S-N curves change during the process chain. Node A and Node B are selected to illustrate the effect of the high-tensile and compressive residual stresses on the fatigue performance. Tables 2 and 3 show the parameters used for the determination of the S-N curves for Node A and Node B, respectively. The residual stresses are obtained from figure 3, while the surface roughness was measured for the L-PBF and HT processes and estimated based on published data for the SP and the LSP. The ultimate strength was assumed to be the same for the entire part. It is anticipated that predictive models will emerge in the near future to predict the material strength in different

FIG. 4 Comparison of experimental and predicted S-N curve of IN718. Experimental data adopted from Yahollahi and Shamsaei.13

.

AFAZOV ET AL., DOI: 10.1520/STP163120190124

TABLE 2 Parameters used for predicting the S-N curves for Node A from figure 3

Ultimate Strength, MPa

Surface Roughness Ra, lm

Residual (Mean) Stress, MPa

Surface Factor

Defect Factor

A

n

L-PBF

1,020

20.5

755

0.55

0.9

12,850

0.382

L-PBF þ HT

1,360

20.5

34

0.55

0.9

4,565

0.191

L-PBF þ HT þ LSP

1,360

2.0

350

0.73

1.0

2,400

0.097

L-PBF þ HT þ SP

1,360

4.0

1,040

0.62

1.0

2,014

0.072

TABLE 3 Parameters used for predicting the S-N curves for Node B from figure 3

Ultimate Strength, MPa

Surface Roughness Ra, lm

Residual (Mean) Stress, MPa

L-PBF

1,020

10.5

465

0.55

0.9

2,293

0.133

L-PBF þ HT

1,360

10.5

24

0.55

0.9

4,374

0.184

Surface Factor

Defect Factor

A

n

L-PBF þ HT þ LSP

1,360

2.0

336

0.73

1.0

2,420

0.099

L-PBF þ HT þ SP

1,360

4.0

984

0.62

1.0

2,062

0.075

locations of the part. The defect factor for the SP and the LSP is assumed to be one, considering that the two processes would fully mitigate any surface defects. However, further research is needed to determine the actual defect factor, including classification of defects and their evolution in the process chain. Figures 5 and 6 show the predicted S-N curves at Nodes A and B of the aerocasing (fig. 3). As expected, the fatigue strength improves by reducing the surface roughness and introducing compressive stresses at the surface. It can be noted that

FIG. 5 Prediction of S-N curves at Node A of the aero-casing (fig. 3).

.

173

174

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 6 Prediction of S-N curves at Node B of the aero-casing (fig. 3).

the fatigue strength at Node B is greater at 10,000,000 cycles after L-PBF than after L-PBF followed by heat treatment. The reason for this is that compressive stresses are induced at Node B after L-PBF. Nevertheless, most of the stresses after L-PBF at the surface are tensile, as shown in figure 3. Therefore, it is recommended that stress-relief heat-treatment solutions are used after L-PBF because the fatigue strength is low for the regions with tensile stresses as predicted for Node A (fig. 5).

Conclusions The following conclusions were reached based on the research conducted in this paper: • Modeling techniques for prediction of residual stresses and distortion in LPBF manufacturing process chains are mature enough to provide support during the product and process development. • A predictive model for fatigue S-N curves in L-PBF process chains is proposed and verified against experimental data from the literature. • The S-N curve predictive model can be adopted in additive manufacturing design workflows for calculation of fatigue performance. ACKNOWLEDGMENTS

The authors wish to acknowledge the High Value Manufacturing Catapult UK and the EU ENCOMPASS project for their financial support.

References 1.

.

A. Triantaphyllou, C. Giusca, G. Macaulay, R. Roerig, M. Hoebel, R. Leach, B. Tomita, and K. Milne, “Surface Texture Measurement for Additive Manufacturing,” Surface Topography: Metrology and Properties 3, no. 2 (2015): 024002, https://doi.org/10.1088/2051672x/3/2/024002

AFAZOV ET AL., DOI: 10.1520/STP163120190124

2.

3.

4.

5.

6.

7.

8. 9.

10. 11.

12. 13.

.

C. Panwisawas, C. Qiu, M. Anderson, Y. Sovani, R. Turner, M. Atallah, J. Brooks, and H. Basoalto, “Mesoscale Modelling of Selective Laser Melting: Thermal Fluid Dynamics and Microstructural Evolution,” Computational Materials Science 126 (2017): 479–490. M. Jamal, M. Morgan, and D. Peavoy, “A Digital Process Optimization, Process Design and Process Informatics System for High-Energy Abrasive Mass Finishing,” International Journal of Advanced Manufacturing Technology 92 (2017): 303–319. A. Rai, C. Korner, and H. Helmer, “Simulation of Grain Structure Evolution during Powder Bed Based Additive Manufacturing,” Additive Manufacturing 13 (2017): 124–134, https:// doi.org/10.1016/j.addma.2016.10.007 A. Yaghi, S. Afazov, A. Holloway, and W. Denmark, “Comparison of Fast Finite Element Modelling Techniques for Prediction of Distortion and Residual Stresses in Laser Powder Bed Fusion” (paper presentation, Design and Manufacturing Simulation of Additive Manufacturing Components, NAFEMS Seminar, Coventry, UK, September 28, 2017). L. Evans, T. Minniti, T. Barrett, A. Muller, and L. Margetts, “Virtual Qualification of Novel Heat Exchanger Components with the Image-Based Finite Element Method” (paper presentation, Ninth Conference on Industrial Computed Tomography, Padova, Italy, February 13–15, 2019). K. Sadananda, A. Vasudevan, and N. Phan, “Analysis of Endurance Limits under Very High Cycle Fatigue Using a Unified Damage Approach,” International Journal of Fatigue 29 (2007): 2060–2071. J. Bannantine, J. Comer, and J. Handrock, Fundamentals of Metal Fatigue Analysis (Upper Saddle River, NJ: Prentice-Hall, Inc., 1990). S. Afazov, W. Denmark, B. Lazaro-Toralles, A. Holloway, and A. Yaghi, “Distortion Prediction and Compensation in Selective Laser Melting,” Additive Manufacturing 17 (2017): 15–22. A. Thomas, M. El-Wahabi, J. Cabrera, and J. Prado, “High Temperature Deformation of Inconel 718,” Journal of Materials Processing Technology 177, nos. 1–3 (2006): 469–472. A. Yaghi, S. Afazov, and M. Villa, “Additive Manufacturing Process Chain Modelling and Simulation” (paper presentation, NAFEMS World Congress, Quebec City, Canada, June 17–20, 2019). S. Afazov, “Modelling and Simulation of Manufacturing Process Chains,” CIRP Journal of Manufacturing Science and Technology 6 (2013): 70–77. A. Yahollahi and N. Shamsaei, “Additive Manufacturing of Fatigue Resistant Materials: Challenges and Opportunities,” International Journal of Fatigue 98 (2017): 14–31.

175

176

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190126

Tommy Hyatt,1 Richard Martin,1 and Rich Fields1

Design of Coupons and Test Methodology for Orthotropic Characterization of FFFProcessed Ultem 9085 Citation T. Hyatt, R. Martin, and R. Fields, “Design of Coupons and Test Methodology for Orthotropic Characterization of FFF-Processed Ultem 9085,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 176–187. http://doi.org/10.1520/STP1631201901262

ABSTRACT

The increasing prevalence of additive manufacturing (AM) methods such as fused filament fabrication (FFF) is motivating the need to reliably predict mechanical behavior of additively manufactured parts. However, predicting mechanical behavior is highly dependent on the availability of accurate property data. Recently published material property data for polyetherimide (PEI, or Ultem) provides a significant step toward quantifying AM process repeatability, but mechanical properties reported are specific to the geometry of the coupons and are therefore less useful for structural analysis of other geometries. In order to quantify material properties in a way that is generically applicable and scalable to any geometry within a finite element model, mechanical testing needs to account for the orthotropy inherent to the additive manufacturing process. Additionally, edge effects and differing extrusion directions that typically exist between the perimeter contour passes and the interior fill passes for each printed layer need to be considered and accounted for. Current industry best practices for coupon fabrication for tensile, compression, and shear were evaluated, tested, and iteratively modified to determine the most effective coupon design and test configurations for capturing the orthotropic properties

Manuscript received October 31, 2019; accepted for publication March 18, 2020. 1 Lockheed Martin MFC, 5600 Sand Lake Rd., Orlando, FL 32819, USA T. H. http://orcid.org/0000-00021863-789X 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

HYATT ET AL., DOI: 10.1520/STP163120190126

of FFF-processed Ultem 9085. Using the improved coupon design and test methodologies, three sets of coupons were fabricated and tested for tensile, compression, and shear in X, Y, and Z directions—both for pure contour construction and also for pure raster construction. Mechanical property results demonstrate typical properties that are meaningfully different than currently published data. Typical property results are presented. Using orthotropic design allowables in finite element models will allow analysts to predict mechanical behavior of arbitrary AM geometry. Keywords PEI, orthotropy, characterization, Ultem, coupon design, B-basis, FFF, FDM

Introduction In most markets, additive manufacturing (AM) is becoming increasingly common thanks to its many advantages, such as low cost, fast design cycles, and ability to easily create parts with complex geometry. Naturally, AM applications are expanding from simple representations of form to actual deliverable hardware. However, with this evolved use case comes a greater design and analysis burden—AM parts must be structurally capable for all expected loads. For most parts manufactured from isotropic materials using traditional manufacturing methods, a finite element (FE) model is constructed and analyzed. This analysis can be used to reliably evaluate margin to requirements, identify areas of concern, predict failure, and generally quantify the expected part performance. One of the key enablers for this process is the highly predictable mechanical behavior of traditionally manufactured metal parts—databases containing mechanical test data are readily available and their isotropic behavior is simple to simulate. In order for AM parts to achieve a similar level of analytical reliability, similar achievements must be reached—specifically, the manufacturing process must be adequately repeatable, and appropriate mechanical properties must be accessible to the analyst. This paper will focus on the collection of mechanical properties for fused filament fabrication (FFF) of a common polyetherimide (PEI): Ultem 9085.

Manufacturing Repeatability Manufacturing repeatability is functionally a prerequisite to measuring mechanical properties because the FFF properties are highly dependent on the manufacturing process. Ambient humidity, build plate temperature, nozzle temperature, cooling rate, extrusion rate, head travel rate, initial z-offset, per-layer z-offset, and backlash tolerances are just a few examples of process variables that can influence the structure and performance of an FFF part. From an industry standpoint, process repeatability is monitored through testing of witness coupons. The America Makes initiative recently published property data for standardized coupon geometry and construction,1 allowing any manufacturer to directly compare their print process .

177

178

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

results to the America Makes results and thereby confirm their manufacturing process parameters are correct.

Mechanical Properties Although the published data from America Makes contain property data for specific witness coupon geometry, those coupons were mostly designed with multitype (both contour and raster in the same coupon) microstructures and are further confounded by possessing different proportions of contour and raster as a function of test specimen and test direction. Those results, therefore, have more limited utility; while they certainly have value for process control, most of the America Makes results cannot easily be applied to an FE model of a part. The most accurate type of structural analysis of an arbitrary FFF part requires orthotropic property data for each geometry construction technique. For context, a brief review of FFF construction techniques and orthotropy is presented here.

FFF Construction Techniques When constructing a given geometry using the FFF process, each individual layer is typically printed in two zones: the contour pass is a continuous perimeter outline of the layer that is usually one filament wide. The raster pass is a set of parallel filaments that fill the interior of the layer (fig. 1). The orientation of the raster passes typically changes by 90 from one layer to the next. The amount of fill can be significantly reduced to save material and shorten print time, but this comes at the cost

FIG. 1 Typical solid FFF construction with perimeter contours and alternating raster in-fill.

.

HYATT ET AL., DOI: 10.1520/STP163120190126

of reduced structural integrity. This paper only evaluates 100% infill for raster passes because that is typical of parts being fabricated with structural intent.

Orthotropy Orthotropic materials have material properties that differ along three mutually orthogonal axes. Orthotropy is often observed in laminated materials because the bond strength between layers tends to differ from the bulk properties of individual layers. Similarly, X and Y properties differ when the layers are composed of directionally aligned filaments. Although many thermoplastics used in AM processes are isotropic as bulk materials, they become orthotropic when FFF processed due to the linear deposition process, which creates a regularly repeating pattern of material discontinuity and different amounts of fusion in different directions.

Project Summary The intent of this project was to obtain orthotropic property data for Ultem 9085 printed on a Fortus 900 þ three-dimensional (3D) printer. Early testing results demonstrated that the current standards for coupon geometry and test methods are not optimized for FFF material. A series of “pathfinding” tests were conducted to evaluate various print settings, coupon geometries, and coupon preparations. Results from the pathfinding tests established the coupon design, fabrication, and testing parameters that were used for collecting orthotropic property data.

Coupon Design-Pathfinding CONTOUR VERSUS RASTER

Although the extrusion tip has a round opening, the as-printed cross section of the filaments is an ellipsoid, which creates a wider and flatter interface from one layer to the next, while typically providing tangent-contact between adjacent filaments within a single layer (fig. 2). Functionally, this means that the layer-to-layer fusion of the contour passes is generally more robust, whereas raster passes end up with intermittent contact between layers. Even though raster passes have less fusion area from one layer to the next, both raster and contour passes are designed to have continuous contact between adjacent filaments within each layer. In order to decouple these effects, all coupons were constructed using either pure contour or pure raster—no combined construction was used. EDGE AND SURFACE EFFECTS

While evaluating some of the initial coupon trials, photomicrographs revealed significant smearing near the end of each print line just before and after the print head changed directions (fig. 3). Additionally, it was observed that the first several layers had less-consistent filament cross section and fusion to adjacent filaments (fig. 4). To prevent these anomalies from influencing the measured property results, Ultem .

179

180

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 Pure contour FFF cross section showing ellipsoid filaments. The two lines are equal length.

was printed into oversized plates, which were subsequently machined using computer numerically controlled (CNC) machining centers down to the intended thickness by removing material from the lower faces, after which the final perimeter was then cut out.

FIG. 3 Smearing observed where the print head changes direction.

.

HYATT ET AL., DOI: 10.1520/STP163120190126

FIG. 4 Inconsistent filament shape and fusion observed at the first layer.

Another surface effect that is common in 3D printing is thermally induced distortion forces during printing. For broad and relatively thin plates, those forces were occasionally able to cause one corner of the printed plate to lift off the build platform and thereby scrap the whole print. To prevent this transient thermal condition from ruining the printed plates, a brim and corner disk strategy was implemented (fig. 5). This is a small extension to the footprint of the entire plate for the first few layers, with an extra circular feature at the corners of each plate. These extended layers provided a thinner and more flexible area for the thermal gradient to resolve. After implementing the brim and corner discs, no new build failures occurred. MACHINING APPROACH

To obtain dimensionally accurate specimens from the Ultem plates, CNC machining was used to cut out the coupons. Given that Ultem is hydroscopic, machining was done without coolant. The coupons were machined using sharp cutting tools with light passes because more aggressive machining tended to result in some smearing of surface material. Photomicrographs of the as-machined surface finish

FIG. 5 Brim and corner disk to minimize thermal distortion.

.

181

182

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

were compared to samples that were mounted in epoxy and polished down through the surface to determine whether the machining process was damaging or altering the outer layers of the coupons. The side-by-side comparison of the as-machined surface versus the mounted-and-polished surface showed that the filament crosssectional shapes, layer-to-layer fusion, and filament-to-filament fusion were nearly indistinguishable. IN-PLANE TENSILE COUPONS

The initial approach to in-plane tensile testing (x and y) was to use the ASTM D638, Standard Test Method for Tensile Properties of Plastics, Type 1 geometry (fig. 6).2 Unfortunately, trial coupons demonstrated that failures frequently occurred at the fillet, and the measured coefficient of variance (CoV) was very tight, indicating that a stress concentration was causing premature failure. After performing some FE modeling-based coupon shape parameter studies and executing some pathfinder preliminary feasibility testing, the streamline-style coupon that was studied in the early 1980s3 was resurrected for use. The resulting coupon remains a dog-bone-style coupon that maintains the ASTM D638 notion of a reduced area test section but with a different transition profile between the gage section and the grip section. The coupon edge curvature follows a particular streamline of a laminar 2D flow field in a channel with an abrupt expansion (much initial work was done on right-angle expansions, but other transitional expansion shapes were also considered by our parameter study). The streamline samples exhibited failures within the gage section at various locations, a slightly higher CoV, and a significantly higher average tensile strength. Based on these results, the streamline geometry was selected for tensile strength evaluations. Z-DIRECTION TENSILE COUPONS

The AM process can be used to print ASTM D638-style samples vertically, but the aspect ratio would necessitate surrounding support structure for the coupon to maintain acceptable print quality. Even if this were not an issue, the relatively small and narrow cross-sectional area was expected to be problematic for raster construction. Two alternatives were explored: ASTM D7291, Standard Test Method for

FIG. 6 ASTM D638, Type 1 (top), and streamline (bottom).

.

HYATT ET AL., DOI: 10.1520/STP163120190126

Through-Thickness “Flatwise” Tensile Strength and Elastic Modulus of a FiberReinforced Polymer Matrix Composite,4 (spool style) and ASTM E8, Standard Test Methods for Tension Testing of Metallic Materials,5 (round dog-bone style). The spools were labor-intensive to fixture and ultimately ended up being overly sensitive to alignment between the upper and lower bonded grip areas. The ASTM E8 coupons were straightforward to produce and test but had several occurrences of failure initiating in the fillet. Two modified ASTM E8 profiles were evaluated—one with very large fillet transitions and one that followed the streamline profile that was downselected to in-plane tensile and just scaled to the ASTM E8 geometry. A set of each was tested, and the streamline version was selected due to its appropriate scatter in the data and failure modes. Surface finish was also a concern—when using typical settings for speeds and feeds on the lathe, the surface appeared super smooth, almost glassy. Upon closer review with an optical microscope, it was confirmed that there was no evidence of individual filaments or layers on the outer surface. This indicated that the selected speeds and feeds were smearing or remelting the Ultem (or both). After testing several different machining settings, we demonstrated that a broader cutting edge and a faster translation per rotation produced an acceptable surface finish without losing the individual filament and layer definition. The resulting surface finish also appeared to be visually “in family” with the coupons that were CNC machined out of Ultem plates. SHEAR COUPONS

ASTM D5379, Standard Test Method for Shear Properties of Composite Materials by the V-Notched Beam Method,6 Iosipescu shear samples were used as the baseline. During concept evaluation testing, it was demonstrated that, for some AM constructions, deformation occurred outside of the gage region, resulting in a poor measure of how much energy was required to cause a shear failure. As such, following guidance from ASTM D5379 (Section 8.2.2.2), G10 phenolic tabs were bonded to the grip area of the Iosipescu samples. This significantly stiffened the body of the coupon and ensured that the gage section reacted the applied shear forces. This resulted in appropriate failure modes and was used in all subsequent shear tests. COMPRESSION COUPONS

The first-pass coupon geometry used on compression testing (both for modulus and for strength) was the specimen type for highly orthotropic, thin-plate laminates as described in ASTM D695, Standard Test Method for Compressive Properties of Rigid Plastics.7 Initial trials revealed that this specimen geometry failed in buckling and did not represent a true compressive strength or modulus. The alternate specimen type selected was a rectangular prism with a square cross section. Two configurations of this coupon, following guidance from ASTM D695, were used where modulus and strength were measured independently. This change allowed modulus and ultimate strength to be measured reliably. .

183

184

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 1 Grips and instrumentation used for mechanical testing

Coupon Type

Grips

Instrumentation

Notes

Mechanical wedge grips

Instron single 2-in. axial

Sandpaper-lined grips

with flat-faced inserts

and transverse

Z-direction tensile

Mechanical wedge grips

Instron dual averaging

(streamline profile,

with v-faced inserts

1-in. axial and transverse

ASTM D5379 test fixture

Micro Measurements

In-plane tensile (streamline profile, flat)

extensometers

extensometer

round) Iosipescu shear (G10

EA-00-062TV-350

tab reinforced grip

strain gages

area) Compression modulus (rectangular prism:

ASTM D695 self-aligning

Instron single 1-in. axial

compression fixture

extensometer

long) Compression strength (rectangular prism:

ASTM D695 self-aligning Test machine compliance compression fixture

corrections

short)

Dimensional Inspection and Mechanical Testing All coupon dimensions were inspected using a digital caliper and micrometers meeting the requirements of the corresponding test methods. All testing was performed using an Instron 5985 load frame, with a 30-kN load cell, and recorded using Bluehill software (Version 3.77). The grips and coupon-specific instrumentation that were used are outlined in table 1.

Minimum Test Matrix Size for Statistical Evaluation To obtain B-basis properties, each measured property needs to be based on at least six replicates from each of three different material batches.8 This is compounded by construction method (contour, raster 6 45, and raster 0/90) and by coupon uniqueness imposed by differences in filament and load directions—for tension and compression, which adds a factor of three (X, Y, and Z), and for shear, which adds a factor of six (XY, YX, XZ, ZX, YZ, and ZY). The complete testing matrix, therefore, calls for 810 coupons, as outlined in table 2. The total replicate count for the test matrix in table 2 can be reduced by taking advantage of material symmetry, as outlined here. Testing that was omitted based on these symmetry relationships is reflected by blanks in table 3. The final test matrix evaluated 576 individual coupons. • Z tensile: 0/90 raster ¼ 6 45 raster • Raster 0/90, 6 45: X tensile ¼ Y tensile .

HYATT ET AL., DOI: 10.1520/STP163120190126

TABLE 2 Coupons required to develop room-temperature-ambient B-basis allowables

Testing Directions

Constructions (Contour, Raster 6 45, Raster 0/90)

Coupons per Batch

Batches

Total

Tension

3

3

6

3

162

Compression

3

3

6

3

162

3

3

6

3

162

6

3

6

3

324

(modulus) Compression (strength) Shear

810

• • • • • • • • • •

Raster 0/90, 6 45: X comp ¼ Y comp Raster 0/90: Shear YZ ¼ Shear XZ Raster 0/90: Shear ZY ¼ Shear ZX Raster 0/90: Shear XY ¼ Shear YX Raster 6 45: Shear YZ ¼ Shear XZ Raster 6 45: Shear ZY ¼ Shear ZX Raster 6 45: Shear XY ¼ Shear YX Contour: Shear XY ¼ Shear XZ Contour: Shear YZ ¼ Shear ZY Contour: Shear YX ¼ Shear ZX

Discussion of Results, Conclusions, and Future Work As utilization of FFF parts continues to increase, the demand for reliable mechanical property data will continue to increase. To support this demand, coupons were designed to align with industry best practices while implementing improved geometries and preparations where necessary. The data gathered demonstrate significant differences in mechanical properties across the spectrum of filament orientations and construction techniques; in other words, FFF-processed polyetherimide is orthotropic. This is consistent with expectations based on the known material continuity differences inherent to the FFF process. Standard coupon design and test methods for FFF-processed materials should be revised to address identified deficiencies. Additionally, a multimachine, multisite characterization effort is necessary to generate allowables with adequate statistical basis for use in FEM across industry. These steps will support and enable further advancements for FFF parts as primary structures. .

185

186

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 3 Typical properties summary

3

Contour

Raster 6 45

Raster 0/90

0.0420

Density X (lbs/in )

0.0419

0.0420

Density Y (lbs/in3)

0.0420

–

–

Density Z (lbs/in3)

0.0426

0.0425

0.0422

Tension X – modulus (msi)

0.363

0.262

0.293

Tension Y – modulus (msi)

0.264

–

–

Tension Z – modulus (msi)

0.301

0.290

–

Tension X – yield (ksi)

6.753

3.633

4.311 N/A

Tension Y – yield (ksi)

N/A

N/A

Tension Z – yield (ksi)

4.900

4.592

–

Tension X – strength (ksi)

12.018

5.863

6.565 –

Tension Y – strength (ksi)

2.223

–

Tension Z – strength (ksi)

7.952

6.922

–

Compression X – modulus (msi)

0.387

0.286

0.312

Compression Y – modulus (msi)

0.283

–

–

Compression Z – modulus (msi)

0.318

0.315

0.308

Compression X – yield (ksi)

9.711

5.504

7.463

Compression Y – yield (ksi)

6.555

–

–

Compression Z – yield (ksi)

7.436

7.616

7.703

Compression X – strength (ksi)

16.086

17.920

18.425

Compression Y – strength (ksi)

28.006

–

–

Compression Z – strength (ksi)

16.459

40.655

33.430

Shear XY – modulus (msi)

0.1099

0.1265

0.1095

Shear YX – modulus (msi)

0.1034

–

–

Shear ZY – modulus (msi)

0.1195

0.1157

0.1089

Shear YZ – modulus (msi)

0.1185

0.1180

0.1117

Shear XY – yield (ksi)

2.201

3.760

2.447

Shear YX – yield (ksi)

2.060

–

–

Shear ZY – yield (ksi)

2.225

3.003

2.548

Shear YZ – yield (ksi)

2.600

2.722

2.580

Shear XY – strength (ksi)

3.103

6.849

4.662

Shear YX – strength (ksi)

2.470

–

–

Shear ZY – strength (ksi)

4.393

5.505

4.933

Shear YZ – strength (ksi)

5.118

5.749

5.070

References 1.

“America Makes Announces Complete, Qualified Database of Material Properties for TM Fused Deposition ModelingV (FDMV) Additive Manufacturing of ULTEM 9085 Resin,” America Makes, February 6, 2019, http://web.archive.org/web/20191031161905/https:/ R

.

R

HYATT ET AL., DOI: 10.1520/STP163120190126

2.

3.

4.

5.

6.

7.

8.

.

www.americamakes.us/america-makes-announces-complete-qualified-database-materialproperties-fused-deposition-modeling-fdm-additive-manufacturing-ultem-9085-resin Standard Test Method for Tensile Properties of Plastics, ASTM D638-14 (West Conshohocken, PA: ASTM International, approved December 15, 2014), https://doi.org/10.1520/ D0638-14 D. W. Oplinger, K. R. Gandhi, and B. S. Parker, Studies of Tension Test Specimens for Composite Material Testing, No. AMMRC-TR-82-27 (Watertown, MA: Army Materials and Mechanics Research Center, 1982). Standard Test Method for Through-Thickness “Flatwise” Tensile Strength and Elastic Modulus of a Fiber-Reinforced Polymer Matrix Composite, ASTM D7291/D729M-07 (West Conshohocken, PA: ASTM International, approved April 1, 2007), https://doi.org/ 10.1520/D7291_D7291M-07 Standard Test Methods for Tension Testing of Metallic Materials, ASTM E8/E8M-16a (West Conshohocken, PA: ASTM International, approved August 1, 2016), https:// doi.org/10.1520/E0008_E0008M-16a Standard Test Method for Shear Properties of Composite Materials by the V-Notched Beam Method, ASTM D5379/D5379M-12 (West Conshohocken, PA: ASTM International, approved July 12, 2012), https://doi.org/10.1520/D5379_D5379M-12 Standard Test Method for Compressive Properties of Rigid Plastics, ASTM D695-15 (West Conshohocken, PA: ASTM International, approved September 1, 2015), https:// doi.org/10.1520/D0695-15 U.S. Department of Defense, “Guidelines for Property Testing of Composites,” in The Composite Materials Handbook, CMH-17, Vol. 1 (Washington, DC: 2002), 2.1–2.92.

187

188

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190114

R. Sunder,1 Ramesh Koraddi,1 and Andrei Gorunov2

Intrinsic Threshold Stress Intensity of Additive Manufactured Metals Citation R. Sunder, R. Koraddi, and A. Gorunov, “Intrinsic Threshold Stress Intensity of Additive Manufactured Metals,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 188–202. http://doi.org/10.1520/STP1631201901143

ABSTRACT

Intrinsic, closure-free, threshold stress intensity, DKth,i, is uniquely related to a certain computable near-tip residual stress, r*; r* is sensitive to applied stress ratio at given Kmax under constant-amplitude loading and, particularly so, to load history under variable-amplitude loading. DKth,i will vary significantly, depending on load conditions, rendering its characterization under controlled r* crucial. Experiments on additive manufactured (AM) PH1 steel and 18Ni300 maraging steel confirm the possibility of characterizing the relationship between DKth,i and r* using small-size compact tension, C(T), specimens. This opens the way for material characterization from miniature specimens cut out from smaller components. Application of the relationship between DKth,i and r* leads to improved assessment of the structural integrity of AM components through residual crack growth analysis as well as specification of allowable defect size. It can serve as an input for optimization of AM process parameters. Keywords intrinsic threshold stress intensity, near-tip residual stress, miniature specimens, test procedure

Manuscript received October 8, 2019; accepted for publication March 16, 2020. 1 BISS Division, ITW-India (P) Ltd, 497E, 14th Cross, 4th Phase, PIA, Bangalore 560058, India R. S. http:// orcid.org/0000-0002-5339-0132 2 Kazan Aerospace Institute (KAI-KNRTU), 10, Karl Marx St., Kazan, Russian Federation http://orcid.org/ 0000-0003-3154-5276 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

SUNDER ET AL., DOI: 10.1520/STP163120190114

Introduction An inevitable consequence of additive manufacturing (AM) is the potential for process-related defects by way of porosity and disbond that may occur randomly.1–5 From a fracture mechanics viewpoint, these constitute preexisting flaws, whose possible growth to failure may be modeled as one of fatigue crack growth.6 The onset of crack growth under service conditions will be governed by threshold conditions of fatigue crack extension. ASTM E647, Standard Test Method for Measurement of Fatigue Crack Growth Rates,7 describes the standard practice for characterization of threshold stress intensity, DKth. DKth is assumed to define conditions for the onset of fatigue crack growth. It is the fracture mechanics equivalent of fatigue limit. DKth values obtained by standard practice are associated with test conditions that involve long cracks and a load shedding process. Together, these induce crack tip conditions that deviate significantly from those, associated with the growth of preexisting flaws. They invariably include a substantial component of crack closure,8 a parameter that is sensitive to load history as well as crack size. Defects in AM materials will start growing in the total absence of crack closure and other associated extrinsic factors that affect the mechanics of the crack tip. To model such a process, one requires knowledge of a certain intrinsic component of DKth that we may refer to as DKth,i. Standard test practices such as ASTM E647 do not provide for estimates of DKth,i. Measurement of crack closure at threshold is complicated by the diminishing magnitude of applied load that makes it difficult to apply the technique prescribed by ASTM E647 to facilitate closure estimates. Because the exact magnitude of the closure component at threshold is typically unknown, one cannot possibly extend estimated DKth to load conditions with a different level of crack closure or to the required case of total absence of closure associated with naturally forming small cracks and process-related defects in AM materials. The problem is made worse by crack closure typically accounting for a large fraction of DKth, particularly at lower applied stress ratios. Attempts to isolate the intrinsic component, DKth,i, have proceeded on the premise that it is a material constant that can be determined at a sufficiently high stress ratio, such as R greater than 0.7, in isolation by compression precracking,9 or using the so-called cyclic R-curve method.10,11 Then, by subtracting the so-called constant intrinsic threshold stress intensity from a new readout of DKth, one can supposedly determine the so-called extrinsic component attributable to closure, oxide debris, fracture surface roughness, and so on. The path to practical implementation in engineering practice would then be restricted to the challenge of estimating these extrinsic components for a given practical case. For reasons that are forthcoming, the assumption of DKth,i being a material constant is deeply flawed. There is also a practical problem with extension of the standard decreasing-K testing practice per ASTM E647 to AM materials if the requirement is to characterize .

189

190

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

component level properties. In the event these components are small, test coupons cut from them to match specified standard crack geometries would be small, too. Then, the available interval of crack extension might not render it possible to reach a valid DKth, given the restrictions imposed by ASTM E647 on the maximum rate at which DK may be decreased. Because these rates cannot exceed specified limits, the associated required minimum interval of crack extension required to achieve threshold conditions might well exceed available specimen dimensions. Recent research established that in atmospheric fatigue, intrinsic, closure-free DKth,i is by no means a material constant.12 As a matter of fact, it appears to be uniquely related to near-tip residual stress, r*, a parameter that will depend on the applied stress ratio at a given Kmax under constant amplitude loading and will be strongly affected by load history under variable-amplitude loading. Available results show that, depending on r*, closure-free, intrinsic threshold, DKth,i can vary by as much as a factor of five! In summary, the unique relationship between DKth,i and r* characterizes threshold properties for a given environment and test frequency. Given such a relationship, in order to determine DKth,i in a practical application, one only needs to compute r* as a function of applied loading conditions. Test procedures to establish the unique relationship between DKth,i and neartip residual stress, r*, were developed that are now highly automated and reproducible12 as shown on a variety of materials;13 r* is computed from a near-tip cyclic hysteretic stress-strain response to applied stress intensity history using cyclic stress-strain properties of the material. The uniqueness of the DKth,i versus r* relationship is accompanied by the lack of any correlation between DKth,i and Kmax as well as between DKth,i and stress ratio, R. The new relationship appears to reflect the connection between crack-tip surface diffusion kinetics and the onset of fatigue crack extension in atmospheric fatigue, a phenomenon whose significance recedes with increasing crack growth rate and one that is absent in high vacuum.14,15 As near-tip stress moderates crack-tip surface diffusion kinetics, r* emerges as the key variable determining DKth,i; r* is sensitive to the effect of load history and therefore reflects cycle-sequence sensitive variation in DKth,i under variable-amplitude loading. This was confirmed by considerably improved engineering estimates of longduration residual fatigue life under service conditions that require the ability to account for load history effects on threshold.13 The objective of this study was to validate the new test technique on AM materials, with particular focus on miniature specimens that might not qualify to deliver valid test results using the conventional testing practice given inadequate available crack growth interval to meet requirements of load shedding. This would open the way for engineering application of the technique to characterize durability of AM components based on fracture mechanics considerations. The next section describes the experimental procedure pursued, including materials studied and testing procedure implemented. This is followed by a description of test results and their comparison with data obtained on other materials. The paper concludes with a discussion of the potential engineering application of the results obtained. .

SUNDER ET AL., DOI: 10.1520/STP163120190114

Experimental Procedure MATERIAL AND TEST COUPONS

The experiments were performed on compact tension C(T) specimens of 5 mm thickness and 20 mm width, made from two AM materials, a PH1 steel and a maraging steel. The PH1 specimens were directly three-dimensionally printed by direct metal laser sintering (DMLS) from raw spherical particles with an average diameter of 20 lm, with a deposition layer thickness of 30 lm at a scan rate of 800 mm/min and direction parallel to loading direction (L-T orientation). The three-dimensional (3D) printed specimens did not require further machining. The maraging steel specimens were wire cut from a 3D printed block manufactured using similar technology but from 18Ni300 powder. A 0.3-mm wire cut notch to a depth of a/ W ¼ 0.25 (a ¼ 5 mm) served as crack initiator. Because the specimens were miniature by nature, a conventional crack opening displacement (COD) gauge, as described by ASTM E647, cannot be used. Therefore, a customized COD gauge was designed to suit miniature specimens in order to measure crack size from unloading compliance. A picture of a test coupon along with a mounted COD gauge appears in figure 1. As shown in the figure, the arms of

FIG. 1 Test specimen with COD gauge mounted between clevis grips. The test specimens were 5 mm thick and 20 mm wide. Cantilever arms of the COD gauge were customized to fit within the 5-mm clevis on the grip.

.

191

192

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

the COD gauge are narrow enough for insertion into the clevis in order to mount on the knife edges at the mouth of miniature specimens. The PH1 steel specimens were of two batches, one as printed and the other after heat treatment to 465 C in order to relieve residual stresses. Test results indicate that heat treatment did marginally affect test results. TEST PROCEDURE

The testing was performed on a high-performance, 250-Hz, 10-kN BISS servohydraulic test system with real-time application software specially designed to suit the new threshold testing practice, as described elsewhere.12,13 Baseline cycling was performed at 100 Hz with adaptive amplitude control to keep loading error less than 1.5% of assigned amplitude, independent of change in specimen stiffness with crack extension. A total of seven specimens, five from PH1 Steel and two from 18Ni300 maraging steel were tested. One each of the specimens was tested to failure at Pmax ¼ 2 kN and stress ratio, R ¼ 0.3. These two tests provided the reference da/dN versus DK data that served as inputs to the settings for the threshold tests. In all the threshold tests, the baseline maximum load, Pmax, was 1.5 kN. The overload-underload conditions were programmed differently for each test, in order to cover a sufficient spread of r* values over which DKth,i was determined. The specimens were precracked from an initial notch size of 5 mm to a crack size of 6 mm at a stress ratio of 0.1 and a cycling frequency of 25 Hz. This was followed by the specially designed process to determine DKth,i under controlled r*.13 As in ASTM E647, the process involved gradual reduction in DK to the point of no discernible growth. However, contrary to standard practice, the test was performed under Pmax ¼ Const with optional periodic overload-underload cycles to induce highly controlled r*. Thus, irrespective of the rate of decrease of DK, maximum applied load ensured that the monotonic plastic zone ahead of the crack tip would always grow in size. This served as a guarantee of the absence of any dK/da-related history effects and also minimized closure. As a rule, stress ratio at threshold was sufficiently high to preclude closure. However, as a safeguard against erroneous estimates, crack closure was measured as per ASTM E647 to ensure that its value was much less than Pmin at threshold. Under slow unloading compliance measurement, the reproducibility of crack size estimates was found to be around 0.008 mm. As per ASTM E647, growth rate estimate is recommended over an interval exceeding ten times the resolution of crack size measurement. In keeping with this guideline, the “no-growth interval” was set to 0.1 mm and “no-growth cycle count” to a million cycles, corresponding to a growth rate of 107 mm/cycle. The chosen no-growth interval also ensures that crack extension during the periodic overloads of the order of 103 mm remains a negligible fraction of this interval. This is an assurance against nonachievement of no-growth conditions due to excessive crack extension during overload cycles, whose designated role is to set r*, rather than advance the crack. .

SUNDER ET AL., DOI: 10.1520/STP163120190114

FIG. 2 Schematic of three successive steps, n, n þ 1, and nþ2, of the loading scheme to determine intrinsic threshold, DKth,i. This illustrates conditions for change in cycle interval and baseline minimum load. The first of two overload cycles with identical POV,max is used to estimate crack size, while the second imposes controlled r* at the commencement of the next step of the baseline that will be determined by POV,min cycling

A schematic description of the loading scheme of baseline cycling combined with periodic overload cycles appears in figure 2. It shows three successive cyclic steps of loading. In terms of test control, two possibilities emerge at the end of each step. Thus, because no crack extension was detected at the conclusion of Step n, cycling interval DN is doubled over the next step. In the event of crack extension, such as after Step n þ 1, applied DK is reduced by 10% over the following Step n þ 2, maintaining the same cycle interval. In this manner, starting with an initial DN of 1,000 cycles, the cycling interval progressively doubles with reducing crack extension between applications of overloads and can reach hundreds of thousands of cycles at threshold. As explained in an earlier work,13 DKth,i, applicable to baseline cycling, is controlled by r* at PBL,min, which in turn is set by the stress intensity sequence associated with the preceding overload-underload sequence given by stress intensity, K, at POV,max and POV,min and governed by equations whose description is forthcoming. Further, it is assumed that as a consequence of certain crack-tip blunting due to the tensile overload, the crack tip will be fully open at POV,min. Given the aforementioned assumptions, the premise of the Masing stress-strain response as simulated by Wetzel,14 the applicability of the Neuber equation for conversion of applied elastic stress to local inelastic stress-strain,15 and the representation of material static and cyclic inelastic response per the Ramberg-Osgood equation in notch fatigue analysis,16 r* for baseline cycling is estimated by sequentially solving equations (1) through (4), where the first one assumes a monotonic .

193

194

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

K-stress response and the subsequent three equations follow the cyclic K-stress relationship as proposed in Sunder:12 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  1 KOV;max ¼

KOV;max  KOV;min ¼ KBL;max  KOV;min ¼ KBL;max  KBL;min ¼

2pr  Er1 ½rE1 þ

n;

r1 K0



(1)

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  1 ffi 2 2pr  Eðr1  r2 Þ½r1 r E þ2

r1 r2 2K 0

n;



rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  1 ffi 2 2pr  Eðr3  r2 Þ½r3 r E þ2

r3 r2 2K 0

n;



rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  1 ffi 

2pr  Eðr3  r Þ½r3 r E þ2

r3 r 2K 0

n;



(2) (3) (4)

where: E is Young’s modulus, K’ is the cyclic strength coefficient, n’ is the cyclic strain hardening exponent, r* is the distance from crack tip, and r1, r2, r3, and r* are the unknowns to be determined by solving equations (1), (2), (3), and (4), respectively. This process determines r* associated with DKth,i. For PH1 steel, E ¼ 180 GPa; K’ ¼ 1,450 MPa; and n’ ¼ 0.26. These constants were established from tests on a miniature low-cycle fatigue (LCF) specimen tested under strain control. For the 18Ni300 maraging steel, E ¼ 180 GPa, K’ ¼ 1,921 MPa, and n’ ¼ 0.21.17 Equations (1) through (4) describe cycle-sequence variation of near-tip stress due to the applied overload-underload sequence between steps of baseline cycling; r*, thus computed, will remain unchanged as long as crack extension over the following step is negligible by comparison to plastic zone size (i.e., the assumption of a stationary crack is valid). The first load excursion in terms of stress intensity is larger than all preceding load cycles. Therefore, r* at KOV,max is given by equation (1), which essentially describes a monotonic K-stress response and is quite independent of preceding load history. Further, KOV,max will not be exceeded over the following three load half-cycles. This renders equations (2) through (4) valid as cracktip stress-strain response will strictly follow cyclic response over these half cycles. Previous work showed that for a given material, the best correlation between DKth,i and r* is achieved at a particular r* that is selected by trial and error.13 This apparent anomaly may be considered in light of aspects that surround the relationship between DKth,i and r*. One, with the crack-tip being a singularity, it is not possible to define actual crack-tip inelastic stress, which would be the defining parameter for DKth,i. And two, the cyclic stress-strain relationship is typically obtained on smooth specimens and cannot be readily applied to the crack-tip surface that is subject to constraint. These two together may explain why r* is not a material-independent constant but a parameter to be selected for best correlation of .

SUNDER ET AL., DOI: 10.1520/STP163120190114

DKth,i versus r*. The excellent correlation between DKth,i and r* does obviously underscore the existence of a unique and reproducible relationship, even if it is at a particular r*. Multiple DKth,i estimates are possible from a single test coupon. As mentioned earlier, each estimate commences with an initial step duration of 1,000 cycles and an initial baseline DK designed to fall into the growth rate band of about 104 mm/ cycle. At the commencement of the next step, the programmed overload cycle is applied at R ¼ 0.1 and frequency 0.1 Hz in order to determine crack size and crack opening stress intensity, Kop. This is followed by the programmed overload/underload sequence to enforce the designated r* at the crack tip that will define growth conditions for the next step. If the crack has grown during the previous step, DK over the next step will be reduced by 10%. If it has not grown, the next step duration will be doubled. And, if, as a consequence of such doubling, more than a million cycles have accumulated without crack extension exceeding 0.1 mm, it is assumed that observed no-growth conditions correspond to a threshold growth rate of 107 mm/cycle. The baseline DK at this point is assumed to correspond to DKth,i, and the test process proceeds to the next DKth,i measurement with increased initial DK corresponding to 104 mm/cycle, initial cycle duration of 1,000 cycles, and a new set of overload-underload parameters. Care is taken in specifying test conditions to ensure that overload magnitude over the next step shall never be less than in the previous step. This ensures absence of overload plastic zone history-related load interaction that may distort near-tip residual stress from computed values. The 10% reduction in DK at each confirmation of even miniscule crack extension ensures rapid retardation of crack growth rate to no-growth conditions. However, in contrast to the process prescribed by ASTM E647, this reduction proceeds with a continuous increase in monotonic plastic zone size, combined with a highly controlled r* as well as suppressed crack closure. The process is essentially rate insensitive. The unloading half of the first overload cycle extends to near-zero load to serve the purpose of keeping crack closure down by squeezing the wake. This is to ensure that DKth,i readout conforms to fully open crack-tip conditions. Kop measured during the first overload cycle serves as a validation of the DKth,i measurement for zero closure. DKth,i estimates involving partial crack closure are to be discarded as invalid. None of the DKth,i estimates in this study were discarded based on this criterion. Individual step conditions by way of overload-underload parameters may be programmed before the commencement of a test. The test can then proceed automatically to completion. As each DKth,i estimate requires between two and four million cycles, a typical experiment can proceed uninterrupted for an extended period of time. About 2 to 3 mm of crack extension is consumed by a single DKth,i estimate. Overload magnitude is restricted to 50% of baseline load level. This is to ensure that the crack closure level is sufficiently low as an assurance that estimated DKth,i .

195

196

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

value is valid (i.e., there is no partial crack closure involved). Also, keeping the overload magnitude under 50% serves as an assurance of negligible crack extension during the overload. Essentially, focus is on the use of the overload cycle exclusively to induce the desired range of controlled variation in r* over which DKth,i will be estimated.

Test Results and Discussion Figure 3 shows a typical crack growth curve from one of the tests on maraging steel, combined with the curve showing variation in baseline DK. Each segment of retarding crack growth is associated with a progressive reduction in DK at Pmax ¼ Const through a steady increase in Pmin, eventually leading to a no-growth condition associated with DKth,i. As shown by figure 3, as many as four DKth,i estimates corresponding to different r* values were estimated over a useful crack growth interval of about 7 mm. This implies that even smaller test coupons may be used for such tests considering that hardly 2 mm of crack extension is adequate to reach the threshold. None of the fracture surfaces bore any visible sign of debris, rubbing, and so on. Also, Kmin at threshold in all cases exceeded estimates of Kop during the associated load/overload cycle, suggesting absence of closure (extrinsic component). The overload-underload sequences programmed for this test resulted in a steady drop in r* over successive DKth,i estimates. For this reason, DKth,i appears to

FIG. 3 Crack growth curve obtained during threshold test on maraging steel specimen. Each segment of steady retardation in growth rate concludes with attainment of no-growth condition corresponding to threshold. This is illustrated by the accompanying variation in baseline DK.

.

SUNDER ET AL., DOI: 10.1520/STP163120190114

increase with crack size. One may note, however, that the two are not related. They could have been programmed to yield decreasing or, for that matter, randomly varying DKth,i with crack size and Kmax. The unique relationship between r* and DKth,i renders the latter independent of both baseline Kmax as well as stress ratio. Figure 4 illustrates, by numerical simulation, how the combination of overloadunderload magnitude and baseline cycle at threshold together determine near-tip residual stress, r*, corresponding to DKth,i. Equations (1) through (4) determine the turning points in curves shown in the figure. Assuming tension-tension loading, the highest possible r* will be determined by Kmax in the absence of overloads. As a

FIG. 4 Numerical simulation based on equations (1) through (4) of near-tip stress versus applied stress intensity excursions, including overloads and underloads. Note the influence of KOV,max and KOV,min on near-tip stress corresponding to KBL,min. Increasing KOV,max will move near-tip stress at KBL,min toward compression. Decreasing KOV,min below KBL,min will have the opposite effect.

.

197

198

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

consequence of the effect of steep elastic unloading following an overload, KOV,max will essentially determine the lowest possible r* that can follow. Note the deleterious effect on r* of the extent of underload, KOV,min, following a tensile overload. Thus, for a given magnitude of overload, KOV,max, Dr* is the margin of possible variation in r* under variable amplitude loading. Figure 5 shows DKth,i values plotted against near-tip residual stress, r*. For both materials, the best fit between the two parameters was obtained at r* ¼ 0.03 mm. The trend in variation of DKth,i with r* is very similar to relationships obtained in earlier works for three materials.13 Both materials show a more than three-fold variation in DKth,i depending on r*. PH1 stainless steel shows a flattening of DKth,i beyond r* ¼ 300 MPa, while in the case of the maraging steel, the curve appears to continue its downward trend with increasing r*. Curiously, data obtained on SS316 in an earlier work13 also indicated a flattening trend in DKth,i beyond a certain r*. This common feature between two different stainless steels appears to deserve further study. Also, more tests with load level and overload sequences set to take r* beyond the range covered in these graphs are likely to reveal whether DKth,i can change beyond the range seen.

FIG. 5 DKth,i versus r* estimates for PH1 stainless steel and 18Ni300 maraging steel. Note that unlike PH1 that appears to suggest a flattening or saturated response at higher r*, the maraging steel appears to indicate a steady and noticeable decrease in DKth,i with an increase in r*. Also, note alleviation of residual stresses by heat treatment appear to render a small but noticeable increase in DKth,i at lower r* in PH1 steel.

.

SUNDER ET AL., DOI: 10.1520/STP163120190114

According to equations (1) and (2), r* at a given DK will steadily increase with increasing stress ratio under constant amplitude loading. Therefore, given the flat DKth,i versus r* response of PH1 steel at high r*, it is possible that if constant amplitude testing per ASTM E647 is performed at sufficiently high stress ratio or Kmax, one may indeed be led to mistakenly conclude that DKth,i is a material constant. However, a similar finding is unlikely in the case of maraging steel, which shows a steady decrease in DKth,i with increasing r*. Obviously, knowledge of how DKth,i varies with r* for individual materials is important from the standpoint of engineering application. Given the apparent absence of crack closure and of other extrinsic components, the data in Figure 5 may serve as inputs to determine minimum loading conditions under which preexisting defects or naturally forming cracks of a given initial size can be expected to grow at the threshold rate of 107 mm/cycle. Stated differently, for a given applied constant amplitude stress level and ratio or, for a given stress history, material properties given in the format of Figure 5 can be used to specify allowable initial defect size and thereby lead to optimized AM process parameters. Similarly, the well-known Kitagawa-Takahashi diagram18 may be extended to engineering application by using the DKth,i value selected from Figure 5 using r* computed for a given loading condition and defect size. The significance of the DKth,i versus r* relationship for engineering estimates of residual life or period between inspections may be judged from Figure 6, citing test results for heat treated PH1 steel as an example. The crack growth rate data points were obtained under constant amplitude loading at R ¼ 0.3. Also shown in the figure are speculative trend lines of crack growth rate tied to the extreme values of DKth,i for the material taken from Figure 5. The two trend lines may be deemed conservative because they are drawn from points of intrinsic threshold and exclude the potential extrinsic component that is inevitable in laboratory testing with load shedding. They describe the possible envelope of near-threshold crack growth rate response determined solely by the margin of observed variation in intrinsic threshold as given by the DKth,i versus r* relationship. In keeping with the receding effect of DKth,i with increasing growth rate, the two trend lines merge into the Paris regime, suggesting that growth rates exceeding 104 mm/cycle are insensitive to DKth,i . Thus, the envelope essentially depicts a gray area that needs to be suitably addressed in order to considerably improve the quality of residual life estimates associated with service loading conditions involving millions of smaller load cycles and falling into the so called high cycle fatigue / very high cycle fatigue (HCF/VHCF) domains.13,19

Conclusions 1. Intrinsic threshold stress intensity, DKth,i, was experimentally characterized for AM PH1 and 18Ni300 maraging steel as a function of computed neartip residual stress, r*. .

199

200

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 6 The da/dN data for heat-treated PH1 steel from a constant amplitude test at R ¼ 0.3. Also shown as circles are extreme DKth,i values for the material from figure 5 that marks its potential variation depending on loading conditions. Trend lines emanating from these two points describe the envelope of potential variation in near-threshold crack growth rate depending on instantaneous DKth,i that will, in turn, depend on load conditions and load history. Note that experimental data points lie close to the trend line corresponding to an extremely high intrinsic threshold.

2. The test technique delivers more conservative estimates of threshold stress intensity than is possible by conventional load-shedding techniques. Being intrinsic by nature, the estimates may apply to naturally forming cracks where extrinsic factors such as crack closure are absent. 3. The test procedure used to characterize the DKth,i versus r* relationship is suitable for miniature C(T) specimens that can be either printed or cut from AM components. As the crack increment to reach threshold is less than 3 mm, specimens as small as 8 mm in width are likely to be adequate to determine at least one DKth,i point. .

SUNDER ET AL., DOI: 10.1520/STP163120190114

4. r* is sensitive to load history and DKth,i can vary by almost five times depending on r*. Given the sensitivity of the near-threshold fatigue response to DKth,i, the quality of residual life estimates in the HCF/VHCF domain is likely to improve by accounting for the effect of load history on r* under service loading conditions. 5. The DKth,i versus r* relationship may be applied to optimize AM process parameters based on no-growth consideration of initial defect size. Similarly, the Kitagawa-Takahashi diagram may be used to leverage permissible design stress against defect size. ACKNOWLEDGMENTS

The 18Ni300 maraging steel material for this study was graciously provided by InTech DMLS Pvt Ltd, Bangalore, India. The tests were meticulously performed by R. Mehanathan.

References 1. 2.

3.

4.

5.

6.

7.

8.

9.

.

P. Edwards and M. Ramulu, “Fatigue Performance Evaluation of Selective Laser Melted Ti–6Al–4V,” Materials Science and Engineering: A 598 (2014): 327–337. E. Wycisk, C. Emmelmann, S. Siddigue, and F. Walther, “High Cycle Fatigue (HCF) Performance of Ti-6Al-4VAlloy Processed by Selective Laser Melting,” Advanced Materials Research 816–817 (2013): 134–139. H. Gong, K. Rafi, T. Starr, and B. Stucker, “Effect of Defects on Fatigue Rests of As-Built Ti-6Al-4V Parts Fabricated by Selective Laser Melting,” in Proceedings of the 23rd Annual International Solid Freeform Fabrication Symposium—An Additive Manufacturing Conference (Austin, TX: University of Texas, 2012), 499–506. D. Greitemeier, V. Holzinger, C. Dalle Donne, J. Eufinger, and T. Melz, “Fatigue Prediction of Additive Manufactured Ti-6Al-4V for Aerospace: Effect of Defects, Surface Roughness,” in Proceedings of the 34th ICAF Conference and 28th ICAF Symposium, ed. A. Siljander (Helsinki, Finland: The International Committee on Aeronautical Fatigue and Structural Integrity, 2015), 12–20. N. Hrabe, T. Gna ¨upel-Herold, and T. Quinn, “Fatigue Properties of a Titanium Alloy (Ti– 6Al–4V) Fabricated via Electron Beam Melting (EBM): Effects of Internal Defects and Residual Stress,” International Journal of Fatigue 94 (2017): 202–210. A. Yadollahi, M. J. Mahtabi, A. Khalili, H. R. Doude, and J. C. Newman, Jr., “Fatigue Life Prediction of Additively Manufactured Material: Effects of Surface Roughness, Defect Size, and Shape,” Fatigue and Fracture of Materials and Structures 41 (2018): 1602–1614. Standard Test Method for Measurement of Fatigue Crack Growth Rates, ASTM E647-15e1 (West Conshohocken, PA: ASTM International, approved May 1, 2015), https://doi.org/ 10.1520/E0647-15E01 W. Elber, “The Significance of Fatigue Crack Closure,” in Damage Tolerance in Aircraft Structures, ed. M. Rosenfeld (West Conshohocken, PA: ASTM International, 1971), 250–242, https://doi.org/10.1520/STP26680S J. C. Newman, Jr., J. Schneider, A. Daniel, and D. McKnight, “Compression Pre-Cracking to Generate Near-Threshold Fatigue Crack Growth Rates in Two Aluminum Alloys,” International Journal of Fatigue 27 (2005): 1432–1440.

201

202

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

10.

11.

12.

13.

14. 15. 16. 17.

18.

19.

.

K. Tanaka and Y. Akiniwa, “Resistance-Curve Method for Predicting Propagation Threshold of Short Fatigue Cracks at Notches,” Engineering Fracture Mechanics 30, no. 6 (1988): 863–876. J. Maierhofer, S. Kolitsch, R. Pippan, H. P. Ganser, M. Madia and U. Zerbst, “The Cyclic R-Curve—Determination, Problems, Limitations and Application,” Engineering Fracture Mechanics 198 (2018): 45–64. R. Sunder, “Characterization of Threshold Stress Intensity as a Function of Near-Tip Residual Stress: Theory, Experiment, and Applications,” Materials Performance and Characterization 4, no. 2 (2015): 105–130. R. Sunder, R. Koraddi, and C. Vishwas, “Threshold Characterization to Support Residual Life Estimates under Variable-Amplitude Service Loading” (paper presentation, ASTM Annual Symposium on Fatigue and Fracture, Denver, CO, May 15–17, 2019). R. M. Wetzel, “A Method for Fatigue Damage Analysis” (Ph.D. thesis, University of Waterloo, 1971). H. Neuber, Kerbspanmungslehre [Theory of Notch Stresses], trans. J. W. Edwards (New York: Springer, 1946). Society for Automotive Engineers, SAE Fatigue Design Handbook, 3rd ed. (Warrendale, PA: SAE, 1997). R. Branco, J. D. M. Costa, F. Berto, S. M. J. Razavo, J. A. M. Ferreira, C. Capela, L. Santos, and F. Antunes, “Low-Cycle Fatigue Behaviour of AISI 18Ni300 Maraging Steel Produced by Selective Laser Melting,” Metals 8, no. 32, (2018): 1–15. H. Kitagawa and S. Takahashi, “Applicability of Fracture Mechanics to Very Small Cracks or the Cracks in the Early Stage,” in Proceedings of the Second International Conference on Mechanical Behavior of Materials (Metals Park, OH: ASM, 1976), 627–631. R. Sunder, “Toward Improved Modeling of Variable-Amplitude Fatigue Crack Growth— Theory and Fractographic Validation,” in Structural Integrity Assessment, Lecture Notes in Mechanical Engineering, (Singapore: Springer Nature, 2020), 3–24.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190135

Yang Li,1 Hongyi Xu,2 Wei-Jen Lai,1 Ziang Li,1 and Xuming Su1

A Multiscale Material Modeling Approach to Predict the Mechanical Properties of Powder Bed Fusion (PBF) Metal Citation Y. Li, H. Xu, W.-J. Lai, Z. Li, and X. Su, “A Multiscale Material Modeling Approach to Predict the Mechanical Properties of Powder Bed Fusion (PBF) Metal,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 203–213. http://doi.org/10.1520/STP1631201901353

ABSTRACT

Metal parts manufactured via the powder bed fusion (PBF) process have drawn tremendous interest in the automotive industry. While numerous studies have shown the unique microstructure of the metal from the PBF process, significant variation of material properties with process parameters has been widely observed, indicating that huge amounts of experiments are required during material characterization. Thus, multiscale material modeling approaches are in great demand so that the properties of the metals via the PBF process can be predicted with confidence, to save costs and time during the design stage. In the present study, a multiscale modeling approach is proposed in which the microscale and mesoscale models are considered in finite element analysis. At the microscale, the model captures the microstructure characteristics within the melt pools to predict the representative properties resulting from epitaxial grain morphology and orientation. The properties are then homogenized and input into a mesoscale model in which the “fish-scale-like” melt pools and boundaries Manuscript received November 1, 2019; accepted for publication January 22, 2020. 1 Research and Innovation Center, Ford Motor Co., 2101 Village Rd., Dearborn, MI 48121, USA Y. L. http:// orcid.org/0000-0003-1461-5623, W.-J. L. http://orcid.org/0000-0002-1508-3019, Z. Li http:// orcid.org/0000-0001-5131-012X 2 Dept. of Mechanical Engineering, University of Connecticut, 191 Auditorium Rd., Storrs, CT 06269, USA http://orcid.org/0000-0001-8641-376X 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

203

204

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

between them are modeled. Stochastic reconstruction of the micro- and mesoscale models are performed based on statistical microstructure information obtained from optical micrographs and scanning electron microscopy (SEM) images. Predicted mechanical properties are compared with experimental data to demonstrate the capability of the approach. The study keeps focus on AlSi10Mg built by selective laser melting (SLM), while universal applicability to other material systems is expected. Keywords additive manufacturing, multiscale modeling, mechanical properties

Introduction Additive manufacturing (AM) technologies have been embraced by a wide range of industry sectors thanks to their unique advantages of achieving design complexity, integrating parts for assembly simplification, and so on.1–3 AM of alloys (e.g., AlSi10Mg and stainless steel 316L) via powder bed fusion (PBF) processes have drawn increasing interest from the automotive industry in recent years. In general, parts built with these alloys possess comparable or even better properties compared to similar materials built via conventional manufacturing processes.4,5 While the performance may not be a critical issue for PBF alloys, cost remains the biggest challenge in engineering practice, which arises in both the high price of raw materials and the elongated cycle time. Even in the design verification stage, the cost of physical testing is significant if a trial-and-error approach is utilized. To address these challenges, integrated computational material engineering (ICME) procedures are in great demand for automotive applications.6,7 With the ingredients of manufacturing process simulation, microstructure prediction, multiscale material modeling, and macroscale computer-aided engineering (CAE) simulation, an ICME workflow is enabled that balances the trade-off between performance and cost with a greatly reduced physical testing requirement. Extensive efforts in the development of the pieces of the puzzle have been reported in the literature,8–18 and a few piloting studies have already provided prototyping solutions for the end-to-end ICME workflow for PBF alloys.19–21 Nonetheless, the most recent advancement of microstructure characterization suggests that some specific microstructure features at different length scales (e.g., subgranular cellular structure in AlSi10Mg22–24 and “fishscale” melt pool and melt pool boundaries25–27) may have significant impact on the mechanical properties of the materials yet are not reflected in the existing multiscale material modeling framework. Therefore, we aim to propose a new microstructure reconstruction-based approach to improve the prediction capability of multiscale microstructure models. The remainder of this paper is organized as follows: The methodology of experiments and microstructure characterizations for the study are discussed, followed by a detailed description of the workflow of the developed multiscale material models. Then, the preliminary results are shown and validations for the developed .

LI ET AL., DOI: 10.1520/STP163120190135

models are provided. The discussions on the limitation and future work of this study are listed in the final section where the conclusions are summarized.

Methodology MECHANICAL TESTING AND MICROSTRUCTURE CHARACTERIZATION

The overview of the experiments and characterization is summarized in figure 1. Quasistatic tensile testing following ASTM E8, Standard Test Methods for Tension Testing of Metallic Materials (superseded) (subsize specimen),28 has been performed for twelve AlSi10Mg coupons built on an SLM Solutions, SLM 500 machine. No postprocessing is applied after the samples are removed from the build plate. The testing setup is shown in figure 1A. Default parameters by the vendor of the machine are followed. The hatching scan strategy is used, with an interval angle of 67 layer by layer. Six of the coupons are oriented vertically in the build chamber, whereas the other six are oriented horizontally. Extensometers are used to measure the strain in the gauge area for all the tests. Characterization of the microstructure is performed in the grip region of one coupon, with vertical orientation after the quasistatic tensile test is finished. Scanning electron microscopy (SEM) imaging (JEOL 6610) and optical microscope

FIG. 1 Overview of mechanical testing and microstructure characterization. The experiment setup is shown in (A). The SEM images of the cellular network in the melt pool and melt pool boundaries are shown in (B) and (C). The postfailure samples loaded along the Z axis are shown in (D). The UTS of samples loaded along the X and Z axis are shown in (E), while the optical microscopic images of the fish-scale-like mesoscale microstructure is shown in (F).

.

205

206

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

imaging (Keyence VHX 2000) are conducted on the mounted samples after polishing and etching. For SEM imaging, a low-resolution scan is first performed to find out the melt pool and melt pool boundary regions with distinct density of cellular structures, which is reported to be composed to aluminum-silicon (Al-Si) eutectics and nanosized Si particles. Zoom-in images are then taken on these areas. At a larger length scale (referred to as the mesoscale in the context of the present study), a “fish-scale” pattern, as shown in figure 1F, with overlapped elliptical melt pools and associated melt pool boundaries is captured, and the width of the melt pool boundaries is manually measured from the images. DEVELOPMENT OF THE MULTISCALE MATERIAL MODELS

To ensure the underlying physics in the PBF alloys is appropriately incorporated into the multiscale material modeling approach, crystal plasticity finite element modeling (CPFEM) is utilized for the simulation at microscale. Most recently, studies have successfully applied the CPFEM approach to PBF alloys, including AlSi10Mg,12 Ti-6Al-4V,13 and Inconel 718,14 to name a few. For practical consideration, a similar approach with simplified constitutive models is developed in this study. However, the properties of a melt pool boundary are very different from that of the melt pool due to remelting and the heat-affected-zone (HAZ) effect.25,26 Therefore, representative volume element (RVE) models with different microstructure characteristics and constitutive parameters in the crystal plasticity model are proposed for the melt pool and the boundary, respectively. A homogenized elastoplastic responses predicted from the microscale models for the melt pool boundary and melt pool are assigned to the mesoscale models to predict the property of the bulk material. Consideration of defects can be added to the mesoscale model, though that is not implemented in the present study. The proposed workflow is shown in figure 2. The SEM electron backscatter diffraction (EBSD) images in figure 1A and 1E in Takata et al.22 are used to provide the granular microstructure and orientations, respectively, for melt pool and melt pool boundaries. Because only crystalline angles can be read from the EBSD images, it is assumed that the major axis of each grain is along the vector of directions projected into the plane, which is reported to be the solidification direction for Al alloys.22 This assumption will be replaced with real EBSD data in the planned future work. MATLAB scripts are developed to extract the geometries and the grain orientations from the EBSD images. For the melt pool models, the grain boundaries in figure 1B and associated grain orientation are directly used, whereas a Voronoi tessellated model is reconstructed to represent the equiaxial grains in the melt pool boundary regions. The grains are assigned with random crystallographic orientations according to the observations in various references.28,29 An automated ABAQUS Python script is developed to generate conformal mesh to retain the smooth grain boundary, which yields fewer numerical issues than the typically used pixelated mesh. The physical dimension of the microscale models for melt pool and melt pool boundaries are 322 lm by 278 lm and 20 lm by 20 lm, respectively. For .

LI ET AL., DOI: 10.1520/STP163120190135

FIG. 2 Schematic of microscale model setup. The workflows for reconstruction of microscale models of melt pool and melt pool boundary model are shown in (A-C) and (E-G), respectively. The reconstructed microstructure is representative for the corresponding regions as shown in the optical microscopic image in (D). The SEM EBSD images in Takata et al.22 in (A) and (E) are used to provide grain morphology and grain orientations. MATLAB scripts are developed to extract the information as shown in (B) and (F), and finally geometries are modeled as shown in (C) and (G) with assigned orientation for each grain using developed ABAQUS Python scripts.

the current models, no fracture is considered, and strain rate is set to quasistatic in the crystal plasticity constitutive laws. Thus, mesh convergency has not been a major concern. Single crystal-type crystal plasticity models are used as the constitutive laws for the grains in the microscale CPFEM models. A simple UMAT subroutine, available online,31,32 is used to define the material properties in the ABAQUS model. Grain orientations are assigned as the input parameters of the UMAT for each grain. As typical solutions for Al alloys, all the twelve < 110 > f111g slip systems in a-Al are activated. Hyperbolic secant hardening law33 is used. The hardening coefficients of all the slip systems are assumed to be the same. The three parameters for the .

207

208

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

hardening law, namely, initial hardening modulus (h0), critical resolved shear stress at initiation (s0), and saturated resolved shear stress (ss), are obtained through model calibration, which searches the optimal parameter values that can provide the best match between experiments in figure 3D and simulations for tensile loading along the build direction. For the mesoscale models, a MATLAB script is developed to generate the melt pool and melt pool boundaries’ geometries as overlapped concentric ellipses. The input parameters are either directly extracted from the SLM 500 machine (e.g., interval angle between layers and hatch spacing) or estimated based on the microscopic images (e.g., melt pool depth and width). The reconstructed mesostructures (fig. 3E and 3F) show visual similarities to the optical microscopic images (fig. 1F). Development of quantifiable descriptors to evaluate the morphology similarity can further improve the validity of the mesoscale microstructure reconstruction, which falls in the future scope of the authors. A quantitative comparison is planned in our future work. The thickness of the melt pool boundaries is measured as 5.0 lm from the optical microscopic images. An ABAQUS Python script is developed to

FIG. 3 Comparison between simulations at microscale (A–C) and mesoscale for loadings along in-plane and build directions (D–F). (A) Comparison between simulated engineering stress and strain of microscale models of melt pool and melt pool boundaries. The contours of nominal stress along build directions at 3% applied strain are shown in (B) and (C), respectively, for melt pool and melt pool boundaries. The simulated stress–strain responses for the mesoscale models loaded along build and in-plane directions are shown in (D), in which Z tensile experimental data are used for parameter calibration in CPFEM. The contours of tensile strain in the mesoscale models at 5% applied strain are shown in (E) and (F). The scales are aligned for (B) and (C) and for (E) and (F).

.

LI ET AL., DOI: 10.1520/STP163120190135

automatically draw the geometry, to perform trimming, and it assigns each of the melt pool and melt pool boundary regions based on the reconstructed mesostructure morphology. Similar to the script for microscale models, the mesoscale script ensures the geometric features retain in ABAQUS during meshing and thus the generated mesh follows the boundaries of different regions.

Results and Discussions RESULTS OF MICROSTRUCTURE CHARACTERIZATION AND MECHANICAL TESTING

The summarized elongation and ultimate tensile strength (UTS) from the quasistatic tensile testing are shown in table 1. Interestingly, it is found that the horizontal samples have slightly lower UTS compared to the vertical samples, though the anisotropy is not significant. To calibrate the parameters in the CPFEM model, one experimental stress–strain curve of a vertical sample is used for training, and another experimental stress–strain curve of the horizontal sample is used for validation. Because modeling of defects is not within the scope of this study, samples with relatively large elongation are picked in this process as they are less likely to be affected by the defects. From the SEM images shown in figure 1B and 1C, the characteristic dimensions of the cells are roughly estimated as 0.5 and 1.0 lm, respectively, for the melt pool and melt pool boundaries. These values are then used to estimate the difference in the hardness parameters of slip systems, which is detailed in the next section of text. A more rigorous quantitative measurement requires more SEM images as well as development on the image analysis scripts in MATLAB, which is in the scope of the future work. The width of the melt pool boundaries is measured as 5.0 lm, which is averaged from the measurement of 13 melt pool boundaries in an optical image with higher magnification than figure 2D. CRYSTAL PLASTICITY PARAMETERS AND COMPARISON BETWEEN PREDICTION AND TESTING

The melt pool and melt pool boundary are differentiated in the proposed model. It has been reported that the dislocations can be trapped by the subgranular cellular structure for as-built selective laser melting (SLM) AlSi10Mg23,24 and thus provides Hall-Petch strengthening to the material. While grain boundaries are also contributing to the Hall-Petch strengthening as in the conventionally manufactured alloys,

TABLE 1 Obtained best-fit parameters for CPFEM

Melt Pool

h0,MP (MPa) 600

.

Melt Pool Boundary

s0, MP (MPa)

ss, MP (MPa)

h0, MB (MPa)

s0, MPB (MPa)

ss, MPB (MPa)

77.5

125.0

700

72.5

146.3

209

210

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

the size of the cell is far smaller and thus overwhelms the effect of the grain boundary. Assuming that all the rest of the strengthening mechanisms stay the same for melt pool and melt pool boundaries, the contribution of the Hall-Petch effect is the sole source of difference for the yield strength of these two regions. Thus, as shown in equation (1):   1 ffi 1 (1) DrY ¼ rYMP  rYMPB ¼ k pffiffiffiffiffi  pffiffiffiffiffiffiffi d d MP

MPB

where: DrY is the difference between the yield strength of melt pool, rYMP , and yield strength of melt pool boundaries, rYMPB ; k is a material constant in Hall-Petch equations, which is taken as 0.04 MPa-m0.5 according to Hadadzadeh et al.,23 and dMP and dMPB are the characteristic dimensions of the cellular structure in melt pool and melt pool boundaries and estimated as 0.5 lm and 1.0 lm, respectively. Thus, DrY is approximately 15.6 MPa. With the typical Taylor factor of 3.1 for Al alloys, the difference in terms of critical resolved shear stress for each slip system is then 5.0 MPa. With this value of DrY as constraint, an optimization process is developed in modeFrontier to reverse engineer the rest of the hardening parameters for the CPFEM models at microscale. When the optimization is terminated, approximately 250 sets of parameters are generated using the pilOPT algorithm in modeFrontier, and the one that can provide the best match (R2 ¼ 0.9988) between the training set stress–strain curve for the selected vertical sample and the mesoscale prediction is listed in table 1. Following the best-fit parameters, the stress–strain responses in the melt pool are shown in figure 3A. Figure 3B further compares the contour of the nominal stress component at 3% of applied strain for both microscale models, where some grains in the melt pool regions are showing clear stress intensifications. The comparisons of simulation and experimental stress–strain for both the training set and the validation set are shown in figure 3D. For the validation set, the loading direction in the microscale model is changed to the horizontal direction to provide the homogenized plastic stress–strain for the mesoscale model. For the validation data, the predicted stress–strain response matches well with the experiments, as shown in figure 3D. Consistent with the observations in the experiments, slight anisotropy is observed for the predictions of stress–strain when loaded along different directions. However, the trend that higher UTS is found for vertical samples is also captured by the simulations, as shown in the inlet figure in figure 3D. Comparison of the strain contour as shown in figure 3E and 3F indicates that the strain in the melt pool boundaries is generally higher than in the melt pools, yet the exact quantity of such a difference seems to vary between horizontal and vertical samples, which might be an indication that the mesoscale microstructure also contributes to the slight anisotropy, though further analysis is required before a firm conclusion can be drawn. This may possibly shed light on explaining the different failure mechanisms when .

LI ET AL., DOI: 10.1520/STP163120190135

samples are loaded along different directions.26,27 Such analysis will be much more valuable in other more anisotropic alloys (e.g., stainless steel 304L and 316L), which is also the interest of the authors. It is noted that a homogenized model without distinguishing the difference in microstructure and material properties of melt pool and melt pool boundaries would not be able to predict the local strain intensification, which possibly sheds light on predicting the different failure mechanisms when the samples are subjected to loading along different directions. LIMITATIONS OF THE CURRENT WORK AND PLANNED IMPROVEMENT

While the multiscale material modeling framework is established in the present study, the authors would like to outline some of the planned improvements to the current work. For example, direct measurement of the local hardness in the melt pool and melt pool boundaries can provide direct validation for the microscale simulations. Moreover, a number of the microscale melt pool models should be reconstructed based on more EBSD data to obtain the more representative properties of the melt pools. To improve the mesoscale microstructure reconstruction process, development of descriptors that can quantitatively measure the similarity between generated mesoscale microstructures and those captured from a microscope would also be of great value to the completeness of this work. Last but not the least important, uncertainty quantifications based on sufficient amount of testing and characterization data would make the developed model more useful in the real-life process and material design practices.

Conclusions A multiscale material modeling framework is presented in this study, with specific focuses on considering the melt pool and melt pool boundaries as distinct constituents, based on the observed microstructural differences. The advantage of the methodology is that the linkage between the microstructure characterized across different length scales is integrated during the property prediction, which provides improved insights when modeling the process-microstructure-property relationships for PBF alloys. Preliminary results indicate that the modeling procedure can predict the work hardening and slight anisotropy observed in the testing of fabricated AlSi10Mg samples, yet further investigations are required for completeness of the study. ACKNOWLEDGMENTS

The authors are grateful for the assistance and insightful discussion from colleagues at Ford Motor Company, specifically Dr. Joy Forsmark, Dr. Yang Huo, Ms. Emily Wolbeck and Mr. James Gavulic.

References 1.

.

W. E. Frazier, “Metal Additive Manufacturing: A Review,” Journal of Materials Engineering and Performance 23, no. 6 (2014): 1917–1928.

211

212

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

2.

3. 4. 5.

6. 7. 8.

9.

10.

11. 12.

13.

14.

15.

16.

17.

18.

19.

.

T. D. Ngo, A. Kashani, G. Imbalzano, K. Nguyen, and D. Hui, “Additive Manufacturing (3D Printing): A Review of Materials, Methods, Applications and Challenges,” Composites Part B: Engineering 143 (2018): 172–196. I. Gibson, D. W. Rosen, and B. Stucker, Additive Manufacturing Technologies, Vol. 17 (New York: Springer, 2014). J. J. Lewandowski and M. Seifi, “Metal Additive Manufacturing: A Review of Mechanical Properties,” Annual Review of Materials Research 46 (2016): 151–186. L. Zhou, A. Mehta, E. Schulz, B. McWilliams, K. Cho, and Y. Sohn, “Microstructure, Precipitates and Hardness of Selectively Laser Melted AlSi10Mg Alloy before and after Heat Treatment,” Materials Characterization 143 (2018): 5–17. J. Allison, M. Li, C. Wolverton, and X. Su, “Virtual Aluminum Castings: An Industrial Application of ICME,” JOM 58, no. 11 (2006): 28–35. J. Allison, “Integrated Computational Materials Engineering: A Perspective on Progress and Future Steps,” JOM 63, no. 4 (2011): 15–18. S. J. Wolff, S. Lin, E. J. Faierson, W. K. Liu, G. J. Wagner, and J. Cao. “A Framework to Link Localized Cooling and Properties of Directed Energy Deposition (DED)-Processed Ti-6Al-4V,” Acta Materialia 132 (2017): 106–117. W. Yan, J. Smith, W. Ge, F. Lin, and W. K. Liu. “Multiscale Modeling of Electron Beam and Substrate Interaction: A New Heat Source Model,” Computational Mechanics 56, no. 2 (2015): 265–276. Y. Lian, S. Lin, W. Yan, W. K. Liu, and G. J. Wagner, “A Parallelized Three-Dimensional Cellular Automaton Model for Grain Growth during Additive Manufacturing,” Computational Mechanics 61, no. 5 (2018): 543–558. A. Rai, H. Helmer, and C. Ko ¨rner, “Simulation of Grain Structure Evolution during Powder Bed Based Additive Manufacturing,” Additive Manufacturing 13 (2017): 124–134. D. Kim, J. Hwang, E. Kim, Y. Heo, W. Woo, and S. Choi, “Evaluation of the Stress-Strain Relationship of Constituent Phases in AlSi10Mg Alloy Produced by Selective Laser Melting Using Crystal Plasticity FEM,” Journal of Alloys and Compounds 714 (2017): 687–697. K. Kapoor, Y. S. Yoo, T. A. Book, J. P. Kacher, and M. D. Sangid, “Incorporating GrainLevel Residual Stresses and Validating a Crystal Plasticity Model of a Two-Phase Ti-6Al4V Alloy Produced via Additive Manufacturing,” Journal of the Mechanics and Physics of Solids 121 (2018): 447–462. V. Prithivirajan and M. D. Sangid, “The Role of Defects and Critical Pore Size Analysis in the Fatigue Response of Additively Manufactured IN718 via Crystal Plasticity,” Materials & Design 150 (2018): 139–153. M. T. Andani, M. R. Karamooz-Ravari, R. Mirzaeifar, and J. Ni, “Micromechanics Modeling of Metallic Alloys 3D Printed by Selective Laser Melting,” Materials & Design 137 (2018): 204–213. S. A. Khairallah, A. T. Anderson, A. Rubenchik, and W. E. King, “Laser Powder-Bed Fusion Additive Manufacturing: Physics of Complex Melt Flow and Formation Mechanisms of Pores, Spatter, and Denudation Zones,” Acta Materialia 108 (2016): 36–45. E. R. Denlinger, M. Gouge, J. Irwin, and P. Michaleris, “Thermomechanical Model Development and In Situ Experimental Validation of the Laser Powder-Bed Fusion Process,” Additive Manufacturing 16 (2017): 73–80. Q. Chen, X. Liang, D. Hayduke, J. Liu, L. Cheng, J. Oskin, R. Whitmore, and A. C. To, “An Inherent Strain Based Multiscale Modeling Framework for Simulating Part-Scale Residual Deformation for Direct Metal Laser Sintering,” Additive Manufacturing 28 (2019): 406–418. J. Smith, W. Xiong, W. Yan, S. Lin, P. Cheng, O. L. Kafka, G. J. Wagner, J. Cao, and W. K. Liu, “Linking Process, Structure, Property, and Performance for Metal-Based Additive Manufacturing: Computational Approaches with Experimental Support,” Computational Mechanics 57, no. 4 (2016): 583–610.

LI ET AL., DOI: 10.1520/STP163120190135

20.

21.

22.

23.

24.

25. 26.

27.

28.

29.

30.

31.

32.

33.

.

W. E. King, A. T. Anderson, R. M. Ferencz, N. E. Hodge, C. Kamath, and S. A. Khairallah, “Overview of Modelling and Simulation of Metal Powder Bed Fusion Process at Lawrence Livermore National Laboratory,” Materials Science and Technology 31, no. 8 (2015): 957–968. R. Martukanitz, P. Michaleris, T. Palmer, T. DebRoy, Z. Liu, R. Otis, T. W. Heo, and L. Chen, “Toward an Integrated Computational System for Describing the Additive Manufacturing Process for Metallic Materials,” Additive Manufacturing 1 (2014): 52–63. N. Takata, H. Kodaira, K. Sekizawa, A. Suzuki, and M. Kobashi, “Change in Microstructure of Selectively Laser Melted AlSi10Mg Alloy with Heat Treatments” Materials Science and Engineering: A 704 (2017): 218–228. A. Hadadzadeh, B. S. Amirkhiz, A. Odeshi, J. Li, and M. Mohammadi, “Role of Hierarchical Microstructure of Additively Manufactured AlSi10Mg on Dynamic Loading Behavior,” Additive Manufacturing 28 (2019): 1–13. A. Hadadzadeh, B. S. Amirkhiz, and M. Mohammadi, “Contribution of Mg2Si Precipitates to the Strength of Direct Metal Laser Sintered AlSi10Mg,” Materials Science and Engineering: A 739 (2019): 295–300. X. Liu, C. Zhao, X. Zhou, Z. Shen, and W. Liu, “Microstructure of Selective Laser Melted AlSi10Mg Alloy,” Materials & Design 168 (2019): 107677. Z. H. Xiong, S. L. Liu, S. F. Li, Y. Shi, Y. F. Yang, and R. D. K. Misra, “Role Of Melt Pool Boundary Condition in Determining the Mechanical Properties of Selective Laser Melting AlSi10Mg Alloy,” Materials Science and Engineering: A 740 (2019): 148–156. D. Dai, D. Gu, H. Zhang, J. Xiong, C. Ma, C. Hong, and R. Poprawe, “Influence of Scan Strategy and Molten Pool Configuration on Microstructures and Tensile Properties of Selective Laser Melting Additive Manufactured Aluminum Based Parts,” Optics & Laser Technology 99 (2018): 91–100. L. Thijs, K. Kempen, J. Kruth, and J. Van Humbeeck, “Fine-Structured Aluminium Products with Controllable Texture by Selective Laser Melting of Pre-Alloyed AlSi10Mg Powder,” Acta Materialia 61, no. 5 (2013): 1809–1819. M. Ghasri-Khouzani, M. H. Peng, R. Attardo, P. Ostiguy, J. Neidig, R. Billo, D. Hoelzle, and M. R. Shankar, “Comparing Microstructure and Hardness of Direct Metal Laser Sintered AlSi10Mg Alloy between Different Planes,” Journal of Manufacturing Processes 37 (2019): 274–280. Standard Test Methods for Tension Testing of Metallic Materials, ASTM E8/E8M-16a (West Conshohocken, PA: ASTM International, approved August 1, 2016), https:// doi.org/10.1520/E0008_E0008M-16A Y. Huang, A User-Material Subroutine Incorporating Single Crystal Plasticity in the ABAQUS Finite Element Program, Mech. Report 178 (Cambridge, MA: Harvard University, Division of Engineering and Applied Sciences, 1991). J. W. Kysar, Addendum to “A User-Material Subroutine Incorporating Single Crystal Plasticity in the ABAQUS Finite Element Program, Mech. Report 178” (Cambridge, MA: Harvard University, Division of Engineering and Applied Sciences, 1997). D. Peirce, R. J. Asaro, and A. Needleman, “Material Rate Dependence and Localized Deformation in Crystalline Solids,” Acta Metallurgica 31, no. 12 (1983): 1951–1976.

213

214

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190144

Chul Y. Park,1 Keith E. Rupel,2 Chelsey E. Henry,3 Kevin F. Malik,1 Sayata Ghose,1 and Upul R. Palliyaguru4

Alternate Method for Determining Yield Strength of Polymer Additive Manufacturing Citation C. Y. Park, K. E. Rupel, C. E. Henry, K. F. Malik, S. Ghose, and U. R. Palliyaguru, “Alternate Method for Determining Yield Strength of Polymer Additive Manufacturing,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 214–228. http://doi.org/10.1520/ STP1631201901445

ABSTRACT

Additive manufacturing (AM) technology offers potential cost savings, production efficiencies, and unique design features that cannot be easily achieved through traditional manufacturing. However, it is known that the final performance of an additively manufactured product can be influenced by various stages of the AM process. For example, the mechanical performance of an additively manufactured product may be influenced by build orientation, location within the processing equipment, and the process parameters utilized in fabrication. In order to understand the level of variation of the mechanical performance, multiple coupon test types are currently used during both AM process development and production of AM parts. Among those, tensile coupon testing is most commonly used. This paper documents the result of a recent investigation on tensile yield strength data reduction methods.

Manuscript received November 7, 2019; accepted for publication January 21, 2020. 1 Boeing Commercial Airplanes, The Boeing Company, P.O. Box 3707 MC 082-51, Seattle, WA 98124-2207, USA, C. Y. P. http://orcid.org/0000-0001-6588-5178 2 Boeing Research & Technology (Retired) 3 Boeing Research & Technology, The Boeing Company, 10026 Prosperity Way, West Jordan, UT 84081, USA 4 National Institute for Aviation Research, Wichita State University, 1845 Fairmount St., Wichita, KS 672600093, USA 5 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

PARK ET AL., DOI: 10.1520/STP163120190144

Keywords mechanical test, coupon test, offset method, yield strength, ultimate strength, data reduction method, true yield, additive manufacturing, ASTM D638

Introduction Coupon level mechanical testing plays an important role in additive manufacturing (AM) during process development, establishment of material and processing specification requirements, material batch acceptance, process quality control, and development of design values or allowables. Tensile coupon testing in accordance with ASTM D638, Standard Test Method for Tensile Properties of Plastics,1 is widely used for polymer AM. The test method works reasonably well for AM processes with some modifications. Investigation on the common challenges of using the ASTM D638 coupon design and alternative coupon designs are documented in a separate paper.2 This paper will focus on the challenges associated with the data reduction aspect of tensile coupon testing.

Motivation of Research The purpose of this research was to investigate an alternative method to more accurately determine the yield strength or strain that truly represent the onset of permanent deformation. During internal development of polymer AM technology, it was discovered that a traditional offset-based yield strength determination method can result in inaccurate yield strength results that may not represent the true permanent deformation. This can be more pronounced for ductile material, especially when tested under elevated temperature. Figure 1 shows an example of a Nylon 11 tensile curve with the 0.2% yield offset line projected on the curve. The alternative yield test method chosen for investigation is based on the work conducted by Datapoint Labs.3 As part of this investigation, Nylon 11 material produced using selective laser sintering (SLS) technology was chosen. It should be noted that the intent of this investigation is to focus on a general data reduction method that can potentially be used for other materials.

Research Approach In order to obtain more accurate yield strength data and evaluate the method proposed by Datapoint Labs,3 a three-step approach was developed: (1) baseline tensile test, (2) multiload tensile test, and (3) yield point check test. The multiload test formed the basis of determining the alternative yield strength data. Baseline tensile test results were intended to provide a reference material behavior. Yield point check test results were used to verify that the overall yield behavior of the multiload and the single-load tests were equivalent. .

215

216

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 An example of a tensile curve for Nylon 11 with a 0.2% offset projected (room temperature ambient).

8000 7000

Stress (psi)

6000 5000 4000 3000 2000 1000 0 0.0

4.0

8.0

12.0

16.0

20.0

24.0

Strain (%)

The ASTM D638 Type I specimen design was chosen for the investigation except for the addition of label tabs. This label tab has been shown to eliminate failures in the grip region associated with the commonly used engraving in the grip area. A typical specimen is shown in figure 2. Two specimen fabrication methods were used—to print to the net shape and to machine the printed blank to the final dimension. Selected specimens were machined with a standard 0.25-in.-diameter end mill at 25 in. per minute (ipm) and 12,000 revolutions per minute (rpm). Double-sided tape was used to fasten the specimens to the base. Air was the only coolant used. For the research, all test specimens were produced to have a ZY orientation (fig. 3). This orientation was chosen to account for Z directional mechanical performance and maximize the number of specimens produced. All specimens were produced from three builds in accordance with Boeing internal material and process specifications. The specimens were located in a nonserial fashion in the build as part of a randomization effort to prevent an entire test grouping from coming from

FIG. 2 Test specimen geometry (refer to ASTM D638 for specimen geometries).

.

PARK ET AL., DOI: 10.1520/STP163120190144

FIG. 3 Specimen build orientation and three-dimensional view of build.

one particular build area and possibly influencing the results. Testing was also conducted in a random, nonsequential fashion to obtain further randomization. Specimen roughness was measured for both as-built and machined specimens. The as-built specimens had the surface roughness (Ra) of approximately 700 lin. and the machined specimens Ra of 38 lin. SPECIMEN CONDITIONING

Specimens used for the investigation were tested at room temperature, 165 F, and 215 F under ambient condition. Prior to testing, specimens were preconditioned a minimum of 40 h at room temperature ambient (RTA) under 50% relative humidity (RH) in accordance with ASTM D618 Procedure A. BASELINE TENSILE TEST

The baseline tensile tests were conducted using both as-built and machined configurations but with only a single replicate of each because it was not the main focus of the research effort given the availability of the historical data. Baseline specimens were loaded at typical ASTM D638 loading rates of 0.2 in./min for all RT and hightemperature test conditions. Displacement data were collected at 10 Hz with an extensometer. MULTILOAD TENSILE TEST

The multiload testing was performed by sequentially loading and unloading the specimens to progressively higher strain levels until the strain level was beyond where permanent deformation was observed. The purpose of this testing was to .

217

218

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

accurately determine the strain level at which permanent deformation began to occur and to measure the degree of permanent deformation at selected strain levels. The following procedure was used in multiload testing: • Stabilize specimens at the test temperature in the chamber prior to testing. • Load test specimen to first defined strain level. • Unload immediately (within less than 2 s) to 5 lb of applied tension. Hold crosshead for 2 min in order to allow the specimen to relax or recover. • Record residual stress and strain at the end of recovery period. • Continue testing for next strain level until specimen completes all strain levels. The intent of this multiload testing was to have several data points collected in each of the regions of the material response load curve: the elastic region, the transition region, and the full plastic region. YIELD POINT CHECK TEST

The yield point check test was designed to verify that multiple loading and unloading cycles did not cause additional permanent deformation compared to a direct loading to a higher strain level. To achieve this, yield point specimens were directly loaded to a strain level in the middle of the projected transition point between elastic and plastic responses so the resultant plastic deformation could be compared against plastic deformation results from multiload testing at the equivalent strain loadings. Target strain levels for the yield point tests at the environmental conditions were between 3% and 6% strain. After loading to the desired applied strain levels and unloading to 5 lb applied loads, specimens were held for 2 min at a constant crosshead position in the same manner as the multiload specimens. The applied load and the extensometer response were recorded through the hold. The following procedure was used for the yield point check test: • Verify the strain level to use for testing: Plot the residual strain versus applied strain results from the multiload testing. Fit a line to the elastic portion of the curve (low strain levels) to verify the strain level where the response fully departs from an elastic response. • Load specimen directly to the defined strain level for yield point check testing. • Unload immediately (within less than 2 s) to 5 lb of applied tension. Hold crosshead for 2 min. • Record residual stress and strain at the end of recovery period.

Results Table 1 summarizes the testing performed for the investigation. Baseline tensile

curves were produced for each test condition followed by multiload testing. Finally, a yield point check test was performed in order to determine if multiload testing and single-loading testing produce equivalent behavior. .

PARK ET AL., DOI: 10.1520/STP163120190144

TABLE 1 Test matrix for yield strength investigation

Number of Specimens Specimen Fabrication

RTA

165 F/ Ambient

215 F/ Ambient

Machined

1

1

1

As-Built

1

1

1

Multiload test

Machined

3

3

3

As-Built

3

2

3

Yield point check

Machined

2

2

2

As-Built

1

2

1

Test Grouping

Baseline tensile test

BASELINE TEST RESULTS

The baseline tensile test established a modulus and an offset yield point to be later compared with the yield strain level determined from the multiload test. Selection of the chord points for the modulus was one of the issues in reducing the baseline test data. Individual baseline ZY tensile results are summarized in table 2. Figures 4, 5, and 6 show the stress strain plots of RTA, 165 F/ambient, and  215 F/ambient, with the 0.2% offset intercepts. Both plots have an expanded view of the low strain region. Key findings are documented here. • There is a significant toe region that would make the use of the typical 0.1% and 0.3% chord modulus method difficult to apply. • The curve is not linear above the toe region. • The 0.2% strain offset is small in proportion to the overall material strain response. The higher temperature data produced additional issues in the low strain region that make it more difficult to select the chord modulus points. Table 2 describes the regions used to more accurately calculate the modulus of each testing. Data quality at the lower strain levels does not impact ultimate stress calculations, but it does

TABLE 2 Baseline test results

Test Condition

RT/ Ambient 165 F/ Ambient 215 F/ Ambient a

6.72

24%

4.09

7.67

40%

4.35

34.0

3.88

> 40%a

1.69

38.0

4.50

> 40%a

1.80

35.1

3.07

> 30%a

1.38

41.4

3.45

>40%a

1.36

Surface Condition

1st Chord Strain (%)

2nd Chord Strain (%)

Modulus (ksi)

As-built

1.5%

1.8%

147.6 155.1

Machined As-built

2.0%

2.3%

Machined As-built

2.0%

2.3%

Machined

Test terminated prior to specimen failure.

.

Elongation (%)

0.2% Offset Yield Strength (ksi)

Ultimate Strength (ksi)

219

220

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 4 Modulus behaviors at low strain levels on RTA as-built specimen. Overall behavior (left) and behavior in highlighted region (right).

FIG. 5 Modulus behaviors at low strain levels on 165 F ambient as-built specimen. Overall behavior (left) and behavior in highlighted region (right).

FIG. 6 Modulus behaviors at low strain levels on 215 F/ambient as-built specimen. Overall behavior (left) and behavior in highlighted region (right).

.

PARK ET AL., DOI: 10.1520/STP163120190144

add variability to modulus calculations and the subsequent offset yield calculations based on the modulus determination for an individual test. Figures 7, 8, and 9 show stress–strain responses of as-built and machined specimens. The percentage curve represents the stiffness comparison between machined and as-built specimens. As can be seen, the stiffness of the machined specimen is noticeably higher than that of the as-built specimen. One of the contributing factors is inaccurate calculation of the cross-sectional area of the as-built specimen. This was documented in another paper.2 This trend was seen on all test conditions. MULTILOAD TEST RESULTS

Utilizing the measured test specimen width and thickness, load was converted to stress at each data collection point as per standard practice. The following steps were used for data reduction:

FIG. 7 RT/ambient baseline tensile stress strain curves. Overall behavior (left) and behavior in highlighted region (right).

FIG. 8 165 F/ambient baseline tensile stress strain curves. Overall behavior (left) and behavior in highlighted region (right).

.

221

222

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 9 215 F/ambient baseline tensile stress strain curves. Overall behavior (left) and behavior in highlighted region (right).

• •

•

•

•

•

.

Record residual stress and strain at the end of each unload hold. Project residual strain at zero stress at the end of each unload hold: Specimen elastic modulus was calculated for each specimen from the first loading cycle, and this elastic modulus was utilized to project from the residual strain after each loading level to what the residual strain would have been if the loading was reduced to zero. Residual strain versus applied strain was then plotted for each specimen with the goal of obtaining the strain response if the specimen was fully unloaded at the end of each progressively higher loading and unloading cycle. Curve fit residual strain versus applied strain curve in elastic and plastic response regions: A reasonable curve fit was obtained without using an excessively high order polynomial by limiting the region of the curve fit to the regions of interest. Calculate yield strain: To calculate the plastic deformation or yield strain, first an acceptable amount of plastic deformation or strain must be selected (typically from 0.05% to 0.5%). The applied strain at which the difference between the actual residual strain response and the projected elastic linear response is equal to the defined acceptable plastic strain at yield is then the yield strain. The yield strain results were determined for each test condition. No conclusion has been reached that would be an appropriate value to utilize for the permitted amount of plastic deformation. The 0.05% strain value was selected as the minimum plastic deformation level used to calculate yield because 0.05% plastic deformation from the linear behavior appeared to be where values repeatedly rose above the noise in the output plot. The maximum strain level of 0.5% of permitted plastic deformation was selected due to its being roughly twice the standard 0.2% offset value used in traditional offset yield calculations.

PARK ET AL., DOI: 10.1520/STP163120190144

The yield testing results via the true yield method for the multiload tensile testing method are provided in tables 3, 4, and 5, with selected plots shown in figures 10, 11, and 12.

TABLE 3 Multiload yield result summary of RTA testing

Yield Strain Plastic Strain Used to Define Yield Strain Surface Condition

Specimen ID

0.05%

0.1%

0.2%

0.3%

0.4%

0.5%

As-built

#1

2.55

2.95

3.45

3.80

4.05

4.30

#2

2.30

2.60

3.00

3.30

3.60

3.85

#3

2.50

2.95

3.45

3.75

4.00

4.20

#1

2.45

2.70

3.10

3.35

3.65

3.90

#2

2.60

2.85

3.20

3.45

3.75

4.05

#3

2.45

2.70

3.10

3.40

3.65

3.95

Average

2.48

2.79

3.22

3.51

3.78

4.04

0.3%

0.4%

0.5%

Machined

TABLE 4 Multiload yield result summary of 165 F/ambient testing

Yield Strain Plastic Strain Used to Define Yield Strain Surface Condition

As-built

Machined

Specimen ID

0.05%

0.1%

0.2%

#1

2.20

2.95

3.90

4.55

5.00

5.45

#2

2.55

3.25

4.25

4.95

5.45

5.50

#1

2.90

3.55

4.35

5.00

5.50

5.50

#2

2.55

3.25

4.15

4.80

5.30

5.50

#3

2.45

3.10

4.00

4.65

5.25

5.70

Average

2.53

3.22

4.13

4.79

5.30

5.53

TABLE 5 Multiload yield result summary of 215 F/ambient testing

Yield Strain Plastic Strain Used to Define Yield Strain Surface Condition

Specimen ID

0.05%

0.1%

0.2%

0.3%

0.4%

As-built

#1

2.25

2.85

3.65

4.25

4.70

5.15

#2

2.95

3.50

4.30

4.90

5.40

5.50

#3

2.50

3.05

3.80

4.40

4.85

5.25

#1

2.75

3.30

4.05

4.55

5.00

5.40

Machined

.

0.5%

#2

2.75

3.30

4.05

4.65

5.10

5.50

#3

2.60

3.20

3.95

4.55

5.05

5.50

Average

2.63

3.20

3.97

4.55

5.02

5.38

223

224

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 10 Yield behavior at RTA based on multiload tests on as-built (top) and machined specimen (bottom).

YIELD CHECK TEST RESULTS

The RTA, 165 F ambient, and 215 F ambient yield point checks consisted of three or four tests and were loaded to 4% or 6% applied strain. The 4% and 6% applied strains were those in the middle of the projected transition point between elastic and plastic responses. The results are summarized in tables 6 through 9. The residual strains with the single-loading tests produced residual strains similar to the multiload results at the same applied strain levels, indicating the use of multiple loading cycles was not biasing the test results.

.

PARK ET AL., DOI: 10.1520/STP163120190144

FIG. 11 Yield behavior at 165 F ambient based on multiload tests on as-built (top) and machined specimen (bottom).

Summary and Conclusion True yield testing was investigated as a possible alternative or complementary test to 0.2% offset yield testing to determine more accurate yield strength values. Previous testing has shown high-strain, low-modulus plastic materials such as SLS Nylon 11 at elevated temperatures produce large variations in yield strength that may be more due to the data reduction method than the material itself. Compared to the traditional offset method, the true yield data reduction method seemed to produce more accurate yield strength data for SLS Nylon 11 .

225

226

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 12 Yield behavior at 215 F ambient based on multiload tests on as-built (top) and machined specimen (bottom).

material. This test method can be used to define the resultant yield strain using a limited number of tests. Main material testing may then be followed in larger quantities using multiple material batches to develop statistically significant yield strength. The purpose of this paper was to provide initial data to support the consideration of an alternate yield strength data reduction method. A more thorough investigation including round-robin testing involving different material types should be further carried out. .

PARK ET AL., DOI: 10.1520/STP163120190144

TABLE 6 RTA yield point check test results (4%)

Projected Residual Strain at 4% Applied Strain Specimen Type

As-built

Multiload Test

Single-Load Yield Check

0.42%

0.35%

0.56%

–

0.41%

–

TABLE 7 RTA yield point check test results (6%)

Projected Residual Strain at 6% Applied Strain Specimen Type

Machined

Multiload Test

Single-Load Yield Check

1.54%

1.41%

1.42%

1.44%

1.42%

–

TABLE 8 165 F/ambient yield point check test results (4%)

Projected Residual Strain at 4% Applied Strain Specimen Type

As-built

Machined

Multiload Test

Single-Load Yield Check

0.27%

0.32%

0.28%

0.29%

0.23%

0.17%

0.15%

0.13%

0.17%

–

TABLE 9 215 F/ambient yield point check test results (6%)

Projected Residual Strain at 6% Applied Strain Specimen Type

As-built

Machined

.

Multiload Test

Single-Load Yield Check

0.75%

0.63%

0.83%

–

0.66%

–

0.65%

0.58%

0.67%

0.52%

0.61%

–

227

228

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

ACKNOWLEDGMENTS

The data presented here are produced by Boeing and the National Institute for Aviation Research (NIAR) under NASA Cooperative Agreement NNL09AA00A, Work Activity 2C27. Disclosure, use, and duplication of the data are governed by NASA Cooperative Agreement NNL09AA00A, Activity 2A38, Exhibit B, 1(b)(5). The authors would like to thank Brian Grimsley and Frank Palmieri of NASA Langley Research Center for supporting this project through the Advanced Composites Project (ACP) program.

References 1.

2.

3.

.

Standard Test Method for Tensile Properties of Plastics, ASTM D638-14 (West Conshohocken, PA: ASTM International, approved December 15, 2014), https://doi.org/10.1520/ D0638-14 C. Y. Park, K. E. Rupel, C. E. Henry, and U. R. Palliyaguru, “Coupon Level Mechanical Testing for Polymer-Based Additive Manufacturing” in Structural Integrity of Additive Manufactured Parts, ed. N. Shamsaei, S. Daniewicz, N. Hrabe, S. Beretta, J. Waller, and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 355–366, https://doi.org/10.1520/ STP162020180102 Francis Barthelat, “Measuring True Yield Point of Plastics,” Datapoint 7.1 (2001): 1–3.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190127

Wei-Jen Lai,1 Ziang Li,1 Avinesh Ojha,1 Yang Li,1 Joy Forsmark,1 Carlos Engler,1 and Xuming Su1

Effects of Surface Roughness and Porosity on Fatigue Behavior of AlSi10Mg Produced by Laser Powder Bed Fusion Process Citation W.-J. Lai, Z. Li, A. Ojha, Y. Li, J. Forsmark, C. Engler, and X. Su, “Effects of Surface Roughness and Porosity on Fatigue Behavior of AlSi10Mg Produced by Laser Powder Bed Fusion Process,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 229–246. http://doi.org/ 10.1520/STP1631201901272

ABSTRACT

Effects of as-built surface roughness and porosity on fatigue performance were investigated for AlSi10Mg produced by the laser powder bed fusion (L-PBF) process. As-built (with original surface roughness) and polished (surface roughness removed) samples were used to study the effect of surface roughness on fatigue strength. Results indicate that roughness has a significant effect on fatigue strength (approximately 40% decrease), and the current empirical model used for conventional materials significantly underestimates the roughness effect. The fatigue strengths in border and hatch regions were studied using polished samples and machined and polished samples. The machined and polished samples show a phenomenal fatigue strength of 154 MPa, which is 50% more compared to the polished-only samples (96 MPa). The samples were then subjected to homogenization to remove residual stress and ensure identical microstructure. The homogenized samples show no difference in fatigue strength in the border and hatch regions. The results suggest that residual stress

Manuscript received October 30, 2019; accepted for publication February 14, 2020. 1 Ford Motor Co., 1 America Rd., Dearborn, MI 48121, USA W.-J. L. http://orcid.org/0000-0002-1508-3019, http://orcid.org/0000-0001-5131-012X, A. O. http://orcid.org/0000-0001-8494-5666, Z. L. Y. L. http://orcid.org/0000-0003-1461-5623, C. E. http://orcid.org/0000-0001-7205-7223 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

229

230

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

might play a role in decreasing fatigue strength in the border region. Different porosity levels were produced to study the effect of defects on fatigue strength. The results show that the modified Murakami’s model can correctly predict the fatigue strength based on defect size. Keywords fatigue, roughness effect, AlSi10Mg, defect

Introduction Metal additive manufacturing (AM) has drawn significant attention in recent years due to its great design freedom for parts that used to be impossible or difficult to make. The high cost and long production time as compared to conventional manufacturing processes are its main disadvantages. However, this process still finds many applications when it comes to low production volume and highcomplexity parts. AlSi10Mg is one of the first few aluminum alloys available for laser-powder bed fusion (L-PBF), and its introduction has caught the automotive industry’s attention as many powertrain components are made of aluminum alloys. The use of AM could potentially achieve more complex designs to increase cooling and combustion efficiency in complex engine parts such as cylinder heads and engine blocks. Due to rapid cooling, AlSi10Mg produced using L-PBF has a very high yield strength and ultimate tensile strength, as compared to cast properties using the same alloy system. However, ductility tends to be lower. In addition, the as-built surface roughness that is inherent in the process could also result in lower fatigue properties because of higher levels of stress concentrations and defects at the surface than would be observed in a cast or wrought surface of the same materials. The effect of surface roughness on fatigue behavior of L-PBF metallic materials has been the focus of many studies.1–10 At least 40 to 50% reduction in fatigue strength was observed in samples with as-built surface condition in most literature. Yadollahi and Shamsaei11 reported no difference in L-PBF Inconel 718 between as-built and machined conditions due to the higher stress concentration of surface defects than surface roughness. Most of the work was focused on the difference in fatigue behavior between the as-built surface and machined surface. The potential residual stress issue was mentioned in various studies1–3 if the specimens were not stress relieved. Unlike stainless steels or Ti-6Al-4V, stress-relieving L-PBF AlSi10Mg results in substantial microstructure change and a decrease in hardness. Therefore, the effect of roughness on fatigue strength in the nonheat-treated state should be fully understood. In this study, the fatigue strength of samples with an as-built surface, polished surface (removing approximately 100 lm layer), and a machined and polished surface were investigated. The effect of roughness on fatigue strength is usually modeled by empirical equations. The FKM method12 is one of the most popular models, and it is used in many commercial fatigue software programs. The model uses a roughness influence .

LAI ET AL., DOI: 10.1520/STP163120190127

factor (fR), which is the ratio of the fatigue strength with roughness to the reference fatigue strength (polished). The equation is shown in equation (1), where RZ,C and RZ;r are the measured surface roughness under the two conditions (as-built and polished, respectively), and rUTS is the ultimate tensile strength (UTS) of the material. The material constants (aR;r , rUTS;sf ;min ) are based on alloy systems and manufacturing methods. Other models such as IABG,13 developed by Siebel and Gaier, and TGL 19340,14 in GDR standards, both use similar formulations. However, they do not distinguish between different materials and are only functions of surface roughness and UTS. These models only consider the surface finish produced by conventional manufacturing methods. Their applicability on AM surface finish needs to be investigated.     1  aR;r  log RZ;C  log 2rUTS =rUTS;sf ;min     fR ¼ (1) 1  aR;r  log RZ;M  log 2rUTS =rUTS;sf ;min

Apart from the empirical models mentioned previously, some researchers found that the surface roughness can be treated as short cracks in a fatigue crack growth model for machined surfaces.15,16 Greitemeier et al.8 proposed a model based on equivalent initial flaw size for fatigue life estimation. Yadollahi et al.9 used a crack closure-based fatigue crack growth model to predict the fatigue life based on surface roughness parameters and found the maximum valley depth would be an appropriate size of initial surface flaw for predicting the fatigue life of AM parts in an as-built surface. These models all require crack growth data and complex constant fitting. Pegues et al.4 proposed a model relating several roughness parameters to effective fatigue notch factor and found good correlation. The method is relatively simple and should be validated by more test data. Defects such as porosity or inclusions are other key factors that control the fatigue strength of metal AM parts.17–20 The defect size distribution and morphology depend on laser process parameters, powder characteristics, and even environmental effects such as gas flow rate. The defect size distribution has been proven to correlate well to the fatigue strength by the Murakami fatigue strength model,21 as shown in equation (2) for surface crack: Dr ¼

1:43  ðHV þ 120Þ pffiffiffiffiffiffiffiffiffi 1=6 ð areaÞ

(2)

where: Dr is the fatigue strength stress amplitude at R ¼ 1, HV is the hardness of the specimen, and pffiffiffiffiffiffiffiffiffi area is the Murakami defect size parameter. However, Murakami calibrated the model constants using fatigue data mostly from steels. Only three data points are from a single wrought aluminum alloy (2017-T4). The model also overestimates these three points by about 7.2% to 14.1%. This .

231

232

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

indicates that the model constants might be material dependent. Another issue is that the model is calibrated using rotating bending fatigue data. Because the fatigue strength measured by rotating bending is always higher than the one measured by uniaxial fatigue testing,22 using the model can significantly overestimate the uniaxial fatigue strength. Ueno et al.23 proposed a new set of material constants for aluminum alloys. The equation was also validated by Tajiria et al.24 However, neither of these studies utilized data from L-PBF samples. Beretta and Romano,17 Romano et al.,18,19 and Romano, Miccoli, and Beretta20 published a series of papers using El-Haddad’s formulation with extreme value statistics to predict the fatigue strength of L-PBF AlSi10Mg. The results show that the fatigue strength is a strong function of defect size but is not sensitive to ultimate tensile strength nor yield strength. Romano et al.19 made the following statement based on the results and literature observation: “When the residual stresses are limited and samples are machined, defects are the main variable affecting the fatigue resistance.” However, Murakami’s formulation clearly indicates that fatigue strength is a function of both defect size and hardness (mechanical strength). AlSi10Mg is known to be sensitive to heat treatment, and the mechanical property can vary a lot depending on the heat treatment. Hence, the effect of heat treatment on fatigue strength needs to be fully understood. This paper presents the preliminary fatigue results to understand the effect of the surface roughness and defect size distribution. The effect of roughness will be evaluated and compared with the FKM guideline that is widely used for surface conditions produced by traditional manufacturing processes.

Experimental Procedure MATERIALS AND LASER PARAMETERS

The fatigue samples were manufactured using SLM 125 by SLM Solutions. All the samples were built with the tensile axis in the Z direction. The laser process parameters are summarized in table 1. An inside-out scan strategy was adopted, which

TABLE 1 Summary of laser parameters and post heat treatment of AlSi10Mg samples

Power (W)

Group 1

Hatch Spacing (mm)

Layer Thickness (mm)

Post Heat Treatment

0.03

NA

Border and contour

300

730

0.2

hatch

350

1,650

0.13

Border and contour

200

730

0.2

hatch

350

1,650

0.13

Group 3

Border and contour

200

730

0.2

hatch

350

1,650

0.13

Group 4

Border and contour

330

730

0.2

hatch

350

1,650

0.13

Group 2

.

Speed (mm/s)

NA 500 C/1 h 500 C/1 h

LAI ET AL., DOI: 10.1520/STP163120190127

hatches the center portion, and then a contour scan and border scan on the outer part to reduce the potential for porosity formation at the edge. Four groups of samples with different geometry and surface conditions were examined. Samples in the same group were produced using the same laser parameters. Group 1 and Group 2 have three sets of samples in each group, as shown in figure 1. The as-built (AB) sample was built to the final dimension without any postsurface treatment. The dimension follows ASTM E466-15, Standard Practice for Conducting Force Controlled Constant Amplitude Axial Fatigue Tests for Metallic Materials.25 The polished sample (P) is the same as the as-built sample except lowstress polishing was performed to remove the as-built surface roughness. Polishing was performed along the fatigue loading direction to ensure a minimal effect on fatigue strength. The machined and polished sample (MþP) was built as 15-mm cylindrical rods and then machined to the final fatigue sample dimension. The machined surface was also polished using the same method used for polished samples. Group 3 has two sets of samples, polished (P) and machined and polished (MþP). They were produced the same way as the ones in Group 2. The only difference is that the samples in Group 3 were homogenized at 500 C for 1 h to remove residual stress and microstructure inhomogeneity between the border scan and hatch scan. Group 4 has two sets of samples as well—polished (P) and machined

FIG. 1 Types of samples in each group.

.

233

234

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

and polished (MþP). The laser parameters of the border and contour scans were adjusted to intentionally produce keyhole porosity. The samples were also homogenized at 500 C for 1 h to ensure no residual stress or microstructure difference is present between the border scan and the hatch scan. Twelve samples were made for as-built, polished, and machined and polished samples for Group 1 (12  3) and Group 2 (12  3). Six samples were made for polished and machined and polished samples in Group 3 (6  2) and Group 4 (6  2). However, not all tests were successful, and results shown in figure 2 might have fewer points than planned. Due to the limited number of samples, the stress levels were focused on the ones near the fatigue strength. Metallographic samples were prepared using the cold mounting method to avoid heat treating the materials. Samples were polished and no etching was performed. The XY planes of the samples were examined and porosity was measured. The Vickers hardness was measured on the same samples using a 300-g force. At least five measurements were collected for each specimen. FATIGUE TEST

Uniaxial fatigue tests were performed at stress ratio R ¼ 1 and 60 to 70 Hz following ASTM E466.25 Multiple servohydraulic fatigue testing machines with a 15 kN load cell were used for the tests. Tests were conducted in lab air at room temperature. The sample geometry is shown in figure 3. Samples were tested until full separation or until 107 cycles (runout).

FIG. 2 S-N curves of samples in Group 1 to Group 4 (A)–(D).

.

LAI ET AL., DOI: 10.1520/STP163120190127

FIG. 3 Fatigue sample geometry.

The model/method used for fatigue strength calculation needs to consider the fatigue behavior of the material tested. Many models were proposed for fitting the fatigue S-N curves. The simplest one is the Basquin method,26 which is widely used for high-cycle fatigue analysis (ASTM E739-10, Standard Practice for Statistical Analysis of Linear or Linearized Stress-Life (S-N) and Strain-Life (e-N) Fatigue Data27) and is shown in equation (3):  b Sa ¼ C 2Nf

(3)

where: sa is the stress amplitude, Nf is the number of cycles to fracture, and C and b are empirical constants. However, the log-log linear assumption cannot correctly characterize most of the SN curves near the high-cycle region. Much of the literature reported log-log bilinear behavior for additive manufactured materials,28–30 which have much lower slopes when approaching 107 cycles. The FKM guideline12 also recommends using a loglog bilinear model for steel and aluminum alloys. Pascual and Meeker31 proposed a random fatigue limit (RFL) model, which is a modified Basquin model and is shown in equation (4).  b Sa  SL ¼ C 2Nf

(4)

The model introduces a new material constant, SL, which is the infinite-life fatigue limit of the material. The model can properly capture the changing slope of the fatigue S-N curve. In this study, the fatigue S-N curves were fitted using an RFL model with the aid of the maximum likelihood method described in Engler-Pinto et al.32 to account for the runout data points. The fatigue strength is then calculated at 107 cycles. .

235

236

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

In order to study the effect of defect size on fatigue strength, the fatigue strength was measured for a single specimen using the procedure proposed by Maxwell and Nicholas.33 The method was used extensively by Murakami21 in his research. The killer defect size was then measured for each specimen using the method proposed by Murakami.21

Results MICROSTRUCTURE AND HARDNESS

The optical micrographs of samples in different groups are shown in figure 4. Hatch patterns and border scans are clearly seen in the polished samples without post heat treatment (fig. 4A and 4B). For the machined and polished samples, the border scan is completely removed. As shown in table 1, the laser parameters in the hatch region are the same for all four groups. These are the optimal laser parameters provided by the equipment manufacturer specifically for the hatch region to minimize porosity. The optimal laser parameters that minimize the porosity for the border/contour region are used in Groups 2 and 3. Increasing the volumetric energy density will

FIG. 4 Optical micrographs of (A) polished sample in Group 1, (B) polished sample in Group 2, (C) polished sample in Group 3 (low magnification), and (D) polished sample in Group 3 (high magnification).

.

LAI ET AL., DOI: 10.1520/STP163120190127

increase keyhole porosity. The laser powers were increased from 300 W to 330 W in Group 1 and Group 4, respectively, to introduce keyhole porosity to study the effect of defect on fatigue strength. The morphology of the defect can have a significant effect on fatigue strength; therefore, the simplest morphology was used for the current study. Figure 4C and 4D shows the polished samples homogenized at 500 C for 1 h. The melt pool cannot be seen, and the silicon eutectic phase has spheroidized. The defect size distribution is shown in figure 5 for the border and hatch regions in all four groups. The average defect area, defect area fraction, and average equivalent defect diameter are summarized in table 2. The equivalent defect diameter is calculated by assuming a round defect with the same defect area. It is clearly seen that the hatch regions in all four groups have similar defect size distributions. As expected, the defect area fractions in the border/contour regions in Groups 1 and 4 are higher than the ones in Groups 2 and 3 due to some large keyhole defects.

FIG. 5 Defect size distribution for hatch and border regions in Groups 1 through 4.

.

237

238

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 2 Summary of defects for hatch and border regions in Groups 1 through 4

Ave. Area (lm2)

Group 1

Group 2

Group 3

Group 4

Area Fraction (%)

Ave. Eq. Diameter (lm)

Border/contour

511.6 6 1373.2

4.88

19.4 6 16.5

hatch

254.2 6 177.2

0.58

17.1 6 5.5

Border/contour

207.2 6 225.1

0.62

15.1 6 6.1

hatch

262.1 6 205.1

0.77

17.2 6 6.0

Border/contour

158.3 6 239.9

0.59

11.4 6 1.2

hatch

167.5 6 208.8

0.31

14.1 6 5.0

Border/contour

456.9 6 983.4

1.93

19.1 6 9.9

hatch

228.7 6 286.1

0.41

13.1 6 3.8

Hardnesses of the border and hatch regions are both around 135 HV. There is no significant difference between the border and hatch regions. The hardnesses of Groups 3 and 4 are around 70 HV after 500 C/1 h solution heat treatment. EFFECT OF SURFACE ROUGHNESS ON FATIGUE PROPERTY

The as-built and polished samples in Group 1 and Group 2 are used for studying the effect of roughness on fatigue strength. Comparing the mechanical properties between the as-built and polished samples gives clear insight on how roughness plays a role in fatigue properties. Surface Roughness Measurement

The surface roughness measurement was performed using the KEYENCE WideArea 3D Measurement System (VR-3000 Series). Ra and Rz were measured following steps illustrated in EN ISO 4288.34 Ra is the arithmetical mean roughness value. It is the arithmetical mean of the absolute values of the profile deviations from the mean line of the roughness profile. Rz is mean roughness depth. It is the mean value of the five peak and valley differences from five sampling lengths in the evaluation length.35 Rt (total height of profile), Rp (maximum profile peak height), and Rv (maximum profile valley depth) are also reported for easy comparing with literature data. For the detailed roughness definition, refer to the Japanese Industrial Standard (JIS) B 0601.36 The roughness measurements for the as-built and polished samples of Group 1 and Group 2 are shown in table 3. It is clear that the roughness measurements are comparable after polishing for both groups. The as-built roughness is slightly higher for Group 2 for both Ra and Rz measurements. However, the reason that causes the difference is uncertain because different laser parameters and different batches of powder were used for the two groups. Fatigue S-N Curves of As-Built and Polished Samples

The fatigue S-N curves for the as-built and polished samples of Group 1 and Group 2 are shown in figure 2A and 2B. The solid circles represent complete failures and .

LAI ET AL., DOI: 10.1520/STP163120190127

TABLE 3 Roughness measurements of samples in Group 1 and Group 2

Group 1

Ra (lm)

Rz (lm)

Rt (lm)

Rp (lm)

Rv (lm)

10.4 6 1.1

As-built

3.5 6 0.2

21.7 6 2.3

30.5 6 5.6

11.3 6 1.8

Polished

0.35 6 0.03

2.0 6 0.2

3.0 6 0.6

1.0 6 0.1

1.0 6 0.1

Machined and

0.51 6 0.09

3.1 6 0.7

5.2 6 2.5

1.6 6 0.3

1.5 6 0.4

As-built

5.5 6 0.8

37.8 6 7.6

61.2 6 22.3

23.3 6 6.8

14.3 6 1.6

Polished

0.38 6 0.04

2.2 6 0.3

3.2 6 0.9

1.1 6 0.2

1.1 6 0.2

Machined and 0.35 6 0.05

2.0 6 0.3

2.8 6 1.0

1.0 6 0.2

1.0 6 0.2

polished Group 2

polished

the hollow circles represent runouts. The fitted lines represent 50% probability of failure based on the random fatigue limit method.32 The fatigue strengths are also calculated using the random fatigue limit method32 at the life of 107 cycles and are shown in table 4. Note that the fatigue strength calculated using the Basquin equation is very close to the one calculated using the random fatigue limit equation except that Basquin does not assume a physical “fatigue limit.” EFFECT OF DEFECTS ON FATIGUE PROPERTIES

Fatigue Results of Group 1 and Group 2

The polished and machined and polished samples in Group 1 and Group 2 are used to study the effect of defects on fatigue strength. In Group 1, a defect was introduced in the border region and the defect was minimized in the hatch region. The polished sample was then tested to obtain the fatigue strength and compare it with the machined and polished (no border) samples. The same procedure was used on a reference set of samples (polished and machined and polished) in Group 2, where the defect was minimized in both the border and hatch regions. The fatigue strength calculated using the random fatigue limit method32 at the life of 107 cycles is shown in table 4.

TABLE 4 Fatigue strengths of samples in Group 1 through Group 4

Fatigue Strength (MPa)

As-built

Group 2

45 6 3

Group 3

Group 4

Polished

96 6 8

93 6 16

104 6 3

89 6 2

Machined and

154 6 9

133 6 15

101 6 3

NA

polished

.

Group 1

50 6 4

239

240

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Fatigue Results of Group 3 and Group 4

Group 3 and Group 4 samples were homogenized to remove potential residual stress and ensure the same microstructure between the hatch and border regions. The hardness dropped to approximately 70 HV after the homogenization. The S-N curves are shown in figure 2C and 2D. The fatigue strength calculated using the random fatigue limit method32 at the life of 107 cycles is shown in table 4.

Discussion EFFECT OF SURFACE ROUGHNESS ON FATIGUE PROPERTY

According to the fatigue test results, the fatigue strength is significantly reduced due to the presence of surface roughness. The polished sample has a fatigue strength that is comparable to typical sand-cast aluminum alloys used for cylinder head applications. For example, the fatigue strength of a typical sand-cast 319-T7 cylinder head at the deck face is around 88 MPa.37 The roughness influence factor (equation [1]) is calculated to compare the impact of surface roughness on fatigue strength. The roughness influence factor for cast aluminum alloys based on the FKM model is plotted in figure 6 (blue curve) with the experimental data from Group 1 (orange, Rz ¼ 21.7 lm) and Group 2 (gray, Rz ¼ 37.8 lm). For the FKM model for cast aluminum alloys, aR;r is 0.32, RZ;M , is 2, and rUTS;sf ;min is 133; rUTS is 429.5 MPa based on the test result of Group 1 samples. Clearly, the FKM model for cast aluminum alloys significantly

FIG. 6 Roughness influence factor of FKM model and experimental data.

.

LAI ET AL., DOI: 10.1520/STP163120190127

underestimates the damaging effect of AM surface roughness. The FKM model predicts an 18% and 22% decrease for the measured surface roughness of Groups 1 and 2, respectively, while the experiment showed a 48% and 52% decrease for Groups 1 and 2, respectively. This could be because the FKM model was calibrated using the roughness created by conventional manufacturing methods and does not account for the local surface feature difference on the AM parts. L-PBF creates different surface features that may have different effects. Kantzos et al.38 used a synchrotron X-ray to reconstruct the detailed surface topography and concluded that the microscale-level stress concentration created by the attached powder particle cannot be captured using a conventional profilometer or optical roughness measurement device. This might be the main reason that L-PBF surface roughness is more detrimental than the conventional as-cast or machined surfaces. Another explanation is the steps created on the surface by the layered structure during the L-PBF process. The steps effectively initiate micro cracks and result in lower fatigue strength. Based on the results, the FKM model was fitted for L-PBF AlSi10Mg using the experimental data and is shown in equation (5).   1  0:32  log RZ;C  logð2rUTS =65Þ fR ¼ (5) 1  0:32  logð2Þ  logð2rUTS =65Þ

The L-PBF process creates a layered structure that can vary with different angles on the part. Therefore, the roughness effect needs further investigation for different build angles and top and down skins. Because larger build angles create higher surface roughness and the roughness influence factors in the current model were fitted by only two data points, the applicability of the model for high roughness needs to be further investigated. EFFECT OF DEFECTS ON FATIGUE PROPERTY

The fatigue strength of machined and polished samples is much higher than the polished samples for both Group 1 and Group 2. This is unexpected for Group 2 because both the hardness and porosity level are the same between the hatch and border regions. It is reasonable that the fatigue strengths are comparable for machined and polished samples in both groups because they use the same laser process parameters. However, the introduction of a large amount of keyhole-type defects, as shown in figure 4A, does not further decrease the fatigue strength in Group 1. To further examine this phenomenon, the samples were homogenized to remove the potential residual stress and to ensure the same microstructure between the hatch and border regions. Group 3 has the same laser parameter as Group 2. Both the hatch and border have similar hardness, microstructure, and porosity level (fig. 4C). The fatigue strengths for the polished and machined and polished samples are also comparable, 104 6 3 MPa and 101 6 3 MPa, respectively. This indicates that residual stress might play a role in the fatigue strength difference in the nonhomogenized samples. Masoomi et al.39 studied the residual stress in vertically built .

241

242

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

L-PBF 17-4 PH stainless steel and found noticeable tensile residual stress (59.8 6 13.4 MPa) at the circumference. A further study on residual stress needs to be performed to understand its effect at the border region. The effect of the defect is also isolated through the homogenization process. Laser power was increased intentionally in the border scan of Group 4 samples to induce porosity. The samples were then homogenized to remove the potential residual stress. The results clearly show that the defect decreases the fatigue strength from approximately 100 MPa to 89 MPa. All the fatigue initiation sites are from surface or near-surface defects, as shown in figure 7, for example. The fatigue strength versus defect size (Kitagawa diagram) was plotted in figure 8 for samples for Group 2 (nonheat-treated) and 4 (heattreated). The fatigue strength is calculated based on Maxwell and Nicholas,33 and the killer size (the size of the defect that causes the crack initiation) was measured based on Murakami21 for each sample. The black (upper) and blue (lower) lines are based on equation (6) proposed by Ueno et al.23 Dr ¼

1:43  ðHV þ 75Þ pffiffiffiffiffiffiffiffiffi 1=6 ð areaÞ

(6)

The results show that fatigue strength depends on both hardness and defect size. The modified Murakami model can account correctly for both factors for LPBF AlSi10Mg. However, care should be taken on the defect morphology, as the model can only correctly predict fatigue strength of “near spherical” defects. The effect of defect morphology will be addressed in a future publication.

FIG. 7 Fatigue crack initiation at a surface keyhole defect (Group 1, polished, tested at 115 MPa, failed at 175,882 cycles).

.

LAI ET AL., DOI: 10.1520/STP163120190127

FIG. 8 Fatigue strength versus defect size for L-PBF AlSi10Mg with/without solution heat treatment (500  C/1 h).

Conclusions The effect of roughness on fatigue strength was evaluated on L-PBF AlSi10Mg. The results indicate that the FKM model significantly underestimates the damaging effect of roughness produced by L-PBF. Hence, the model needs recalibration. The effect of roughness of different build angles needs to be studied as well. Potential residual stress was observed in the border indirectly. The presence of the residual stress needs further confirmation. The presence of defects at the surface has a negative impact on fatigue strength, as expected. A modified Murakami’s fatigue strength model was validated for L-PBF AlSi10Mg under as-built and heattreated conditions. The results show that fatigue strength depends on both hardness and defect size. The modified Murakami model can correctly account for both factors for L-PBF AlSi10Mg. Artificial defects will be introduced to validate the model for larger defect size.

References 1.

.

R. Shrestha, J. Simsiriwong, and N. Shamsaei, “Fatigue Behavior of Additive Manufactured 316L Stainless Steel Parts: Effects of Layer Orientation and Surface Roughness,” Additive Manufacturing 28 (2019): 23–38, https://doi.org/10.1016/j.addma.2019.04.011

243

244

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

2. 3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

14. 15.

16. 17.

.

R. Molaei and A. Fatemi, “Fatigue Design with Additive Manufactured Metals: Issues to Consider and Perspective for Future Research,” Procedia Engineering 213 (2018): 5–16. J. W. Pegues, M. Roach, R. S. Williamson, and N. Shamsaei, “Surface Roughness Effects on the Fatigue Strength of Additively Manufactured Ti-6Al-4V,” International Journal of Fatigue 116 (2018): 543–552, https://doi.org/10.1016/j.ijfatigue.2018.07.013 J. W. Pegues, N. Shamsaei, M. D. Roach, and R. S. Williamson, “Fatigue Life Estimation of Additive Manufactured Parts in the As-Built Surface Condition,” Material Design and Processing Communications 1 (2019): e36, https://doi.org/10.1002/mdp2.36 J. H. Rao, K. Zhang, P. Rometsch, A. Huang, and X. Wu, “The Influence of Surface Roughness on the Fatigue Performance of Selective Laser Melted Aluminum Alloy A357,” in Proceedings of the 16th International Aluminum Alloys Conference (ICAA16) (Montreal, Quebec, Canada: Canadian Institute of Mining, Metallurgy and Petroleum, 2018). R. Molaei and A. Fatemi, “Crack Paths in Additive Manufactured Metallic Materials Subjected to Multiaxial Cyclic Loads Including Surface Roughness, HIP, and Notch Effects,” International Journal of Fatigue 124 (2019): 558–570, https://doi.org/10.1016/j.ijfatigue.2019.03.007 T. M. Mower and M. J. Long, “Mechanical Behavior of Additive Manufactured, PowderBed Laser-Fused Materials,” Materials Science & Engineering A 651 (2016): 198–213, http://dx.doi.org/10.1016/j.msea.2015.10.068 D. Greitemeier, C. Dalle Donne, F. Syassen, J. Eufinger, and T. Melz, “Effect of Surface Roughness on Fatigue Performance of Additive Manufactured Ti–6Al–4V,” Materials Science and Technology 32 (2016): 629-634, http://dx.doi.org/10.1179/1743284715Y.0000000053 A. Yadollahi, M. J. Mahtabi, A. Khalili, H. R. Doude, and J. C. Newman, Jr., “Fatigue Life Prediction of Additively Manufactured Material: Effects of Surface Roughness, Defect Size, and Shape,” Fatigue & Fracture of Engineering Materials & Structures 41 (2018): 1602–1614, https://doi.org/10.1111/ffe.12799 E. Wycisk, C. Emmelmann, S. Siddique, and F. Walther, “High Cycle Fatigue (HCF) Performance of Ti-6Al-4V Alloy Processed by Selective Laser Melting,” Advanced Materials Research 816–817 (2013): 134–139, https://doi.org/10.4028/www.scientific.net/AMR.816817.134 A. Yadollahi and N. Shamsaei, “Additive Manufacturing of Fatigue Resistant Materials: Challenges and Opportunities,” International Journal of Fatigue 98 (2017): 14–31, http:// dx.doi.org/10.1016/j.ijfatigue.2017.01.001 Forschungskuratorium Maschinenbau (FKM), Analytical Strength Assessment of Components in Mechanical Engineering, FKM Guideline, 5th rev. ed. (Frankfurt, Germany: FKM, 2003). E. Siebel and M. Gaier, “Untersuchungen u ¨ber den Einfluß der Oberfla ¨chenbeschaffenheit auf die Dauerschwingfestigkeit metallischer Bauteile [Studies on the Influence of the Surface Quality on the Fatigue Strength of Metallic Components],” VDI Journal 98 (1956): 1751–1774. Dauerfestigkeit der Maschinenbauteile [Fatigue Strength of Machine Elements], TGL 19340/01–04 (Berlin, Germany: GDR State Publishers, 1983). M. Suraratchai, J. Limido, C. Mabru, and R. Chieragatti, “Modelling the Influence of Machined Surface Roughness on the Fatigue Life of Aluminium Alloy,” International Journal of Fatigue 30 (2008): 2119–2126, https://doi.org/10.1016/j.ijfatigue.2008.06.003 D. Arola and C. Williams, “Estimating the Fatigue Stress Concentration Factor of Machined Surfaces,” International Journal of Fatigue 24 (2002): 923–930. S. Beretta and S. Romano, “A Comparison of Fatigue Strength Sensitivity to Defects for Materials Manufactured by AM or Traditional Processes,” International Journal of Fatigue 94 (2017): 178–191, http://dx.doi.org/10.1016/j.ijfatigue.2016.06.020

LAI ET AL., DOI: 10.1520/STP163120190127

18.

19.

20.

21. 22. 23.

24.

25.

26. 27.

28.

29.

30.

31. 32.

33.

.

S. Romano, A. Branda ˜o, J. Gumpinger, M. Gschweitl, and S. Beretta, “Qualification of AM Parts: Extreme Value Statistics Applied to Tomographic Measurements,” Materials & Design 131 (2017): 32–48, http://dx.doi.org/10.1016/j.matdes.2017.05.091 S. Romano, A. Bru ¨ckner-Foit, A. Branda ˜o, J. Gumpinger, T. Ghidini, and S. Beretta, “Fatigue Properties of AlSi10Mg Obtained by Additive Manufacturing: Defect-Based Modelling and Prediction of Fatigue Strength,” Engineering Fracture Mechanics 187 (2018): 165–189, https://doi.org/10.1016/j.engfracmech.2017.11.002 S. Romano, S. Miccoli, and S. Beretta, “A New FE Post-Processor for Probabilistic Fatigue Assessment in the Presence of Defects and Its Application to AM Parts,” International Journal of Fatigue 125 (2019): 324–341, https://doi.org/10.1016/j.ijfatigue.2019.04.008 Y. Murakami, Metal Fatigue: Effects of Small Defects and Nonmetallic Inclusions (London, UK: Elsevier, 2002). J. A. Bannantine, Fundamentals of Metal Fatigue Analysis (Englewood Cliffs, NJ: Prentice Hall, 1989). A. Ueno, M. Nishida, S. Miyakawa, K. Yamada, and S. Kikuchi, “Fatigue Limit Estimation of Aluminum Die-Casting Alloy by Means of HArea Method,” Journal of the Society of Materials Science, Japan 63 (2014): 844–849, https://doi.org/10.2472/jsms.63.844 A. Tajiri, T. Nozaki, Y. Uematsu, T. Kakiuchi, M. Nakajimac, Y. Nakamura, and H. Tanaka, “Fatigue Limit Prediction of Large Scale Cast Aluminum Alloy A356,” Procedia Materials Science 3 (2014): 924–929. Standard Practice for Conducting Force Controlled Constant Amplitude Axial Fatigue Tests for Metallic Materials, ASTM E466-15 (West Conshohocken, PA: ASTM International, approved November 1, 2015), https://doi.org/10.1520/E0466-15 O. H. Basquin, “The Exponential Law of Endurance Tests,” ASTM Proceedings 10 (1910): 625–630. Standard Practice for Statistical Analysis of Linear or Linearized Stress-Life (S-N) and Strain-Life (e-N) Fatigue Data, ASTM E739-10(2015) (West Conshohocken, PA: ASTM International, approved November 1, 2015), https://doi.org/10.1520/E0739-10R15 N. T. Aboulkhair, I. Maskery, C. Tuck, I. Ashcroft, and N. M. Everitt, “Improving the Fatigue Behaviour of a Selectively Laser Melted Aluminium Alloy: Influence of Heat Treatment and Surface Quality,” Materials and Design 104 (2016): 174–182, https://doi.org/10.1016/ j.matdes.2016.05.041 E. Brandl, U. Heckenberger, V. Holzinger, and D. Buchbinder, “Additive Manufactured AlSi10Mg Samples Using Selective Laser Melting (SLM): Microstructure, High Cycle Fatigue, and Fracture Behavior,” Materials and Design 34 (2012): 159–169, https:// doi.org/10.1016/j.matdes.2011.07.067 E. M. Tang and P. C. Pistorius, “Oxides, Porosity and Fatigue Performance of AlSi10Mg Parts Produced by Selective Laser Melting,” International Journal of Fatigue 94 (2017): 192–201, http://dx.doi.org/10.1016/j.ijfatigue.2016.06.002 F. G. Pascual and W. Q. Meeker, “Estimating Fatigue Curves with the Random FatigueLimit Model,” Technometrics 41 (1999): 277–290. C. Engler-Pinto, J. Lasecki, R. Frisch, M. DeJack, and J. Allison, Statistical Approaches Applied to Fatigue Test Data Analysis, SAE Technical Paper 2005-01-0802 (Warrendale, PA: SAE International, 2005), https://doi.org/10.4271/2005-01-0802 D. Maxwell and T. Nicholas, “A Rapid Method for Generation of a Haigh Diagram for High Cycle Fatigue,” in Fatigue and Fracture Mechanics: 29th Volume, ed. T. L. Panontin and S. D. Sheppard (West Conshohocken, PA: ASTM International, 1998), https:// doi.org/10.1520/STP14974S

245

246

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

34.

35.

36.

37.

38.

39.

.

Geometrical Product Specifications (GPS)—Surface Texture: Profile Method—Rules and Procedures for the Assessment of Surface Texture, EN ISO 4288:1996 (Geneva: ISO International, 1996). Geometrical Product Specifications (GPS)—Surface Texture: Profile Method—Terms, Definitions and Surface Texture Parameters, EN ISO 4287:1997 (Geneva: ISO International, 1997). Geometrical Product Specifications (GPS)—Surface Texture: Profile Method—Terms, Definitions and Surface Texture Parameters, JIS B 0601:2013 (Tokyo: Japanese Standards Association, 2013). W. Li, H. Cui, W. Wen, X. Su, and C. Engler-Pinto, “Very High Cycle Fatigue of Cast Aluminum Alloys under Variable Humidity Levels,” SAE International Journal of Materials and Manufacturing 8 (2015): 444–449, https://doi.org/10.4271/2015-01-0556 C. A. Kantzos, R. W. Cunningham, V. Tari, and A. D. Rollett, “Characterization of Metal Additive Manufacturing Surfaces Using Synchrotron X-Ray CT and Micromechanical Modeling,” Computational Mechanics 61 (2018): 575–580, https://doi.org/10.1007/ s00466-017-1531-z M. Masoomi, N. Shamsaei, R. A. Winholtz, J. L. Milner, T. Gna ¨upel-Herold, A. Elwany, M. Mahmoudi, and S. M. Thompson, “Residual Stress Measurements via Neutron Diffraction of Additive Manufactured Stainless Steel 17-4 PH,” Data in Brief 13 (2017): 408–414, https://doi.org/10.1016/j.dib.2017.06.027

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190138

Jan Dzˇugan,1 Daniel Melzer,1 Martina Koukolı´kova´,1 Jaroslav Vavrˇ´k, ı 1 and Mohsen Seifi2,3

Characterization of Functionally Graded Materials Based on Inconel 718 and Stainless Steel 316L Manufactured by DED Process Citation J. Dzˇugan, D. Melzer, M. Koukolı´kova ´, J. Vavrˇ´k, ı and M. Seifi, “Characterization of Functionally Graded Materials Based on Inconel 718 and Stainless Steel 316L Manufactured by DED Process,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 247–256. http://doi.org/ 10.1520/STP1631201901384

ABSTRACT

Additive manufacturing (AM) processes are being widely investigated and gradually applied for engineering applications. Presently, the main focus is on single material systems deposition, where there are still many issues with process stability and repeatability. However, a further huge leap in the field of AM process development will be the design of multiple material–heterogeneous components. The current study presents investigation of an experimental build consisting of multiple layers of austenitic stainless steel 316L and Inconel 718, which was created by a powder blow direct energy deposition system allowing simultaneous multiple materials’ deposition. Results report details of the microstructure investigation using optical metallography and scanning electron microscopy (SEM) analysis focusing on transition regions between the materials, Manuscript received November 1, 2019; accepted for publication February 11, 2020. 1 COMTES FHT, Inc. Pru ˚myslova´ 995, 334 41 Dobrˇany, Czech Republic J. D. http://orcid.org/0000-00018996-9341 2 ASTM International, 1850 M St. NW, Suite 1030, Washington, DC 20036, USA http://orcid.org/00000001-8385-2337 3 Case Western Reserve University, 10900 Euclid Ave., Cleveland, OH 44106, USA 4 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

247

248

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

where a significant difference was found in transition steel to Inconel and Inconel to steel. Hardness profile across deposited layers is established. Mechanical property assessment of the materials within the single material layers over the materials’ transition in the horizontal direction as well as across multimaterial layers in the building direction was carried out with the use of minitensile specimens. Results obtained in this study demonstrate the importance of the order of deposited materials and the effect on their mechanical properties. Keywords DED, heterogeneous materials, local properties, minisample testing, mechanical properties

Introduction Current technological development and efforts toward higher efficiency of technical systems bring a demand for materials with a good combination of corrosion resistance and strength. Often a combination of stainless steels and Inconel alloys gives not only corrosion resistance, due to their high chromium (Cr) content, but also good strength properties. These conventionally processed alloys are commonly used in nuclear power plants, aerospace engines, or in the gas industry.1,2 Among the modern ways to create functionally graded materials (FGMs) are additive manufacturing (AM) techniques. In the last decade, considerable numbers of different processes were developed. AM technologies provide an “unlimited degree” of freedom for geometry where very complex structures can be created. One of these processes is direct energy deposition (DED).3 With the use of AM, functionally graded structures may be created, as has been shown in previous studies4,5 where different powder mixtures were applied. Formed transition regions between different materials can be the weakest links of the structure. As a result of chemical composition inconsistency, considerable microstructural changes, carbide and element diffusion, intermetallic phase precipitation, and grains nucleation mechanism may affect properties and usability of produced components.6 The current study investigates the experimental blocks consisting of multiple layers of stainless steel 316L (SS316L) and Inconel 718 (IN718) that were created by a powder blown (PB) DED system allowing simultaneous multiple materials’ deposition. There are investigated local and orientation-related mechanical and microstructural properties of the single material layer as well as the transition regions.

Experimental The PB DED deposition system INSSTEK MX 600 with a 2-kW Ytterbium fiber laser was used for the experimental material deposition. The size of powder particles ranged from 50 to 150 lm. Inconel 718 powder was produced by AP&C, and steel 316 L was produced by Sandvick. Chemical composition is presented in table 1. A cubic block of 35-mm edge length with 5-mm-thick layers of SS316L and IN718 .

TABLE 1 Chemical composition of input powder materials wt%

C

Mn

P

S

Si

Cr

Mo

Ni

Cu

Al

B

Co

Nb

Ti

Fe

316L

0.04

0.02

0.006

0.001

0.07

19.00

3.04

52.89

0.10

0.54

0.00

0.10

5.04

0.90

Balance

IN718

0.21

1.33

0.020

0.006

0.82

17.20

2.30

10.40

-

-

-

-

-

-

Balance ˇ UGAN ET AL., DOI: 10.1520/STP163120190138 DZ

Note: C ¼ carbon; Mn ¼ manganese; P ¼ phosphorus; S ¼ sulfur; Si ¼ silicon; Mo ¼ molybdenum; Ni ¼ nickel; Cu ¼ copper; Al ¼ aluminum; B ¼ boron; Co ¼ cobalt; Nb ¼ niobium; Ti ¼ titanium; Fe ¼ iron.

249

.

250

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 Deposited block consisting of multiple layers of SS316L and IN718.

was produced. The deposited block is depicted in figure 1. The subsequent material layer was always deposited after the previous one was fully finished. The block was produced with set deposition parameters as presented in table 2. Scan strategy contour–filling (CF) was used for deposition. Deposition was realized under the protective atmosphere of argon gas. After the deposition process, the block was cut vertically for subsequent examinations. Chemical composition measurement, using an optical emission spectrometer (Bruker Q4 TASMAN), of powder and produced block was performed, and only negligible deviations from composition of input powders were detected. In order to measure porosity, a whole-block cross-sectional area was polished. Image analysis via light microscope (Nikon ECLIPSE MA200) and NIS Elements 5.2 software was carried out. The chosen area of interest was subsequently etched using Glyceregia etchant (15 mL hydrochloride [HCl], 10 mL glycerol, and 5 mL nitric acid (HNO3) in order to reveal transition regions between single material layers. In addition, energy-dispersive X-ray spectroscopy (EDS) analysis of transition regions was performed using a JEOL IT 500 HR scanning electron microscope.

TABLE 2 Deposition parameters for IN718 and SS316L

Material

.

Laser Power (W)

Feeding Rate (g/min)

IN718

375

3.5

8

SS316

417

3

8.5

Scanning Speed (l/min)

ˇ UGAN ET AL., DOI: 10.1520/STP163120190138 DZ

Hardness measurements were performed according to ISO 6507-1: Vickers hardness measurement using a laboratory hardness tester (Struers Durascan 50) with a normal force of FN ¼ 49.03 N and a hold time of 20 s. Hardness was measured across the deposited layers with a 1-mm step between indentations. A set of miniaturized tensile tests (MTTs) was performed to assess the mechanical properties.7,8 All tests were carried out on the testing machine with a linear drive and load capacity of 5 kN at ambient temperature and a strain rate of e_ ¼ 0.0009 s1. All tests were recorded using a digital image correction (DIC) system with a single telecentric camera mode. Ultimate tensile strength (UTS), yield stress (YS), and total elongation and reduction of area were evaluated for each specimen.

Results POROSITY

A porosity level of 0.011% was measured for a full block cross section. Globular pores with a maximum diameter of 20 lm were observed as well as small cracks located exclusively near transition regions. The porosity level was also measured in the immediate vicinity of the transition regions. The different porosity levels of 0.068% and 0.034% for transition SS316L to IN718 and IN718 to SS316L, respectively, suggest different fusion quality between materials with respect to the material order. LIGHT MICROSCOPY

Two kinds of transitions were observed in the area of material type change after etching, depicted in figure 2. A sharp transition at the interface of SS316L to IN718 was detected, whereas transition on the IN718 to SS316L interface was

FIG. 2 Light microscopy observation of etched cross section of deposited block: (A) IN718 to SS316L and (B) SS316L to IN718.

.

251

252

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

characterized by a gradual change. This finding indicates that the temperature intervals of the solidus and liquidus of the steel and nickel alloy and applied laser power may affect the character of the formed transition region. EDS ANALYSIS

Results of EDS analysis, presented in figure 3, further expand the light microscopy findings. The chemical composition in the transition area is changing corresponding to the character of interface rather sharply or gradually despite the fact that the transition regions always represent the material type change between a couple of the same materials. The dashed gray line represents the border of different material structures, namely IN718 and SS316L. HARDNESS MEASUREMENT

The hardness profile across the deposited layers is presented in figure 4. Single material layers show a hardness of 230HV5 and 280HV5 for SS316L and IN718,

FIG. 3 Map of chemical composition change of EDS analysis of transitions: (A) IN718 to SS316L and (B) SS316L to IN718.

FIG. 4 Hardness profile across the deposited block.

.

ˇ UGAN ET AL., DOI: 10.1520/STP163120190138 DZ

respectively. The transition region hardness exhibits very similar results to SS316L, with a slight decrease of hardness to 220 HV5. The presented hardness profile clearly demonstrates that hardness within a single 316L layer continuously decreases when deposited on previous IN718. In addition, the hardness changes rapidly when going from IN718 to SS316L, while the change from SS316L to IN718 is gradual. TENSILE TESTS

Specimen extraction and geometry is visualized in figure 5. Examples of selected engineering stress–engineering strain curves are depicted in figure 6. Negligible differences in tensile properties were observed for X- and Y-oriented specimens for a single material. Z-oriented specimens exhibit about 10% lower tensile values,9,10 which is associated with a significant decrease in plasticity in the case of SS316L. Transition regions show a different character throughout the deposited block. Transition from SS316L to IN718 yielded higher strength and plasticity in comparison to transition from IN718 to SS316L. Results of Z-oriented specimens located in transition from SS316L to IN718 are similar to SS316L plane-oriented specimens.

Results Discussion Results of mechanical tests suggest that the mechanical characteristics do not depend on the deposition height. That is demonstrated by a very good agreement of the mechanical properties of single material layers across the whole block. Light microscopy observations also clearly demonstrate the formation of two types of transition regions. A sharp transition is formed when SS316L is not fully melted due to the relatively low temperature of deposited IN718. Compared to that,

FIG. 5 (A) MTT specimen extraction visualization and (B) MTT specimen geometry.

.

253

254

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 6 Representative engineering stress–engineering strain curves for tested MTT.

a gradual transition is observed when SS316L is deposited on IN718. In that case, the melting pool is a mixture of these two materials because the laser power melts the substrate of IN718. Pores randomly but equally spread all over the deposited block. However, the porosity measurement revealed that pores and mainly cracks are located exclusively in transition regions or in their vicinity. In addition, the most crack sensitivity is in the transition from IN718 to SS316L because the higher porosity level and more cracks were observed in this area. Crack formation seems to be a result of the complex structure formation in the moment of fusion of two different types of materials. This is also observable by EDS analysis where transition regions are viewed from a chemical point of view rather sharply or gradually. If the melting point of a newly formed phase significantly differs from the melting points of the base materials, this may cause cracking.11 In order to avoid crack formation, numerical simulation tools or in situ monitoring might help in better understanding the fusion processes and transition region formation, as was shown in studies.12 Tensile test results demonstrate that a sharp transition exhibits higher values of tensile properties in comparison to a gradual transition. According to the fractographic observation results, the fracture surface of gradual transition specimens contains significantly greater lack of fusion and crack type defects in comparison to a sharp transition. This points out that the gradual transition has a higher tendency for defects to form.

Conclusion In the presented work, an experimental block consisting of multiple layers of stainless steel 316L and Inconel 718 was successfully created by a PB DED system. .

ˇ UGAN ET AL., DOI: 10.1520/STP163120190138 DZ

Achieved results point out that functionally graded structures can be produced by a DED system with very great reproducibility. This is supported by a set of mechanical tests that yielded very similar results over the whole block for certain characteristic areas. However, the mechanical and structural properties are strongly related to the material deposition order. Different deposition parameters (laser power) for certain materials have a strong effect on the formed transition region types as well as their quality and mechanical integrity. ACKNOWLEDGMENTS

This work was done within the project FV40166: Evaluation of Degraded Steels for the Construction of Turbines and Superheaters of Power Generating Boilers (2019–2022, sponsored by the Ministry of Industry and Trade of the Czech Republic).

References 1.

2.

3. 4.

5.

6.

7.

8.

9. 10.

.

A. Bansal, A. K. Sharma, S. Das, and P. Kumar, “On Microstructure and Strength Properties of Microwave Welded Inconel 718/Stainless Steel (SS-316L),” Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications 230, no. 5 (2016): 939–948, https://doi.org/10.1177/1464420715589206 F. Caiazzo and V Alfieri, “Simulation of Laser-Assisted Directed Energy Deposition of Aluminum Powder: Prediction of Geometry and Temperature Evolution,” Materials 12, no. 13 (2019), https://doi.org/10.3390/ma12132100 Additive Manufacturing—General Principles—Terminology, ISO/ASTM 52900:2015 (Geneva, Switzerland: ISO International, 2015). K. Shah, I. U. Haq, A. Khan, S. A. Shah, M. Khan, and A. J. Pinkerton, “Parametric Study of Development of Inconel-Steel Functionally Graded Materials by Laser Direct Metal Deposition,” Materials and Design 54 (2014): 531–538, https://doi.org/10.1016/j.matdes.2013.08.079 L. Yan, Y. Chen, and F. Liou, “Additive Manufacturing of Functionally Graded Metallic Materials Using Laser Metal Deposition,” Additive Manufacturing 31 (2020), https:// doi.org/10.1016/j.addma.2019.100901 F. Dokme, M. K. Kulekci, and U. Esme, “Microstructural and Mechanical Characterization of Dissimilar Metal Welding of Inconel 625 and Aisi 316l,” Metals 8, no. 10 (2018), https:// doi.org/10.3390/met8100797 J. Dzˇugan, R. Procha´zka, and P. Konopı´k, “Micro-Tensile Test Technique Development and Application to Mechanical Property Determination,” in Small Specimen Test Techniques: 6th Volume, ed. M. Sokolov and E. Lucon (West Conshohocken, PA: ASTM International, 2015), 12–30, https://doi.org/10.1520/STP157620140022 J. Dzˇugan, M. Seifi, R. Prochazka, M. Rund, P. Podany, P. Konopik, and J. J. Lewandowski, “Effects of Thickness and Orientation on the Small Scale Fracture Behaviour of Additively Manufactured Ti-6Al-4V,” Materials Characterization 143 (2018): 94–109, https:// doi.org/10.1016/j.matchar.2018.04.003 M. Brandt, Laser Additive Manufacturing: Materials, Design, Technologies, and Applications (Boston, MA: Elsevier/Woodhead, 2017). V. Popovich, E. Borisov, A. A. Popovich, V. Sufiiarov, D. V. Masaylo, and L. Alzina, “Functionally Graded Inconel 718 Processed by Additive Manufacturing: Crystallographic Texture, Anisotropy of Microstructure and Mechanical Properties,” Materials & Design 114 (2016), https://doi.org/10.1016/j.matdes.2016.10.075

255

256

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

11. 12.

.

S. Kou, “Solidification and Liquation Cracking Issues in Welding,” JOM 55, no. 6 (2003): 37–42, https://doi.org/10.1007/s11837-003-0137-4 B. Lo ´pez, I. Gutie ´rrez, and J. J. Urcola, “Study of the Microstructure Obtained after Diffusion Bonding Inconel 625 to Low Alloy Steel by Hot Uniaxial Pressing or Hipping,” Materials Characterization 28, no. 1 (1992): 49–59, https://doi.org/10.1016/10445803(92)90028-G

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190134

Inigo Bacaicoa,1 Sascha Horn,1 Angelika Brueckner-Foit,1 Julia Richter,1 and Thomas Niendorf1

Fretting Fatigue Characterization in Press-Fit Joints of AM Parts by X-Ray Tomography and Digital Image Correlation Citation I. Bacaicoa, S. Horn, A. Brueckner-Foit, J. Richter, and T. Niendorf, “Fretting Fatigue Characterization in Press-Fit Joints of AM Parts by X-Ray Tomography and Digital Image Correlation,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 257–270. http:// doi.org/10.1520/STP1631201901342

ABSTRACT

Complex structures made by additive manufacturing (AM) have to be integrated into larger components. In these cases, contact conditions at the press fit joint can induce crack initiation and accelerate damage accumulation when the system is subjected to cyclic loading. Surface-finishing techniques can improve the surface quality of AM structures. However, surface finishing can be a challenge when the geometry is extremely complex. Moreover, electropolishing, which is often applied to AM parts, is unable to reduce the waviness associated with the AM process, at least when reasonable polishing times are considered. Furthermore, the surface finishing of AM parts can lead to the opening of subsurface pores, which are associated with the AM process and are among the main causes of fatigue failure in AM parts. Cracks initiate from the micronotches related to the rough surface. However, at the same time, roughness tips can lead to superior form fit and additional adhesive forces, especially if the AM part has a higher hardness than the conventionally manufactured part. Consequently, new testing procedures have to be developed for press-fit joints involving AM

Manuscript received November 1, 2019; accepted for publication March 3, 2020. 1 Institute for Materials Engineering, University of Kassel, Moenchebergstrasse 3, 34109 Kassel, Germany 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

257

258

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

parts subjected to vibrational loading as the classical fretting test cannot address the benefit of an additional form fit. In this study, the internal defect characteristics of AM specimens such as pores and lack of fusions, as well as the three-dimensional surface morphology, were characterized by high-resolution microcomputed tomography. The basic idea of the AM specific testing procedure presented in the present work is to use a classical flat dog-bone fatigue specimen with a press-fit joint in the center. This specimen is then subjected to fatigue loading in order to monitor the deformation field in the region of the joint by digital image correlation. This procedure enables the early detection of the onset of damage. Keywords additive manufacturing, fatigue, X-ray tomography, digital image correlation, fretting, SLM

Introduction Complex three-dimensional (3D) parts can be manufactured by selective laser melting (SLM) without additional fixing and assembling,1–3 enabling manufacturing of complex on-demand parts and, thus, lightweight components.4 Even if surface finishing is a common practice to improve the mechanical properties, it may not always be possible to conduct surface finishing for different reasons, such as production costs, geometry constraints, or process limitations. Moreover, different scales of roughness can be a challenge for some surface-finishing techniques (i.e., electropolishing, in which the waviness and the micronotches associated with the AM process may not be fully removed). Furthermore, parts manufactured by AM have to be integrated into more complex structures. A typical scenario could be a press-fit joint, in which a part made by AM is pressed into a structural part that has been conventionally manufactured. Under operational conditions, the joint is subjected both to static and vibrational (i.e., fatigue) loading. The surface roughness of the additively manufactured parts can strongly influence the failure mechanisms5,6 of the system because a rough contact surface implies that significant micronotches prevail from which cracks may initiate. Moreover, subsurface pores located close to contact zones can accelerate crack initiation and damage evolution.7 Surface finishing may also reduce the surface distance of subsurface defects, promoting surface crack initiation as well as the interaction of the subsurface pore with the micronotches in the rough surface. In this study, microcomputed X-ray tomography (l-CT) and digital image correlation (DIC) were used in order to characterize the damage behavior of a press-fit joint realized using SLM cylinders made from stainless steel and case-hardening steel pressed into a conventionally manufactured steel. The 3D geometries of the SLM parts were imported into a finite element (FE) software in order to simulate the stress concentrations in the press-fit joint considering the three-dimensional topology of the rough surface as well as the 3D morphology of subsurface pores following the methodology developed by Bacaicoa et al.8 .

BACAICOA ET AL., DOI: 10.1520/STP163120190134

The results obtained by FE simulation were compared with the DIC measurements of strain and stress concentrations during fatigue tests, and the fracture surfaces were analyzed by scanning electron microscopy (SEM) in order to characterize the damage behavior during fatigue.

Materials and Methods Two different material and process systems were considered in the current study: (1) AISI 316L (X2CrNiMo17-12-2) stainless steel cylinders manufactured by SLM with a diameter of 3.4 mm and a length of 15 mm thermally pressed into a conventionally manufactured X38CrMoV5-1 fatigue specimen and (2) 16MnCr5 steel cylinders manufactured by SLM with the same diameter and a length of 35 mm mechanically pressed into a conventionally manufactured X5CrNi18-10 fatigue specimen. The AM parts were manufactured using an SLM machine (SLM Solutions, Lu¨beck) applying a 400 W laser. For the 16MnCr5, the laser power was set at 300 W, the layer thickness was 50 lm, the hatch distance was 100 lm, and the scan speed was 750 mm/s, eventually leading to an energy density of 80 J/mm3. The stainless steel 316L was molten with a different set of parameters following strategies reported elsewhere.9 The geometry shown in figure 1 was used to cut the specimens from a conventionally forged and milled steel sheet. The cylinder-shaped AM parts were then pressed into these specimens. The specified geometry led to a homogeneous stress distribution in the center of the specimen. This is required to avoid influencing the fatigue behavior by the specimen geometry. Before polishing the specimen, a 3.4 mm diameter hole was drilled in its center. In the case of the AISI 316L SLM cylinder, it was cooled in liquid nitrogen, pressed into the hole of the X38CrMoV5-1 specimen, and heated to a temperature of 300 C (fig. 2). In the case of the 16MnCr5 steel, no thermal expansion was used. Instead, the cylinder was fixed in a

FIG. 1 Geometry of the fatigue specimens with highlighted region of interest for DIC.

.

259

260

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 Three-dimensional model of the AM pin (gray) being pressed into the fatigue specimen (yellow).

servohydraulic tensile testing machine (MTS Criterion) and pressed into the X5CrNi18-10 fatigue specimen, which was float mounted. Maximum pressing forces between 1.1 and 1.6 kN were recorded. Specimens with the press-fit joints were sprayed with silicon oxide particles in order to generate a reference pattern for DIC in the region of interest. Tensile and fatigue tests (R ¼ 1, 50 Hz) were conducted and images captured at different load levels to calculate the strain distributions by DIC. The deformation field was measured and compared in two different DIC programs (VEDDAC* and DaVis{). Finally, the fracture surfaces were analyzed in the SEM in order to study the failure mechanism. The AM cylindrical pin was analyzed by l-CT in a ZEISS Xradia Versa 520 at a voltage of 160 kV, power of 10 W, source distance of 20 mm, and detector distance of 25 mm, leading to a voxel size of 3.74 lm. The pin geometry with the 3D topology of the rough surface and the subsurface pores was segmented using a gray-thresholding tool employing a commercial segmentation software (Avizo{). The surface model and mesh were edited using a reverse engineering tool (Geomagic§). The computer-aided design (CAD) model was generated based on the prevailing surface topology and exported into a commercial FE software (Abaqus**).

*

Chemnitzer Werkstofftmechanik GmbH, VEDDAC. Lavision GmbH, DaVis. { Thermo Fisher Scientific, Avizo 9.2. § 3D SYSTEMS, Geomagic Wrap. ** Dassault Systemes, SIMULIA Abaqus FEA. {

.

BACAICOA ET AL., DOI: 10.1520/STP163120190134

Results and Discussion MATERIAL CHARACTERIZATION

AISI 316L Steel

Optical microscopy revealed a significant eccentricity of the pin as well as different roughness levels at different positions (fig. 3). For 3D analysis, a l-CT scan of the cut-free pin was conducted. The topography, as shown in figure 4, revealed a partial smoothing of the surface as a result of the plastic deformation during pressing. Two-dimensional cross sections of the scan were statistically analyzed using ImageJ (National Institutes of Health), and relatively high scatter in terms of roughness and the circularity level of the cylindrical SLM parts can be seen in the statistical distribution (fig. 5). Ra values were considered for the statistical analysis of the roughness. These values are most significant for the qualitative description of the roughness here as for a detailed quantitative assessment of the roughness, other parameters may be required.10,11 Obviously, the deviations in circularity of the cylindrical SLM parts lead to eccentricity between the SLM part and the hole drilled in the specimen. This effect is even more pronounced since different roughness levels are found within different areas of the AM pin surface. As can be revealed from the 3D segmentation of inner pores in figure 6, elongated, radially oriented pores in the center as well as tangential ones in the subsurface regime were present. The statistical distribution of the maximum Feret size (fig. 7) showed large pores with a maximum Feret length of up to 700 lm. Elongated pores radially oriented in the inner part together with tangential pores in the subsurface can be seen in figure 6.

FIG. 3 Micrograph of the AM pin pressed into the fatigue specimen.

.

261

262

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 4 Three-dimensional geometry of the surface topology of the AM AISI 316L steel pin, acquired by μ-CT (pin diameter is 3.4 mm).

FIG. 5 Statistical distribution of the roughness levels in the different CT cross sections.

For the simulation of the stress concentrations associated with the rough surface and the subsurface pores, the rough AM pin with the subsurface pores was embedded in the hole of the specimen CAD geometry and the contact interaction was modeled considering rough contact. 16MnCr5 Steel

For the 16MnCr5 steel mechanically pressed into the specimen, globular features were detected on the rough surface (fig. 8). In this case, the mechanical pressing enabled robust control of the pressing process. Lower circularity defects associated with the AM process were observed as well as lower roughness levels and scatter of .

BACAICOA ET AL., DOI: 10.1520/STP163120190134

FIG. 6 Three-dimensional geometry of the pores located in the AM AISI 316L steel pin (diameter is 3.4 mm).

FIG. 7 Statistical distribution of the maximum Feret sizes of pores.

FIG. 8 Three-dimensional geometry of the surface topology of the AM 16MnCr5 steel pin, acquired by μ-CT (diameter is 3.4 mm).

.

263

264

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 9 Statistical distribution of the roughness levels in the different CT cross sections.

these values (fig. 9). No significant eccentricity caused by circularity defects during the pressing process was detected, leading to well-defined contact conditions at different positions of the specimen. In figure 10, the 3D geometry of the subsurface pores is revealed by l-CT. This analysis clearly reveals that subsurface pores are mostly tangentially oriented. The statistical analysis of the pore sizes points at the significantly lower fraction of large inner pores with an average maximum Feret size of 125 lm (fig. 9).

FIG. 10 Three-dimensional geometry of the pores located in the AM 16MnCr5 steel pin (diameter is 3.4 mm).

.

BACAICOA ET AL., DOI: 10.1520/STP163120190134

FIG. 11 Evolution of the deformation field after applying different loads: (A) zero cycles, (B) one cycle, and (C) four cycles. VEDDAC software, pixel size: 0.5 μm, subset size: 300 px, step size: 50 px, median filter.

DAMAGE BEHAVIOR

AISI 316L Steel

The DIC measurements revealed high local strain concentrations caused by the micronotches in the contact zone associated with the rough surface of the 316L specimen (fig. 11). Multiple locations were found with high strain concentrations. In these regions, significant evolution of damage was monitored and characterized during the cyclic loading at 305 MPa (nominal stress level calculated for the total area of the pin-specimen system). Linear-elastic FE simulations were conducted with cylindrical AM pins embedded in the hole of the outer specimen geometry in a first step considering rough contact interaction (ABAQUS). Generally, it can be deduced from figure 12 that the maximum stress concentrations are located in specific areas of the contact zones between the AM pin and the specimen upon monotonic loading. The eccentricity between the AM pin and the specimen hole had a significant influence on the positions where the maximum stress concentrations are located in the contact zones. Afterward, the AM pin with the 3D rough surface topology was embedded into the specimen geometry, and the maximum stress concentrations were calculated in the contact zone between the AM pin and the specimen (fig. 12). The fracture surfaces of the specimens show crack initiation in specific positions of the contact zone (fig. 13). A shear-induced crack initiation process can be observed, which is typical in case of fretting damage. The damage process is characterized by a combination of roughness-induced friction together with pressure caused by the interference fit and eccentricity between the AM pin and the specimen hole. The damage positions coincide with the FE results for the scenario considering eccentricity. 16MnCr5 Steel

Homogenous stress distribution in the region with the highest contact pressure was found by strain analysis based on DIC (fig. 14). In contrast to the thermally pressed .

265

266

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 12 Von Mises stress distribution after applying tensile loading, calculated by linear elastic FE simulation: (A) specimen with press-fit joint and (B) AM pin.

AISI 316L AM steel pin, the effect of micronotches on the stress concentration is not as significant, and the highest strain concentrations are only seen in early stages of fatigue. The FE simulation performed in ABAQUS with the 3D CAD model of the AM pin, including the 3D topography of the part and the subsurface pores showed that the highest local stress concentrations are located on the edges of the subsurface pores due to their tangential arrangement (fig. 15). It can also be seen that an interaction between the local stress fields of pores and the rough surface occurs as well as an increase of the local stress concentration on the surface due to the subsurface pores. Moreover, it can be observed that interaction takes place between individual subsurface pores due to their proximity to each other. On the fracture surfaces, it can be seen (fig. 16) that crack initiation occurs on one specific site of the contact zone. In this case, analysis revealed that a locally prevailing superposition of roughness-induced friction and pressure due to eccentricity (even if less pronounced in the 16MnCr5) and the damage position show a very good match with the simulation results for the scenario considering eccentricity.

.

BACAICOA ET AL., DOI: 10.1520/STP163120190134

FIG. 13 Fracture surface of the X38CrMoV5-1 fatigue specimen with the AISI 316L steel AM pin: (A) global view (crack initiation sites marked) and (B) detail view.

Conclusions A newly designed classical flat dog-bone specimen with an AM press-fit joint in the center was used in order to monitor the deformation field in the region of the joint by DIC under fatigue loading. This novel approach enabled the early detection of the onset of damage. Based on the results discussed, the following conclusions can be drawn: 1. Microcomputed X-ray tomography is a useful tool for characterization of the 3D topology of the roughness and the 3D geometry of the subsurface pores. 2. The high resolution of the l-CT enabled the simulation of the effect of subsurface pores and roughness on the local stress concentration. 3. It was found that subsurface pores significantly increase the local stress concentration on the rough surface. 4. The analysis of the fracture surfaces revealed fretting fatigue as the main damage mechanism.

.

267

268

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 14 Evolution of the deformation field after applying a cyclic stress of 140 MPa: (A) 0 cycles, (B) 50,000 cycles, (C) 100,000 cycles, and (D) 150,000 cycles. The sample failed at 160,000 cycles. DaVis, pixel size: 0.77 μm2, subset size: 65 px, step size: 15 px, Gaussian filter.

FIG. 15 Von Mises stress distribution along the subsurface pores and the 3D rough surface within the contact zone between the AM pin and the outer specimen calculated by FE.

.

BACAICOA ET AL., DOI: 10.1520/STP163120190134

FIG. 16 Fracture surface of the X5CrNi18-10 fatigue specimen with the 16MnCr5 steel AM pin: (A) global view (crack initiation sites marked) and (B) detail view.

References 1. 2.

3. 4.

.

S. Dadbakhsh, L. Hao, and N. Sewell, “Effect of Selective Laser Melting Layout on the Quality of Stainless Steel Parts,” Rapid Prototyping Journal 18, no. 3 (2012): 241–249. Standard Terminology for Additive Manufacturing Technologies, ASTM F2792-12a (West Conshohocken, PA: ASTM International, approved 2012, withdrawn 2015), https:// doi.org/10.1520/F2792-12A Z. Sun, X. Tan, S. B. Tor, and W. Y. Yeong, “Selective Laser Melting of Stainless Steel 316L with Low Porosity and High Build Rates,” Materials & Design 104 (2016): 197–204. W. E. Frazier, “Metal Additive Manufacturing: A Review,” Journal of Materials Engineering and Performance 23, no. 6 (2014): 1917–1928.

269

270

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

5.

6.

7. 8.

9.

10.

11.

.

J. Gu ¨nther, D. Krewerth, T. Lippmann, S. Leuders, T. Tro ¨ster, A. Weidner, H. Biermann, and T. Niendorf, “Fatigue Life of Additively Manufactured Ti–6Al–4V in the Very High Cycle Fatigue Regime,” International Journal of Fatigue 94 (2017): 236–245. J. Gu ¨nther, S. Leuders, P. Koppa, T. Tro ¨ster, S. Henkel, H. Biermann, and T. Niendorf, “On the Effect of Internal Channels and Surface Roughness on the High-Cycle Fatigue Performance of Ti-6Al-4V Processed by SLM,” Materials & Design 143 (2018): 1–11. T. M. Mower and M. J. Long, “Mechanical Behavior of Additive Manufactured, PowderBed Laser-Fused Materials,” Materials Science and Engineering: A, 651 (2016): 198–213. I. Bacaicoa, M. Wicke, M. Luetje, F. Zeismann, A. Brueckner-Foit, A. Geisert, and M. Fehlbier, “Characterization of Casting Defects in a Fe-Rich Al-Si-Cu Alloy by Microtomography and Finite Element Analysis,” Engineering Fracture Mechanics, 183 (2017): 159–169. O. O. Salman, F. Brenne, T. Niendorf, J. Eckert, K. G. Prashanth, T. He, and S. Scudino, “Impact of the Scanning Strategy on the Mechanical Behavior of 316L Steel Synthesized by Selective Laser Melting,” Journal of Manufacturing Processes 45 (2019): 255–261. J. Pegues, M. Roach, R. S. Williamson, and N. Shamsaei, “Surface Roughness Effects on the Fatigue Strength of Additively Manufactured Ti-6Al-4V,” International Journal of Fatigue 116 (2018): 543–552. A. Yadollahi, M. J. Mahtabi, A. Khalili, H. R. Doude, and J. C. Newman, Jr., “Fatigue Life Prediction of Additively Manufactured Material: Effects of Surface Roughness, Defect Size, and Shape,” Fatigue & Fracture of Engineering Materials & Structures 41, no. 7 (2018): 1602–1614.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190125

Dalia Mahmoud,1 Mohamed A. Elbestawi,1 and Kassim S. Al-Rubaie1

Effect of Microstructure and Internal Defects on the Mechanical Properties of Ti6Al4V Gyroid Lattice Structures for Biomedical Implants Citation D. Mahmoud, M. A. Elbestawi, and K. S. Al-Rubaie, “Effect of Microstructure and Internal Defects on the Mechanical Properties of Ti6Al4V Gyroid Lattice Structures for Biomedical Implants,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 271–288. http:// doi.org/10.1520/STP1631201901252

ABSTRACT

Selective laser melting (SLM) is a laser powder bed fusion (L-PBF) technique that can be used to print lattice structures with fine complicated features. Much effort has been made to choose a lattice design that enhances the mechanical and biological functions for biomedical implants. Triply periodic minimal surface (TPMS) lattice structures, namely gyroids, have shown a great potential to match the mechanical and biological properties of bone tissue. Although the design plays a major role in determining the properties of lattice structures, the effect of the SLM process on the lattice structure quality is often overlooked. This work focuses on the relationship between the resultant microstructure and the mechanical properties of Ti6Al4V gyroid lattice structures. Different process parameter combinations were used to develop a wide range of volumetric energy density (VED). The gyroid design was then printed at three VED levels: 43, 103, and 192 J/mm3. The apparent density, morphology, and internal defects Manuscript received October 29, 2019; accepted for publication February 11, 2020. 1 Mechanical Engineering, McMaster University, 1280 Main St., W. Hamilton, ON L8S 4L8, Canada D. M. http://orcid.org/0000-0002-4654-2273, M. A. E. http://orcid.org/0000-0003-0982-6127, K. S. A.-R. http://orcid.org/0000-0003-4507-2852 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

271

272

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

were analyzed. Microcomputed tomography (microCT) was used for characterizing the morphology of the samples. The results showed that the apparent density was highly dependent on the VED level; the density of the parts printed with a VED of 192 J/mm3 was 150% higher than that of those printed with VED of 43 J/mm3. The percentage of internal defects ranged from 0.3 to 2.1% and was directly proportional to the VED level. The mechanical strength was more dependent on the overall density rather than the internal defects. Thus, parts printed at VED of 192 J/mm3 had an almost 200% higher apparent compressive modulus and peak strength compared to those printed at VED of 43 J/mm3. In addition, a finite element model has been developed using R . The numerical results were in good agreement with the experimental ABAQUSV data and may be used to make predictions for different gyroid designs. Keywords lattice structures, selective laser melting, mechanical properties, internal defects, gyroids, finite element analysis

Introduction Titanium (Ti) and its alloys have excellent biocompatibility, low stiffness, high specific strength, and high corrosion resistance. These advantages make these alloys superior for biomedical applications over other metallic alloys such as stainless steels and cobalt-chromium (Co-Cr) alloys.1 Moreover, the breakthrough in additive manufacturing technology, specifically selective laser melting (SLM), makes titanium alloys excellent candidates for manufacturing biomedical implants.2 Since titanium alloys are hard to machine and relatively expensive, SLM offers a much easier way to manufacture implants. Conventional processes, such as casting and forging, are energy consuming and incapable of producing the required customization that is easily provided by SLM processes.3 Furthermore, SLM provides shorter production time and more advanced functionality, such as the option of printing the parts in the form of lattice structures or functionally graded materials.4 A major challenge in the SLM process is the rapid solidification that results in highly nonequilibrium microstructures.5 Therefore, much work is directed to correlate SLM process parameters to the resulting microstructure and internal defects of the printed titanium parts.6–8 Numerical modeling can also help in predicting the microstructure features of metallic parts printed by SLM.9 The most critical parameters usually considered are those involved in the volumetric energy density (VED),10 namely, laser power, scan speed, layer thickness, and hatch spacing. Due to the high variability of the process, feedstock powder, and shape of the printed parts, it might be difficult to optimize a set of process parameters of volumetric energy density.5 For example, Kasperovich et al.11 defined a VED value of 117 J/mm3 to obtain fully dense Ti6Al4V parts, whereas Pal et al.12 reported the highest density of bulk parts printed at a VED level of 68 J/mm3. .

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

Some important considerations are the geometry and temperature of melt pool created by the interaction between the laser and metallic powder. Yang et al.13 classified the melting mode in SLM into two types: conduction mode and keyhole mode. The conduction mode is defined by a low to moderate VED, whereas the keyhole mode is associated with high VED capable of penetrating previously melted layers. While each mode has its pros and cons, a trade-off between both modes should be reached to ensure the adequate performance of printed parts. The melt pool mode affects the formation of internal porosities, which are crucial to the performance of the printed parts. In this regard, the effect of shielding gas flow velocity on the melt pool size has been studied.14 It was found that below a certain threshold, the melt pool becomes unstable and its width increases significantly. Pegues et al.15 investigated the influence of internal defects on the tensile properties of bulk parts and concluded that their effect on the ductility was greater than on the yield and tensile strength. Gong et al.6 concluded that internal porosity less than 1% has no effect on the mechanical properties; however, a higher percentage (1% to 5%) may significantly affect tensile strength, fatigue, and hardness of the parts. A major difference between printing lattice structures and bulk parts is that the lattice part contains porosities designed for certain applications. These intended porosities make the laser scan vector relatively shorter, thus resulting in less residual stresses in lattice structures.16 Some efforts have been made to investigate the effect of SLM process parameters on the dimensional and mechanical properties of computer aided design (CAD)-based lattice structures.17–19 It can be concluded that laser power plays a critical role in determining the lattice strut size because it affects the melt pool temperature and size. However, the effect of the SLM process parameters on the quality of complicated structures such as triply periodic minimal surface (TPMS) lattices is usually overlooked. TPMS lattice structures are identified by having a zero-mean curvature at all points. They are known to have biomimetic features.20,21 They have interconnected porosities, which make them a good candidate to be used as a bone replacement. Much effort is being made to investigate the pore size22 and functionally graded porosity23,24 of TPMS structures on the mechanical properties of these parts. However, very few studies25 cover how SLM process parameters affect the microstructure, geometry, and internal defects. A gap in understanding the effect of the defects on the mechanical properties is present. The main contribution of this work is to study the effect of VED on the quality of complicated gyroid lattice structures. The microstructure and geometrical and internal defects resulting from different VED values have been analyzed. Furthermore, the effect of these defects on the mechanical properties of lattice structures has been studied and compared to similar Ti6Al4V lattice structures from the literature. Finally, an FEA model was proposed to be used to predict the mechanical properties of those gyroid structures.

.

273

274

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Materials and Methods DESIGN AND 3D PRINTING

TPMS lattice structure designs were tested in this work, namely the cellular gyroid unit cells. These designs are biomimetic and well suited for biomedical implant applications for various reasons, including high surface area to volume ratio, interconnected porosity, and adequate mechanical strength.26 Figure 1 represents the cylindrical-shaped gyroid lattice structures possessing a diameter of 15 mm and a length of 23.5 mm. Each unit cell was designed to have a length of 1.5 mm; the design relative density is 0.20%. Plasma atomized Ti6Al4V powder having an average particle size of 45 lm was used as a feedstock material; the powder composition is given in table 1. An AM400 Renishaw machine equipped with an Nd:YAG laser with a maximum power of 400 W was used to print the lattice structures. The Renishaw machine works by a pulsed modulated laser, where the laser path is defined by a series of exposures overlapping each other. A pulsed laser is recommended for small components such as lattice structures with small features because this technique provides more control over the heat input compared to a continuous laser.27 The

FIG. 1 (A) Gyroid part design and as-built parts using SLM process at different VED levels: (B) 43 J/mm3, (C) 103 J/mm3, and (D) 192 J/mm3.

TABLE 1 Chemical composition (wt.%) of the Ti6Al4V powder

Al

5.50–6.50

V

Fe

O

C

N

H

Y

Ti

3.50–4.50

 0.25

 0.13

 0.08

 0.05

 0.012

 0.005

Bal.

Note: Al ¼ aluminum; V ¼ vanadium; Fe ¼ iron; O ¼ oxygen; C ¼ carbon; N ¼ nitrogen; H ¼ hydrogen; Y ¼ yttrium. .

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

TABLE 2 Process parameters and calculated VED for each sample

Samples

VED (J/mm3)

Laser Power (W)

Exposure Time (ls)

Low VED

100

40

43

Med VED

200

50

103

High VED

300

65

192

distance between the exposure points is known as a point distance. In this work, a combination of laser power and exposure time values are used to obtain three levels of VED. The effects of these levels on the microstructure, internal defects, and—in turn—on the mechanical properties of gyroid lattice structures of SLM-Ti6Al4V, have been studied in detail. According to Karimi et al.,28 VED can be calculated using equation (1). Table 2 shows the calculated VED levels. In this study, a value of 10 ls was added to each exposure time to account for delay time between each exposure and the other. VED ¼ Point distance Exposure time

Laser Power  hatch distance  layer thickness

(1)

Microstructure Characterization

According to the metallographic preparation standard, samples were sectioned from the parts built at three VED levels, mounted, ground, mechanically polished, and then etched by Kroll’s reagent. For the microstructural characterization, a JOEL SEM microscope and a Nikon optical microscope were used. Moreover, the X-ray diffraction (XRD) was carried out by a Bruker D8 DISCOVER XRD instrument provided with a cobalt sealed tube source. Morphology and Internal Defects Characterization

A Skyscan 1172 was used to perform microcomputed tomography (microCT) scanning for the printed parts. In this work, 100 kV was used, with a 0.5-mm Al and 0.038 Cu filter combination. The size and the position of the printed part relative to the X-ray source and the detector determined the pixel size. Although less pixel size can be obtained by scanning a small bisected part of the sample, difficulties associated with part size, beam hardening, and scattering are some limiting factors. In this study, a pixel size of 8 lm was achieved to scan the whole part. The specimen was rotated around the z-axis in 1,600 steps, with a 3-s exposure for each orientation. The analysis of the CT scans was performed using CTAn and CTvol software suites. MECHANICAL TESTING

Uniaxial compression tests were carried out using an MTS testing machine with a 50-kN load cell, in accordance with the ISO 13314 standard. Crosshead .

275

276

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

displacement was set to be 2 mm/min (strain rate of 0.001 mm/s). Three samples were tested at each VED level to ensure the accuracy of the collected data. The stresses were calculated by multiplying the forces recorded from the load cell by the cross-section area. The strain was measured using crosshead displacement (taking into consideration compliance correction) and a video camera for some specimens. The difference between both results was less than 3%. Therefore, all the data collected in this work were obtained by the crosshead displacement after a compliance correction. FINITE ELEMENT ANALYSIS

To understand the effect of microstructure and morphology on the stress-strain relationship of the gyroids structures, FEA was performed using ABAQUS/ EXPLICIT-6.14 to simulate the quasistatic compression. In this study, an elasticplastic model was used in the simulation for the material model based on the experimental results for as-built properties obtained by Cain et al.;29 future studies would consider the failure criteria. Two rigid plates simulated the compression platens. The displacement of the upper plate was defined to match that of the experimental and maintain the same strain rate, whereas the lower plate was fixed. The gyroid models were meshed using voxel meshing techniques. Voxel mesh is usually used to model biomedical parts because the complexity and highly detailed parts are usually hard to model with conventional mesh shapes.30 Another advantage of voxel meshing is that it requires less computational time and gives a straightforward solution for modeling highly complex structures.31 Finite-sized models are usually suggested for lattice structures,32 in which instead of modeling the whole part, only a small number of elements are modeled. For example, Quevedo Gonza´lez and Nun˜o33 found that modeling eight-by-eight unit cells minimizes the difference between the whole model and the finite model. In this work, to reduce the model size, only a quarter of the sample height was considered. The model was composed of 3,002,786 elements, with an element size of 35 lm. The element size chosen was based on a mesh sensitivity analysis that ensured the convergence of the stiffness and peak strength.

Results and Discussion MICROSTRUCTURE

The optical micrographs of the SLM fabricated Ti6Al4V gyroid parts, printed at three VED levels along the build direction are shown in figure 2A–C. Very fine laths representing acicular a0 -martensite can be observed within the columnar structures. The formation of these fine laths is due to the very high cooling rates of the SLM process. The elongated columnar structures represent the prior b grains formed upon solidification. The optical micrographs for the top surface of the printed gyroid parts are seen in figure 2D–F. It can be noted that the microstructures of the top surface and the cross-sectional surface are similar. .

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

FIG. 2 Optical micrographs (100) of the Ti6Al4V gyroids printed at three VED levels: (A), (D) at 43 J/mm3; (B), (E) at 103 J/mm3; and (C), (F) at 192 J/mm3. (A), (B), and (C) were taken along the build direction. (D), (E), and (F) were taken from the top side. Scale bars are equal to 100 μm.

The SEM micrographs of the Ti6Al4V gyroid parts printed at three VED levels are illustrated in figure 3. Generally, the microstructure of as-built parts consists of a lath-type a0 -martensite; however, its size slightly increases as the VED increases. This behavior may be attributed to a reduction in the cooling rate as the VED increases.34 According to its size,35 a0 -martensite can be classified into four types: primary, secondary, tertiary, and quaternary. The laths of primary martensite are the biggest, whereas those of the quaternary are the finest, being distributed among

FIG. 3 SEM images (5,000) of the gyroid parts printed at different VED levels: (A) 43 J/mm3, (B) 103 J/mm3, and (C) 192 J/mm3.

.

277

278

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

the laths of the tertiary martensite. The laths of secondary martensite, comparatively denser and smaller than those of the primary martensite, lie perpendicular or at about 45 to them. All these types of a0 -martensite can be seen in figure 3. The results of this work are in agreement with those obtained by Gong et al.,6 indicating that the VED does not have a significant effect on the microstructure evolution, which is basically governed by the cooling rate. The XRD spectrum of the samples processed at three VED levels is presented in figure 4. Similar diffraction patterns can be seen in which all the peaks can be identified as a0 -martensite having the same crystal structure of hexagonal closepacked (hcp) type. This may be expected due to extremely fast cooling rates occurring during the solid-state phase transformation, thus leading to a diffusionless transformation of the high-temperature b-phase into a fully a0 -martensite. However, the peaks in the sample processed at a VED of 43 J/mm3 are slightly broader relative to the same peaks of the samples processed at 103 and 192 J/mm3. This broadening of the peaks may be related to very fast cooling rates that occur during the SLM process, thus hindering the diffusion of vanadium from a0 -martensite. The supersaturation of vanadium leads to a crystal deformation and broadening of the related peaks.36 MORPHOLOGY

MicroCT scanning was used to evaluate the internal porosity of the parts. The parts were cut into smaller lengths (10 mm) to make sure adequate resolution could be obtained. The struts of the gyroid parts were all characterized using microCT; the trabecular thickness measurement for each strut was carried out using the

FIG. 4 XRD analysis for the Ti6Al4V gyroid parts printed at different VED levels.

.

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

R , whereas the bone to tissue two-dimensional (2D) analysis tool pack in CTAnV ratio was measured using the 3D analysis tool pack to evaluate the relative density of each part. Figure 5 shows the design of a cross section of the gyroid parts along with the microCT slices; all the slices were taken at a constant height from the base of the parts (height ¼ 9.5 mm). As the VED level increased, the white area percentage increased. This indicates that the percentage of melted Ti6Al4V increased. This was noted by the increase in relative density and strut size listed in table 3. However, at a VED level of 43 J/mm3, some broken and incomplete struts can be seen. This might be attributed to the fact that the energy was insufficient to melt the powder. Therefore, some balling may occur along the melt track as well as lack of fusion between successive layers that might not be visible in the cross section but that will result in incomplete struts or empty gaps instead of solid struts. As the VED level increases, more energy is provided to the powder; and therefore, the melt pool size increases, leading to a thicker strut and higher relative density for the entire part.

Internal Porosity Figure 6 represents a reconstructed gyroid unit cell from the different parts printed at different VED levels. The internal porosity was calculated from the whole parts and reported in table 3. A 3D object analysis tool pack in the software was used to

FIG. 5 MicroCT slices taken from a constant height ¼ 9.5 mm from base of the gyroid parts at different VED levels: (A) design, (B) 43 J/mm3, (C) 103 J/mm3, and (D) 192 J/mm3.

TABLE 3 Apparent density, strut size, and internal defect percentage of Ti6Al4V gyroid parts at three VED levels

Apparent Density (%)

Average Strut Size (lm)

St. Dev. (lm)

43

17

271.3173

8.8607

0.323

103

31

293.7477

11.7801

0.626

193

40

302.1761

1.3431

2.11

VED (J/mm3)

.

Internal Porosity (%)

279

280

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 6 Reconstructed gyroid unit cells at different VED levels: (A) 43 J/mm3, (B) 103 J/mm3, and (C) 192 J/mm3.

analyze equivalent pore diameters. A histogram of the pore frequency distribution along the whole part is illustrated in figure 7, where a random distribution with no specific pattern was observed. Figure 8 represents the histogram distribution for the equivalent pore diameter and the frequency in each part along the build direction. Parts printed at VED 43 J/mm3 were found to have the least frequency of internal porosities. As the VED increased to 103 J/mm3, the strut size increased, and internal porosity started to increase as well. As the VED reached 192 J/mm3, the strut size and the internal porosity became bigger. It should be noted that the resolution of the microCT was 8 lm for each pixel in this study. Thus, the minimum diameter of internal porosity that could be detected for these parts was limited to two times the resolution

FIG. 7 Pore frequency distribution along with the build directions of gyroid lattice structures printed at three VED levels: (A) 43 J/mm3, (B) 103 J/mm3, and (C) 192 J/mm3.

.

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

FIG. 8 Internal pore size distribution for equivalent pore diameter in gyroid lattice structures printed at different VED levels: (A) 43 J/mm3, (B) 103 J/mm3, and (C) 192 J/mm3.

(16 lm), which is the cutoff usually used in the literature.37,38 The most common repeatable equivalent pore diameter was 40 lm; the highest frequency was around 7,000 in the gyroid parts printed with high VED level. According to Vilaro, Colin, and Bartout,39 most pores between 10 and 50 lm may be attributed to entrapped gas during the melting process. These errors can be reduced by increasing the apparent density of the powder bed or by reducing the inert gas pressure while printing. In addition, the aspect ratio of these pores within the three parts was analyzed as well; aspect ratio was defined as the minimum diameter divided by the maximum diameter.40 Therefore, a perfect spherical porosity would have an aspect ratio of one. As the shape of the pore becomes distorted, the aspect ratio decreases. For the parts printed at a VED of 43 J/mm3, the aspect ratio of 0.9 to 1 was noted to be the most common ratio of these porosities, as shown in figure 9. As the level of VED increased, the aspect ratio dropped. Parts printed at 192 J/mm3 were reported to have an aspect ratio reaching 0.6. Although melt pool temperature and geometry have not been measured in this study, the change in the frequency and shape of

FIG. 9 Aspect ratio for internal pores of the different gyroid lattice structures printed at different VED levels: (A) 43 J/mm3, (B) 103 J/mm3, and (C) 192 J/mm3.

.

281

282

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

internal porosities is usually attributed to the change in melt pool condition from conduction to keyhole.41 This is related to the amount of energy delivered to the powder and the laser power; high laser power creates a spatter in the molten metal, causing a pore formation that looks like a keyhole. MECHANICAL PROPERTIES Figure 10 depicts the stress-strain diagrams of the samples of Ti6Al4V gyroid lattice

structures printed at different VED levels. The typical stress-strain diagram of lattice structures is divided into three zones: elastic zone, plateau zone, and densification zone. For biomedical applications, four important values are usually considered: the apparent modulus of elasticity, the apparent compressive strength, the elastic strain, and the densification strain. Parts printed at a low VED of 43 J/mm3 had the lowest apparent modulus of elasticity and compressive strength but the highest elastic strain. As the VED level increased, both the apparent modulus of elasticity and compressive strength were seen to increase. On the other hand, elastic strain and densification strain were noted to decrease. In our previous study,42 we investigated the effect of morphology on the mechanical properties. It was noted that as the VED increased, the strut thickness increased, thereby enhancing the strength of the gyroid structure. In this study, we have examined the microstructure and phase analysis of the gyroid parts printed at different VED levels. It can, therefore, be assumed that the internal defects do not affect the apparent modulus of elasticity and compressive strength; however, they do affect the elastic and densification strain.

FIG. 10 Stress–strain diagram of the gyroid parts at different VED levels.

.

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

TABLE 4 Apparent modulus of elasticity and compressive strength of the gyroid parts at different VED levels

VED (J/mm3)

Apparent Modulus of Elasticity (GPa)

Compressive Strength

43

1.61 6 0.17

46.34 6 2.34

103

2.63 6 0.32

73.45 6 0.67

193

5.34 6 25

127.54 6 1.11

The apparent modulus of elasticity and peak compressive strength of the printed gyroids are listed in table 4. By comparing other TPMS lattice structures, it was noted that the mechanical properties were highly related to the relative density. For example, Maszybrocka et al.25 printed TPMS lattice structures (Schwartz diamonds) with a relative density between 34% and 60%. The printed structures showed an apparent modulus of elasticity between 3.1 and 12.9 GPa and a compressive strength ranging from 62 to 200 MPa. Moreover, the effect of using Ti6Al4V gyroid sheets lattice structures was investigated by Yang et al.43 They obtained a wide range of apparent modulus of elasticity between 0.5 and 15 GPa and a compressive strength between 21.7 and 338 MPa. In addition, the electron beam melting (EBM) process was used to print gyroid lattice structures44 with relative density around 20%; the apparent modulus of elasticity and compressive strength were found to be 1 GPa and 20 MPa, respectively. Depending on the requirements of the implant and its position in the body, the right choice of relative density, apparent modulus of elasticity, and compressive strength can be determined. The mechanical properties of as-built SLM parts can be improved through postprocessing operations such as sandblasting45 and solution heat treatment.46 The FEA model was developed to predict the mechanical properties of the printed gyroids. The FEA results were compared to the gyroid parts printed at the VED level of 103 J/mm3 because these parts had the closest relative density to the design value. The results of the numerical analysis are in agreement with the experimental ones and are illustrated in figure 11; an error of 10% is found between the results. The material model might be one of the factors contributing to the difference between the FEA model and the experimental results. Most of the material models used right now in the literature are adopted from materials manufactured by forging or casting. Very limited work focuses on obtaining the material model based on SLM-fabricated Ti6Al4V. Another factor contributing to the difference is that the model size was finite; infinite models would give more accurate results but are computationally expensive. Finally, more studies need to be done on voxel meshing and comparing it to tetrahedral meshes to evaluate which represents the material and lattice geometries in more depth. Moreover, the internal defects were not accounted for in this model. In our future studies, the internal defects and more appropriate material models representing the printed Ti6Al4V material will be adopted to enhance the accuracy of the developed model. .

283

284

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 11 Experimental versus FEA stress–strain diagram for gyroid parts.

Conclusions In this work, the effects of three VED levels (43, 103, and 192 J/mm3) on the microstructure, internal porosities, and mechanical properties of the SLM fabricated Ti6Al4V gyroid lattice structures have been studied. The energy levels have been chosen carefully from the literature to cover different varieties and widen the research scope. The main conclusions are as follows: 1. The VED levels included in this study had no significant effect on the microstructure of the printed gyroid lattice structures. The expected martensitic structure is related to the fast cooling rate of the SLM process. 2 The relative density and strut size are related to the amount of VED provided to the powder. A VED level of 43 J/mm3 was found to be insufficient based on the incomplete struts observed. At a VED of 192 J/mm3, the measured apparent density was doubled compared to the design value, whichwas related to the oversized melt pool geometry. 3. The gyroid lattice structures printed at a VED of 192 J/mm3 had the highest internal porosities, reaching 2.1%, as well as the highest frequency of irregular pores. These irregularly shaped porosities are usually attributed to the keyhole formation due to both the high laser power and VED level. 4. The lattice structures fabricated at a VED level of 192 J/mm3 showed the highest apparent modulus of elasticity and compressive strength due to the thickest strut size and greatest apparent density of the parts printed at that .

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

VED level. These parts also had the least elastic strain due to the presence of the highest frequency of internal defects. These defects act as stress concentrators and facilitate crack initiation. Future work would include the update of the material model to mimic the properties of printed Ti6Al4V. Moreover, the inclusion of internal defects in the FEA model would help in obtaining more accurate results and would reduce the differences between the numerical and experimental work. ACKNOWLEDGMENTS

The authors would like to acknowledge the support of the Natural Sciences and Engineering Research Council of Canada (NSERC; funding reference number 518494).

References 1.

2.

3.

4.

5.

6.

7.

8.

9.

.

L.-C. Zhang and L.-Y. Chen, “A Review on Biomedical Titanium Alloys: Recent Progress and Prospect,” Advanced Engineering Materials 21 (2019): 1801215–1801244, https:// doi.org/10.1002/adem.201801215 W. S. W. Harun, M. S. I. N. Kamariah, N. Muhamad, S. A. C. Ghani, F. Ahmad, and Z. Mohamed, “A Review of Powder Additive Manufacturing Processes for Metallic Biomaterials,” Powder Technology 327 (2018): 128–151, https://doi.org/10.1016/J.POWTEC.2017.12.058 L.-C. Zhang and H. Attar, “Selective Laser Melting of Titanium Alloys and Titanium Matrix Composites for Biomedical Applications: A Review,” Advanced Engineering Materials 18, no. 4 (2016): 463–475, https://doi.org/10.1002/adem.201500419 D. Mahmoud and M. Elbestawi, “Lattice Structures and Functionally Graded Materials Applications in Additive Manufacturing of Orthopedic Implants: A Review,” Journal of Manufacturing and Materials Processing 1, no. 2 (2017): 13–32, https://doi.org/10.3390/ jmmp1020013 H. Shipley, D. McDonnell, M. Culleton, R. Coull, R. Lupoi, G. O0 Donnell, and D. Trimble, “Optimisation of Process Parameters to Address Fundamental Challenges during Selective Laser Melting of Ti-6Al-4V: A Review,” International Journal of Machine Tools and Manufacture 128 (2018): 1–20, https://doi.org/10.1016/j.ijmachtools.2018.01.003 H. Gong, K. Rafi, H. Gu, G. D. Janaki Ram, T. Starr, and B. Stucker, “Influence of Defects on Mechanical Properties of Ti-6Al-4V Components Produced by Selective Laser Melting and Electron Beam Melting,” Materials and Design 86 (2015): 545–554, https://doi.org/ 10.1016/j.matdes.2015.07.147 X. Zhao, S. Li, M. Zhang, Y. Liu, T. B. Sercombe, S. Wang, Y. Hao, R. Yang, and L. E. Murr, “Comparison of the Microstructures and Mechanical Properties of Ti–6Al– 4V Fabricated by Selective Laser Melting and Electron Beam Melting,” Materials & Design 95 (2015): 21–31, https://doi.org/10.1016/j.matdes.2015.12.135 H. Gong, K. Rafi, H. Gu, T. Starr, and B. Stucker, “Analysis of Defect Generation in Ti-6Al4V Parts Made Using Powder Bed Fusion Additive Manufacturing Processes,” Additive Manufacturing 1 (2014): 87–98, https://doi.org/10.1016/j.addma.2014.08.002 J. Heang Kuan Tan, S. Leong Sing, and W. Yee Yeong, “Microstructure Modelling for Metallic Additive Manufacturing: A Review,” Virtual and Physical Prototyping 15, no. 1 (2020), https://doi.org/10.1080/17452759.2019.1677345

285

286

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

.

J. Ciurana, L. Hernandez, and J. Delgado, “Energy Density Analysis on Single Tracks Formed by Selective Laser Melting with CoCrMo Powder Material,” The International Journal of Advanced Manufacturing Technology 68 (2013), 1103–1110, https://doi.org/ 10.1007/s00170-013-4902-4 G. Kasperovich, J. Haubrich, J. Gussone, and G. Requena, “Correlation between Porosity and Processing Parameters in TiAl6V4 Produced by Selective Laser Melting,” Materials and Design 105 (2016), 160–170, https://doi.org/10.1016/j.matdes.2016.05.070 S. Pal, N. Gubeljak, R. Hudak, G. Lojen, V. Rajtukova, J. Predan, V. Kokol, and I. Drstvensek, “Tensile Properties of Selective Laser Melting Products Affected by Building Orientation and Energy Density,” Materials Science and Engineering A 743 (2019), 637–647, https://doi.org/10.1016/j.msea.2018.11.130 J. Yang, J. Han, H. Yu, J. Yin, M. Gao, Z. Wang, and X. Zeng, “Role of Molten Pool Mode on Formability, Microstructure and Mechanical Properties of Selective Laser Melted Ti6Al-4V Alloy,” Materials and Design 110 (2016), 558–570, https://doi.org/10.1016/ j.matdes.2016.08.036 J. Reijonen, A. Revuelta, T. Riipinen, K. Ruusuvuori, and P. Puukko, “On the Effect of Shielding Gas Flow on Porosity and Melt Pool Geometry in Laser Powder Bed Fusion Additive Manufacturing,” Additive Manufacturing 32 (2020), https://doi.org/10.1016/ j.addma.2019.101030 J. Pegues, K. Leung, A. Keshtgar, L. Airoldi, N. Apetre, N. Iyyer, and N. Shamsaei, “Effect of Process Parameter Variation on Microstructure and Mechanical Properties of Additively Manufactured Ti-6Al-4V,” in Proceedings of the 28th Annual International Solid Freeform Fabrication Symposium (Austin, TX: University of Texas, 2017), 61–74. O. Rehme and C. Emmelmann, “Rapid Manufacturing of Lattice Structures with Selective Laser Melting,” in Proceedings Volume 6107, Laser-Based Micropackaging (Bellingham, WA: International Society for Optics and Photonics, 2006), https://doi.org/10.1117/ 12.645848 S. L. Sing, F. E. Wiria, and W. Y. Yeong, “Selective Laser Melting of Lattice Structures: A Statistical Approach to Manufacturability and Mechanical Behavior,” Robotics and Computer-Integrated Manufacturing 49 (2018), 170–180, https://doi.org/10.1016/ j.rcim.2017.06.006 S. M. Ahmadi, R. Hedayati, R. K. Ashok Kumar Jain, Y. Li, S. Leeflang, and A. A. Zadpoor, “Effects of Laser Processing Parameters on the Mechanical Properties, Topology, and Microstructure of Additively Manufactured Porous Metallic Biomaterials: A Vector-Based Approach,” Materials and Design 134 (2017), 234–243, https://doi.org/10.1016/ j.matdes.2017.08.046 H. Liang, D. Xie, Y. Mao, J. Shi, C. Wang, L. Shen, and Z. Tian, “The Size Effect on Forming Quality of Ti–6Al–4V Solid Struts Fabricated via Laser Powder Bed Fusion,” Metals 9, no. 4 (2019), https://doi.org/10.3390/met9040416 O. Al-Ketan, D. W. Lee, R. Rowshan, and R. K. Abu Al-Rub, “Functionally Graded and Multi-Morphology Sheet TPMS Lattices: Design, Manufacturing, and Mechanical Properties,” Journal of the Mechanical Behavior of Biomedical Materials 102 (2020), https:// doi.org/10.1016/j.jmbbm.2019.103520 L. Li, “Early Osteointegration Evaluation of Porous Ti6Al4V Scaffolds Designed Based on Triply Periodic Minimal Surface Models,” Journal of Orthopaedic Translation 19 (2019), 94–105, https://doi.org/10.1016/j.jot.2019.03.003 Z. Cai, Z. Liu, X. Hu, H. Kuang, and J. Zhai, “The Effect of Porosity on the Mechanical Properties of 3D-Printed Triply Periodic Minimal Surface (TPMS) Bioscaffold,” BioDesign and Manufacturing 2 (2019), 242–255, https://doi.org/10.1007/s42242-01900054-7

MAHMOUD ET AL., DOI: 10.1520/STP163120190125

23.

24.

25.

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

.

C. Peng and P. Tran, “Bioinspired Functionally Graded Gyroid Sandwich Panel Subjected to Impulsive Loadings,” Composites Part B: Engineering 188 (2020), https://doi.org/ 10.1016/j.compositesb.2020.107773 L. Yang, M. Ferrucci, R. Mertens, W. Dewulf, C. Yan, Y. Shi, and S. Yang, “An Investigation into the Effect of Gradients on the Manufacturing Fidelity of Triply Periodic Minimal Surface Structures with Graded Density Fabricated by Selective Laser Melting,” Journal of Materials Processing Technology 275 (2020), https://doi.org/10.1016/j.jmatprotec.2019.116367 J. Maszybrocka, B. Gapin ´ski, M. Dworak, G. Skrabalak, and A. Stwora, “The Manufacturability and Compression Properties of the Schwarz Diamond Type Ti6Al4V Cellular Lattice Fabricated by Selective Laser Melting,” The International Journal of Advanced Manufacturing Technology 105 (2019), 3411–3425, https://doi.org/10.1007/s00170-01904422-6 A. Ya ´nez, A. Herrera, O. Martel, D. Monopoli, and H. Afonso, “Compressive Behaviour of Gyroid Lattice Structures for Human Cancellous Bone Implant Applications,” Materials Science and Engineering C 68 (2016), 445–448, https://doi.org/10.1016/ j.msec.2016.06.016 A. G. Demir, P. Colombo, and B. Previtali, “From Pulsed to Continuous Wave Emission in SLM with Contemporary Fiber Laser Sources: Effect of Temporal and Spatial Pulse Overlap in Part Quality,” International Journal of Advanced Manufacturing Technology 91, nos. 5–8 (2017), 2701–2714, https://doi.org/10.1007/s00170-016-9948-7 P. Karimi, T. Raza, J. Andersson, and L. E. Svensson, “Influence of Laser Exposure Time and Point Distance on 75-:m-Thick Layer of Selective Laser Melted Alloy 718,” International Journal of Advanced Manufacturing Technology 94, nos. 5–8 (2018), 2199–2207, https://doi.org/10.1007/s00170-017-1019-1 V. Cain, L. Thijs, J. Van Humbeeck, B. Van Hooreweder, and R. Knutsen, “Crack Propagation and Fracture Toughness of Ti6Al4V Alloy Produced by Selective Laser Melting,” Additive Manufacturing 5 (2015), 68–76, https://doi.org/10.1016/j.addma.2014.12.006 M. Babich and V. Kublanov, “Voxel Based Finite Element Method Modelling Framework for Electrical Stimulation Applications Using Open Source Software,” in Proceedings of the 2019 Ural Symposium on Biomedical Engineering, Radioelectronics and Information Technology, USBEREIT 2019 (Piscataway, NJ: Institute of Electrical and Electronics Engineers, 2019), 127–130, https://doi.org/10.1109/USBEREIT.2019.8736569 Q. Mao, K. Su, Y. Zhou, M. Hossaini-Zadeh, G. S. Lewis, and J. Du, “Voxel-Based MicroFinite Element Analysis of Dental Implants in a Human Cadaveric Mandible: Tissue Modulus Assignment and Sensitivity Analyses,” Journal of the Mechanical Behavior of Biomedical Materials 94 (2019), 229–237, https://doi.org/10.1016/j.jmbbm.2019.03.008 H. Mehboob, F. Tarlochan, A. Mehboob, and S. H. Chang, “Finite Element Modelling and Characterization of 3D Cellular Microstructures for the Design of a Cementless Biomimetic Porous Hip Stem,” Materials and Design 149 (2018), 101–112, https://doi.org/ 10.1016/j.matdes.2018.04.002 F. J. Quevedo Gonza ´lez and N. Nun ˜o, “Finite Element Modelling Approaches for WellOrdered Porous Metallic Materials for Orthopaedic Applications: Cost Effectiveness and Geometrical Considerations,” Computer Methods in Biomechanics and Biomedical Engineering 19, no. 8 (2016), 845–854, https://doi.org/10.1080/10255842.2015.1075009 D. K. Do and P. Li, “The Effect of Laser Energy Input on the Microstructure, Physical and Mechanical Properties of Ti-6Al-4V Alloys by Selective Laser Melting,” Virtual and Physical Prototyping 11, no. 1 (2016), 41–47, https://doi.org/10.1080/17452759.2016.1142215 J. Yang, H. Yu, J. Yin, M. Gao, Z. Wang, and X. Zeng, “Formation and Control of Martensite in Ti-6Al-4V Alloy Produced by Selective Laser Melting,” Materials and Design 108 (2016), 308–318, https://doi.org/10.1016/j.matdes.2016.06.117

287

288

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

36.

37.

38.

39.

40.

41.

42.

43.

44.

45.

46.

.

E. Sallica-Leva, R. Caram, A. L. Jardini, and J. B. Fogagnolo, “Ductility Improvement Due to Martensite a0 Decomposition in Porous Ti-6Al-4V Parts Produced by Selective Laser Melting for Orthopedic Implants,” Journal of the Mechanical Behavior of Biomedical Materials 54 (2016), 149–158, https://doi.org/10.1016/j.jmbbm.2015.09.020 S. Tammas-Williams, H. Zhao, F. Le´onard, F. Derguti, I. Todd, and P. B. Prangnell, “XCT Analysis of the Influence of Melt Strategies on Defect Population in Ti-6Al-4V Components Manufactured by Selective Electron Beam Melting,” Materials Characterization 102 (2015), 47–61, https://doi.org/10.1016/j.matchar.2015.02.008 G. Zio ´łkowski, E. Chlebus, P. Szymczyk, and J. Kurzac, “Application of X-Ray CT Method for Discontinuity and Porosity Detection in 316L Stainless Steel Parts Produced with SLM Technology,” Archives of Civil and Mechanical Engineering 14, no. 4 (2014), 608– 614, https://doi.org/10.1016/j.acme.2014.02.003 T. Vilaro, C. Colin, and J. D. Bartout, “As-Fabricated and Heat-Treated Microstructures of the Ti-6Al-4V Alloy Processed by Selective Laser Melting,” Metallurgical and Materials Transactions A 42, no. 10 (2011), 3190–3199, https://doi.org/10.1007/s11661-011-0731-y I. Maskery, N. T. Aboulkhair, M. R. Corfield, C. Tuck, A. T. Clare, R. K. Leach, R. D. Wildman, I. A. Ashcroft, and R. J. M. Hague, “Quantification and Characterisation of Porosity in Selectively Laser Melted Al-Si10-Mg Using X-Ray Computed Tomography,” Materials Characterization 111 (2016), 193–204, https://doi.org/10.1016/j.matchar.2015.12.001 R. Cunningham, C. Zhao, N. Parab, C. Kantzos, J. Pauza, K. Fezza, T. Sun, and A. D. Rollett, “Keyhole Threshold and Morphology in Laser Melting Revealed by Ultrahigh-Speed X-Ray Imaging,” Science 363, no. 6429 (2019), 849–852, https://doi.org/10.1126/ science.aav4687 D. Mahmoud, M. A. Elbestawi, and B. Yu, “Process–Structure–Property Relationships in Selective Laser Melting of Porosity Graded Gyroids,” Journal of Medical Devices 13, no. 3 (2019), 031005, https://doi.org/10.1115/1.4043736 E. Yang, M. Leary, B. Lozanovski, D. Downing, M. Mazur, A. Sarker, A. M. Khorasani, et al., “Effect of Geometry on the Mechanical Properties of Ti-6Al-4V Gyroid Structures Fabricated via SLM: A Numerical Study,” Materials and Design 184 (2019), https://doi.org/ 10.1016/j.matdes.2019.108165 A. Ataee, Y. Li, D. Fraser, G. Song, and C. Wen, “Anisotropic Ti-6Al-4V Gyroid Scaffolds Manufactured by Electron Beam Melting (EBM) for Bone Implant Applications,” Materials and Design 137 (2018), 345–354, https://doi.org/10.1016/j.matdes.2017.10.040 L. Yang, C. Yan, W. Cao, Z. Liu, B. Song, S. Wen, C. Zhang, Y. Shi, and S. Yang, “Compression–Compression Fatigue Behaviour of Gyroid-Type Triply Periodic Minimal Surface Porous Structures Fabricated by Selective Laser Melting,” Acta Materialia 181 (2019), 49–66, https://doi.org/10.1016/j.actamat.2019.09.042 S. L. Sing, S. Huang, and W. Y. Yeong, “Effect of Solution Heat Treatment on Microstructure and Mechanical Properties of Laser Powder Bed Fusion Produced Cobalt-28Chromium-6Molybdenum,” Materials Science and Engineering A 769 (2020), https://doi.org/ 10.1016/j.msea.2019.138511

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190164

Matthew Wayne Sanders,1 Adam Rowe,1 and Suresh Divi1

Full-Scale High-Load, Thermal, and Fatigue Testing of Additive Manufactured Powder Bed Fusion Component for Oil Field Applications Citation M. W. Sanders, A. Rowe, and S. Divi, “Full-Scale High-Load, Thermal, and Fatigue Testing of Additive Manufactured Powder Bed Fusion Component for Oil Field Applications,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 289–307. http://doi.org/10.1520/ STP1631201901642

ABSTRACT

As the usage of additive manufacturing (AM) expands into more critical applications, the need to establish confidence in the expected performance and reliability of AM components also becomes more critical. Significant research and efforts have been made public related to the qualification of AM components for aerospace and medical applications; however, very little information has been presented with regard to the use of AM within the oil and gas industry. The harsh and demanding environments of oil and natural gas production present unique and challenging conditions for AM components to withstand. To help address this lack of information, a case study AM component was created to showcase the types of features that can be created using the AM process while designing for oil field conditions. An Alloy 625 laser powder bed fusion printed component was created and analyzed via a finite element model (FEM) and then statically load tested and fatigue tested to simulate typical oil field conditions. Various properties, including hardness, were documented along with the microstructure. Corrosion testing was also performed to compare the critical pitting temperature

Manuscript received December 21, 2019; accepted for publication February 6, 2020. 1 Stress Engineering Services, Inc., 13800 Westfair East Dr., Houston, TX 77041, USA 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

289

290

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

of the Alloy 625 AM material to a traditional wrought Alloy 625 material. Fullscale tests performed included axial compression loading to more than 80,000 lb, rotational bending fatigue testing to more than 10 million cycles, combined load testing of 5,000 ftlb torque and bending, flame impingement, and rapid cryogenic temperature cyclizing. After each testing stage, the part was inspected for crack indications. The compression test was monitored using advanced digital image correlation (DIC) to monitor the strain deformation of the part during testing. The results of the testing were compared to the FEM using the DIC data and found to be in good agreement. Keywords load testing, fatigue testing, thermal shock loading, additive manufacturing, finite element analysis (FEA), digital image correlation (DIC), corrosion testing, metallurgy, Alloy 625, oil field

Introduction In 2018, more than 67% of the energy consumed in the United States was derived from petroleum or natural gas.1 Even with the important advancements in nonpetroleum-based energy sources such as wind, solar, and nuclear, oil and natural gas will continue to supply a large proportion of the energy consumed in the United States for the foreseeable future. As a result, significant effort will continue to be focused on making oil and natural gas production as efficient as possible. Many new technologies are being explored by the industry. One such new technology is the use of additive manufacturing (AM) processes, which are sometimes referred to as three-dimensional (3D) printing. AM is the process of joining successive layers of material to gradually construct a 3D object. AM is based on an approach that directly contrasts with traditional manufacturing, where manufacturing begins with a volume of material that is larger than the final component and then removes material to produce the desired final geometry. AM is known to provide a number of benefits as compared to traditional manufacturing techniques. AM can typically produce components faster and with more efficient use of material, resulting in less waste. One of the greatest advantages of AM is its ability to produce complex geometric shapes that cannot be machined using traditional techniques. The option to design geometric features that would be impractical to fabricate with traditional techniques means structures can be optimized with highly favorable properties, such as weight reduction and improved fluid flow. AM also offers the possibility of consolidating multiple assembled parts into one build, resulting in reduced part counts and easier assembly. The aerospace industry was one of the first to take advantage of the benefits of AM. More recently, advancements in the build size and build quality of AM parts have unlocked new potential uses in the oil and gas industry. The oil and gas industry could greatly benefit from AM’s ability to optimize flow characteristics in components and to produce parts on demand, resulting in shorter lead times and reduced inventory. .

SANDERS ET AL., DOI: 10.1520/STP163120190164

Because pressure containment is typically a significant concern for oil field equipment, a study was performed to investigate the performance of AM parts in pressure-containment applications. Stainless steel tubes were printed and then pressured to burst. As a control, commercially available ASTM A213, Standard Specification for Seamless Ferritic and Austenitic Alloy-Steel Boiler, Superheater, and HeatExchanger Tubes, annealed tubes were tested and found to burst near 25,217 psi, as shown in figure 1A. One type of laser powder bed fusion (LPBF) printed tube was found to burst near 31,673 psi (fig. 1B). However, another type of LPBF printed tube was observed to leak through the wall thickness (fig. 1C) at very low pressure (50 psi). This result was unexpected because the AM tubes were reportedly manufactured to meet the chemical composition and mechanical property requirements of ASTM F3184-16, Standard Specification for Additive Manufacturing Stainless Steel Alloy (UNS S31603) with Powder Bed Fusion.2 Even though material test coupons during the build showed acceptable yield strength (fig. 2), the material was observed to exhibit large amounts of porosity (fig. 3), which led to the inability of the tube to perform as intended (unbeknownst to the manufacturer). Based on the results of the pressure-containment study and the increased interest in AM for the oil and gas industry, it was desired to create a case study AM component for metallurgical examination as well as to investigate its performance under various oil field loading scenarios.

Materials and Methods A case study AM component was created that contained many unmachinable features (such as internal converging and diverging channels) that were only achievable through an AM process. A drawing of the case study component with some of the external surfaces removed is shown in figure 4.

FIG. 1 Photographs of failure modes for commercially available seamless tubing (A), AM tube Type 1 (B), and AM tube Type 2 (C).

.

291

292

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 2 Chart showing 0.2% yield strength of AM material compared to minimum requirements of ASTM F3184-16.

FIG. 3 Photograph of metallurgical AM specimen showing a large degree of porosity.

.

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 4 Two views of the case study component showing complex internal geometric features and traditionally unmachinable flow paths.

Figure 5 shows a photograph of the case study part after printing with its dimensions. A photograph of a different side of the part is shown in figure 6. Two sample parts were printed; one was subjected to metallurgical evaluation and corrosion testing, while the second was used only for full-scale testing. The parts were printed using the LPBF technique on an EOS M290 machine from Ni-325 powder. After printing, the parts were stress relieved per AMS 2774E at 190  F for 1 h. METALLURGICAL CHARACTERIZATION

One of the untested LPBF AM Alloy 625 samples was selected for metallurgical characterization and corrosion testing. A disk specimen was taken from the end of the part furthest from the build plate. Metallurgical characterization consisted of chemical analysis, microstructural examination, and hardness testing. The bulk chemical composition was measured using a portable X-ray fluorescence (XRF) analyzer. Table 1 lists the measured composition as compared to UNS N06625 (Alloy 625), which is identical for wrought (ASTM B446, Standard Specification for Nickel-Chromium-Molybdenum-Columbium Alloy [UNS N06625], NickelChromium-Molybdenum-Silicon Alloy [UNS N06219], and Nickel-ChromiumMolybdenum-Tungsten Alloy [UNS N06650] Rod and Bar) and PBF AM (ASTM F3056, Standard Specification for Additive Manufacturing Nickel Alloy [UNS N06625] with Powder Bed Fusion) parts. Carbon, sulfur, and phosphorus were below the limits of detection for the XRF system and are not shown. Future work should employ chemical analysis techniques capable of detecting these elements, as well as oxygen, which can be introduced through powder fusion processes. While .

293

294

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Photograph of case study AM component showing its dimensions (left). A second part was printed and sectioned to show its complex internal features (right).

FIG. 6 Photograph of case study component after printing.

the composition of the AM powder consumable itself is proprietary, elements of the finished component as analyzed via XRF satisfied the requirements for Alloy 625. Two cross-section specimens were prepared for metallographic examination, one oriented longitudinal to the build direction and one oriented transverse to the build direction, as shown in figure 7. The specimens were mounted, ground, and polished to examine the build density. Both the longitudinal and transverse cross sections exhibited only minor porosity; image threshold analyses reported porosity fractions of less than 0.2% in each specimen. .

SANDERS ET AL., DOI: 10.1520/STP163120190164

TABLE 1 Chemical composition of AM component

LPBF Part

UNS N06625 (ASTM B446 or ASTM F3056)

Nickel

65.9

58.0 min / Balance

Chromium

20.7

20.0–23.0

Iron

0.1

5.0 max

Molybdenum

8.7

8.0–10.0

Niobium þ Tantalum

3.9

3.15–4.15

Silicon

0.5

0.50 max

Manganese

a

0.50 max

Cobalt

a

1.0 max

Titanium

a

0.40 max

Aluminum

a

0.40 max

a

Not detected.

FIG. 7 Prepared metallographic specimens from AM component. Numbered scale divisions are inches.

The specimens were etched with a Kallings No. 2 acid solution to reveal the microstructures, examples of which are shown in figure 8 and figure 9 at low and high magnification, respectively. The microstructure consisted of irregular facecentered cubic (FCC) grains with dispersed carbides. It is normal to find carbides in Alloy 625, and they are most likely in the form of MC or M6C. No evidence was observed of remnant solidification microstructures from the fusion process, indicating that the 1,900 F heat treatment had effectively annealed the microstructure. There was no clear microstructural difference between the transverse and longitudinal sections. Mixtures of small and large grains were present in both orientations, sometimes referred to as a duplex microstructure, suggesting that partial .

295

296

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 8 Microstructure of AM component.

FIG. 9 Microstructures of AM component.

recrystallization had occurred. There was also no evidence of excessive secondary phases along grain boundaries. Figure 10 compares the LPBF Alloy 625 microstructure from this project to that of an annealed wrought Alloy 625 bar taken from Stress Engineering Service’s metallography library. Although the grain sizes are different, this feature is a function of annealing time and temperature. Such grain sizes can vary substantially even among wrought products. The noteworthy microstructural difference between the AM product and the wrought product was the uniformity of grain shape and distribution. This grain size and shape distribution is not in itself expected to be deleterious, but it may be indicative of other texture or fine precipitate differences that could affect corrosion performance. Advanced characterization techniques such as scanning electron microscopy, electron backscatter diffraction, or transmission electron microscopy (or any combination thereof) would be needed to further characterize these differences. .

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 10 LPBF Alloy 625 from this project (left) compared to annealed wrought Alloy 625 (right) from a metallography library.

FIG. 11 Hardness testing results.

Vickers (10 kg) hardness testing was conducted across the transverse specimen to assess the consistency of through-thickness properties. The results are shown in figure 11. The hardness ranged from 214 HV 10 to 223 HV 10, with an average hardness of 219 HV 10. The hardness values showed good through-wall consistency and were marginally lower toward the internal bore. The hardness of the LPBF AM material was comparable to typical hardness values for annealed wrought Alloy 625. Brinell hardness measurements (HRB) were also taken near the build plate side and the final laser pass side. This would be the top and bottom of the part in the build (Z) direction. The build plate side showed values of 95.6, 96.4, and 94.8 HRB while the last pass side showed values of 95.9, 94.7, and 94.2 HRB. This comparison shows good consistency in hardness values on the extremes of the build direction. .

297

298

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Overall, the LPBF material was found to be sound and free of metallurgical defects. Porosity was low, and there was no evidence of flaws that would be expected to compromise mechanical performance as compared to wrought parts. The anticipated corrosion response was not clear due to the pronounced inhomogeneity of the grain structure. Thus, mechanical testing and corrosion testing were performed on these components, the results of which are discussed in the following section. ELECTROCHEMICAL CORROSION TESTING

To further characterize the expected performance of these AM components, potentiodynamic polarization1 and potentiostatic electrochemical corrosion testing were performed on AM Alloy 625 and wrought 625 in ASTM G48, Standard Test Methods for Pitting and Crevice Corrosion Resistance of Stainless Steels and Related Alloys by Use of Ferric Chloride Solution,3 Test Solution C (acidified 6% FeCl3 with pH ¼ 0.6). Commercially available wrought Alloy 625 cylinder electrodes (20 mm in diameter by 10 mm long) and lab-prepared square (25 mm by 25 mm by 10 mm) electrodes for AM Alloy 625 with 1 cm2 exposed area were used for electrochemical corrosion testing. Potentiodynamic polarization is a technique where the potential of the electrode is varied at a specific rate by application of a current through the electrolyte. It is a commonly used testing method for measuring corrosion resistance and passive behavior of metal/alloys. A potentiodynamic polarization scan was performed at room temperature on working electrodes (WEs) of wrought and AM Alloy 625 material by scanning from 0.1 V to þ 1.5 V with respect to a reference electrode (RE) (saturated calomel electrode [SCE]) at a scan rate of 2 mV/s. Graphite was used as the counter electrode (CE). A schematic of the electrochemical corrosion test setup is in figure 12. ASTM G150, Standard Test Method for Electrochemical Critical Pitting Temperature Testing of Stainless Steels and Related Alloy,4 is used for evaluating the resistance of passive alloys to pitting corrosion based on the concept of the determination of a potential independent critical pitting temperature (CPT). This potentiostatic experiment imposes a constant passive potential on the working electrode for a specific time period while the solution is heated at a rate of 1.8 F/min (1 C/min). This potential is held constant during the entire temperature scan. The current is monitored during the temperature scan. The CPT is defined as the temperature at which the current increases rapidly, which for practical reasons is defined as the temperature at which the current density exceeds 100 lA/cm2 for 60 s. Pitting on the specimen is then confirmed visually after the test. In testing the AM specimen, a passive potential (þ550 mV) was selected from both potentiodynamic plots. ASTM G150 testing was performed on both the AM and wrought Alloy 625 specimens in ASTM G48 solution C (acidified ferric chloride) at þ550 mV constant potential at 1 C/min temperature ramp.3,4 .

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 12 Schematic of electrochemical corrosion test setup.

Potentiodynamic polarization scans for wrought and AM Alloy 625 material are shown in figure 13. Both specimens exhibited anodic passivation in the ASTM G48 Test Solution C. The AM Alloy 625 showed relatively higher current density at all potentials than the wrought Alloy 625. A passive potential (550 mV) was selected from both potentiodynamic plots for the ASTM G150 potentiostatic (constant potential) test. CPT scans for wrought and AM Alloy 625 are plotted together in figure 14. CPT was measured at 100 mA/cm2 current density; the values are listed in table 2. Wrought Alloy 625 exhibited a higher CPT than AM Alloy 625. Visual examination

FIG. 13 Potentiodynamic polarization scans of wrought and AM Alloy 625.

.

299

300

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 14 Potentiostatic scans for wrought and AM Alloy 625 in ASTM G150 testing.

TABLE 2 Critical pitting temperatures of wrought and AM Alloy 625

Specimen

Critical Pitting Temperature ( C)

AM Alloy 625

64

Wrought Alloy 625

81

of the electrode surfaces showed pitting corrosion on the exposed surface of the AM Alloy 625 specimen (fig. 15). FULL-SCALE TESTING

To help evaluate how the AM component would perform under oil field conditions, three full-scale load tests were conducted. The first test was to apply axial compression to the part, as shown in figure 16A, to 80,000 lb. The second test was to apply a combined load of torque and bending (fig. 16B) of 5,000 ft-lb, while the third was to apply a sinusoidal alternating bending moment across the part by rotating it (fig. 16C) to 10 million cycles at a peak von Mises equivalent (VME) stress range of 1,750 psi. These three tests represent common oil field loading conditions to which downhole tools are subjected. Due to the highly complex geometric features of the case study component, monitoring the induced stresses resulting from full-scale testing was impractical using traditional strain-gage monitoring techniques. For this reason, during compression testing, the part was monitored via advanced digital image correlation (DIC). DIC can be used to capture and record the deformation of high-strain-gradient loading conditions for geometries where traditional strain gages are difficult to install or cannot accurately measure. .

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 15 Corrosion pits (arrows) on surface of AM Alloy 625 electrode after ASTM G150 testing. Original magnification: 30.

FIG. 16 Schematic of applied loads for three full-scale tests.

.

301

302

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 17 Schematic of test setup incorporating digital image correlation.

A schematic of the DIC test setup is shown in figure 17. Two cameras are mounted stereoscopically to view a test article that is painted with a stochastic back and white contrasting speckle pattern. Advanced DIC software algorithms track the relative movements of the speckle pattern and produce a contour plot of the 3D displacement field of the viewed surface. Following the three load tests, the AM part was subjected to thermal shock and cycling. The part was submerged in a bath of liquid nitrogen at roughly 195 C (320 F) and allowed to soak until thermal equilibrium was reached. The part was then removed from the liquid nitrogen bath and immediately subjected to flame impingement until a thermocouple reading of roughly 982 C (1,800 F) was achieved. The part was then immediately submerged in the liquid nitrogen bath, resulting in a rapid thermal shock from cooling (1,800 F to 320 F in seconds). This process was then repeated for a total of 2.5 thermal cycles (returning to room temperature after the third liquid nitrogen soak). Photographs of the part at each of the thermal cycles are shown in figure 18.

Results and Discussion The DIC data from the axial compression test are shown in figure 19. These results show very high stress gradients throughout the part. This empirical strain data from real-world testing were then compared to an analytical simulation of the same .

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 18 Photographs of AM sample during thermal cycle testing.

FIG. 19 Digital image correlation strain data from compression test.

.

303

304

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 20 Comparison of DIC results (A and C) to FEM results (B and D).

axial compression loading obtained via a finite element model (FEM) (fig. 20). The results show generally good agreement for the strain gradient between the analytical model and the test data; however, differences can be noted. After the compression test, the part was inspected using a dye-penetrant method; no externally visible crack indications were observed. After the compression test, the AM part was tested with a combined load of torque and bending (fig. 16B) and then fatigue tested to 10 million cycles (fig. 16C). The AM part was then reinspected using a dye-penetrant method, with no externally visible crack indications observed. The part was then subjected to thermal shock and cycle testing. After the thermal shock and cycling, areas of plastic deformation were visible at thin features on the outside surfaces of the part. Photographs of this permanent deformation are shown in figure 21. After load testing and thermal cycling, the sample part was sectioned via wire electrical discharge machining (EDM) to enable inspection of the internal cavities and features. Dye-penetrant inspection was again performed, and no crack indications were found. A photograph of the sectioned test part after dye-penetrant inspection is shown in figure 22. Figure 23 shows a close-up view of an area of sharp notches that were thought to produce cracks during testing. However, posttest inspection showed no crack initiation at this location. In summary, a case study part was created to represent typical components that might be designed for oil field applications. A detailed metallurgical analysis was performed on the as-printed material, and the material exhibited low amounts of .

SANDERS ET AL., DOI: 10.1520/STP163120190164

FIG. 21 Photographs of plastic deformation seen after thermal shock cycle testing.

porosity and consistent hardness throughout. The microstructure of the AM sample was found to be comparable to that of typical wrought material. Electrochemical corrosion testing was performed on specimens of AM Alloy 625 and wrought Alloy 625. Potentiodynamic polarization scans and ASTM G150 corrosion testing indicated that the AM Alloy 625 showed relatively higher current density at all potential levels as compared to wrought Alloy 625. Further, the AM material exhibited a lower critical pitting temperature. Thus, the electrochemical corrosion performance of AM Alloy 625 was found to be inferior to a comparable wrought 625 Alloy. The AM sample component was subjected to axial compression, a combination of bending with torque, bending fatigue, thermal shock, and thermal cycling. No catastrophic failure or inability to support load was observed. After each round of testing, no external surface cracks were indicated via dye-penetrant inspection. After thermal shock testing, some plastic deformation was visible in the very thin wall thickness regions (0.13 in. thick). DIC and FEA results for the compression tests show generally good correlation; however, some differences were indicated. The DIC data can then be used to adjust the FEA assumptions and boundary conditions to produce a better analytical model. Following all testing, the AM sample part was sectioned for examination; no crack-like indications were found on the internal cavities. The results of this study indicate that this type of AM component would be expected to represent a suitable and advantageous alternative for use in a wide range of demanding oil field applications. .

305

306

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 22 Photograph of sectioned AM part after magnetic particle inspection (MPI) showing no indications.

FIG. 23 Close-up photograph of an area of sharp notches showing no indications.

.

SANDERS ET AL., DOI: 10.1520/STP163120190164

ACKNOWLEDGMENTS

The authors would like to thank Patrick Whalen and Ryan Brister for their support and collaboration with this effort as well as Stress Engineering Services, Inc., for allowing the authors to perform this study.

References 1. 2.

3.

4.

5.

.

U.S. Energy Information Administration, Monthly Energy Review, Table 1.3, April 2019 Preliminary Data (Washington, DC: U.S. Energy Information Administration, 2019). Standard Specification for Additive Manufacturing Stainless Steel Alloy (UNS S31603) with Powder Bed Fusion, ASTM F3184-16 (West Conshohocken, PA: ASTM International, approved September 1, 2016), https://doi.org/10.1520/F3184-16 Standard Test Methods for Pitting and Crevice Corrosion Resistance of Stainless Steels and Related Alloys by Use of Ferric Chloride Solution, ASTM G48-11(2015) (West Conshohocken, PA: ASTM International, approved November 1, 2015), https://doi.org/ 10.1520/G0048-11R15 Standard Test Method for Electrochemical Critical Pitting Temperature Testing of Stainless Steels and Related Alloys, ASTM G150-18 (West Conshohocken, PA: ASTM International, approved May 1, 2018), https://doi.org/10.1520/G0150-18 Standard Test Method for Conducting Potentiodynamic Polarization Resistance Measurements, ASTM G59-97(2014) (West Conshohocken, PA: ASTM International, approved May 1, 2014), https://doi.org/10.1520/G0059-97R14

307

308

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190136

Nand Kishore Singh,1 Shashi Kant Kumar,2 Satish K. S. N. Idury,2 K. K. Singh,3 and Ratneshwar Jha1

Dynamic Compression Response of Porous Zirconium-Based Bulk Metallic Glass (Zr41Ti14Cu12.5Ni10Be22.5) Honeycomb: A Numerical Study Citation N. K. Singh, S. K. Kumar, S. K. S. N. Idury, K. K. Singh, and R. Jha, “Dynamic Compression Response of Porous Zirconium-Based Bulk Metallic Glass (Zr41Ti14Cu12.5Ni10Be22.5) Honeycomb: A Numerical Study,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 308–321. http://doi.org/10.1520/STP1631201901364

ABSTRACT

Bulk metallic glasses (BMGs) are a unique class of materials that possess high yield strength and elastic limit. In view of their high yield strength and elastic limit, BMG honeycombs are attractive for mechanical energy absorption applications. However, the inability to synthesize BMGs in bulk form hinders their practical applications. In this context, additive manufacturing techniques provide a promising route to fabricate BMG honeycomb in bulk form. Because additive manufactured BMGs are porous, the manner in which a porous BMG honeycomb absorbs energy at various strain rates needs to be probed to suit this material for diverse practical applications. In this numerical study, we explore the effect of

Manuscript received November 4, 2019; accepted for publication February 6, 2020. 1 Dept. of Mechanical Engineering, School of Engineering, Rowan University, 201 Mullica Hill Rd., Glassboro, NJ 08028, USA N. K. S. https://orcid.org/0000-0002-4887-9873, R. J. https://orcid.org/0000-00030614-549X 2 School of Engineering, Dept. of Mechanical Engineering, Dayananda Sagar University-Innovation Campus, Kudlu Gate, Bengaluru 560 068, India 3 Dept. of Mechanical Engineering, Indian Institute of Technology (ISM) Dhanbad, Jharkhand, 826 004, India https://orcid.org/0000-0002-7186-3167 4 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

SINGH ET AL., DOI: 10.1520/STP163120190136

pore density (0, 5, 1.0, 15, and 20% by volume), strain rate (10, 100, and 1,000/s), and slenderness ratio (edge length to height: 0.5, 1, and 1.5) of a zirconium (Zr)-based BMG (Zr41Ti14Cu12.5Ni10Be22.5) honeycomb on its compression response through finite element simulations. The results are depicted in terms of stress–strain curves and energy–time curves. The energy absorption ability of the honeycomb with higher slenderness ratio increased from 98.6 kJ to 336.71 kJ at 20% porosity, while at 0% porosity, it increased from 118 kJ to 419.1 kJ as the strain rate was increased from 10 to 1,000/s. However, at 10% porosity, honeycomb of intermediate slenderness ratio (i.e., 1.0) exhibited the largest energy absorption to the order of 258 kJ at the strain rate of 1,000/s. Keywords bulk metallic glasses, porosity, mechanical behavior, finite element simulations

Introduction Bulk metallic glasses (BMGs) are a unique class of materials that possess high yield strength and elastic limit.1 On account of their high yield strength, elastic limit, excellent fracture strength, and superior resistance to wear and corrosion, BMGs are promising candidates for future structural and functional applications.2 However, during room temperature deformation, BMGs fail in a brittle manner after yielding, which restricts their application in the structural domain. In order to deal with such inherent limitation in room temperature deformation of BMG, various strategies such as fabricating BMG matrix composites with a reinforcing crystalline phase,3 synthesis of nano-grained MGs with unique interfaces,4 fabrication of cellular BMG structures,5 and production of porous BMG structures have been proposed.6 Among those methodologies, cellular and porous BMG structures attracted huge attention for technological applications on account of their ability to absorb higher energy and profound ductility.1,5,6 Sarac et al.1 demonstrated the superior energy absorption ability of BMG honeycomb synthesized through lithography and a thermoplastic forming technique. Liu et al.5 fabricated MG honeycomb architectures of various relative densities and showed they have remarkable energy absorption ability compared to that of conventional metals and ceramics. It was further revealed that by tuning the geometrical attributes of a BMG cellular structure—such as the thickness of the cell, cell size, and shape—its energy absorption capability can be remarkably tailored.7,8 Similarly, porous BMG foams synthesized through various processing routes were reported to exhibit remarkable plasticity, strength, and energy absorption ability.9,10 In BMG foams, too, the configuration of the pore, the volume fraction of pores, and diameter to the spacing ratio of the pores aids in optimizing the plasticity11,12 and subsequent energy absorption during deformation. However, the tunable plasticity obtained for micro- and nanoscopic BMG structures13,14 and energy absorption characteristics of millimeter-sized BMG honeycombs1,5 could not be mimicked for bulk structural applications because BMGs .

309

310

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

are known to exhibit size effects13,14 during deformation. Also, the inability to synthesize an amorphous structure into bulk shapes due to cooling rate limitations15 poses a serious limitation on these materials from an application point of view. In this scenario, the advent of additive manufacturing techniques holds a lot of promise to create a fully amorphous structure of significantly larger dimensions as compared to conventional processing methods of BMGs.16 Various studies reported the fabrication of complex and fully amorphous zirconium (Zr)-based BMG structures through additive manufacturing routes.17–19 Moreover, additive manufacturing processes resulted in the manufacturing of open-cell BMG structures of extraordinarily ordered interporosity,20 which are impractical to generate through conventional processing methods. Therefore, a BMG honeycomb of much larger dimensions compared to that reported in existing literature can be manufactured and its energy absorption capability can be engineered remarkably through manipulation of its pore configuration if processed through additive manufacturing. To the best of the authors’ knowledge, there are no experimental and simulationbased reports that explore the mechanical behavior of inherently porous BMG honeycombs. Therefore, probing the deformation behavior of porous honeycombs aids in tailoring the mechanical properties of significantly large and porous cellular structures generated through additive manufacturing. It is pertinent to mention that there are very few reports that deal with the mechanical behavior of BMG honeycombs,1,5,21 and all of them address honeycomb deformation only during quasistatic compression. But the manner in which BMG honeycombs (both porous and fully dense) respond during dynamic compression at various strain rates is yet to be explored. Also, there are no studies as to how the aspect ratio of a BMG honeycomb affects its compressive deformation; this probably is due to the sample size limitations of BMG honeycombs in Sarac et al.,1 Liu et al.,5 and Jayakumar and Hanan.21 Since additive manufacturing demonstrated the potential to eliminate such sample size restrictions for BMG, the effect of aspect ratio on BMG honeycomb compressive deformation needs to be probed. In this numerical study, we explored the effect of the pore volume fraction, strain rate, and aspect ratio of the BMG honeycomb on its compressive response. The constitutive behavior of BMG is modeled based on elastic failure material cards in LS DYNA.* The deformation response is captured in terms of the stress–strain curve and energy–time curves.

Simulation Methodology Three-dimensional (3D) finite element simulations were performed on a honeycomb structure through LS DYNA. The length and width of the honeycomb structure are 45 mm and 28 mm, respectively, as shown in figure 1. Each unit cell in the honeycomb has a 10-mm edge length. However, the height of the honeycomb is *

.

LS-DYNA keyword user’s manual, Volume II Material Models, Version R7.0.

SINGH ET AL., DOI: 10.1520/STP163120190136

FIG. 1 Schematic diagram of honeycomb.

varied as 5, 10, and 15 mm to probe the effect of aspect ratio on the honeycomb’s deformation response. Accordingly, the honeycomb samples were named H5, H10, and H15, respectively. Honeycomb was meshed with solid hexahedral elements and the size of each element was 1 mm. To simulate compression at a strain rate of 10, 100, and 1,000/s, nodes at the lower surface of honeycomb mesh were fixed while nodes at the upper surface were given different velocities, as shown in figure 1. Isotropic elastic failure implemented through material card Mat_13 (isotropic-elasticfailure model) in LS DYNA was adopted as the constitutive model for BMG. Material properties of Zr-based BMG with 0% porosity were taken from Lu, Ravichandran, and Johnson22 as input for isotropic elastic failure material card, Mat_13 of LS-Dyna, and in order to simulate a porous material honeycomb, material properties were evaluated by rule of mixture with an assumption that the porosity was distributed uniformly throughout the structure. Five different porosities were considered (table 1). All the elastic properties of the porous honeycomb were calculated based on the rule of mixtures. The material properties are appended in table 1.

Result and Discussion STRESS–STRAIN ANALYSIS

The details pertaining to peak stress obtained in all the simulations is tabulated in table 2. Though simulation was carried out for five different volume percentages of porosities (0, 5, 10, 15, and 20%), results are only explained for 0, 10, and 20% volume porosities because the trend in data is similar for intermediate porosities (5 and 15%). .

311

312

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 1 Material properties utilized for simulation

Mechanical Properties of Zr41Ti14Cu12.5Ni10Be22.5

Pore Density (%)

Compressive Yield Strength (GPa)

Elastic Modulus (GPa)

Shear Modulus (GPa)

Bulk Modulus (GPa)

Poisson’s Ratio

Density (Kg/m3)

0

2

101.2

37.4

114.1

0.352

6,125

5

1.9

96.14

36.03

96.8

0.334

5,818

10

1.8

91.08

34.58

82.92

0.317

5,512.5

15

1.7

86.02

33.11

71.33

0.299

5,206.2

20

1.6

80.96

31.57

61.96

0.282

4,900

TABLE 2 Peak stress in MPa obtained at varying strain rates for different slenderness ratio and volume percentages of porosities Strain Rate of 10/s 0% porosity

5% porosity

10% porosity

15% porosity

20% porosity

H5

2,000

1,900

1,800

1,700

1,600

H10

1,460

1,390

1,315

1,245

1,175

H15

1,175

1,128

880

1,021

958

0% porosity

5% porosity

10% porosity

15% porosity

20% porosity

H5

2,000

1,900

1,800

1,700

1,175

H10

1,570

1,488

1,417

1,346

1,268

H15

1,578

1,485

1,320

1,336

1,246

0% porosity

5% porosity

10% porosity

15% porosity

20% porosity

H5

2,000

1,900

1,800

1,700

1,600

H10

1,975

1,888

1,791

1,688

1,587

H15

1,981

1,883

1,320

1,686

1,586

Strain Rate of 100/s

Strain Rate of 1,000/s

Highlighting Effects of Varying Porous Volume Percentage and Slenderness Ratios at a Strain Rate of 10/s Figure 2 shows the stress–strain curve at a strain rate of 10/s for 0, 10, and 20% vol-

ume porosities. At a strain rate of 10/s, maximum load is taken by the honeycomb with the least slenderness ratio (i.e., 0.5 [H5]) for all volume percentage of porosities. At 0% volume porosity and 10/s strain, the maximum load taken by H5 is 2,000 MPa, which decreases to 1,460 MPa and 1,175 MPa for H10 and H15, respectively. Similarly, at 10% and 20% volume porosity, maximum loads taken by H5 are 1,800 MPa and 1,600 MPa, respectively, which decreases to 1,315 MPa and 1,175 MPa, respectively, for H10. Similarly, for H15, it decreases to 880 MPa and 958 .

SINGH ET AL., DOI: 10.1520/STP163120190136

FIG. 2 Stress–strain curve at strain rate of 10/s with varying volume percentage of porosities—(A) 0%, (B) 10%, and (C) 20%—and slenderness ratio.

MPa for 10% and 20% porosity volumes, respectively. Hence, it can be concluded that for all volume percentage of porosity at the strain rate of 10/s, the maximum load is taken by H5, which decreases for H10 and further decreases for H15. This trend can be attributed to the increase of buckling with the increase in the slenderness ratio. Maximum load (2,000 MPa) among all is taken by H5 with 0% volume porosities, which decreases to 1,800 MPa and 1,600 MPa for 10% and 20% porosities, respectively. A similar trend is displaced by H10 and H15 with the variation of porous volume percentage. Highlighting Effects of Varying Strain Rate at 20% Volume of Porosities Figure 3 shows the stress–strain curve at a strain rate of 10/s, 100/s, and 1,000/s for 20% volume porosities. At a strain rate of 10/s and a volume porosity of 20%, the peak stress for H5 is 1,600 MPa, which decreases to 1,175 MPa and 958 MPa for .

313

314

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 3 Stress–strain curve for 20% volume porosities at various strain rates: (A) 10/s, (B) 100/s, and (C) 1,000/s.

H10 and H15, respectively. At the low strain rate (10/s), the slenderness ratio clearly affects peak stress, whereas at the intermediate strain rate (100/s) and high strain rate (1,000/s), the slenderness ratio does not have many effects on peak stress. It is interesting to note that at the high strain rate, all the honeycombs fail due to fracture of intercellular walls, as compared to significant buckling of intercellular walls noticed for BMG foams.10,23 At a strain rate of 1,000/s, maximum loads taken by H5, H10, and H15 are similar (i.e., 1,600 MPa, 1,587 MPa, and 1,586 MPa, respectively). This can be attributed to the rapid propagation of stress waves at a higher strain rate, which reduces the buckling effect. A similar trend of peak stress is also observed at the strain rate of 100/s, where peak stresses for H5, H10, and H15 are 1,175 MPa, 1,268 MPa, and 1,246 MPa, respectively. It is interesting to note that in figure 3 the peak stress at 10/s strain rate is 1,600 MPa and the peak stress at 100/s is 1,175 MPa for H5 but again, the peak stress is 1,600 MPa for 1,000/s. This phenomenon occurs only in the case of H5 with 20% porosity and is possibly attributed to the interaction of porosities by propagating stress wave, large amount of porosity, and crushing mechanism as the only failure mode (buckling effect is very low for H5 due to the low slenderness ratio). .

SINGH ET AL., DOI: 10.1520/STP163120190136

TABLE 3 Maximum energy absorption (in KJ) at varying strain rates for different volume percentages of porous and slenderness ratio 0% porosity

5% porosity

10% porosity

15% porosity

20% porosity

125.25

Strain Rate of 10/s

H5

155

151.12

142.1

133

H10

114.5

109.28

103.87

97.93

91.94

H15

118.08

111.78

110

106 .01

98.66

H5

177.07

164.42

152.52

140.34

91.94

H10

127.39

119.63

114.15

107.74

100.62

H15

158.50

146.4

158

144.16

132.36

125.25

Strain Rate of 100/s

Strain Rate of 1,000/s

H5

183.28

171.89

159.47

148.13

H10

258.83

238.98

218.68

204.54

217.75

H15

419.1

399.53

158

361.01

336.71

ENERGY–TIME CURVE ANALYSIS

Details pertaining to maximum energy absorbed in simulation at varying conditions is tabulated in table 3. Though table 3 shows energy absorbed by honeycomb with 0, 5, 10, 15, and 20% volume of porosities, results are discussed only for 0, 10, and 20% porosities because of similar trends in intermediate porosities. Highlighting Effect of Varying Strain Rates for 20% Volume Porosities in a Honeycomb

As shown in figure 4 and tabulated in table 3, maximum energy among all is absorbed by H15 (336.71 KJ) at a strain rate of 1,000/s with 20% porosity. This energy absorption decreases to 217.75 KJ and 125.25 KJ for H10 and H5, respectively. Whereas at a strain rate of 100/s and 20% porosity, the maximum energy absorbed by H15 is 132.36 KJ, which reduces to 100.62 KJ and 91.94 KJ for H10 and H5, respectively. In contrast to the previously explained trend, at a strain rate of 10/s, the maximum energy absorbed by H15 is 98.66 KJ, which increases to 125.25 KJ for H5. This variation in trend can be summarized as follows: At a high strain rate, the maximum energy is absorbed by the honeycomb with the highest slenderness ratio (H15) because of the cushioning effect by trapped gas and the rapid propagation of a stress wave covering the entire bulk material before the formation of a new surface or buckling. At a low strain rate, the maximum energy is absorbed by the honeycomb with the least slenderness ratio (H5), which can be attributed to slow propagation of a stress wave, which is not able to reduce the buckling effect. At the low strain, the rate of trapped air does not give a cushioning

.

315

316

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 4 Energy–time relation for 20% volume porosities at various strain rates: (A) 10/s, (B) 100/s, and (C) 1,000/s.

effect and acts as a defect. In all other cases, energy absorbed depends on the combined effect of rapid coverage of the entire bulk material by a stress wave, reducing the buckling effect and the response of trapped gas as a cushioning effect or defect. However, the authors of this paper recommend more work be done before using these results for any practical application. The cushioning effect is not so prominent for a 10% volume of porosity at a strain rate of 1,000/s. Hence, at a strain rate of 1,000/s and 10% volume porosity, the maximum energy is absorbed by H10 (instead of H15, as in case of 20% porosity) as a combined effect of bulk material and cushioning effect. At 0% volume of porosity, no cushioning effect occurs, but the maximum energy among all is absorbed by H15 at a strain rate of 1,000/s. This can be attributed to higher bulk materials in H15, which is covered by a rapidly propagating stress wave at 1,000/s before failure. Hence, for a better understanding of porosity volume effect, in the upcoming section, we will discuss the effect of varying porosity at a particular strain rate.

.

SINGH ET AL., DOI: 10.1520/STP163120190136

Highlighting Effect of Variation in Porous Volume Percentage at a Strain Rate of 10/s

As shown in figure 5, at the strain rate of 10/s, the maximum energy (155 KJ) is absorbed by H5 with 0% porosities, which decreases to 142.1 KJ and 125.25 KJ for 10% and 20% porosities, respectively. This decrease is attributed to the reduction of bulk material with an increase of porosities at the same slenderness ratio. At a strain rate of 10/s for all percentage of porosities, the maximum energy absorbed by H10 and H15 is less when compared to that of H5 because of the buckling effect. Comparing the energy absorbed by H10 and H15, H15 absorbs more energy for all percentages of porosities. This is a combined effect of available bulk material and buckling effects. As a summary, energy absorption at a low strain rate depends on the combined effect of bulk material and buckling effect. A similar trend is observed at a strain rate of 100/s, so the next section includes a discussion related to a strain rate of 1,000/s.

FIG. 5 Energy–time curve at strain state of 10/s for various volume percentages of porosity: (A) 0%, (B) 10%, and (C) 20%.

.

317

318

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Highlighting Effect of Variations in Porous Volume Percentages at the Strain Rate of 1,000/s

As shown in figure 6 and tabulated in table 3, at a strain rate of 1,000/s and 0% porosity, the maximum energy (419.1 KJ) is absorbed by H15, which decreases to 258.83 KJ and 183.28 KJ for H10 and H5, respectively, at the same strain rate and porosity. H15 absorbs the maximum energy because rapid stress wave propagation covers more bulk material before the formation of any new surface or buckling (in fact, no buckling was observed). At a strain rate of 1,000/s and 10% porosity, H10 absorbs the maximum energy (218.68 KJ) in comparison to 152.52 KJ and 158 KJ energy absorbed by H5 and H15, respectively. This can be attributed to the combined effect of bulk material and cushioning effect. In H5, bulk material is less as compared to H10, whereas in H15, bulk material is more. But 10% porosity is not able to give a prominent cushioning effect, which means trapped gas will act as a defect. So more defect in H15 leads to less energy absorption as compared to H10.

FIG. 6 Energy–time curve at strain rate of 1,000/s for various volume percentages of porosity: (A) 0%, (B) 10%, and (C) 20%.

.

SINGH ET AL., DOI: 10.1520/STP163120190136

Maximum energy absorbed at the strain rate of 1,000/s and 20% porosity is also attributed to the combined effect of bulk material and cushioning effect. Trapped air in 20% porosities gives a good cushioning effect; hence, H15 absorbs a maximum energy of 336.71 KJ as compared to 217.75 KJ and 125.25 KJ energy absorbed by H10 and H5, respectively. As a summary, maximum energy at the strain rate of 1,000/s is absorbed by H15 because it has maximum material, which is covered by rapid propagation of a stress wave, reducing the buckling effect. And with increasing porosities, maximum energy absorption depends on the combined effect of the bulk material and the cushioning effect of trapped gas. Though it is intuitive that an increase in porosity decreases strength, the increase in strength with strain rate is counterintuitive because BMGs are known to exhibit negative strain rate sensitivity during deformation.24

Conclusions In this paper, a systematic investigation was carried out on a BMG honeycomb structure to explore the effect of porosity, strain rate, and slenderness ratio on its compression deformation through 3D finite element simulations. The deformation response was captured through stress–strain curves and energy time curves, which were explained by possible failure mechanisms. The results can be summarized as follows: • The peak stress of the honeycomb decreased with an increase in porosity at all strain rates and slenderness ratios. • The effect of the slenderness ratio is clear at a low strain rate, whereas at high strain rates and high porosities, peak stress is similar for all slenderness ratios. • The energy absorbed depends on the coverage of bulk materials by the propagation of the stress wave reducing the buckling effect and the response of trapped gas giving a cushioning effect or acting as a defect: • At a low strain rate (10/s) and an intermediate strain rate (100/s), stress wave propagation is slow, and trapped gas acts as a defect giving no cushioning effect. Hence, maximum energy at low and intermediate strain rates is absorbed by the honeycomb with a low slenderness ratio (H5). • At a high strain rate (1,000/s), stress waves propagate rapidly, which reduces the buckling effect, supporting a higher slenderness ratio for more energy absorption. But the cushioning effect depends on the volume percentage of porosity. At 10% porosity, the cushioning effect is not prominent; so it acts as a defect and the maximum energy is absorbed by H10. However, at 20% porosity, the cushioning effect is prominent, and maximum energy is absorbed by H15 as a combined supportive effect of the cushioning effect and more coverage of bulk material by rapid stress propagation.

.

319

320

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

References 1. 2. 3.

4.

5. 6. 7.

8.

9.

10.

11.

12. 13. 14. 15. 16.

17.

18.

.

B. Sarac, J. Ketkaew, D. O. Popnoe, and J. Schroers, “Honeycomb Structures of Bulk Metallic Glasses,” Advanced Functional Materials 22 (2012): 3161–3169. C. A. Schuh, T. G. Huffnagel, and U. Ramamurty, “Mechanical Behavior of Amorphous Alloys,” Acta Materialia 55, no. 12 (2007): 4067–4109. D.C. Hofmann, J. Y. Suh, A. Wiest, G. Duan, M. L. Lind, M. D. Demetriou, and W. L. Johnson, “Designing Bulk Metallic Glass Matrix Composite with Toughness and Tensile Ductility,” International Journal of Science 451 (2008): 1085–1089. F. C. Li, T. Y. Wang, Q. F. He, Y. Yang, B. A. Sun, C. Y. Guo, and T. Feng, “Micromechanical Mechanism of Yielding in Dual Nano-Phase Metallic Glass,” Scripta Materialia 154 (2018): 186–191. Z. Liu, W. Chen, J. Carstensen, J. Ketkaew, R. M. Ojeda Mota, J. K. Guest, and J. Schroers, “3D Metallic Glass Cellular Structures,” Acta Materialia 105 (2016): 35–43. A. H. Brothers and D. C. Dunand, “Plasticity and Damage in Cellular Amorphous Metals,” Acta Materialia 53 (2005): 4427–4440. S. H. Chen, K. C. Chen, F. F. Wu, and L. Xia, “Achieving High Energy Absorption Capacity in Cellular Bulk Metallic Glasses,” Nature Science Reports 5 (2015): 10302, https:// doi.org/10.1038/srep10302 J. C. Zhang, C. Chen, Q. X. Pei, Q. Wan, W. X. Zhang, and Z. D. Sha, “Deformation and Failure Mechanism of Nanoscale Cellular Structures of Metallic Glasses,” RSC Advances 6 (2016): 100899, https://doi.org/10.1039/C6RA22112B M. D. Demetriou, J. P. Schramm, C. Veazey, and W. L. Johnson, “High Porosity Metallic Glass Foam: A Powder Metallurgy Route,” Applied Physics Letters 91 (2007): 161903, https://doi.org/10.1063/1.2799248 M. D. Demetriou, C. Veazey, J. S. Harmon, J. P. Schramm, and W. L. Johnson, “Stochastic Metallic-Glass Cellular Structures Exhibiting Benchmark Strength,” Physical Review Letters 101 (2008): 145702, https://doi.org/10.1103/PhysRevLett.101.145702 B. Sarac, B. Klusemann, T. Xiao, and S. Bargmann, “Materials by Design: An Experimental and Computational Investigation on the Microanatomy Arrangement of Porous Metallic Glasses,” Acta Materialia 77 (2014): 411–422. D. Sopu, C. Soyarslan, B. Sarac, S. Bargmann, M. Stoica, and J. Erkert, “Structure-Property Relationships in Nanoporous Metallic Glasses,” Acta Materialia 106 (2016): 199–207. D. Jang and J. R. Greer, “Transition from a Strong-Yet-Brittle to a Stronger-and-Ductile State by Size Reduction of Metallic Glasses,” Nature Materialia 9 (2010): 215–219. R. Liontas and J. R. Greer, “3D Nano-Architected Metallic Glass: Size Effect Suppresses Catastrophic Failure,” Acta Materialia 133 (2017): 393–407. A. Inoue and A. Takeuchi, “Recent Development and Application Products of Bulk Glassy Alloys,” Acta Materialia 59 (2011): 2243–2267. Z. Mahbooba, L. Thorsson, M. Unosson, P. Skoglund, H. West, T. Horn, C. Rock, E. Vogli, and O. Harrysson, “Additive Manufacturing of an Iron-Based Bulk Metallic Glass Larger than the Critical Casting Thickness,” Applied Materials Today 11 (2018): 264–269. X. P. Li, M. P. Roberts, S. O’Keeffe, and T. B. Sercombe, “Selective Laser Melting of Zr-Based Bulk Metallic Glasses: Processing, Microstructure, and Mechanical Properties,” Materials & Design 112 (2016): 217–226. P. Bordeenithikasem, Y. Shen, H.-L. Tsai, and D. C. Hofmann, “Enhanced Mechanical Properties of Additively Manufactured Bulk Metallic Glasses Produced through Laser

SINGH ET AL., DOI: 10.1520/STP163120190136

19. 20.

21.

22.

23. 24.

.

Foil Printing from Continuous Sheetmetal Feedstock,” Journal of Additive Manufacturing 19 (2018): 95–103. Y. Shen, Y. Li, C. Chen, and H.-L. Tsai, “3D Printing of Large, Complex Metallic Glass Structures,” Materials & Design 117 (2017): 213–222. C. Zhang, X.-M. Li, S.-Q. Liu, H. Liu, L.-J. Yu, and L. Liu, “3D Printing of Zr-Based Metallic Glasses and Components for Potential Biomedical Applications,” Journal of Alloys and Compounds 790 (2019): 963–973. B. Jayakumar and J. C. Hanan, “Modeling the Axial Mechanical Response of Amorphous Fe45Ni45Mo7B3 Honeycombs,” Metallurgical and Materials Transactions A 43A (2012): 2669–2675. J. Lu, G. Ravichandran, and W. L. Johnson, “Deformation Behavior of the Zr41.2Ti13.8Cu12.5Ni10Be22.5 Bulk Metallic Glass over a Wide Range of Strain-Rates and Temperatures,” Acta Materialia 51, no. 12 (2003): 3429–3443. J. G. Wang, K. C. Chan, J. C. Fan, L. Xia, G. Wang, and W.-H. Wang, “Buckling of Metallic Glass Bars,” Journal of Non-Crystalline Solids 387 (2014): 1–5. F. H. D. Torre, A. Dubach, M. E. Siegrist, and J. F. Lo ¨ffler, “Negative Strain Rate Sensitivity in Bulk Metallic Glass and Its Similarities with the Dynamic Strain Aging Effect during Deformation,” Applied Physics Letters 89 (2006): 091918, https://doi.org/10.1063/ 1.2234309

321

322

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190139

Erik Woodard,1 Zach Post,1 and Mark Morrison2

Preclinical Testing of a Novel, Additive-Manufactured, ThreeDimensional Porous Titanium Structure Citation E. Woodard, Z. Post, and M. Morrison, “Preclinical Testing of a Novel, Additive-Manufactured, Three-Dimensional Porous Titanium Structure,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 322–339. http://doi.org/10.1520/STP1631201901393

ABSTRACT

When considering orthopedic implants for bone growth, several factors such as porosity, pore size, stiffness, friction, and strength can affect bone growth and contribute to the long-term success of the implant. Additive manufacturing is one tool to help achieve the ideal factors for implant structures and materials. SmithþNephew has developed an additive manufactured (AM), Ti-6Al-4V advanced porous structure designed to be similar to cancellous bone with up to 80% porosity. This structure is currently used as part of both acetabular shells and augments. This paper describes the preclinical testing of this advanced porous structure that comprised coupon-level and device-level testing. The critical parameters that can influence bone ingrowth, such as pore size (mean void intercept length, or MVIL) and porosity, were measured. The ability of the three-dimensional porous structure to withstand compressive, tensile, and shear forces was evaluated in static (monotonic) testing. Finally, bone ingrowth was assessed in a load-bearing ovine model. Clinically relevant device-level fatigue testing was conducted in foam blocks with a cavity and adjacent rim defect to

Manuscript received November 4, 2019; accepted for publication March 4, 2020. 1 SmithþNephew, Inc., 1450 East Brooks Rd., Memphis, TN 38116, USA E. W. http://orcid.org/0000-00017723-9034 2 SmithþNephew, Inc., 7135 Goodlett Farms Pkwy., Cordova, TN 38016, USA http://orcid.org/0000-00031956-0932 3 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

WOODARD ET AL., DOI: 10.1520/STP163120190139

simulate the acetabulum. The strength of the locking screw hole features was assessed using static and fatigue cantilever bending and pull-through strength. Acetabular constructs were also fatigue tested in an unsupported model with an adjacent augment and corresponding defect. Constructs completed all clinically relevant fatigue testing with no fractures. Keywords laser powder bed fusion, porous structure, orthopedic implants

Introduction Orthopedic implants depend on several factors for long-term clinical success. Implants must have good initial and long-term stability to minimize the risk of aseptic loosening and enable the potential for osseointegration. It has been shown that even small motions (less than 150 lm) at the bone-implant interface can discourage biological fixation.1 Pore sizes must be large enough to allow for bone ingrowth and vascularization, with pore sizes larger than 100 lm appearing necessary to accomplish this goal.2,3 Traditionally, porous structures such as sintered titanium beads have been used to create a bone-interfacing surface with relatively high friction and open pores allowing for biologic growth into the coating, resulting in good initial stability and long-term fixation. Orthopedic implants must also avoid stress shielding of surrounding bone. This occurs when the elastic modulus, or stiffness, of the implant is much greater than that of the bone into which it is implanted. This does not allow forces placed on the implant to be transferred to the bone. Wolff’s law states that bone will adapt to the stresses placed on it over time.4 If the load placed on the bone is increased, it will remodel itself to become stronger and resist the loading. Likewise, a lack of load transfer will cause a decrease in bone density and strength over time.5 Therefore, it is desirable to have an implant composed of a structure that closely matches the modulus of the surrounding bone. Since traditional porous structures consist of a coating applied to a solid substrate, it is difficult to produce an implant with a modulus similar to that of cancellous bone. Additive manufacturing is a helpful tool to accomplish several of the goals associated with orthopedic implants. It allows for the design and manufacture of complex porous structures that cannot be produced with traditional sintering techniques, and these structures can be integrated into implants in unique ways because the entire device can be manufactured as a single component. The structure parameters can be customized to ensure a desirable porosity, pore size, strength, and flexibility are achieved. This also allows for design flexibility because additive manufacturing can be used to produce differing structures at any region of the implant. To accomplish some of the desired goals associated with orthopedic implants, SmithþNephew has developed a fully randomized additive-manufactured, Ti-6Al4V advanced porous structure with properties designed to be similar to cancellous bone manufactured using laser powder bed fusion (LPBF) (fig. 1). During .

323

324

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 1 SmithþNephew’s fully randomized, additive-manufactured, Ti-6Al-4V advanced porous structure.

development, various lattice geometries were explored and a structure with the desired combination of features (porosity, flexibility, strength) was ultimately chosen. This structure is currently used as part of both acetabular shells and augments. These include the REDAPT Fully Porous Acetabular Shell, REDAPT Modular Acetabular Shell, and REDAPT Acetabular Augments (fig. 2). This paper describes a subset of the preclinical testing of this advanced porous structure that comprised coupon level and device-level testing. This testing is critical to ensure that the relatively novel methods used to manufacture the components produce the desired properties and performance. Several challenges and opportunities unique to testing additive manufactured structures and devices are detailed.

Material and Methods MATERIAL

.

All components (coupons and devices) were manufactured on LPBF systems from Ti-6Al-4V powder. Coupons and devices were postprocessed (cleaned, heat treated, and so on) in an identical manner.

WOODARD ET AL., DOI: 10.1520/STP163120190139

FIG. 2 Acetabular devices that incorporate the CONCELOC porous structure. The REDAPT Fully Porous Acetabular Shell (left) is made primarily from the porous structure, the REDAPT Modular Acetabular Shell (center) incorporates the CONCELOC structure on the outer surface, and the REDAPT Slice and Staple Augments (right) support the acetabular shells in the case of adjacent defects.

MATERIAL TESTING—COUPONS

The flexibility of additive manufacturing (AM) allowed for the use of various coupons for the material testing of the porous structure. All material testing was performed on these coupons, which were considered representative of the basic properties of the porous structure at the device level. Industrial additive manufacturing machines from Electro Optical Systems (EOS, Munich, Germany) were used to produce all samples used for testing. MATERIAL TESTING—PORE MORPHOLOGY

Porosity and pore size (mean void intercept length, or MVIL) were measured using the methodology in ASTM F1854-15, Standard Test Method for Stereological Evaluation of Porous Coatings on Medical Implants,6 and were compared to ranges reported for cancellous bone. MATERIAL TESTING—MECHANICAL PROPERTIES

Compressive properties of the porous structure including modulus, 0.2% offset yield strength, and peak strength were tested with six Ø0.40 in. by 0.46 in. (Ø10.2 mm by 11.7 mm) cylinders at a displacement rate of 1.00 in./min (25.4 mm/min). The peak compressive strength was defined as the strength prior to the first negative slope on the stress versus strain plot, which is indicative of the beginning of the plateau region in which the cells of the porous structure begin to collapse.7 Tensile properties of the porous structure including modulus, 0.2% offset yield strength, and ultimate strength were tested with eight Ø0.39 in. by 2.25 in. (Ø10 mm by 57.2 mm) .

325

326

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

cylinders with solid ends and a porous gage sections at a displacement rate of 0.1 in./min (2.54 mm/min). Shear properties of the porous structure including modulus, 0.2% offset yield strength, and ultimate strength were tested with six Ø0.75 in. by 2.00 in. (Ø19.05 mm by 50.8 mm) cylinders with solid ends and porous gage sections at a displacement rate of 0.1 in./min (2.54 mm/min). A subset of the measured values for mechanical properties was compared to values available in the literature for cancellous bone. MATERIAL TESTING—BONE INGROWTH

Bone ingrowth was assessed in a validated, load-bearing, subarticular ovine model with ethical review process approval. Bilateral defects were created in the cancellous bone of proximal sheep tibias parallel to and 3 mm below the medial tibia plateau (fig. 3). This region of the bone exhibits higher stresses than the more standard implantation sites in the femoral metaphysis or diaphysis.8 A clinically successful porous bead coating was used as a control. Eight coupons of each group were pressfit into a defect for a 12-week implantation time. The push-out force for removal of the coupons was measured, and these forces were compared with a one-sided paired t-test. For push-out testing, the harvested tibias were sectioned and the coupons were pushed out of the bone from the flat backside of the coupon using an Instron test frame (Norwood, MA). The depth of bone ingrowth was not directly measured as part of this study. DEVICE TESTING—ACETABULAR SHELL DEFORMATION

The REDAPT Fully Porous Acetabular Shell is made primarily from the CONCELOC porous structure, with internal solid reinforcement features in critical locations

FIG. 3 (A) Coupon used for the bone ingrowth study, (B) a model of the defect, and (C) a radiograph of the coupon implanted in the defect.

.

WOODARD ET AL., DOI: 10.1520/STP163120190139

FIG. 4 A cross section of the REDAPT Fully Porous Acetabular Shell is shown, with the solid internal reinforcement features indicated.

(fig. 4), which are only possible due to the use of additive manufacturing. The goal of this configuration is to create an acetabular shell that is flexible while maintaining the strength requirements for this device. To analyze the flexibility of the shell, two types of deformation test were conducted. The first test consisted of creating a simulated acetabular cavity in the center of a polyurethane foam block with a density of 0.48 g/cm3, which is in the reported range for cancellous bone (0.1 to 0.7 g/ cm3).9 The simulated cavity consisted of a hemisphere with a diameter 2 mm less than that of the corresponding acetabular shell outer diameter, to create a severe press-fit condition.10 A defect was created adjacent to the simulated cavity by removing 25% of the rim to a depth of 5 mm (fig. 5). The shells were impacted into the foam blocks using a 10-lb (4.5 kg) drop weight from a height of 2.75 in. (6.99 cm) twenty times until fully seated. The inner surface of the shell near the rim was analyzed using a coordinate measurement machine (CMM), and the total roundness (RONt) was calculated and compared to the liner clearance with the shell and liner at maximum material condition (MMC) using a tolerance stackup. .

327

328

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

FIG. 5 Schematic of the setup for shell deformation testing in a foam block.

The shells were removed from the foam blocks and the roundness was measured again to determine if the initial RONt was likely due to permanent deformation. The roundness of the shells was also measured prior to any testing using the same methods to establish a baseline for deformation and to determine how much deformation was due to the test method. For the second deformation test, the shell was placed on its side and a compressive load was applied at a constant displacement rate of 0.1 in./min (0.25 cm/min) up to 250 lbf (1,112 N) near the rim using a rounded steel fixture (fig. 6). The slope of the linear region of the force versus displacement curve was calculated, and this was designated as the shell stiffness. The stiffness of the REDAPT Fully Porous Acetabular Shell was compared to other SmithþNephew shell designs tested using the same method. DEVICE TESTING—FINITE ELEMENT ANALYSIS AND WORST-CASE DETERMINATION

To determine the worst-case combination of devices to use for mechanical fatigue testing, finite element analysis (FEA) was performed using ABAQUS/CAE 2016 to compare the maximum principal stress for various shell and augment size combinations and orientations. Since the porous structure could not be directly modeled in the simulation, the bulk elastic modulus and Poisson’s ratio were used for analysis, and the porous regions were treated as a flexible solid region. For all other regions, the mechanical properties for wrought Ti-Al-6V were used.11 The material properties used for the components in the analysis are shown in table 1. The component models were virtually assembled together in the foam block model. A hemispherical cavity matching the shell’s outer diameter was created in the block model. Surface-to-surface contact was simulated between the block and shell outer diameter (friction coefficient ¼ 1.1). The cement mantle between the .

WOODARD ET AL., DOI: 10.1520/STP163120190139

FIG. 6 Image of the setup for shell deformation testing to measure the shell stiffness.

TABLE 1 List of material properties for the component used in the finite element analysis

Component

Foam block

Material

Elastic Modulus (E, GPa)

Poisson’s Ratio

0.48 g/cm3 Density

10.9

0.3

Polyurethane foam Cement Liner Shell and augment

PMMA bone cement

2.45

0.35

Cross-linked polyethylene

0.7

0.4

Ti-6Al-4V

114

0.3

(solid regions)

shell and liner was fixed to all surfaces using a tie constraint. A matching femoral head was virtually placed into the inner surface of the liner, and a frictionless contact condition was specified between these components. The femoral head was treated as a rigid body for this analysis. The outer surfaces of the block (excluding the shell-contacting surface) were fixed in all degrees of freedom, and a 1,200-lbf (5,338 N) load was placed on the center of the femoral head at a 50 angle to the block face to simulate the mechanical test conditions and worst-case expected clinical conditions. Quadratic tetrahedral (C3D10) elements were used to mesh all components. The maximum principal stress in the acetabular shell was analyzed for various shell/augment size combinations and shell orientations. Each simulation was run with three mesh sizes of increasing density in the critical stress region, until a mesh convergence of less than 5% was achieved following the guidance of ASTM F2996-13, Standard Practice for Finite Element Analysis (FEA) of Non-Modular Metallic Orthopaedic Hip Femoral Stems. .

329

330

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

Due to the complexity of analyzing both the shell and augment in a construct setup, a three-step method was used to determine the worst-case combination. First, the area of the acetabular shell that remained unsupported when assembled with an augment was used to select a subset of worst-case combinations. This region was used because the shell was the thinnest porous component in the construct, and this unsupported region was directly in line with the applied load. Therefore, this was considered the critical region for testing. The predicted worst-case combinations were analyzed for maximum principal stress using a finite element model, and then a subset of these worst-case combinations was selected for construct fatigue testing. The cycle count at fracture and fracture location from the construct fatigue testing were used to verify the results of the finite element and engineering analysis. DEVICE TESTING—ACETABULAR SHELL FATIGUE

The REDAPT Fully Porous Shell and REDAPT Modular Acetabular Shell were tested for fatigue strength using a construct setup to simulate expected clinical conditions (fig. 7). Servohydraulic single axis test frames (Material Test Systems, Eden Prairie, MN) were used for all fatigue testing. Polyurethane foam blocks with a density of 0.48 g/cm3 within the range reported for cancellous bone9 (0.1–0.7 g/cm3) were used as a test medium to support the acetabular construct. A hemisphere was created in the center of the blocks using an acetabular reamer, and shells and liners were assembled into the blocks. A defect representative of a Paprosky Type IIB classification12 was created in the superior/posterior quadrant. The acetabular

FIG. 7 Representative image of the setup used for fatigue testing of acetabular shells.

.

WOODARD ET AL., DOI: 10.1520/STP163120190139

constructs were fatigue tested using a sinusoidal cyclic compressive load of 120/1,200 lbf (534/5,338 N) at an inclination angle of 50 and a frequency of 10 Hz for 10 million cycles (runout) or construct failure. Failure was defined as fracture or loosening of any component. Constructs were tested in air, at room temperature. After the completion of testing, the acetabular constructs were removed from the blocks and inspected visually and microscopically at 10 magnification for fractures or damage. DEVICE TESTING—SHELL AND AUGMENT FATIGUE

In addition to testing the fatigue strength of the acetabular shells, the shells were also tested in conjunction with respective acetabular augments. These augments are used to fill defects encountered during total hip arthroplasty surgery. The augment is used to support the acetabular shell and prevent shell motion and migration. In the case of a severe acetabular defect, a lack of support could cause either fracture of the acetabular shell or shell migration leading to aseptic loosening. Because no standard currently exists for fatigue testing an acetabular shell in conjunction with an augment, the parameters for this test were based on internal protocols for worstcase expected clinical conditions. The same foam blocks were used for the shell/augment combination testing, and a larger defect was created adjacent to the acetabular cavity using a secondary reamer. This simulated defect was used to place the augment adjacent to the acetabular shell (fig. 8). Feasibility testing showed that the worst-case positioning of the augment occurred at a 45 orientation, with the augment in the superior/posterior quadrant. This resulted in a lack of support for the acetabular shell in line with the fatigue loading. Testing was conducted at an inclination angle of 50 and a frequency of 10 Hz. Failure was defined as fracture or

FIG. 8 Representative setup and schematic of the test setup used for acetabular shell and augment fatigue testing.

.

331

332

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

loosening of any component. Constructs were tested in air, at room temperature. The runout load was defined as the maximum fatigue load at which six of the constructs completed 10 million cycles of testing with no failures. The runout load of the acetabular shell and augment constructs in this configuration was compared to that of a clinically successful SmithþNephew acetabular device with similar indications. DEVICE TESTING—SCREW HOLE FEATURE STRENGTH

The REDAPT Acetabular Shells and REDAPT Augments have variable angle screw hole features that can accept locking or nonlocking screws. The use of locking screws results in increased cantilever bending stiffness and may limit the motion of the acetabular construct. To test the strength of the locking screw hole features, AM coupons were produced with all relevant features (fig. 9). These coupons allowed for consistent testing that would have been challenging to achieve using a full acetabular shell construct. Additive manufacturing was a useful tool to simplify testing and manufacturing for these features of interest. For a worst-case test condition, all features of interest were manufactured at least material condition (LMC). Static cantilever bending testing was conducted by securing locking and nonlocking screws in the coupons and through a 20-mm (0.974 in.) cube of 15 lb/ft3 (0.24 g/cm3) density block of polyurethane foam. An electromechanical (EM) test frame (MTS, Eden Prairie, MN) was used for this monotonic testing. A 45 chamfer was placed on the underside of the foam block to provide relief and reduce support from the foam while allowing for testing of nonlocking screws (fig. 10). The tip of the screw was loaded at a constant displacement rate of 0.1 in./min (0.25 cm/min) until failure. Failure was defined as a drop of 90% from the peak load. The moment arm for each screw/coupon construct was defined as the distance from the loading location on the screw to the flat surface of the coupon. The maximum supported load for each screw was multiplied by the moment arm to determine the cantilever bending moment. The static bending strength of the REDAPT locking screw

FIG. 9 Representative image of the coupons used to isolate the REDAPT screw hole features and perform cantilever bending screw testing.

.

WOODARD ET AL., DOI: 10.1520/STP163120190139

FIG. 10 Image of the setup for screw cantilever bending testing.

features was compared to that of a clinically successful acetabular nonlocking screw hole design using a Fisher’s F-test for variances and a Student’s t-test for differences in means, with a ¼ 0.05. The stiffness of the locking or nonlocking screw constructs was also calculated by taking the slope of the force versus displacement curve in the linear region. The force was normalized to the moment arm distance for each tested sample, so stiffness was reported as a moment/displacement.

Results Porosity of up to 67% was measured in the bulk of the porous structure and up to 80% near the surface. These values are within the broad range from 30 to 90% reported for cancellous bone.13 Pore size (MVIL) between 202 and 342 lm was measured in the bulk structure and between 484 and 934 lm near the surface. These values are within the broad range from 148 to 5,085 lm reported for cancellous bone13 and are larger than the pore size of about 100 lm, which is considered beneficial for biological fixation.14 Compressive modulus of the porous structure was 4.3 GPa, which is near the range reported for cancellous bone of 0.6 to 3.2 GPa.15 The 0.2% offset compressive yield strength was 76.5 MPa, which is higher than the range reported for cancellous bone of 3 to 17 MPa.15 Peak compressive strength was 106.3 MPa. Tensile modulus was 2.8 GPa, which is near the range reported for cancellous bone of 0.6 to 2.7 GPa.16 The 0.2% offset tensile yield strength and ultimate tensile strength were 27.3 MPa and 100.1 MPa, respectively, which are higher than the ranges reported for cancellous bone of 2 to 11 MPa and 0.9 to 23.1 MPa, respectively.15–17 The 0.2% offset shear yield strength was 39.0 MPa; ultimate shear strength was 63.3 MPa. A summary of the static test results is shown in table 2. .

333

334

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 2 Compression and tensile properties of the porous structure and cancellous bone as well as shear properties of the porous structure

Porous Structure Compression

Modulus (GPa) 0.2% Offset yield

4.3 6 0.1

Cancellous Bone Compression

0.6–3.2

15

Porous Structure Tensile

Cancellous Bone Tensile 15

Shear

2.8 6 0.22

0.6–2.7

76.5 6 2.0

3–1715

27.3 6 5.2

2–1115

39.0 6 2.5

–

–

–

100.1 6 2.6

0.9–23.116,17

63.3 6 2.5

106.3 6 3.5

–

–

–

–

strength (MPa) Ultimate strength (MPa) Peak strength (MPa)

In the bone ingrowth animal model, a pushout force 23% higher than the bead coating control (p ¼ 0.013) was measured for the AM structure. The average pushout force for the AM structure coupons was 1,633 6 283 N compared to 1,329 6 221 N for the sintered bead coupons. For the acetabular shell deformation testing, the RONt of the shell inner diameter did not exceed the liner clearance under worst-case press-fit condition. After removal, the shell RONt was within the range prior to impact into the foam blocks, indicating that the measured RONt was not due to permanent deformation of the shell. The shell deformation prior to impaction in the foam blocks was negligible compared to the deformation caused by press fit into the blocks, so it was determined that the majority of shell deformation was due to the test method. The shell RONt in the foam blocks and after removal as a percentage of liner clearance is shown in figure 11. When comparing the calculated shell stiffness based on the slope of the force versus displacement curve, it was determined that the REDAPT Fully Porous Acetabular Shell was 54.7% less stiff than a comparable traditionally manufactured SmithþNephew acetabular shell design. The results of the FEA showed that the worst-case acetabular shell and augment configuration occurred with the augment at a 45 angle to the superior/inferior axis, in the superior/posterior quadrant. This resulted in an unsupported shell span directly in line with the applied loading. Mechanical fatigue testing of the acetabular shell and augment combination confirmed that the critical stress region was located in the unsupported span of the shell. Fractures of the acetabular shell in mechanical fatigue testing with an augment corresponded to the peak stress location from the FEA study, which provided a validation of the FEA for the context of use of choosing the worst-case device combinations for mechanical testing. The maximum principal stress distribution in the critical region of the smallest shell size (48 mm) using FEA is shown in figure 12. This region was used for all engineering analysis and was therefore the region of interest for mechanical fatigue testing to determine the worst-case shell and augment size combinations. Based on the results of the FEA, it was determined that the maximum principal stress in the smallest .

WOODARD ET AL., DOI: 10.1520/STP163120190139

FIG. 11 Results of the acetabular shell deformation in foam blocks. The results of the RONt measurements are shown as a percentage of the acetabular liner clearance at MMC. The error bars represent standard deviations.

FIG. 12 Maximum principal stress distribution in the acetabular shell when combined with an augment. The region of interest based on the results of initial fatigue testing is circled.

.

335

336

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

shell and augment combination (48 mm shell with 50 mm by 8 mm augment) was 91.3% greater than that of the largest size combinations (76 mm shell with 74 mm by 8 mm augment). Analysis of other shell/augment size combinations showed that the maximum principal stress for all other sizes was between the ranges of these two combinations. The results of the fatigue testing with only the acetabular shell (no augment) indicated that six (n ¼ 6) constructs completed 10 million cycles of fatigue testing at 120/1,200 lbf (534/5,338 N) with no signs of fracture or failure. This load corresponds to the femoral neck fatigue strength requirement for femoral prostheses in ASTM F2068-15, Standard Specification for Femoral Prostheses—Metallic Implants.18 The results of the fatigue testing with combined shell and augment indicated that the runout load was at least 800 lbf (3,559 N), which was greater than that of a clinically successful SmithþNephew predicate acetabular device. When comparing the results of the acetabular shell and augment construct fatigue testing, a good correlation was found between the predicted strength of the construct based on engineering analysis and the cycle count to failure (R2 ¼ 0.97). The results of the screw hole feature cantilever bending strength showed that the bending strength of the REDAPT screw hole features was 216% greater than that of a clinically successful predicate acetabular screw hole design for nonlocking screws, and the difference was statistically significant (p < 0.05). Nonlocking screws had a 117% greater ultimate cantilever bending strength in the variable angle screw hole features compared to locking screws. However, locking screws resulted in greater resistance to motion, or stiffness, compared to nonlocking screws. The average cantilever bending stiffness of the locking screws in the variable angle screw holes was 741 6 97 in.-lbf/in., compared to 95 6 19 in.-lbf/in. for the nonlocking screws. The results of the cantilever bending testing are representative of components at LMC condition and should therefore be taken as the lowest expected values. Components manufactured using nominal dimensions would be expected to result in greater cantilever bending values.

Discussion Additive manufacturing can enable the production of complex porous structures not possible with traditional manufacturing techniques. However, this introduces another level of complexity to analysis of these structures, particularly regarding testing of orthopedic implants. The primary challenge for device-level testing of additive manufactured orthopedic implants can be the determination of the worstcase devices or device combinations. Numerical simulations of additive manufactured porous structures have been conducted in the literature on lattices with repeating and well-defined three-dimensional shapes.19,20 However, the application of these methods to device-level simulation as well as randomized complex lattice structures is unknown, and the validation of such applications likely presents a significant challenge. .

WOODARD ET AL., DOI: 10.1520/STP163120190139

In the absence of a validated numerical simulation for complex lattice structures used as part of orthopedic implants, FEA can still be a useful tool for predicting stresses for construct testing. Caution should be taken in the interpretation of the results of these simulations, and they usually should be combined with engineering analysis and rigorous mechanical testing to confirm the results. Typically, validation of the FEA method would include comparison to stresses in the physical components using strain gaging or another method to directly measure stress or strain (or both). The challenges of using these methods are greatly increased due to the complexity of analyzing additive manufactured lattice structures. For the context of use of selecting the worst-case device combinations, a partial validation may be accomplished by matching the region of greatest stress in the simulation to the expected physical failure location. The risk of selecting the incorrect component combination can be further minimized by selecting a subset of worst-case devices for testing and verifying the consistency of the results. Additive manufacturing can be used to produce unique test coupons that can isolate features of interest and simplify testing. This can be accomplished when device-level testing is overly complex or unnecessary. Several standards exist for testing the attachment strength of porous coatings.21–23 These methods can be adapted for testing additive manufactured porous structures, but some adaptations may be useful. These standard methods typically involve fixing test coupons with the porous structure together using a strong bonding adhesive. If the adhesive strength of the porous coating to the substrate is stronger than the bonding agent, the true properties of the porous structure may not be able to be determined. In these cases, it can be useful to utilize additive manufacture coupons composed of porous or solid material in select regions, instead of bonding together multiple coupons or test stubs. Using this method, failures of the porous structure can be obtained and the true properties (shear, tension, and compression) can be calculated. Additive manufacturing can be used to produce porous structures with similar properties to cancellous bone, which is useful in orthopedic applications. These structures can be incorporated into orthopedic devices to extend the benefits of the porous structure to multiple applications. In summary, for the most part, existing standards can be utilized or adapted for testing additive manufactured structures or orthopedic devices. However, additive manufacturing does introduce a layer of testing complexity beyond that required for devices manufactured with traditional methods. The effect of process parameters on the additive manufactured structure must be understood, and manufacture of complex structures made possible by additive manufacturing presents further challenges to modeling and simulation. Until standard methods are available for analysis, several analysis methods should be combined with a rigorous test protocol to ensure the safety and efficacy of additive manufactured orthopedic devices.

.

337

338

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

References 1.

2.

3. 4. 5.

6.

7. 8.

9.

10. 11.

12.

13.

14.

15. 16. 17.

.

M. Jasty, C. Bragdon, D. Burke, D. O’Connor, J. Lowenstein, and W. H. Harris, “In Vivo Skeletal Responses to Porous-Surfaced Implants Subjected to Small Induced Motions,” The Journal of Bone and Joint Surgery 79, no. 5 (1997): 707–714. J. D. Bobyn, R. M. Pilliar, H. U. Cameron, and G. C. Weatherly, “The Optimum Pore Size for the Fixation of Porous-Surfaced Metal Implants by the Ingrowth of Bone,” Clinical Orthopaedics and Related Research 150 (1980): 263–270. V. Karageorgiou and D. Kaplan, “Porosity of 3D Biomaterial Scaffolds and Osteogenesis,” Biomaterials 26, no. 27 (2005): 5474–5491. H. M. Frost, “Wolff’s Law and Bone’s Structural Adaptations to Mechanical Usage: An Overview for Clinicians,” The Angle Orthodontist 64, no. 3 (1994): 175–188. J. M. Wright, P. M. Pellicci, E. A. Salvati, B. Ghelman, M. M. Roberts, and J. L. Koh, “Bone Density Adjacent to Press-Fit Acetabular Components: A Prospective Analysis with Quantitative Computed Tomography,” The Journal of Bone and Joint Surgery 83, no. 4 (2001): 529–536. Standard Test Method for Stereological Evaluation of Porous Coatings on Medical Implants, ASTM F1854-15 (West Conshohocken, PA: ASTM International, approved March 15, 2015), https://doi.org/10.1520/F1854-15 L. J. Gibson and M. F. Ashby, Cellular Solids: Structure and Properties (Cambridge, MA: Cambridge University Press, 1997). S. Fenwick, D. Wilson, M. Williams, J. Penmetsa, K. Ellis, M. Smith, J. Dodd, J. Pitcher, and M. Scott, “A Load Bearing Model to Assess Osseointegration of Novel Surfaces—A Pilot Study,” in Proceedings of the Eighth World Biomaterials Congress (Red Hook, NY: Curran Associates, 2008), 1240. D. C. Wirtz, N. Schiffers, T. Pandorf, K. Radermacher, D. Weichert, and R. Forst, “Critical Evaluation of Known Bone Material Properties to Realize Anisotropic FE-Simulation of the Proximal Femur,” Journal of Biomechanics 33, no. 10 (2000): 1325–1330. E. Adler, S. A. Stuchin, and F. J. Kummer, “Stability of Press-Fit Acetabular Cups,” The Journal of Arthroplasty 7, no. 3 (1992): 295–301. E. Brandl, C. Leyens, and F. Palm, “Mechanical Properties of Additive Manufactured Ti6Al-4V Using Wire and Powder Based Processes,” IOP Conference Series: Materials Science and Engineering 26, no. 1 (2011), https://doi.org/10.1088/1757-899X/26/1/012004 W. G. Paprosky, P. G. Perona, and J. M. Lawrence, “Acetabular Defect Classification and Surgical Reconstruction in Revision Arthroplasty: A 6-Year Follow-Up Evaluation,” The Journal of Arthroplasty 9, no. 1 (1994): 33–44. W. C. Hayes and M. L. Bouxsein, “Biomechanics of Cortical and Trabecular Bone: Implications for Assessment of Fracture Risk,” in Basic Orthopaedic Biomechanics, ed. V. C. Mow and W. C. Hayes (Philadelphia, PA: Lippincott-Raven Publishers, 1997), 69–112. J. D. Bobyn, R. M. Pilliar, H. U. Cameron and G. C. Weatherly, “The Optimum Pore Size for the Fixation of Porous-Surfaced Metal Implants by the Ingrowth of Bone,” Clinical Orthopaedics and Related Research 150 (1980):263–270. E. F. Morgan and T. M. Keaveny, “Dependence of Yield Strain of Human Trabecular Bone on Anatomic Site,” Journal of Biomechanics 34, no. 5 (2001): 569–577. L. Røhl, E. Larsen, F. Linde, A. Odgaard, and J. Jørgensen, “Tensile and Compressive Properties of Cancellous Bone,” Journal of Biomechanics 24, no. 12 (1991): 1143–1149. T. M. Keaveny, E. F. Wachtel, C. M. Ford, and W. C. Hayes, “Differences between the Tensile and Compressive Strengths of Bovine Tibial Trabecular Bone Depend on Modulus,” Journal of Biomechanics 27, no. 9 (1994): 1137–1146.

WOODARD ET AL., DOI: 10.1520/STP163120190139

18.

19.

20.

21.

22.

23.

.

Standard Specification for Femoral Prostheses—Metallic Implants, ASTM F2068-15 (West Conshohocken, PA: ASTM International, approved March 15, 2015), https:// doi.org/10.1520/F2068-15 A. Zargarian, M. Esfahanian, J. Kadkhodapour, and S. Ziaei-Rad, “Numerical Simulation of the Fatigue Behavior of Additive Manufactured Titanium Porous Lattice Structures,” Materials Science and Engineering: C 60 (2016): 339–347. R. Hedayati, H. Hosseini-Toudeshky, M. Sadighi, M. Mohammadi-Aghdam, and A. A. Zadpoor, “Computational Prediction of the Fatigue Behavior of Additively Manufactured Porous Metallic Biomaterials,” International Journal of Fatigue 84 (2016): 67–79. Standard Test Method for Shear Testing of Calcium Phosphate Coatings and Metallic Coatings, ASTM F1044-05(2017)e1 (West Conshohocken, PA: ASTM International, approved December 1, 2017), https://doi.org/10.1520/F1044-05R17E01 Standard Test Method for Shear and Bending Fatigue Testing of Calcium Phosphate and Metallic Medical and Composite Calcium Phosphate/Metallic Coatings, ASTM F116014(2017)e1 (West Conshohocken, PA: ASTM International, approved December 1, 2017), https://doi.org/10.1520/F1160-14R17E01 Standard Test Method for Tension Testing of Calcium Phosphate and Metallic Coatings, ASTM F1147-05(2017)e1 (West Conshohocken, PA: ASTM International, approved May 1, 2017), https://doi.org/10.1520/F1147-05R17E01

339

340

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED MATERIALS AND PARTS

STP 1631, 2020 / available online at www.astm.org / doi: 10.1520/STP163120190133

Guney Mert Bilgin,1 Cansinem Tuzemen,1 Cemre Tigli,1 and Yesim Nur Gulcan1

Investigation of Microstructure and Mechanical Properties of SLM-Produced Inconel 718 and Hastelloy-X Alloys Citation G. M. Bilgin, C. Tuzemen, C. Tigli, and Y. N. Gulcan, “Investigation of Microstructure and Mechanical Properties of SLM-Produced Inconel 718 and Hastelloy-X Alloys,” in Structural Integrity of Additive Manufactured Materials and Parts, ed. N. Shamsaei and M. Seifi (West Conshohocken, PA: ASTM International, 2020), 340–351. http://doi.org/10.1520/ STP1631201901332

ABSTRACT

Inconel 718 (IN718) and Hastelloy-X (HX) samples were manufactured by utilizing the selective laser melting (SLM) method in two different specimen orientations: parallel and perpendicular to the build direction; moreover, IN718 samples were subjected to precipitation hardening treatment. Microstructural investigation, densification behavior, and hardness values were studied along with an examination of uniaxial tensile properties at room temperature and 649 C. Microstructural investigations showed that grain orientations strictly depended on heat flow direction where the columnar grains were formed perpendicular to the base plate. A subsequent heat treatment led to the disappearance of melt pool boundaries while laser tracks were still visible for IN718. Hardness properties were not affected considerably by build direction for both IN718 and HX materials in as-built and heat-treated conditions. Yield and ultimate tensile strengths (UTS) were improved 84% and 43%, respectively, for IN718 samples after subsequent precipitation hardening heat treatment along with a

Manuscript received November 1, 2019; accepted for publication April 1, 2020. 1 Tusas Engine Industries Inc., Esentepe Mahallesi, Cevreyolu Blv 356, TR 26210, Eskisehir, Turkey G. M. B. http://orcid.org/0000-0002-8909-0633, C. T. http://orcid.org/0000-0002-8484-4168 2 Fourth ASTM Symposium on Structural Integrity of Additive Manufactured Materials and Parts on October 7–10, 2019 in National Harbor, MD, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

.

BILGIN ET AL., DOI: 10.1520/STP163120190133

remarkable decrease of 43% in elongation at room temperature. A consistent increase in yield and UTS for heat-treated samples was observed when the uniaxial tensile tests were carried out at an elevated temperature. The room temperature and elevated temperature tensile properties were also investigated for HX samples in as-built condition within this study. It was found that HX is a very promising alloy, particularly for high-temperature aerospace applications. Keywords selective laser melting (SLM), Inconel 718, Hastelloy-X, microstructure, mechanical properties, additive manufacturing (AM), heat treatment

Introduction Additive manufacturing (AM) is a technological process in which complex threedimensional (3D) parts are built layer by layer directly from computer aided design (CAD) data. AM has remarkable advantages over conventional manufacturing methods such as freedom to produce complex geometries, removal of expensive tooling, short lead time, optimum material usage, and manufacturing of near net shape parts with good dimensional accuracy.1–3 Selective laser melting (SLM) is an innovative powder bed additive manufacturing technology that can be used directly for fabrication of metallic parts from the CAD model by using a high-power, focused laser beam. Furthermore, near net shape or complex geometries (or both), which are especially used in the aerospace industry, could be produced by controlling the melting of subsequent powder layer by layer under an inert atmosphere. The SLM process offers several advantages in comparison to traditional manufacturing technologies. For instance, it reduces production steps, increases flexibility, provides high efficiency, and allows the production of parts with high relative density (99.99%) without expensive tooling or mold costs.3 Inconel 718 (IN718) and Hastelloy-X (HX) are nickel-based superalloys that have been widely used in stationary and rotational hot section aero-engine components such as disks, combustion chambers, blades, and vanes up to 600 to 700 C. 4–6 IN718 is a precipitation hardenable material that is especially used for production of high-temperature parts in gas turbine engines due to its superior creep and fatigue properties.7 IN718 is mainly strengthened by the precipitation of c0 and c00 , which are coherent secondary phases within a face-centered cubic (FCC) c matrix; c0 has a FCC crystal structure with a composition of Ni3(Al,Ti), and c00 has a bodycentered tetragonal (BCT) crystal structure with a composition of Ni3Nb.7,8 HX is another member of the nickel-based superalloy family that is a solid solution strengthened alloy exhibiting high oxidation resistance and high strength at elevated temperatures along with high formability properties. HX alloy is mainly used within combustion chambers in gas turbine engines because of its outstanding elevated temperature performance.9,10 Conventional production of complex IN718 and HX aerospace parts is inconvenient and expensive because of the subtractive fabrication nature of these .

341

342

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

manufacturing methods. Therefore, nowadays, there are many studies about SLM of IN718 and HX parts. Although SLM provides crucial advantages over conventional methods, high cooling rates obtained in the SLM process lead to the formation of a nonequilibrium microstructure.11,12 Tensile properties have not been linked to microstructural properties in detail, although there are several studies in the literature on laser powder bed fusion (LPBF) of IN718 and HX.3,13 Moreover, there is no systematic work on the high-temperature tensile response of IN718 and HX alloys in their service conditions where these alloys are mainly used in hightemperature applications within the aerospace industry.3,13–15 Based on the literature review presented here, current works have not offered a comprehensive study on microstructure-related high-temperature tensile properties. In this study, IN718 and HX samples were produced via the SLM method in two different orientations that were parallel and perpendicular to the build direction. Afterward, precipitation hardening heat treatment was applied to IN718 samples in compliance with AMS 5663 to investigate their behavior in service conditions. The microstructures of the samples were examined with an optical microscope to observe the effect of build direction for the material and the effect of subsequent heat treatment for IN718. In addition, relative densities of samples were analyzed by utilizing the optical image analysis method and were compared to each other to investigate the effect of the build orientation on the density. With the purpose of evaluating mechanical properties in their high-temperature service conditions, room temperature and elevated temperature uniaxial tensile tests were carried out on as-built/heat-treated IN718 and HX samples along with hardness measurements. The tensile behavior of IN718 material was compared to ASTM F3055, Standard Specification for Additive Manufacturing Nickel Alloy (UNS N07718) with Powder Bed Fusion, while the mechanical properties of HX were linked to AMS 5536 because a powder bed fusion standard has not been revealed yet for HX.

Materials and Methods IN718 and HX powders were used to produce as-built samples in this study. IN718 and HX powders were provided by LPW Technology Ltd. and EOS GmbH, respectively. The SEM images and particle size distribution (PSD) results of IN718 and HX powders are seen in figure 1, and the chemical compositions are given in table 1. Powder samples were investigated by using the FEI NOVA 430 NanoSEM. PSD analysis was carried out using a Microtrac SYNC device with a FlowSync unit. According to PSD results of the starting materials, both of the IN718 and HX powders have a Gaussian distribution with D50:33 lm. Whereas the prealloyed IN718 powders have a PSD between 16 and 45 lm for D10 to D90 values, the PSD of prealloyed HX powders is 18 to 51 lm. Cylindrical, rectangular prism, and cubic samples were built in XY and Z directions for utilizing uniaxial tensile tests and metallographic characterization. A total .

BILGIN ET AL., DOI: 10.1520/STP163120190133

FIG. 1 SEM images of (A) IN718 powders, (B) HX powders, and (C) PSD results.

of 16 IN718 and 8 HX samples were produced by an EOS M290 SLM machine. In this experimental study, the following default EOS parameter sets were used for manufacturing both IN718 and HX samples: 295-W laser power, 960-mm/s velocity, 0.11-mm distance hatch spacing, and 10-mm stripe width.16 Cylindrical specimens have a 14.5-mm diameter and an 81-mm height, and prismatic specimens measuring 15 by 14.5 by 80 mm were built to investigate uniaxial tensile behavior in the Z and XY direction, respectively. In addition to these specimens, cubic specimens were manufactured with the dimensions 15 by 15 by 15 mm in order to investigate several properties, including density, hardness, and microstructure. After sample production was completed, blank parts were separated from the base plate by wire electrical discharge machining (EDM), and cut surfaces were cleaned using sandblasting. Vacuum heat treatment according to AMS 5663 was applied to eight of the IN718 specimens, including each type of specimen, in order to investigate the effect of the thermal history on the microstructure and the mechanical properties in service conditions. The chemical composition analysis was done with half of the cubic samples using the optical emission spectrometry (OES) method in accordance with .

343

344

STP 1631 On Structural Integrity of Additive Manufactured Materials and Parts

TABLE 1 Chemical composition of powders and as-built samples for IN718 and HX

IN718

HX

Elements (wt%)

Powder

As-Built

Elements (wt%)

Powder

Ni

52.86

53.2

Cr

20.98

As-Built

21.6

Cr

19.13

19.1

Fe

18.25

18.4

Nb

5.14

5.22

Mo

8.54

8.5

Mo

3.1

3.02

Co

1.44

2.57 0.93

Ti

0.94

0.856

W

0.70

Al

0.44

0.464

Si

0.15

0.1

Mn

0.09

0.058

Ti

0.05

0.09 0.049

Co