Structural integrity of additive manufactured parts 9780803176867, 0803176864

Citation preview

Shamsaei | Daniewicz | Hrabe | Beretta | Waller | Seifi

Structural Integrity of Additive Manufactured Parts STP 1620

ASTM INTERNATIONAL Helping our world work better ISBN: 978-0-8031-7686-7 Stock #: STP1620 www.astm.org

ASTM INTERNATIONAL Selected Technical Papers

Structural Integrity of Additive Manufactured Parts STP 1620 Editors: Nima Shamsaei Steve Daniewicz Nik Hrabe

Stefano Beretta Jess Waller Mohsen Seifi

SELECTED TECHNICAL PAPERS STP1620

Editors: Nima Shamsaei, Steve Daniewicz, Nik Hrabe, Stefano Beretta, Jess Waller, and Mohsen Seifi

Structural Integrity of Additive Manufactured Parts ASTM STOCK #STP1620 DOI: 10.1520/STP1620-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. | ASTM International, issuing body. Title: Structural integrity of additive manufactured parts / editors, Nima Shamsaei, Steve Daniewicz, Nik Hrabe, Stefano Beretta, Jess Waller, Mohsen Seifi. Description: West Conshohocken, PA : ASTM International, 2020. | Series: Selected technical papers ; STP1620 | Includes bibliographical references. | Summary: "This compilation of Selected Technical Papers, STP1620, Structural Integrity of Additive Manufactured Parts, contains peer-reviewed papers that were presented at a symposium held November 6-8, 2018, in Washington, DC, USA. The symposium was sponsored by ASTM International Committee F42 on Additive Manufacturing Technologies, with co-sponsorship from Committees E08 on Fatigue and Fracture and E07 on Nondestructive Testing"-- Provided by publisher. Identifiers: LCCN 2019055470 (print) | LCCN 2019055471 (ebook) | ISBN 9780803176867 (paperback) | ISBN 9780803176874 (pdf) Subjects: LCSH: Three-dimensional printing--Congresses. | Machine parts--Testing--Congresses. | Machine parts--Design and construction--Congresses. Classification: LCC TS171.95 .S77 2020 (print) | LCC TS171.95 (ebook) | DDC 621.9/88--dc23 LC record available at https://lccn.loc.gov/2019055470 LC ebook record available at https://lccn.loc.gov/2019055471 ISBN: 978-0-8031-7686-7 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 May, 2020

Foreword THIS COMPILATION OF Selected Technical Papers, STP1620, Structural Integrity of Additive Manufactured Parts, contains peer-reviewed papers that were presented at a symposium held November 6–8, 2018, in Washington, DC, USA. The symposium was sponsored by ASTM International Committee F42 on Additive Manufacturing Technologies, with co-sponsorship from Committees E08 on Fatigue and Fracture and E07 on Nondestructive Testing. Symposium Chairs and STP Editors: Nima Shamsaei Auburn University Auburn, AL, USA Steve Daniewicz The University of Alabama Tuscaloosa, AL, USA Nik Hrabe National Institute of Standards and Technology Boulder, CO, USA Stefano Beretta Politecnico di Milano Milan, Italy Jess Waller National Aeronautics and Space Administration, HX5 Las Cruces, NM, USA Mohsen Seifi ASTM International Washington, DC, USA

iii

Contents

Overview

ix

Metal Additive Manufacturing Defect Formation and Nondestructive Evaluation Detectability Ben Dutton, Wilson Vesga, Jess Waller, Steve James, and Mohsen Seifi Evaluation of Nondestructive Volumetric Testing Methods for Additively Manufactured Parts Anne-Françoise Obaton, Bryan Butsch, Stephen McDonough, Ewen Carcreff, Nans Laroche, Yves Gaillard, Jared B. Tarr, Patrick Bouvet, Rodolfo Cruz, and Alkan Donmez NDE-Based Quality Assurance of Metal Additively Manufactured Aerospace Parts at NASA, JAXA, and ESA Jess M. Waller, Eric R. Burke, Douglas N. Wells, Charles T. Nichols, ˜o, Johannes Gumpinger, Martin Born, Tommaso Ghidini, Ana D. Branda Tsuyoshi Nakagawa, Akio Koike, Masami Mitsui, and Tsuyoshi Itoh Eddy Current Technologies for Real-Time Monitoring and Post-Process Examination of Laser Powder Bed Fusion Process Evgueni Iordanov Todorov Nondestructive Evaluation of Additive Manufactured Parts Using Process Compensated Resonance Testing Richard A. Livings, Eric J. Biedermann, Chen Wang, Thomas Chung, Steven James, Jess M. Waller, Scott Volk, Ajay Krishnan, and Shane Collins Nondestructive Evaluation of Programmed Defects in Ti-6Al-4V L-PBF ASTM E8-Compliant Dog-Bone Samples Jeong K. Na, John Middendorf, Michael Lander, Jess M. Waller, and Richard W. Rauser

v

1

51

92

129

165

206

Expression of Additive Manufacturing Surface Irregularities through a Flaw-Based Assessment ˜o, Emilie Beevers, Johannes Gumpinger, Ana D. Branda Thomas Rohr, Tommaso Ghidini, Stefano Beretta, and Simone Romano Development and Testing of 316L Stainless Steel Metal Additive Manufacturing Test Articles for Powder Bed Fusion and Directed Energy Deposition Processes Jessica L. Coughlin, Trevor G. Hicks, Patrick S. Dougherty, and Steven A. Attanasio

234

250

Effects of the Shielding Argon Gas Flow on the Mechanical and Physical Properties of Selective Laser Melted Ti-6Al-4V Solid Material Oscar A. Quintana, Nia Hightower, and Robert Rybolt

278

Influence of Intentionally Induced Porosity and Postprocessing Conditions on the Mechanical Properties of Laser Powder Bed-Fused Inconel 625 Jean-Rene Poulin, Morgan Letenneur, Patrick Terriault, and Vladimir Brailovski

294

Elucidating the Impact of Processing Parameters of Additive Manufactured Materials Using Microscale Tensile Samples Michael Duffy, Salahudin Nimer, Marc Zupan, and Steven Storck

313

Additive Manufactured Polymer Test Coupons: Issues and Solutions Rachael Andrulonis, Royal Lovingfoss, and John Tomblin

334

Coupon-Level Mechanical Testing for Polymer-Based Additive Manufacturing Chul Y. Park, Keith E. Rupel, Chelsey E. Henry, and Upul R. Palliyaguru

355

Characterization of Stress Fields Near Pores and Application to Fatigue Lives James C. Sobotka and R. Craig McClung

367

Fatigue and Fracture Behavior of Additively Manufactured Austenitic Stainless Steel Chris San Marchi, Thale R. Smith, Joshua D. Sugar, and Dorian K. Balch

381

Fatigue Behavior of Laser Metal Deposited 17-4 PH Stainless Steel Lakshmi J. Vendra, Anjani Achanta, and Eric Sullivan

399

Fatigue Assessment of 17-4 PH Stainless Steel Notched Specimens Made by Direct Metal Laser Sintering Filippo Berto, Ali Fatemi, Nima Shamsaei, and Seyed Mohammad Javad Razavi

415

A Comparison of Stress Corrosion Cracking Susceptibility in Additively Manufactured Ti-6Al-4V for Biomedical Applications Michael Roach, R. Scott Williamson, Jonathan Pegues, and Nima Shamsaei

423

Fatigue Properties of Additive Manufactured Ti-6Al-4V and CoCr-Alloy Total Ankle Components Dirk Scholvin, Doug Linton, David Tuttle, Ben Walters, Scott Bible, Jesse Fleming, and Jon Moseley

vi

437

Primary Processing Parameter Effects on Defects and Fatigue in Alloy 718 Luke Sheridan, Bo Whip, and Joy Gockel Role of Post-Fabrication Heat Treatment on the Low-Cycle Fatigue Behavior of Electron Beam Melted Inconel 718 Superalloy ¨nther, Florian Brenne, and Thomas Niendorf Thomas Wegener, Johannes Gu An Efficient Test Method for the Quantification of Technology-Dependent Factors Affecting the Fatigue Behavior of Metallic Additive Manufacturing Components Gianni Nicoletto Probabilistic Computational Fatigue and Fracture Modeling of Additive Manufactured Components Robert G. Tryon, Robert McDaniels, Andrew Chern, Michael Oja, Animesh Dey, Ibrahim Awad, and Chad Duty Probabilistic Framework for Defect Tolerant Fatigue Assessment of Additively Manufactured Parts Applied to a Space Component Simone Romano, Stefano Beretta, Stefano Miccoli, and Michael Gschweitl Rapid Additive Qualification Using Probabilistics (RAQUP) Genghis Khan, Natarajan Chennimalai Kumar, Liping Wang, Vipul Gupta, Voramon Dheeradhada, Tim Hanlon, Laura Dial, and Joseph Vinciquerra Performance Signature Qualification for Additively Manufactured Parts under Conditions Emulating In-Service Loading John G. Michopoulos, John C. Steuben, Athanasios P. Iliopoulos, Trung Nguyen, and Nam Phan NASA’s Efforts for the Development of Standards for Additive Manufactured Components Richard Russell

vii

450

465

484

507

526

540

550

573

Overview To ensure the structural integrity of additively manufactured (AM) parts, it is critical to establish feedstock-process-structure-property-performance relationships, particularly where these components and structures are being used in safety-critical applications. ASTM International (ASTM), in coordination with partners from government, industry, and academia, has organized a series of annual symposia to help address these needs. The 3rd ASTM Symposium on Structural Integrity of Additive Manufactured Parts—which took place from November 6–8, 2018, in Washington, DC—was designed to provide a forum for the exchange of ideas regarding the structural integrity of parts fabricated using AM, with a focus on the lack of industry standards and design principles, as well as qualification and certification challenges. The event was co-sponsored by ASTM Committee F42 on Additive Manufacturing Technologies, along with Committees E08 on Fatigue and Fracture and E07 on Nondestructive Testing. It was supported by the ASTM Additive Manufacturing Center of Excellence (AM CoE), National Aeronautics and Space Administration (NASA), National Institute of Standards and Technology (NIST), European Structural Integrity Society (ESIS) TC15, and the U.S. National Congress for Theoretical and Applied Mechanics. Approximately 250 people attended the 3-day event, representing 11 different nations around the world. The symposium included 56 talks by experts from industry, government, and academia, and 3 panel discussions, which covered nondestructive testing and the challenges of AM standardization and adoption in the medical device and aerospace industries. The 27 selected technical papers in this collection 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 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 and in situ monitoring of AM parts, and the below articles are in this category:  

Metal Additive Manufacturing Defect Formation and Nondestructive Evaluation (NDE) Detectability Evaluation of Nondestructive Volumetric Testing Methods for Additively Manufactured Parts

ix

   

NDE-Based Quality Assurance of Metal Additively Manufactured Aerospace Parts at NASA, JAXA, and ESA Eddy Current Technologies for Real-time Monitoring and Post Process Examination of Laser Powder Bed Fusion Process Nondestructive Evaluation of Additive Manufactured Parts Using Process Compensated Resonance Testing Nondestructive Evaluation of Programmed Defects in Ti-6Al-4V L-PBF ASTM E8-Compliant Dog-Bone Samples

As the lack of understanding and standardization to capture the structural integrity of AM parts is still a major roadblock against the adoption of AM in load-bearing, safety critical applications, many papers presented relating 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 mechanical test methods to AM parts and innovative approaches to standardization were emphasized in this symposium. The following papers focus on related topics for this category:             

Expression of Additive Manufacturing Surface Irregularities through a FlawBased Assessment Development and Testing of 316L Stainless Steel Metal Additive Manufacturing Test Articles for Powder Bed Fusion and Directed Energy Deposition Processes Effects of the Shielding Argon Gas Flow on the Mechanical and Physical Properties of Selective Laser Melted Ti-6Al-4V Solid Material Influence of Intentionally Induced Porosity and Postprocessing Conditions on the Mechanical Properties of Laser Powder Bed-Fused Inconel 625 Elucidating the Impact of Processing Parameters of Additive Manufactured Materials Using Microscale Tensile Samples Additive Manufactured Polymer Test Coupons—Issues and Solutions Coupon-Level Mechanical Testing for Polymer Based Additive Manufacturing Characterization of Stress Fields Near Pores and Application to Fatigue Lives Fatigue and Fracture Behavior of Additively Manufactured Austenitic Stainless Steel Fatigue Behavior of Laser Metal Deposited 17-4 PH Stainless Steel Fatigue Assessment of 17-4 PH Stainless Steel Notched Specimens Made by Direct Metal Laser Sintering A Comparison of Stress Corrosion Cracking Susceptibility in AdditivelyManufactured Ti-6Al-4V for Biomedical Applications Fatigue Properties of Additive Manufactured Ti-6Al-4V and CoCr-Alloy Total Ankle Components

x

    

Primary Processing Parameter Effects on Defects and Fatigue in Alloy 718 Role of Post-Fabrication Heat Treatment on the Low-Cycle Fatigue Behavior of Electron Beam Melted Inconel 718 Superalloy An Efficient Test Method for the Quantification of Technology-Dependent Factors Affecting the Fatigue Behavior of Metallic AM Components Probabilistic Computational Fatigue and Fracture Modeling of Additive Manufactured Components Probabilistic Framework for Defect Tolerant Fatigue Assessment of AM Parts Applied to a Space Component

Finally, recent advances in standardization, qualification and certification were presented:   

Rapid Additive Qualification Using Probabilistics (RAQUP) Performance Signature Qualification for Additively Manufactured Parts under Conditions Emulating In-Service Loading NASA’s Efforts for the Development of Standards for Additive Manufactured Components

We would like to extend our gratitude to everyone who made this symposium possible. The hard work of the authors, coauthors, our symposium co-chairs and coeditors, peer reviewers, and personnel of ASTM and its AM CoE partners, as well as the support of NASA, NIST, ESIS TC15, and the U.S. National Congress for Theoretical and Applied Mechanics 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 PARTS

STP 1620, 2020 / available online at www.astm.org / doi: 10.1520/STP162020180136

Ben Dutton,1 Wilson Vesga,1 Jess Waller,2 Steve James,3 and Mohsen Seifi4,5

Metal Additive Manufacturing Defect Formation and Nondestructive Evaluation Detectability Citation B. Dutton, W. Vesga, J. Waller, S. James, and M. Seifi, “Metal Additive Manufacturing Defect Formation and Nondestructive Evaluation Detectability,” 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), 1–50. http://doi.org/10.1520/ STP1620201801366

ABSTRACT

Depending on input material, process method, process parameters, and postprocessing, the resulting defect state in as-built and finished additive manufactured (AM) parts can be highly variable and complex. To complicate matters further, the terminology used to describe specific defect types can be archaic or user specific and is in need of global harmonization. A common understanding of the root causes of defects and the effect of defects on relevant properties continues to evolve. In powder bed processing, for example, potential defects can be very small, down to the powder particle size. Defects also can occur because of single or multiple causes. Even when there are multiple causes,

Manuscript received December 14, 2018; accepted for publication May 17, 2019. 1 The Manufacturing Technology Centre Ltd., Pilot Way, Ansty Park, Coventry, CV9 7JU, UK B. D. http://orcid.org/0000-0002-4867-3170, W. V. http://orcid.org/0000-0002-1799-0245 2 NASA-JSC White Sands Test Facility, 12600 NASA Rd., Las Cruces, NM 88012, USA http://orcid.org/ 0000-0002-7847-8376 3 134 Dickenson Ave., Newbury Park, CA 91320, USA http://orcid.org/0000-0002-5799-0151 4 ASTM International, 1850 M St NW #1030, Washington, DC 20036, USA http://orcid.org/0000-00018385-2337 5 Case Western Reserve University, Dept. of Materials Science & Engineering, 10900 Euclid Ave., White 335, Cleveland, OH 44106, USA 6 ASTM Symposium on Structural Integrity of Additive Manufactured Parts on November 6–8, 2018 in Washington, DC, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

1

2

STP 1620 On Structural Integrity of Additive Manufactured Parts

single defect types can be produced that fail by a single failure mode. Alternatively, a single defect type can have several different failure modes. The objective of this paper is to classify and identify types of technologically important defects that occur in AM parts produced by powder bed fusion (PBF) and directed energy deposition (DED). A breakdown of technologically important defects is presented in three sections: the cause, the defect, and detection by nondestructive evaluation (NDE). The effect-of-defect on relevant end-use properties is addressed whenever possible. For example, the effect of lack-of-fusion flaws on ultimate tensile properties and high cycle fatigue life is discussed, thus demonstrating the need to be able to detect such flaws. Thus, although the causes of the defects occurring in PBF and DED parts can be quite different, the actual defects can have some similarities. In general, reliable detection of defects by NDE does not depend on the process cause, but depends more on the size, geometry, and location (and, potentially, the morphology) of the defect as well as the complexity, density, and surface finish of the part. Keywords additive manufacturing, defect/flaw detection, powder bed fusion, directed energy deposition, nondestructive evaluation, fatigue, structural integrity

Introduction As the metal additive manufacturing (AM) industry moves toward industrial production, the need for robust qualification and certification (Q&C) procedures, including those related to nondestructive evaluation (NDE) of as-built and finished parts, becomes more pressing. An important subset of this evolving Q&C domain pertains to critical property measurements of AM parts. Although such measurement procedures are well documented, with various legacy standards for conventional metallic material forms, such as cast or wrought structural alloys, fewer standards are available that allow for systematic evaluation of those properties for AM-processed parts. This is due in part to the general lack of AM-specific standards and specifications for AM parts and processes,1 which are a logical precursor to the material characterization standards, including NDE standards that can be used for a given material system. This paper summarizes some of the important standardization and testing activities centered on AM defect detection and characterization, as well as some of the challenges and limitations associated with using NDE to characterize metal AM parts. In particular, the paper examines parts used in mission- and flight-critical applications in which high consequence of failure, structural demand, and issues associated with part inspectability and defect detectability play major roles. Although the specific definitions relative to commercial aircraft applications may vary depending on the application type (e.g., airframe structures versus propulsion engine components), there is a common denominator in that appropriate

DUTTON ET AL., DOI: 10.1520/STP162020180136

damage tolerance assessment needs to be performed for parts of high criticality.2,3 Two elements of such assessment, fatigue crack growth analysis and NDE, traditionally have relied on industry-accepted standards and methods. The latter area, NDE, is elaborated on in this paper. Note that this paper uses the term “defect” to indicate any anomaly with any size, geometry, or distribution that may occur in a metal AM part, but rigorously, “defects” are to be determined by the effect they have on design structural loads and other conditions a part experiences over its service life. As the level of criticality of AM parts in aviation is expected to increase continuously3 more efforts need to be focused on understanding the fatigue and fracture properties of AM materials, and the corresponding inspection and test qualification methodologies that rely mainly on NDE and destructive tests. In addition to “conventional” fatigue crack initiation mechanisms in homogeneous substrate materials, crack initiation resulting from the presence of inherent AM material anomalies. such as porosity, lack of fusion (LOF), or inclusions, also needs to be considered along with the effect of material anomalies on mechanical4–8 and fatigue properties.9–13 Because of the random nature of material defects (not specific to AM materials), the Federal Aviation Administration (FAA) Advisory Circular 33.70-1 defining damage tolerance requirements for engine life-limited parts (LLPs) states that “the probabilistic approach to damage tolerance assessment is one of two elements necessary to appropriately assess damage tolerance.”14 To support such an assessment, the appropriate detection of material anomalies is needed, in addition to conventional fatigue and fracture property testing. Such detection should focus on developing the size distribution and frequency of occurrence of material anomalies. This information can be used to define an exceedance curve15,16 for a given class of material defects, which is the key input into probabilistic fracture mechanics-based assessment, such as the one defined in the FAA Advisory Circulars 33.14-117 and 33.70-218 for specific types of material or manufacturing defects. Note that inspection of fracture-critical AM components for defects needs to be based on a realistic variation of materials and processes so that the material defect state generated in the prototype part or witness coupon is characteristic of that generated in the final production part by the full-scale manufacturing environment.

Harmonized Flaw Terminology As a new technology, AM is in need of industry agreement on definitions of specific terms to communicate flaws and flaw types. In the absence of common agreement about the precise meaning of specific flaws and flaw types, individuals and organizations risk poor product quality, misaligned acceptance requirements, and confusing outcomes related to controlling defects in production builds. As an example, the terms “void” and “pore” in common parlance are synonymous but, in metallurgy or fracture mechanics, they are not. Although both denote a similar defect class, each term has different implied specificity as a distinct class or subclass.

3

4

STP 1620 On Structural Integrity of Additive Manufactured Parts

Although current NDE terminology distinguishes between a flaw (an imperfection or discontinuity that is not necessarily rejectable) and a defect (a flaw that does not meet specified acceptance criteria and is rejectable),19 harmonized terminology for specific AM flaw types is still emerging. As discussed in the Standardization Roadmap for Additive Manufacturing,20 there is a high-priority need for industry-driven standardization with input from experts in metallurgy, NDE, and AM fabrication to identify flaws or flaw concentrations that can jeopardize an AM part’s intended use. As a first step, consensus on flaw terminology must be reached. Preliminary definitions for technologically important flaw types needed in NDE accept/reject of AM parts are included in recently released voluntary consensus standards.21,22 As matters stand, these definitions will need to undergo further vetting by ASTM F42, Standard Test Methods for Conductivity Type of Extrinsic Semiconducting Materials, and ISO TC261 before inclusion in future revisions of ISO/ASTM 52900.23

Defect Causes In each of the seven distinct AM processes categories23–26 the use of different input materials, power sources, energies, and process parameters can yield a variety of “flaw” types. As alluded to earlier, a myriad of terms currently exist that commonly are used to describe these “flaws,” often incorrectly. A literature review of the information available to the public domain was conducted to collage information from globally published works to categorize the different “flaws” produced during the powder bed fusion (PBF) process and directed energy deposition (DED), based on their appearance and location.1,27 The causes and effects of a number of AM flaws have been reported in the European project AMAZE.28 Explanations of the mechanisms by which these flaws are generated are given and those mechanisms are linked to the process parameters selected and the resulting processing conditions. Understanding the conditions under which flaws are generated and simplifying the terminology used to describe these flaws will hopefully aid the drive for quality improvement required for widespread implementation of the technology. The flowchart displayed in figure 1 shows an idea of the complexity of flaw generation within the PBF and DED processes categories.28 As can be seen, the generation of one flaw type can result in an anomalous processing condition, which in turn generates a second flaw. For example, the presence of a thick layer or low laser (or electron beam) power can lead to under-melting, which in turn can lead to unconsolidated powder. Coupled with the tendency of the power source to decrease the surface energy of unconsolidated powder under the action of surface tension, an ensuing ball formation may arise because of shrinkage and worsened wetting, leading to pitting, an uneven build surface, or an increase in surface roughness.29 Therefore, even when there are multiple causes, a single flaw type or conditions can be generated (excessive surface roughness) causing failure by a single failure mode (surface cracking leading to reduced fatigue properties). Alternatively, it is also conceivable that a single flaw type or condition can cause failure by several different failure modes.

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 1 Flowchart showing the complexity of defect generation within the laser PBF process. The legend is as follows: machine inputs/choices (bold), process resulting conditions (italics), AM part resulting defect/flaw (bold italics) and common type of failure (capitals). Source: Courtesy AMAZE EU Project.28

POWDER BED FUSION

The flaws state present in PBF parts can be affected by the power used, the type of energy source (laser or electron beam), the build atmosphere, the processing parameters chosen, and any post-processing steps performed. In fact, more than 50 factors have been identified that may need to be considered when setting up a laserPBF (L-PBF) build. These factors can be grouped into powder, set-up, processing, and part parameter categories.30 In this paper, L-PBF “defects” are emphasized; however, given that the electron beam-PBF (EB-PBF) process is inherently similar, comparable “defects” are likely to occur. Distinct aspects of the PBF process, as compared with conventional metal fabrication processes, such as welding, forging, or casting, include the rapid melting and cooling rate, formation of metastable or nonequilibrium microstructures, narrow melt pool, planar deposition, powder morphology, and the need for support structures. Surface conditions typically feature a layer of partially fused powder and irregularities, such as stair stepping. The as-deposited material in the bulk displays

5

6

STP 1620 On Structural Integrity of Additive Manufactured Parts

banding and planar (horizontal layer) variations in microstructure and may include regions of unfused or partially fused powder. The as-built flaw state will be altered by post-processing procedures, including heat treatment (HT), hot isostatic pressing (HIP), machining, and surface finishing. The biggest challenge in the application of NDE to metal parts made by PBF processes is flaw size variation (lm to mm scale) and location (potentially everywhere). Compared with L-PBF, EB-PBF typically is performed at higher temperatures, which reduces conditions of residual stress and distortion. A vacuum chamber environment offers a high level of purity for reactive metals, reducing defects associated with contamination and pickup of oxygen, nitrogen, and other impurities. Postprocessing removal of powder from internal volumes may be more challenging than with L-PBF. As with L-PBF, HT, HIP, machining, and surface finishing may be required to facilitate the successful application of NDE. DIRECTED ENERGY DEPOSITION

Directed energy deposition (DED) systems are differentiated from PBF systems by the ability to process larger build volumes (>1,000 mm3) with minimal tooling and by the ability to produce parts with either composition gradients or hybrid structures consisting of multiple materials having different compositions and structures. The use of articulated energy sources and feedstock delivered directly to the melt pool is also possible. In contrast to the layer-by-layer PBF screening and melt pass process, powder or wire feedstock for DED is delivered to the melt pool in coordination with the energy source under a shield gas or vacuum. In powder-based DED methods, surface and subsurface conditions and the types of flaws generated are similar to those found in PBF-melted deposits and conventional laser welds. Flaws with the AM-deposited alloy may occur with a dissimilar combination of the base feature alloy. The relatively small and distributed flaw size and rougher surface conditions in DED compared with PBF parts may create challenges for the application of existing NDE methods and for the definition of allowable flaw sizes and locations. Parts made by wire-based DED methods are similar to electron beam welding in conduction mode melting (verses keyhole mode melting). They feature similar flaw morphology and are amenable to evaluation using similar NDE methods. The process creates a large three-dimensional (3D) weld clad build-up to a near-net shape requiring post-deposition machining and often HT or HIP to achieve a final part. Even though HT and HIP may be useful to reduce residual stresses and increase part density, respectively, special care must be taken to ensure that part quality is not decreased because of defects becoming more crack like, in particular with HIP.31,32 Compared with PBF processes, unique aspects of the wire-based DED process include the high-purity shield gas or vacuum environment of the chamber, the extended time at temperature during a build cycle, and associated effects on the asdeposited grain structure, alloy segregation, distortion and residual stress. Operation in a vacuum ensures a clean process environment and eliminates the need for a

DUTTON ET AL., DOI: 10.1520/STP162020180136

consumable shield gas. Flaws are similar to those found in inert gas arc welds and those associated with dissimilar combinations of the base feature alloy (e.g., plate or shaft substrates) versus wire feedstock alloy. DEFECT COMPARISON BETWEEN PBF AND DED

Because of the similarity in manufacturing, defects from welding and casting bear some resemblance to defects from AM processes, such as PBF and DED. Defects in post-built PBF and DED parts are identified and listed in various references.29,33–37 As shown in table 1, both technologies have common defects, such as porosity, inclusions, geometry, LOF, and a rough surface texture. The mechanisms for PBF and DED defect generation are quite different, and more important, the relative abundance of each defect type will be different because if the melting and solidification mechanisms involved.25,36–38 Direct deposition involves imparting a momentum into the melt pool rather than melting the powder that is already present. The important difference between the two methods is that of timescales. In PBF, it is generally described that there is a balance of timescales between melting and resolidification. If the melt rate is insufficient because of deficient laser energy, it results in inadequate melting and weak bonding between layers producing LOF pores.27,39,40 If the melt rate is too high, vaporization of the powder is caused and material is pulled out, which can cause balling or gas porous.27,41,42 Gas porous usually exhibits spherical or elliptic shapes that can be attributed to inert gas entrapment, inert gas dissolved in liquid melt and released on solidification, and metal gas evaporated.43–47 In DED, this balance is not relevant, but the powder (or wire) that is fed into the melt pool must melt sufficiently quickly. The issue of adding cold material (with a given momentum) to a melt pool is not well understood, but it clearly has a large effect on the Marangoni convection direction and thermal gradients present.48 It is likely that the melt-pool depth will be much

TABLE 1 AM (DED and PBF) defect classes and subclasses21

Defect Class

Defect Subclass

Surface

roughness, under/overfill, crater, sag, stair stepping, worm track,

Porosity

gas porosity, keyhole porosity, chained porosity, surface breaking porosity,

Cracking

hot tear, cold craking, crater, HAZ as in DED to substrate, delamination

contour separation, spatter, balling microporosity

Lack of Fusion Part Dimensions Density Inclusions Discoloration Residual Stress Hermetic Sealing

cold lap, trapped powder, oxide lap, linear, planar, post HIP external, internal, lattice, custom (minimum feature size) weight, volume, meets partial density specification feedstock impurity, reaction with chamber contaminants, segregation, banding, planar oxidation reduced mechanical properties, out-of-tolerance dimensions, warping vacuum, pressure

Note: DED ¼ directed energy depositon; HAZ ¼ heat affected zone; HIP ¼ hot isostatic pressing.

7

8

STP 1620 On Structural Integrity of Additive Manufactured Parts

shallower (which may balance the lack of powder surrounding the melt pool) and that the thermal gradients will be less severe (which causes a flatter melt pool), although this depends on the wetting between substrate (which has no surrounding powder) and the melt pool. This difference in the melt-pool dynamics affects its shape. This has two important consequences: grain growth and bubble dynamics. Internal defects are caused by cracking, pores, or lack of material. Cracking has many causes but generally is related to the grain boundary (apart from solidification cracking). Note that the issue of “spattering” that is believed to be prominent in direct deposition (or indeed welding) is still a significant issue in PBF. For L-PBF (e.g., selected laser metling) the issue is that of ablation at the surface of the melt pool caused by the large thermal gradients. For EB-PBF (e.g., electron beam melting) the problem occurs from two mechanisms: ablation and charging of the powder.

AM Flaws Metal parts fabricated by AM can have porosity, LOF flaws, trapped powder, inclusions, stop- and start-type flaws, cracks, residual stress, and poor dimensional accuracy (see figs. 10 through 12 and table 1 in ASTM E3166;21 see table 1 in ISO/ ASTM DTR 5290522). Flaws can be introduced by post-processing or by damage caused by qualification testing before being placed into service. Once in service, additional damage can be incurred as a result of impact; cuts, scratches, and abrasion; loading stresses; thermal cycling; physical aging; and environment effects, such as exposure to air, ultraviolet (UV) radiation, and aerospace media. These factors will lead to complex damage states in the part that can be visible or invisible, macroscopic or microscopic. These damage states can be characterized by the presence of de-densification, depressions, chemical modification, microstructural variation, and foreign object debris (inclusions). Often these discontinuities can placed into four main categories: (1) manufacturing; (2) scratches, scuffs, and abrasion; (3) mechanical damage; and (4) cosmetic damage (no measureable effect on mechanical or end use properties). Because the primary purpose of NDE is to detect flaws produced by manufacturing, technologically important AM flaws produced during manufacturing are summarized in the next section. LIST OF TECHNOLOGICALLY IMPORTANT AM FLAWS

Technologically important AM flaws produced during manufacturing are summarized in table 2, along with definitions, process causes, effect-of-defect, illustrative pictures (micrographs or X-ray computed tomography [XCT] images), and literature references. Flaws can also be grouped into several categories: (1) general bulk flaws (e.g., process-induced porosity, gas porosity, keyhole porosity, cracks/delamination, inclusions); (2) poor dimensional accuracy, warping, and residual stress; (3) unique PBF flaws (e.g., trapped powder, skipped layers, unconsolidated power/

delamination

Crack,

Porosity, keyhole

Porosity, gas

(modulus and Poisson’s ratio), or fatigue life (near surface pores).

nants (moisture or hydrocarbons) in the feedstock or fused part.

species, may be intermittent

within the deposit or elon-

melt pool resulting in hydrodynamic instability of the surrounding liquid metal and subsequent collapse, leav-

formed due to instability of

the vapor cavity during

processing.

tion perpendicular to build axis (Z) can occur leading to delamination or layer separation.

large thermal gradients leading to residual stresses and delamination of a part from the build plate, or cracking in the part, especially in large components. More severe in PBF than in DED. Separation of

transgranular in metals; in

severe cases can result in

two-dimensional (planar)

separation (delamination)

between adjacent build

layers.

incomplete melting.

viously solidified material or

propriate melting overlap with pre-

successive layers because of inap-

separation. Preferential propaga-

and fast cooling rates can lead to

Occurs as embedded or surface

performance.

may be intergranular or

Separation of material which High-intensity (focused) beams

ing a void at the root of the keyhole.

sufficiently high to cause a deep

ized by a circular depression

A type of porosity character- Created when the energy density is

solidification front.

chained due to the moving Minor or no observed effect on

changes in material properties

solidification, or volatile contami-

cient sources of gaseous

gated, interconnected, or

fully dense part, and/or cause in

species (nitrogen, oxygen) during

faceted in shape; with suffi-

Effect-of-Defect

If significant, can cause a less than

Process Cause

Absorption/desorption of gaseous

Definition

Voids that are spherical or

General Bulk Defects

Flaw

TABLE 2 Representative technologically important AM flaws produced during manufacturing Appearance

(continued)

Kempen et al.49

ASTM E3166-20.21

ASTM E3166-20.21

Ref.

DUTTON ET AL., DOI: 10.1520/STP162020180136

9

with surrounding material.

may not have coherency

bination thereof; may or

hydrides, carbides, or a com-

density of the powder bed or build platform shift. Laser-PBF: scan head/optics problems. Electron

tion, while smaller out-of-

tolerance conditions must

be determined by CMM or

cal techniques.

interference.

other appropriate metrologi- beam-PBF: presence of EMF

geometry, beam intensity and the

detected by visual examina-

cal properties

is not performed.

stress especially if post-processing

offset factors are effected by part

The presence of visible steps heating of the build plate. Scaling/

residual stress,

and gross variation can be

mechanical properties due to pre-

fast cooling rates, inadequate pre-

sion from the CAD model.

accuracy, warping,

reduced mechani-

Out-of-tolerance part, reduced

Large thermal gradients, excessive

Variation of the part dimen-

Poor dimensional

properties).

failure (degraded mechanical

material, or build chamber (soot).

rated into the deposit typi-

cally as oxides, nitrides,

centrators and the locus for part

tion of the chamber gas, input

metallic material incorpo-

Effect-of-Defect

Inclusions can serve as stress con-

Process Cause

Formed as a result of contamina-

Definition

Foreign nonmetallic or

Dimensional, Residual Stress, and Warping Defects

Inclusions

Flaw

TABLE 2 Representative technologically important AM flaws produced during manufacturing (continued)

Appearance

Ref.

(continued)

Moylan et al.50

ASTM E3166-20.21

10 STP 1620 On Structural Integrity of Additive Manufactured Parts

component.

cleanliness are required.

cells, and cavities into the part

stress in finished part is perpendicular to the plane of the flaw.

region skipped by the laser or low laser power, whereas stop/start process anomalies are caused by interruption of feedstock supply or short feed (powder starved build area).

flaws are unique to PBF and

both are attributed to inter-

ruption of the build, powder,

or laser such that new layers

are improperly deposited

into previously solidified

(layer defects)

layers.

Deleterious if direction of principle

Skipped layers are attributed to a

Skipped layer and stop/start

stop/start flaws

powder.

design, which trap unfused

be deleterious if high levels of

nal features such as channels, open

cavities.

compared to incorporation of inter- fatigue properties; however, may

trapped within internal part

intended for the part but is

Unmelted powder that is not Process causes play less of a role

Skipped layer and

Trapped Powder

Little/no effect on mechanical or

fatigue properties of the

ing of powder in the hopper or

internally. poor flow properties.

may affect the mechanical or

caused by powder run out, bridg-

powder

uted to incomplete powder feed

Unmelted powder that due

Unconsolidated

melted and became trapped

Effect-of-Defect

to process failure was not

Process Cause

If big size or levels are presented, it

Definition

Unconsolidated powder is attrib-

(continued)

Unique PBF Flaws

Flaw

TABLE 2

Appearance

(continued)

ASTM E3166-20.21

ASTM E3166-20.21

Vesga et al.51

Ref.

DUTTON ET AL., DOI: 10.1520/STP162020180136

11

mechanical properties.

penetration of a deposited layer into the previous layer, or (3)

Can be an empty cavity or

contain unconsolidated

vertical LOF)

are caused by low power/fast scan

detection difficult. LOF can

roughness are correlated, postprocess surface finishing may be necessary to preserve fatigue

correlated with laser or electron beam power and velocity. In general, as or power (P) and build rate (V) increase, feature quality/resolu- properties. tion decrease and surface roughness increases.

der, surface irregularities

such as waviness. Similar to

spatter, undercut, irregular

top bead, ropey bead, and

slumping noted for welded

parts.

Includes partially fused pow- Like keyhole porosity, roughness is

Appearance

Ref.

Seifi et al.2

Vesga et al.51

Note: CAD ¼ computer aided design; CMM ¼ coordinate measuring machine; DED ¼ directed energy deposition; EBM ¼ electron beam melting; EMF ¼ electromagnetic field; LOF ¼ lack of fusion; PBF ¼ powder bed fusion; SLM ¼ selected laser melting.

Surface roughness

Surface and Near-Surface Defects

(vertical LOF, right).

tal LOF) or multiple layers

occur across single (horizon- rate settings.

two successive passes. (1) and (2)

in the bulk, making its

Since fatigue life and surface

less than fully dense, or degrade

ited solidified layer, (2) inadequate

tered (fully melted) parts.

cross-layer defect:

powder. LOF typically occurs improper hatch spacing between

voids, LOF can cause a part to be

ited layer onto a previously depos-

porosity specific to unsin-

horizontal LOF;

Effect-of-Defect

Like other forms of porosity and

Process Cause

(1) Incomplete melting of a depos-

Definition

A type of process-induced

LOF layer defect:

Flaw

TABLE 2 Representative technologically important AM flaws produced during manufacturing (continued)

12 STP 1620 On Structural Integrity of Additive Manufactured Parts

DUTTON ET AL., DOI: 10.1520/STP162020180136

LOF, stop/start flaws); and (4) surface and near-surface defects (e.g., excessive roughness, balling, worm track, small core offset, contour separation, core bleed through). Illustrative examples of the underlined terms are given in table 2. Note that “voids” is a general term encompassing both irregularly shaped and elongated cavities (process-induced porosity), spherically shaped cavities (gas-induced and keyhole porosity), cracks, delaminations, and LOF. These cavities can be empty (gas) or filled with partially or wholly unfused powder (LOF). Voids are distinct from intentionally added open cells to reduce weight. DEFECT CLASSIFICATION STRATEGIES FOR AM

As pointed out elsewhere1 and in this Selected Technical Papers (STP) publication,52 there are longstanding NDE standard defect classes for conventionally manufactured cast, wrought, forged, and welded production parts. The defects produced by these conventional processes generally are not similar to those produced by AM processes. In addition, the NDE signal attenuation characteristics in AM parts may differ from those in conventional parts. Therefore, legacy physical reference standards and NDE procedures must be used with caution when inspecting AM parts.53 This implies that until an accepted AM defect classification such as that discussed here, and associated NDE detection limits for technologically relevant AM defects are established, the NDE methods and acceptance criteria used for AM parts will remain part-specific point designs. Variation of AM process parameters and disruptions during build may induce a variety of defects (anomalies) in AM parts that can be detected, sized, and located by NDE.1 To a large extent, the NDE procedures used to characterize metal AM parts21,22 are based on legacy procedures developed for conventionally manufactured cast, wrought, or welded production parts. Table 1 in ASTM E366121 represents a much-needed AM defect classification that lists technologically important AM defect types present or that may evolve in metal AM parts during build, postprocessing, and service. Some defects are unique to metal AM processes, such as PBF or DED, whereas others are common across all metal manufacturing techniques, including closely related conventional manufacturing processes, such as welding. Other defects such as layer defects, cross-layer defects, trapped powder, and unconsolidated powder are unique to parts made by PBF22 (fig. 2). Application of the NDE procedures21,22 to as-built and post-processed parts can be used to establish a quality record of the defect state before service, thus flagging areas of concern and reducing the likelihood of material or component failure during service. This, in turn, mitigates or eliminates the attendant risks associated with loss of function, and possibly, the loss of ground support personnel, crew, or mission. In addition to defect classification strategies based on NDE detection limits for technologically relevant defects, or acceptance criteria for the minimum allowable defect sizes, a classification strategy based on the physical attributes possessed by defects is also possible and perhaps, is more intuitive. For example, defect morphology, orientation, size, and location have been found to be useful attributes for

13

14

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 2 Unique and non-unique defects in metal parts made by PBF and DED deposition.22

classifying defects and are detailed here—namely, the morphology of a defect pertains to the local shape of a defect as defined by its boundary. For instance, irregularly shaped or elongated voids (process-induced porosity) will have irregular or elongated morphologies, whereas circular voids (gas-induced and keyhole porosity) will have a spherical morphology in the bulk. In contrast, cracks can have a dendritic or, in other cases, planar morphology (as in a delamination). Although these differently shaped voids, cracks, and delaminations may exist as bulk or surface breaking, and the stresses that accumulate around them may be marginally different, each flaw type may contribute similarly to part failure. The orientation of a defect describes the position of a defect relative to some arbitrary direction, usually taken as the machine build or Z-direction. For example, a cross-layer defect that propagates through several adjacent layers can have a different local morphology in any given layer. In addition, because AM processes tend to prohibit generation of volumetric defects with significant height in the

DUTTON ET AL., DOI: 10.1520/STP162020180136

Z-direction, the orientation of defects must be known before NDE inspection, especially when detection sensitivity depends on the defect orientation relative to the inspection direction.54 For this reason, detection of planar defects such as aligned or chained porosity or laminar cracks that form along the build plane can be difficult to detect by incremental step inspection methods, such as XCT. Last, the generation of local defects has many causes, as shown in figure 1, but nevertheless, is fairly well understood.31,33 The mechanism for converting local defects with a given morphology into discontinuities with an orientation, however, is unknown and remains difficult to explain.55 The size of a defect has relevance from both an NDE inspection perspective (the minimum detectable flaw size) and fracture mechanics perspective (the critical initial flaw size, CIFS). It is incumbent upon the fracture control community to define the CIFS in order to define the objectives of the NDE. Knowledge of the CIFS will allow the NDE and fracture control communities to evaluate risks and make recommendations regarding the acceptability of risk. The location of a defect is relevant to the defect’s effect on properties and performance, and whether it can be detected with appropriate volumetric and surfacesensitive NDE methods. Parts with inaccessible regions that preclude or inhibit interrogation with NDE will have high “AM risk” as defined in the Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals.54 Such parts subject to fracture control rules will require not only identification of the inaccessible regions but also knowledge of the CIFS to evaluate risks successfully and to make recommendations regarding the acceptability of risk. Defect location can be further subcategorized as being in the bulk, near the surface (subsurface), or at the surface, and further ranked in terms of effect as little of no effect, unknown effect, out-of-tolerance condition because of poor dimensional accuracy, or degradation of a relevant property56 (see fig. 7 in Waller et al.52). In this regard, defect location will be critically important when assessing the effectof-defect, especially when effects of morphology, orientation, and size are unclear on part properties and performance are unclear. Together, physical defect attributes such as morphology, orientation, size, and location provide a powerful framework for classifying defects and can be used to complement defect classification strategies delimited by NDE capability (minimum detectable flaw size) or acceptance criteria (critical initial flaw size). Ultimately, the goal is to determine which of the physical defect attributes play a prominent role in influencing properties and performance. Further refinement of NDE is possible by looking at still other physical defect attributes related to morphology, orientation, size, and location. For example, total defect volume, average defect volume, defect volume fraction, spatial location, number of defects, defect length (aspect ratio), equivalent spherical diameter, cross-sectional area, and nearest neighbor distance have been correlated with tensile yield strength of 17-4 PH stainless-steel AM dogbone specimens, showing that the number of defects exhibited the strongest correlation to yield strength compared with the other attributes.57

15

16

STP 1620 On Structural Integrity of Additive Manufactured Parts

NDE for AM AM can present challenges to NDE when the geometrical complexity is high, the material is dense, or the part is large. A number of NDE methods have been tried, but for complex geometries, XCT is the preferred option. If sufficient test articles are available to provide a signature for in-family versus out-of-family response, whole-body resonance methods, such as process compensated resonance testing (PCRT), hold promise for inspection of parts with complex geometries. Other wellestablished NDE methods have partial capabilities and may need to be combined to achieve full coverage. These methods include visual testing (VT), radiographic testing (RT), ultrasonic testing (UT), eddy current testing (ET), penetrant testing (PT), thermographic testing (TT), and metrology (MET). As stated on NDE procedures for AM by ISO,21 these NDE methods do not cover all of the inspection requirements that AM parts often need. Advanced in-process monitoring methods, such as laser ultrasonic testing (LUT), infrared (IR) thermographic melt-pool monitoring, and acoustic spectroscopy, hold promise for qualifying parts during build. For the extreme cases in which parts are too large, an alternative or perhaps the only solution would be to perform in-process inspection. DETECTION BY NDE

In addition to the defect attributes of morphology, orientation, size, and location discussed above, the selection of an appropriate NDE method is governed by a range of practical and material considerations. Practical considerations include (1) special equipment or facilities requirements, (2) cost of examination, (3) personnel and facilities qualification, (4) geometrical complexity of the part, (5) part size and accessibility of the inspection surface or volume relative to NDE used (e.g., the ability to detect embedded flaws), and (6) process history and post-processing. Material considerations include (1) defect type and sensitivity of the NDE for that defect; (2) the presence of dissimilar metals (gradient structures); (3) the presence of ferrous or nonferrous metals; (4) surface roughness or finish; (5) residual stress; (6) unique, nonequilibrium microstructures caused rapid cooling; (7) density variations because of thickness variations in parts; and (8) anisotropy because of layer structure in parts. The desired NDE output should be clearly separated from responses originating from surrounding material and configurations and should be applicable to the general range of material conditions, environment, and operational restraints. Complex designs can have regions that cannot be inspected by NDE methods requiring line-of-sight access to the inspection surface or volume. In these cases, as pointed out earlier, inspectors can use NDE techniques based on whole-body response, such as PCRT.58 Alternatively, designers and NDE inspectors can develop a functional rapport to determine part acceptance relative to NDE used and the areas most in need of inspection. Part design features, such as radiographic windows and uniform wall sections, can be specifically added to enhance post-build NDE. One of the advantages of AM is the ability to make design-to-constraint parts, leading to the fabrication of shapes that

DUTTON ET AL., DOI: 10.1520/STP162020180136

cannot be produced using conventional processing methods. Although these parts can be lighter, they can have embedded or intricate features with line-of-sight issues that make them difficult to characterize by NDE.24,38 According to AFRLRX-WP-TR-2014-0162,59 geometric complexity is a primary factor governing the ability to apply post-process NDE to AM parts. Although application of conventional NDE techniques is possible for AM parts with simple geometries, topology optimized AM parts with more complex geometries require specialized NDE techniques. Following AFRL-RX-WP-TR-2014-0162,59 the effect of design complexity on NDE method selection is summarized in table 3. Table 4 lists industrial NDE procedures that are used to inspect AM parts. The ability of each technique to detect different types of defects as well as to locate them in the interior or exterior surface of a part is listed. Finally, the NDE techniques are further characterized by the ability to globally screen or detect and locate a defect. Several reviews of current NDE standards that have potential applicability to some AM defects to determine the type of defects (surface and internal) have been conducted.28,34,36 As part of the background research undertaken, existing standards for welding and castings were reviewed because the manufacturing processes (or the end products) are similar to DED and PBF. Moreover, their applicability to AM defects also have been discussed,34,60–62 and a summary of this work is presented in table 5. Notably, defects 8 to 11 are not typical defects occurring in the current processes and occur only in AM; therefore, the NDE standards require special attention.

TABLE 3 AM design complexity groupings59 Group

1

Design

Description

Simple parts

These parts are simple with well-established designs that do not capitalize on the advantages of AM. Such parts may already have consensus NDE inspection procedures.

2

Optimized standard parts

These parts are based on conventional designs, but some the advantages of AM, such as lighter weight or fewer parts, are incorporated into the design.

3

Parts with embedded

The added features add complexity to a part, thereby

features

decreasing NDE inspectability. Access is limited to internal

Design-to-constraint

These parts appear to be free-formed without straight lines or

parts

parallel surfaces and have no analogue made by conventional

inspection surfaces. 4

subtractive techniques. The presence of detailed external and internal features greatly reduces NDE inspectability because the amount of inspection surface has increased and the vast majority of the structure is detailed and embedded. 5

Lattice structures

These parts consist of a freeform metallic lattices that have a high strength-to-weight ratios, increased surface areas, and tailored stiffness and damage tolerance. The structures pose the greatest challenge for existing NDE technologies, requiring the use of new or creative NDE techniques.

17

Material and Defect Types Detected

velocity and/or sensor-part juxtaposition

affecting sound attenuation, propagation, acoustic

In any solid material, any condition or defect

affecting heat conduction

In any solid material, any condition or defect

reduced mechanical properties

affecting x-ray absorption, with the exception of

In any solid material, any condition or defect

pores, nicks, others)

Any solid material. Discontinuities (e.g., cracks,

Any solid material. Any defect or condition

visible, structured and laser light reflection

In any solid material, any condition r defect affecting

(e.g., porosity)

for local defects (e.g., cracks) and distributed flaws

In electrically conducting and/or magnetic materials

Surface and subsurface

Surface and subsurface

Surface and subsurface

Surface breaking

Surface and subsurface

Surface

Surface and near subsurface

Detects and location

metals

Detects and images location but limited penetration in

Detects and images location but no depth information

Detects and identifies location

Global screening

Detects and images location

Detects and images location

resolution at the expense of scan speed

minimized, focal spot may be optimized for

resolution for parts 10 to 200 mm thick

reduced mechanical properties

affecting x-ray absorption, with the exception of

article and imaging plane Detects and images defect location; field of view

Surface and subsurface, typically 10–200 :m

In any solid material, any condition or defect

dictated by detector size and distance between test

reduced mechanical properties

Global Screening or Detect Location

Detects and images defect location; field of view

Surface or Interior Defect Sensitivity

Surface and subsurface, > 200 :m resolution

affecting x-ray absorption, with the exception of

In any solid material, any condition or defect

Note: XCT ¼ X-ray computed tomography; ET ¼ eddy current testing; MET ¼ metrology; PCRT ¼ process compensated resonance testing; PT ¼ penetrant testing; RT ¼ radiographic testing; TT ¼ thermographic testing; UT ¼ ultrasonic testing.

UT

TT

RT

PT

PCRT

MET

ET

XCT, microfocus

XCT, macro

Method

TABLE 4 General inspection capabilities for selected conventional post-build NDE techniques for additive manufactured parts21

18 STP 1620 On Structural Integrity of Additive Manufactured Parts

incomplete fusion, spatter, sagging,

excess penetration, crater crack, incomplete root penetration,

surface finish, which should be

covered by dimensional

measurement (surface roughness)

Incomplete fusion (DED only)

Post machined

inclusions, undercut, overlap,

porosity), exclusing excludes poor

4

As-built

Surface defect (cracks, voids,

Surface defect (cracks, crater, voids,

characteristic

Nonuniform weld bead and fusion

steps in a part)

groove, excess weld metal,

Post machined

As-built

(continued)

for complex geometry.

For simple geometry but limited

is important.

In DED, direction of radiography beam

RT UT

External defects only.

Yes

Applicable for external defects only.

Yes, ferromagnetic.

Yes

Yes

Yes

geometry is limited.

VT

RT

VT

MPT

PT

VT

VT

Yes Simple geometry is possible but complex

UT

Yes

RT

Post machined

geometry is limited. RT

As-built

3

Porosity

Porosity

Yes Simple geometry is possible, complex

UT

Yes

RT

Post machined

Applicability to AM Defects

RT

NDE

As-built

Condition

2

Void

Defect in Welded or Cast Part

Void

Defect in As-Build AM Part

1

No.

TABLE 5 Review of NDE methods for detection of post-build AM defects34

DUTTON ET AL., DOI: 10.1520/STP162020180136

19

7

6

Undercuts at the toe of the

5

Applicability to AM Defects

complex geometry.

fusion characteristic

Nonuniform weld bead and

Nonuniform weld bead

Post machined

As-built

Post machined

will be difficult to detect.

UT

(continued)

etry; limited for complex geometry.

Only for simple external and internal geom-

get coverage of the defects.

Applicable for external defects only. Direction of X-ray beam is important to

RT

coverage on the defects.

is important to get

Direction of X-ray beam

Applicable for external defects only.

complex geometry.

For simple geometry. Limited for

VT

RT

VT

UT

RT

Inclusions with the same density as the part

will be difficult to detect. Not possible because of surface roughness.

copper)

UT

metallic inclusion other than

Inclusions with the same density as the part

from wire feedstock)

RT

For simple geometry; limited for

coverage of the defects.

Direction of X-ray beam is important to get

RT UT

Applicable for external defects only.

is important.

In DED, direction of radiographic beam

Applicable for external defects only.

VT

RT

VT

NDE

contamination in powder or

As-built

Post machined

As-built

Condition

Inclusion (slag, flux, oxide and

Undercuts

Defect in Welded or Cast Part

Inclusion in part (from

welds between adjoining weld beads

Defect in As-Build AM Part

No.

TABLE 5 Review of NDE methods for detection of post-build AM defects34 (continued)

20 STP 1620 On Structural Integrity of Additive Manufactured Parts

Unconsolidated powder

Cross layer

Layer

Trapped powder

8

9

10

11

None

None

None

None

Defect in Welded or Cast Part

TBD TBD

As-built Post machined

TBD TBD

As-built Post machined

TBD TBD

Post machined

TBD

As-built

TBD

Applicability to AM Defects

Post machined

NDE

As-built

Condition

Note: AM ¼ additive manufactured; DED ¼ directed energy deposition; MPT ¼ magnetic particle testing; NDE ¼ nondestructive evaluation; PT ¼ penetrant testing; RT ¼ radiographic testing; TBD ¼ to be determined; UT ¼ ultrasonic testing; VT ¼ visual testing.

Defect in As-Build AM Part

(continued)

No.

TABLE 5

DUTTON ET AL., DOI: 10.1520/STP162020180136

21

22

STP 1620 On Structural Integrity of Additive Manufactured Parts

In this paper, defects categorized as “non-NDE” are those that would be covered by other types of inspection, for example, metallography. In addition, although the ASTM standard guide on NDE of metal AM parts covers dimensional measurements using laser scanning, structured light, and photogrammetric methods, dimensional measurements are not currently considered in the ISO JG59 standard.22 Notably, XCT, RT, UT, and neutron radiography/tomography (NR) have varying degrees of utility as metrological tools. The use of XCT, in particular, to measure internal dimensions of complex topology-optimized metal AM parts, although deemed important, is still evolving.19 In this regard, AM is helping to spur development and refinement of NDE. In the ISO draft guide for NDE of AM products,22 a new structure for NDE standards for AM is proposed, as shown in figure 3. This structure is based on the current welding standards structure shown by the white boxes, whereas the required development of new standards tailored specifically for AM are shown by the black boxes. Because the evaluation of the AM part integrity can be done either after or during the build, it is important to define which NDE methods would be applicable in both conditions or if specific techniques applied to each situation,27,34,28,35 A study to establish the potential detection capability of the NDE method for post-build and in-process for AM is also included in Brierly et al.34 and Everton et al.35 This study produced a tool to estimate the potential of NDE methods to detect AM defects depending on part complexity and surface finish. When part geometries are not internally intricate and the external surface is not highly complex, NDE methods such as UT will have better capability to detect defects.60,61,63Additionally, if the part has been machined, it will increase the capability of UT and allow for other NDE methods, such as dye PT, to be able to detect surface-breaking defects.5,64 X-ray computed tomography along with acoustic and ultrasonic resonance methods are the only suitable NDE methods for inspecting extremely complex geometries, such as topology-optimized parts and lattices.60,65–67 A study to establish the potential detection capability of NDE methods for AM is also included in Masuo et al.,31 Brierly et al.,34 and Everton et al.35 Previous work developed a selector tool to demonstrate the capability of each NDE method to inspect AM parts for each type of AM defect. This tool considers the micro- and macro-complexity of the part, rather than the potential defects of interest. In addition, it includes a range of component structures and surface finishes, ranging from simple geometries to as-built lattice components. The tool is demonstrated in two extreme cases. In one case, the part is a simple block with machined surface finish. Table 6 shows that commonly used NDE methods such as contact or immersion UT and dye PT for bulk and surfacebreaking defects, respectively, would be capable. In the second case, the part is a complex lattice structure with as built surface finish. Table 7 shows a drastic reduction in the NDE methods capable of this type of detection (only XCT and resonance testing).

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 3 Proposed NDE standards structure for AM. The black boxes indicate development of new NDE standards for AM.22

NDE CAPABILITY DEMONSTRATION SAMPLES

Representative images of trials performed in support of the AMAZE project28 on a titanium alloy calibration artefact are presented in figure 4. Comparison of XCT slice data (fig. 4, left) with a corresponding micrograph of a polished cross section taken from the same location (fig. 4, right) shows that XCT missed flaws smaller than about 45 lm. This result highlights the importance of knowing what the

23

Radiographic

light

Optical / visible

Mechanical

Class

X-ray

Dye-penetrant

recognition

Laminography

(cone beam) CT /

Tomography

Diffraction

Realtime / Digital 2D (fan beam) / 3D

2D

Film / Computed /

visible

Fluorescent /

boroscope etc.

lighting /

Aids such as

Computed

Conventional,

testing

Simple

Resonance

analysis

Acoustic pattern

Diffraction

(air-coupled)

Immersion

Single / twin / array probe, Time of Flight

Contact or

Including

near-contact

Sub-type

Vibration

Ultrasonic

Type

Surface breaking voids

Surface breaking cracks / lack-of-fusion

TABLE 6 Post-built NDE method potential for a simple block with machined surface finish

Near surface residual stress

(continued)

Sub-surface residual stress

Sub-surface microstructure variation

Near surface microstructure variation

Trapped powder (Powder Bed Fusion only)

Inclusions

Internal voids, incl. cross-layer defects

Isolated/ clustered porosity

Internal cracks / lack-of-fusion / layer defects

24 STP 1620 On Structural Integrity of Additive Manufactured Parts

Eddy current

Magnetic field

excited

Vibrationally

excited

Electrically

Measurement

Alternating

Current Field

particle

Barkhausen

Magnetic

Flash

Laser

Sub-type

excited

Type

Optically

(continued)

Electromagnetic

Thermal

Class

TABLE 6

Single / array probe

Thermosonics

Induction-heated

Including Surface breaking cracks / lack-of-fusion Surface breaking voids Internal cracks / lack-of-fusion / layer defects Isolated/ clustered porosity Internal voids, incl. cross-layer defects

Inclusions

Trapped powder (Powder Bed Fusion only)

Near surface microstructure variation

Sub-surface microstructure variation

(continued)

Near surface residual stress Sub-surface residual stress

DUTTON ET AL., DOI: 10.1520/STP162020180136

25

Sub-type

Acoustic

Incidence

Grazing

Photometry

Laser Speckle

Shearography

Spectroscopy

Acoustic

Resolved

Spatially

Laser

Ultrasound

Mechanical

Ultrasound

Transducer

Optical-

Mechanical

Eletromagnetic- Electromagnetic

Type

Including

interferometry

pattern

Electronic speckle

Note: Shaded areas indicate the level of method capability. Source: Courtesy AMAZE EU Project.28

Mixed

Class

Internal cracks / lack-of-fusion / layer defects

Surface breaking voids

Surface breaking cracks / lack-of-fusion

Post-built NDE method potential for a simple block with machined surface finish (continued)

Isolated/ clustered porosity

TABLE 6

Sub-surface residual stress

Near surface residual stress

Sub-surface microstructure variation

Near surface microstructure variation

Trapped powder (Powder Bed Fusion only)

Inclusions

Internal voids, incl. cross-layer defects

26 STP 1620 On Structural Integrity of Additive Manufactured Parts

Thermal

Radiographic

light

Optical / visible

Mechanical

Class

recognition

Tomography

Flash

Thermosonics

excited

Induction-heated

Vibrationally

Laser

Laminography

(cone beam) CT /

Computed

Diffraction

Real-time / Digital 2D (fan beam) / 3D

2D

Film / Computed /

Fluorescent / visible

boroscope etc.

Aids such as lighting /

Conventional,

Resonance testing

Acoustic pattern

Diffraction

(air-coupled)

Immersion

Single / twin / array probe, Time of Flight

Contact or

Including

nearcontact

Sub-type

Electrically excited

Optically excited

X-ray

Dye-penetrant

Simple

Vibration analysis

Ultrasonic

Type

Internal cracks/ lack-of-fusion/ layer defects

Surface breaking voids

Surface breaking cracks / lack-of-fusion

TABLE 7 Post-built NDE method potential for a complex lattice structure with as-built surface finish

Sub-surface residual stress

Near surface residual stress

(continued)

DUTTON ET AL., DOI: 10.1520/STP162020180136

27

Sub-surface microstructure variation

Near surface microstructure variation

Trapped powder (Powder Bed Fusion only)

Inclusions

Internal voids, incl. cross-layer defects

Isolated/ clustered porosity

Ultrasound Microscopy

Grazing Incidence

Photometry

Laser Speckle

Shearography

Acoustic Spectroscopy

Laser Ultrasound

Spatially Resolved

Optical-

Transducer Ultrasound

Acoustic

Electromagnetic

Measurement

Current Field

Alternating

Barkhausen

Magnetic particle

Sub-type

Mechanical

Mechanical

Eletromagnetic-

Magnetic field

Eddy current

Type

Note: Shaded areas indicate the level of method capability. Source: Courtesy AMAZE EU Project.28

Mixed

Electromagnetic

Class

interferometry

speckle pattern

Electronic

Single / array probe

Including

Isolated/ clustered porosity

Internal cracks/ lack-of-fusion/ layer defects

Surface breaking voids

Surface breaking cracks / lack-of-fusion

TABLE 7 Post-built NDE method potential for a complex lattice structure with as-built surface finish (continued)

Sub-surface residual stress

Near surface residual stress

Sub-surface microstructure variation

Near surface microstructure variation

Trapped powder (Powder Bed Fusion only)

Inclusions

Internal voids, incl. cross-layer defects

28 STP 1620 On Structural Integrity of Additive Manufactured Parts

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 4 Computed tomogram (left) and corresponding micrograph of a cross-section (right) of a titanium alloy artifact showing location of defects present in both (circled in green and pointed by arrow) and present only on the cross-section (other circled in red). The limit of detection for this particular sample using XCT was estimated at 3 times the voxel size or 123 μm. Source: Courtesy AMAZE project.28

minimum detectable flaw size is that can be reliably detected relative to the critical initial flaw size, thus ensuring flaws larger than the CIFS are detected and that appropriate part screening by NDE is performed. It also highlights the need to develop voluntary consensus organization (VCO) standards for the design and fabrication of artifacts or phantoms appropriate for demonstrating NDE capability as identified in the ANSI/American Makes standardization roadmap.20 These artifacts or phantoms can conceivably contain intentionally added naturally occurring flaws (LOF, porosity, skipped layers, stop/start flaws) or programmed features, such as watermarks or embedded geometrical features simulating the noted natural flaws. Given the fact that many of the flaws inherent to AM processes are still being identified, their effects on properties and performance are being assessed, and their process causes are being determined, fabrication of artifacts or phantoms with naturally occurring flaws represents a lower state of technological readiness. For this reason, development of first-generation NDE capability demonstration samples has focused on incorporating programmed features, such as embedded geometrical features simulating natural flaws.

Samples with Programmed Features Extensive knowledge gained by the Manufacturing Technology Centre (MTC) from the AMAZE project on NDE capability and typical PBF defect types encouraged MTC to carry out other projects in this area. Projects included ISO TC261 JG59 standard development,22 along with the design, fabrication, and characterization of artifacts to demonstrate NDE capability. The objective of the artifacts design is to

29

30

STP 1620 On Structural Integrity of Additive Manufactured Parts

simulate AM defect types with geometric features not covered by current NDE physical reference standards or XCT/RT image quality indicators (IQIs), namely, layer, cross-layer, unconsolidated powder/trapped powder, and inclusions. Some of these defects (layer flaw and cross-layer flaw) also are being simulated in Inconel 718 XCT capability demonstration standards (see fig. 11 in “Visual Guide to the Most Common Defects in Powder Bed Fusion Technology”68) fabricated by Concept Laser and circulated to XCT users to develop and refine XCT procedures in support of ASTM Committee E07 standard development. Two approaches are considered: the star artifact and an a` la carte approach. The first approach uses a generic geometry star, which will provide initial indication of defect detectability using a number NDE methods, and may be used for relatively simple geometries. The three design are as follows: version S0: No nominal defects, version S1: seeded defects are positioned close to the tip and on the surface of the internal pentagon, whereas version S2 contains the same defects as S1 but are positioned at the corner, away from the tip. Figure 5 shows the design and figure 6 shows the builds in two different materials.69 These samples provide an initial capability calibration of the NDE method, which can be compared with the results obtained in the ISO TC 261 JG59 standard.22 These samples will be available after the JG59 standard is published so that capability comparisons can be performed. An example of the XCT results on an Inconel start artifact is presented in figure 7. The 3D XCT scan data show build limitations for feature sizes smaller than about 200 lm, and XCT capability limitations caused by beam hardening, especially at hard corners, as indicated by shadowing and haloing effects. Other NDE methods examined in trials conducted in support of the development of the JG59 standard have included PCRT, thermographic testing, and nonlinear acoustic (NLA) methods, which have been carried out using Hastelloy and Maraging alloys.22 Results show that PCRT has demonstrated the capability to differentiate between designs with defects (S1 and S2) and design without defects (S0) for both materials using the vibrational pattern recognition (VIPR) analysis. Hastelloy results are shown in figure 8. NLR results were inconclusive for all the samples tested, however, and it was concluded that either the samples do not exhibit significant nonlinear resonance features, or higher excitation levels are required. This method, however, has shown some capability on a different geometry inspected in the AMAZE project28 and lately on the RASCAL project to detect close (touching cracks). An example of RASCAL results is shown in figures 9A–9C, in which the frequency shifts for the defective part compared with the nondefective one. Figure 9C shows the nonlinearity factors (NLF) for each hinge, demonstrating clear discrimination between intact and defective samples: it is possible to set a threshold (e.g., at NLF ¼ 100) above which the high-strain data for the defective sample are unequivocally located and below which the high-strain data for the intact sample clearly fall. An advantage of this method is that it does not rely on a high volume of parts and statistical data to make the pass/fail decision. For cases in which NDE capability needs to be demonstrated for an AM part with a geometry that is considerably more complex than that of the star artifact

FIG. 5 Star artifact design S1 and S2 developed for ISO TC 261 JG59 standard.69

DUTTON ET AL., DOI: 10.1520/STP162020180136

31

32

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 6 Example of built star artifacts for ISO TC 261 JG59 standard built to demonstrate general NDE capabilities.69

R

FIG. 7 X-ray computed tomography example from InconelV star artifact for JG59 standard.

22

discussed previously, a second approach is provided. An a` la carte design, which provides specific steps to follow to demonstrate NDE capability for complex AM parts, is provided for illustration. The main steps are as follows: (1) identify

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 8 PCRT results obtained on Hastelloy alloy star artifact for ISO TC 261 JG59 standard using Vibrational Pattern Recognition analysis (VIPR).69

structurally critical defect locations and sizes for which detection is required, ideally through modeling or mechanical test; (2) fabricate a capability demonstration artifact with programmed features, such as embedded cylindrical holes, cones, and slots simulating defects in the complex production part to be inspected; and (3) inspect with NDE methods (e.g., UT, XCT, resonance, and other emerging NDE methods that show promise) to verify the method is capable of inspecting the part according to part inspection requirements for defect detection. The geometry shows where defects were incorporated at geometrically critical locations and at required defect sizes as displayed in figure 10. An additional reason for an a` la carte design is that all AM designs are different, and it is not feasible to generate a specific IQI or physical reference standard for all possible geometries. GE Power provided to MTC an airfoil design as part of a single-client project with their approval to use it for the standard development. The customer selected embedded cylindrical holes only, with different diameters that opened to the surface. These holes were located on various locations on the airfoil, as per their part inspection requirements. The primary focus was to determine the sensitivity of the XCT detection of the same defects (sizes and shapes) at different locations in the airfoil. Results are shown in figure 11. SAMPLES WITH NATURAL FLAWS

The effect of a given defect type, size, and distribution on end-use properties (effect-of-defect) cannot be determined unless test specimens containing the defect

33

34

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 9 3D XCT images of the two Nacelle Hinges inspected. (A) Sample ND1 (planned without defects); (B) sample ND2 exhibiting a process crack on one of its arms; (C) Nonlinear Resonance curves of both samples showing that ND2 exhibits a decreasing Nonlinearity factor compared with ND1. The envelope curves represent þ/ 1 standard deviation calculated using 10 consecutive shots. The tests were performed by Theta Tech. Source: Courtesy RASCAL Project.51

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 10 Example of a` la carte design based on generic airfoils (design provided by GE Power).22

FIG. 11 Computed tomography scan of a` la carte design based on generic airfoils. Cylindrical holes representing layer defects with diameter 100, 300, 500, and 700 μm are shown.22

of interest are tested to failure. Natural flaws, for example, horizontal LOF, skipped layers, stop/start flaws, inclusions, gas porosity, or excessive surface roughness, can degrade mechanical properties such as strength or fatigue life. In addition, NDE

35

36

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 12 Relationships among AM process parameters, thermal history solidification, microstructure and mechanical behavior.70

detection of an indication attributable to a natural defect is not necessarily grounds for rejection unless the structural effect of the defect on the designed loads has been demonstrated. Therefore, the relevance of detection of a given flaw type, size, and distribution in a production part can be in doubt unless physical reference flaw specimens with the same flaw types, sizes, and distributions are evaluated using the same NDE procedure.62 Figure 12 illustrates the interrelations existing among AM process parameters (top), thermal history, solidification, microstructure, and mechanical behavior (bottom). The objective is to fabricate such flaw specimens, beginning with known service loads and microstructural models using a reverse engineering approach.70 In addition to incorporating programmed geometric feature simulating defects into an NDE capability demonstration part, such as was shown for the star and air foil artifacts discussed earlier, demonstration parts can be made that contain seeded natural flaws, for example, controlled porosity, skipped layers, and stop/start flaws. When manufactured as a permanent asset, these parts can be used to demonstrate the NDE capability for a given defect type, size, and distribution. An example is presented in figure 13, which shows a natural defect that has been detected by LUT and indicates the potential of such NDE method for in-process inspection.71 Conversely, when these seeded flaws are incorporated into a sacrificial coupon, for example, an ASTM E8, Standard Test Methods for Tension Testing Methods of Metallic Materials–compliant dog-bone specimen,72 these parts can be used to determine effect-of-defect.1 Development of a VCO standard has been proposed under the joint jurisdiction of ASTM F42 and ISO TC 261 to provide guidance for making

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 13 Natural defects generated during DED process, where left image is the part photo, middle is a vertical XCT slice and right is the LUT B-Scan. Parabolas indicate the presence of the subsurface natural voids. The yellow arrow highlights Rayleigh wave, longitudinal wave (4L) with a red arrow, shear wave (2S) with blue arrow and amber arrow indicate the longitudinal-shear wave (LS). Source: Courtesy OpenHybrid project.71

NDE capability demonstration parts and sacrificial coupons with intentional or programmed defectsthat would address Gap NDE2 identified in the ANSI/American Makes standardization roadmap.20 RELIABILITY OF NDE DETECTION

NDE has become an integrated component of production systems to assess quality and integrity of the parts manufactured or in-service. The capabilities of various NDE methods are quantified by their reliability, which is defined as the capability of detecting a defect or crack in a given size group under the inspection conditions and procedure specified.73 The reliability in NDE is created based on the following elements: applicability, reproducibility, repeatability, and capability. The latter element, capability, is accounted for by the probability of detection (POD) of a flaw or defect for the method used.74,75 The use of the POD functions as a metric for quantifying the NDE capability that underwent considerable development in the late 1960s and early 1970s. This concept was established primarily at NASA in the 1970s, and it became the fundamental element of engineering design, risk, and lifetime assessment and prediction. The reliability of NDE is expressed in terms of the defect size or crack (a) having a detection probability of 90%. The previously mentioned crack size is known as the a90 crack size.76 A certain statistical uncertainty is associated with the value of a90, which is represented by a 95% confidence interval. This crack size is represented as a90/95 and serves as an important parameter to quantify NDE reliability77–80 for cast and wrought metals parts. This criterion may

37

38

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 14 POD curve as a function of six different heights of defect (0.5, 1.25, 2, 2.75, 3.5 and 4.25 mm with a90/95 value.81

not apply to AM, as stated elsewhere.21,54 An example of a POD curve is shown in figure 14.81 Such curves are determined either through an experimental approach80–82 or through numerical simulation.82,83 The former large-scale experiments make the process timely and costly, whereas the latter, known as modelassisted POD (MAPOD), can reduce time and cost but is not widely recognized as a reliability methodology. Since its introduction by NASA, POD has been improved progressively by the aeronautic industry in the United States to generate guide MIL-HDBK-1823A, which was published in 2009.84,85 At present, this document is considered to be the state-of-the-art guidance for conducting POD studies. For example, the nuclear industry under the European Network for Inspection and Qualification (ENIQ) has released several reports to introduce this technology in Europe.86 and in the oil and gas industry, the qualification of NDE is performed by Det Norske Veritas.87 The methods have been standardized by ASTM E2862, Standard Practice for Probability of Detection Analysis for Hit/Miss Data,88 and ASTM E3023, Standard Practice for Probability of Detection Analysis for aˆ Versus a Data,89 which are congruent with the current MIL-HDBK-1823A methodology.85 Two related probabilistic methods can be used to analyze reliability data and produce POD curves as a function of the flaw size a. The way of recording the data define the method to be used. If the data are recorded in terms whether or not the flaw was detected, it is a discrete data and is called “hit/miss” data. If there is more

DUTTON ET AL., DOI: 10.1520/STP162020180136

information response, such as voltage, signal amplitude, or light intensity, this signal response can be interpreted as the perceived flaw size. This type of data is called aˆ (“a hat data”) or “signal response” data and is continuous data.76,90 Numerous references.76,78,91–95 provide more detail covering the theoretical POD methodology aspect, which is beyond the scope of this paper. Although several advanced NDE methods are available for use, the vast majority of the NDE techniques used for quality control and in-service inspection are conventional. Therefore, it is vital to compare the capability of these methods on various attributes as shown in table 8.96 It is clear that reliable (i.e., “special”) NDE methods are needed to assess manufactured parts or components made by AM, which contain smaller defects with sizes, for example, less than 0.5 mm. The methodology of POD reliability studies had been widely applied for the aerospace industry and adopted by other industries, such as oil and gas, railways, steelmaking, and nuclear. Aircraft structures manufactured in titanium have been assessed by X-ray radiographs for the detection of ceramic inclusions in thick titanium structures manufactured using HIP.90 The X-ray radiography results have been analyzed in terms of POD as a function of the shell diameter, using each estimated shell thickness as the titanium thickness for four different face coat formulations. These results were used to improve the face coat formulations, and to improve detectability, as shown in figure 15. The Programme for Inspection of Steel Components (PISC) trials, which have applications to steels used in the nuclear sector, showed that the flaw characteristics (e.g., flaw shape, geometry, and orientation) had a relatively larger influence on the final POD results compared with other physical parameters. Offshore tubular joints trials carried at the University College London considered the detection of fatigue cracks. The a90/95 POD confidence limit combination was achieved for cracks with typical lengths 100 mm. In general, the vast majority of the literature on POD methodology covers applications in which the size of the defects, particularly cracks, are in the order of millimeters. For AM, however, the size of the defects, especially for parts produced by powder-based processes, are in the order of micrometer rather than millimeters. The lack of POD data available for AM is one of the topics that NASA has identified as a challenge to establish the NDE reliability of the AM sector.97 Although some studies have been carried out on POD, the POD methodology is not yet developed for AM because of two main factors: lack of adequate or routine NDE process integration into AM fabrication or inspection, and the specific features present in parts made by AM. The corresponding lack of reliable NDE methods currently available in the market is a key barrier to AM expansion because of the established deficiencies to validate and certify the quality of the components fabricated using AM technologies.98–100 Some effort to reduce this gap has been addressed by NASA and their partners through the Office of Safety and Mission Assurance NDE Program to reduce the gaps associated with NDE application for AM in the aerospace sector, but it remains an issue.97,101 XCT is the NDE

39

3–7

13–25

7–13

3–7

7–13

UT

RT

ET

MPT

PT

>25

>25

>25

13–25

7–13

Joints

G

G

G

F

G

Vibration

F

G

G

G

G

Temperature

F

F

E

P

F

Access

Adaptability for Field Use

H

M

L

L

M

Surface Quality

L

L

M

M

M

Complexity

P

P

G

P

F

Automation

Instrumentation of Method

H

H

M

H

H

Operator Skill

Note: E ¼ excellent; G ¼ good; F ¼ fair; P ¼ poor; L ¼ low; M ¼ moderate; H ¼ high; UT ¼ ultrasonic testing; RT ¼ radiographic testing; ET ¼ eddy current testing; MPT ¼ magnetic particle testing; PT ¼ penetrant testing.

Welds

NDE Method

Sensitivity, Crack Length, mm

TABLE 8 Comparison of the conventional NDE methods used for fatigue cracks assessment in steel highway bridges96

40 STP 1620 On Structural Integrity of Additive Manufactured Parts

DUTTON ET AL., DOI: 10.1520/STP162020180136

FIG. 15 Graph showing the results of assessment of four different facecoat (shell) formulations (A, B, C, and D) as read by four independent operators at the 0.904-in. titanium thickness.90

technology used widely to assess the quality of parts made by AM, and despite its limitations, XCT remains the method that has the potential for POD work.

Fracture Control Considerations As the use of metal AM parts in fracture-critical applications becomes more accepted, the need for Q&C standards covering all aspects of the part life cycle becomes increasingly prevalent. Fracture-critical properties are affected by a variety of factors and two of the dominant factors are defect density and microstructural variation. Although the goal of producing defect-free parts in as-deposited materials remains an area of extreme interest, NDE inspection procedures and various post-processing techniques may continue to be needed in fracture-critical applications. Noncritical locations with no structural demand (i.e., low stress, strain) in fracture critical parts, however, may not require the same damage tolerance or properties as required in highly stressed areas. The type and rigor Q&C procedures, including NDE, will be driven by the relevant industrial sector and end-use application (aerospace, defense, medical, energy, or automotive). For example, the Q&C requirements for high-value, limited-quantity production-run aerospace parts will be more stringent than for high-quantity production-run commodity parts. In aerospace, the added rigor of

41

42

STP 1620 On Structural Integrity of Additive Manufactured Parts

Q&C requirements in large part can be attributed to the safety concerns associated with human space flight and with military and commercial aviation. As is noted in Seifi et al.,1 the key challenge confronting NASA, the Federal Aviation Administration, the Department of Defense, and the commercial aerospace sector is the qualification of fracture-critical AM parts using inspection (NDE) and testing (typically fatigue testing), especially in applications in which structural margins are low and the consequence of failure is high. Such parts use a damage-tolerant rationale and require careful attention. At this time, it is not clear whether defect sizes from NASA-STD-5009.102 which were based principally on an NDE capability study conducted on flat, fatigue-cracked 2219-T87 aluminum panels early in the Space Shuttle program, are applicable to AM hardware, particularly when the as-built AM part surface is still present and surface-sensitive NDE techniques, such as ET and PT, are used. Therefore, care must be taken when using the starting crack sizes from NASA STD 5009 as the departure point for estimating AM part life. To quantify the risks associated with the AM parts, it is incumbent on the structural assessment community, including ASTM Committee E08 on Fracture and Fatigue, to define CIFS for the part to establish the objectives of the NDE. Specifically, more efforts need to be focused on the characterization and understanding of fatigue and fracture properties of AM materials, and the corresponding testing methodologies. In addition to conventional crystallographic fatigue crack-initiation mechanisms in homogeneous substrate materials, crack initiation resulting from inherent AM material anomalies, such as porosity, LOF defects, or inclusions, also needs to be considered.4,5,10,103–106 For fracture-critical parts, when a damage tolerance analysis is used, physical reference samples with simulated CIFS defects may be necessary to demonstrate NDE detection capability. To determine the appropriate level of NDE sensitivity relative to part category (e.g., prototype parts, production parts, nonstructural parts, primary structure fracture critical parts), part categories have been established that fix the required level of NDE sensitivity (i.e., the size of the defect that needs to be detected). Examples of NASA and JAXA part categories have been published.54,107

Conclusions and Recommendations A review of defects in additive manufactured parts, defect terminology and classification, defect causes, the detection of those defects using consensus NDE methods, and the reliability associated with NDE was carried out on available literature and current work. Representative efforts conducted to date, including representative data acquired using consensus NDE methods are summarized, focusing primarily on NDE applied on as-built or post-processed metal parts fabricated using PBF and DED processes. The main conclusion and recommendations are as follows.

DUTTON ET AL., DOI: 10.1520/STP162020180136

•

•

•

•

•

•

•

Defect and flaw definitions are being harmonized but more effort is required to finalize. There is significant concern about ensuring the integrity of metal parts produced by AM, which is impeding the pace of implementation of this technology. For complex AM geometries, only a few NDE methods demonstrate the necessary capability for detection of technologically important flaws, mainly, XCT, RT, and resonance testing. More work is needed on these methods to assess their capability (i.e., POD studies). To reduce the NDE inspection burden and associated costs and to make NDE more effective, it is recommended to run structural models to determine critical geometrical locations and defect sizes that must be detected in AM parts. The results of this work would be used to select the most appropriate NDE method based on capability and cost effectiveness. Post-processing methods, such as HT and HIP, may help reduce residual stresses and increase part density, respectively, but care must be taken to ensure that part quality is not decreased as a result of defects becoming more crack like, in particular with HIP. Relying on NDE performed on parts after build is not ideal because this results in high part cost and high scrap rates as well as difficulties identifying some flaws; however, in-process monitoring methods have not yet fully matured and commercial systems have only recently become available. Until further engineering experience is obtained on commercial in-process monitoring capabilities, and those techniques evolve, reliance on NDE performed after build will continue. In-process monitoring has the value to monitor and potentially control the process to reduce quality issues. In addition, in-process inspection (NDE/ metrology) has the added value of being able to inspect the part as it is built, which for some complex and large AM parts may be the only NDE-based capability solution. The progression of NDE will be the integration of in-process inspection within AM processes, thus enabling immediate fault detection and the potential to initiate corrective actions.

ACKNOWLEDGMENTS

The authors would like to acknowledge the support from the Manufacturing Technology Centre (MTC) through AMAZE, Core Research Program, OpenHybrid, and RASCAL projects. Special thanks to Dominic Nussbaum at GE Power for his continual interest and support, and for providing the generic airfoil for the a` la carte design. The additional support received through the standard committees ISO TC261/ASTM F42 JG59 and ASTM E07 are also appreciated. Thanks to Vibrant NDT Germany and Theta Technologies UK for their work on PCRT and NLA, respectively.

43

44

STP 1620 On Structural Integrity of Additive Manufactured Parts

References 1.

2.

3. 4.

5. 6.

7.

8. 9.

10.

11. 12.

13.

14.

15.

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,” Journal of the Minerals, Metals, and Materials Society 69, no. 3 (2017): 439–455. M. Seifi, A. Salem, J. Beuth, O. Harrysson, and J. J. Lewandowski, “Overview of Materials Qualification Needs for Metal Additive Manufacturing,” Journal of the Minerals, Metals, and Materials Society 68, no. 3 (2016): 747–764. M. Gorelik, “Additive Manufacturing in the Context of Structural Integrity,” International Journal of Fatigue 94, Pt. 2 (2017): 168–177. H. Gong, K. Rafi, H. Gu, G. D. J. Ram, T. Starr, and B. Stucker, “Influence of Defects on Mechanical Properties of Ti-GA1-4 Components Produced by Selective Laser Melting and Electron Beam Melting,” Materials and Design 86 (2015): 545–554. J. J. Lewandowski and M. Seifi, “Metal Additive Manufacturing: A Review of Mechanical Properties,” Annual Review of Materials Research 46, no. 1 (2016): 151–186. M. Awd, J. Tenkamp, M. Hirtler, S. Siddique, M. Bambach, and F. Walther, “Comparison of Microstructure and Mechanical Properties of Scalmalloy Produced by Selective Laser Melting and Laser Metal Deposition,” Materials 11, no. 1 (2017): 17. M. Fousova ´, D. Vojte ˇch, K. Doubrava, M. Daniel, and C.-F. Lin, “Influence of Inherent Surface and Internal Defects on Mechanical Properties of Additively Manufactured Ti6Al4V Alloy: Comparison between Selective Laser Melting and Electron Beam Melting,” Materials 11, no. 4 (2018): 537. W. E. Frazier, “Metal Additive Manufacturing: A Review,” Journal of Materials Engineering and Performance 23, no. 6 (2014): 1917–1928. 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. N. Hrabe, T. Gna¨upel-Herold, and T. Quinn, “Fatigue Properties of a Titanium Alloy (Ti6Al4V) Fabricated via Electron Beam Melting (EBM): Effects of Internal Defects and Residual Stress,” International Journal of Fatigue 94 (2017): 202–210. W. Mohr, Assessment of Literature on Fatigue Performance of Additively Manufactured Metals (Columbus, OH: EWI, 2016). S. Romano, S. Beretta, A. Branda ˜o, J. Gumpinger and T. Ghidini, “HCF Resistance of AlSi10Mg Produced by SLM in Relation to the Presence of Defects,” Procedia Structural Integrity 7 (2017): 101–108. M. Seifi, D. Christiansen, J. L. Beuth, O. Harrysson, and J. J. Lewandowski, “Process Mapping, Fracture and Fatigue Behavior of Ti-6Al-4V Produced by EBM Additive Manufacturing,” in Proceedings of the 13th World Conference on Titanium, ed. A. Pilchak (San Diego, CA: TMS, 2015), 1373–1377. Guidance Material for Aircraft Engine Life-Limited Parts Requirements: Including Change 1, Advisory Circular 33.70-1 (Washington, DC: U.S. Department of Transportation, Federal Aviation Administration, February 24, 2017). M. Gorelik, Y. Lenets, and M. N. Menon, “GT2005-68770. Development of Probabilistic Lifing System for Advanced Turbine Rotor Alloys,” in Proceedings of the ASME Turbo Expo 2005: Presented at the 2005 ASME Turbo Expo, Volume 4: Industrial and Cogeneration. Oil and Gas Applications, Structures and Dynamics (New York: NY: ASME, 2005): 463–466.

DUTTON ET AL., DOI: 10.1520/STP162020180136

16.

17.

18.

19. 20. 21.

22.

23.

24.

25.

26.

27.

28.

29.

30. 31.

32.

R. Corran, M. Gorelik, D. Lehmann, and S. Mosset, “GT2006-90843: The Development of Anomaly Distributions for Machined Holes in Aircraft Engine Rotors,” in Proceedings of the ASME Turbo Expo 2006: Presented at the 2006 ASME Turbo Expo.Volume 5: Marine Microturbines and Small Turbomachinery Oil and Gas Applications, Structures and Dynamics, Parts A and B (New York, NY: ASME, 2006), 941–950. Damage Tolerance for High Energy Turbine Engine Rotors—Including Change 1, Advisory Circular 33.14-1 (Washington, DC: U.S. Department of Transportation, Federal Aviation Administration, March 7, 2017). Damage Tolerance of Hole Features in High-Energy Turbine, Advisory Circular 33.70-2 (Washington, DC: U.S. Department of Transportation, Federal Aviation Administration, August 28, 2009). Standard Terminology for Nondestructive Examinations, ASTM E1316-19a (West Conshohocken, PA: ASTM International, approved August 23, 2019). Additive Manufacturing Standardization Collaborative, Standardization Roadmap for Additive Manufacturing, Version 2.0 (ANSI and NDCMM/America Makes, June 2018). Standard Guide for Nondestructive Examination of Metal Additively Manufactured Aerospace Parts after Build, ASTM E3166-20 (West Conshohocken, PA: ASTM International, approved February 1, 2020), http://doi.org/10.1520/E3166-20 Additive Manufacturing—General Principles—Non-destructive Testing of Additive Manufactured Products, ISO/ASTM DTR 52905 (Geneva, Switzerland: International Organization for Standardization, under development). Additive Manufacturing—General Principles—Terminology, ISO/ASTM 52900 (ASTM F2792) (Geneva, Switzerland: International Organization for Standardization, approved December 2015). S. A. Tofail, E. P. Koumoulos, A. Bandyopadhyay, S. Bose, L. O’Donoghue, and C. Charitidis, “Additive Manufacturing: Scientific and Technological Challenges, Market Uptake and Opportunities,” Materials Today 21, no. 1 (2018): 22–37. V. Bhavar, P. Kattire, V. Patil, S. Khot, K. Gujar, and R. Singh, “A Review on Powder Bed Fusion Technology of Metal Additive Manufacturing,” in Additive Manufacturing Handbook: Product Development for the Defense Industry, ed. A. B. Badiru, V. V. Valencia, and D. Liu (New York: Taylor & Francis, 2017), 751–753. A. Gebhardt and J.-S. Ho ¨tter, “6–Direct Manufacturing: Rapid Manufacturing,” in Additive Manufacturing: 3D Printing for Prototyping and Manufacturing (Munich, Germany: Verlag, 2016): 395–450. E. Yasa, J. Deckers and J. Kruth, “The Investigation of the Influence of Laser Re-melting on Density, Surface Quality and Microstructure of Selective Laser Melting Parts,” Rapid Prototyping Journal 17, no. 5 (2011): 312–327. AMAZE, “Additive Manufacturing Aiming towards Zero Waste and Efficient Production of High-Tech Metal Products,” Manufacturing Technology Centre, 2017, https://perma. cc/X9GV-MZBP R. Li, J. Liu, Y. Shi, L. Wang, and W. Jiang, “Balling Behavior of Stainless Steel and Nickel Powder during Selective Laser Melting Process,” International Journal of Advanced Manufacturing Technology 59, no. 9–12 (2011): 1025–1035. J. P. Kruth and M. Van Elsen, Complexity of Selective Laser Melting: A New Optimisation Approach (Leuven, Belgium: KU Leuven, 2007). 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. V. Popov, A. Katz-Demyanetz, A. Garkun, G. Muller, E. Strokin, and H. Rosenson, “Effect of Hot Isostatic Pressure treatment on the Electron-Beam Melted Ti-6Al-4V Specimens,” in Procedia Manufacturing 21 (2018): 125–132.

45

46

STP 1620 On Structural Integrity of Additive Manufactured Parts

33. 34.

35.

36.

37.

38. 39.

40. 41.

42.

43.

44.

45.

46.

47.

48.

49.

C. Qiu, G. Ravi and M. M. Attallah, “Microstructural Control during Direct Laser Deposition of a b-Titanium Alloy,” Materials and Design 81 (2015): 21–30. N. Brierly, B. Dutton, M. V. Felice, K. Milne, N. Turner, and S. Everton, “NDE as an Enabler for Additive Manufacturing” (paper presentation, 54th Annual Conference of the British Institute of Non-destructive Testing, Telford, UK, September 8–10, 2015). S. Everton, P. Dickens, C. Tuck, and B. Dutton, “Using Laser Ultrasound to Detect Subsurfaces Defects in Metal Laser Powder Bed Fusion Components,” Journal of the Minerals, Metals, and Materials Society 70, no. 3 (2018): 378–383, htps://doi.org/10.1007/s11837017-2661-7 S. M. Kelly and S. L. Kampe, “Microstructural Evolution in Laser-Deposited Multilayer Ti6Al-4V Builds: Part I. Microstructural Characterization,” Metallurgical and Materials Transactions A 35, no. 6 (2004): 1861–1867. P. C. Collins, L. J. Bond, H. Taheri, T. A. Bigelow, M. R. B. M. Shoaib, and L. W. Koester, “Powder-Based Additive Manufacturing: A Review of Types of Defects, Generation Mechanisms, Detection, Property Evaluation and Metrology,” International Journal of Additive and Subtractive Materials Manufacturing 1, no. 2 (2017): 172–209. S. Singh, S. Ramakrishna, and R. Singh, “Material Issues in Additive Manufacturing: A Review,” Journal of Manufacturing Processes 25 (2017): 185–200. C. Zhao, K. Fezzaa, R. W. Cunningham, H. Wen, F. De Carlo, L. Chen, A. D. Rollett, and T. Sun, “Real-Time Monitoring of Laser Powder Bed Fusion Process Using High-Speed Xray Imaging and Diffraction,” Scientific Reports 7, no. 1 (2017). S. Liu and Y. C. Shin, “Additive Manufacturing of Ti6Al4V Alloy: A Review,” Materials and Design 164, no. 107552 (2019): 23. P.-H. Li, W.-G. Guo, W.-D. Huang, Y. Su, X. Lin, and K.-B. Yuan, “Thermomechanical Response of 3D Laser-Deposited Ti–6Al–4V Alloy over a Wide Range of Strain Rates and Temperatures,” Materials Science and Engineering: A 647 (2015): 34–42. 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. H. Attar, M. Bo ¨nisch, M. Calin, L.-C. Zhang, S. Scudino, and J. Eckert, “Selective Laser Melting of In Situ Titanium–Titanium Boride Composites: Processing, Microstructure and Mechanical Properties,” Acta Materialia 76 (2014): 13–22. X. Zhou, D. Wang, X. Liu, D. Zhang, S. Qu, J. Ma, G. London, Z. Shen, and W. Liu, “3DImaging of Selective Laser Melting Defects in a Co–Cr–Mo Alloy by Synchrotron Radiation Micro-CT,” Acta Materialia 98 (2015): 1–16. S. Pang, W. Chen, and W. Wang, “A Quantitative Model of Keyhole Instability Induced Porosity in Laser Welding of Titanium Alloy,” Metallurgical and Materials Transactions A 45, no. 6 (2014): 2808–2818. 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. L. E. Murr, S. A. Quinones, S. M. Gaytan, M. I. Lopez, A. Rodela, E. Y. Martinez, D. H. Hernandez, E. Martinez, F. Medina, and R. B. Wicker, “Microstructure and Mechanical Behavior of Ti-6Al-4V Produced by Rapid-Layer Manufacturing, for Biomedical Applications,” Journal of the Mechanical Behavior of Biomedical Materials 2, no. 1 (2009): 20–32. K. Antony and N. Arivazhagan, “Studies on Energy Penetration and Marangoni Effect during Laser Melting Process,” Journal of Engineering Science and Technology 10, no. 4 (2015): 509–525. K. Kempen, B. Vrancken, S. Buls, L. Thijs, J. Van Humbeeck, and J.-P. Kruth, “Selective Laser Melting of Crack-Free High Density M2 High Speed Steel Parts by Baseplate Preheating,” Journal of Manufacturing Science and Engineering 136, no. 6 (2014): 061026– 061032.

DUTTON ET AL., DOI: 10.1520/STP162020180136

50.

51.

52.

53.

54.

55.

56. 57.

58.

59.

60.

61.

62.

63.

64.

S. Moylan, J. Slotwinski, A. Cooke, K. Jurrens, and M. A. Donmez, Lessons Learned in Establishing the NIST Metal Additive Manufacturing Laboratory (Gaithersburg, MD: U.S. Department of Commerce, National Institute of Standards and Technology, 2013). W. Vesga, B. Dutton, J. R. Wright, D. Rodriguez-Sanmartin, M. Potter, N. Lunn, M. Baratoui, and T. Martin, “Automated Complementary NDT Inspection System for Production of AM Parts– RASCAL” (paper presentation, 58th Annual Conference of the British Institute of Non-destructive Testing, Telford, UK, September 3–5, 2019). J. Waller, E. Burke, D. Wells, C. Nichols, A. Branda ˜o, J. Gumpinger, M. Born, T. Ghidini, T. Nakagawa, A. Koike, M. Mitsui, and T. Itoh, “NDE-Based Quality Assurance of Metal Additively Manufactured Aerospace Parts at NASA, JAXA, and ESA,” 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) J. Waller, B. Parker, K. Hodges, E. Burke, J. Walker, and E. Generazio, Nondestructive Evaluation of Additive Manufacturing State-of-the-Discipline Report, NASA/TM-2014218560 (technical report, NASA Johnson Space Center, Houston, TX, November 2014). Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals, MSFC-STD-3716 (Huntsville, AL: NASA Marshall Space Flight Center, October 81, 2017). B. Zhang, Y. Li, and Q. Bai, “Defect Formation Mechanisms in Selective Laser Melting: A Review,” Chinese Journal of Mechanical Engineering 30, (2017): 515–527, https:// doi.org/10.1007/s10033-017-0121-5 A. Brown, Z. Jones, and W. Tilson, Classification, Effects, and Prevention of Build Defects in Powder-bed Fusion Printed Inconel 718 (Huntsville, ALL NASA MSFC, Huntsville, AL, 2016). J. Madison, L. Swiler, O. Underwood, B. Boyce, B. Jared, J. Rodelas, and B. Salzbrenner, Identification of Defect Signatures in an Additively Manufactured PrecipitationHardened Stainless Steel (Albuquerque, NM: Sandia National Laboratories, 2017). R. A. Livings, E. J. Biedermann, C. Wang, T. Chung, S. James, J. M. Waller, S. Volk, A. Krishnan, and S. Collins, “Nondestructive Evaluation of Additive Manufactured Parts Using Process Compensated Resonance Testing,” in Structural Integrity of Additive Manufactured Parts (West Conshohocken, PA: ASTM International, 2020), 165–205, http:// doi.org/10.1520/STP162020180111 E. Todorov, R. Spencer, S. Gleeson, M. Jamshidinia, and S. M. Kelly, America Makes: National Additive Manufacturing Innovation Institute (NAMII) Project 1: Nondestructive Evaluation (NDE) of Complex Metallic Additive Manufactured (AM) Structures, AFRL-RXWP-TR-2014-0162 (Cincinnati, OH: Air Force Research Laboratory, Materials and Manufacturing Directorate, Wright-Patterson Air Force Base, June 2014). B. M. Sharratt, “Non-Destructive Techniques and Technologies for Qualification of Additive Manufactured Parts and Processes: A Literature Review,” DRDC-RDDC-2015C0352015 (Sharratt Victoria, BC: Research and Consulting, Inc., 2015). Q. Y. Lu and C. H. Wong, “Applications of Non-Destructive Testing Techniques for PostProcess Control of Additively Manufactured Parts,” Virtual and Physical Prototyping 12, no. 4 (2017): 301–321. D. Chauveau, “Review of NDT and Process Monitoring Techniques Usable to Produce High-Quality Parts by Welding or Additive Manufacturing,” Welding in the World 62, no. 5 (2018): 1097– 1118. J. M. Na, J. K. Middedorf, J. Lander, and J. Waller, “Nondestructive Evaluation of Programmed Defects in Ti-6Al-4V L-PBF ASTM E8-Compliant Dog-Bone Samples,” in Structural Integrity of Additive Manufactured Parts (West Conshohocken, PA: ASTM International, 2020), 206–233, http://doi.org/10.1520/STP162020180095 Q. Y. Lu and C. H. Wong, “Additive Manufacturing Process Monitoring and Control by Non-destructive Testing Techniques: Challenges and In-process Monitoring,” Virtual and Physical Prototyping 13, no. 2 (2018): 39–48.

47

48

STP 1620 On Structural Integrity of Additive Manufactured Parts

65.

66.

67.

68. 69.

70. 71. 72. 73.

74. 75. 76. 77. 78. 79. 80.

81.

82.

83.

M. Albakri, L. Sturm, C. B. Williams, and P. Tarazaga, “Non-destructive Evaluation of Additively Manufactured Parts via Impedance-Based Monitoring,” Solid Freeform Fabrication Symposium 8 (2015): 1475–1490. D. Kiefel, M. Scius-Bertrand, and R. Sto ¨ßel, “Computed Tomography Inspection of Additive Manufactured Components in Aeronautic Industry,” in Proceedings of the 8th Conference on Industrial Computed Tomography (2018). V. Aloisi and S. Carmignato, “Influence of Surface Roughness on X-ray Computed Tomography Dimensional Measurements of Additive Manufactured Parts,” Case Studies in Nondestructive Testing and Evaluation 6 (2016): 104–110. “Visual Guide to the Most Common Defects in Powder Bed Fusion Technology,” 2014, https://perma.cc/Q2KV-2F8E B. Dutton, “Non-Destructive Evaluation & Inspection for AM” (presentation, Second ASTM Additive Manufacturing Center of Excellence Workshop, Senlis, France, September 16, 2019). A. Yadollahi and N. Shamsaei, “Additive Manufacturing of Fatigue Resistant Materials: Challenges and Opportunities,” International Journal of Fatigue 98 (2017): 14–31. B. Duttton, “In-Process Industrial Applications of Laser Ultrasound” (paper presentation, Sixth International Symposium on Laser Ultrasonics, Nottingham, UK, July 12, 2018). Standard Test Methods for Tension Testing Methods of Metallic Materials, ASTM E8/E8M16a (West Conshohocken, PA: ASTM International, approved August 1, 2016). W. D. Rummel, “Transfer of POD Performance Capabilities from Simple Shapes to Complex Shapes,” Review of Progress in Quantitative Nondestructive Evaluation, 18 (1999): 2305–2310. G. Dobmann, D. Cioclov, and J. H. Kurz, “The Role of Probabilistic Approaches in NDT Defect-Detection, -Classification, and -Sizing,” Welding in the World 51, no. 5 (2007): 9–15. P. Berens and A. P. Hovey, Evaluation of NDE Reliability Characterisation (Dayton, OH: Wright-Patterson Air Force Base, 1981). G. Georgiou, Probability of Detection (POD) Curves—Derivation, Applications and Limitations (London, UK: Health and Safety Executive, 2006). A. Keprate, “Probability of Detection: History, Development and Future,” Pipeline Technology Journal 8, no. 2 (2016): 41–45. C. Annis, L. Gandossi, and O. Martin, Optimal Sample Size for Probability of Detection Curves, IAEA-CN-194 (Vienna, Austria, 2012). A. Berens, NDE Reliability Data Analysis, Vol. 17 (Cleveland, OH: ASM International, 1989). W. D. Rummel, “Nondestructive Inspection Reliability—History, Status and Future Path,” in Proceedings of the 18th World Conference on Nondestructive Testing (Durban, South Africa, 2010): 608–622. D. Tisseur, “POD Calculation on a Radiographic Weld Inspection with CIVA 11 RT Module,” in Proceedings of the 10th International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurized Components (Cannes, France: JRC-EU, 2013): 123–129. R. B. Thompson, J. L. Brasche, D. Forsyth, E. Lindgren, E., P. Swindell, and W. Winfree, Recent Advances in Model-Assisted Probability of Detection (Washington, DC: NASA, 2009). P. Calmon, B. Chapuis, F. Jenson, and E. Sjerve, “The Use of Simulation in POD Curves Estimation: An Overview of the IIW Best Practices Proposal,” in Proceedings of the 19th World Conference on Nondestructive Testing (WCNDT), (Red Hook, NY: Curran Associates, Inc., 2016), 3392–3398.

DUTTON ET AL., DOI: 10.1520/STP162020180136

84.

A. P. Berens and P. Hovey, “Statistical Methods for Estimating Crack Detection Probabilities,” in Probabilistic Fracture Mechanics and Fatigue Methods: Applications for Structural Design and Maintenance, ed. J. Bloom and J. Ekvall (West Conshohocken, PA: ASTM International, 1983): 79–94. 85. Department of Defense Handbook, MIL-HDBK-1823A: Nondestructive Evaluation System Reliability Assessment (Philadelphia, PA: Department of Defense, 2009). 86. European Network for Inspection and Qualification, European Methodology for Qualification of Non-destructive Testing: Third Issue, EUR 22906 EN (Luxembourg: European Commission, 2007). 87. DNV-RP-G101, 2010. Risk Based Inspection of Offshore Topsides Static Mechanical Equipment, DNV-RP-G101 (Høvik , Norway: Det Norske Veritas, 2010). 88. Standard Practice for Probability of Detection Analysis for Hit/Miss Data, ASTM E286218 (West Conshohocken, PA: ASTM International, approved April 5, 2018). 89. Standard Practice for Probability of Detection Analysis for aˆ Versus a Data, ASTM E3023-15 (West Conshohocken, PA: ASTM International, approved June 15, 2015). 90. F. R. Child, D. H. Phillips, L. W. Liese and W. D. Rummel, “Quantitative Assessment of the Detectability of Ceramic Inclusions in Structural Titanium Castings by X-Radiography,” in Review of Progress in QNDE 18B (Boston, MA: Springer, 1999), 2311–2317. 91. M. Hirsch, R. Patel, W. Li, G. Guan, R. K. Leach, S. D. Sharples, and A. T. Clare, “Assessing the Capability of In-Situ Nondestructive Analysis during Layer Based Additive Manufacture,” Additive Manufacturing 13 (2017): 135–142. 92. C. R. A. Schneider, R. M. Sanderson, C. Carpentier, L. Zhao, and C. Nageswaran, “Estimation of Probability of Detection Curves Based on Theoretical Simulation of the Inspection Process" (paper presentation, BINDT Annual Conference, Cambridge, UK, 2012). 93. I. Virkkunen and M. Ylitalo, “Practical Experiences in POD Determination for Airframe ET Inspection” (paper presentation, 8th Symposium on NDT in Aerospace, Bangalore, India, November 3–5, 2016). 94. E. R. Generazio, Binomial Test Method for Determining Probability of Detection Capability for Fracture Critical Applications, NASA/TP-2011-217176 (Hampton, VA: NASA Langley Research Center, 2011). 95. E. R. Generazio, Interrelationships between Receiver/Relative Operating Characteristics Display, Binomial, Logit, and Bayes’ Rule Probability of Detection Methodologies, NASA/ TM–2014-218183 (Hampton, VA: NASA Langley Research Center, 2014). 96. P. E. Hartbower and P. J. Stolarski, eds., Structural Materials Technology: An NDT Conference, (Lancaster, PA: Technomic Publishing Company, Inc., 1996). 97. J. Waller, D. Wells, S. James, and C. Nichols, Additive Manufactured Product Integrity. NASA NTRS - 2017-0002071- JSC-CN-38994 (Cape Canaveral, FL: NASA, 2017). 98. D. Kanzler, “How Reliable Are the Results of My NDT Process? A Scientific Answer to a Practical Everyday Question” (paper presentation, ESIS TC24 Workshop: Integrity of Railway Structures, Wittenberge, Germany, September 25–26, 2017). 99. D. Kanzler and C. Mu ¨ller, “Evaluating RT Systems with a New POD Approach” (paper presentation, Nineteenth World Conference on Non-Destructive Testing, Munich, Germany, June 13–17, 2016). 100. D. Kanzler and C. Mu ¨ller, “How Much Information Do We Need? A Reflection of the Correct Use of Real Defects in POD-Evaluations,” AIP Conference Proceedings 1706, no. 1 (2016): 200008-1–200008-7. 101. J. Waller, R. Saulsberry, B. Parker, K. Hodges, E. Burke, and K. Taminger, “Summary of NDE of Additive Manufacturing Efforts in NASA,” Review of Progress in Quantitative Nondestructive Evaluation 1650, no. 51 (2015): 51–62.

49

50

STP 1620 On Structural Integrity of Additive Manufactured Parts

102. Nondestructive Evaluation Requirements for Fracture-Critical Metallic Components, NASA-STD-5009 (Washington, DC: NASA Technical Standard, 2019). 103. P. Li, D. H. Warner, A. Fatemi, and N. Phan, “Critical Assessment of the Fatigue Performance of Additively Manufactured Ti–6Al–4V and Perspective for Future Research,” International Journal of Fatigue 85 (2016): 130–143. 104. G. Nicoletto, “Anisotropic High Cycle Fatigue Behavior of Ti–6Al–4V Obtained by Powder Bed Laser Fusion,” International Journal of Fatigue 94, Pt. 2 (2017): 255–262. 105. D. Greitemeier, F. Palm, F. Syassen, and T. Melz, “Fatigue Performance of Additive Manufactured TiAl6V4 Using Electron and Laser Beam Melting,” International Journal of Fatigue 94, Pt. 2 (2017): 211–217. 106. 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, Pt. 2 (2017): 178–191. 107. R. Russell, D. Wells, J. Waller, B. Poorganji, E. Ott, T. Nakagawa, J. Sandoval, N. Shamsaei, and M. Seifi, “Qualification and Certification of Metal Additive Manufactured Hardware for Aerospace Applications,” in Additive Manufacturing for the Aerospace Industry, ed. F. Froes and R. Boyer (Amsterdam, The Netherlands: Elsevier, 2019), 33–66.

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED PARTS

STP 1620, 2020 / available online at www.astm.org / doi: 10.1520/STP162020180099

Anne-Franc¸oise Obaton,1,2 Bryan Butsch,3 Stephen McDonough,4 Ewen Carcreff,5 Nans Laroche,5 Yves Gaillard,6 Jared B. Tarr,1 Patrick Bouvet,6 Rodolfo Cruz,4 and Alkan Donmez1

Evaluation of Nondestructive Volumetric Testing Methods for Additively Manufactured Parts Citation A.-F. Obaton, B. Butsch, S. McDonough, E. Carcreff, N. Laroche, Y. Gaillard, J. B. Tarr, P. Bouvet, R. Cruz, and A. Donmez, “Evaluation of Nondestructive Volumetric Testing Methods for Additively Manufactured Parts,” 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), 51–91. http://doi.org/10.1520/STP1620201800997

ABSTRACT

Additive manufacturing enables the production of customized and complex parts. These two aspects are attractive for the aerospace and medical sectors. In these critical sectors, however, governed by strict safety requirements, the quality of the parts is of paramount importance, and the technology has advanced at a much faster pace than regulations and quality controls. The reliability of the parts must be guaranteed, and hence quality control is needed. Considering the complexity of additively manufactured part shapes, the inspection methods need to be nondestructive, three-dimensional, and volumetric. X-ray Manuscript received October 26, 2018; accepted for publication February 15, 2019. 1 Engineering Laboratory, National Institute of Standards and Technology, 100 Bureau Dr., MS 8220, Gaithersburg, MD 20899, USA A.-F. O. http://orcid.org/0000-0002-5509-3203 2 Laboratoire National de Me ´trologie et d’Essais, 1 Rue Gaston Boissier, Paris, France 3 The Modal Shop, Inc., 3149 E. Kemper Rd., Cincinnati, OH 45241, USA http://orcid.org/0000-0001-57930864 4 OKOS Solutions, LLC, 7036 Tech Cir., Manassas, VA 20109, USA S. M. http://orcid.org/0000-0002-44073952, 5 The Phased Array Company, 8 Bis Rue de la Garde, 44300, Nantes, France E. C. http://orcid.org/0000http://orcid.org/0000-0001-5726-4689 0002-0599-1048 , N. L. 6 Centre Technique des Industries de la Fonderie, 44 Avenue de la Division Leclerc, 92318 Se`vres, France 7 ASTM Symposium on Structural Integrity of Additive Manufactured Parts on November 6–8, 2018 in Washington, DC, USA. This work is not subject to copyright law. ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959.

51

52

STP 1620 On Structural Integrity of Additive Manufactured Parts

computed tomography is presently the most appropriate method, but the relatively high cost and testing duration make routine inspection difficult. Thus, alternative nondestructive volumetric methods are required. In this paper, four alternative methods utilizing acoustic waves (resonant acoustic method, process compensated resonance testing) and ultrasonic waves (conventional ultrasonic testing, phased array ultrasonic testing combined with total focusing method) are investigated and compared with X-ray computed tomography using synchrotron radiation. Keywords nondestructive testing (NDT), volumetric NDT, additive manufacturing (AM), acoustic methods, ultrasound methods, X-ray computed tomography (XCT)

Introduction One of the main advantages of additive manufacturing (AM) is to allow the fabrication of complex-shaped parts, including with inner cavities, that cannot be manufactured with conventional methods from a single block of material. Another main advantage of AM is the possibility to build customized parts for roughly the same price as some high-volume parts. These advantages are highly appreciated in the aerospace and medical sectors. But AM has been shown to produce specific types of defects in as-built parts, such as cross-layer defects, layer-specific defects, and unconsolidated or trapped powder defects. In these critical aerospace and medical sectors, the quality requirements of the parts are particularly severe, and the parts must be certified before they can be used in their intended application. This is a challenge for quality control of parts with such complex geometry and inner cavities, especially those with the high surface roughness levels typical of AM surfaces. The inspection methods not only must be nondestructive, volumetric to screen inner cavities, and appropriate for high roughness and complex shapes, but also inexpensive and fast for routine inspections. Among existing volumetric nondestructive testing (NDT) methods, X-ray tomography is presently the most appropriate method to inspect the inner features and defects of complex AM parts. X-ray tomography is expensive, however, and the scanning process takes a long time. In addition, the resulting scan files are large; thus, they are not easily viewable. So, even if X-ray tomography is recommended for first article inspection (FAI), it is not suitable for routine inspection of high-volume production of customized parts, for example. Alternative methods to X-ray tomography must be explored.1,2 All the NDT volumetric methods that currently are used for the characterization of conventionally manufactured parts require a new understanding of how they interact with AM parts. Then, further development of existing NDT methods, or novel ones, are greatly needed to handle the level of complexity that AM presents. The resonant acoustic and ultrasound NDT linear volumetric methods are potentially valuable techniques that need to be explored for AM parts.

OBATON ET AL., DOI: 10.1520/STP162020180099

This paper presents the capabilities of some of the existing acoustic and ultrasonic volumetric NDT methods to inspect AM parts with specific defects. The artifacts proposed by ISO TC261/ASTM F42 Joint Group 59 (JG59) “NDT for AM parts” are used in this study. The paper also presents results obtained with X-ray computed tomography (XCT) using synchrotron radiation as a comparison. The paper is structured as follows: the first section is dedicated to the presentation of JG59 and their standards development regarding NDT inspections using a specially designed artifact. The second section describes the fabrication of this artifact in different materials, and the last section presents the principles, inspection results on the artifacts, and benefits and drawbacks of the following five volumetric NDT methods: • whole-body inspection methods using acoustic waves: (1) resonant acoustic method (RAM) and (2) process compensated resonance testing (PCRT), • selective inspection methods using ultrasonic waves: (3) conventional ultrasonic testing (CUT) and (4) phased array ultrasonic testing (PAUT) methods, and • X-ray for comparison: (5) XCT using synchrotron radiation (SRXCT).

STANDARDIZATION GROUP ISO TC261-ASTM F42 ON AM: JOINT GROUP JG59 ON NDT FOR AM PARTS

Because AM is causing a revolution in manufacturing capabilities, many industries are interested in taking advantage of these methods. Each time a new technology grows in capability and use, however, standardization issues arise that must be overcome for the technology to continue its advancement. Several standards developing organizations (SDOs) are addressing this subject. The International Organization for Standardization (ISO) and ASTM International have implemented a formal agreement to codevelop a series of AM standards through several joint groups (JG). JG59 of ISO Technical Committee (TC) 261 and ASTM F42 is dedicated to NDT for AM parts. Goal of the JG59

Several standards dedicated to NDT of manufactured parts already exist, and the JG59 does not aim to duplicate these standards. AM is different from other manufacturing processes like casting, forging, and machining. With AM, parts are built layer by layer from raw feedstock, such as metal powder, so one might say that the part is manufactured simultaneously with the material. This also means that the microstructure of the built material and the potential for embedded defects are specific to the AM technology. As mentioned previously, AM enables manufacturers to build complex geometries but with rough surfaces. Therefore, existing NDT methods must be evaluated for AM applicability. These methods might need to be adapted to accommodate AM characteristics.

53

54

STP 1620 On Structural Integrity of Additive Manufactured Parts

Furthermore, new NDT methods also may be needed. Thus, a primary goal of the JG59 “NDT for AM parts” is to write a best practice guide presenting potential NDT methods suitable for post-process inspection of AM metallic parts, referring to existing NDT standards when appropriate. This best practice guide will contain a list of typical AM defects that can be found in two of the main types of metal additive manufacturing process categories used in industry: power bed fusion (PBF) and directed energy deposition (DED). It will give recommendations for NDT methods suitable for the special features of AM. The first version of this guide should be complete by the end of 2020. To develop this guide, joint group members have designed and fabricated several artifacts with typical AM defects in different materials. These artifacts were then characterized by different NDT methods by the JG59 members (table 1) to evaluate the behaviors and potential for each NDT method for use with AM parts.

TABLE 1 NDT technologies investigated by JG59 with star artifacts in different materials Material

Hastelloy X

NDT Technologies

Center/Company/University

X-Ray Computed Tomography

MTC, UK

(XCT) Phased Array Ultrasonic Testing

University of Bristol, UK

(PAUT) Process Compensated Resonance

Vibrant, Germany

Testing (PCRT)

Maraging Steel

Cobalt Chrome

Nonlinear acoustic method (NLA)

Theta Tech, UK

Thermography testing (TT)

University of Bath, UK

Digital Radiography testing (RT)

FujiFilm, USA

XCT

MTC, UK

PAUT

University of Bristol, UK

PCRT

Vibrant, Germany

NLA

Theta Tech, UK

TT

University of Bath, UK

RT

FujiFilm, USA

Resonance Acoustic Method

NIST/LNE and The Modal Shop,

(RAM)

USA/France

XCT

NIST/LNE, USA/France

Synchrotron XCT (SRXCT)

Novitom, CTIF, LNE, France and NIST, USA

Stainless Steel

XCT

EWI, USA

PAUT

EWI, USA

XCT-Low Energy

ISS, Germany

XCT

MTC, UK

RAM

NIST/LNE and The Modal Shop,

PCRT

NIST/LNE and Vibrant, USA/France

USA/France

(continued)

OBATON ET AL., DOI: 10.1520/STP162020180099

TABLE 1

Material

Aluminum

(continued)

NDT Technologies

Center/Company/University

XCT

MTC, UK

Conventional Ultrasonic Testing

NIST/LNE and OKOS, USA/France

(CUT) PAUT–Total Focusing Method

TPAC, LNE, France and AOS, NIST,

(TFM)

USA

Titanium

XCT

MTC, UK

Titanium

SRXCT and Neutron diffraction ND

ESRF and ILL, France

Star Artifact with Typical AM Defects

This artifact was designed by JG59 to be optimized for laboratory 450-kV XCT. The size of the artifact depends on the density of the material. It is in the shape of a star (fig. 1), hence, its name: star artifact. The artifact contains the following defects unique to AM parts: • Cross-layer defects that are represented by vertical cylinders of different diameters but the same length (These cylinders are connected to each other and open to outside at the bottom of the star, so that powder is released at the largest diameter cylinder.) • Layer-specific defects that are represented by horizontal cylinders of different diameters but the same length, with an open end to release powder • Unconsolidated/trapped powder defects that are represented by spheres of different diameters and internal cylinders in various orientations Voids and porosities are common defects also found in conventional manufacturing and are mostly covered by existing NDT standards. These defects are located in critical areas such as deep sections and hard-toreach areas, in five different regions. Two versions of the star design, designated as S1 and S2, were used in the evaluation of NDT methods (fig. 1). In these two

FIG. 1 Schematics of the AM star artifacts (design S1 and S2) proposed by JG59 (where R is a Region in the artifact, h and a define the height and width of the artifact, respectively, and numerical values are given in table 2).

55

56

STP 1620 On Structural Integrity of Additive Manufactured Parts

versions, the defects are of the same size, same height along the part, and the same orientation; the only difference is their locations. The S2 design has the defects in thinner sections of the star branches, whereas the S1 design has the defects in the thicker sections. All defects are in the range of 100 lm to 800 lm in diameter. At NIST, we evaluated statistical NDT methods that required a large number of samples. Considering this fact, we decided to build half-size stars also. The sizes of the defects, however, remained the same. STAR ARTIFACT FABRICATION

The star artifacts were manufactured by different members of the JG59, depending on their capabilities in terms of AM equipment and materials. NIST built star artifacts in cobalt-chromium (Co-Cr) (figs. 2A and 2B) and in stainless steel (SS) (figs. 2C and 2D) using AM machines belonging to the laser PBF process category. Zodiac Aerospace built aluminum star artifacts using also a laser PBF process. The characteristics of the manufactured star artifacts are shown in table 2 and a picture representing the finished star artifacts is shown in figure 3. Because some of the NDT methods investigated require statistical analysis, significantly more artifacts (specifically, SS artifacts) were built for those methods to establish the baseline for artifacts with no defects. Furthermore, to evaluate the capability of such techniques to sort parts based on the numbers of defects they contain, stars with different numbers of defects were manufactured in two similar builds (i.e., the same location on

FIG. 2 Photos of the manufacturing of the Co-Cr (A and B) and SS (C and D) star artifacts, during the build (A and C) and after the build on the machine platform (B and D).

OBATON ET AL., DOI: 10.1520/STP162020180099

TABLE 2 Characteristics of the manufactured star artifacts Material

Version

Size: full/half (HF)

Full (a ¼ 60 mm, h ¼ 45 mm)

Quantity

Cobalt-chrome-

S0

molybdenum-based

S2

superalloy Co-Cr MP1

S0

Half (size of defects/2 also

S2

(a ¼ 30 mm, h ¼ 22.5 mm)

1

Aluminum AlSi10Mg

S1

Full (a ¼ 114 mm, h ¼ 45 mm)

1

Stainless steel 17-4

S0

Full (a ¼ 60 mm, h ¼ 45 mm)

2

S2

1

1

S1

2

S2 S0

1 3

2 Half (a ¼ 30 mm, h ¼ 22.5 mm)

S1

40 8

S2

8

S2

HF without defect in Region 1

8

S2

HF without defect in Regions 1

8

and 2 S2

HF without defect in Regions

8

1, 2, and 3 S2

HF without defect in Regions

8

1, 2, 3, and 4

Note: S0 ¼ a fully dense part, without inner features representing the defects; a ¼ the width of the star; h ¼ the height of the star; see figure 1.

FIG. 3 Photos of the manufactured star artifacts, from left to right: Al, SS, and Co-Cr.

57

58

STP 1620 On Structural Integrity of Additive Manufactured Parts

the AM build platform, with two or three marginally different parts added on the platform between the first and second build): • eight similar parts with defects in the 5 Regions, such as defined by the JG59; • eight similar parts with defects in only Region 1; • eight similar parts with defects in only Regions 1 and 2; • eight similar parts with defects in only Regions 1, 2, and 3; and • eight similar parts with defects in only Regions 1, 2, 3, and 4. Four of the eight stars are from one build, and the other four are from a second build. VOLUMETRIC NDT METHODS EVALUATED

Nondestructive inspection methods can be separated into two main groups: (1) surface and subsurface methods enabling only a surface and subsurface inspection of the test object (e.g., by visual, dye penetrant, magnetic particles, eddy current, thermography for metal) and (2) volumetric methods enabling volumetric inspection of the test object (e.g., by ultrasound, radiography, thermography for polymer, acoustic emission, X-ray tomography).3,4 Some volumetric methods, however, can be limited in terms of the inspectable thickness according to the material of the part tested. Volumetric methods also can be divided into two main groups: whole-body inspection methods and selective inspection methods. The whole-body inspection methods do not provide information about defect locations, whereas selective inspection methods do provide such information. We have investigated volumetric methods from both groups to evaluate their respective potential for inspecting AM parts. Whole-Body Inspection Methods

Resonant Ultrasound Spectroscopy (RUS) Methods Principle of RUS Methods—There are several variants of RUS methods, as described in ASTM E2001, Standard Guide for Resonant Ultrasound Spectroscopy for Defect Detection in Both Metallic and Non-Metallic Parts,5 but their basic principles are similar. These whole-body inspection approaches compare the frequency spectrum of the mechanical resonances of a set of reference parts, known as “good parts,” with the frequency spectrum of the mechanical resonances of the test parts. These are pass/fail assessments. Two similar objects will have similar resonant frequency spectra. A shift of the frequency peaks of a part, compared with the statistical variation from the similar good parts, will be the signature of a structural change of the part (e.g., changes in part geometry or material properties associated with the mass, stiffness, and damping). A characterization using RUS methods includes several steps. The first step consists of a mechanical impulse of the test object to make it vibrate as free as possible. Under this excitation, the object vibrates at its natural resonant frequencies. At the second step, the response of the test object is monitored by a sensor, and the

OBATON ET AL., DOI: 10.1520/STP162020180099

frequency spectrum is recorded. At the third step, this spectrum is compared with the spectrum of the good parts, which are obtained following the same first two steps. For this comparison, the well-defined resonant peaks for each good part are identified. A subset of such resonant peaks that are consistent for all reference parts peaks and have distinct separation with bad parts are selected for further consideration. The ranges (boundaries) in variations in each resonant peak frequency of the good parts are evaluated, these are called “criteria.” Outliers of these variations must be excluded from these ranges. The measured resonant peak frequencies of the test parts are compared against the corresponding ranges in frequencies of the good parts. All test parts with frequency peaks outside of these boundaries are rejected as faulty parts. RAM6 and PCRT, described in ASTM E2534, Standard Practice for Process Compensated Resonance Testing via Swept Sine Input for Metallic and Non-Metallic Parts,7 are two types of RUS methods. Principles of RAM—In the RAM, the excitation is done by an impact hammer (first step) and the resulting acoustic frequency data are obtained with a microphone. No contact probes need to interact with the test part. This allows the parts to resonate in as free a state as possible. The output of the microphone is converted into a frequency spectrum using a high-speed analog-to-digital converter (24 bits) performing a fast Fourier transform to determine resonant peaks (second step). Finally, a statistical analysis conducted with software reveals whether or not the test parts fall within the defined criteria (third step). If at least one of the measured resonant peak frequencies of the test part lays outside of the established frequency ranges of the good part, the test part is considered bad. For a part to pass RAM testing, it must pass every individual criteria point. For a part to fail RAM testing, it needs to fail only one criteria point. RAM Inspection of the SS Half-Size Star Artifacts—Figure 4 shows the setup used for the RAM measurements8 of eighty-eight SS half-size stars (fig. 5), resulting from the

FIG. 4 (A) The Modal Shop RAM set-up (A) and (B) inspection of the half-size SS star artifacts using the set-up with the automatic hammer.

59

60

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 5 CAD of the SS half-size star artifact, S2 design, seen from different sides (R refers to the Region in the star).

two successive builds (table 2). Among them, 20 from each build were manufactured with no defects and were considered as reference parts to define the baseline for the criteria. The stars were systematically weighed before the test as the inspection software enables weight compensation. Then, the star artifacts were hit with an automatic hammer on the sides of their bottom faces, as these positions were experimentally found to provide the least amount of damping. To determine optimum impact positions, a part was hit in several different positions and orientations. The impact location and orientation resulting in the greatest signal-to-noise ratio of the resonant peaks as well as the greatest frequency separation between good and bad parts was selected as the optimum position and orientation for the subsequent tests. When performing the tests, each part was hit three consecutive times at three different branches of the star at the base. This procedure allowed us to check for repeatability and reproducibility. For consistent impacts on every part, the same positions on the same branches of the stars were always struck. The frequency window considered for analysis ranged from 500 Hz to 50 kHz, and a resolution of 3.5 Hz was chosen, which gives an acquisition time of 0.23 s. We considered the first and second build separately. The data on the 20 reference parts were first gathered, and then the remaining 24 stars with defects were tested. The tests were repeated on two consecutive days. The acceptance criteria to define the baseline for comparison of the resonant peaks, from this first set of data, were set up over six frequency bands, as shown in figure 6. This figure shows the complete spectrum for a single part test. The vertical lines labelled 1 to 6 show the bands of the resonant peaks chosen for evaluation. These criteria are used by the software to sort the bad parts from the good. The data for a part must pass all the criteria to pass resonance testing. There is a constant shift, for all peaks, in the frequency spectrum between build 1 and build 2 (table 3). The same criteria (same width), however, could be chosen between build 1 and build 2. It is common (especially in a powder metal build process) to see small shifts in resonant spectra from build to build because of variations

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 6 Typical half-size SS star artifact RAM spectrum showing the subset of the six resonant peaks used for comparison criteria.

TABLE 3 RAM criteria, center frequencies, and shift between build 1 and build 2 for the half-size SS star artifacts Center Frequency

Frequency Shift (Hz)

Build 1 (Hz)

Build 2 (Hz)

Criterion 1

27,742.5

27,549.5

Criterion 2

28,072.5

28,017.5

55

Criterion 3

31,489.5

31,455

34.5

Criterion 4

34,027.5

34,001

26.5

Criterion 5

34,489

34,601

112

Criterion 6

39,401

39,366.5

34.5

193

in the raw material, manufacturing conditions, and environmental conditions. Note that the general shape of the frequency spectrum does not change from build to build, just a constant shift in frequency is observed, which is based on changes to the mass or stiffness of the parts. Also note that the spectrum shift is a global shift for the entire build and does not affect the ability to sort the good parts from the defective parts. The outcome of the resonant acoustic test for the eighty-eight SS half-size star artifacts is displayed in table 4.

TABLE 4 RAM overall pass/fail results for the 88 half-size SS star artifacts Build 1

Build 2

Pass

Fail

Pass

Good parts

20

0

19

1

Bad parts with

0

24

0

24

defects

Fail

61

62

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 7 Half-size SS star artifact RAM spectra of a reference part in grey and of a part with defects in black.

We defined criteria that allowed us to differentiate the good parts from parts with internal defects. The reference parts showed a significant difference in resonance compared with the parts with defects (fig. 7). One reference part out of 20 was rejected, however, as its resonance peaks were too different from the other reference parts. This is probably due to an unintentional organic defect in that part resulting from small variations in the manufacturing process. In the range of frequencies analyzed, the parts could not be sorted based on the number of defects they contain. The main benefits of RAM are as follows: • easy to use • fast • simple • no restriction in size • no restriction in shape • no roughness restriction • no part preparation or fixturing required • suitable for medium to high volumes • able to identify defective parts • suitable for production end of line testing or routine quality inspection The main drawback of the method is that it is not possible to identify the type or location of defects in the part. PCRT Principle of PCRT—PCRT7 uses piezoelectric transducers in contact with the part, but otherwise is similar to RAM. Excitation of the part uses a piezoelectric transducer to generate a low-energy swept-sine wave, generally with ultrasonic frequencies (5 kHz to 500 kHz). Then the frequency response is recorded with other piezoelectric transducers. Each test requires configuring a specific fixture and optimizing data collection settings for each new part geometry.

OBATON ET AL., DOI: 10.1520/STP162020180099

The PCRT system used to characterize the star artifacts is fully automated with one transmitter and two receiver transducers. For data analysis, the system uses a Z-score (ASTM E3081, Standard Practice for Outlier Screening Using Process Compensated Resonance Testing via Swept Sine Input for Metallic and Non-Metallic Parts)9 to identify outliers of resonance peak variations within a population of parts. In addition, when a simple resonance analysis is insufficient due to the interlacing of the spectra for good and bad parts, the system uses a pattern recognition tool, such as the Mahalanobis-Taguchi system (MTS)10 to sort the parts. PCRT Inspection of the SS Half-Size Star Artifacts—Figure 8 shows the setup used for the PCRT inspection of 74 of the 88 SS half-size stars resulting from the two successive builds (table 2). We considered 20 and 18 parts as reference parts from builds 1 and 2, respectively. For each part, data were collected between 25 kHz and 225 kHz in less than 1 min and then processed using Z-score and MTS analysis. The tests scanned six resonant peaks (60 Hz, 70 Hz, 110 Hz, 116 Hz, 132 Hz, 180 Hz) and the calculations are based on those six frequencies. The data collection settings and scoring criteria were identical for both builds in the sample population. The repeatability of resonance measurements was evaluated using one reference part, with more than 30 repetitions. The variations, expressed as the standard deviations, in each measured peak frequency with more than 30 repetitions were calculated. The frequency peaks with large variations were eliminated from the set of peaks to be

FIG. 8 PCRT set-up used for inspection of the half-size SS star artifacts.

63

64

STP 1620 On Structural Integrity of Additive Manufactured Parts

used for the defect detection criteria. The standard deviations for the remaining 26 resonant peaks were between 0.05% and 0.013% of the mean peak frequencies. PCRT partial spectra for four good and four bad parts are compared in figure 9, which shows that the shift in frequency between good and bad parts is not the same for every resonant peak. To deal with this, a proprietary pattern recognition algorithm (Vibrational Pattern Recognition11) is used to differentiate the parts with defects from the parts with no defects. To confirm repeatability, these tests were run three times. Each test run gave the same results. The overall results are presented in table 5. As with RAM, we were able to sort the good parts from the ones with internal defects. The parts, however, could not be sorted by the number of defects they contain. The main benefits of PCRT are as follows: • fast • no restriction in size • no restriction in shape • no roughness restriction

FIG. 9 Typical half-size SS star artifact PCRT spectrum. The four plots at the top represent four good parts and the four at the bottom represent four bad parts.

TABLE 5 PCRT overall pass/fail results for the 74 half-size SS star artifacts Build 1 Pass

Build 2 Fail

Pass

Fail

Good parts

18

0

20

0

Bad parts with

0

24

0

12

defects

OBATON ET AL., DOI: 10.1520/STP162020180099

suitable for medium to high volume able to identify defective parts • suitable for production end of line testing, or routine quality inspection The main drawbacks of PCRT are as follows: • part set-up required • not possible to identify the type or location of defects in the part • •

Selective Inspection Methods

Selective inspection methods use scanning techniques to inspect only selected regions of a part, possibly from multiple directions. CUT Principle of CUT—CUT, as described in ASTM E1001, Standard Practice for Detection and Evaluation of Discontinuities by the Immersed Pulse-Echo Ultrasonic Method Using Longitudinal Waves,12 and explained by Olympus,13,14 generally uses one or more piezoelectric transducers, which can act as a transmitter, receiver, or both, to produce ultrasound into the test part. The emitting frequencies are theoretically ranging from 20 kHz upward to 1 GHz, depending on the density of the material and on the size of the smallest defect to be detected. The ultrasound wavelength (k ¼ v/f, where v is the velocity of sound in the material of the tested part and f is the sound frequency) must be roughly the size of the smallest defect. Decreasing the wavelength allows the detection of smaller defects but reduces the penetration depth because attenuation increases significantly. When the CUT is performed in contact mode, the thin film of a couplant must be spread on the surface of the transducer to increase the transmission of the ultrasonic waves into the part. This eliminates variations due to microscopic air pockets, which greatly impede ultrasound. The acoustic impedance of the couplant should be similar to that of the test part. CUT, however, also can be performed in the noncontact mode. In that case, the transducer and part are immersed in a tank of liquid (generally water), which also serves as the couplant. This allows the ultrasound beam to be scanned over various regions of a part with minimal variations in the coupling. When sound waves transmitted into the part, reflection, transmission, or refraction occurs at each boundary separating a defect from the surrounding material, and at the boundaries defining the part itself. The resultant echoes are detected by a transducer and processed to determine the presence of the defects and their approximate locations. The detailed scanning procedure and signal processing must compensate for confounding phenomena, such as mode conversion, scattering, and shadowing. These phenomena can prevent observation of some defects. There are three possible ways of presenting the information (fig. 10): 1. A-scan (x-axis: time, y-axis: reflected wave amplitude): shows peaks on a graph (oscilloscope). The graph represents the amplitude of the reflected sound wave as a function of time. Position on the x-axis reflects the depth of

65

66

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 10 Schematic representation of A, B, and C scans used in ultrasound inspection.

each reflecting feature or flaw. Position on the y-axis reflects the amplitude of the echo from each reflecting feature or flaw. 2. B-scan (x-axis: linear position of the transducer, y-axis: depth of the reflector, grey levels or colors: reflected wave amplitude): shows an image with different grey levels or colors corresponding to the amplitude of the reflection from each reflecting feature or flaw. Position on the x-axis shows the position of the transducer along its scan path. Position on the y-axis reflects the depth of each reflecting feature or flaw. 3. C-scan (x-axis: linear position of the transducer on the x-axis, y-axis: linear position of the transducer on the y-axis, grey levels or colors: reflected wave amplitude, time of flight or depth): shows an image viewed from the top (planar view) of a region of interest within the volume of the test specimen. The images display different grey levels or colors corresponding to the amplitude, time of flight, or depth of the signal for different positions of the transducer scanning the surface of the part. A C-scan image is formed in a plane normal to a B-scan image. The CUT system used to inspect the star artifacts is associated with a software system (ODIS from OKOS), including amplitude, time domain, and frequency domain, and imaging in a real-time or post-collection review. The software enables simultaneous A-B-C-scan collection. The color code on the images indicates the amplitude of the signal (i.e., change in attenuation) such as displayed in figure 11, where black (in the center) corresponds to low amplitude and white (on the borders) to high amplitude. CUT Inspection of the Aluminum S1 Star Artifact—The computer-aided design (CAD) of the S1 aluminum star artifact is shown in figure 12. This part was tested in

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 11 Color code or grey code on the C-scan images (black in the center: low amplitudes, white on the borders: large amplitudes).

the configuration shown in figure 13. One transducer, acting as a transmitter or receiver, was used with the artifact immersed in water. The part was first scanned from the top, and then from each side of its branches (fig. 14) using the longitudinal waves. The schematics of the scanning strategies are shown in figure 14. The characteristics of the transducer used for both scans are displayed in table 6. The respective water path distances between the transducer and the test artifact during the scan were 132 mm for the top surface scan, and 83 mm for the side surface scans. The gain was adjusted to 39 dB to obtain the 100% threshold reflection of the back wall of the star artifact. Then, to outline the important features inside the artifact, two time-domain windows (data 1 and data 2) were defined on the A-scan, such as the one shown in figure 15. These windows

FIG. 12 CAD of the aluminum star artifact, design S1, seen from different sides (R refers to the Region in the star).

67

68

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 13 CUT side inspection of the aluminum star artifact.

correspond to two different regions (depths) of interest in the artifact. Their widths (with corresponding ranges of depths) were determined based on the speed of sound in the material. The speed of sound given in the literature for wrought aluminum was chosen (as 6375 m/s). This value, however, might be different in layer-by-layer manufactured aluminum. Consequently, the uncertainty on the corresponding depths of interest is not negligible. Data 1, the longer rectangle on figure 16, excludes the top and rear surfaces of the part, whereas data 2, the shorter rectangle on figure 16, selects the region around the rear surfaces. The C-scan images shown in figure 16 through figure 20 are based on these two timedomain windows focusing on two different regions of depth. Because all of the defects in the artifact are in the same position along its height, the first defect does not allow the rest of the defects to be inspected by CUT from the top surface of the star. In contrast, the C-scan images recorded from the side

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 14 Scanning strategies of the aluminum star artifact using the CUT method.

TABLE 6 Transducer characteristics and experimental parameters used for the CUT inspection of the aluminum artifact Scan

Type

Frequency (MHz)

c

Focal legnth (mm)

Diameter (mm)

f

Ø

Wavelength (mm)

Spot size (mm)

k¼ s/c

1.22  f/ Øk

Resolution (mm)

0.707  1.22 f/Øk

from top

Ceramic

25

6

9.5

0.3

0.2

0.1

from side

Ceramic

10

75

9.5

0.6

6.1

4.3

Sound velocity in aluminum

s

6.4

(mm/:s)

surfaces of the star artifact and presented in figure 16 through figure 20 reveal more information: • The cylinders with different orientations (d12, between Regions 1 and 2) cannot be observed in near field from sides labeled 1A and 2B on figure 16, probably because the gain was constant with distance. The larger ones can be seen in the far field from the side labeled 3B (fig. 17), but not from the side labeled 5A (fig. 20). • The vertical cylinders with different diameters (d23, between Regions 2 and 3) cannot be seen from the side labeled 1A (fig. 16) because the cylinders with different orientations (d12) hide them. The larger cylinders can be seen from all sides, in near and far fields.

69

70

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 15 Time-domain windows (rectangles) defined on the A-scan plot of the reflected signal for data analysis of CUT.

•

•

•

The spheres of different diameters (d34, between Regions 3 and 4) cannot be seen from any sides because the sphere shape is too complex to detect with ultrasound waves. Indeed, this geometry scatters the waves in every direction. Furthermore, these spheres are full of powder. Thus, the reflected wave reaching the transducer is very low in energy. In this experiment, the energy of the reflected wave might be too low to be detected. Although the spheres probably could have been detected with shear wave methodology, only longitudinal wave imaging was used during the present evaluation. The horizontal cylinders of different diameters located on an outside edge of the star (d45out, between Regions 4 and 5) are not seen from most sides. This is probably due to the fact that they are positioned right at the borders of the time windows. So, because the widths of these windows are not accurate, they might be outside. They can all be seen and discretized, however, in the far field from the side labeled 3A (fig. 18). They appear right in the middle of the image rather than on an edge probably because of diffraction at the R4 edge of the star. The horizontal cylinders of different diameters located on an inside edge of the star (d45in, between Regions 4 and 5) are not seen from most sides

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 16 CUT inspection from sides 1A and 2B of the aluminum star artifact. The parallel lines represent the longitudinal waves propagating from sides 1A/2B, “data” refers to the time domain window, and “APA” stands for “Absolute Peak Amplitude.”

FIG. 17 CUT inspection from sides 2A and 3B of the aluminum star artifact. The parallel lines represent the longitudinal waves propagating from sides 2A/3B, “data” refers to the time domain window, and “APA” stands for “Absolute Peak Amplitude.”

71

72

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 18 CUT inspection from sides 3A and 4B of the aluminum star artifact. The parallel lines represent the longitudinal waves propagating from sides 3A/4B, “data” refers to the time domain window, and “APA” stands for “Absolute Peak Amplitude.”

probably because they are very small (2 mm long) and located on an edge. The larger ones, however, can be seen from the side labeled A4 (fig. 19) in the near field. • There was a mistake in the CAD data used to manufacture the aluminum S1 star artifacts. One of the cylinders with a different orientation is isolated from the others (d5 in Region 5, fig. 20). We might see it from the side labeled 1B (fig. 20) in the near field. The reason it cannot be seen from the other sides is probably due to its orientation. Except for the spheres, all other types of larger defects were detected from at least one side, although they were not discretized in size. The interpretation of the image, however, is not straightforward and requires experience. In addition, the defects and their localization were known in advance. In the case of blind inspection, the analysis would have been even more difficult. Furthermore, the spatial resolution of the images, in that case, is not high enough to perform accurate dimensional measurements of the defects even though the standard object would have been used for comparison. Despite the surface roughness of the AM parts, we were able to scan the star artifacts from as-built surfaces, even from side surfaces, which generally are rougher than the top surfaces in AM. So, the roughness of the surface did not preclude CUT. Moreover, note that the size and thickness of the aluminum star do not allow for inspection with laboratory 225-kV XCT. Laboratory

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 19 CUT inspection from sides 4A and 5B of the aluminum star artifact. The parallel lines represent the longitudinal waves propagating from sides 4A/5B, “data” refers to the time domain window, and “APA” stands for “Absolute Peak Amplitude.”

FIG. 20 CUT inspection from sides 5A and 1B of the aluminum star artifact. The parallel lines represent the longitudinal waves propagating from sides 5A/1B, “data” refers to the time domain window, and “APA” stands for “Absolute Peak Amplitude.”

73

74

STP 1620 On Structural Integrity of Additive Manufactured Parts

450-kV XCT is required. This underlines an advantage of UT methods for large parts—that is, 450-kV XCT systems are relatively rare. The main benefits of the CUT method are as follows: • few roughness restrictions • real-time imaging • less restriction in size than XCT • ability to identify the type and location of defects in the part The main drawbacks of the CUT method are as follows: • part set-up required • experience required • shape restriction • not suitable for high volume and routine quality control • no real discretization in size PAUT Principle of PAUT—The principle of PAUT is similar to that of CUT.15 Instead of consisting of a single piezoelectric transducer, however, the PAUT probe is composed of an array of piezoelectric transducers. Each transducer of this array can be pulsed independently with respect to the others to generate several waves that will interfere. The waves that are in phase will be added together (constructive interference), whereas the ones that are out of phase will cancel each other (destructive interference). This results in a unique wavefront constituting a synthesized focused beam that can be oriented (controllable angle), shaped (controllable focal distance and focal spot size), and swept electronically. Dynamic control of the beam enables examining the complex shapes across a range of different perspectives without moving the probe. The spatial resolution of the images is linked to the number of elements in the array. The larger the number of elements, the better the resolution. It is possible to improve the performance of the above-mentioned standard PAUT system by implementing a data acquisition method called full matrix capture (FMC) with a post-processing reconstruction algorithm known as total focusing method (TFM).16–18 In the generic FMC acquisition method, a single transducer of the array is pulsed and the time domain signals (A-scans) are captured by all transducers in the array and stored for later processing. This is done for all transducers of the array. The number of data sets acquired is the square of the number of transducers. Thus, a 128-transducer probe will produce 16,384 raw A-scans. The offline data processing uses the complete set of time-domain data from all combinations of transmitting and receiving transducers. The reconstructed image from the A-scans is equivalent to making coherent summations over all transducers to focus at each point in the target region. Thus, FMC combined with TFM enables offline reconstruction of a more detailed image (with better spatial resolution, perspective, and defect definition) than a standard PAUT imaging method in which all array elements create a unique wavefront to form a beam with a fixed focus.

OBATON ET AL., DOI: 10.1520/STP162020180099

The instrument19 that was used to perform the PAUT of the star artifacts implements both FMC and TFM. It also permits the acquisition of standard PAUT data. The software associated with the hardware enables simultaneous A-B-C-scan displays as well as a D-scan, which is similar to the C-scan but in a perpendicular direction. If the C-scan displays an image in the plane (XY), then the D-scan will display the image in the orthogonal plane (ZY). During the translation of the PAUT probe, the beam inside the probe is swept in the perpendicular direction to scan the entire surface. The color code or grey variations on the images is the same as the one described earlier for the CUT images (fig. 11). PAUT–TFM of the Aluminum S1 and S2 Star Artifacts—Both of the S1 and S2 designs of the aluminum star artifacts, with the corresponding CAD shown in figures 12 and 21, could be tested using PAUT-TFM. Only one linear probe acting as a transmitter or receiver was used to induce the longitudinal waves. The probe is 32 mm long and composed of 128 transducers separated from each other by 0.25 mm (center to center). The star artifact, immersed in a tank of water for the coupling (fig. 22), was scanned only from one side of its branches (fig. 22), from the top to

FIG. 21 CAD of the aluminum star artifact, design S2, seen from different sides (R refers to the Region in the star).

75

76

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 22 PAUT-TFM side inspection of the aluminum star.

the bottom of the star (y-axis; i.e., sideways as shown in fig. 23). The transducers were operated at a center frequency of 10 MHz (bandwidth of 6 MHz), which corresponds to a wavelength of 0.62 mm in aluminum. The water path distance between the probe and the scanned surface of the artifact during the scan was 12.5 mm. Finally, the analog gain was set manually to 40 dB to properly visualize all of the defects and to acquire the data. The digital gain can be subsequently adjusted on the reconstructed image (300 pixels  400 pixels) offline. The A-, B-, C-, and D-scan presented in figure 23 through figure 27 for the S1 design and in figure 28 through figure 32 for the S2 design show interesting results: • Three of the cylinders with different orientations can be seen from side 1A (fig. 23). The orientations of the other cylinders prevent detection. • Five vertical cylinders with different diameters can be detected from side 2A (fig. 24) using the three optimized digital gains chosen to display the images. The cylinders with smaller diameters, down to 100 lm, can be detected with increased gain (fig. 24 bottom images) at the expense of the reduced quality of the image (lower signal to noise ratio). • Three of the spheres of different diameters can be seen from side 3A (fig. 25).

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 23 PAUT-TFM inspection from side 1A of the S1 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 1A.

FIG. 24 PAUT-TFM inspection from side 2A of the S1 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 2A. The two bottom scans correspond to higher digital gains.

77

78

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 25 PAUT-TFM inspection from side 3A of the S1 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 3A.

Three of the horizontal cylinders of different diameters (10 mm long) located on an outside edge of the star can be seen from side 4A (fig. 26). Their position on the edge hides the others. • Some horizontal cylinders of different diameters located on an inside edge of the star are seen from side 4A (fig. 27). They are located on an edge and are short (2 mm long), so they are difficult to differentiate from the reflection of the edge. • It does not seem that we can see the single cylinder manufactured by mistake. The defects are more visible on PAUT-TFM images from the S2 design than from the S1 design. • Six cylinders with different orientations (instead of three in the S1 design) can be seen in Region 5 from side 4A (fig. 32). This means that we can visualize defects down to 200 lm in diameter in various orientations. •

FIG. 26 PAUT-TFM inspection from side 4A of the S1 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 4A.

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 27 PAUT-TFM inspection from side 4A of the S1 aluminum star artifact with different scan parameters and scan area than in Fig. 26. The parallel lines represent the longitudinal waves propagating from side 4A.

All vertical cylinders can be seen in Region 4 from the side labeled 3A (fig. 31) like in the S1 design, although with increased gain. This means that we can visualize defects down to 100 lm in diameter. • All spheres (instead of three in the S1 design) can be seen in Region 4 from the side labeled 3A (fig. 30). This means that we can visualize defects down to 100 lm in diameter. • The horizontal cylinders are easier to see when located inside the star artifact (fig. 28) than when located outside the star artifact (fig. 29). Four of these cylinders are seen in figure 29. Thus, because of its high spatial resolution when used with FMC/TFM, the PAUT method reveals individual defects and even allows rough dimensional •

FIG. 28 PAUT-TFM inspection from side 5A of the S2 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 5A.

79

80

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 29 PAUT-TFM inspection from side 1A of the S2 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 1A.

FIG. 30 PAUT-TFM inspection from side 2A of the S2 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 2A.

measurements by comparison to reference objects. The interpretation of the images is easier than in CUT but is still not straightforward. Like in CUT, however, the surface roughness did not prevent the part from being tested, and the size of the part was not an obstacle. Apart from the better resolution than in CUT, another real advantage of the PAUT-TFM methods is to be able to perform the static inspection by steering the beam instead of scanning the probe. This is of particular interest for complex shapes that cannot be scanned. The main benefits of the PAUT–TFM are as follows: • possibility of inspection without scanning by steering the beam • less restriction in shape than CUT • less restriction in size than XCT • fast scanning (real-time), faster than CUT

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 31 PAUT-TFM inspection from side 3A of the S2 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 3A. The two bottom scans correspond to higher digital gains.

FIG. 32 PAUT-TFM inspection from side 4A of the S2 aluminum star artifact. The parallel lines represent the longitudinal waves propagating from side 4A.

• • •

few roughness restrictions ability to identify the type and location of defects in the part discretization in size

81

82

STP 1620 On Structural Integrity of Additive Manufactured Parts

rough dimensional measurements by comparison to reference objects The main drawbacks of the PAUT–TFM are as follows: • part set-up required • requires experience • requires more experience and training than CUT • more expensive than CUT •

SRXCT Principle of XCT—Tomography is a volumetric imaging method. It consists of probing a test part with electromagnetic waves (e.g., microwaves, terahertz waves, X-ray) or elastic waves (e.g., sound, ultrasound, etc.) at different angular positions of the part, recording the interactions of these waves with the test part for each angle independently, and post-processing these data with a reconstruction algorithm to give a volumetric image. A tomographic device includes a wave source, a sensor that detects the effects of interaction between the waves and the test part (e.g., attenuation, reflection, diffusion, diffraction), and a translation stage with a rotating platform on which the test part is positioned. The most commonly used waves are X-rays, as the short wavelength allows for better spatial resolution of the image. The best-known method of X-ray tomography is XCT.20–22 In a laboratory XCT system, the X-ray source consists of a high-voltage power supply (usually the voltage (U) ranges between 10 kV and 450 kV) and an X-ray tube made of two electrodes: a filamentary cathode, usually of tungsten, and an anode, which is the target. By applying an electric current (I) to the cathode, electrons are extracted; then, by applying a high potential difference (U) between the anode and the cathode, the electrons are accelerated to collide with the target, focused on a spot on the target, to emit X-rays. ASTM E1695-95, Standard Test Method for Measurement of Computed Tomography (CT) System Performance,23 describes two factors that affect the quality of an XCT image: geometrical unsharpness, which limits the spatial resolution, and random noise, which limits the contrast sensitivity. Spatial resolution influences the ability to detect small geometrical features in the object, whereas contrast sensitivity influences the ability to detect inhomogeneities in the object. The geometrical unsharpness is linked to the size of the focused spot. This size depends on the required power for the measurement (P ¼ UI), which depends on the thickness and density of the test part. The thicker and higher density of the part, the higher the power required, which creates a larger focus spot. An increase in the size of the focus spot induces more geometrical unsharpness. Thus, a larger focus spot results in a decrease in spatial resolution. The random noise is linked to the brilliance of the source, which is proportional to the number of photons produced per second. The higher the brilliance, the lower the random noise. Thus, a higher brilliance results in higher contrast sensitivity.

OBATON ET AL., DOI: 10.1520/STP162020180099

The produced X-ray beam is polychromatic (multiwavelength), which is not ideal because artifacts (an artificial characteristic on the image) can appear on the image as a result of beam hardening (i.e., loss of the low energies in the source spectrum due to their absorption by the test part). The beam’s shape is commonly conical. This shape causes the magnification (mag) of the part on the detector to depend on its position between the source and the detector. The closer the part to the source, the larger its magnification on the detector, which usually is a flat panel composed of a matrix of pixels. This magnification and the size of a pixel of the detector (pixelsize) will define the spatial resolution of the image. Indeed, the spatial resolution is related to the size of the voxel (equivalent in a three-dimensional (3D) image of a pixel) defined as follows: voxelsize ¼ pixelsize =mag:

(1)

The detector records the attenuated intensity of a ray, which is proportional to the thickness and density of material crossed. This enables the ability to acquire two-dimensional (2D) images (i.e., gray-scale images that show the radiation attenuation through the material) called projections. These projections, acquired over 360 around the test part, enable the reconstruction of the image in 3D. Even though laboratory XCT is the more widespread X-ray tomographic system, tomography is also performed with a SRXCT source. Principle of SRXCT—Synchrotron radiation is generated by electrons confined in a large loop and accelerated in a magnetic field. Because of the bending of their trajectory, the electrons accelerate or decelerate losing energy and thus produce electromagnetic radiation with a wide polychromatic spectrum from far infrared to hard X-rays. The use of a monochromator allows for the selection of a monochromatic X-ray beam used to perform CT. The detector is generally a scintillator combined with a camera. Otherwise, the SRXCT principle of image acquisition and reconstruction is the same as for laboratory XCT. Compared with laboratory XCT, in addition to being monochromatic instead of polychromatic, the synchrotron beam is collimated instead of conical, and its brilliance is much higher than laboratory XCT. These are the three main differences between sources from laboratory XCT and synchrotron radiation. These differences are of paramount importance as they enable better spatial resolution, better contrast sensitivity, and fewer artifacts. Indeed, the benefit of a monochromatic beam is that it prevents artifacts on the image. The first benefit of a parallel beam is that the part does not need to be scanned over 360 but simply over 180 . The information acquired is identical from one side to the other. The second benefit is that there is no focused spot that would create geometric unsharpness and thus decrease spatial resolution. The last benefit is that there is no magnification, which means that the spatial resolution is defined only by the pixel size of the detector. The benefit of high brilliance is that it decreases random noise and thus increases contrast sensitivity.

83

84

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 33 CAD of the Co-Cr full-star artifact, S2 design, seen from different sides (R refer to the Region in the star).

TABLE 7 Experimental parameters used for the XCT inspection using synchrotron of the Co-Cr star

Acquisition Parameters Synchrotron radiation source Detector

Acquisition mode (pixel)

Energy (keV)

195

camera

PCO Edge 4.2

scintillator

LuAg 2000

image size (pixel)

2048  400

pixel size (:m)

23.3

Half-acquisition 500

Exposure time (ms)

15

Image per projection

20

Number of projections Number of scans at different vertical

5,000 10

locations Distance between sample and detector (m)

13

SRXCT of the Co-Cr S2 Star Artifact—An XCT beamline (ID19),24 from the European Synchrotron Radiation Facility (ESRF) located in Grenoble, France, was used to characterize the S2 design of the Co-Cr star artifact, with the corresponding CAD shown in figure 33. The acquisition parameters are given in table 7 and the different images, processed with the ImageJ software, are displayed in figure 34 through figure 38. The spatial resolution of the images is indisputably better than the ones from the UT methods. All defects are clearly located and imaged. Inspection of the 3D images also indicated an organic defect (hot tear or crack) at the base of the star (fig. 39). The resolution of the images allows for a dimensional comparison with the CAD model (fig. 40).

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 34 Star artifact images obtained by SRXCT for Region 1.

FIG. 35 Star artifact images obtained by SRXCT for Region 2.

85

86

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 36 Star artifact images obtained by SRXCT for Region 3.

FIG. 37 Star artifact images obtained by SRXCT for Region 4.

OBATON ET AL., DOI: 10.1520/STP162020180099

FIG. 38 Star artifact images obtained by SRXCT for Region 5.

The main benefits of XCT are as follows: • no restriction in shape • no roughness restrictions • ability to identify the type and location of defects in the part • discretization in size • dimensional measurements by comparison to reference objects The main drawbacks of XCT are as follows: • part set-up required • experience required • not suitable for high volume and routine quality control • more expensive than other NDT methods

Conclusion The paper provides a comparison of NDT methods for typically expected defect sizes and shapes resulting from PBF processes. These methods can be used as the baseline for further analysis, including the application-dependent critical size information. We have investigated four volumetric NDT methods as potential alternatives to XCT for faster and cheaper characterization for routine control and inspection of additively manufactured metal parts: two whole-body inspection methods using acoustic waves (RAM and PCRT) and two selective inspection methods using

87

88

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 39 Organic defect (hot tear or crack) in Region 2.

FIG. 40 Superposition of the CAD and the X-ray scan in Region 2.

ultrasound waves (CUT and PAUT-TFM). The two whole-body inspection methods gave similar results. The parts with defects could be separated from the parts without defects, but the parts with defects could not be sorted based on the number of defects they contained. The results of the two selective inspection methods were significantly

OBATON ET AL., DOI: 10.1520/STP162020180099

different. The CUT method allowed for detection of most of the defects with lowspatial resolution, although defects 5 mm apart could not be discriminated. Furthermore, the defects were not detected from every direction; successful results depended on the side inspected. The PAUT-TFM resulted in higher resolution images than CUT, allowing for detection and separation of nearly all defects for any given side inspection. As a comparison, we performed SRXCT, which gave better spatial resolution, better contrast sensitivity, and fewer artifacts than laboratory XCT. The whole-body inspection methods allowed for only the identification of defective parts, whereas the selective inspection methods allowed for identification of the type of the defect. In applications like routine inspections, it is sufficient to determine that a part is nonconforming to reject it. Then, if localization and size of defects are needed to evaluate the defect criticality and impact on part performance, a secondary evaluation could use a selective inspection method. Further volumetric NDT methods, such as nonlinear acoustic methods, should be investigated to boost the application of AM in critical sectors and to guarantee the reliability of parts for critical applications. ACKNOWLEDGMENTS

We would like to particularly thank Wilson Vesga, Ben Dutton (convener of the JG59), and David Ross-Pinnock from the Manufacturing Technology Centre (MTC), United Kingdom, for the valuable information provided to help us. We are also grateful to Zodiac-Aerospace Corp., France, for building the star artifacts in aluminum; to Sofiane Terzi and Barbara Fayard from Novitom Corp., France, for the SRXCT inspection; and to Julieanne Heffernan from Vibrant Corp., United States, for the PCRT inspection. DISCLAIMER

Certain commercial entities, equipment, or materials may be identified in this document to describe an experimental procedure or concept adequately. Such identification is not intended to imply recommendation or endorsement by the National Institute of Standards and Technology, nor is it intended to imply that the entities, materials, or equipment are necessarily the best available for the purpose.

References 1.

2.

A. F. Obaton, M.-Q. Le ˆ, V. Prezza, D. Marlot, P. Delvart, P. A. Huskic, S. Senck, E. Mahe ´, and C. Cayron, “Investigation of New Volumetric Non-Destructive Techniques to Characterise Additive Manufacturing Parts,” Welding in the World 62, no. 5 (2018): 1049–1057, https://doi.org/10.1007/s40194-018-0593-7 J. B. Perraud, A. F. Obaton, J. Bou-Sleiman, B. Recur, H. Balacey, F. Darrack, J. P. Guillet, and P. Mounaix, “Terahertz Imaging and Tomography as Efficient Instruments for Testing Polymer Additive Manufacturing Objects,” Applied Optics 55, no. 13, (2016): 3462–3467, https://doi.org/10.1364/AO.55.003462

89

90

STP 1620 On Structural Integrity of Additive Manufactured Parts

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13. 14. 15. 16.

E. Todorov, R. Spencer, S. Gleeson, M. Jamshidinia, and S. M. Kelly, America Makes: National Additive Manufacturing Innovation Institute (NAMII) Project 1: Nondestructive Evaluation (NDE) of Complex Metallic Additive Manufactured (AM) Structures, AFRL-RXWP-TR-2014-0162, EWI (interim report, Air Force Research Laboratory, Materials and Manufacturing Directorate, Wright-Patterson Air Force Base, Dayton OH, June 2014). B. M. Sharratt, Non-Destructive Techniques and Technologies for Qualification of Additive Manufactured Parts and Processes: A Literature Review,” DRDC-RDDC-2015-C035 (Defence Research Report, Government of Canada, March 1, 2015) http://web.archive.org/web/20200108154044/https://cradpdf.drdc-rddc.gc.ca/PDFS/unc200/p801800_ A1b.pdf Standard Guide for Resonant Ultrasound Spectroscopy for Defect Detection in Both Metallic and Non-Metallic Parts, ASTM E2001-18 ASTM International, (West Conshohocken, PA: ASTM International, approved December 6, 2018), https://doi.org/10.1520/ E2001-18 M. Schiefer and L. Sjoeberg, “Physical Basis of the Resonant Acoustic Method for Flaw Detection,” http://web.archive.org/web/20200108153712/http://www.modalshop.com/ filelibrary/Physical%20Basis%20of%20the%20Resonant%20Acoustic%20Method%20for% 20Flaw.pdf Standard Practice for Process Compensated Resonance Testing via Swept Sine Input for Metallic and Non-Metallic Parts, ASTM E2534-15 (West Conshohocken, PA: ASTM International, approved 2015), https://doi.org/10.1520/E2534-15 G. Stultz, R. Bono, and M. Schiefer, “Fundamentals of Resonant Acoustic Method NDT,” http://www.modalshop.com/techlibrary/Fundamentals%20of%20Resonant%20 Acoustic%20Method%20NDT.pdf Standard Practice for Outlier Screening Using Process Compensated Resonance Testing via Swept Sine Input for Metallic and Non-Metallic Parts, ASTM E3081-16 (West Conshohocken, PA: ASTM International, approved December 1, 2016), https://doi.org/10.1520/ E3081-16 W. H. Woodall, R. Koudelik, K.-L. Tsui, S. Bum Kim, Z. G. Stoumbos, and C. P. A. Carvounis, “A Review and Analysis of the Mahalanobis—Taguchi System,” Technometrics 45, no. 1 (2003): 1–15, https://doi.org/10.1198/004017002188618626 E. Biedermann, J. Heffernan, A. Mayes, G. Gatewood, L. Jauriqui, B. Goodlet, T. Pollock, C. Torbet, J. C. Aldrin, and S. Mazdiyasni, “Process Compensated Resonance Testing Modeling for Damage Evolution and Uncertainty Quantification,” in AIP Conference Proceedings 1806, 090005 (New York: American Institute of Physics, 2017), 090005-1– 090005-10, https://doi.org/10.1063/1.4974649 Standard Practice for Detection and Evaluation of Discontinuities by the Immersed Pulse-Echo Ultrasonic Method Using Longitudinal Waves, ASTM E1001-16 (West Conshohocken, PA: ASTM International, approved December 1, 2016), https://doi.org/10.1520/ E1001-16 Olympus, “Ultrasonic Transducers Technical Notes,” 2006, http://web.archive.org/save/ https://mbi-ctac.sites.medinfo.ufl.edu/files/2017/02/ultrasound-basics.pdf Olympus, “Ultrasonic Flaw Detection Tutorial,” http://web.archive.org/web/20200 108154230/https://www.olympus-ims.com/en/ndt-tutorials/flaw-detection/ Olympus, “Phased Array Tutorial,” http://web.archive.org/web/20200108155804/ https://www.olympus-ims.com/en/ndt-tutorials/phased-array/ M. Karaman, P.-C. Li, and M. O’Donnell, “Synthetic Aperture Imaging for Small Scale Systems,” IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, 42, no. 3 (1995): 429–442, http://doi.org/10.1109/58.384453

OBATON ET AL., DOI: 10.1520/STP162020180099

17.

18.

19.

20.

21.

22.

23.

24.

C. Holmes, B. W. Drinkwater, and P. D. Wilcox, “Post-Processing of the Full Matrix of Ultrasonic Transmit-Receive Array Data for Non-Destructive Evaluation,” NDT&E International 38, no. 8 (2005): 701–711, http://doi.org/10.1016/j.ndteint.2005.04.002 R. Spencer, R. Sunderman, and E. Todorov, “FMC/TFM Experimental Comparisons,” in 44th Annual Review of Progress in Quantitative Nondestructive Evaluation, Vol. 37 (New York: American Physics Institute, 2018), 020015-1–020015-5, https://doi.org/10.1063/ 1.5031512 E. Carcreff, G. Dao, and D. Braconnier, “Total Focusing Method for Flaw Characterization in Homogeneous Media” (paper presentation, Fourth International Symposium on Nondestructive Characterization of Materials, Marina Del Rey, FL, June 22–26, 2015), https:// www.researchgate.net/profile/Ewen_Carcreff/publication/282654039_Total_focusing_ method_imaging_for_flaw_characterization_in_homogeneous_media/links/56162ac70 8ae4ce3cc65a588/Total-focusing-method-imaging-for-flaw-characterization-in-homogeneous-media.pdf A.-F. Obaton, J. Fain, M. Djemaı¨, D. Meinel, F. Le ´onard, E. Mahe ´, B. Le ´cuelle, J-.J. Fouchet, and G. Bruno, “in vivo XCT Bone Characterization of Lattice Structured Implants Fabricated by Additive Manufacturing: A Case Report,” Heliyon 3, no. 8 (2017): E00374, http://doi.org/10.1016/j.heliyon.2017.e00374 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, Pt. B (2016): 122–127, http://doi.org/10.1016/ j.csndt.2016.05.003 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, http://doi.org/10.1016/j.addma.2017.06.011 Standard Test Method for Measurement of Computed Tomography (CT) System Performance, ASTM E1695-95(2013) (West Conshohocken, PA: ASTM International, approved June 1, 2013), http://doi.org/10.1520/E1695-95R13 ID19—Microtomography Beamline (Grenoble, France: European Synchrotron Radiation Facility), http://web.archive.org/web/20200108161206/https://www.esrf.eu/ home/UsersAndScience/Experiments/StructMaterials/ID19.html

91

92

STRUCTURAL INTEGRITY OF ADDITIVE MANUFACTURED PARTS

STP 1620, 2020 / available online at www.astm.org / doi: 10.1520/STP162020180106

Jess M. Waller,1 Eric R. Burke,2 Douglas N. Wells,3 Charles T. Nichols,1 Ana D. Branda˜o,4 Johannes Gumpinger,4 Martin Born,4 Tommaso Ghidini,4 Tsuyoshi Nakagawa,5 Akio Koike,5 Masami Mitsui,6 and Tsuyoshi Itoh5

NDE-Based Quality Assurance of Metal Additively Manufactured Aerospace Parts at NASA, JAXA, and ESA Citation J. M. Waller, E. R. Burke, D. N. Wells, C. T. Nichols, A. D. Branda ˜o, J. Gumpinger, M. Born, T. Ghidini, T. Nakagawa, A. Koike, M. Mitsui, and T. Itoh, “NDE-Based Quality Assurance of Metal Additively Manufactured Aerospace Parts at NASA, JAXA, and ESA,” 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), 92–128. http:// doi.org/10.1520/STP1620201801067

ABSTRACT

As metal additive manufactured spaceflight hardware moves closer to use in upcoming missions, the need for appropriate nondestructive evaluation (NDE) Manuscript received December 17, 2018; accepted for publication August 27, 2019. 1 NASA-Johnson Space Center White Sands Test Facility, 12600 NASA Rd., Las Cruces, NM 88012, USA J. M. W. https://orcid.org/0000-0002-7847-8376 2 NASA Langley Research Center, 1 NASA Dr., Langley, VA 23666, USA 3 NASA Marshall Space and Flight Center, Martin Rd. SW, Huntsville, AL 35808, USA D. N. W. http:// orcid.org/0000-0001-5595-2429 4 ESA/ESTEC, European Space Research and Technology Center, Keplerlaan 1, 2201 AZ Noordwijk, The http://orcid.org/0000-0002-3120-1513, J. G. https://orcid.org/0000-0001-6478Netherlands A. D. B. 3065 T. G. http://orcid.org/0000-0003-0431-4800 5 Japanese Aerospace Exploration Agency, 2-1-1 Sengen, Tsukuba, Ibaraki 305-8505, Japan T. N. http:// orcid.org/0000-0002-0419-555X, A. K. https://orcid.org/0000-0001-5983-9319, T. I. https:// orcid.org/0000-0002-5314-0811 6 Japan Quality Assurance Organization, 1 Chome-25 Kanda Sudacho, Chiyoda City, Tokyo 101-8555, Japan https://orcid.org/0000-0002-4620-4290 7 ASTM Symposium on Structural Integrity of Additive Manufactured Parts on November 6-8, 2018 in Washington, DC, USA. C 2020 by ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA 19428-2959. Copyright V

WALLER ET AL., DOI: 10.1520/STP162020180106

procedures to qualify and certify parts becomes more pressing. Traditionally, the level and specificity of certification requirements for aerospace parts is linked to part criticality, which is determined by structural margin, consequence of failure, and part inspectability. Unfortunately, rule-level (performance-based) certification requirements often do not define specific or acceptable NDE procedures or requirements. This level of detail usually falls on the original equipment manufacturer (OEM), which is unwise in the case of new technology, such as additive manufacturing (AM). Instead, it is incumbent upon the end user to communicate risks and with the OEM and to provide oversight to ensure risks are controlled. These risks arise from many sources, including uncertainty about property optimization for rapidly evolving AM processes, lack of engineering experience, and limited operational histories. From the NDE perspective, the main risk arises from uncertainty in the NDE procedure to detect the requisite critical flaw type, size, and distribution. Because the level of criticality of AM aerospace parts is expected to increase, more effort is needed to characterize and understand fatigue and fracture properties of AM materials. Crack initiation resulting from the presence of AM flaws must be considered. This requires knowledge of the critical initial flaw size (CIFS) and the appropriateness of NASA-STD-5009 flaw sizes. Knowledge of the CIFS for a given AM flaw type will allow for the fracture control and NDE communities to evaluate risks and communicate recommendations regarding the acceptability of risk. Toward this goal, this paper discusses NDE-related activities at the National Aeronautics and Space Administration, Japan Space Exploration Agency, and European Space Agency. Current NDE best practices for AM hardware are discussed, along with tailoring NDE according to part criticality. Keywords additive manufacturing, nondestructive evaluation, aerospace, metals, qualification and certification, quality control, fracture critical, parts categories, risk mitigation

Introduction Additive manufacturing (AM) is revolutionizing the traditional aerospace part design and manufacturing paradigm. For existing designs, AM offers the ability to reduce cost, especially for one-of-a-kind or limited-production-run quantities common in the aerospace sector. For new designs, high cost and long lead times associated with production of complex hardware by conventional manufacturing routes have pushed manufacturers to rely on meticulous analysis to mitigate or eliminate the chance of failure. In the case of National Aeronautics and Space Administration (NASA) rocket engines, which can contain a variety of metal AM parts, including turbopumps, injectors, combustion chambers, and nozzles, significant reductions are anticipated in the design, development, testing, and evaluation (DDT&E) time (from 7–10 years to 2–4 years), hardware lead time (from 3–6 years to 6 months), and costs (order-of-magnitude reductions).1 To realize these gains and exploit the full potential of AM, robust quality control and qualification procedures are needed,

93

94

STP 1620 On Structural Integrity of Additive Manufactured Parts

especially for fracture-critical metal AM hardware. This, in turn, entails a comprehensive understanding of the factors responsible for and control of dimensional inaccuracy, defects such as porosity and voids, microstructural variation, excessive surface roughness, and residual stress. Additionally, because location-dependent properties in as-built AM parts are affected by complex interactions between defect-dominated and microstructure-dominated failure mechanisms, the interaction between defects and microstructure needs to be better understood.2,3 Realizing these gains, and improving the current engineering and scientific understanding is predicated on the advancement of the measurement science associated with nondestructive evaluation (NDE) (synonymous with nondestructive testing [NDT] or nondestructive inspection [NDI]). NDE is one of the engineering disciplines used to verify the structural integrity of high-value spaceflight and aerospace parts. More specifically, robust NDE procedures tailored specifically to AM parts are needed, so that the defect state of the AM part can be properly characterized and monitored during service. The wide range of AM processes, process parameters, input materials (feedstock), process equipment, and post-processing can cause significant property variation in finished AM parts. Against this backdrop of property variation, mitigation against part failure relies mainly on NDE to detect defects and on destructive (e.g., proof) tests to assess the effect of defects on part performance. Challenges associated with the application of mechanical (destructive) tests to AM hardware are discussed elsewhere.4 For example, the coordinate system and nomenclature developed for testing of conventional metal parts,5–7 do not adequately cover the full spectrum of parts produced by AM,8–10 although the available literature and standards11–13 do address some of the specific aspects that must be considered when testing AM parts. In addition, properties measured using a standard subscale test coupon might not be the same as those present in the final AM part. The NDE inspector also must overcome additional challenges—namely, AM part complexity, surface roughness, access to the inspection surface or inspection volume, and the lack of clear accept or reject criteria based on allowable defect type, size, and distribution.14–16 As pointed out elsewhere,4 longstanding NDE standard defect classes apply to welds and castings. Although similarities may exist, the defects produced by conventional metallurgical processes generally will not be same as those produced by AM processes. Therefore, despite processing and defect similarities, it is not recommended that existing physical reference standards for welding and casting be used to determine NDE capability when inspecting AM parts.17 This implies that until an accepted AM defect catalog18 and associated NDE detection limits for technologically relevant AM defects are established, the NDE methods and acceptance criteria used for AM parts will remain part-specific point designs. As noted previously,4 the variation in AM processing parameters and disruptions during a build may induce a variety of defects (anomalies) in AM parts that can be detected, sized, and located by NDE. In general, the NDE procedures used to

WALLER ET AL., DOI: 10.1520/STP162020180106

characterize metal AM parts19,20 are based on procedures developed for conventionally manufactured cast, wrought, or welded production parts. Some of the more established NDE procedures that can be used for NDE-based quality assurance (QA) of AM parts are computed tomography (CT), eddy current testing (ET), metrology (MET), penetrant testing (PT), process compensated resonance testing (PCRT), radiographic testing (RT), infrared thermographic testing (IRT), and ultrasonic testing (UT).19 Table 1 in Seifi et al.4 (co-developed by ASTM Committee E07 on Nondestructive Testing19 and ISO TC 261 Joint Group 59) is a much-needed AM defects catalog listing common AM defect types present in as-built metal AM parts and that may be altered by subsequent post-processing and service. The root causes (process origins), hot isostatic pressing (HIP) recoverability, and applicable NDE methods for specific types of AM defects are summarized. Some AM defects are unique to specific metal AM processes, for example, directed energy deposition (DED) or powder bed fusion (PBF), whereas other defects are common across all metal manufacturing techniques, including conventional processes, such as welding. For example, layer defects, cross-layer defects, trapped powder, and unconsolidated powder are unique to PBF.20 In addition to AM processes, processing parameters,21,22 and post-processing23 also govern a part’s defect state. Application of the NDE procedures described in ASTM E3166, Standard Guide for Nondestructive Examination of Metal Additively Manufactured Aerospace Parts after Build,19 to asbuilt and post-processed parts is intended to establish a quality record of the defect state before service, thus reducing the likelihood of failure during service. This, in turn, mitigates the attendant risks associated with loss of function, and possibly, the loss of ground support personnel, crew, or mission. The range of NDE and optical measurement techniques relative to the defect location that is interrogated and spatial resolution that is obtained is illustrated in 24 figure 1. Any given NDE method such as those shown in figure 1, will have its advantages and limitations, which are described in more detail in ASTM E3166.19 Process history, flaw type, flaw size, flaw distribution, part dimensions, effects of material or microstructure on signal attenuation and part complexity are all relevant considerations for the selection of appropriate NDE methods as part of an inspection plan. One promising NDE method to assess the structural integrity of complex metal AM parts is X-ray CT. Detectability of defects by CT is directly correlated with the size, thickness, and complexity of the object being analyzed. Recent work has shown the significant progression of quantitative CT technology as a tool for assessing the integrity of structural AM components25,26 and precursor materials.27 Figure 2 illustrates the dependence of CT resolution (voxel size) on object size and provides relevant information regarding the limitations of defect detection in typical AM parts.28 Simply put, the smaller the AM part is, the better the resolution will be. This places extreme demands on the practicality of CT to scan large parts and emphasizes the need to understand the defect types, sizes, and distribution responsible for failure when developing an appropriate inspection strategy. These considerations determine which CT technique should be used (fig. 3) and may

95

process conditions, process parameters, develop

during Build (PC16)

NDE equipment or demonstrate NDE capability for

detection of naturally occurring flaws (lack of fusion,

for Demonstrating NDE Capability

(NDE2)

marks, embedded geometrical features).

porosity), or intentionally added features (water-

Design and make artifacts or phantoms to calibrate

Design and Make Artifacts or Phantoms

during build (see NDE3)

methods for sensor-based monitoring of the part

In addition to standards for monitoring feedstock,

level of NDE (see NDE8)

appropriate qualification requirements such as the

associated with a part and used as a metric to gauge

Develop a part classification system to describe risk

371616 (see PC16)

practice where needed, for example, in MSFC-STD-

In-Process Monitoring of AM Parts

AM Part Classification System (QC2)

subject matter experts and supported by the AM

after Build (NDE3)

community to assess and augment current best NDE

Develop industry-driven standards19,20 led by NDE

mately captured in ISO/ASTM 5290077

in ASTM E316619 ISO/ASTM DTR 5290519 and ulti-

of AM flaws detectable by NDE methods, developed

Adopt uniform terminology for the identification

Recommendation

Standard Guide for NDE of AM Parts

Common Defects Catalog (NDE1)

Gapa

TABLE 1 Prioritized list of NDE-related standardization gaps for AM aerospace parts

medium

medium

(important but low TRL)

high

high

high

Priorityb Organization(s)

(continued)

ASTM F42/ISO TC 261

ASTM E07.10

NASA, AWS

NIST

ASTM E07, ASTM F42/ISO TC 261, ASME,

ASME, AWS D20, NIST

ASTM E07, ASTM F42/ISO TC 261, SAE,

96 STP 1620 On Structural Integrity of Additive Manufactured Parts

(continued)

anticounterfeiting

Develop NDE methods and standards for

polymeric and nonmetallic AM parts

assess and augment current best NDE practice for

experts and supported by the AM community to

Develop standards led by NDE subject matter

critical AM parts (see QC2)

Establish NDE acceptance classes for fracture

provide a simple, unified interpretation of results

of additional or complementary NDE data to

Develop industry standards to allow combination

internal features in AM parts

standards (E1570, E1695) for measurement of

Address the applicability of current and draft CT

Recommendation

low

low

medium

ASTM F42/ISO TC 261, ASTM, SAE

ASTM D20

ASTM F42/ISO TC 261, ASTM E07,

ASTM E08

ASTM F42/ISO TC 261, ASTM E07,

ASTM

ASTM

medium

medium

Organization(s)

Priorityb

Note: AM ¼ additive manufacturing; ASME ¼ American Society of Mechanical Engineers; ASTM ¼ ASTM International; AWS ¼ American Welding Society; ISO TC ¼ International Organization for Standardization Technical Committee; NASA ¼ National Aeronautics and Space Administration; NDE ¼ nondestructive evaluation; NIST ¼ National Institute of Standards and Technology; SAE ¼ SAE International; TRL ¼ technical readiness level. a Analogous gaps in the Standardization Roadmap for Additive Manufacturing41 are given in parentheses. b Rankings based on four criteria: (1) criticality, (2) achievability, (3) scope, and (4) effect.42

NDE of Counterfeit AM Parts (NDE7)

Materials (NDE6)

NDE of Polymers and Other Nonmetallic

Critical AM Parts (NDE8)

NDE Acceptance Criteria for Fracture

Data Fusion (NDE5)

Features (NDE4)

Dimensional Metrology of Internal

Gapa

TABLE 1

WALLER ET AL., DOI: 10.1520/STP162020180106

97

98

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 1 Comparison of NDE techniques and optical measurement techniques according to detectable defect location and spatial resolution.24

FIG. 2 Resolution versus object size limitations.28

WALLER ET AL., DOI: 10.1520/STP162020180106

FIG. 3 Semiquantitative interdependence of X-ray energy, penetration, and resolution for steel using computed tomography.19

justify the use of high-energy synchrotron-based CT systems for larger or thicker parts with hidden embedded features.29 Although high-powered 2 to 5 MeV and linear detector array (LDA) and synchrotron-based CT systems are not practical for industrial applications, this capability does exist at NASA and the Japanese Aerospace Exploration Agency (JAXA) as well as at its contractor facilities. Although high-resolution micro-focus and nano-focus X-ray CT systems can be used on subscale samples (e.g., cubes and witness coupons) to detect micrometer-size or submicrometer-size defects, these defects may not be relevant to the life-limiting defects in actual production parts. In addition, as with any NDE-based defect detection strategy, it is also important to ask the following questions: (1) what is the largest defect that can go undetected, (2) what is the effect of a given defect type, size and distribution on part performance and safety, and (3) what is the consequence of catastrophic part failure stemming from such inspection misses? In addition to a much-needed AM defect catalog,4,18 which tells the NDE inspector which defects are important and need to be detected, other NDE-related technological gaps15 prevent full infusion of AM in NASA, JAXA, the European Space Agency (ESA), and related commercial space applications. The technology

99

100

STP 1620 On Structural Integrity of Additive Manufactured Parts

gaps identified in 201415 are still relevant today. Some of the more important gaps include (1) developing NDE methods tailored specifically to AM parts, (2) using NDE to understand effect-of-defect problems, (3) establishing NDE detection limits and accept or reject criteria for technologically important AM defects, and (4) fabricating AM physical reference standards to verify and validate NDE equipment and AM processes. Waller et al.15 also cataloged NASA agency and prime contractor AM activity, which recently has included fabrication of turbopumps, injectors, combustion chambers, and nozzles (fig. 4).10 Another important milestone governing how NDE is used to qualify metal laser-PBF (L-PBF) spaceflight hardware is the release of two NASA Marshall Space Flight Center (MFSC) quality documents.16,30 It is notable that these quality documentations, while calling out NDE for

FIG. 4 Additive manufactured activity (top15) and hardware such as injectors (bottom10) fabricated by NASA.

WALLER ET AL., DOI: 10.1520/STP162020180106

FIG. 5 Evolving sequence of additive manufacturing nondestructive testing and related standardization gaps (arrows indicate current activity).42

qualification and certification (Q&C) of spaceflight hardware as part of the general part production plan, do not provide explicit, step-by-step procedural detail on how the NDE methods are performed. This is despite the fact that NDE is a key decisional point for accept or reject of fracture critical L-PBF parts (fig. 5), the findings of which determine whether a part is approved for service or returned to an internal material review board for final disposition.

Considerations for Fracture-Critical Hardware Perhaps the main challenge confronting NASA,16,30 its international space partners, and the aerospace sector31 is the qualification of fracture-critical AM parts using a combination of NDE inspection and destructive test, especially in applications in which the structural margins are low and the consequence of failure is high. Such parts use a damage-tolerance rationale and require careful attention. At this time, it is not clear whether or not defect sizes given in table 1 in NASA-STD5009,32 which were derived from conventional aluminum crack panels made in the 1980s, are applicable to AM hardware, particularly when the as-built AM part surface is still present and surface-sensitive NDE techniques, such as ET and PT, are used (fig. 1). A demonstration of adequate life starting from the NASA-STD-5009 flaw sizes generally is deemed inappropriate for fracture-critical, damage-tolerant AM parts. Knowledge of the critical initial flaw size (CIFS) will allow the fracture control and NDE community to evaluate risks and communicate meaningful recommendations regarding the acceptability of the risk. It is recognized that parts

101

102

STP 1620 On Structural Integrity of Additive Manufactured Parts

with high AM risk, as defined in MSFC-STD-3716, Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals,16 may have regions inaccessible to NDE. To understand these risks, it is important that inaccessible regions are identified along with the corresponding CIFS. Many AM parts will require the use of multiple NDE techniques to achieve full coverage of a part. Therefore, a combination of both volumetric and surface-sensitive NDE methods may be necessary. Surface inspection techniques also may require the as-built surface to be improved to render a successful inspection, depending on the defect size of interest, surface curvature, signal-to-noise ratio, and other factors. In addition, removal of the as-built AM surface merely to a level of visually smooth may be insufficient to reduce the NDE noise floor because of near-surface porosity or boundary artifacts. It is incumbent upon the structural assessment community, including ASTM Committee E08 on Fracture and Fatigue, to define the CIFS for AM parts if the risks associated with spaceflight parts and the objectives of the NDE are to be adequately addressed. Specifically, more effort needs to be focused on the characterization and understanding of fatigue and fracture properties of AM materials, including appropriate use of the corresponding mechanical test methods.5–7,12,33,34 In addition to “conventional” crystallographic fatigue crack initiation mechanisms in homogeneous substrate materials, crack initiation resulting from the presence of inherent AM material anomalies, such as porosity, lack of fusion defects, or inclusions, also needs to be considered.35–41 Note that characterization of defects in fracture-critical AM parts must be based on realistic variation in material properties, microstructure, and material defect characteristics representative of the full-scale production environment. Failure to do so may result in “lessons learned” similar to those experienced during the early days of powder metallurgy (PM), when an undetected nonmetallic inclusion in a PM turbine disk was found to be responsible for the failure of a fracture-critical component that caused the crash of an F-18 aircraft.31 The AM process offers a unique opportunity to build hardware that will allow for defect detection directly in the part. For example, a demonstration part with simulated CIFS, surface-connected, and volumetric defects could be built. Partspecific NDE detection capability then could be demonstrated at the accepted probability of detection (POD) for a given defect size to establish the acceptability of the NDE for screening fracture-critical AM parts. The physics of the layered AM process will place additional constraints on the NDE performed and its acceptability. Because volumetric defects with significant height in the build (z) direction are prohibited, the concern for the NDE inspector instead will be the detection of planar defects, such as aligned or chained porosity or even laminar cracks that can form along the build plane. The layered structure of AM parts has four important implications from an NDE and fracture and fatigue perspective:16 • Planar defects are particularly well-suited for growth.

WALLER ET AL., DOI: 10.1520/STP162020180106

The primary defect orientation of concern is defined, which may be meaningful in analysis or with NDE detection methods dependent on alignment with volumetric defects. • Planar defects will generally exhibit very low contained volume. • The limited z-height of planar defects can be demanding on incremental step inspection processes such as CT. The level of criticality of AM aerospace parts is expected to increase as this technology matures and gains widespread acceptance.31 As the use of metal AM parts in fracture-critical applications becomes more accepted, the need for Q&C standards covering all aspects of the part life cycle will become more prevalent. Fracture-critical properties are affected by a variety of factors, and two of the dominant factors already mentioned are defect density and microstructural variation.2,3 Although the goal of producing defect-free parts in as-deposited materials remains an area of extreme interest, NDE screening methods and various post-processing techniques (e.g., HIP) may continue to be needed in fracture-critical applications. Noncritical locations, however, with no structural demand (i.e., areas of low stress and strain) in fracture-critical parts may not require the same level of damage tolerance or ultimate properties as would be needed in highly stressed areas. The type and rigor of Q&C procedures also are driven by the relevant industrial sector (e.g., aerospace, defense, medical, energy, or automotive) and end-use application (e.g., primary structure, secondary structure, spare, prototype part). For example, the Q&C requirements for high-value, limited production quantity aerospace parts are more stringent than for high-production quantity commodity parts. This is due to the unique safety concerns associated with human space flight and military and commercial aviation, which impose added rigor and stringency than can be justified in other sectors. •

CURRENT NDE STANDARDIZATION GAP ANALYSIS

Although the aerospace, medical, and other industrial sectors commonly use NDE methods, the scope of a given method may be limited to a specific application, material, and characterization goal. The U.S. military first controlled many of the standards used for inspection of products provided by the aerospace industry. During the military acquisition reform of the 1990s, changes shifted control from the military as owners to industry as owners.42 Two primary receiving organizations for emerging industry standards were ASTM International and SAE International. These organizations continue to create, revise, and release national NDE standards used by U.S. industry. NDE methods promulgated in national standards are often differentiated by the defect location in the part for which the inspection method is best suited. These defect locations are often referred to as embedded, subsurface, surface, or surface breaking, as elaborated on in detail in Dutton et al.18 and later in this paper. Embedded defect detection methods include acoustic emission (AE), CT, ET, RT, IRT, and UT; surface and near-surface defect detection methods include ET and PT (fig. 1).

103

104

STP 1620 On Structural Integrity of Additive Manufactured Parts

Given the number of commercially available AM processes (PBF and DED have the highest level of technical maturity and capacity to produce aerospace parts with detailed features) and materials (metals, polymers, ceramics, composites, and gradient materials), each part will have its own level of complexity and challenges from an NDE inspector’s perspective. A determination as to whether to separate or combine different AM processes into one or more standards should provide a coordinated answer to both the NDE inspector and AM-processing equipment user. Industrial sector NDE standardization needs or gaps have been evaluated and are summarized in the American National Standards Institute (ANSI)–America Makes Standardization Roadmap for Additive Manufacturing,42 which lists 95 national standards that are in progress or slated for development. Figure 6 shows NDErelated gaps relevant to the aerospace sector, along with corresponding draft standards currently in work. Of those standards, eight pertain directly to NDE of aerospace AM parts (Gaps NDE1-8). The eight gaps and two other closely affiliated gaps corresponding to AM part categories (Gap QC2) and in-process monitoring (Gap PC16) are listed in table 1 and are shown in figure 6. Part categories will be discussed in more detail later in this paper. For in-process monitoring of AM parts during a build, the reader is referred elsewhere.42–45 The list of standardization gaps given in table 1 is intended to be representative of the NDE-related standardization needs in the aerospace sector. Note that the standardization needs for polymeric aerospace parts (Gap NDE6 in table 1 and fig. 6), which exhibit good specific strength, low weight, and similar optimized design opportunities compared with metal parts, have not been designated for fracture-critical applications. Consequently, the development of related NDE-related standards for polymeric AM parts is deemed to be of lower priority.

INTERNATIONAL SPACE AGENCY APPROACHES FOR NONDESTRUCTIVE TESTING

NASA Approach

The development of NASA and national industry standards such as those listed in table 1 is required to properly balance the benefits of AM with its inherent risks. Although significant effort is focused on developing national standards for AM by voluntary consensus organizations (VCOs), such as ASTM and SAE,41 the content and scheduled release of many of the standards needed to ensure proper qualification of AM hardware in many cases does not support the near-term programmatic needs of NASA, or by analogy, its international space partners. Given that AM parts were slated for NASA spaceflight use as early as 2019,46 NASA Marshall Space Flight Center (MSFC) personnel released a NASA Center-level standard16 to establish uniform practices for the L-PBF process. The MSFC standard has been used as a basis for L-PBF process implementation for each of the manned spaceflight programs, including Commercial Crew, the Space Launch System, and the Orion Multi-Purpose Crew Vehicle programs. The MSFC standard and its companion

WALLER ET AL., DOI: 10.1520/STP162020180106

FIG. 6 Key products and processes in MSFC-STD-3716 (top) and symbol legend (bottom), including nondestructive evaluation (NDE) decisional activities (bottom right triangle).16

105

106

STP 1620 On Structural Integrity of Additive Manufactured Parts

specification provide a consistent framework for the development, production, and evaluation of L-PBF parts for NASA’s spaceflight applications. On the basis of the principles of MSFC-STD-3716, Standard for Additively Manufactured Spaceflight Hardware by Laser Powder Bed Fusion in Metals,16 the development of NASA-level standards to meet NASA AM-related program needs is underway for manned spaceflight, non-crewed spaceflight, and aeronautics. It is currently unknown whether any of the new NASA standards will overlap voluntary consensus standards identified in the ANSI–America Makes Standardization Roadmap for Additive Manufacturing,42 table 1, or figure 6. Because many of the NDErelated standardization activities require extensive research and agency resources, for example, effect-of-defect studies or fabrication of AM physical reference standards, the goal instead has been to combine resources and promote collaboration between NASA, its international space partners, and partners in industry, government, and academia, whenever possible or practical. The NDE standards developed, whether under NASA or VCO jurisdiction, will be applicable to mature technologies (technology pull). For NASA, this entails a focus on NDE of metal parts made by PBF and DED processes (wire-fed DED is used by NASA). AM materials determined to be out of scope with current NASA NDE and AM priorities include polymers, composites, ceramics, regolith, and printed circuits. For these AM materials, reliance on VCO standards is anticipated unless no existing VCO standards exist, or if existing VCO standards do not serve the public interest or are incompatible with NASA’s missions, authorities, priorities, and budget resources.47 NASA Qualification Requirements

As mentioned earlier, the MSFC Technical Standard16 and Specification30 provide the framework and establish general qualification requirements for the development and production of spaceflight hardware produced using the L-PBF process. Figure 5 illustrates the key products and processes controlled by MSFC-STD-3716 and shows, figuratively, how each product or process is related. General requirements of an additive manufacturing control plan (AMCP), which governs foundational process controls and part production controls, also are shown. Together, these controls provide the basis for reliable part design and production. Foundational controls include a qualified metallurgical process (QMP) for each machine, an equipment control plan (ECP), personnel training, and material property development through a material property suite (MPS). Part production controls are typical of aerospace operations and include part categories, a part production plan (PPP), a qualified production plan (QPP), and other miscellaneous production controls. The flowchart symbols in figure 5 indicate the type of product or action, such as internal documents, documents requiring approval, databases, or in the case of NDE, decisional actions. The AMCP also defines how an active quality management system (QMS) is integrated throughout the process. A QMS commonly used by the aerospace industry is SAE AS9100D48 (or an approved equivalent). Key points of QMS integration are illustrated with a

WALLER ET AL., DOI: 10.1520/STP162020180106

triangles symbol in figure 5. The AMCP and the QMS govern the engineering and QA disciplines, respectively, from start to finish. NASA Integrated Structural Integrity Rationale To ensure the design state of an AM part is mature before use in service, all L-PBF parts shall have an integrated structural integrity rationale summarized in the PPP that ensures that the structural integrity of the part is commensurate with its consequences of failure and associated requirements. The integrated structural integrity rationale describes, in succinct fashion, how the QA activities imposed on the part, including NDE, when considered as a whole, form sufficient rationale for structural integrity. In addition to NDE (inspection), other QA activities commonly include but are not limited to L-PBF process controls, proof testing, leak testing (if applicable), and component-level functional acceptance testing. Whether quantitative and qualitative test and inspection (T&I) rationale are levied on a part depends on the part category (vide infra). For critical flight applications, the responsibility to evaluate technical aspects of the integrated structural integrity rationale for an L-PBF part generally will rest with the fracture control community, where the disciplines of material and processes, structural assessment, NDE, and safety and mission assurance intersect. NASA NDE Inspection Requirements NASA L-PBF parts may be subject to multiple QA activities to achieve coverage of the part, particularly those classified with low structural margin, high consequence of failure, or high L-PBF risk. For example, Class A parts with a high consequence of failure shall receive comprehensive NDE for surface and volumetric defects within the limitations of current NDE best practice and part geometry unless otherwise stated. The rationale also identifies areas or volumes of the part relying solely on process controls, that is, areas not verifiable by post-build inspection, as risk areas for further consideration. As alluded to earlier, VCO or internal NASA standards containing NDE acceptance criteria developed for welded or cast hardware may not be applicable to L-PBF hardware. Nonconformance NDE findings are handled through the QMS. For Class A parts, NDE indications of cracks, crack-like defects, or other AM defects known to affect performance adversely are elevated to senior review and disposition by the appropriate cognizant engineering organization as required by applicable fracture control policy. This approach not only ensures proper disposition of the nonconforming part, but also is instrumental in identifying the process causes that created the condition. A summary list or table with all NDE production steps, including NDE reports detailing part acceptance certificate-of-compliance information is also required as part of the quality documentation for L-PBF parts. NASA In-Process Monitoring Requirements Currently, findings of build disruptions or product anomalies by in-process monitoring during the build must be evaluated by the cognizant engineering

107

108

STP 1620 On Structural Integrity of Additive Manufactured Parts

organization within NASA before being used as part of the integrated structural integrity rationale governing acceptance of a part for subsequent use. Care is taken to ensure that any in-process monitoring used by NASA does not alter the qualified L-PBF process. In other words, incorporating or integrating in-process monitoringsensing techniques into the L-PBF process is not permitted to alter the process and must be passive. Additionally, preference is given to current commercially available in-process monitoring methods, such as infrared melt-pool monitoring with higher demonstrated technology readiness levels. These methods will be better suited for NASA production work versus research and development. NASA NDE and In-Process Monitoring Limitations Until production experience matures, the L-PBF process faces added risks associated with unknown or insufficiently mitigated failure modes. As noted, available process controls, such as sampling of witness coupons, are useful for identifying systemic lapses in the L-PBF process, but they may not correlate directly to the structure integrity of finished hardware. In-process monitoring methods for active feedback control or post-build play-back verification are emerging and have the potential to produce qualified as-built parts,15,49 yet they remain in development, and their inclusion in NASA’s certification processes is premature. Beyond assurances of process stability, mitigation against local process anomalies and defects relies mainly on NDE methods and structural acceptance proof testing. Such mitigations are emphasized in MSFC-STD-3716,16 but they can be significantly challenged by the design freedom inherent to the AM process. NASA and Related Effect-of-Defect Studies AM flaws known to be detrimental to AM part performance include surface roughness50 and horizontal lack-of-fusion (H-LOF) flaws.51 Studies at NASA MSFC on the effect of bulk and surface defects in L-PBF parts also have shown a correlation between defect location (surface, near surface, or bulk) and effect-of-defect (little or no effect, out-of-tolerance part, degradation, of relevant property) (fig. 7). In addition, examination of the effect of H-LOF defects shows a significant deterioration of tensile properties, most notably in the ultimate elongation of Inconel 718 L-PBF specimens51 (fig. 8). Although subsurface defects often can be eliminated by postprocessing, such as HIP,2,11,52 the resulting microstructural changes or coarsening can produce strength reductions,53 while subsequent heat treatment may lead to the reappearance of defects.54 It is also known that surface roughness negatively affects fatigue resistance of metallic materials. Thus, reducing the surface roughness by polishing should improve the fatigue resistance of AM materials, whether in lowor high-cycle-life fatigue test regimes.55,56 NASA Additive Manufactured Part Categories To allow requirements to be tailored for a particular NASA application, a classification system has been devised to define and communicate the risk associated with a

WALLER ET AL., DOI: 10.1520/STP162020180106

FIG. 7 Surface and bulk defect consequences noted for Laser-Powder Bed Fusion InconelV 718 specimens.51 R

given AM part and to levy an appropriate level of QA through sufficient process control, inspection, and testing. The current NASA MFSC classification system16 shown in figure 9 is based on three key decisions: consequence of failure, structural demand, and AM risk. AM risk, in turn, is calculated based on five ratings criteria: (1) Can all surfaces and volumes be reliably inspected, or does the design permits adequate proof testing based on stress state? (2) Can as-built surface be fully removed on all fatigue-critical surfaces? (3) Are surfaces interfacing with sacrificial supports fully accessible and improved? (4) Are structural walls or protrusions 1 mm in cross-section? and (5) Can it be assured that critical regions of the AM part do not require sacrificial supports? This decision tree leads to eight distinct part categories. Alternatives to the classification scheme shown in figure 9 could allow for parts with a low consequence for failure to be placed into a “do no harm” category designated as Class C.46 JAXA Approach

The importance of NDE as a tool to screen AM products of varying criticality has similar relevance to JAXA, as mentioned for NASA. Like its international space

109

110

STP 1620 On Structural Integrity of Additive Manufactured Parts

R

FIG. 8 Large drop in the tensile elongation of Laser-Powder Bed Fusion InconelV 718 specimens containing horizontal lack-of-fusion flaws such as those pictures in the inset.51

FIG. 9 NASA MSFC-STD-3716 additive manufactured part classification system.16

WALLER ET AL., DOI: 10.1520/STP162020180106

partner peers, JAXA is actively involved in making AM spaceflight hardware, also with an emphasis on PBF and DED parts (powder DED is used by JAXA). Consistent with the goals of a national Japanese Technology Research Association for Future Additive Manufacturing (TRAFAM) effort sanctioned by the Japanese Ministry of Economy, Trade and Industry,57 JAXA Safety and Mission Assurance (S&MA) experts are conducting research on domestic NDE capabilities to detect flaws unique to AM. More specifically, current work is focusing on clarifying the flaw detection capability limit used for microfocus X-ray CT (micro-CT). JAXA recently reported on its QA-related activities, including NDE, at an ESANASA-JAXA trilateral S&MA meeting.58 In that meeting, JAXA shared information on activities related to AM standardization, establishing NDE quality assurance guidelines and establishing requirements for AM parts. As for AM standardization, TRAFAM57 is participating as a representative of Japan with ISO TC 261, which has a collaborative agreement with ASTM Committee F42 on Additive Manufacturing Technologies to adopt international standards for AM under joint ISO/ASTM jurisdiction. Standards related to AM being developed in Europe and in the United States thus will have an opportunity for Japanese participation and concurrence. As for NDE QA guidelines, analogous to the NASA MSFC approach, NDE is considered mandatory for JAXA Category A spaceflight hardware (fig. 10). Efforts to date in NDE performed by JAXA include application of a novel technique to reduce micro-CT beam-hardening artifacts in images of AM parts by using an HCFC-225 screening liquid (fig. 11). These findings have been incorporated into the ASTM Committee E07 standard mentioned earlier.19 The use of tin and copper metal filters and appropriate beam-hardening correction factors also have been investigated by JAXA. Taken as a whole, these measures require trial and error to achieve the best result for improving CT contrast-to-noise ratios under standard resolution imaging conditions. JAXA is actively continuing its NDE research and discussions and is accumulating technical data associated with the interrelationship between flaw size and fatigue failure to establishing meaningful acceptance criteria for AM defects. JAXA Qualification Requirements

Quality assurance aspects for rockets and satellites built and provided by JAXA are managed by applying JAXA QA program standards.58,59 Although AM has similarities with welding, and welding calls out NDE to confirm an article’s quality level, JAXA’s QA program standards do not yet include specific requirements for controlling the unique aspects of the AM process and corresponding hardware. Similar to the challenges faced by NASA and ESA, the content and scheduled release of many of the industry standards needed to ensure proper QA of AM hardware may not support the near-term programmatic needs of JAXA. In addition, because AM is considered to be a special process by JAXA, special process control per JAXA’s quality assurance program standards59,60 is required. As was the case for NASA, the approach to establish specific requirements for AM technology, and more

111

STP 1620 On Structural Integrity of Additive Manufactured Parts

FIG. 10 Draft JAXA classification system for metal additive manufactured parts showing categories A1-A3 where nondestructive evaluation is mandatory.46 In-Process (Manufacturing plan development) study

N ① Flight Hardware

Y

Use In-Process Evaluaon Table

N

② High Strength Structural Parts

Y

Visualize Crical Parameter

N

③ Complex Shape

Y N

④HIP Not Performed [HIP Not Performed]

Use 3D X-ray CT Quality Evaluaon Table

Y [CT Cannot Be Used]

Opmize Imaging Parameter

Y

[HIP Performed]

⑤Consider Convenonal

⑤Same

⑤Same

NDE (ET, MT, PT, RT, UT)

as le

as le

Post-Process (NDE) study ⑧ CT can’t be used due to size/machine constraints

⑥Consider 3D X-ray CT

N [Can be used] Y

⑨ Flaw

detecon accuracy 63 :m: < 1%

>53 :m: 2%

>53 :m: 1%

>45 :m: 4%

254 STP 1620 On Structural Integrity of Additive Manufactured Parts

COUGHLIN ET AL., DOI: 10.1520/STP162020180109

(SE), representing powder in front of a recoater during spreading. The D10, D50, and D90 values, which were measured using laser diffraction, describe the powder particle diameters below which 10%, 50%, and 90% of the particles exist. The compositions of the powders used for the L-PBF builds are provided in table 2; parameters used to fabricate the L-PBF builds are given in table 3. The build platform material for builds L-PBF-A and L-PBF-B was TOOLOX 33 quenched and tempered tool steel; for build L-PBF-C, the build platform material was 1.2083 stainless tool steel. Builds L-PBF-A and L-PBF-B, which were fabricated using a 3D Systems ProX DMP 320 machine, consisted of an array of tension test specimens built in each of the eight corners of the build volume. This test article build design is intended provide a model for a qualification method that is based on ensuring acceptable material performance through the assessment of variability in tensile properties related to specimen surface finish, thickness, orientation, and location within the build volume. Each tension test specimen array in the build consisted of a “thick” (6.35 mm) and “thin” (0.76 mm) subsize rectangular tension test specimen in three orientations relative to the build platform: horizontal, angled (53 ), and vertical. The support structures for the builds were designed to have minimal contact with the gauge regions of the specimens. A rectangular prism “witness specimen” that spanned the full build height was fabricated at each corner of the build platform. A photograph of one of the lower corners of build L-PBF-A is shown in figure 1A. A photograph of an upper region of build L-PBF-A is shown in figure 1B, where it can be seen that the vertically oriented specimens in this region were not successfully completed. The thin and thick specimens in each orientation were separated by 0.76 mm; this space can be seen for the horizontal specimens in figure 1A. The tension test specimens in build L-PBF-B were built with an additional 0.13 mm on all surfaces compared with the L-PBF-A tension test specimens to account for post-process surface finishing. Both builds underwent a stress relief heat treatment at 900 C for 75 min in vacuum, followed by a nitrogen quench. The specimens were then removed from the build platforms, and the tension test specimens from build L-PBF-B were subjected to an isotropic superfinishing (REM ISF) process. This finishing process involves circulating the parts within ceramic media inside a vibratory bowl while chemicals are introduced. The tension test specimens from build L-PBF-A did not undergo any surface finishing treatment, aside from the minimum required to remove support structure. All specimens that were successfully built were tension tested at room temperature in accordance with ASTM E8, Standard Test Methods for Tension Testing of Metallic Materials.21 Build L-PBF-C, which was fabricated using an EOS M290 machine, consisted of “thick” (6.35 mm) and “thin” (0.76 mm) rectangular tension test specimens in three orientations relative to the build platform: horizontal, angled (50 ), and vertical. This build was designed to assess potential variability in tensile properties based on specimen thickness, orientation, and location on the build platform, as well as inter-specimen spacing. The thin horizontal and angled specimens were built in

255

Balance

L-PBF-C

Fe

66.29

L-PBF-A, B

Builds

18.13

17.62

Cr

14.79

12.78

Ni

0.007

0.02

C

TABLE 2 Powder chemical composition for L-PBF builds, weight % Mo

2.78

2.42

Mn

1.16

0.41

Si

0.33

0.39

P