So You Want to Design Aircraft: Robots on the Floor 2017946556, 9780768084245, 9780768084252, 9780768084450, 9780768084443, 0768084245

Citation preview

So You Want to Design Aircraft: Robots on the Floor

Other SAE Books of Interest: Fundamentals of Engineering HighPerformance Actuator Systems By Kenneth Hummel (Product Code: R-459)

Successful Prediction of Product Performance By Lev Klyatis

(Product Code: R-448)

Automated/Mechanized Drilling and Countersinking of Airframes By George Nicholas Bullen (Product Code: R-416)

For more information or to order a book, contact: SAE International 400 Commonwealth Drive Warrendale, PA 15096, USA Phone: 1+877.606.7323 (U.S. and Canada only) or 1+724.776.4970 (outside U.S. and Canada) Fax: 1+724.776.0790 Email: [email protected] Website: books.sae.org

So You Want to Design Aircraft: Robots on the Floor Edited by Jean L. Broge

Warrendale, Pennsylvania, USA

400 Commonwealth Drive Warrendale, PA 15096 USA E-mail: [email protected] Phone: +1 877.606.7323 (inside USA and Canada) +1 724.776.4970 (outside USA) Fax: +1 724.776.0790 Copyright © 2018 SAE International. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, distributed, or transmitted, in any form or by any means without the prior written permission of SAE International. For permission and licensing requests, contact SAE Permissions, 400 Commonwealth Drive, Warrendale, PA 15096-0001 USA; email: [email protected]; phone: 1+724.772.4028; fax: 1+724.772.9765. SAE Order Number SYWD-0001 http://dx.doi.org/10.4271/SYWD-0001 Library of Congress Cataloging-in-Publication Data 2017946556 Information contained in this work has been obtained by SAE International from sources believed to be reliable. However, neither SAE International nor its authors guarantee the accuracy or completeness of any information published herein and neither SAE International nor its authors shall be responsible for any errors, omissions, or damages arising out of use of this information. This work is published with the understanding that SAE International and its authors are supplying information, but are not attempting to render engineering or other professional services. If such services are required, the assistance of an appropriate professional should be sought. ISBN-Print ISBN-PDF ISBN-epub ISBN-prc

978-0-7680-8424-5 978-0-7680-8425-2 978-0-7680-8445-0 978-0-7680-8444-3

To purchase bulk quantities, please contact SAE Customer Service: Email: [email protected] Phone: 1+877.606.7323 (inside USA and Canada) 1+724.776.4970 (outside USA) Fax: 1+724.776.0790 Visit the SAE International Bookstore at books.sae.org

Contents Introduction: Robots on the Floor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii Chapter 1: A Human Walks into an Assembly Cell... . . . . . . . . . . . 1 A mix of different external safety systems can allow the whole human-machine interaction (HMI) assembly cell to work properly in an industrial context. The scenario for HMI, in this case, is that an operator enters a robot working area with the aim to perform an assembly task that requires two hands. There are different support systems that could be applied to this assembly application, but every one of these systems needs to coincide with official standards to be applied in an approved industrial HMI assembly cell.

What is Safety?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Support Systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Test Setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maximizing Productivity and Safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 3 4 6

Chapter 2: Skills-Based Human-Robot Cooperation . . . . . . . . . 11 Experience has led to current assembly concepts necessarily consisting of “customized automation.” The authors offer a structured approach to the analysis of assembly processes to extract the required information necessary for a skills-based task assignment to either humans or robots. The analysis begins with the assessment of a product and follows the process through to the assembly of aircraft fuselages. An advantage of this method is that the proportions of manual and automated tasks can get reassigned quickly and easily. Thus, the executing workforce can get chosen appropriately according to the relevant conditions.

State-of-the-Art Review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 A Structured Approach for a Task Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Demonstration Scenario Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

Chapter 3: Modular Gripper for Flexible Assembly Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 The demand for passenger aircraft is growing steadily, with an annual increase in passenger numbers of around 4.7% expected by the year 2032. In this context, Airbus expects around 29,200 new orders up through that time frame. To meet this future demand leading aerospace companies must embrace the development of new, adaptable, and modular assembly systems. The use of intelligent and networked industrial robots is a very promising approach for highperformance, flexible, and complex assembly operations.

Challenges of Major Component Assembly. . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Flexible Aircraft Assembly Concept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 Laboratory Setup. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 v

Contents

Modular, Lightweight Gripper System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Transfer to Aircraft Production Applications . . . . . . . . . . . . . . . . . . . . . . . . . 38

Chapter 4: Embracing Reconfigurable Assembly Cells for Aerospace . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 There are efforts within industry to apply reconfigurable techniques to reduce operating costs and time-to-market. Despite the apparent benefits to aerospace and the reported savings and lessons learned from the automotive industry, a great deal of research is needed to suit the specific challenges of aerospace assembly. Most aerostructure assembly systems are designed to produce one variant only. Assembly systems that produce more than one variant do exist, but have long changeover or involve extensive retrofitting. This chapter presents novel contributions toward Reconfigurable Assembly Systems (RASs) for aerospace assembly and aims to advance development of the next generation of aircraft assembly systems.

State-of-the-Art RAS Review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . RAS Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Product Family and RAS Design Framework. . . . . . . . . . . . . . . . . . . . . . . . . Industrial Implementation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Limitations of the Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

42 47 48 50 51

Chapter 5: Optimizing Robot-Based Machining Systems . . 53 Three different methods are presented for the assessment of performance and accuracy of robots for machining processes or processes with path accuracy in general. By abstracting and generalizing the process of path realization for machining, the three methods were designed with respect to effort of realization and information output. In particular, besides methods available on the market, a new approach called “virtual machining,” which needs no real machining process and no additional measurement device, except the robot itself, is emphasized.

Methods for Performance Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 Optimization Strategies for Robot Machining. . . . . . . . . . . . . . . . . . . . . . . . . 59

About the Editor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

vi

Introduction: Robots on the Floor

Introduction: Robots on the Floor In the summer of 2017, Boeing released its Current Market Outlook 2017-2036, projecting that the global commercial jet fleet would double in size by the end of 2036, with 41,030 new airplanes delivered within a 19-year period. Airbus’s forecast for the same period was also substantial, projecting about 34,900 new aircraft— without accounting for regional jets in its forecasts. Both companies likewise agreed that approximately 60% of those new aircraft will be the result of organic growth in the market, while about 40% will consist of replacement aircraft for current fleets. There is also consensus that 40% of the overall aircraft will go to the Asian market, with 35% distributed between Europe and North America. There are plenty of other conclusions that can be drawn from the Boeing and Airbus market reports, especially for industry analysts, curious suppliers, and eager competitors. However, it doesn’t take a seasoned data scientist to realize two inevitabilities of the coming years. First off, that is a lot of new aircraft, and they will require a lot of new pilots. Second, and more pertinent here, all those aircraft will need to be made as quickly, efficiently, and economically as possible for companies to remain competitive and meet market demands. It will require—by all accounts—more than what is humanly possible. It is ironic that as aircraft have become more sophisticated, much of aircraft manufacturing has remained simplified and manual. However, as orders for commercial aircraft have dramatically increased over the past years (and are universally expected to remain on that trajectory), the competition and conversation has shifted from “how quickly new technologies can be incorporated into aircraft,” to “how quickly the aircraft can be manufactured and delivered.” Enter ever-increasing automation and robotic. To remain (or become) competitive in the aerospace industry, it is necessary to adapt new ways and methodologies to increase automation in manufacturing. For both aircraft on the flight line and on the production line, safety is paramount. To that point, the first chapter starts with the basics of defining safety and safety systems when it comes to both humans and robots working on the assembly floor.

vii

Introduction: Robots on the Floor This book, the first in a series that will explore different aspects of advanced manufacturing, then examines a variety of other topics that range from the risks and rewards of increased cooperation between humans and robots within manufacturing systems, to introducing a process that enables the determination of either human or robot tasking, to the configuration and optimization of flexible assembly cells. These topics will continue to evolve and mature as the industry ramps up to mass produce aircraft that are destined to become increasingly complex, even more complex as the systems that build them.

viii

Chapter 1

A Human Walks into an Assembly Cell... Rickard Olsen and Kerstin Johanses Linköping University

Magnus Engstrom Saab AB Increased collaboration between humans and robots is considered by many to be the next big challenge to address in manufacturing and assembly operations. As computing power becomes more potent, the possibilities of achieving safer working environments increase, as all safety signals demand fast data management. Such information could be the enabling power that leads to a human working closely or directly with a robot, using the robot as a third hand. There are different levels of interaction between humans and machines. The term human-machine interaction (HMI) encompasses all types of interactions. These interactions could be on such a low level where the human performs all actions and decision-making or at the highest level where the machine is autonomous. In this chapter, HMI is used to describe when the operator interacts with a robot. Additionally, the term human-robot collaboration (HRC) is used to designate when the operator is using the robot either in a collaborative operation or in a collaborative workspace. Therefore, when implementing HRC, the challenges are safety and dependability.

1

Chapter 1

What is Safety? There are a multitude of ways to define safety. In a manufacturing workspace, managing safety means adhering to various standards and regulations, especially regarding machines used in the manufacturing process. These standards and regulations aim to protect the operator from injury. The International Organization for Standardization (ISO) defines a safe state for a machine: “Condition of a machine or piece of equipment where it does not present an impendent hazard.” This is important to manage when implementing the next generation of HRC solutions. There are different kinds of safety systems, from the classic fence-guarded systems to pre-collision systems and post-collision systems. The classic fence-guarded systems traditionally consists of a robot surrounded by a fence. Here, the fence should be designed to prevent humans from accessing the robot workspace where hazards may be present. The fence should restrict robot motion if or when personnel access the workspace. Another way to prevent accidents between humans and robots is to utilize a precollision system. This method integrates external support systems with the communication system of a robot to monitor the workspace. Different types of sensors may be used for identifying personnel entering the workspace (e.g., vision systems, force sensors on the floor).

 An overview of different safety control systems is depicted, from the classic fence-guarded systems to pre-collision system and post-collision systems.

2

A Human Walks into an Assembly Cell... The third main type of safety system is the post-collision system. Here, integrated sensors, lightweight structures, or software-created barriers are used to prevent or minimize collision damage. Regarding ISO, there are three major applicable robotic safety standards, and these three standards cite other standards as well. One of these standards—the Technical Specification (TS) focusing on HRC—is under revision and has not yet been released. Therefore, the content in that TS can only be identified and referred to through other research and work, including interviews with key persons from that TS development workgroup. To use a robot in a collaborative mode, a visual indicator and one or more of the following is required: • Safety-rated monitored stop: A category 0 stop, or a decelerated to a category 2 stop, but with a safety catch that when it doesn't work it automatically goes into a category 0 stop • Hand guiding: This should be equipped with an emergency stop and an enabling device and it should operate with a safety-rated monitored speed • Speed and separation monitoring • Power and force limiting by inherent design or control In the robot system, the developer of the robotic cell defines the different requirements to ensure that the environment in the collaborative workspace is safe. A risk analysis is required to identify all the hazards that could be present in each workspace. The ISO standards set the safety limits (i.e., speed limits when the operator enters the robot working area) for an HRC setup.

Support Systems Concerning assembly tasks, HRC is considered mandatory for improved flexibility and adaptability. With the assistance of different support systems, the programming methods can be changed in the future (i.e., a simplified way for teaching the robot different paths). The cooperation between humans and robots will provide new concepts in the design of industrial robots, such as dual-armed robots or lightweighted robots. Today's focus on accuracy could be changed to safety in the future and support systems could help the robot system with both accuracy and safety. A 3D vision sensor and a force sensor are two different kinds of support systems. With the 3D vision sensor, positional errors may be identified; it is sensitive to motion and/or parts that are observed in a 3D volume. A computer manages received images from the vision system and then, using algorithms, creates a 3D environment of the vision systems’ subjected area.

3

Chapter 1

 There are multiple possible setups for an human-robot collaboration (HRC) cell, with three concepts depicted here. When an object enters the subjected 3D environment, the vision system should react in different ways depending on how close the new object is to the robot and what kind of speed vector it has. On the signal from the vision system, the robot reaction could be a change of path, a category 0 stop, or speed reduction, depending on the situation. Force sensors are used either externally or internally on a robot to ensure that it does not crash into a product, operator, or other foreign object that may have entered the robot’s projected movement area. Impact tests such as the head-injury criterion (HIC) test have evaluated robot system limitations and potential in relation to injury and velocity. Robot velocities up to 2 m/s is below current limit thresholds according to robot impact tests.

The Test Setup Based on an interview with a key individual with deep working knowledge concerning the future TS, it is clear that the TS will contain a more detailed description about collaborative robot systems including workspaces. Therefore, a demonstrator of the HMI assembly cell was developed to test two different safety setups. In the first setup, the robot performs a movement with a component to be assembled (the rib) and then the stops. Thereafter, the operator enters the workspace and performs an assembly task. In this case, the safety solution is straightforward, since the robot is stopped and in a controlled position. In the second setup, the robot and the operator share the same workspace at the same time (i.e., the robot can move while the operator is close to it). This case is more demanding from a safety viewpoint and requires more complex support systems to achieve a safe environment.

4

A Human Walks into an Assembly Cell... This layout can be used for testing both setups: 1. Robot positions the rib and stops. The operator enters the cell and performs assembly tasks while the robot secures the rib in position. 2. The robot and operator share the same workspace. The robot can move while the operator performs assembly tasks. The demonstrator cell evaluated different safety solutions both physically and virtually: 1. Camera safety solution (physically) 2. Reachability between operator and robot (physically) 3. Light curtain safety (virtually and physically) As a sub-project in this demonstrator development, a study on a safety monitoring solution using Microsoft’s Kinect was performed. This equipment monitored the area around the robot; locating moving objects, calculating the closest point of the moving object, and displaying it. It uses different color codes on a flat screen to communicate proximity: red for too close, yellow for collaborative mode, and green for normal mode. The system is controlled by Industrial ROS (Robot Operating System), a widely-used open source tool for robot programming.

 Researchers developed a rib assembly application that uses an external force sensor. The force sensor is controlled with Matlab Stateflow to create a production cell that is more flexible and cheaper than a production cell that uses dedicated equipment and jigs.

5

Chapter 1

 Computer-aided design (CAD) drawing of a portion of the demonstrator cell. The robot used is a Yaskawa SDA10; but in the demonstrator, only one arm was used. Light curtains were demonstrated in the virtual cell. The light curtains were oriented horizontally approximately 300 mm over the floor such that they covered the maximum space needed to avoid harmful collisions between the operator and the robot with the work piece. The choice of 300 mm also considered the length of a jumping human, while at the same time making it nearly impossible for an adult to crawl under the light curtain. Compared to a classic fence-guard system, this setup provided a more transparent view of the work area. A small fixture held the light curtains in the proper position. A controller was placed so the operator would need to leave the working area and push a button as an extra safety feature. The controller also gives the robot and safety system a signal when the operator is finished with the operation.

Maximizing Productivity and Safety During the demonstrator testing, it became obvious that reachability in combination with safety solutions that maximize productivity is a vital variable when implementing HMI. The demonstrator assembly cell mock-up is based on an early design in the overall research project. Therefore, the evaluations presented are based on experiences where the robot grips the rib from “the wrong side” compared to the final solution in the overall project. However, this does not affect the results regarding safety in the HMI solution. The first test was positive; the reachability and flexibility of the robot was adequate for the task. The Kinect solution showed that with a cheap and open source programming solution (ROS), it is possible to implement a safety monitoring system. However, it is important to ensure that the ROS is compatible with the selected robot and safety equipment. The developed tracking algorithm operated properly when one

6

A Human Walks into an Assembly Cell... object entered the Kinect’s collaboration area; but when two or more objects entered the area, the computer had problems processing the data in real time. Adequate computing capacity is important when implementing a camera safety solution. The different color coding in the visualization of the Kinect data signaled the actual robot location to the operator. If connected to the robot Technology Readiness Level system as an active safety system, the red would Technology Readiness Level (TRL) symbolize a safety-rated stop, yellow would is a measurement that assesses the indicate that the speed of the robot should be set maturity and helps to do comparison into collaborative speed, and green would signal between technologies. The levels grade from 1 to 9, with 9 as the most mature the system to operate at full speed. technology level. Roughly the levels

can be described as follows: the lower The virtual cell uses light curtains as a safety levels, 1-3, are identified as covering solution. This was chosen because the alternadifferent basic research issues, levels 4-6 tive, safety mats, are more expensive and do not are pre-industry implementation levels, and levels 7-9 are different levels for a have a high enough safety classification as light technology implemented in industry. curtains for use in an industrial environment. Furthermore, safety mats need to cover a large area of floor if tasked to operate without physical fences. Therefore, light curtains are, as of today, considered to be more cost effective.

Different evaluation trials for safety and layout solutions for a future HRC assembly cell need to be related to technology readiness levels (TRLs) in the overall project. Since the TS is not yet released and the threshold TRLs for the LOCOMACHs project is TRL 5 or 6, the results must be related to a flexible implementation of HRC. The TRLs in the overall project are defined as: • TRL 5 - small-scale partial integration of technology (with partially representative interfaces) • TRL 6 - large-scale major integration, highly representative implementation of technology in target elements The different trials in the demonstrator cells in this paper have different TRLs and different possibilities to reach higher TRLs. The physical cell setup in the lab should have reached TRL 5, though as the TS was not released yet, the safety solution only reached TRL 4. The virtual cell reached TRL 3 when it came to HRC. However, converting the virtual cell that visualizes an HMI-cell into an HRC cell could be achieved. The transition of the virtual HMI-cell into an HRC cell includes adding an enabling switch, a 3-stage safety button, and a camera safety solution (i.e., the Kinect solution). The Kinect solution that has been developed and evaluated here has, according to the overall project TRL definitions, reached a TRL 4.

7

Chapter 1

 The virtual cell developed for the demonstrator cell of HMI. A controller is placed so the operator needs to leave the working area and push a button as an extra safety feature. This controller gives the robot and safety system a signal when the operator is finished.

 In this image showing the positioning of the rib by the robot in the physical demonstrator of the test cell, it was desired to present the robot positioning the rib in relation to potential free space for an operator. Based on the performed trials in this demonstrator cell, current market equipment and technology can be combined with custom-made components to develop an HRC cell. It is possible to ensure operator safety when the robot moves; however, it is crucial to ensure safety when the operator and the robot share the same workspace.

8

A Human Walks into an Assembly Cell...

 The LOCOMACHs projects defined technology readiness levels (TRLs); different trials in the demonstrator cells presented here have different TRLs and different possibilities to reach higher TRLs. The solution must combine several different support systems and this will demand computer capacity and redundant system solutions. Furthermore, an enabling device must be used according to old robot safety standards and preliminary information about the content of the TS. Force sensors can be utilized to mitigate or minimize harm if the robot would collide with the operator, but this is recommended to be part of the redundant solution in the first implementation phases of HRC in assembly. Future research and development involves building and testing a physical HRC cell for a rib assembly in the aerospace industry in a complete demonstrator combining these results. The risk for implementing HRC in near future is related to the release of the new TS. Therefore, this research is focusing on several different supporting systems to have knowledge about the anticipated needs and possibilities to implement HRC stepwise. One solution is based on the existing standards, the physical cell setup; and one that is relevant for the future, the virtual cell. This stepwise implementation could gain the technology implementation in the future, since the operators are learning step by step how to collaborate with a robot in a safe way. The main conclusion is that safety solutions, which are reliable and do not reduce productivity, will be the key for implementing HRC in industrial assembly cells. This chapter is based on and imagery is taken from SAE International Technical Paper “On Safety Solutions in an Assembly HMI-Cell” by Olsen, R., Johansen, K., and Engstrom, M., doi:10.4271/2015-01-2429.

9

Chapter 2

Skills-Based HumanRobot Cooperation Rainer Mueller, Matthias Vette, and Ortwin Mailahn ZeMA Against the background of globalized markets, it is an ongoing challenge for manufacturing companies to maintain competitiveness in high-wage locations. Meeting high quality requirements, maintaining the ability to respond quickly to market challenges, and reducing production costs are just some key strategies to survive in the market. Historically, the assembly process has had a higher proportion of manual work performed during production compared to other areas. Particularly for large component assembly, such as aircraft fuselages, there is much potential for streamlining manufacturing. However, this consists of more than just broadly automating the maximum number of processes, as was the case in the 1990s. Past experience has led to current assembly concepts consisting of “customized automation.” Through the ability of robotic systems to work in cooperation with humans, a whole new area of automation has come about over the last years. The decision for or against an automated process no longer needs to be carried out for each basic task, but instead can be tailored to specific tasks according to specific human or robotic skills. Thus, tasks can be assigned on a needs-oriented basis and designed to respond flexibly to changes in the state of production, depending on the availability of equipment. Skills-oriented assembly sequence planning is essential for planners faced

11

Chapter 2 with the challenge of efficient resource management in the context of reconfiguration planning or changing boundary conditions. Previous research on this issue has included extensive modeling of possible assembly sequences that were then mapped, supported by software, on possible cooperation types. In practice, it turns out that the majority number of human-robot cooperation (HRC) forms are executed in autonomous or synchronized operation. Cooperating and collaborating operation forms have hardly been implemented mainly due to the difficulties in safeguarding the processes. This chapter will focus on a structured approach to the analysis of assembly processes to extract the required information necessary for a skills-based task assignment to humans or robots.

State-of-the-Art Review The state of the art still corresponds mostly to the strict separation of working spaces of humans and robots. Through the market availability of a new robot generation that should be qualified for direct HRC (due to a reduction of kinetic energy, a twochannel monitoring of the control, and different sensor concepts for collision avoidance), the concept of HRC has gained momentum. There are now many robot producers with different models on the market. Some of the most prominent representatives are the UR10 and budget-priced UR5 from Universal Robot; the two-armed YuMi from ABB; the APAS from Bosch, which is equipped with a capacitive skin; and the LBR iiwa from KUKA, which disposes of highly sensitive force-torque sensors in all axes. As industrial robots always constitute an incomplete machine, becoming only complete in combination with peripheral devices (e.g., grippers, tools, sensors, etc.), according to DIN EN ISO 10218, they can be assessed against the background of a defined application scenario.

 A new generation of robots is just one of the breakthroughs that has led to increased potential to optimize human-robot cooperation (HRC).

12

Skills-Based Human-Robot Cooperation

 Experts have distinguished four operating modes of HRC. The temporal and spatial separation of the work of humans and robots represents the current state of technology. Thus, the implementation of HRC critically depends on whether the necessary security conditions can be met. The preview of the TS ISO15066 at this stage provides the specification for risk assessment procedures and supplies technical and biomechanical limits and guidelines for the technical documentation of HRC. Here, the specific features of robots designed to work with humans are considered. Regarding the type of cooperation between humans and robots, the temporal and spatial separation of the work of humans and robots represents the current state of technology. Working humans and robots at the same time, but spatially separated so there is an autarkic operating mode since the work content can be matched to each other. Working at the synchronized operation form, humans and robots execute their tasks after each other in the same workspace. A cooperating operation form is used when humans and robots work simultaneously in the same workspace. A special form is the simultaneous execution of a common task. This hand-in-hand work is called collaboration. In literature, this term is often also used as a collective term for HRC. The cooperation of humans and robots enables the use of the respective advantages of these resources. The linking of these factors is the goal of HRC. Meanwhile, there are increasing pilot projects in the industry that want to leverage these accumulated benefits. In most cases, pilot projects are laid out to sequential operation forms. At an Audi production plant, a robot takes components from a box and hands it to the human at an ergonomic level. These forms of direct interaction are currently very rare.

13

Chapter 2

 The cooperation of humans and robots enables the use of the respective advantages of these resources. The linking of these factors is the goal of HRC. The decision on who has to do what work units shall be taken in the assembly planning. Current research efforts are aimed to use the information that is generated in the assembly planning procedure consistently throughout the real assembly stations. For this, descriptive models are developed, linking the virtual world with the real planning assembly system. In planning, this description model will ensure the compatibility of product requirements and skills of the equipment to perform the required assembly processes. This comparison can be extended to the skills of the human resource. In relation to humans, it’s important to assure that the resources are used in the context of ergonomically acceptable limits. To take account of this principle, digital human models are integrated into planning systems to perform an ergonomic assessment. An example here is the human model “Jack,” which is available as part of Siemens Technomatix. Due to the variety of different planning objectives, assembly scheduling is hardly feasible without software support. To quickly and flexibly meet task-specific decisions regarding the assignment of tasks to human or robot, below a structured approach is presented that is suitable especially for stations that need to be reconfigured frequently.

A Structured Approach for a Task Analysis The approach is divided into two sections. First, the description of the approach is established that structures the procedure methodically. Later, an application

14

Skills-Based Human-Robot Cooperation example of the aircraft industry is presented with respect to the demonstration scenario, which was implemented to evaluate the approach. The description model follows the methodological approach "product → process → resources.” First, a product analysis is performed, since the product is the direct connection to the customer and thus, the central driver for assembly tasks. The product sets the requirements for the process and the resources that will be analyzed afterwards. Finally, a procedure describes how the findings from the analysis can be incorporated in a description model so that the assignment of work tasks to human or robot can be made comprehensive and transparent.

Product Analysis The product analysis aims to detect the requirements set by the product to the components and resources. A product can be a module that is composed of several components. Furthermore, it can be composed of modules that get formed by single components or sub-modules that are brought to a higher order. Sub-modules are pre-produced in subassembly stations. The product analysis therefore begins at the level of the individual component. Six characteristics of a product must be analyzed. These properties have a direct influence on the allocation of tasks to human or robot: • Number of parts • Delivery condition • Handling characteristics • Joining direction • Joining method • Quality requirements The number of parts indicates how often the component is installed in the product. If the product is getting produced in different versions, the number of parts may vary for each version. The delivery condition is of particular relevance to the automation of processes. “Bin picking” to this day is still a challenge for mechatronic systems. Products can be delivered in bulk, sorted and packed, in magazines, as flowables, or taped. Among the handling characteristics, several aspects can be summarized. When components get handled, they interact with a gripper. The ensuing forces must not overload the device. Furthermore, areas need to be defined in which the gripper can hold the component by force, form, or material fit. Other conditions, such as the weight of a component, sharp-edged geometries, temperature, or safety to health of the components are also analyzed in terms of the handling characteristics.

15

Chapter 2 The joining direction is best derived from a graphical representation of the product structure. The product structure is also considered in the context of process analysis in detail to derive the assembly priorities. The joining method has a major impact on the resources. Whether assembling, attachment or pressing, primary forming, forming, welding, soldering, gluing, or filling, the components are laid out here according to the derived requirements. The quality requirements are represented by quality attributes which set conditions for a manual or automated assembly. This may be, for example, the purity of surfaces or the observance of tolerances.

Process Analysis The process analysis, which follows the product analysis, considers the assembly process in the assembly system and its functions. To identify the assembly process, the assembly priorities must be determined; this results from the product structure and the resulting joining situation. The product structure has already been used with reference to a graphic representation within the product analysis. Through the visual representation of the components in detail and the target status, the joining situation can be determined. The joining situation includes all conditions under which a component is assembled. This results in an assembly priority chart of predecessor and successor relationships, under which adherence the components can be assembled to form the product. This may also imply that joining processes are combined to pre-assembly processes for sub-modules. The assembly process can take place piecewise or in sets. In the piecewise assembly only one product of a lot is assembled at once and started only after completion of the previous product. For the assembly in sets, the work is first loaded with a defined set of identical components and the product is then assembled in parallel in one sequence. Within a set, the assembly processes repeat in this manner and can be more economical depending on the available product base area and the volume of production. Furthermore, in an assembly planning system where the component support structure is to be analyzed as well, the component support is the interface from the product to the assembly system. The assembly priority chart consists of the individual assembly processes, which can be derived from the function sequences. A function sequence typically consists of the operations sorting, supplying, joining, and testing. Since these processes, both manual as well as automated versions, can get aggregated with a time, a first estimation of a time regime can be part of the functional analysis. As part of the process analysis former planning results of earlier planning processes can be included.

16

Skills-Based Human-Robot Cooperation Generally, it should be noted that overproduction, waiting, transportation, inventory, transport, and production defects should be kept to a minimum.

Resources Analysis The analysis of the resources can be divided into three stages: • Working environment • Working place • Human/Technical resources Impact on the work environment includes noise, mechanical vibration, radiation, climate, lighting, and materials. The workplace itself must be designed for the realization of HRC according to ergonomic criteria. An assembly station that is adapted to a spatial concentration on the joining area, is one such example. Humans are generally analyzed in terms of capability, skills, and motivation. Capability is the overarching concept that enables a person to acquire a skill. Movement of the finger, for example, is a requirement for learning to play the piano. Motivation is another criterion that must be analyzed. While the analyzation of skills can be transferred to technical systems, the motivation remains exclusively human. The information regarding the strengths and weaknesses of human or robot lead to the objective that each one carries out a work task that is best suited for each. In this review, the robot is always considered as a complete machine, including the appropriate tool for the task. From the technical specifications of the complete robot system the skills of the mechatronic system can be derived. While the analysis of the technical characteristics of mechatronic systems is well suited to determine the skills of the system, the survey of human skills is challenging. Most methods for assessing human skills originate in the operational integration management. In Germany, cognitive and anesthekinesic skills are essential for the allocation of an assembly working place. Those are normally provided by the employees in the form of qualification sheets and skills profiles. But these can’t give an integral view on the complete skills of the human. For employees with disabilities, the skills assessment includes in addition to the above mentioned some medically coined criteria. For a decision regarding HRC, an integral acquisition of the human skills is needed, so that the procedure described here draws on work from integration management. Thus, first a self-assessment is done via a Work Ability Index, which can give an initial general statement about the working skills of a person. Complementary to the self-assessment, an external assessment with respect to work-related skills is performed by a company doctor or occupational health professional.

17

Chapter 2

 Shown is one example for an approach of a skills-based task assignment, considering the working environment and working place.

 This graphic depicts the comparison of the skills that take place against the background of the technology. Per character there is an assessment of the criteria performed against the background of the technology features, whether a human or a robot is better suited for the task execution.

Skills-Based Task Assignment Here, the analytical results of the previous sections are processed to reach a decision regarding the allocation of tasks. The decision on the allocation of tasks is based on a profile comparison. Within the comparison, the profile of the workplace is checked against the profiles of the available resources. The comparison of the skills takes place against the background of the technology features cycle time, additional investment, process safety, and quality of work. Thereafter, for example, the assembly operation “insertion of a component” is described by the criterion “joining movement,” which has the character “by running

18

Skills-Based Human-Robot Cooperation a joining tool.” Per character there is an assessment of the criteria performed against the background of the technology features, whether a human or a robot is better suited for the task execution. The criteria may be multiplied by a weighting factor to emphasize certain activities.

 This image illustrates how individual components are assembled into sub-assemblies, which then form the sections and finally an aircraft fuselage.

 As a validation environment for the skills-based approach to the assignment of tasks, a demonstration site was set up at ZeMA (Centre for Mechatronics and Automation GmbH).

The assessment provides the basis for a decision on the allocation of tasks. When a decision has been taken, appropriate work instructions or robot programs can be written. For the implementation of a flexible use of resources, the use of assistance systems that accelerate the reconfiguration of assembly systems is recommended.

19

Chapter 2

Demonstration Scenario Results This section describes the application of the proposed description model on the assembly of fuselage skin sections within the manufacturing of airplanes.

Product Analysis The product “fuselage shell element” consists of four main product groups and other components. The first component is the shell skin. A real fuselage shell element used as reference has the dimension of 2 x 4 m. For this demonstration scenario, a smaller model of this shell element was used. The shell skin is made of carbon fiber reinforced plastic (CFRP) rovings. The skin is curved and, due to the matrix that surrounds the CFRP rovings, has an average rigidity. In this way, a shape tolerance is possible and must be considered as part of an assembly process. The surface of the skin may contain residues of lubricants and dirt from storage. The weight of the skin is about 24 kg.

 This actual reference fuselage shell element had dimensions of 2 x 4 m. For the demonstration scenario, a smaller model of this shell element was used. Next, the stringer is to be considered. This is also constructed of CFRP rovings, which are embedded in a matrix. The geometry of this is already relatively torsionresistant. Along the longitudinal direction, the rigidity is medium, so that for large lengths, a deflection due to its own weight is possible, if it is not adequately supported during handling. The stringer has a length of 4 m, a width of 0.08 m, and a height of 0.05 m in the demonstration scenario. It weighs 2160 g. For the surfaces, the same applies as for the skin. The clip is made of organic sheet and a rectangular CFRP component of 0.1 m length, 0.04 m width, and 0.06 m in height. It weighs 30 g and is of high rigidity. For the surfaces, the same applies as for the skin.

20

Skills-Based Human-Robot Cooperation

 A fuselage segment of the demonstration scenario consists of a skin, three stringers, four clips and two frames. Shown are the product and its components. The frame is made of CFRP rovings, which has corresponding recesses for the stringers. It is the component that gives the curvature to the shell skin and is made accordingly. The frame has a high rigidity due to the geometry and is 2.5 m long, 0.05 m wide, and 0.1 m high. Each frame overlaps the skin with 0.5 m in the final assembly to connect the fuselage shell elements and form a barrel. It weighs 1125 g. For the surfaces, the same applies as for the skin. All components are stored in a supply shelf. Regarding the handling properties, it must be ensured that the components are not structurally overloaded during handling, either with static or dynamic forces. The gripping points are determined by the geometry of the component and its structure. Recesses need to be considered when choosing the gripping points. The joining direction of all the components is from above. Only the frame is also fitted transversely to the clips. Adhesive bonding is used for joining of all components. Gluing requires an activation for an optimal adhesion through the generation of free radicals on the CFRP surface of the components. The purity of the surfaces is therefore an essential quality criterion. A fuselage segment of the demonstration scenario consists of a skin, three stringers, four clips, and two frames.

Process Analysis As part of the process analysis, the product’s structure was determined in the first step based on a technical drawing. Accordingly, the shell skin was first inserted into a kinematic fixture that adapted itself to the geometry of the future shell element. Next, the stringers were assembled. The geometric arrangement of the stringers determines where the clips are placed. Only when the clips are assembled, can the frame elements be added to the shell and clips.

21

Chapter 2

 As part of the process analysis, the structure of the fuselage shell element was determined based on technical drawing in the first step.

 Result of the task assignment for the demonstration site. After evaluating all four basic processes, the assignment of tasks arises, regarding the application scenarios. The joining situation for the individual components was derived from the information of the product structure that is not critical at the small dimensions of the demonstration product. For larger components it is, with respect to the adhesive joining process, relevant that the intended position and orientation over the entire contact surface away is reached directly. Subsequent corrections could lead to scattering of the adhesive on the surface and no uniform thickness in the contact zone. As recommended, an assistant support with the help of laser projectors was used. Next, the subtasks were determined, and based on those an assembly priority chart was created and an initial assessment of the cycle times was carried out.

22

Skills-Based Human-Robot Cooperation

Resources Analysis For working environments within German factories, there are guidelines that define which way work environments have to be laid out in terms of noise, mechanical vibration, radiation, climate, lighting, and materials. These requirements were part of the evaluation of the demonstrator at ZeMA, but not critical under the condition of a research institute. One aspect that can be reproduced very realistically, even under research conditions, was the ergonomic design of the workplace. For this purpose, a software tool was used that reproduces the assembly cell in a virtual environment and evaluates the ergonomics according to the current state of the work sciences. As a mechatronic resource, a lightweight UR10 robot from Universal Robots was used in the assembly cell. It has six axes and can rotate all joints ±360°. Foot and shoulder joints can have a speed of 120°/s and elbow and wrist joints of 180°/s. Points are approached with a repeatability of ±0.1 mm and the reach radius is 1300 mm. Due to the path-dependent velocity and acceleration profiles, the execution times of assembly operations can only be estimated. A good guideline provides simulation results of virtual reproductions of assembly processes. If these are not available the experience of similar processes can be used. In the scenario described here, the assembly cell was virtually rebuilt in a simulation environment and the robot driven directly from the simulation environment.

 This image depicts the actual demonstrator site in HRC-operation at ZeMA.

23

Chapter 2 The skills of humans were determined based on a self and an external assessment based on a variety of relevant features, as described earlier. As the self- and external assessment were not congruent, a variance analysis was carried out so that features that varied considerably in self- and external assessment were analyzed again. Other features that can only be assessed by experts as fine anesthekinesic skills, reaction rate, etc., were collected exclusively by the external assessment. After evaluating all four basic processes, the assignment of tasks was applied to the application scenarios. A process type classified as “A” results in a synchronized operating mode, in which the human first activates the surface of the component’s skin with a plasma torch, the robot applies adhesive, and the human manually joins the components. The curing, inspection, and documentation is again performed by the robot. In process type “B,” a higher proportion of automated components is realized at the expense of processing time. Based on the required set-up time for the robot, the process of adhesive application can be carried out here after the plasma activation of the component. It was shown that HRC processes can be enabled through structured planning, without having to rely solely on “gut feeling.” The thematic area that needs to be considered for HRC assembly planning is still very spacious, and probably the main reason for the high number of HRC processes that still get implemented solely by someone’s gut feeling. Whether planning in the future can be as flexible as the applications of the new lightweight robot generation allows for depends crucially on how well planning methods can be provided that support a suitable, easy-to-use HRC method. Thus, it will be essential to simplify the planning methods to such an extent that they represent a sufficiently accurate, reliable, and impartial as possible planning process. This chapter is based on and imagery is taken from SAE International Technical Paper "Empowering of Assembly Processes for Human-Robot-Cooperation in Terms of Task Assignment,” by Mueller, R., Vette, M., and Mailahn, O., doi:10.4271/2016-01-2093.

24

Chapter 3

Modular Gripper for Flexible Assembly Operations Marc Fette CTC GmbH

Kim Schwake and Jens Wulfsberg HSU/UniBw Hamburg

Frank Neuhaus Airbus UK

Manila Brandt Airbus Operations GmbH Aircraft production is in the throes of an interesting challenge. A quick look at the current orders from Airbus and Boeing reveals 12,000 aircraft deliveries in the coming years. Every six and a half hours a new aircraft must be completed. They must deliver those aircraft on time and make money while doing so. New and innovative technologies are the key to success throughout the entire aerospace supply chain; however, challenges are ever present. For example, complex assembly processes lead to a relatively low level of automation. Furthermore, most of the current automation solutions in the aviation industry are realized by special

25

Chapter 3 machine engineering, which is associated with high investment costs and low flexibility of the production systems. In this context, the use of intelligent and networked industrial robots is a very promising approach for high-performance, flexible, and complex assembly operations.

Challenges of Major Component Assembly An increasing trend among civil aircraft manufacturers is to offer customers different versions of an aircraft type. Usually the cockpit, tail, and wing sections remain the same within the product family, but the diameter and length of the fuselage may change to increase or decrease the maximum number of passengers. Such a scenario necessitates a flexible production system to process the different fuselages. Each aircraft type has its own assembly requirements. Status quo for production planning is every aircraft type gets a unique manufacturing line with unique assembly processes. Aircraft assembly machines are custom-designed to meet the specific requirements based on the tight tolerances defined by the aircraft engineers. As a result, the degree of standardization is low, increasing the number of other devices necessary for completing the process, such as customized and expensive jigs and tools that are designed for a specific aircraft type. For every production line, there is a limited number of spare parts or replacement systems. If one system fails, it is a major issue for the available capacity of production. Thus, adapting one production system to assume tasks for production of other aircraft variants is simply not practical.

 As passenger projections continue to rise throughout the world, the leading aerospace companies become involved in the development of new, adaptable and modular assembly systems.

26

Modular Gripper for Flexible Assembly Operations Further difficulty arises if the aircraft is subjected to design changes or modifications, particularly in the early phases of the life cycle, or, as in the case of the Airbus A320neo, even in the middle of the life cycle. There is an overall lack of flexibility within current production systems to fulfill the upcoming requirements for competitive production in the aircraft industry. Standardization helps to increase flexibility, and flexible systems lead to robust productions. Commercial aircraft structural parts generally have a length of 2-20 m. As in many fields of aircraft production, the quality is relevant to product safety, and especially in major component assembly, the required close tolerances need to be met. The following is an example regarding the positioning, aligning, and shaping of shell parts in preparation for a joining process by either riveting or some other future joining process. In large-part assembly, shape and position tolerances of ±2 mm are acceptable. Quality control while processing and after end of work is mandatory for increasing assembly efficiency, product safety, and documentation. An aircraft section in this example is subdivided into a four-shell part division. Even though stringers and frames are integrated for supporting the structure, the shell parts cannot be described as inherently stiff. For manufacturing reasons, the shell part has to be rotated after allocation at the station, which leads to three factors as main causes for necessary adjustments during assembly: random deviations of nominal dimensions, effects of gravity, and handling operations during transportation. Aside from those factors, newer materials such as carbon fiber reinforced plastics (CFRPs) that are on the rise in the aircraft industry are more difficult to process than prevalent aluminum because they are highly sensitive to any force inductions. The process consists of handling large shell parts for setup, shaping for fitting to geometric appearance, and positioning of shell parts to each other in preparation of the joining. For the joining process, the necessary mounting holes cannot be precast. After the shaping process, the drilling process occurs, which ensures congruence with the mounting holes without strain. Then, after the drilling operation, the parts are separated to provide work space for cleaning the shell parts and applying sealant. To finish the joining process, the shell parts must be returned to the exact position from drilling operation, which results in a yet another positioning operation. The complexity of the assembly system requires a rigid steel construction and a space-consuming layout. Its use is restricted as it concentrates on one aircraft type and a limited number of variants. This scenario undermines the need for quick reactions to changing or emerging needs.

27

Chapter 3

 The complexity of the assembly system—in this case a current major component assembly station for an Airbus A350XWB at Premium Aerotec in Nordenham, Germany—requires a rigid steel construction and space-consuming layout. Due to unpredictable economic or social incidences, customer demand can suddenly change. The ability to react to these events with the use of scalable production resources is crucial to aircraft manufacturers and will influence the assembly process chain.

Flexible Aircraft Assembly Concept Flexibility and cost reduction are the ultimate goals in production and manufacture. This is one reason why such great value is placed on commercial-of-the-shelf (COTS) solutions. COTS are quickly available and most of them have a high level of maturity—allowing for independence from certain system suppliers—and are comparatively cheap. It is of particular interest in the aircraft industry because such components can shorten a long certification process if they are already used and certified in another application. Looking toward future production, standard six-degrees-of-freedom industrial robots are key. There are several robots from different manufacturers available on the market with appropriate load capacity and range. The handling task consists of taking up the shell part and the provision of the shell part in the starting position for the process steps. For fixation of the shell part, a system based on vacuum cups is used. This principle is used in many fields of industry and has also proved its worth in the aircraft sector. The vacuum cups are mounted on a special end-effector.

28

Modular Gripper for Flexible Assembly Operations The shaping operation is performed with aid of the special end-effector. It consists of a pre-curved framework; however, it provides a high flexibility for the adaptability to other geometries due to its actuating elements. With respect to the shaping operation, linear actuators are installed for force induction into the shell part at defined points on the frame to manipulate the geometry of the shell part. The linear actuators can perform a one-dimensional motion in the direction of the barrel center. To cling to the curved work pieces, vacuum cups with ball joints are equipped. The positioning operation is also implemented by the industrial robots. The previously clung and shaped work pieces are placed in the correct position with aid of calculated coordinates based on the actual coordinates measured with an appropriate measurement instrument. In contrast to the clinging of the shell parts where the motion is taught, it can be preset. To achieve required accuracy, an iterative procedure for positioning and shaping is needed. After the drilling operation and separation of shell parts for cleaning, the previously saved position can easily be approached again so that the parts are in their original position after cleaning. At this point, the advantage of the separation of functions is apparent because the shell part maintains in its shape. Since no re-shaping takes place in this scenario compared to the current process, a process step is eliminated that consequently results in significantly shorter process times. Additionally, there are advantages by lowering stress on the shell part while shaping and by reducing the risk of damage to the shell part with unnecessary complex movements. For these reasons, a higher accuracy is to be expected. The system is designed for the processing of shell parts of different materials. The built-in sensor technology is particularly suitable for CFRP. During processing, it has to be taken into account that CFRP is quite sensitive in contrast to aluminum. Manipulation exerts external forces on the shell part; without knowledge of the absolute value of the forces, the security against material failure is not ensured. Because it is not possible yet, at least not for production-wide application, to equip the shell parts with strain gauges for force control, force measurement must be done externally as close as possible to the shell part. For this reason, force sensors are attached to the vacuum cups. Continuous monitoring detects dangerous stress situations early. The manipulation process is automatically stopped in such cases. In addition, the force sensors can also be used to document abnormalities of the process. Impact loads as a result of collision can be detected and provide an indication if damages can be presumed. It is also conceivable that the process flow can be monitored and average process times calculated by analyzing characteristic vibration patterns to deduce drilling operations.

29

Chapter 3 As the designed assembly system is configured with a decentralized control, an external absolute measurement system must be used. It provides absolute coordinates for the calculation of the shell shape and position, and for assembly system setup. This system must be highly accurate over large distances. Laser trackers are provided in this concept as a global measurement instrument. The advantage of this measurement technology and method is the existing application within production. This provides a high knowledge base and robust processes. The assembly process begins before the first shell part arrives at the assembly station. When the shell parts are clocked, their actual shape and position are measured-in based on defined reference features to identify deviations from the target shape and position as well as to identify the orientation and position about the assembly station and aircraft date. The recorded data is used to calculate manipulation operations. During the process, the coordinates are iteratively re-recorded to correct any deviations. A higher degree of automation compared to the current system enables a reduction in quality variance. At the same time, a leaner process can be generated by eliminating unnecessary process steps, which leads to lower cycle times. Nevertheless, it can be assumed that joining operations will continue to be conducted manually.

Laboratory Setup To understand the new assembly technology and evaluate its potential, an experimental laboratory setup was composed. For this purpose, a system that is scaled and limited to its functional elements was sufficient. For practical reasons, one industrial robot (KUKA KR125/2 with a load capacity of up to 125 kg and an accuracy of ±0.2 mm) was used instead of several robots. Consequently, a smaller shell part was used that was also adapted in its rigidity and comparable to a full-scaled shell part. Likewise, the end-effector framework was scaled and consisted of standardized profiles. The frame construction can be changed quickly and easily due to the selected profile system previously proven in many experimental setups. The first tests showed that a planar arrangement of the linear actuators has tight limits in respect of the curvature of the shell part. A second framework was built with a precurved layout. Vacuum cups are also installed on ball joints to allow the adaption to the shell part surface and subsequently a safe gripping the shell part in the laboratory test. Directly behind the vacuum cups the force sensors were attached. These are piezoelectric 3D measuring sensors used to determine tractive, compressive, and shear forces. Although these sensors are especially useful for particularly stiff

30

Modular Gripper for Flexible Assembly Operations constructions with very dynamic loads, they can easily be applied in this experimental setup. In the current experimental setup, the primary measurement device for absolute coordinates is substituted by a measuring arm with sufficient range and comparable accuracy. A Romer Absolute Arm 7545SI with a probing volumetric accuracy of ±0.119 mm is used. For the active motion process during positioning, the absolute encoders of the robot are used. For the active shaping process, the linear actuator’s built-in incremental encoder is used. The relevant absolute coordinate measuring features are attached at the end effector and distributed over the shell part. This results in process-relevant measurement points that are used during the active process and checkpoints that document the final shape and final position after the process. They are also used for quality assurance. The processing of the measured values is provided by a software application. It processes the calculations of the necessary shaping and positioning operations considering the tolerance-loaded work pieces by compensation and optimization calculations. The complex calculations are realized with help of MathWorks’ MATLAB. In the conceptual design of the software system, emphasis was placed on modularity and easy extensibility. The software created in the .NET programming environment can be easily adapted to changes in the end-effector design. Extensions to the number of end effectors or changes in frame geometry are carried out quickly. The goal of the experimental scenario is a shape and position deviation of no more than ±2 mm. This tolerance interval is typical for the large component assembly. Of course, the test shell part with its approximate length of 2 m is at the bottom of the major component assembly spectrum since the lab setup is only intended to demonstrate the feasibility of the technical principle. If using more accurate components, especially the linear actuators, closer tolerance intervals are possible. The process flow of the experimental scenario begins with the grabbing of the shell part from a shell tray. The shell is not clamped and thus the shell part is undefined in shape and position so the grabbing is not defined. This leads to different initial conditions for each test cycle. The robot then moves in a taught-in position which allows for the calibration of the system components. First the position and orientation of the end effector is measured. In the experiment, it must be assumed that the shell part does not meet the specification for size and shape in its delivery state. Therefore, the actual shape, position, and orientation are taken from reference points. This generates a model of the shell part in the calculation software.

31

Chapter 3 The end effector is limited by its linear actuators to a shaping by one-dimensional traversing cylinder. This must be considered in the shaping strategy because it is possible that intolerable shape deviations in non-influenceable directions quickly occur by one-dimensional manipulations. That is why attention must be paid to optimum setting parameters with the smallest traverse distance. In the calculation of the output parameters of the linear actuators, the previously geometric uncertainty of the shell part can be used to generate an optimal work piece coordinate system for traversing the shell part. After the first shaping cycle is complete, the result is checked against the data points. Then, a second shaping cycle is initiated, if necessary. This situation can be caused by deformations of the end effector or by inaccuracies of the linear actuators. If the measurement points are in the respective range of tolerance, the shaping process is checked with help of the checkpoints and lead back to the geometric shape of the test shell part. Subsequently, the test shell part is positioned between two rigidly suspended adjacent shell parts which simulates a longitudinal joint. The necessary motion parameters are calculated and then sent to the robot controller, thereby minimizing the gap between the joint partners. Thereafter the shell part can be attached and joined. During the process, the force sensors constantly monitor the work piece condition. When defined force overruns, or unanticipated stress states occur, the process stops and alerts the user about the problem. Thus, the work piece can be protected from critical forces. The process time and process-relevant setting parameters are recorded in the background and documented after the process. For the first tests, there was a dummy shell part with undocumented demands. Since the maximum angle of the ball joints on the linear actuators are quickly exceeded because of the planar structure of the first end effector version, the realizable diameters are quite large. The second prototype must be pre-curved. This also improves the possibilities of shaping with one-dimensional linear actuators that are not in one plane, but are aligned concentrically. For the next shell part, which was more realistic with specific requirements concerning shape and position, the second end effector version with pre-curved design was built. The analysis of the shape and position accuracy was realized via the checkpoints distributed on the shell part. In this case, an ideal shell part, deposited as model in the program, served as a reference. The deviations of the checkpoints to the ideal values were calculated, while in the evaluation of the result the edge areas have a higher relevance regarding the following joint operations. Among the geometric

32

Modular Gripper for Flexible Assembly Operations data, the force curves for the shell part were recorded and compliance with force limits were monitored. The close tolerances of large component assembly from ±2mm could be met in terms of shape and positioning accuracy after a few iteration cycles. Studies revealed that the default connection profiles of the framework do not provide high rigidity and are elastically deformed and visible during the shaping process. The fact that it is still possible to meet the tolerances is based on the absolute measuring device and the iterative procedure. Deviations from desired shape can be accurately determined because the measured values are independent of the end-effector and thus the deformed framework. Any inaccuracies while shaping can be compensated for in the following iterative designed process. In this way, deformation of the end effector as well as inaccuracy of the components can be eliminated. The separation of the function with an industrial robot for positioning and shaping with the end effector proved successful. Thus, the process is reduced in its complexity and it is leaner. The robot can compensate for the deviations of orientation of the shell part. The limitation to one degree of freedom for shaping in combination with the standardized industrial robot and its technically mature control reduces the control effort compared to the present process significantly. For potential savings, the independence from hall cranes, as well as elimination of unnecessary process steps which were necessary due to a complex kinematic with simultaneous shaping and positioning operations can be identified. With the help of the force sensors, material critical loads could be prevented and process abnormalities like collisions could be detected. It is also conceivable that possible damages from manufacturing of the work piece material can identified by analysis of force curves. During design, much attention was paid to guarantee that the system provides high flexibility. The flexibility extends to various types of aircraft (force application points for shaping, body diameter, and fuselage lengths), different aircraft variants (length of the shell parts), substitutability of components (independence of single suppliers, availability of system components), and expandability/scalability (modularity of the system in terms of hardware and software). As in the field of machine tools, such universally oriented machines often have disadvantages compared to special machines. Lower accuracy and longer lead times are typical symptoms. These disadvantages are observed especially in mass production. In the present case; however, it is a quasi-unique production and the benefits of such a universal machine are apparent.

33

Chapter 3 For positioning and handling, an off-the-shelf industrial robot has proven advantageous. The application is special because in this case the robot is not used for automated production with repetition-intensive activities, but as a universal handling tool where the human and robot occupy the same workspace. For now, it is established that the human is outside the danger zone while the robot is in motion. However, since the required speeds of the robot falls in an extremely low range, with the appropriate safety measures, the human could remain in the work area. The low velocity does not lead to process-technical disadvantages because the proportion of time for robot movement while positioning and handling compared to the total process time is negligible. General disadvantages of industrial robots to special kinematics such as lower accuracy and lower stiffness are not relevant due to the process-related iteration loops. This contrasts with large technical and economic advantages in terms of flexibility. Industrial robots are equipped with matured control with defined interfaces. The procedure is robust and accurate. Due to the large spread of industrial robots, the number of possible operators is large. Thus, the maintainability and extensibility are high. Simultaneously, the anticipated costs for adjustments and the dependence of specific system suppliers decrease. In addition to the hardware-technical flexibility, the system provides a high degree of flexibility when considering software. The software has a modular design and with open standards. The MATLAB calculations facilitate the processing of shell parts with varying quality. The also enable handling of the shell parts which were inaccurately allocated with help of the initial measuring of relative position and orientation of the shell part to the end effector and following calculations of deviations. This fact is advantageous and supports flexibility because costly and component-specific jigs can be eliminated (under certain circumstances). The limited load capacity and stiffness of the robot requires the use of several robots including several end-effectors. This increases the overall system and makes it more complex. However, the advantage of a plurality of robots is the continuous support of the shell part over the entire length of the component. In contrast to the currently used system, it is possible to better keep the shape of the shell part than in the case of a wide support due to a small number of support towers. It has been shown that the industrial robot is also useful in large component assembly and can bring about improvements. The economic benefits include a leaner process, the cost advantages for components, and maintenance. The missing stiffness of the framework may be a problem for a later investigation. The first matter is the increased amount of iterations of shaping operations to compensate the deviations arising from the framework deformation in the

34

Modular Gripper for Flexible Assembly Operations background. This leads to inefficient processes and high lead times. Looking a step further, the scalability of the concept with the aim to process very large shell parts could be very difficult by using standard aluminum profiles. On the other hand, the use of conventional welded steel jigs (which is quite common in the aircraft industry) in impractical due to weight and robot capacity.

 Researchers used a demonstrator of an assembly system with a scaled-down test shell part to achieve a deeper knowledge about the new assembly technology and to evaluate its potential and limits.

Modular, Lightweight Gripper System The innovative lightweight gripper and assembly system is made of CFRP and has been developed along a methodological development approach. The system is called the X-profile system. It consists of complex, pultruded profiles made of CFRP and flexible, exactly positionable adapter clamps made of composite materials with a thermoset matrix. The motivation for developing this gripper system was the substitution of components of the so-called Euro-Greifer-Tooling (EGT), a lightweight gripper system

35

Chapter 3 for industrial robots in the automotive industry. The EGT kit consists of different octagonal and round aluminum carrier profiles, load attachment plates, and several aluminum or steel connector clamps. By using the new CRFP gripper system, a significant reduction of the total weight can be achieved with integration into the existing EGT system. A reduction of the loads, the use of the next-size-smaller industrial robots, and higher acceleration and speeds result in higher flexibility and increased production rates.

 The motivation for developing the gripper system was the substitution of Euro-Greifer-Tooling (EGT), a lightweight gripper system for industrial robots in the automotive industry. The CFRP carrier profile has a total mass of about 1.8 kg/m and a deflection of 5.74 mm if it is fixed at one side, with a length of one meter and applying a point load of 100 kg at the other end. In this case, the most unfavorable bending stress was observed. In addition to the bending characteristics of the CRFP carrier profile, torsional loads and other specifications are simulated by finite element method (FEM) and investigated by experimental testing. The carrier profiles form in combination with various adapter clamps, such as corner or cross connectors, a universally mountable, modular assembly and gripper system for high-performance applications. The mass of the adapter clamps is between 0.18 kg and 0.43 kg, depending on the type of material. Through the future use of this modular gripper system for aircraft assembly, the total weight of the gripper or mounted system components can be kept relatively low with a high level of mechanical properties. The resulting lightweight potential is necessary to scale an adaptable and modular production and

36

Modular Gripper for Flexible Assembly Operations assembly system based on the networked interaction of flexible industrial robots, especially for large-scale structures in the aircraft industry.

(a)

(b)  In addition to the bending characteristics (a) of the carbon fiber reinforced plastic (CFRP) carrier profile, torsional loads (b) and other specifications are simulated by finite element method (FEM) and investigated by experimental testing.

Furthermore, the weight reduction can prevent the application of heavy-duty robots. This advantage, but also the possibility of using the next smallest industrial robot class, obtains lower investment costs and a holistic increased productivity due to accelerated motion. In addition, energy, maintenance, and failure-related costs due to higher loads can be reduced by using the CFRP gripper system. Increased flexibility due to a higher reach of industrial robots, the flexible design of production lines, and more functional gripper systems is a further advantage. Due to the modularity,

37

Chapter 3 scalability, and high adaptability of the new composite gripper and assembly system, high-performance applications for industrial robotics and the realization of flexible handling systems in the aircraft production are possible.

Transfer to Aircraft Production Applications For the construction of the first laboratory demonstrator for the robot-based positioning and shape correction system, a low-cost solution based on framework made of aluminum profiles was implemented. The simple aluminum profile system is within the concept validation because it allows for optimizations or entirely new configurations in a cost efficient and expedient manner compared to welded constructions. The high freedom of design and a wide range of standard connectors of the modular system have proved successful. However, the simple profile system is not designed for high-performance applications with strict rigidity and accuracy requirements. Low mechanical properties result in process-related disadvantages with this standard system. Consequently, it comes to significant elastic deformations. By using an independent absolute measuring system, the inaccuracies caused by deformations can be compensated. In terms of industrial use, these elasticities lead to cycle time extensions due to the implementation of several iteration loops. This is incompatible regarding highly efficient production processes. The use of the new X-profile promises an improvement of the efficiency of the current test setup.

 An early demonstrator consists of the X-profile system, a possible assembly system of a robot-controlled device for more efficient production.

38

Modular Gripper for Flexible Assembly Operations Regarding modern aircraft production, changing requirements and customer demands will increasingly cause product and production changes. Product life cycles are being shortened in the aircraft industry accordingly. The aerospace industry is subjected to constant changes even if the products have still long lifecycles. Lean production and a higher level of automated production are among the highest priorities of the aviation industry. Therefore, it is necessary to implement flexible and reconfigurable production systems. The devices currently used cannot satisfy these demands. Furthermore, there is a need for many handling and transportation devices in due to the distributed production. Jigs and fixtures are of great importance. Nowadays, rigidly welded, inflexible steel structures with high weight are dominating the assembly lines. The advancement of the lightweight gripper and assembly system made of CFRP for use in the aviation industry can contribute to significant cost reductions. The reduced weight in comparison to metal constructions can help to ensure that standard production devices such as industrial robots can be brought into application. Regarding the process-oriented use of assembly systems, current production strategies have to be changed due to the rising demand on civil aircraft. Manufacturers must be able to react to market changes in the short term, in light of the increasing volatility of the aircraft industry: without production adaptability and flexibility this is not possible. Production systems that can be reconfigured quickly and easily are necessary to meet this goal. Besides the flexibility and freedom of design, the gripper and assembly system must provide robustness. The manufacturing and assembly process in the aviation industry is often characterized by large-scale components or structure with no inherent rigidity. The modular X-profile system must prove its suitability for the construction of large self-supporting assembly systems, especially those whose assembly results are influenced by gravity. In addition, there are disturbances of the production process by excited oscillations or pulses that need to be partially compensated for by the profile system and its structural design. The damping behavior of the lightweight assembly system made of CFRP and the connection elements are of great importance for the usability and efficiency. Due to the flexibility and the cost efficiency of the whole assembly kit, it is also important to keep the number of elements to a minimum. This chapter is based on and imagery is taken from SAE International Technical Paper "Use of an Innovative Modular Gripper System for Flexible Aircraft Assembly Operations," by Fette, M., Schwake, K., Wulfsberg, J., Neuhaus, F. et al., doi:10.4271/2016-01-2108.

39

Chapter 4

Embracing Reconfigurable Assembly Cells for Aerospace Thomas G. Jefferson, Richard Crossley, Anthony Smith, and Svetan Ratchev University of Nottingham An unprecedented number of orders for commercial aircraft and the impact of climate change has placed great urgency on aerospace manufacturing to make gains in efficiency throughout aircraft production and operation. Current manufacturing systems are using technologies and production methods unsuited to a dynamic future market. New techniques are required to bring about quick time-to-market to fill niches and introduce more efficient technologies. Given the likelihood of change, the aerospace industry must seize the opportunity to innovate and readdress approaches to aerospace manufacture to ensure its prosperity. Assembly is one such area where cycle times are in the order of hundreds of hours and commissioning assembly cells takes up to 24 months. Assembly is cited to be 15-70% of total manufacturing time and assembly costs are typically 25-50% of the total cost of manufacturing. These systems are not performing at their optimum and can be replaced with new technologies and methods. Adopting new strategies for assembly could lead to substantial gains for aerospace.

41

Chapter 4 Reconfigurable Assembly Systems (RASs) are a solution to increase manufacturing agility. An RAS for automotive body-in-white (BiW) production shows costs reductions between 5-14% compared to separate lines. The efforts for planning, engineering, installation, and production start-up are between 50-80% less compared to individual production lines. Another example reports cost reduction of up to 50% in development and operation. An RAS is designed at the outset for rapid adjustment of its production capacity and functionality, with the aim of satisfying the requirements of the market through change or rearrangement of its components. An RAS embodies up to six characteristics with customization and scalability considered as essential: • Customization (flexibility limited to part family) • Convertibility (design for functionality changes) • Scalability (design for capacity changes) • Modularity (components are modular) • Integrability (interfaces for rapid integration) • Diagnosability (design for easy diagnostics) Increasing the commonality between assemblies is yet another solution. Wellexecuted commonality strategies can produce savings of 15-50%. Firms have cut costs by 30% and reduced lead times by 50% by employing commonality. Commonality is a key enabler to reconfigurability and both are employed simultaneously in automotive development. There are efforts in aerospace to apply reconfigurable techniques to reduce operating costs and time-to-market. Despite the apparent benefits to aerospace and the reported savings in automotive, a great deal of research is needed to suit the specific challenges of aerospace assembly. The research aims to advance development of the next generation of aircraft assembly systems. This chapter presents novel contributions toward RASs for aerospace assembly. A framework is specified for integrated design of a wing family and RAS. A new application of a novel assembly method using metrology-enhanced robotics is then defined for hole-to-hole assembly. An RAS demonstrator cell is presented with early work and findings. This research sets a precedent for the future of aerospace, which must adopt a more agile approach to assembly to cope with modern demands.

State-of-the-Art RAS Review This section reviews research on reconfigurable assembly and product family design to identify the latest approaches and technical state-of-the-art that will be adapted and applied to aerospace. Discussion of practical examples takes precedence where possible to bear most relevance to industrial implementation.

42

Embracing Reconfigurable Assembly Cells for Aerospace

Product Family Design A product family is defined as “a set of common components, modules, or parts from which a stream of derivative products can be efficiently developed and launched.” Many companies utilize product families to capture larger segments of the market, reduce costs, and shorten time-to-market. Product families are particularly common in electronic consumer products such as laptops or smartphones. The automotive industry also uses product families extensively to reduce manufacturing costs and increase customer satisfaction by offering an extensive range of customization options. A recent example is the Volkswagen Group MQB platform where most components are shared across Skoda, Volkswagen, Audi, etc. Such commonality drastically reduces tooling costs and production time. Scale-based platforms are evident in the aerospace industry but not to the extent seen in automotive. Notable examples include Lockheed Martin military transports and some Rolls-Royce engines. Embraer exploited scaling and commonality to reduce development and production costs. The 170 and 175 models have 95% commonality among subsystems as do the 190 and 195 models. The extent to which the assembly system has been considered is unclear, however.

 Commonality has increased with each new generation of vehicles under the Volkswagen Group. The majority of subassemblies are identical regardless model and marque. The left image depicts the prior situation where models had little commonality. The right image depicts the revised arrangement where all marques share a common platform. The common platform means assembly processes are shared and so is much of the tooling. Corresponding reconfigurable assembly systems (RASs) designed from the outset are capable of producing all models according to demand. Volkswagen Group are continuing to achieve cost savings in the order of 30% and lead time reductions in the order of 50%.

43

Chapter 4

Design Methods A robust, systematic method is fundamental to product family design. Equally, an RAS should be systematically designed from the outset to embody reconfigurable characteristics. Commonality between products simplifies the assembly system and so there is mutual benefit in concurrent part family design and RAS design. Common aircraft design methods only give minor consideration to the assembly process, instead focusing on aerodynamic and structural performance. This has a costly downstream effect on assembly which is in part responsible for assembly requiring up to 50% of the manufacturing costs despite being less than half of the production life cycle. Around 2014, researchers introduced the first product family formation method that specifically addresses RAS design. The method uses similarity coefficients of the assembly sequences to cluster components based on commonality. The work is yet to be implemented physically; presenting a theoretical example. Use of this method should yield benefits to production efficiency by way of improving system utilization and productivity. Others have proposed a co-evolution methodology which sets out parameters to simultaneously design a product family and assembly system. The methodology outlines a process for the design of subsequent generations of product and assembly system to maximize profits. Another method discusses a systematic approach to RAS design but falls short of a real-world implementation. The method uses open architecture to select the most appropriate configuration to suit market demand. A recent publication compares product design approaches and manufacturing system design approaches to derive an integrated systematic design approach to changeable manufacturing systems. The design approach itself analyzes the requirements of current and future product ranges along with changeability requirements that are satisfied by systematically specifying appropriate reconfigurable solutions before simulation and evaluation before implementation. Finally, other research cites product family as a central issue in RASs and commonality as a key attribute. The work discusses various mathematical methods for product family formation before introducing a novel matrix-based method to quantify commonality and compatibility to group products for production.

Reconfigurable Assembly Systems RASs covers a broad area of research. The discussion here is limited to physical realization in industry while still possessing the characteristics outlined in the original definition. Assembly tooling is discussed in particular, which is most relevant to

44

Embracing Reconfigurable Assembly Cells for Aerospace aerospace assembly. Due to the increased commonality of product families in the car industry, RASs are more feasible and frequent.

 The reconfigurable tooling where bodyside tooling for vehicle A and B are on either side of a base plate which rotates and indexes to body structure. Volume and model type can be adjusted according to demand. Nissan operates a notable real-world example. Up to eight types of vehicle are produced on a single line, whereas a conventional line can only produce two types of vehicles. The capital expenditure (CapEx) for a new model is reduced by 80% compared to a conventional line. The lead time for the tooling is approximately three months, whereas a conventional system needs approximately one year before launching. Subsequent numerical control (NC) systems have a Process Capability Index (Cp) ratio of 20-30% higher than that of a conventional system. RASs are less prominent in aerospace assembly, which is likely due to many factors. Legacy commercial aircraft produce fewer models in much lower volumes than automotive. The automotive sector produces new models roughly every three years which means automotive tooling has a much shorter life cycle. Aircraft models are in production for 20 years or more; tooling is used for the life of the aircraft and the tooling design is duplicated to increase production rate. Conventional assembly of aerostructures is carried out using large, dedicated fixtures to hold the design tolerances. Assembly fixtures require extensive reinforcement to maintain the tolerances across a large area. Moving parts of the fixture, as with a reconfigurable fixture, opens the risk of assemblies falling out of tolerance. Aircraft also have much longer life cycles than cars and so the demand for change occurs less frequently, though some believe this is likely to change. Cost of flight certification is also a major challenge specific to aerospace RASs. Despite fewer examples of production-demonstrated RASs in aerospace, the technologies to implement such a system that embody RAS characteristics are available.

45

Chapter 4

 The latest body-in-white (BiW) assembly tooling uses numerical control (NC) actuators to position tooling for a variety of models. Modern metrology systems can now be used to verify position or be physically integrated and adapt in real time. Recent reductions in setup complexity and costs mean metrology systems technologies congruent to RASs are now becoming viable for industrial use. Notable examples of aerospace RASs research include the demonstration of reconfigurable flexible tooling that uses a laser tracker for accurate assembly. A similar principle is employed to position a robotic end-effector and hexapods. This approach was adopted to assemble structures of the SAAB Neuron aircraft. For more compliant parts, a reactive reconfigurable tool demonstrates the use of load cells to assemble fuselage panels. A flexible tooling fixture has also been demonstrated for assembling 17 different fuselage panels. Detailed reviews find that in-production technologies can, with careful design, be used to design an RAS. A recent study presented a demonstrator that uses reconfigurable hexapods mounted to modular box joint tooling to position ribs inside a wing box. A further advancement that enables reconfigurability is mobile robotics. Mobile robotics and mobile automation in general provide the necessary requirements for scalability and capacity increase by adding additional automation that has the capability to move around the factory. Automation for drilling the A400M wing is moved around the factory without detriment to process capability. Advances have been made in integrating metrology systems with conventional industrial robots. This body of work has culminated into mobile robotic technology.

46

Embracing Reconfigurable Assembly Cells for Aerospace

 A mobile robot platform can integrate metrology systems with conventional industrial robots to perform multiple assembly tasks in different locations on a variety of assemblies. Several of the authors of this chapter developed a novel RAS design method to deal with specific aerospace issues. The methodology was applied to design a RAS for cost-effective wing structure assembly specifically designed for rate increase. The case study results were positive despite being applied to an existing wing that was not designed for product family. Estimated CapEx was of similar order-of-magnitude to a conventional assembly system but RASs require higher initial investment, although, overall production costs are reduced in the long term with significant reduction in backlog. The RAS design method can potentially be applied to simultaneous product family and RAS design but this has not yet been tested. Other recent design methods use similar principles and could be applied to system design.

RAS Summary Reviews of current research for product family design and RASs have been presented regarding adoption for aerospace assembly. Several sources mark use of product platforms in aerospace. However, there is little evidence of downstream consideration of the assembly system and even less so for RASs where there is no effort in designing aerospace components as a family for reconfigurable assembly. The latest research shows product family formation has emerged as a key issue for RASs which has been addressed to some extent with novel methods, but there is little evidence of physical implementation. RASs are successfully demonstrated in the automotive sector with significant improvements to cost and production rate.

47

Chapter 4

 An RAS design methodology for a cost-effective wing structure assembly can be specifically designed for rate increase. Automotive RASs are supported with designing part families for reconfigurable assembly. Technologies suitable for aerostructure RASs are available but fall short of full production implementation. The literature review found no research on concurrent design of product family and RASs for aerostructures; despite the potential gains evident in the automotive sector. Of course, applying automotive RASs to aerospace is unfeasible because of the different requirements but the same commonality and reconfigurability principles can be applied. This chapter discusses the implementation of such a system using real-world products and constraints to elicit challenges and open new research streams. The work is presented in an industrial context and elucidates issues particular to aerospace for future resolution. The early phases of a wingbox and reconfigurable assembly cell demonstrators are presented as part of ongoing research at The University of Nottingham and provide insight into the future of aerostructure design and assembly.

Product Family and RAS Design Framework The automotive sector has shown product families designed with common platforms enable assembly of all variants using a single assembly system designed to reconfigure itself to specific changes that has reduced production costs and timeto-market. Aerospace assembly has yet to adopt this design practice partly because of priorities given to aerodynamic performance. With cheaper production costs of new competitors in emerging economies, manufacturing agility may soon be central to maintaining competitiveness. The changing market and advances in technology

48

Embracing Reconfigurable Assembly Cells for Aerospace mean aerostructures can be designed in such a way to improve manufacturability without cost to flight performance. No approach exists thus far which is applicable for the unique challenges of aerospace. A design framework was developed to address challenges particular to aerospace. The framework takes a novel approach to symbiotic design of product family and RASs. Previous methodologies integrate product and process design but do not fit aerospace. Nor do systematic approaches to product family formation and RAS design. These works have been adapted to suit aerospace design approaches and constraints. The framework introduces two new corollaries to develop common aerostructures and an assembly system which also differentiates the framework from previous attempts. The first corollary is that of common assembly processes instead of common platform architecture seen in automotive. With advances to robotics, vision systems, tool changers, and metrology systems it is possible for one system to perform assembly operations in a variety of situations. The second corollary is that of integrated collaboration between design and manufacturing engineers for structural design. Historically, design is passed to manufacture without iteration and assembly input. Concurrent engineering has been implemented but is generally not applied effectively in practice because there is no formal protocol for the interaction between design and manufacturing centered on structural design.

 A framework can be established for design of aerostructure product family and reconfigurable assembly system.

49

Chapter 4 Structural design is the point in design that fixes aerodynamic geometry and assembly geometry. Therefore, this should be allocated the most time for design iteration and should involve both sets of stakeholders. Given that aerodynamic or structural performance is not negatively affected, manufacturing and assembly can take precedence and employ design for manufacture and assembly (DfMA) and reconfigurability strategies. Intrinsically linked to the first corollary, common processes, commonality in the structure and interfaces is possible providing the structure passes stress analysis. The framework structures an overlap in the design phase where aircraft design and manufacturing design can iterate effectively and converge on a final structural design without significant trade-off. Ultimately, these modular underpinnings mean assembly can use one set of flexible assembly processes embodied in one reconfigurable system.

Industrial Implementation High-value research and development is underway at The University of Nottingham to assemble a set of aerostructures using an RAS. Components will be composite and metallic structures for a flight-test research aircraft. As such, details of the components and assembly system are restricted at this time until appropriate declassification. The framework has been employed for the design phase and implementation is underway. Several non-flying aerostructure components have been manufactured to demonstrate the assembly process capabilities. The assembly cell will be reconfigured for future aerostructure sets to demonstrate reconfigurable capabilities.

 A non-flying wingbox rib for assembly using the reconfigurable assembly cell is depicted.

50

Embracing Reconfigurable Assembly Cells for Aerospace

 A non-flying wingbox spar for assembly using the reconfigurable assembly cell is depicted. The build philosophy for assembly is partially based on measurement-assisted determinate assembly. The overarching philosophy is assembly of components with full size holes pre-drilled to remove disassembly operations. To assemble the different aerostructures several technological innovations are being developed to provide reconfigurability characteristics at the required accuracies. Metrology-enhanced robotics, flexible handling, and vision systems are combined to assemble components according to on-part tolerances with compensatory processes used to manage manufacturing deviations. The fixturing system is positioned using a vision system to computer aided design (CAD) nominal to account for tolerance stack-up in the fixture components. Similarly, critical key characteristics (KCs) are measured and manufacturing errors are compensated for in the NC paths. A set of aerostructures are in tandem development under the design framework centering on structural design to develop geometry with common assembly processes without trade-off for flight performance. Developing the geometry with assembly in mind has simplified hole-to-hole assembly by interaction with stress analysis and assembly engineers. The ongoing work is showing comparable results to industry and details are forthcoming. Likely metrics will be based on a wide range of capabilities including accuracy and repeatability, reconfigurability, production rate and capital expenditure.

Limitations of the Approach The design framework introduces a new approach to aerostructure assembly with novel insight into structural design. Several advantages can be claimed in using the design framework, but there are some inherent limitations. While the framework rationale provides key insight into overcoming aerostructure design for assembly success it is dependent on careful execution in practice. Such work would involve detailed work package structuring and positive involvement from all stakeholders. The broader context may involve drastic changes to

51

Chapter 4 industry protocol which should be pursued on the grounds of competitiveness. Reorganization of personnel is also a possibility. Practical concerns are extensive; including but not limited to: re-evaluation of tolerances carried forward from older programs, greater number of iterations between design and manufacturing for design changes, extra stress analysis, and required advances in structures optimization. In regard to the demonstrator, extensive technical development is still needed. Overall, aerostructures with a common platform for reconfigurable assembly are well within technical possibility. And though the framework is a radical strategy that represents significant departure from the current approach, it has been demonstrated in automotive products and other industries. This chapter is based on and imagery is taken from SAE International Technical Paper "Design of a Reconfigurable Assembly Cell for Multiple Aerostructures," by Jefferson, T., Crossley, R., Smith, A., and Ratchev, S., doi:10.4271/2016-01-2105.

52

Chapter 5

Optimizing Robot-Based Machining Systems Simon Kothe, Sven Philipp von Stürmer, Hans Christian Schmidt, and Christian Boehlmann Fraunhofer IFAM

Jörg Wollnack and Wolfgang Hintze Hamburg University of Technology Though the manufacture of carbon fiber reinforced plastic (CFRP) aircraft structures results in a part that is close to the final contour, trimming and drilling processes are still a very important part of the process chain to achieve the tolerances for assembly. Currently those drilling and milling processes are conducted mainly by gantry machining centers. Because of their basic design and mission, they must be bigger than the structures being machined. This leads to huge systems since CFRP aircraft structures are typically large, highly integrated components. A new approach that is becoming more and more important for machining processes are robot-based machining systems, which up to now, have mainly been used in the automotive industry. The advantages of this approach include less investment costs, since robots are standard mass automation components, and more flexibility by re-arrangement of the systems. Other benefits include a reduction in lead-time due to several robots machining one structure at the same time and the possibility of mobilizing the systems, leading to versatile production environments. The advantage of shorter lead-times is especially important, as the productivity of aircraft manufacturing processes simply must be improved. Over the last ten years,

53

Chapter 5 continuously growing air traffic has resulted in significantly increased aircraft order numbers (~+70%), and as those number have grown so has the general use of CFRP.

 Machining system experiments were performed with a KUKA KR300 Quantec Ultra robot (KRC4) mounted on a Güdel TMF-4 rail.

Methods for Performance Assessment In comparison to conventional computer numerical control (CNC) machining centers, the main problem of robots is their positioning and path accuracy. Furthermore, their low position and orientation-dependent stiffness, resulting from the serial rotary axes kinematics, leads to varying tool deflections in reaction to dynamic process forces. The Fraunhofer IFAM works on these challenges to qualify industrial robots for reliable machining processes. This chapter offers three different methods for the assessment of performance and accuracy of robots for machining processes or processes with path accuracy in general. By abstracting and generalizing the process of path realization for machining, the three methods were designed with respect to effort of realization and information output. The first method is based on a real machining process of a specimen while the others only track a specified path by external or internal measurement systems. In particular, the third approach is based on the internal encoder information of the robot itself. It represents an innovative and easy to use strategy since there is no need to

54

Optimizing Robot-Based Machining Systems perform a real machining process and no additional measurement systems (e.g., Laser Tracker) are needed. Due to this fact every operator of a robotic system is able to perform this method. A typical strategy to test the machining performance of a robot is the machining of a specimen according to VDI/NCG 5211-2. It was formally designed to check the performance of 5-axis CNC machining centers. The geometry consists of different form elements so that typical machining failures like path accuracy, re-orientationbehavior, or wrong-tool compensation can be visually checked without the need of measurement devices. Since the easy-to-cut polyurethane-based material Necuron 1007 was used and only low material-removal-rates were performed the active process force was only 10-40 N. Due to this fact, failures caused by the machine and controller itself and not by the force-tool-deflection can be investigated. The tool used for the milling test was a 6-mm solid carbide end mill with four cutting edges. The machining results showed robot-typical deviations of the form elements that are caused by several influences (e.g., interpolation inaccuracies, control errors, and calibration failures) like wrong kinematic parameters and spindle alignment. Control errors and interpolation inaccuracies can typically be seen in the machined surface and geometrical shape, while calibration and tool alignment failures are particularly observed during re-orientation-movements of the “dome,” “tilted box edges,” or “spikes.”

 By abstracting and generalizing the process of path realization for machining, three methods were designed with respect to effort of realization and information output. The top layer describes the access to the information for assessment depending on the progress of path realization.

55

Chapter 5

 A typical strategy to test the machining performance of a robot is the machining of a specimen according to VDI/NCG 5211-2. The specimen geometry consists of different form elements so that typical machining failures like path accuracy, re-orientationbehavior or wrong tool-compensation can be visually checked without need of measurement devices.

 The machining result of the specimen with the KUKA KR300 robot system shows robot-typical deviations of the form elements that are caused by several influences (e.g., interpolation inaccuracies, control errors and calibration failures) like wrong kinematic parameters and spindle alignment.

56

Optimizing Robot-Based Machining Systems Test specimens provide a good and easy way to display, compare, and judge the quality of machining systems, though it is nearly impossible to derive the reason for failures. Machining performance is a result of several different influences caused by different drivers. Due to the robot kinematics, that is built up of six rotation axes, movements are typically interpolated by several axes at the same time, which makes an association difficult. Another method to judge the quality of a machining system without the necessity of machining is tracking of a path according to ISO 9283. For this type of investigation, a continuously acquiring coordinate measurement system to track and analyze the path (e.g., a Laser Tracker system) is necessary. The path is built-up by different linear and circular movements varying in orientation and size with several continuous and discontinuous block changes; reverse movements are also included. The measurement of the ISO 9283 path shows typical errors of an industrial robot leading the tool center point (TCP) on a path in Cartesian space. The actual path will show systematic and static as well as dynamic errors. The error in absolute translation and orientation of the whole path in space is visible, and typically influenced by calibration and alignment errors. Further errors occur at discontinuous segments of the path, where higher accelerations for quasi-stepwise changes in direction would be necessary. The observed errors give a first hint toward fields of suitable actions for optimization. Based on the findings a retuning of the robot control parameters and an external guidance with Laser Tracker were chosen for improvement.

 Another method to judge the quality of a machining system without the necessity of machining is tracking of a standard path according to ISO 9283. The actual path shows systematic, static, and dynamic errors.

57

Chapter 5

 The new “virtual machining” approach gives an outlook to the result of a performance assessment method based on a virtual part, calculated on basis of a measured path. This gives the advantage to image and identify errors before machining. The third “virtual machining” approach gives an outlook to the result of a performance assessment method based on a virtual part, calculated on basis of a measured path. This introduces the advantage to image and identify errors before machining. A smart possibility to build up the virtual part is given by the cheap and easy way using internal resolver data of the robot system itself. A set of actual TCP-poses in the coordinate system of the part is traced. The measured poses will be saved and converted to an numerical control- (NC) program. After that the NC-program is imported into computer-aided monitoring (CAM) software. A material removal simulation then calculates the virtual part based on the measured data. The virtual part is made available as a triangulated surface (stereo lithography (STL) format). This data format is typically usable with most third-party software solutions (e.g., for target/actual part comparison). By comparing a virtual part derived from encoder data and a real part using the example of the VDI/NCG test specimen the usability of this approach as a prediction-tool becomes apparent. Errors, like those at the dome of the specimen, are clearly mapped on the virtual part and corresponding to the deviations of the real

58

Optimizing Robot-Based Machining Systems part. The investigations show that by using the resolver data of the robot system, control errors in particular can be recognized and become visible in the virtual part. To capture further errors the method can also be expanded by using external measurement system data, (e.g., path tracking by Laser Tracker). Depending on the state of acquisition nearly all errors can be projected into the virtual part.

 This zoom-view of the “dome” provides an example for calibration and alignment-failures, leading to wrong tool-paths. The image also shows chatter marks in the surface, indicating control errors.

Optimization Strategies for Robot Machining Two strategies are offered here for robot positioning and path accuracy optimization. One is real-time closed-loop guidance with a Laser Tracker. The second is optimization of control parameters.

Closed-Loop External Guidance with LaserTracker One method for optimization of a robot machining process is closed-loop external guidance by a real-time capable Laser Tracker. The setup for the presented realization of this strategy used a Leica AT901 with real-time extension and T-Mac. An EtherCat-Bridge from ESD Electronics was integrated to allow a Master-MasterEtherCat-Communication between the internal KUKA robot bus system and an industrial PC (IPC) from Beckhoff. The IPC was used to calculate axis angle correction values based on the actual pose measured by the Laser Tracker and the target

59

Chapter 5

 To deal with the dynamic behavior of the closed-loop, and integral controller was implemented on the industrial PC (IPC). Shown is the setup for external guidance. pose received from the robot controller. To compensate the position and orientation errors these correction values were fed back to the robot controller via the KUKA Remote Sensor Interface (RSI). To deal with the dynamic behavior of the closed-loop an integral controller was implemented on the IPC. The first performance check conducted with the Laser Tracker real-time guidance was the standard-path according to ISO 9283. The position and orientation of the path according to the global reference coordinate system was optimized. Without external guidance, the absolute Euclidean position error reached up to 1.5 mm. This failure was reduced to less than 0.13 mm, and results showed that quasi-static failures from calibration and alignment are substantially reduced. In contrast to this the shape of the path itself cannot be optimized significantly because of the limited dynamic behavior of the robotic system. Only significant macroscopic deviations with a low control effort can be reduced. Since the standard-path check according to ISO 9283 involves no external forces, an additional machining test was conducted. The objective was to investigate the ability of the external Laser Tracker guidance regarding compensation of process-forceindicated tool deflections. For this purpose, the Necuron material from the VDI/NCG 5211-2 specimen-check and a 14.9 mm solid carbide end mill were used. Two notches were machined in different feed directions. To induce relevant forces into the robot structure, certain material removal rates were needed. For this purpose, the depth of the notches was set to 12 mm.

60

Optimizing Robot-Based Machining Systems

 The first performance check conducted with the Laser Tracker real-time guidance was the standard-path according to ISO 9283. With external guidance. The position and orientation of the path according to the global reference coordinate system was optimized. The deflection of the tool, mainly caused by the induced process force of 180 N, was approximately 0.6 mm. Additionally, a static offset due to calibration and alignment errors was visible leading to a complete deviation of ~1.1mm. The sum of deflection and static offset was reduced to 0.1 mm when the Laser Tracker control loop was activated.

 To improve the control performance, a more aggressive set of optimized control parameters was chosen and the path was analyzed with the virtual machining method

61

Chapter 5

Optimization of Control Parameters Compared to the ISO 9283 path, the VDI/NCG 5211-2 specimen path is even more challenging concerning machine dynamics. With the approach of virtual machining the effects of the contouring errors on the internally measured path are mapped on the virtual specimen. The quality of the movement is restricted by the limited control performance. Standard industrial robots are not designed for small and dynamic movements that are required for machining processes. Even the control setup is not optimized for such types of motion so that a less conservative control setup is a promising optimization action.

 The comparison is shown between default and optimized control parameters with tool-paths (top), virtual machined specimen (center), and 3D-microscope image of real specimen (bottom).The analysis of the internally measured path and comparison to the default control parameter set shows that the actual/target deviation is reduced by the optimized control parameters. The “optimized path” gets closer to the target path.

62

Optimizing Robot-Based Machining Systems To improve the control performance a more aggressive set of control parameters was chosen and the path was analyzed with the virtual machining method. The analysis of the internally measured path and comparison to the default control parameter set showed that the actual/target deviation was reduced by the optimized control parameters. The “optimized path” got closer to the target path. This was also observable in the shape of the virtual specimens tilted box edges. The solid specimen machined with default parameters showed a part of not removed material at the right chamfer while the left chamfer was too deep. In contrast to that, the specimen with optimized parameters showed a more homogenous shape closer to the target. The images of the element, gathered by a Keyence 3D microscope, verified this optimization. The tilted box edge as an example illustrated the better path accuracy and resulting shape all over the specimen by using optimized control parameters. By directly getting a picture of the shape, the advantage of the virtual machined approach became significant. Compared to a simple 3D path plot (x|y|z), where effects of errors on the result are much harder to identify, the volumetric visualization made the analysis more intuitive. Next steps will include the development of an optimization strategy based on the virtual machining technique. For this purpose, a tuning-software for robotmachining-programs will be worked out. To further improve the performance, it will also include an enhanced mathematic robot kinematics model including a linear rail. This chapter is based on and imagery is taken from SAE International Technical Paper "Accuracy Analysis and Error Source Identification for Optimization of Robot Based Machining Systems for Aerospace Production," by Kothe, S., von Stürmer, S., Schmidt, H., Boehlmann, C. et al., doi:10.4271/2016-01-2137.

63

About the Editor Jean is currently the Editorial Director and Content Strategist for SAE International’s Aerospace Product Group in Warrendale, PA. Previously, she worked as the Managing Editor for all of SAE Magazines and as the Editor of Aerospace Engineering and Off-Highway Engineering magazines. Prior to that, Jean worked in product development in industry and served as an Air Weapons Controller in the United States Air Force with primary responsibility as an Air Surveillance Officer.

65