Artificial Neural N etworks for Intelligent Manufacturing

Artificial Neural N etworks for Intelligent Manufacturing

Intelligent Manufacturing Series Ser ies Editor: Andrew Kusiak Department of Industrial Engineering The University of Iowa, USA Manufacturing has been issued agreat challenge - the challenge of Artificial Intelligence (AI). We are witnessing the proliferation of applications of AI in industry, ranging from finance and marketing to design and manufacturing processes. AI tools have been incorporated into computer-aided design and shop-floor operations software, as well as entering use in logistics systems. The success of AI in manufacturing can be measured by its growing number of applications, releases of new software products and in the many conferences and new publications. This series on Intelligent Manufacturing has been established in response to these developments, and will include books on topics such as: design for manufacturing concurrent engineering process planning production planning and scheduling programming languages and environments design, operations and management of intelligent systems Some of the titles are more theoretical in nature, while others emphasize an industrial perspective. Books dealing with the most recent developments will be edited by leaders in the particular fields. In areas that are more established, books written by recognized authors are planned. We are confident that the titles in the series will be appreciated by students entering the field ofintelligent manufacturing, academics, design and manufacturing managers, system engineers, analysts and programmers. Titles available Object-oriented Software for Manufacturing Systems Edited by S. Adiga Integrated Distributed Intelligence Systems in Manufacturing M. Rao, Q. Wang and 1. Cha Artificial Neural Networks for Intelligent Manufacturing Edited By C.R. Dagli

Artificial Neural Networks for Intelligent Manufacturing Edited by Cihan H. Dagli Associate Professor Department of Engineering Management University of Missouri-Rolla USA Springer-Science+Business Media, B.V.

First edition 1994 Springer Science+Business Media Dordrecht 1994 Originally published by Chapman & Hali in 1994 Softcover reprint of the hardcover Ist edition 1994 Typeset in 10/12 pts Times by Thomson Press (India) Ltd, New Delhi ISBN 978-94-010-4307-6 ISBN 978-94-011-0713-6 (ebook) DOI 10.1007/978-94-011-0713-6 Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the UK Copyright Designs and Patents Act, 1988, this publication may not be reproduced, stored, or transmitted, in any form or by any means, without the prior permission in writing of the publishers, or in the case of reprographic reproduction only in accordance with the terms of the licences issued by the Copyright Licensing Agency in the UK, or in accordance with the terms of licences issued by the appropriate Reproduction Rights Organization outside the UK. Enquiries concerning reproduction outside the terms stated here should be sent to the publishers at the London address printed on this page. The publisher makes no representation, express or implied, with regard to the accuracy of the information contained in this book and cannot accept any legal responsibility or liability for any errors or omissions that may be made. A catalogue record for this book is available from the British Library Library of Congress Cataloging-in-Publication data Artificial neural networks for intelligent manufacturing / edited by Cihan H. Dagli. - ist ed. p. cm. Includes index. ISBN 0-412-48050-6 (acid-free paper) 1. Neural networks (Computer science) 1. Dagli, Cihan H., 1949- QA76.87.A7432 1994 670'.285'63 - dc20 2. Manufacturing processes. 93-35422 CIP ~Printed on permanent acid-free text paper, manufactured in accordance with the proposed ANSljNISO Z 39.48-199X and ANSI Z 39.48-1984

To the memory of my father Kenan Dagli To my mother Zuhre Dagli and My wife Refia and my sons Kenan Cagri and Mehmet Ediz

Contents Contributors Preface xiii xv PART ONE Intelligent manufacturing: Basic concepts and tools Intelligent manufacturing systems Cihan H. Dagli 1.1 Manufacturing systems and strategies 1.2 Hierarchical levels in manufacturing 1.3 Characteristics of intelligent manufacturing systems 1.4 Summary References 2 Intelligent systems architecture: Design techniques Deborah Stacey 2.1 Introduction 2.2 Knowledge-based systems 2.3 Artificial neural networks 2.4 Hybrid intelligent systems 2.5 Manufacturing systems implementations 2.6 Summary References 3 Basic artificial neural network architectures Cihan H. Dagli and Pipatpong Poshyanonda 3.1 Basic concepts 3.2 Percept ron 3.3 Backpropagation 3.4 Adaptive resonance theory 3.5 Summary References 4 Hybrid intelligent systems: Tools for decision making in intelligent manufacturing Gregory R. Madey, Jay Weinroth and Vijay Shah 4.1 Overview 3 4 7 13 15 16 17 17 18 24 30 34 35 37 39 39 43 48 56 64 64 67 67

Vlll Contents 4.2 Manufacturing decision-making problems: 68 organization, coordination and executing levels 4.3 Hybrid intelligent systems developed 71 out of neural networks 4.4 Survey of neural network hybrid 77 intelligent systems 4.5 Case study of the development of a 81 hybrid intelligent system for decision making in manufacturing 4.6 Summary 86 References 87 PART TWO Neurocomputing for intelligent manufacturing: 91 Organization and coordination level applications 5 Conceptual design problem 93 Ali Bahrami and Cihan H. Dagli 5.1 Characteristics of design problem 93 5.2 Introduction to fuzzy sets and binary 98 relationships between functions and structures 5.3 Sample problem definition 101 5.4 Fuzzy knowledge representations 102 5.5 Mapping fuzzy functional requirements 105 to design structure by F AM 5.6 Implementation and input/output 106 representations 5.7 Experimental results 108 5.8 Summary 109 References 109 6 Machine-part family formation 111 Cesar O. Malave and Satheesh Ramachandran 6.1 Characteristics of group technology 111 6.2 Neural network approach 117 6.3 Discussion 138 References 141 7 Process planning 143 M adhusudhan Posani and Cihan H. Dagli 7.1 Characteristics of process planning 143 7.2 Sample problem definition 149 7.3 Development of network architecture 150 7.4 Artificial neural network implementation 154 7.5 Performance of the intelligent system 154 architecture

Contents IX 7.6 Summary 156 References 157 8 Scheduling 159 John Y. Cheung 8.1 Characteristics of scheduling problems 159 8.2 The Hopfield net approach 161 8.3 Simulated annealing 174 8.4 Other neural network techniques 182 8.5 Summary 186 References 186 9 Automated assembly systems 195 Cihan H. Dagli and Mahesh Kumar Vellanki 9.1 Automated assembly 196 9.2 Generic assembly cell 198 9.3 Power supply board assembly: 211 A case study 9.4 Summary 227 References 228 10 Manufacturing feature identification 229 Mark R. Henderson 10.1 Characteristics of manufacturing features 229 10.2 Sample problem definition 239 10.3 Development of network architecture 241 10.4 Artificial neural network implementation 247 10.5 Performance of the intelligent system 253 architecture 10.6 Summary 260 Acknowledgements 263 References 263 11 Vision based inspection 265 J oydeep Ghosh 11.1 Introduction 265 11.2 Characteristics of vision based inspection 267 systems 11.3 Representation of 3D objects 269 11.4 Modeling and matching strategies 272 11.5 Artificial neural networks (ANNs) for 273 vision-based inspection 11.6 Viewer-centered object recognition 280 11.7 Direct, object-based ANN approaches 292 11.8 Concluding remarks 293 Acknowledgements 294 References 294

x Contents 12 Performance analysis of artificial neural 299 network methods Benito Fernandez R. 12.1 Introduction 299 12.2 Artificial neural systems in 300 man ufacturing 12.3 The power of neural networks 301 12.4 Artificial neural network paradigms in 309 manufacturing 12.5 Performance analysis 314 12.6 Benchmarks 321 12.7 Simulation paradox 338 12.8 Performance measures 341 12.9 Decision functions 345 12.10 Metrics from measure 347 12.11 Cluster analysis 348 12.12 ANN paradigm selection in 352 manufacturing 12.13 Tools that increase performance 353 12.14 Summary 363 References 363 PART THREE Neurocomputing for intelligent manufacturing: 369 Execution level applications 13 Process monitoring and control 371 Michel Guillot, Riadh Azouzi and Marie-Claude Cote 13.1 Introduction to process monitoring 371 and control 13.2 Neural, network models for process 376 monitoring and control 13.3 Neural network approaches to process 378 monitoring 13.4 Neural network approaches to process 380 control 13.5 Implementation cases 384 13.6 Summary 396 References 396 14 Adaptive control in manufacturing 399 Yung Shin 14.1 Characteristics of adaptive control 399 systems 14.2 Sample problem definition 403 14.3 Adaptive neuro-control architecture 406

Contents Xl 14.4 Performance of adaptive neuro-control 410 systems 14.5 Conclusions 411 References 411 15 Fuzzy neural control 413 L.H. Tsoukalas. A. Ikonomopoulos and R.E. Uhrig 15.1 Problem of fuzzy control 413 15.2 Fuzzy neural architectures 417 15.3 Development of system architecture 422 15.4 Fuzzy neural network implementation 430 and performance 15.5 Summary 432 References 433 16 Neural networks in continuous process 435 diagnostics N ajwa S. M erchawi and Soundar R. T K umara 16.1 Introduction 436 16.2 Neural networks for diagnostics 436 16.3 Problem description 437 16.4 Knowledge representation for continuous 439 process diagnostics by a neural network 16.5 Example problem: The TMI-2 nuclear 442 reactor 16.6 Implementation and simulation results 447 16.7 Summary and conclusions 460 References 461 Index 463

Contributors Riadh Azouzi, Mechanical Engineering Department, Laval University, Quebec, Canada, GIK 7P4. Ali Bahrami, Computer Information Systems, Department of Economics and Management, Rhode Island College, Providence, Rhode Island 02908. John Y. Cheung, School of Electrical Engineering & Computer Science, The University of Oklahoma, Norman, OK 70319-036l. Marie-Claude Cote, Mechanical Engineering Department, Laval University, Quebec, Canada, GIK 7P4. Cihan H. Dagli, Engineering Management Department, University of Missouri-Rolla, Rolla, Missouri 65401 Benito Fermindez R., Neuro-Engineering Research and Development (NERD) Laboratory, The University of Texas at Austin, Department of Mechanical Engineering, Austin, Texas, 78712-1063. Joydeep Ghosh, Department of Electrical and Computer Engineering, The University of Texas at Austin, Austin, Texas 78712-1084. Michel Guillot, Mechanical Engineering Department, Laval University, Quebec, Canada, GIK 7P4. Mark R. Henderson, Associate Director, Computer-Integrated Manufacturing Systems Research Center, Associate Professor of Mechanical/Aerospace Engineering, Arizona State University. A. Ikonomopoulos, Department of Nuclear Engineering and Center for Neural Engineering, The University of Tennessee and Oak Ridge National Laboratory, Knoxville, TN 37996-2300. Soundar R.T. Kumara, Intelligent Design and Diagnostics Research Laboratory, Department of Industrial and Management Systems Engineering, The Pennsylvania State University, University Park, PA 16802 Gregory R. Madey, Kent State University, Graduate School of Management, Kent State University, Kent, Ohio 44242, USA. Cesar O. Malave, Department of Industrial Engineering, Texas A & M University, College Station, TX 77843-3131.

XIV Contributors Najwa S. Merchawi, Intelligent Design and Diagnostics Research Laboratory, Department of Industrial and Management Systems Engineering, The Pennsylvania State University, University Park, PA 16802 Madhusudhan Posani, University of Missouri, University Extension, Columbia, Missouri 65211. Pipatpong Poshyanonda, Engineering Management Department, University of Missouri-Rolla, Rolla, Missouri 65401. Satheesh Ramachandran, Department of Industrial Engineering, Texas A & M University, College Station, TX 77843-3131. Vijay Shah, Kent State University, Graduate School of Management, Kent State University, Kent, Ohio 44242, USA. Yung Shin, School of Mechanical Engineering, Purdue University, West Lafayette, IN 47907. Deborah Stacey, Department of Computing and Information Science, University of Guelph, Guelph, Ontario Canada N1G 2W1. L.H. Tsoukalas, Department of Nuclear Engineering and Center for Neural Engineering, The University of Tennessee and Oak Ridge National Laboratory, Knoxville, TN 37996-2300. R.E. Uhrig, Department of Nuclear Engineering and Center for Neural Engineering, The University of Tennessee and Oak Ridge National Laboratory, Knoxville, TN 37996-2300. Mahesh Kumar Vellanki, AT & T Bell Laboratories, Engineering Research Center, Hopewell, NJ 082525. Jay Weinroth, Kent State University, Graduate School of Management, Kent State University, Kent, Ohio 44242. USA.

Preface The quest for building systems that can function automatically has attracted a lot of attention over the centuries and created continuous research activities. As users of these systems we have never been satisfied, and demand more from the artifacts that are designed and manufactured. The current trend is to build autonomous systems that can adapt to changes in their environment. While there is a lot to be done before we reach this point, it is not possible to separate manufacturing systems from this trend. The desire to achieve fully automated manufacturing systems is here to stay. Manufacturing systems of the twenty-first century will demand more flexibility in product design, process planning, scheduling and process control. This may well be achieved through integrated software and hardware architectures that generate current decisions based on information collected from manufacturing systems environment, and execute these decisions by converting them into signals transferred through communication network. Manufacturing technology has not yet reached this state. However, the urge for achieving this goal is transferred into the term 'Intelligent Systems' that we started to use more in late 1980s. Knowledge-based systems, our first efforts in this endeavor, were not sufficient to generate the 'Intelligence' required - our quest still continues. Artificial neural network technology is becoming an integral part of intelligent manufacturing systems and will have a profound impact on the design of autonomous engineering systems over the next few years. This book introduces this newly-emerging technology and demonstrates its use, with examples, in intelligent manufacturing systems. After introducing the basic components of intelligent manufacturing systems and intelligent system architecture design techniques, the text provides sufficient coverage of basic artificial neural network architectures to be used successfully in manufacturing applications. The subsequent two parts of the book, discussed in more detail below, cover the use of neural networks in organization, coordination and execution levels of the manufacturing system hierarchy. The second part of the book covers the development of artificial neural network architectures for solving fundamental problems of manufacturing; conceptual design, group technology, process planning, scheduling, automated assembly, inspection and manufacturing feature identification. It concludes with an excellent chapter that summarizes the performance analysis of artificial neural network methods. The third part concentrates on execution

XVI Preface level applications that generally require extensive data collection. The use of artificial neural networks in process monitoring and control, adaptive control, fuzzy neurocontrol and continuous process diagnostics are examined. As a whole, Artificial Neural Networks for Intelligent Man~facturing provides a major reference for all those interested in intelligent manufacturing systems. I believe that this is another step in our quest to build autonomous systems. I would like to thank all the contributors who have made this book possible. Cihan H. Dagli Rolla, Missouri, USA Octo ber 1993