FUZZY ENGINEERING EXPERT SYSTEMS WITH NEURAL NETWORK APPLICATIONS

Transcription

1 FUZZY ENGINEERING EXPERT SYSTEMS WITH NEURAL NETWORK APPLICATIONS ADEDEJI B. BADIRU Department of Industrial Engineering University of Tennessee Knoxville, TN JOHN Y. CHEUNG School of Electrical and Computer Engineering University of Oklahoma Norman, OK JOHN WILEY & SONS, INC.

2

3 FUZZY ENGINEERING EXPERT SYSTEMS WITH NEURAL NETWORK APPLICATIONS

4

5 FUZZY ENGINEERING EXPERT SYSTEMS WITH NEURAL NETWORK APPLICATIONS ADEDEJI B. BADIRU Department of Industrial Engineering University of Tennessee Knoxville, TN JOHN Y. CHEUNG School of Electrical and Computer Engineering University of Oklahoma Norman, OK JOHN WILEY & SONS, INC.

6 This book is printed on acid-free paper. Copyright 2002 by John Wiley & Sons, New York. All rights reserved. Published simultaneously in Canada. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except as permitted under Sections 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, 222 Rosewood Drive, Danvers, MA 01923, (978) , fax (978) Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 605 Third Avenue, New York, NY , (212) , fax (212) , PERMREQ@WILEY.COM. This publication is designed to provide accurate and authoritative information in regard to the subject matter covered. It is sold with the understanding that the publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional person should be sought. Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. For more information about Wiley products, visit our web site at Library of Congress Cataloging-in-Publication Data: Badiru, Adedeji Bodunde, 1952 Fuzzy engineering expert systems with neural network applications / Adedeji B. Badiru, John Y. Cheung. p. cm. ISBN Expert systems (Computer science) 2. Fuzzy systems. 3. Neural networks (Computer science) I. Cheung, John Y. II. Title. QA76.76.E95 B dc Printed in the United States of America

7 To Our Spouses, Iswat Amori Badiru, and Rose Kwan-Fun Cheung

8

9 CONTENTS Preface Acknowledgments xv xvii 1 Artificial Intelligence Origin of Artificial Intelligence Human Intellligence versus Machine Intelligence The First AI Conference Evolution of Smart Programs Branches of Artificial Intelligence Neural Networks Emergence of Expert Systems Embedded Expert Systems 12 2 Fundamentals of Expert Systems Expert Systems Process Expert Systems Characteristics Domain Specificity Special Programming Languages Expert Systems Structure The Need for Expert Systems Benefits of Expert Systems Transition from Data Processing to Knowledge Processing Heuristic Reasoning Search Control Methods Forward Chaining Backward Chaining User Interface Natural Language Explanations Facility in Expert Systems Data Uncertainties 23 vii

10 viii CONTENTS Application Roadmap Symbolic Processing 24 3 Problem Analysis Problem Identification Problem Analysis Scope of the Problem Symbolic Nature of the Problem Solution Time Frequency of Problem Occurrence Optimization versus Satisficing Data and Knowledge Availability Data Requirement Analysis Expert System Justification Problem-Selection Guidelines 47 4 Knowledge Engineering Knowledge-Acquisition Phases The Knowledge Engineer Knowledge Characteristics Choosing the Expert Knowledge Extraction versus Knowledge Acquisition Methods of Extracting Knowledge from Experts Interviews Open-Ended Interviews Advantages and Disadvantages of Interviews Task Performance and Protocols Analyzing the Expert s Thought Process Constrained Task Tough Case Method Questionnaires and Surveys Documentation and Analysis of Acquired Knowledge Expert s Block Knowledge-Acquisition Meetings Group Knowledge Acquisition Brainstorming Delphi Method Nominal Group Technique 64

11 CONTENTS ix 4.5 Knowledge-Acquisition Software Knowledge Elicitation Tools Induction-by-Example Tools Characteristics of Knowledge Types of Knowledge Knowledge-Representation Models Semantic Networks Frames Scripts Rules Predicate Logic O-A-V Triplets Hybrids Specialized Representation Techniques Concept of Knowledge Sets Properties of Knowledge Sets Reasoning Models 97 5 Probabilistic and Fuzzy Reasoning Human Reasoning and Probability Bayesian Approach to Handling Uncertainty Logical Relations Plausible Relations Contextual Relations Decision Tables and Trees Dempster Shafer Theory Support Function Plausibility Uncertainty of A Belief Interval Focal Elements Doubt Function Certainty Factors Fuzzy Logic Definition of Fuzzy Set Fuzzy Systems Crisp Logic versus Fuzzy Logic Crisp Sets 133

12 x CONTENTS Fuzzy Sets Fuzzy Set Construction Fuzzy Set Operations Cuts Extension Principle Fuzzy Operations Fuzzy Complement Fuzzy Intersection Fuzzy Union Duality Fuzzy Implication Fuzzy Aggregation Fuzzy Numbers Fuzzy Number Representation Fuzzy Number Operation Fuzzy Ordering Fuzzy Relation Definition of a Fuzzy Relation Binary Relation Operations with Relations Evidence Theory Believability and Plausibility Uncertainty Fuzziness Nonspecificity Fuzzy Logic Multivalued Logic Unconditional Fuzzy Propositions Conditional Fuzzy Propositions Selection of Implication Operator Multiconditional Reasoning Summary Neural Networks Introduction Definition of a Neurode Variations of a Neurode Single Neurode 183

13 CONTENTS xi The McCulloch Pitts Neurode McCulloch Pitts Neurodes as Boolean Components Single Neurode as Binary Classifier Single-Neurode Perceptron Single-Layer Feedforward Network Multicategory SLP Associative Memory Correlation Matrix Memory Pseudoinverse Memory Widrow Hoff Approach Least-Mean-Squares Approach Adaptive Correlation Matrix Theory Error-Correcting Pseudoinverse Method Self-Organizing Networks Principal Components Clustering by Hebbian Learning Clustering by Oja s Normalization Competitive Learning Network Multiple-Layer Feedforward Network Multiple-Layer Perceptron XOR Example Back-Error Propagation Variations in the Back-Error Propagation Algorithm Learning Rate and Momentum Other Back-Error Propagation Issues Counterpropagation Network Radial Basis Networks Interpolation Radial Basis Network Single-Layer Feedback Network Single-Layer Feedback Network A Discrete Single-Layer Feedback Network Bidirectional Associative Memory Hopfield Network Summary References 220

14 xii CONTENTS 8 Neural-Fuzzy Networks Technology Comparisons Neurons Performing Fuzzy Operations Neurons Emulating Fuzzy Operations Neurons Performing Fuzzy Operations Neural Network Performing Fuzzy Inference Regular Neural Network with Crisp Input and Output Regular Neural Network with Fuzzy Input and Output Fuzzy Inference Network ANFIS Applications Clustering and Classification Classification Multilayer Fuzzy Perceptron Evolutionary Computing Introduction Binary Genetic Algorithm Genetic Representation Population Fitness Check and Cost Evaluation Mating Pool Pairing Mating Mutation The Next Generation Performance Enhancements Continuous Genetic Algorithm Genetic Representation Mating Mutation Performance Enhancements Evolutionary Programming Evolutionary Strategies 259

15 CONTENTS xiii Evolutionary Programming Summary Intelligent Strategy Generation in Complex Manufacturing Environments Introduction Model Description Evolutionary Algorithm Process Simulation Fuzzy Logic Evaluation Product Demand Forecasting Using Genetic Programming Introduction Algorithm Description Chromosome Structure Fitness Evaluation Reproduction and Generation Evolution Experiments and Results Conclusion 281 References 282 Index 289