Soft computing and intelligent systems design : theory, tools, and applications/ Fakhreddine O. Karray and Clarence de Silva.

Includes bibliographical references and index.

Part I Fuzzy logic and fuzzy control
Chapter 1 Introduction to intelligent systems
and soft computing
Introduction
Intelligent systems
1.2.1 Machine intelligence
1.2.2 Meaning of intelligence
1.2.3 Dynamics of intelligence
1.2.4 Intelligent machines
Knowledge-based systems
1.3.1 Architectures of knowledge-based systems
1.3.2 Production systems
1.3.2.1 Reasoning strategies
1.3.2.2 Conflict resolution methods
1.3.3 Frame-based systems
1.3.4 Blackboard systems
1.3.5 Object-oriented programming
1.3.6 Expert systems
1.3.6.1 Development of an expert system
1.3.6.2 Knowledge engineering
1.3.6.3 Applications Knowledge representation and processing
1.4.1 Semantic networks
1.4.2 Crisp logic
1.4.2.1 Crisp sets
1.4.2.2 Operations of sets
1.4.2.3 Logic
1.4.2.4 Correspondence between sets and logic 32
1.4.2.5' Logic processing (reasoning and inference) 33
1.4.2.6 Laws of logic 33
1.4.2.7 Rules of inference 35
1.4.2.8 Propositional calculus and predicate calculus 37
1.5 Soft computing 38
1.5.1 Fuzzy logic 38
1.5.2 Neural networks
1.5.3 Genetic algorithms ^3
1.5.4 Probabilistic reasoning 44
1.5.5 Approximation and intelligence 45
1.5.6 Technology needs 30
Chapter 2 Fundamentals of fuzzy logic systems 57
2.1 Introduction 37
2.2 Background 38
2.2.1 Evolution of fuzzy logic 60
2.2.1.1 Popular applications 61
2.2.2 Stages of development of an intelligent product 63
2.2.3 Use of fuzzy logic in expert systems 64
2.3 Fuzzy sets ^3
2.3.1 Membership function 66
2.3.2 Symbolic representation 66
2.4 Fuzzy logic operations 68
2.4.1 Complement (negation, NOT) 69
2.4.2 Union (disjunction, OR) 70
2.4.3 Intersection (conjunction, AND) 72
2.4.4 Basic laws of fuzzy logic 73
2.5 Generalized fuzzy operations 76
2.5.1 Generalized fuzzy complement 76
2.5.2 Triangular norms 77
2.5.2.1 T-norm (generalized intersection) 77
2.5.2.2 S-norm or triangular conorm
(generalized union) 78
2.5.3 Set inclusion (A c B) 80
2.5.3.1 Grade of inclusion 81
2.5.4 Set equality (A = B) 82
2.5.4.1 Grade of equality 82
2.6 Implicatiion (if-then) 82
2.6.1 Considerations of fuzzy implication 83
2.7 Some definitions 89
2.7.1 Height of a fuzzy set 89
2.7.2 Support set 89
2.7.3 a-cut of a fuzzy set 90
2.7.4 Representation theorem 90
2.8 Fuzziness and fuzzy resolution 91
2.8.1 Fuzzy resolution 91
2.8.2 Degree of fuzziness 93
2.8.2.1 Measures of fuzziness 93
2.9 Fuzzy relations 97
2.9.1 Analytical representation of a fuzzy relation 99
2.9.2 Cartesian product of fuzzy sets 100
2.9.3 Extension principle 101
2.10 Composition and inference 105
2.10.1 Projection 105
2.10.2 Cylindrical extension 109
2.10.3 Join 115
2,10.4 Composition 116
2.10.4.1 Sup-product composition 116
2.10.5 Compositional rule of inference 117
2.10.5.1 Composition through matrix
multiplication 119
2.10.6 Properties of composition 121
2.10.6.1 Sup-t composition 121
2.10.6.2 Inf-s composition 121
2.10.6.3 Commutativity 121
2.10.6.4 Associativity 122
2.10.6.5 Distributivity 123
2.10.6.6 DeMorgan's Laws 123
2.10.6.7 Inclusion 123
2.10.7 Extension principle 125
2.11 Considerations of fuzzy decision-making 126
2.11.1 Extensions to fuzzy decision-making 127
Chapters Fuzzy logic control 137
3.1 Introduction
3.2 Background ^38
3.3 Basics of fuzzy control 141
3.3.1 Steps of fuzzy logic control 145
3.3.2 Composition using individual rules 146
3.3.3 Defuzzification 151
3.3.3.1 Centroid method 151
3.3.3.2 Mean of maxima method 151
3.3.3.3 Threshold methods 152
3.3.3.4 Comparison of the defuzzification methods 152
3.3.4 Fuzzification 153
3.3.4.1 Singleton method 154
3.3.4.2 Triangular function method 154
3.3.4.3 Gaussian function method 155
3.3.4.4 Discrete case of fuzzification 156
3.3.5 Fuzzy control surface 156
3.3.6 Extensions of Mamdani fuzzy control 162
3.4 Fuzzy control architectures 162
3.4.1 Hierarchical fuzzy systems 164
3.4.2 Hierarchical model 166
3.4.2.1 Feedback/filter modules 168
3.4.2.2 Functional/control modules 169
3.4.3 Effect of information processing 169
3.4.4 Effect of signal combination on fuzziness 171
3.4.5 Decision table approach for a fuzzy tuner 172
3.5 Properties of fuzzy control 180
3.5.1 Fuzzy controller requirements 180
3.5.2 Completeness 181
3.5.3 Continuity 181
3.5.4 Consistency 182
3.5.5 Rule validity 185
3.5.6 Rule interaction 185
3.5.7 Rule base decoupling 186
3.5.7.1 Decision-making through a
coupled rule base 187
3.5.7.2 Decision-making through an
uncoupled rule base 189
3.5.7.3 Equivalence condition 190
3.6 Robustness and stability 191
3.6.1 Fuzzy dynamic systems 191
3.6.2 Stability of fuzzy systems 192
3.6.2.1 Traditional approach to stability
analysis 192
3.6.2.2 Composition approach to stability
analysis 195
3.6.3 Eigen-fuzzy sets 200
3.6.3.1 Iterative method 200
Part lI Connectionist modeling and neural networks 221
Chapter 4 Fundamentals of artificial neural networks 223
4.1 Introduction 223
4.2 Learning and acquisition of knowledge 224
4.2.1 Symbolic learning 224
4.2.2 Numerical learning 224
4.3 Features of artificial neural networks 226
4.3.1 Neural network topologies 227
4.3.1.1 The feedforward topology 227
4.3.1.2 The recurrent topology 227
4.3.2 Neural network activation functions 228
4.3.3 Neural network learning algorithms 230
4.3.3.1 Supervised learning 230
4.3.3.2 Unsupervised learning 231
4.3.3.3 Reinforcement learning 232
4.4 Fundamentals of connectionist modeling 233
4.4.1 McCulloch-Pitts models 233
4.4.2 Perceptron 234
4.4.3 Adaline 243
4.4.4 Madaline 244
Chapter 5 Major classes of neural networks
5.1 Introduction
5.2 The multilayer perceptron
5.2.1 Topology
5.2.2 Backpropagation learning algorithm
5.2.3 Momentum
5.2.4 Applications and limitations of MLP
5.3 Radial basis function networks
5.3.1 Topology
5.3.2 Learning algorithm for RBF
5.3.3 Applications
5.4 Kohonen's self-organizing network
5.4.1 Topology
5.4.2 Learning algorithm
5.4.3 Applications
5.5 The Hopfield network
5.5.1 Topology
5.5.2 Learning algorithm
5.5.3 Applications of Hopfield networks
Industrial and commercial applications of ANN 281
5.6.1 Neural networks for process monitoring
and optimal control 282
5.6.2 Neural networks in semiconductor
manufacturing processes 282
5.6.3 Neural networks for power systems 284
5.6.4 Neural networks in robotics 285
5.6.5 Neural networks in communications 286
5.6.6 Neural networks in decision fusion and
pattern recognition 288
Chapter 6 Dynamic neural networks and their
applications to control and chaos prediction 299
6.1 Introduction 299
6.2 Background 300
6.2.1 Basic concepts of recurrent networks 300
6.2.2 The dynamics of recurrent neural networks 301
6.2.3 Architecture 301
6.3 Training algorithms 304
6.3.1 Backpropagation through time (BPTT) 304
6.3.2 Real-time backpropagation learning 305
6.4 Fields of applications of RNN 306
6.5 Dynamic neural networks for identification and control 307
6.5.1 Background 307
6.5.2 Conventional approaches for identification and control 308
6.5.2.1 Systems identification 310
6.5.2.2 Adaptive control 311
6.6 Neural network-based control approaches 313
6.6.1 Neural networks for identification 315
6.6.2 Neural networks for control 320
6.6.2.1 Supervised control 320
6.6.2.2 Inverse control 321
6.6.2.3 Neuro-adaptive control 322
6.7 Dynamic neural networks for chaos time series prediction
6.7.1 Background 324
6.7.2 Conventional techniques for chaos system prediction and control 324
6.7.3 Artificial neural networks for chaos prediction 325
6.7.3.1 Conventional feedforward networks 325
6.73.2 Recurrent neural networks (RNNs)-based predictors 327
Chapter 7 Neuro-fuzzy systems 337
7.1 Introduction 337
7.2 Background 338
7.3 Architectures of neuro-fuzzy systems 339
7.3.1 Cooperative neuro-fuzzy systems 340
7.3.1.1 Neural networks for determining
membership functions 341
7.3.1.2 Adeli-Hung algorithm (AHA) 342
7.3.1.3 Learning fuzzy rules using neural networks 343
7.3.1.4 Learning in fuzzy systems using neural networks 344
7.3.1.5 Identifying weighted fuzzy rules using neural networks
7.3.2 Neural network-driven fuzzy reasoning 344
7.3.3 Hybrid neuro-fuzzy systems 345
7.3.3.1 Architecture of hybrid neuro-fuzzy systems 346
7.3.3.2 Five-layer neuro-fuzzy systems 347
7.3.3.3 Four-layer neuro-fuzzy systems (ANFIS) 350
7.3.3.4 Three-layer neuro-fuzzy approximator 350
7.4 Construction of neuro-fuzzy systems 355
7.4.1 Structure identification phase 355
7.4.1.1 Grid-type partitioning 355
7.4.1.2 Clustering 356
7.4.1.3 Scatter partitioning 357
7 4^2 Parameter learning phase 357
7.4.2.1 The backpropagation learning algorithm 358
7.4.2.2 Hybrid learning algorithms 359
Part III Evolutionary and soft computing 363
Chapters Evolutionary computing 365
8.1. Introduction 365
8.2 Overview of evolutionary computing 366
8.2.1 Evolutionary programming 369
8.2.2 Evolutionary strategies 370
8.2.3 Genetic programming 370
8.2.4 Genetic algorithms 371
8.3 Genetic algorithms and optimization 372
8.3.1 Genotype 373
8.3.2 Fitness function 374
8.4 The schema theorem: the fundamental theorem of genetic algorithms 375
8.5 Genetic algorithm operators 376
8.5.1 Selection 377
8.5.2 Crossover 377
8.5.3 Mutation 378
8.5.4 Mode of operation of GAs 378
8.5.5 Steps for implementing GAs 381
8.5.6 Search process in GAs 381
8.6 Integration of genetic algorithms with neural networks 388
8.6.1 Use of GAs for ANN input selection 388
8.6.2 Using GA for NN learning 389
8.7 Integration of genetic algorithms with fuzzy logic 390
8.8 Known issues in GAs 391
8.8.1 Local minima and premature convergence 391
8.8.2 Mutation interference 392
8.8.3 Deception 392
8.8.4 Epistasis 392
8.9 Population-based incremental learning 393
8.9.1 Basics of PBIL 393
8.9.2 Generating the population 393
8.9.3 PBIL algorithm 394
8.9.4 PBIL and learning rate 395
8.10 Evolutionary strategies 395
8.11 ES applications 400
8.11.1 Parameter estimation 400
8.11.2 Image processing and computer vision systems 400
8.11.3 Task scheduling hy ES 400
8.11.4 Mobile manipulator path planning by ES 400
8.11.5 Car automation using ES 401
Part IV Applications and case studies 405
Chapter 9 Soft computing for smart machine design 407
9.1 Introduction 407
9.1.1 Intelligent machines 408
9.1.2 Intelligent control 408
9.1.3 Hierarchical architecture 409
9.1.4 Development steps 411
9.2 Controller tuning 413
9.2.1 Problem formulation 414
9.2.1.1 Rule base 415
9.2.1.2 Compositional rule of inference 417
9.2.2 Tuning procedure 418
9.2.2.1 Rule dissociation 418
9.2.2.2 Resolution relations 419
9.2.2.3 Tuning inference 421
9.2.2.4 Accuracy versus fuzzy resolution 421
9.2.3 Illustrative example 422
9.2.3.1 Resolution relation 423
9.2.3.2 Stability region 426
9.2.3.3 Tuning results 427
9.3 Supervisory control of a fish processing machine 427
9.3.1 Machine features 430
9.3.2 Supervisory control system 432
9.3.3 Information preprocessing 435
9.3.3.1 Image preprocessing 435
9.3.3.2 Servomotor response preprocessing 436
9.3.3.3 Cutter load preprocessing 439
9.3.3.4 Conveyor speed preprocessing 443
9.3.4 Knowledge-based decision-making 443
9.3.4.1 Knowledge acquisition 444
9.3.4.2 Decision-making 446
9.3.4.3 Servo tuning 449
9.3.4.4 Product quality assessment 452
9.3.4.5 Machine tuning 453
9.3.5 System implementation 453
9.3.5.1 System modules 455
9.3.5.2 User interface of the machine 456
9.3.6 Performance testing 457
9.3.6.1 Servomotor tuning examples 457
9.3.6.2 Machine tuning example 461
9.3.6.3 Product quality assessment
Chapter 10 Tools of soft computing in real-world applications
Case study 1: Expert parameter tuning of DC motor controller Case study 2: Stabilizing control of a high-order power system
by neural adaptive feedback linearization 497
Case study 3; Soft computing tools for solving a class of facilities layout planning problem 510
Case study 4: Mobile position estimation using an RBF network in CDMA cellular systems 522
Case study 5: Learning-based resource optimization in ATM
networks