ISBN: 3790814571
TITLE: Evolving Rule-Based Models
AUTHOR: Angelov
TOC:

Preface vii
1 Introduction 1
1.1 Flexible Models - An Opportunity for Control Theory 1
1.2 Flexible Models and their Identification 2
1.2.1 Expert Knowledge and Parameters Tuning 3
1.2.2 Data-driven Techniques 4
1.2.3 Precision and Transparency 5
1.2.4 The Need for On-line Algorithms 6
1.3 Intelligent Adaptive Systems - A Higher Level of Control 7
1.4 Structure of the Book 7
Part I System Modelling: Basic Principles 11
2 Conventional Models 13
2.1 First Principles Models 13
2.1 l Heating/cooling Coil Model 14
2.1.2 Fermentation Process Model 15
2.2 Black-box Models 17
2.2.1 Linear Black-box Models 17
2.2.2 Polynomial Models 18
2.2.3 Regression Models 19
2.2.4 Neural Networks 20
2.2.4.1 Radial-basis Functions (RBF) Neural Networks 21
2.2.4.2 Hybrid NN-First Principles Model of a Fermentation Process 22
2.3 Conclusion 22
3 Flexible Models 25
3.1 Fuzzy Set Theory: Basic Introduction 26
3.1.1 Fuzzy Set Definition 27
3.1.2 Basic Operations over Fuzzy Sets 28
3.1.2.1 T-norms 28
3.1.2.2 S-norms 29
3.1.2.3 Negation 29
3.1.2.4 De-fuzzification 30
3.1.2.5 Degree of Similarity Between Fuzzy Sets 31
3.2 Models with Flexible Parameters or (In)equalities 31
3.2.1 Models with Flexible Parameters 31
3.2.2 Models with Flexible (In)equalities 32
3.3 Flexible Rule-based Models 34
3.3.1 Flexible Relational Models 35
3.3.2 Mamdani type Models 36
3.3.3 Takagi-Sugeno type Models 37
3.4 Conclusion 41
Part II FLEXIBLE MODELS IDENTIFICATION 43
4 Non-linear Approach to (Off-line) Identification of Flexible Models 47
4.1 Identification Problem Formulation 47
4.1.1 Identification Criteria 48
4.2 GA -Brief Introduction 49
4.3 Centre-of-Gravity-based Crossover Operator 52
4.3.1 CoG-based Cross-over Operator - How It Works 54
4.3.2 CoG-based Operator - Why It Works 54
4.3.3 Test Examples 57
4.4 Encoding and Decoding Indices of Flexible Rules and Linguistic Terms 57
4.4.1 Encoding Procedure 59
4.4.2 Decoding a Flexible Rule 59
4.5 Algorithm of the Non-linear Approach 62
4.6 Conclusion 63
5 Quasi-linear Approach to FRB Models (Off-line) Identification 67
5.1 Data Space Clustering 67
5.2 Subtractive Clustering 71
5.3 Parameters (of the Consequent Part) Estimation 72
5.4 Flexible Rule-based Model Refinements 73
5.4.1 Model Structure Simplification 73
5.4.2 Model Parameter's Refinement/Optimisation 74
5.5 Algorithm for (Off-line) Quasi-linear Identification of FRB Models 75
5.6 Conclusion 75
6 Intelligent and Smart Adaptive Systems 79
6.1 Intelligent Systems 79
6.1.1 Loose Definition 79
6.1.2 Problems 80
6.1.3 Importance 80
6.1.4 Specifics 81
6.2 Smart Adaptive Systems 81
6.2.1 The Issue of Smart Adaptive Systems 81
6.2.2 Features of a Smart Adaptive System 82
6.2.3 Practical Implications 82
6.2.4 Intelligent Indoor Climate Control System 83
6.3 Conclusion 84
7 On-line Identification of Flexible TSK-type Models 87
7.1 The Concept 87
7.2 Basic Phases of the Procedure 88
7.3 Potentials Up-date in On-line Mode 89
7.4 Rule-base Innovation and Modification Mechanism 92
7.5 Parameters Up-date 96
7.6 FRB Model Up-grade; 'Learning trough Experience' 99
7.7 Rule Structure and Parameters Tuning and Refinement 99
7.7.1 Similarity-based Simplification of Linguistic Terms 100
7.7.2 Parameters Refinement (Tuning) by Non-linear Optimisation 101
7.8 Flow-chart of the Algorithm 102
7.9 eR Control Algorithm 104
7.10 Conclusion 108
Part III Engineering Applications 111
8 Modelling Indoor Climate Control Systems 115
8.1 Modelling Components of HVAC Systems 116
8.1.1 Heating/cooling Coil Modelling 117
8.1.1.1 Modelling Outlet (from the Coil) Air Temperature 119
8.1.1.2 Modelling Heat Transfer in a Heating/Cooling Coil 123
8.1.2 Ducted Fan Modelling 124
8.1.3 Modelling Efficiency of Boilers: Hybrid Model Approach 130
8.2 Modelling the Thermal Load of a Building 131
8.3 Learning trough Experience (VL Strategy) 138
8.4 On-line Modelling Dynamical Signals 142
8.5 Model Simplification by Linguistic Term's Reduction 147
8.6 Refinement of Linguistic Terms' Parameters 151
8.7 Testing the New CoG-based Crossover Operator 154
8.7.1 Numerical Test Functions (NF1-NF5) 154
8.7.1.1 DeJong's Function (NFI) 155
8.7.1.2 Rastrigin's Function (NF2) 156
8.7.1.3 Sum of Different Powers (NF3) 157
8.7.1.4 Schwefel's Function (NF4) 158
8.7.1.5 Griewangk's Function (NF5) 159
8.7.2 Optimal Scheduling of a Hollow Core Ventilated Slabs (AC) 160
8.8 ICC System - Open or Closed Loop? A System Approach 164
8.9 Conclusion 166
9 On-line Modelling of Fermentation Processes 169
9.1 Bio-processes - Specifics of their Modelling 169
9.2 eR Model of a Fermentation Process 171
9.2.1 Lactose Oxidation Process Specifics 171
9.2.2 Experimental Data 172
9.2.3 Modelling the Process 173
9.2.3.1 First Principles-based Model 173
9.2.3.2 eR Model 174
9.2.3.3 Analysis of the Results 178
9.3 Conclusion 178
10 Intelligent Risk Assessment 181
10.1 Application of eR Models in Creditworthiness Assessment 181
10.1.1 Creditworthiness Assessment: Problem Specifics 182
10.1.2 Flexible Rule-based System 182
10.1.3 Credit Risk Assessment by a Flexible Rule-based System 182
10.2 Intelligent Evolving System for Risk Assessment in Civil Aviation 187
10.2.1 Specifics and Importance of the Problem 187
10.2.2 Intelligent Technologies for Risk Assessment in Civil Aviation 188
10.3 InTEND: Evolving Distributed Intelligence System for Evaluation of TENDering 190
10.4 Conclusion 191
11 Conclusions 193
References 199
Index 209
END
