Advanced Dynamic-System Simulation: Model Replication and Monte Carlo Studies [With CDROM]

2. Auflage 2013
ISBN: 978-1-118-39735-0
Verlag: WILEY

Buch, Englisch, 280 Seiten, CDROM, 004, Format (B × H): 163 mm x 239 mm, Gewicht: 590 g

A unique, hands-on guide to interactive modeling and simulation of engineering systems

This book describes advanced, cutting-edge techniques for dynamic system simulation using the DESIRE modeling/simulation software package. It offers detailed guidance on how to implement the software, providing scientists and engineers with powerful tools for creating simulation scenarios and experiments for such dynamic systems as aerospace vehicles, control systems, or biological systems.

Along with two new chapters on neural networks, Advanced Dynamic-System Simulation, Second Edition revamps and updates all the material, clarifying explanations and adding many new examples. A bundled CD contains an industrial-strength version of OPEN DESIRE as well as hundreds of program examples that readers can use in their own experiments. The only book on the market to demonstrate model replication and Monte Carlo simulation of real-world engineering systems, this volume:

* Presents a newly revised systematic procedure for difference-equation modeling
* Covers runtime vector compilation for fast model replication on a personal computer
* Discusses parameter-influence studies, introducing very fast vectorized statistics computation
* Highlights Monte Carlo studies of the effects of noise and manufacturing tolerances for control-system modeling
* Demonstrates fast, compact vector models of neural networks for control engineering
* Features vectorized programs for fuzzy-set controllers, partial differential equations, and agro-ecological modeling

Advanced Dynamic-System Simulation, Second Edition is a truly useful resource for researchers and design engineers in control and aerospace engineering, ecology, and agricultural planning. It is also an excellent guide for students using DESIRE.

Autoren/Hrsg.

Korn, Granino A.

Fachgebiete

• Mathematik | Informatik Mathematik Geometrie Dynamische Systeme
• Technische Wissenschaften Technik Allgemein Modellierung & Simulation

Inhaltsverzeichnis

PREFACE xiii

CHAPTER 1 DYNAMIC-SYSTEM MODELS AND SIMULATION 1

SIMULATION IS EXPERIMENTATION WITH MODELS 1

(a) Difference-Equation Models 2

(b) Differential-Equation Models 2

(c) Discussion 3

ANATOMY OF A SIMULATION RUN 8

(a) Euler Integration 10

(b) Improved Integration Rules 10

SIMPLE APPLICATION PROGRAMS 12

(a) Linear Oscillator 12

(b) Nonlinear Oscillator: Duffing's Differential Equation 14

INRODUCTION TO CONTROL-SYSTEM SIMULATION 21

(a) Guided Torpedo 24

(b) Complete Torpedo-Simulation Program 26

STOP AND LOOK 28

References 29

CHAPTER 2 MODELS WITH DIFFERENCE EQUATIONS, LIMITERS, AND SWITCHES 31

SAMPLED-DATA SYSTEMS AND DIFFERENCE EQUATIONS 31

(a) Introduction 31

(b) Difference Equations 31

(c) A Minefield of Possible Errors 32

(a) General Difference-Equation Model 32

(b) Simple Recurrence Relations 33

TWO MIXED CONTINUOUS/SAMPLED-DATA SYSTEMS 37

DYNAMIC-SYSTEM MODELS WITH LIMITERS AND SWITCHES 40

(a) Limiter Functions 40

(b) Switching Functions and Comparators 42

EFFICIENT DEVICE MODELS USING RECURSIVE ASSIGNMENTS 48

References 55

CHAPTER 3 FAST VECTOR-MATRIX OPERATIONS AND SUBMODELS 57

ARRAYS, VECTORS, AND MATRICES 57

(a) Improved Modeling 57

(b) Array Declarations, Vectors, and Matrices 57

(c) State-Variable Declarations 58

VECTORS AND MODEL REPLICATION 59

(a) Vector Assignments and Vector Expressions 59

(b) Vector Differential Equations 60

(c) Vector Sampled-Data Assignments and Difference Equations 60

(a) Definition 61

(b) Simple Example: Resonating Oscillators 61

(a) Definition 63

(b) Preview of Significant Applications 63

MORE VECTOR OPERATIONS 65

(a) Sums and DOT Products 65

(b) Euclidean, Taxicab, and Hamming Norms 65

(a) Maximum/Minimum Selection 66

(b) Masking Vector Expressions 66

VECTOR EQUIVALENCE DECLARATIONS SIMPLIFY MODELS 67

MATRIX OPERATIONS IN DYNAMIC-SYSTEM MODELS 67

(a) Matrix Expressions and DOT Products 68

(b) Matrix Differential Equations 68

(c) Matrix Difference Equations 69

VECTORS IN PHYSICS AND CONTROL-SYSTEM PROBLEMS 69

USER-DEFINED FUNCTIONS AND SUBMODELS 72

References 75

CHAPTER 4 EFFICIENT PARAMETER-INFLUENCE STUDIES AND STATISTICS COMPUTATION 77

MODEL REPLICATION SIMPLIFIES PARAMETER-INFLUENCE STUDIES 77

(a) Simple Repeated-Run Study 78

(b) Model Replication (Vectorization) 78

(a) Measures of System Performance 80

(b) Program Design 81

(c) Two-Dimensional Model Replication 81

(d) Cross-Plotting Results 82

(e) Maximum/Minimum Selection 83

(f) Iterative Parameter Optimization 83

STATISTICS 84

COMPUTING STATISTICS BY VECTOR AVERAGING 85

(a) Simple Probability-Density Estimate 86

(b) Triangle and Parzen Windows 87

(c) Computation and Display of Parzen-Window Estimates 88

REPLICATED AVERAGES GENERATE SAMPLING DISTRIBUTIONS 91

(a) Introduction 91

(b) Demonstrations of Empirical Laws of Large Numbers 93

(c) Counterexample: Fat-Tailed Distribution 95

RANDOM-PROCESS SIMULATION 95

(a) Modeling Continuous Noise 98

(b) Continuous Time Averaging 99

(c) Correlation Functions and Spectral Densities 100

SIMPLE MONTE CARLO EXPERIMENTS 100

References 106

CHAPTER 5 MONTE CARLO SIMULATION OF REAL DYNAMIC SYSTEMS 109

INTRODUCTION 109

REPEATED-RUN MONTE CARLO SIMULATION 109

VECTORIZED MONTE CARLO SIMULATION 113

Cannon Shot 113

SIMULATION OF NOISY CONTROL SYSTEMS 119

ADDITIONAL TOPICS 123

(a) Introduction 123

(b) Dynamic Systems with Random Perturbations 123

(c) Mean-Square Errors in Linearized Systems 124

References 125

CHAPTER 6 VECTOR MODELS OF NEURAL NETWORKS 127

ARTIFICIAL NEURAL NETWORKS 127

SIMPLE VECTOR ASSIGNMENTS MODEL NEURON LAYERS 130

(a) Pattern Normalization 131

(b) Contrast Enhancement: Softmax and Thresholding 131

(a) Computing Successive Neuron-Layer Outputs 132

(b) Input from Pattern-Row Matrices 133

(c) Input from Text Files and Spreadsheets 133

SUPERVISED TRAINING FOR REGRESSION 134

(a) Problem Statement 134

(b) Linear Mean-Square Regression and the Delta Rule 135

(c) Nonlinear Neuron Layers and Activation-Function Derivatives 136

(d) Error-Measure Display 136

(a) The Generalized Delta Rule 137

(b) Momentum Learning 139

(c) Simple Example 139

(d) The Classical XOR Problem and Other Examples 140

MORE NEURAL-NETWORK MODELS 140

(a) Basis-Function Expansion and Linear Optimization 142

(b) Radial Basis Functions 143

PATTERN CLASSIFICATION 146

(a) Simple Linear Classifiers 147

(b) Softmax Classifiers 148

(c) Backpropagation Classifiers 148

(d) Functional-Link Classifiers 149

(e) Other Classsifiers 149

(a) Classification Using an Empirical Database: Fisher's Iris Problem 149

(b) Image-Pattern Recognition and Associative Memory 151

PATTERN SIMPLIFICATION 155

(a) Bottleneck Layers and Encoders 156

(b) Principal Components 156

NETWORK-TRAINING PROBLEMS 157

(a) Introduction 157

(b) Adding Noise 158

(c) Early Stopping 158

(d) Regularization 159

UNSUPERVISED COMPETITIVE-LAYER CLASSIFIERS 159

(a) Template Patterns and Template Matrix 159

(b) Matching Known Template Patterns 160

(c) Template-Pattern Training 160

(d) Correlation Training 162

(a) Pattern Classification 164

(b) Vector Quantization 164

SUPERVISED COMPETITIVE LEARNING 167

EXAMPLES OF CLEARN CLASSIFIERS 168

(a) Image Recognition 168

(b) Fast Solution of the Spiral Benchmark Problem 169

References 174

CHAPTER 7 DYNAMIC NEURAL NETWORKS 177

INTRODUCTION 177

NEURAL NETWORKS WITH DELAY-LINE INPUT 178

(a) Linear Combiners 180

(b) One-Layer Nonlinear Network 181

(c) Functional-Link Network 181

(d) Backpropagation Network with Delay-Line Input 182

STATIC NEURAL NETWORKS USED AS DYNAMIC NETWORKS 183

RECURRENT NEURAL NETWORKS 185

(a) Conventional Model of a Jordan Network 185

(b) Simplified Jordan-Network Model 186

(c) Simplified Models for Other Feedback Networks 187

(a) Delay-Line Feedback 187

(b) Neural Networks with Both Input and Feedback Delay Lines 188

PREDICTOR NETWORKS 189

(a) Off-Line Prediction Using Stored Time Series 189

(b) Off-Line Training System for Online Predictors 189

(c) Example: Simple Linear Predictor 190

OTHER APPLICATIONS OF DYNAMIC NETWORKS 199

(a) Introduction 201

(b) Example: Program for Matching Narendra's Plant Model 201

MISCELLANEOUS TOPICS 204

References 204

CHAPTER 8 MORE APPLICATIONS OF VECTOR MODELS 207

VECTORIZED SIMULATION WITH LOGARITHMIC PLOTS 207

MODELING FUZZY-LOGIC FUNCTION GENERATORS 209

(a) Fuzzy Sets and Membership Functions 210

(b) Fuzzy Intersections and Unions 210

(c) Joint Membership Functions 213

(d) Normalized Fuzzy-Set Partitions 213

(a) Gaussian Bumps: Effects of Normalization 215

(b) Triangle Functions 215

(c) Smooth Fuzzy-Basis Functions 216

(a) Problem Statement 217

(b) Experiment Protocol and Rule Table 217

(c) DYNAMIC Program Segment and Results 220

PARTIAL DIFFERENTIAL EQUATIONS 221

(a) Introduction 221

(b) Using Differentiation Operators 221

(c) Numerical Problems 224

FOURIER ANALYSIS AND LINEAR-SYSTEM DYNAMICS 229

(a) Using the Index-Shift Operation with Analog Integration 232

(b) Linear Sampled-Data Systems 235

(c) Example: Digital Comb Filter 236

REPLICATION OF AGROECOLOGICAL MODELS ON MAP GRIDS 237

References 242

APPENDIX: ADDITIONAL REFERENCE MATERIAL 245

References 248

USING THE BOOK CD 251

INDEX 253

Über Autor(innen)

GRANINO A. KORN, PhD, is Professor of Electrical and Computer Engineering at the University of Arizona and a partner with G.A. and T.M. Korn Industrial Consultants, a company that designs systems for interactive simulation of dynamic systems and neural networks. He is the author of fifteen books, a Fellow of the IEEE, and the recipient of several awards for his work on computer simulation.