Computationally Intelligent Hybrid Systems

SEPPO J. OVASKA, DSc (Tech), is a professor in the Department of Electrical and Communications Engineering at Helsinki University of Technology, Finland. He is a senior member of the IEEE, and has published more than 180 papers in peer-reviewed journals and international conferences.

"…the only book to examine the practical issue involved in the creation of high-performance, cost-effective applications using a synthesis of neural networks, fuzzy systems and evolutionary computation with traditional computing methods." (International Journal of General Systems, June 2005) "...accessible for practical engineers and at the same time quite interesting for theoretical computer scientists...it will inspire more people to use the (currently under-utilized) fusion techniques." (Journal of Intelligent & Fuzzy Systems, Vol. 16, No. 3, 2005)"…these well-written papers serve to offer insight into the powerful combination of soft and hard computing that is now being applied…to real-world applications." (Computing Reviews.com, June 10, 2005)"This is the first book to treat the subject. With this work, the editor hopes to bridge the gap between the proponents of soft computing and hard computing." (E-STREAMS, May 2005)

• Contributors xvForeword xviiDavid B. FogelPreface xixEditor’s Introduction to Chapter 1 11 Introduction to Fusion of Soft Computing and Hard Computing 5Seppo J. Ovaska1.1 Introduction 51.1.1 Soft Computing 5a1.1.2 Fusion of Soft-Computing and Hard-Computing Methodologies 71.2 Structural Categories 91.2.1 Soft Computing and Hard Computing Are Isolated from Each Other 101.2.2 Soft Computing and Hard Computing Are Connected in Parallel 111.2.3 Soft Computing with Hard-Computing Feedback and Hard Computing with Soft-Computing Feedback 121.2.4 Soft Computing is Cascaded with Hard Computing or Hard Computing is Cascaded with Soft Computing 121.2.5 Soft-Computing-Designed Hard Computing and Hard-Computing-Designed Soft Computing 131.2.6 Hard-Computing-Augmented Soft Computing and Soft-Computing-Augmented Hard Computing 141.2.7 Hard-Computing-Assisted Soft Computing and Soft-Computing-Assisted Hard Computing 151.2.8 Supplementary Categories 161.2.9 General Soft-Computing and Hard-Computing Mapping Functions 191.3 Characteristic Features 191.3.1 Proportional Integral Derivative Controllers 201.3.2 Physical Models 201.3.3 Optimization Utilizing Local Information 211.3.4 General Parameter Adaptation Algorithm 221.3.5 Stochastic System Simulators 221.3.6 Discussion and Extended Fusion Schemes 221.4 Characterization of Hybrid Applications 241.5 Conclusions and Discussion 25References 27Editor’s Introduction to Chapter 2 312 General Model for Large-Scale Plant Application 35Akimoto Kamiya2.1 Introduction 352.2 Control System Architecture 362.3 Forecasting of Market Demand 372.4 Scheduling of Processes 392.4.1 Problem Decomposition 392.4.2 Hybrid Genetic Algorithms 422.4.3 Multiobjective Optimization 432.5 Supervisory Control 452.6 Local Control 472.7 General Fusion Model and Fusion Categories 492.8 Conclusions 51References 51Editor’s Introduction to Chapter 3 573 Adaptive Flight Control: Soft Computing with Hard Constraints 61Richard E. Saeks3.1 Introduction 613.2 The Adaptive Control Algorithms 623.2.1 Adaptive Dynamic Programming 633.2.2 Neural Adaptive Control 643.3 Flight Control 673.4 X-43A-LS Autolander 683.5 LOFLYTEw Optimal Control 733.6 LOFLYTEw Stability Augmentation 763.7 Design for Uncertainty with Hard Constraints 823.8 Fusion of Soft Computing and Hard Computing 853.9 Conclusions 85References 86Editor’s Introduction to Chapter 4 894 Sensorless Control of Switched Reluctance Motors 93Adrian David Cheok4.1 Introduction 934.2 Fuzzy Logic Model 954.2.1 Measurement of Flux Linkage Characteristics 954.2.2 Training and Validation of Fuzzy Model 974.3 Accuracy Enhancement Algorithms 1014.3.1 Soft-Computing-Based Optimal Phase Selection 1024.3.2 Hard-Computing-Based On-Line Resistance Estimation 1044.3.3 Polynomial Predictive Filtering 1054.4 Simulation Algorithm and Results 1084.5 Hardware and Software Implementation 1094.5.1 Hardware Configuration 1094.5.2 Software Implementation 1104.6 Experimental Results 1114.6.1 Acceleration from Zero Speed 1124.6.2 Low-Current Low-Speed Test 1134.6.3 High-Speed Test 1144.6.4 Test of Step Change of Load 1184.7 Fusion of Soft Computing and Hard Computing 1194.8 Conclusion and Discussion 122References 122Editor’s Introduction to Chapter 5 1255 Estimation of Uncertainty Bounds for Linear and Nonlinear Robust Control 129Gregory D. Buckner5.1 Introduction 1295.2 Robust Control of Active Magnetic Bearings 1305.2.1 Active Magnetic Bearing Test Rig 1325.3 Nominal H1 Control of the AMB Test Rig 1335.3.1 Parametric System Identification 1335.3.2 Uncertainty Bound Specification 1355.3.3 Nominal H1 Control: Experimental Results 1375.4 Estimating Modeling Uncertainty for H1 Control of the AMB Test Rig 1385.4.1 Model Error Modeling 1405.4.2 Intelligent Model Error Identification 1415.4.3 Uncertainty Bound Specification 1465.4.4 Identified H1 Control: Experimental Results 1475.5 Nonlinear Robust Control of the AMB Test Rig 1485.5.1 Nominal Sliding Mode Control of the AMB Test Rig 1485.5.2 Nominal SMC: Experimental Results 1505.6 Estimating Model Uncertainty for SMC of the AMB Test Rig 1515.6.1 Intelligent System Identification 1515.6.2 Intelligent Model Error Identification 1555.6.3 Intelligent SMC: Experimental Results 1565.7 Fusion of Soft Computing and Hard Computing 1595.8 Conclusion 162References 162Editor’s Introduction to Chapter 6 1656 Indirect On-Line Tool Wear Monitoring 169Bernhard Sick6.1 Introduction 1696.2 Problem Description and Monitoring Architecture 1726.3 State of the Art 1766.3.1 Monitoring Techniques Based on Analytical Models 1766.3.2 Monitoring Techniques Based on Neural Networks 1786.3.3 Monitoring Techniques Based on Fusion of Physical and Neural Network Models 1816.4 New Solution 1846.4.1 Solution Outline 1846.4.2 Physical Force Model at Digital Preprocessing Level 1856.4.3 Dynamic Neural Network at Wear Model Level 1876.5 Experimental Results 1896.6 Fusion of Soft Computing and Hard Computing 1926.7 Summary and Conclusions 194References 195Editor’s Introduction to Chapter 7 1997 Predictive Filtering Methods for Power Systems Applications 203Seppo J. Ovaska7.1 Introduction 2037.2 Multiplicative General-Parameter Filtering 2057.3 Genetic Algorithm for Optimizing Filter Tap Cross-Connections 2077.4 Design of Multiplierless Basis Filters by Evolutionary Programming 2117.5 Predictive Filters for Zero-Crossings Detector 2137.5.1 Single 60-Hz Sinusoid Corrupted by Noise 2137.5.2 Sequence of 49-, 50-, and 51-Hz Sinusoids Corrupted by Noise 2177.5.3 Discussion of Zero-Crossings Detection Application 2227.6 Predictive Filters for Current Reference Generators 2237.6.1 Sequence of 49-, 50-, and 51-Hz Noisy Sinusoids 2257.6.2 Sequence of 49-, 50-, and 51-Hz Noisy Sinusoids Corrupted by Harmonics 2297.6.3 Artificial Current Signal Corrupted by Odd Harmonics 2307.6.4 Discussion of Current Reference Generation Application 2327.7 Fusion of Soft Computing and Hard Computing 2337.8 Conclusion 234References 237Appendix 7.1: Coefficients of 50-Hz Sinusoid-Predictive FIR Filters 239Editor’s Introduction to Chapter 8 2418 Intrusion Detection for Computer Security 245Sung-Bae Cho and Sang-Jun Han8.1 Introduction 2458.2 Related Works 2478.2.1 Neural Computing 2488.2.2 Genetic Computing 2498.2.3 Fuzzy Logic 2518.2.4 Probabilistic Reasoning 2538.3 Intrusion Detection with Hybrid Techniques 2538.3.1 Overview 2548.3.2 Preprocessing with Self-Organizing Map 2548.3.3 Behavior Modeling with Hidden Markov Models 2568.3.4 Multiple Models Fusion by Fuzzy Logic 2598.4 Experimental Results 2618.4.1 Preprocessing 2618.4.2 Modeling and Intrusion Detection 2638.5 Fusion of Soft Computing and Hard Computing 2678.6 Concluding Remarks 268References 270Editor’s Introduction to Chapter 9 2739 Emotion Generating Method On Human–Computer Interfaces 277Kazuya Mera and Takumi Ichimura9.1 Introduction 2779.2 Emotion Generating Calculations Method 2799.2.1 Favorite Value Database 2809.2.2 Calculation Pleasure/Displeasure for an Event 2829.2.3 Favorite Value of Modified Element 2849.2.4 Experimental Result 2859.2.5 Complicated Emotion Allocating Method 2869.2.6 Dependency Among Emotion Groups 2949.2.7 Example of Complicated Emotion Allocating Method 2969.2.8 Experimental Results 2979.3 Emotion-Oriented Interaction Systems 2989.3.1 Facial Expression Generating Method by Neural Network 2989.3.2 Assign Rules to the Facial Expressions 3019.4 Applications of Emotion-Oriented Interaction Systems 3029.4.1 JavaFaceMail 3029.4.2 JavaFaceChat 3079.5 Fusion of Soft Computing and Hard Computing 3089.6 Conclusion 310References 311Editor’s Introduction to Chapter 10 31310 Introduction to Scientific Data Mining: Direct Kernel Methods and Applications 317Mark J. Embrechts, Boleslaw Szymanski, and Karsten Sternickel10.1 Introduction 31710.2 What is Data Mining? 31810.2.1 Introduction to Data Mining 31810.2.2 Scientific Data Mining 32010.2.3 The Data Mining Process 32110.2.4 Data Mining Methods and Techniques 32210.3 Basic Definitions for Data Mining 32310.3.1 The MetaNeural Data Format 32310.3.2 The “Standard Data Mining Problem” 32610.3.3 Predictive Data Mining 32910.3.4 Metrics for Assessing Model Quality 33310.4 Introduction to Direct Kernel Methods 33510.4.1 Data Mining and Machine Learning Dilemmas for Real-World Data 33510.4.2 Regression Models Based on the Data Kernel 33810.4.3 Kernel Transformations 33910.4.4 Dealing with Bias: Centering the Kernel 34010.5 Direct Kernel Ridge Regression 34210.5.1 Overview 34210.5.2 Choosing the Ridge Parameter 34310.6 Case Study #1: Predicting the Binding Energy for Amino Acids 34410.7 Case Study #2: Predicting the Region of Origin for Italian Olive Oils 34610.8 Case Study #3: Predicting Ischemia from Magnetocardiography 35010.8.1 Introduction to Magnetocardiography 35010.8.2 Data Acquisition and Preprocessing 35110.8.3 Predictive Modeling for Binary Classification of Magnetocardiograms 35110.8.4 Feature Selection 35810.9 Fusion of Soft Computing and Hard Computing 35910.10 Conclusions 359References 360Editor’s Introduction to Chapter 11 36311 World Wide Web Usage Mining 367Ajith Abraham11.1 Introduction 36711.2 Daily and Hourly Web Usage Clustering 37211.2.1 Ant Colony Optimization 37211.2.2 Fuzzy Clustering Algorithm 37411.2.3 Self-Organizing Map 37611.2.4 Analysis of Web Data Clusters 37711.3 Daily and Hourly Web Usage Analysis 37811.3.1 Linear Genetic Programming 37911.3.2 Fuzzy Inference Systems 38211.3.3 Experimentation Setup, Training, and Performance Evaluation 38711.4 Fusion of Soft Computing and Hard Computing 38911.5 Conclusions 393References 394Index 397About the Editor 409