Guo / Wang Stochastic Distribution Control System Design

A Convex Optimization Approach
1. Auflage 2010
ISBN: 978-1-84996-030-4
Verlag: Springer
Format: PDF
Kopierschutz: 1 - PDF Watermark

E-Book, Englisch, 196 Seiten

Reihe: Advances in Industrial Control

A recent development in SDC-related problems is the establishment of intelligent SDC models and the intensive use of LMI-based convex optimization methods. Within this theoretical framework, control parameter determination can be designed and stability and robustness of closed-loop systems can be analyzed. This book describes the new framework of SDC system design and provides a comprehensive description of the modelling of controller design tools and their real-time implementation. It starts with a review of current research on SDC and moves on to some basic techniques for modelling and controller design of SDC systems. This is followed by a description of controller design for fixed-control-structure SDC systems, PDF control for general input- and output-represented systems, filtering designs, and fault detection and diagnosis (FDD) for SDC systems. Many new LMI techniques being developed for SDC systems are shown to have independent theoretical significance for robust control and FDD problems.

Author 1: Professor Lei Guo:2003-present: Full professor with research activities on stochastic control, nonlinear control, filter design and fault detection, in Institute of Automation, BUAA, Beijing, P R China. He is also affiliated as a full professor with Research Institute of Automation, Southeast University, China.2002-2003: Research fellow at Dept. Paper Science, UMIST, Manchester, UK.2001-2002: Research associate in Department of Automobile and Aeronautical Engineering, Loughborough University, UK;2000-2001: Research associate in Department of Mechanical Engineering, Glasgow University, UK;1999-2000: Postdoctoral research fellow in IRCCyN, CNRS, Nantes, France, sponsored by Pays de la Loire project.Following Professor Guo's previous work on stochastic distribution control at UMIST, his recent research is mainly focused on the new developments of non-Gaussian filtering algorithms for signal processing and the shape control of stochastic distributions using LMIs. This includes the developments of nonlinear observers and LMI techniques for adaptive tuning rules for nonlinear systems.Professor Lei Guo will (with John Bailleul of Boston University) be general chair of the IEEE Conference on Decision and Control and Chinese Control Conference being held jointly in Shanghai in December 2009.Author 2: Professor Hong Wang:1982: Received the BSc (first class) degree in Electrical Engineering from Huainan University of Technology, Anhui, P.R. China1984: Received the MEng (first class) in Automatic Control from Huazhong Univ.Science & Tech, Wuhan, P.R. China1987: Received the PhD degree in Power Systems Automation from Huazhong Univ. Science & Tech., Wuhan, P.R. China, received an outstanding PhD thesis award and three best papers awards.2004-present: Professor in Process Control, Director of the Control Systems Centre, School of Electrical and Electronics Engineering, The University of Manchester (formally UMIST), Manchester, working on the control of stochastic distributions for stochastic systems, fault diagnosis and fault tolerant control andcomplex systems modeling.2002-2003: Professor in Process Control, Control Systems Centre, UMIST, Manchester.1999-2002: Reader in Process Control at UMIST, working on stochastic distribution control, fault diagnosis and complex systems modeling.1997-1999: Senior lecturer in Process Control at UMIST, working on stochastic distribution control, fault diagnosis and complex systems modeling.Prof Wang is a fellow of IEE, fellow of InstMC and IEEE Senior Member, and acted as an associate editor for the leading control theory journal (IEEE Transactions on Automatic Control), board member for 4 international journals, and a member of the IFAC Safeprocess Committee, the IFAC Adaptive and Learning Systems Commitee and a member of the IFAC Stochastic Systems Committee. He is the originator of probability density function shape control and has published 190 papers in international journals and conferences (25 invited papers). He is the leading author of 3 books.His research activites have been focused on: i. stochastic distribution control and filtering of general non-Gaussian dynamic stochastic systems and closed-loop entropy minimization; ii. fault detection, diagnosis and fault-tolerant control for dynamic systems; iii. artificial-neural-network-based control systems design and applications to complex systems such as papermaking, combusion systems and systems biology; and iv. plant-wide modeling, fault diagnosis and optimization.

Autoren/Hrsg.

Guo, Lei

Wang, Hong

Inhaltsverzeichnis

1;Foreword;9
2;Preface;11
3;Contents;13
4;Abbreviations and Notation;17
5;Chapter 1 Developments in Stochastic Distribution Control Systems;19
5.1;1.1 Introduction;19
5.1.1;1.1.1 Paper Web Formation Systems;21
5.1.2;1.1.2 Flame Distribution Control;22
5.1.3;1.1.3 Challenging Issues;24
5.2;1.2 Stochastic Distribution Control when Output PDFs are Measurable;26
5.3;1.3 Stochastic Distribution Control when Output PDFs are Unmeasurable;29
5.4;1.4 Minimum Entropy Control;30
5.5;1.5 Stochastic Distribution Filtering Design;31
5.6;1.6 Conclusions;32
6;Part I - Structural Controller Design for Stochastic Distribution Control Systems;33
6.1;Chapter 2 Proportional Integral Derivative Control for Continuous-time Stochastic Systems;35
6.1.1;2.1 Introduction;35
6.1.2;2.2 Problem Formulation;36
6.1.2.1;2.2.1 Model Representation;36
6.1.2.2;2.2.2 Pseudo-PID Controller Structure;37
6.1.3;2.3 Pseudo-PID Controller Design;39
6.1.3.1;2.3.1 Solvability Condition;39
6.1.3.2;2.3.2 Feasible Design Procedures;41
6.1.3.3;2.3.3 Robust Pseudo-PID Controller;44
6.1.4;2.4 Simulations;45
6.1.5;2.5 Conclusions;47
6.2;Chapter 3 Constrained Continuous-time Proportional Integral Derivative Control Based on Convex Algorithms;49
6.2.1;3.1 Introduction;49
6.2.2;3.2 Problem Formulation;50
6.2.2.1;3.2.1 B-spline Expansion and Dynamic Weight Model;50
6.2.2.2;3.2.2 Generalized PID Controller Design;52
6.2.3;3.3 Constrained PID Controller Design Based on LMIs;53
6.2.3.1;3.3.1 Peak-to-peak Performance Control;53
6.2.3.2;3.3.2 Peak-to-peak Tracking Performance;56
6.2.3.3;3.3.3 Peak-to-peak Constrained Tracking Control;58
6.2.4;3.4 Simulations;59
6.2.5;3.5 Conclusions;60
6.3;Chapter 4 Constrained Discrete-time Proportional Integral Control Based on Convex Algorithms;62
6.3.1;4.1 Introduction;62
6.3.2;4.2 Problem Formulation;62
6.3.2.1;4.2.1 B-spline Expansion and Discrete-time Weight Dynamical Model;62
6.3.2.2;4.2.2 Target Weight Model and Discrete-time PI Controller;64
6.3.3;4.3 Robust Constrained Tracking Control;65
6.3.3.1;4.3.1 Solvability for Peak-to-peak Performance;65
6.3.3.2;4.3.2 Peak-to-peak Tracking Performance;68
6.3.3.3;4.3.3 Constrained Peak-to-peak Tracking Control;70
6.3.4;4.4 Simulations;71
6.3.5;4.5 Conclusions;72
7;Part II - Two-step Intelligent Optimization Modeling and Control for Stochastic Distribution Control Systems;75
7.1;Chapter 5 Adaptive Tracking Stochastic Distribution Control for Two-step Neural Network Models;78
7.1.1;5.1 Introduction;78
7.1.2;5.2 Output PDF Model Using B-spline Neural Network;79
7.1.3;5.3 Dynamic Neural Network Identification;80
7.1.4;5.4 Adaptive Tracking Control forWeight Reference Model;83
7.1.5;5.5 Illustrative Examples;86
7.1.5.1;5.5.1 Example 1;86
7.1.5.2;5.5.2 Example 2;87
7.1.6;5.6 Conclusions;91
7.2;Chapter 6 Constrained Adaptive Proportional Integral Tracking Control for Two-step Neural Network Models with Delays;95
7.2.1;6.1 Introduction;95
7.2.2;6.2 Output PDF Model Using B-spline Neural Network;96
7.2.3;6.3 Time Delay DNNs Identification;97
7.2.4;6.4 Constrained PI Tracking Control forWeight Vector;101
7.2.5;6.5 Illustrative Examples;105
7.2.5.1;6.5.1 Example 1;105
7.2.5.2;6.5.2 Example 2;107
7.2.6;6.6 Conclusions;111
7.3;Chapter 7 Constrained Proportional Integral Tracking Control for Takagi-Sugeno Fuzzy Model;114
7.3.1;7.1 Introduction;114
7.3.2;7.2 Problem Statement and Preliminaries;114
7.3.2.1;7.2.1 Output PDF Model Using B-spline Expansion;114
7.3.2.2;7.2.2 PI Controller Design Based on T-S Fuzzy Model;116
7.3.3;7.3 Main Results;117
7.3.3.1;7.3.1 Stability Analysis with L1 Measure Index;117
7.3.3.2;7.3.2 Peak-to-peak Tracking Performance;119
7.3.3.3;7.3.3 Constrained Peak-to-peak Tracking Control;121
7.3.4;7.4 An Illustrative Example;122
7.3.5;7.5 Conclusions;123
8;Part III - Statistical Tracking Control – Driven by Output Statistical Information Set;125
8.1;Chapter 8 Multiple-objective Statistical Tracking Control Based on Linear Matrix Inequalities;127
8.1.1;8.1 Introduction;127
8.1.2;8.2 Problem Formulation and Preliminaries;127
8.1.3;8.3 Formulation for the H2/H8 Tracking Problem;129
8.1.4;8.4 Illustrative Examples;134
8.1.5;8.5 Conclusions;137
8.2;Chapter 9 Adaptive Statistical Tracking Control Based on Two-step Neural Networks with Time Delays;141
8.2.1;9.1 Introduction;141
8.2.2;9.2 Problem Formulation and Preliminaries;141
8.2.3;9.3 Time Delay DNNs Identification;142
8.2.4;9.4 Variable Structure Tracking Control forWeight Vectors;144
8.2.5;9.5 An Illustrative Example;148
8.2.6;9.6 Conclusions;150
9;Part IV - Fault Detection and Diagnosis for Stochastic Distribution Control Systems;154
9.1;Chapter 10 Optimal Continuous-time Fault Detection Filtering;156
9.1.1;10.1 Introduction;156
9.1.2;10.2 Problem Formulation;157
9.1.3;10.3 Main Result;158
9.1.4;10.4 Numerical Examples;163
9.1.5;10.5 Conclusions;166
9.2;Chapter 11 Optimal Discrete-time Fault Detection and Diagnosis Filtering;167
9.2.1;11.1 Introduction;167
9.2.2;11.2 Formulation of the FDD Problem with Guaranteed Cost Performance;168
9.2.3;11.3 The Optimal Fault Detection Filter Design;170
9.2.4;11.4 Fault Diagnosis Filter Design;172
9.2.5;11.5 Simulation Examples;175
9.2.6;11.6 Conclusions;176
10;Part V - Conclusions;180
10.1;Chapter 12 Summary and Potential Applications;181
10.1.1;12.1 Summary;181
10.1.2;12.2 Potential Applications of PDF Control;182
10.1.2.1;12.2.1 PDF Control in Data-driven Modeling;182
10.1.2.2;12.2.2 PDF Control in Data Dimension Reduction;183
10.1.2.3;12.2.3 PDF Control in Multi-scale Plant-wide Modeling for Fault Diagnosis and Fault Tolerant Control;183
11;References;186
12;Index;199

Leseproben

"Part IV Fault Detection and Diagnosis for Stochastic Distribution Control Systems (p. 136-137)

Safety and reliability are of paramount importance for practical processes. As a result, Fault detection and diagnosis (FDD) theory has been developed in the past three decades (see [6, 41, 48, 84–86, 121, 192, 197] and [201] for surveys). For stochastic systems, the approaches used thus far include the system identi?cation technique [84] and statistical approaches based on the Bayesian theorem, likelihood methods, and hypothesis test techniques [6]. Also, ?lters or observers have been widely used to generate the residual signal for fault detection and estimation purposes (see [20, 48]). Generally, the observer-based or ?lter-based FDD methodologies have been developed along with the observer or ?lter design theory, and many of them have been applied to practical processes successfully.

Until now, most of the existing ?lter-based or observer-based FDD results for stochastic systems have been concerned only with Gaussian variables, where mean or variance was the objective for optimization. However, non-Gaussian variables exist widely in many complex stochastic systems due to the presence of nonlinearities, for which the classical ?ltering approaches are insuf?cient, especially for variables with asymmetric and multiple-peak stochastic distributions (see [62, 63, 76] and [157]).

This is particularly true for stochastic distribution systems where the output is represented as the measured PDFs. Indeed, along with the continued and fast improvements of advanced instruments and data processing techniques, in practice the measurements for general ?lter design can be the stochastic distributions of the system output rather than its values. Typical examples include the particle-size distribution systems in chemical processing and the combustion ?ames distribution process [63, 157, 161]. As such, new ?lter- or observer-based FDD design algorithms are required for general stochastic systems using the output stochastic distributions.

For the SDC systems discussed in Part II, two main procedures were included [157]. The ?rst is to use a B-spline expansion technique to model the measurable output PDFs, where PDFs can be represented by the weight dynamics corresponding to some basis functions. The second step is to establish a further dynamical model relating the input and dynamical weight vector of the B-spline expansion. In this part, square root B-spline expansions are adopted for the measured output PDFs of general stochastic systems, and nonlinear weight dynamical models are considered instead of linear ones.

In Chapter 10, a new fault detection method using an augmented Lyapunov functional approach is presented. With the guaranteed cost performance index used as the objective function, an optimization algorithm with LMI constraint is applied to minimize the threshold value. This can improve residual signal sensitivity to the faults. Up to now few results have been seen focusing on FDD problems for the discrete-time SDC systems, where the nonlinearity and time delays are included and the measurement will be a nonlinear function of the state."