Systems and control theory has experienced significant development in the past few decades. New techniques have emerged which hold enormous potential for industrial applications, and which have therefore also attracted much interest from academic researchers. However, the impact of these developments on the process industries has been limited.
The purpose of Multivariable System Identification for Process Control is to bridge the gap between theory and application, and to provide industrial solutions, based on sound scientific theory, to process identification problems. The book is organized in a reader-friendly way, starting with the simplest methods, and then gradually introducing more complex techniques. Thus, the reader is offered clear physical insight without recourse to large amounts of mathematics. Each method is covered in a single chapter or section, and experimental design is explained before any identification algorithms are discussed. The many simulation examples and industrial case studies demonstrate the power and efficiency of process identification, helping to make the theory more applicable. Matlab M-files, designed to help the reader to learn identification in a computing environment, are included.


For academic and industrial engineers in all areas of systems and control engineering, especially process control.

Table of Contents

Chapter headings. Foreward. Preface. Symbols and Abbreviations. Introduction. What is process identification? The hierarchy of modern automation systems. Multivariable model-based process control. Outline of the book. Models of Dynamic Processes and Signals. SISO Continuous-time models. SISO Discrete-time models. MIMO models. Models of signals. Linear processes with disturbances. Nonlinear models. Identification Test Design and Data Pretreatment. Controller configuration; selection of MV's, CV's and DV's. Preliminary process tests. Test signals for final test, persistent excitation. Test for model identification, final test. Sampling frequency and anti-aliasing filter. Pre-treatment of data. When is excitation allowed? Concluding remarks. Identification by the Least Squares Method. The principle of least-squares. Estimating models of linear processess. Industrial case studies. Properties of the least-squares estimator. Conclusions. Extensions of the Least-Squares Method. Modifying the frequency weighting by prefiltering. Output error method. Instrumental Variable (IV) methods. Prediction error methods. User's choice in identification for control. More on order/structure selection. Model validation. Identifying the glass tube drawing process. Recursive parameter estimation. Conclusions and discussion. Asymptotic Method; SISO Case. The Asymptotic theory. Optimal test signal spectrum for control. Parameter estimation and order selection. Model validation using upper error bound. Simulation studies and conclusion. Asymptotic Method; MIMO Case. MIMO version of the asymptotic theory. Asymptotic method. Identification of the glass tube drawing processes. Conclusions. Subspace Model Identification of MIMO Processes. Introduction. Definition of the state space identification problem. Definition of the data equation. Analy


No. of pages:
© 2001
Elsevier Science
eBook ISBN:
Print ISBN:
Print ISBN:

About the author

Y. Zhu

Affiliations and Expertise

Faculty EE, Eindhoven University of Technology, P O Box 513, 5600 MB Eindhoven, The Netherlands.