Pearson R.K. Discrete-time Dynamic Models

Oxford: Oxford University Press, 1999. — 481 p.This book is intended as a reference for anyone interested in the development of discrete-time dynamic models to approximate or predict the evolution of complex physical systems. Because I spent 17 years with the DuPont company, most of the examples discussed in this book have a strong process control flavor, but the ideas presented here certainly have much wider applicability, particularly in areas like nonlinear digital signal processing, model-based fault detection, or the newly emerging field of mechatronics, involving integrated microcomputer-based control of electromechanical systems. The general subject of model development is a large one, and this book makes no attempt to cover all of it. Instead, the principal focus here is on the first step in the model development process: that of model structure selection. In fact, the original motivation for this book was the observation that many papers appearing in the literature proceeded roughly as follows: "We investigate the utility of nonlinear model structure X in developing controllers for physical system P..." The question of why this particular model structure was chosen—or indeed, what other model structures might be reasonable alternatives—was too seldom addressed, yet these questions seemed in many ways the most difficult. The intent of this book is to provide some useful guidance in dealing with these questions.
More specifically, this book has two primary objectives: first, to describe some of the useful model structures that have been proposed in the process control, time-series analysis and digital signal processing literature and second, to describe some of the important qualitative behavior of these model structures to provide a rational basis for matching classes of potentially useful mathematical models with specific applications.