Identification of Multivariable Industrial Processes: For Simulation, Diagnosis and Control

ISBN-13: 9781447120605 / Angielski / Miękka / 2011 / 187 str.

Yucai Zhu; Ton Backx

Opis

Identification of Multivariable Industrial Processes presents a unified approach to multivariable industrial process identification. It concentrates on industrial processes with reference to model applications. The areas covered are experiment design, model structure selection, parameter estimation as well as error bounds of the transfer function. This publication is intended to fill the gap between modern systems and control theory and industrial application. It is based on the results of 10 years of research and application experiences. The theories and models discussed are fully explained and illustrated with case studies. At an early stage the reader is introduced to real applications.

Spis treści

1 Introduction.- 1.1 Some Preliminary Concepts.- 1.2 Digital Control of Industrial Processes.- 1.3 Outline of the Book.- 2 Linear Models of Dynamic Processes and Signals.- 2.1 SISO Continuous-Time Models.- 2.2 SISO Discrete-Time Models.- 2.3 MIMO Models.- 2.4 Models of Signals.- 2.5 Linear Processes with Disturbances; Conclusion.- 3 Identification Experiments and Data Pre-treatmet.- 3.1 Selection of Inputs/Outputs and Preliminary Experiments.- 3.2 Experiment for Model Estimation.- 3.3 Pre-treatment of Data.- 3.4 Conclusions.- 4 Identification by the Least-Squares Method.- 4.1 The Principle of Least-Squares.- 4.2 Estimating Models of Linear Processes.- 4.2.1 Finite Impulse Response (FIR) Models.- 4.2.2 Transfer Operator Models.- 4.2.3 Model Validation and Order Selection; A Simulation Approach.- 4.3 Two Industrial Case Studies.- 4.3.1 Identification and Control of a Single-Stand Rolling Mill.- 4.3.2 Identification and Control of a Glass Tube Process.- 4.4 Properties of the Least-Squares Estimator.- 4.5 Conclusions.- 5 Extensions of the Least-Squares Method.- 5.1 Modifying the Frequency Weighting by Prefiltering.- 5.2 A Natural Choice of Criterion — Output Error Method.- 5.3 Using Correlation Techniques — Instrumental Variable (IV) Methods.- 5.4 Obtaining White Residuals — Prediction Error Methods.- 5.4.1 Generalized Least-Squares (GLS) Method.- 5.4.2 General Properties of the Family of Prediction Error Methods.- 5.5 Identifying the Glass Tube Process Using a Prediction Error Method.- 5.6 Conclusions and Discussion.- 6 MIMO Process Identification: A Markov Parameter Approach.- 6.1 Rationale of the Method.- 6.2 The Identification Procedure.- 6.3 Identification of the Glass Tube Manufacturing Process.- 6.4 Conclusions.- 7 Identification for Robust Control; SISO Case.- 7.1 Asymptotic Properties of Prediction Error Models.- 7.2 The Identification Method.- 7.2.1 The Procedure.- 7.2.2 Model Order Selection.- 7.2.3 Optimal Experiment Design for Simulation.- 7.3 Recursive Estimation.- 7.3.1 The Recursive LS Method.- 7.3.2 A Recursive Output Error Method.- 7.3.3 A Recursive Identifier for Robust Adaptive Control.- 7.4 A Simulation Study.- 7.5 Conclusions.- 8 Identification for Robust Control; MIMO Case.- 8.1 The MIMO Version of the Asymptotic Theory.- 8.2 The Identification Method.- 8.2.1 The Procedure.- 8.2.2 Model Structure Selection.- 8.2.3 Input Design for Simulation.- 8.2.4 Determining a State Space Realization.- 8.3 Identification of Two Industrial Processes.- 8.3.1 Identification of the Glass Tube Process.- 8.3.2 Identification of a Four-Effect Evaporator.- 8.4 Closed Loop Identification of Coprime Factors.- 8.5 Conclusions.- 9 Identification and Robust Control of the Glass Tube Process.- 9.1 From Identification to Robust Control; Guidelines.- 9.2 Identification and Control of the Glass Tube Process; Control Results.- 9.3 Conclusions.- 10 Identification for Fault Diagnosis; Estimation of Continuous-Time Models.- 10.1 An Indirect Method of Continuous-Time Model Estimation.- 10.2 Enhancing a Parameters Subset by Input Design.- 10.3 A Simulation Study.- 10.4 Conclusions.- Symbols and Abbreviations.- References.