Advances in Industrial Control

Other titles published in this series: Digital Controller Implementation and Fragility Robert S.H. Istepanian and James F. Whidborne (Eds.)

Optimisation of Industrial Processes at Supervisory Level Doris Sáez, Aldo Cipriano and Andrzej W. Ordys

Robust Control of Diesel Ship Propulsion Nikolaos Xiros

Hydraulic Servo-systems Mohieddine Jelali and Andreas Kroll

Model-based Fault Diagnosis in Dynamic Systems Using Identification Techniques Silvio Simani, Cesare Fantuzzi and Ron J. Patton

Strategies for Feedback Linearisation Freddy Garces, Victor M. Becerra, Chandrasekhar Kambhampati and Kevin Warwick

Robust Autonomous Guidance Alberto Isidori, Lorenzo Marconi and Andrea Serrani

Dynamic Modelling of Gas Turbines Gennady G. Kulikov and Haydn A. Thompson (Eds.)

Control of Fuel Cell Power Systems Jay T. Pukrushpan, Anna G. Stefanopoulou and Huei Peng

Fuzzy Logic, Identification and Predictive Control Jairo Espinosa, Joos Vandewalle and Vincent Wertz

Optimal Real-time Control of Sewer Networks Magdalene Marinaki and Markos Papageorgiou

Process Modelling for Control Benoît Codrons

Computational Intelligence in Time Series Forecasting Ajoy K. Palit and Dobrivoje Popovic

Modelling and Control of Mini-Flying Machines Pedro Castillo, Rogelio Lozano and Alejandro Dzul

Ship Motion Control Tristan Perez

Hard Disk Drive Servo Systems (2nd Ed.) Ben M. Chen, Tong H. Lee, Kemao Peng and Venkatakrishnan Venkataramanan

Measurement, Control, and Communication Using IEEE 1588 John C. Eidson

Piezoelectric Transducers for Vibration Control and Damping S.O. Reza Moheimani and Andrew J. Fleming

Manufacturing Systems Control Design Stjepan Bogdan, Frank L. Lewis, Zdenko Kovačić and José Mireles Jr.

Windup in Control Peter Hippe

Nonlinear H2/H∞ Constrained Feedback Control Murad Abu-Khalaf, Jie Huang and Frank L. Lewis

Practical Grey-box Process Identification Torsten Bohlin

Control of Traffic Systems in Buildings Sandor Markon, Hajime Kita, Hiroshi Kise and Thomas Bartz-Beielstein

Wind Turbine Control Systems Fernando D. Bianchi, Hernán De Battista and Ricardo J. Mantz

Advanced Fuzzy Logic Technologies in Industrial Applications Ying Bai, Hanqi Zhuang and Dali Wang (Eds.)

Practical PID Control Antonio Visioli

(continued after Index)

Mohieddine Jelali • Biao Huang Editors with M.A.A. Shoukat Choudhury, Peter He, Alexander Horch, Manabu Kano, Nazmul Karim, Srinivas Karra, Hidekazu Kugemoto, Kwan-Ho Lee, S. Joe Qin, Claudio Scali, Zhengyun Ren, Maurizio Rossi, Timothy Salsbury, Sirish L. Shah, Ashish Singhal, Nina F. Thornhill, Jin Wang and Yoshiyuki Yamashita

Detection and Diagnosis of Stiction in Control Loops

State of the Art and Advanced Methods

123

Mohieddine Jelali, Dr.-Ing Department of Plant and System Technology VDEh-Betriebsforschungsinstitut GmbH (BFI) Sohnstraße 65 40237 Düsseldorf Germany mohieddine.jelali@bfi.de

Biao Huang, PhD Department of Chemical and Materials Engineering University of Alberta 536 Chemical and Materials Engineering Building Edmonton, Alberta T6G 2G6 Canada bhuang@ualberta.ca

ISSN 1430-9491 ISBN 978-1-84882-774-5 e-ISBN 978-1-84882-775-2 DOI 10.1007/978-1-84882-775-2

British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Control Number: 2009936772 © Springer-Verlag London Limited 2010 DuPont™ and Performance Surveyor™ are trademarks of E.I. du Pont de Nemours and Company.http://www2.dupont.com Excel® is a registered trademark of Microsoft Corporation in the United States and other countries.http://www.microsoft.com Genetic Algorithm and Direct Search Toolbox™, MATLAB®, Optimization Toolbox™, Signal Processing Toolbox™, Simulink® and System Identification Toolbox™ are all trademarks or registered trademarks of The MathWorks, Inc., 3 Apple Hill Drive, Natick, MA 01760-2098, U.S.A. http://www.mathworks.com Intel® and Pentium® are registered trademarks of Intel Corporation in the United States and other coun-tries. http://www.intel.com TOMLAB® is a registered trademark of Tomlab Optimization Inc. 1260 SE Bishop Blvd Ste E, Pullman, WA 99163-5451, USA. http://www.tomopt.com Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may only be re-produced, stored or transmitted, in any form or by any means, with the prior permission in writing ofthe publishers, or in the case of reprographic reproduction in accordance with the terms of licencesissued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those termsshould be sent to the publishers. The use of registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant laws and regulations and thereforefree for general use. The publisher makes no representation, express or implied, with regard to the accuracy of the infor-mation contained in this book and cannot accept any legal responsibility or liability for any errors oromissions that may be made. Cover design: eStudioCalamar, Figueres/Berlin Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Springer London Dordrecht Heidelberg New York

Advances in Industrial Control

Series Editors Professor Michael J. Grimble, Professor of Industrial Systems and Director Professor Michael A. Johnson, Professor (Emeritus) of Control Systems and Deputy Director

Industrial Control Centre Department of Electronic and Electrical Engineering University of Strathclyde Graham Hills Building 50 George Street Glasgow G1 1QE United Kingdom

Series Advisory Board Professor E.F. Camacho Escuela Superior de Ingenieros Universidad de Sevilla Camino de los Descubrimientos s/n 41092 Sevilla Spain

Professor S. Engell Lehrstuhl für Anlagensteuerungstechnik Fachbereich Chemietechnik Universität Dortmund 44221 Dortmund Germany

Professor G. Goodwin Department of Electrical and Computer Engineering The University of Newcastle Callaghan NSW 2308 Australia

Professor T.J. Harris Department of Chemical Engineering Queen’s University Kingston, Ontario K7L 3N6 Canada

Professor T.H. Lee Department of Electrical and Computer Engineering National University of Singapore 4 Engineering Drive 3 Singapore 117576

Professor (Emeritus) O.P. Malik Department of Electrical and Computer Engineering University of Calgary 2500, University Drive, NW Calgary, Alberta T2N 1N4 Canada

Professor K.-F. Man Electronic Engineering Department City University of Hong Kong Tat Chee Avenue Kowloon Hong Kong

Professor G. Olsson Department of Industrial Electrical Engineering and Automation Lund Institute of Technology Box 118 S-221 00 Lund Sweden

Professor A. Ray Department of Mechanical Engineering Pennsylvania State University 0329 Reber Building University Park PA 16802 USA

Professor D.E. Seborg Chemical Engineering 3335 Engineering II University of California Santa Barbara Santa Barbara CA 93106 USA

Doctor K.K. Tan Department of Electrical and Computer Engineering National University of Singapore 4 Engineering Drive 3 Singapore 117576

Professor I. Yamamoto Department of Mechanical Systems and Environmental Engineering The University of Kitakyushu Faculty of Environmental Engineering 1-1, Hibikino,Wakamatsu-ku, Kitakyushu, Fukuoka, 808-0135 Japan

Fur Doris, Yasmin und DunjaM.J.

To Yali and LindaB.H.

ix

Series Editors’ Foreword

The series Advances in Industrial Control aims to report and encourage technology transfer in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. New theory, new controllers, actuators, sensors, new industrial processes, computer methods, new applications, new philosophies…, new challenges. Much of this development work resides in industrial reports, feasibility study papers and the reports of advanced collaborative projects. The series offers an opportunity for researchers to present an extended exposition of such new work in all aspects of industrial control for wider and rapid dissemination.

There are two approaches to process performance assessment analysis. One is to base the analysis on a top-down viewpoint that seeks to assess performance through a decomposed structure for the process system. Fig. 0.1 shows such a decomposition comprised of the process, its instrumentation and the controller. Each of these areas can be subdivided further and the lower levels of the decomposition then rigorously assessed to ensure each component is delivering optimum performance – the idea being that achieving optimised performance at component level will translate into good process operation at global level.

Instrumentation

Process Performance Assessment

Process

Design

Controller

Actuators Design Sensors Tuning Characteristics

Fig. 0.1 Process decomposition

A quite different, complementary, bottom-up, approach to process performance assessment which depends on the installation of adequate data acquisition technology, is to pursue the analysis of performance using data-driven methods and algorithms. The data acquisition devices provide access to, and

x Series Editors’ Foreword

measurements of, process parameters, control signals, setpoints, process trajectory information, and process output signals. Features that might be sought in the data include offsets, biases, drifts, oscillations, noise effects, and shape distinctive disturbances. This information then has to be linked to some form of root-cause analysis to determine the possible location in a decomposed process structure (Fig. 0.1) and the cause of the performance deterioration. The techniques of statistical process control and the routines of controller performance assessment (from a field initiated by Professor Thomas Harris) are examples of two different groups of data-driven techniques for investigating process performance issues.

Within the decomposition of Fig. 0.1, one large class of actuators is that of pneumatic and hydraulic valves, and one of the possible performance-inhibiting consequences of their use is to introduce stiction (static-friction) into a control loop. However, stiction in a control loop can lead to oscillations or limit cycles. Thus the measurement data from an oscillating control loop can be used as a possible diagnostic signal to (a) detect the presence of valve stiction, and (b) provide an estimate of the stiction magnitude. This Advances in Industrial Control volume aims to establish a comprehensive and fundamental scientific framework for the detection and diagnosis of stiction in a range of industrial control loops. Early chapters of the book concentrate on understanding and modelling the physical origins of stiction. The central part of the book concentrates on the different approaches to constructing automated analysis procedures that provide robust stiction detection and possible stiction magnitude estimates. This is followed by an extended chapter reporting an exhaustive comparative assessment of the different methods and containing guidelines for the performance and use of the routines presented. A chapter on new possible research issues arising from the results reported brings the book to a satisfying conclusion.

The Editors of the Advances in Industrial Control series have always been aware of the importance of process performance assessment to industrial process and control engineers, and the series was the first to publish a book-length presentation of the rapidly growing subject of controller performance assessment, namely the monograph Performance Assessment of Control Loops (ISBN 978-1-85233-639-4, 1999) by Biao Huang and Sirish L. Shah. By way of an update, a second contribution to this particular field was published in 2007, Process Control Performance Assessment (ISBN 978-1-84628-623-0, 2007) edited by Andrzej W. Ordys, Damien Uduehi and Michael A. Johnson. Process nonlinearity analysis, stiction detection and diagnosis are new strands in the development of process performance assessment routines. Recently, M.A.A. Shoukat Choudhury, Sirish L. Shah and Nina F. Thornhill contributed the monograph Diagnosis of Process Nonlinearities and Valve Stiction (ISBN 978-3-540-79223-9, 2008) to the series. Now the Editors are very pleased to add this new and thorough scientific study to the growing book literature of process performance assessment techniques and applications in the series. These are books that can really be said to be essential volumes in every industrial process and control engineer’s library.

Industrial Control Centre M.J. Grimble Glasgow M.A. Johnson Scotland, UK 2009

Preface

Aim of the Book

Control performance monitoring (CPM) technology has progressed steadily sincethe key research step taken by Harris [41]. CPM has emerged as an important area ofscientific and technological development in the last decade. Considerable effort hasbeen devoted to the development and application of different CPM methodologiesin various industrial fields, such as refining, chemicals, and petrochemicals, mineralprocessing, mining and metal processing, as well as pulp and paper.

The diversity of causes of poor performance in industrial control loops has pro-vided the main motivation for exploring different CPM techniques. The cause ofpoor performance is not limited to controller design and tuning; other elements inthe control systems, such as sensors and actuators, are often responsible for the poorperformance. In a recent study by Paulonis and Cox [92], it was reported that theperformance of 42% control loops was in the categories of “fair” and/or “poor”.In other studies by Bialkowski [10] and Ender [26], it was found that in 30% ofcontrol loops, output cycling and increased output variability was due to instrumentperformance issues (such as valve backlash, dead time, etc.). Thus, a solution to anycontrol-loop performance problem will all too often involve diagnosis and analysisof oscillation and valve-stiction problems. Besides the detection and diagnosis ofsticky control valves, it is also important to be able to quantify stiction so that a listof sticky valves in order of their maintenance priority can be prepared.

This book is a collection of contributions from several internationally leading re-searchers on automatic detection and diagnosis of static friction (stiction) in controlloops manipulated by valves, and contains by now the most comprehensive surveyof state-of-the-art and advanced techniques. Not only does it present the principles,assumptions, strengths and drawbacks, but also provides guidelines and detailedworking procedures for the implementation and application of each method. An ex-haustive comparison of the described approaches on nearly 100 control loops fromdifferent industrial fields (building, chemicals, pulp and paper, mineral processing,mining, and metal processing) is included.

xi

xii Preface

Stiction is a sophisticated non-linear phenomenon. Its detection and quantifi-cation has been identified as a highly challenging academic as well as industrialproblem. There does not exist a single “perfect” solution; each method has its as-sumptions, strengths, and weaknesses. Thus, it is imperative for readers to haveaccess to different options, and compare and choose the most appropriate one fora given process-operating condition. In this respect, the book provides a good re-source for researchers and industrial practitioners to find a variety of basic as wellas advanced methods for solving valve-stiction problems. The methods presentedinclude techniques for diagnosis plus quantification of stiction with both open-loopand closed-loop methods. The reader will learn how the different methods work,what key issues should be considered, and how to parametrise them. The compre-hensive comparative study gives readers useful guidelines in choosing the appro-priate method. Moreover, users will have access to MATLAB® software associatedwith the book so that they can directly benefit from the book by applying the meth-ods to their data. The software are available to the public and can be downloadedfrom http://www.ualberta.ca/˜bhuang/Stiction-Book/.

Readership

The algorithms presented in this book are the state-of-the-art, and are demonstratedand compared on industrial case studies from different industrial fields (building,chemicals and petrochemicals, mining, mineral and metal processing). The bookwill thus serve academic and industrial staff working in all these industries on con-trol systems design, maintenance or optimisation.

The book presents – in one place – some material published in the archival litera-ture over the last several years as well as advanced material developed most recently.It is aimed primarily at researchers and engineers in process control and industrialpractitioners in process automation, but it is also accessible to postgraduate studentsand final-year students in process-control engineering.

Outline of the Book

The book, which has 14 chapters, is divided into two parts. Chapter 1 is an intro-duction to the general topics and problems treated in the book. Part I is devotedto reviewing the state-of-the-art in stiction modelling and oscillation detection, andis divided into three chapters. Chapter 2 reviews stiction models with emphasis onsimple data-driven relationships between controller output (OP) and valve position(VP), i.e. one- and two-parameter models, and gives examples that investigate thevalidity and compare the behaviour of three representative stiction models avail-able in the literature. New alternative data-driven models of stiction are proposedin Chap. 3 and compared with earlier modelling approaches. Oscillation detection

Preface xiii

is one of the first tasks to be carried out when analysing control-loop performance.The different techniques for oscillation detection are described and discussed in de-tail in Chap. 4. Their applicability to industrial control-loop diagnosis is criticallyevaluated on various simulation and industrial case studies.

Part II of the book presents basic and advanced methodologies for stiction di-agnosis and quantification, and contains ten chapters. In Chap. 5, pattern-basedstiction-detection and quantification methods that assume the availability of the ma-nipulated variable (MV), i.e. VP, are treated. In all subsequent chapters, the stictiondetection is based only on the knowledge of OP and PV, which reflects industrialpractice, where MV is usually not known, except for flow control loops, where PVand MV are considered to be coincident. Chapter 6 describes the simplest stiction-detection method based on cross-correlation between OP and PV. Since this methodis only applicable to self-regulating processes, another method based on shape anal-ysis (histogram) is also presented in this chapter for detecting stiction in integrat-ing plants. Chapter 7 discusses a method for stiction detection based on curve fit-ting of the output signal of the first integrating component located after the valvein a control loop, i.e. OP for self-regulating processes or PV for integrating pro-cesses. Considering the similarity between the oscillation shapes of PV in loopsaffected by valve stiction and those obtained by a relay operating in closed loopon a first-order-plus-time-delay process, a technique based on shape determinationis described in Chap. 8. Chapter 9 presents a method for detecting stiction-like be-haviour in feedback control loops based on extracting features from a time seriesrecord of PV, delivering the probability that stiction is present (ranged 0–100 %).Chapter 10 discusses a new procedure for quantifying valve stiction in control loopsusing separable least-squares and global optimisation algorithms, assuming that theprocess dynamics can be described by a Hammerstein model. In a similar context,a novel stiction-detection strategy based on closed-loop identification is proposedusing closed-loop operating data along with a general analysis on identifiability ofthe stiction loops in Chap. 11. Chapter 12 presents a novel methodology that bridgesgaps in oscillation diagnostic methods. A power spectral density (PSD) based os-cillation detection method followed by a model-based approach for identifying andquantifying the root cause of the oscillations is proposed. In all these chapters, thedescribed methodologies are illustrated with simulated data and industrial data. InChap. 13, an exhaustive comparative study of all methods presented is given, involv-ing a large number of data sets gathered from different industrial plants, assessingthe relative efficiency of the techniques, and delivering guidelines for selecting theright method for the application at hand. Chapter 14 provides a summary and a lookforward to future research challenges within the topic treated in the book.

In Appendix A, some information about the industrial control loops consideredthroughout the book is given. Appendix B contains a short overview of the twoprominent methods for non-linearity detection in the CPM area, i.e. the bicoherencetechnique and the surrogate analysis method.

xiv Preface

Acknowledgements

This book is an outcome of a project initiated by the principal editor to establishan international database of industrial loops from different fields, to implement acommon graphical user interface (GUI) for oscillation and stiction detection, and tocompare oscillation and stiction-detection methods.

The editors would like to thank all the contributors for their very spontaneousenthusiasm and agreement to contribute to the book and particularly to the com-parative study in Chap. 13. For this purpose, industrial data (Appendix A) wereprovided by courtesy of: Ashish Singhal and Timothy Salsbury for loops BAS 1–8; Alexander Horch for loops CHEM 1–3 and PAP 1–10; Peter He and Joe Qinfor loops CHEM 4–6; Biao Huang for loops CHEM 7–12; Nina Thornhill for loopsCHEM 13–17 and CHEM 40–64; Claudio Scali for loops CHEM 18–28 and CHEM32–39; Shoukat Choudhury and Sirish Shah for loops CHEM 29–31, PAP 11–13,POW 1–5 and MIN 1.

As far as possible, we have tried to achieve a common framework for all chapterspresented. The contributors were all asked to produce precise algorithms that can beeasily implemented by the reader, to demonstrate their methods in simulation andindustrial studies, and to use similar notation. Also, overlap in text and figures wasminimised, and cross-references between the chapters were created. All this willhopefully help readers to easily understand and compare the techniques proposed.We are grateful to all contributors for their forbearance in meeting the many requestsfor clarity and consistency.

Special thanks are due to Alexander Horch, Claudio Scali, Nina Thornhill andShoukat Choudhury for numerous discussions and valuable suggestions, which haveimproved the presentation of many chapters of the book.

We also acknowledge Alexander Horch, Claudio Scali, Shoukat Choudhury,Timothy Salsbury and Peter He for providing software (either as m- or p-code) to beincluded in the oscillation- and stiction-detection GUI implemented by the principaleditor.

Last but not least, we would like to acknowledge Oliver Jackson, Aislinn Bun-ning (Springer), Sorina Moosdorf and Katja Roser (le-tex publishing services) fortheir editorial comments and detailed examination of the book. We also thank Pro-fessor Michael A. Johnson for providing many valuable comments and suggestions.

Dusseldorf, Germany Mohieddine JelaliEdmonton, Alberta, Canada Biao HuangMay 2009

Copyright Acknowledgements

Parts of the book appeared in the archived literature and the authors gratefully ac-knowledge permissions to re-use material from the following papers.

Chapter 2 contains extracts and figures from Choudhury M.A.A.S., Shah S.L.,Thornhill N.F., Shook D., Automatic detection and quantification of stiction in con-trol valves, Control Engineering Practice 14:1395–1412, © 2006, Elsevier Ltd.,with permission from Elsevier.

Parts of Chap. 6 are reprinted (with revision) from Horch A., A simple method fordetection of stiction in control loops, Control Engineering Practice 7:1221–1231,© 1999, Elsevier Ltd., with permission from Elsevier.

Chapter 7 is reprinted in large part with permission from He Q.P., Wang J., PottmannM., Qin S.J., A curve fitting method for detecting valve stiction in oscillating controlloops, Ind. Eng. Chem. Res. 2007, 46(13), 4549–4560. Copyright 2007 AmericanChemical Society.

Chapter 8 is reprinted (with revision) from Rossi M., Scali C., A comparison of tech-niques for automatic detection of stiction: simulation and application to industrialdata, Journal of Process Control 15:505–514, © 2005, Elsevier Ltd., with permis-sion from Elsevier.

Chapter 9 is an expanded and revised version of Singhal A., Salsbury T.I., A simplemethod for detecting valve stiction in oscillating control loops, Journal of ProcessControl 15:371–382, © 2005, Elsevier Ltd., with permission from Elsevier.

Chapter 10 is an expanded and revised version of Jelali M., Estimation of valvestiction in control loops using separable least-squares and global search algorithms,Journal of Process Control 18:632–642, © 2008, Elsevier Ltd., with permissionfrom Elsevier.

xv

Contents

List of Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxv

Abbreviations and Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxix

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1Mohieddine Jelali and Biao Huang1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Typical Valve-controlled Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21.3 Stiction Phenomenon and Related Effects . . . . . . . . . . . . . . . . . . . . . 41.4 Input–Output Relation of Valves Under Stiction . . . . . . . . . . . . . . . . 61.5 Limit Cycles due to Stiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.6 Typical Observations in Control Loops with Sticky Valves . . . . . . . 121.7 Industrial Examples of Loops with Stiction . . . . . . . . . . . . . . . . . . . . 151.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

Part I Stiction Modelling and Oscillation Detection

2 Stiction Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21M.A.A. Shoukat Choudhury, Nina F. Thornhill, Manabu Kano andSirish L. Shah2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.2 Physics-based Stiction Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.3 Data-driven Stiction Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.3.1 One-parameter Stiction Model . . . . . . . . . . . . . . . . . . . . . . . 242.3.2 Two-parameter Stiction Model . . . . . . . . . . . . . . . . . . . . . . 242.3.3 Choudhury’s Stiction Model . . . . . . . . . . . . . . . . . . . . . . . . 252.3.4 Simulation of the Stiction Model . . . . . . . . . . . . . . . . . . . . . 282.3.5 Kano’s Stiction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

2.4 Comparison Between Choudhury’s and Kano’s Stiction Models . . 322.4.1 Similarities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322.4.2 Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

xvii

xviii Contents

2.4.3 Comparisons Using an Industrial Example . . . . . . . . . . . . . 322.5 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3 An Alternative Stiction-modelling Approach and Comparison ofDifferent Stiction Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37Q. Peter He, Jin Wang and S. Joe Qin3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.2 He’s Two-parameter Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.3 Three Data-driven Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3.3.1 Implementation of the First-principles Model . . . . . . . . . . 413.3.2 Comparison of Data-driven Models . . . . . . . . . . . . . . . . . . 43

3.4 Further Investigation of Valve Stiction . . . . . . . . . . . . . . . . . . . . . . . . 453.5 He’s Three-parameter Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493.6 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513.7 An Industrial Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 573.9 Appendix: Proof of the Equivalence Between He’s

Two-parameter and Three-parameter Model . . . . . . . . . . . . . . . . . . . 58

4 Detection of Oscillating Control Loops . . . . . . . . . . . . . . . . . . . . . . . . . . 61Srinivas Karra, Mohieddine Jelali, M. Nazmul Karim and AlexanderHorch4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.2 Root-causes for Oscillatory Control Loops . . . . . . . . . . . . . . . . . . . . 62

4.2.1 Poor Process and Control System Design . . . . . . . . . . . . . . 624.2.2 Aggressive Controller Tuning . . . . . . . . . . . . . . . . . . . . . . . 634.2.3 Non-linearities in Control-loop Hardware . . . . . . . . . . . . . 634.2.4 External Oscillatory Disturbances . . . . . . . . . . . . . . . . . . . . 64

4.3 Characterisation of Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 644.3.1 Auto-covariance Function . . . . . . . . . . . . . . . . . . . . . . . . . . 644.3.2 Power Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.3.3 Strength of Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

4.4 Techniques for Detection of Oscillations in Control Loops . . . . . . . 674.4.1 Detection of Spectral Peaks . . . . . . . . . . . . . . . . . . . . . . . . . 674.4.2 Regularity of Large Enough Integral of Absolute Error . . 694.4.3 Regularity of Upper and Lower IAEs and Zero-crossings 724.4.4 Decay-ratio Approach of the Auto-correlation Function . . 744.4.5 Regularity of Zero-crossings of the Auto-correlation

Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.4.6 Spectral Envelope Method . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.5 Critical Evaluation of Oscillation-detection Methods . . . . . . . . . . . . 804.5.1 Features of Industrial Control-loop-oscillation Detection . 804.5.2 Detection Test Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 814.5.3 Signal with Coloured Noise . . . . . . . . . . . . . . . . . . . . . . . . . 834.5.4 Signal with One Predominant Oscillation . . . . . . . . . . . . . . 844.5.5 Signal with Dampened Oscillation . . . . . . . . . . . . . . . . . . . 85

Contents xix

4.5.6 Signal with Multiple Oscillations . . . . . . . . . . . . . . . . . . . . 874.5.7 Signal with Intermittent Oscillations . . . . . . . . . . . . . . . . . . 90

4.6 Comprehensive Oscillation Characterisation . . . . . . . . . . . . . . . . . . . 934.7 Industrial Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.7.1 Oscillating Flow Control Loop . . . . . . . . . . . . . . . . . . . . . . 954.7.2 Unit-wide Oscillation Caused by a Sensor Fault . . . . . . . . 964.7.3 Plant-wide Oscillation Caused by a Valve Fault . . . . . . . . 97

4.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

Part II Advances in Stiction Detection and Quantification

5 Shape-based Stiction Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Manabu Kano, Yoshiyuki Yamashita and Hidekazu Kugemoto5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1035.2 Method Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

5.2.1 Method A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045.2.2 Method B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1055.2.3 Method C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106

5.3 Key Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1075.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1085.5 Application to Industrial Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1115.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

6 Stiction Detection Based on Cross-correlation and Signal Shape . . . . 115Alexander Horch6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1156.2 The Cross-correlation Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1176.3 Industrial Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

6.3.1 Loop Interaction I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1206.3.2 Loop Interaction II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1216.3.3 Flow Control Loop I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1236.3.4 Flow Control Loop II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1256.3.5 Level Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

6.4 Theoretical Explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1266.4.1 Correlation for Oscillating External Disturbances . . . . . . . 1276.4.2 Tight Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1296.4.3 Correlation in the Presence of Stiction . . . . . . . . . . . . . . . . 129

6.5 Conclusions (Cross-correlation Method) . . . . . . . . . . . . . . . . . . . . . . 1316.6 Stiction Detection for Integrating Processes . . . . . . . . . . . . . . . . . . . 1326.7 Detection in Integrating Loops – Basic Idea . . . . . . . . . . . . . . . . . . . 132

6.7.1 Differentiation and Filtering . . . . . . . . . . . . . . . . . . . . . . . . . 1336.7.2 Sample Histogram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1356.7.3 Distribution for the Stiction Case . . . . . . . . . . . . . . . . . . . . 1366.7.4 Distribution for the Non-stiction Case . . . . . . . . . . . . . . . . 138

6.8 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1426.8.1 Level Control Loop with Stiction . . . . . . . . . . . . . . . . . . . . 142

xx Contents

6.8.2 Level Control Loop Without Stiction . . . . . . . . . . . . . . . . . 1436.8.3 Level Control Loop with Deadband . . . . . . . . . . . . . . . . . . 143

6.9 Self-regulating Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1436.9.1 Flow Control Loop with Stiction . . . . . . . . . . . . . . . . . . . . . 1446.9.2 Flow Control Loop Without Stiction . . . . . . . . . . . . . . . . . . 1456.9.3 Loops with Dominant P-control . . . . . . . . . . . . . . . . . . . . . 145

6.10 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

7 Curve Fitting for Detecting Valve Stiction . . . . . . . . . . . . . . . . . . . . . . . 149Q. Peter He and S. Joe Qin7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1497.2 Method Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151

7.2.1 Sinusoidal Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1527.2.2 Triangular Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1537.2.3 Stiction Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153

7.3 Key Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1547.5 Application to Industrial Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1597.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

8 A Relay-based Technique for Detection of Stiction . . . . . . . . . . . . . . . . 165Claudio Scali and Maurizio Rossi8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1668.2 Trends of Different Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1688.3 Method Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

8.3.1 Basic Idea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1708.3.2 Stiction Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1718.3.3 Fitting Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1728.3.4 Fitting Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172

8.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1758.4.1 Nominal Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1758.4.2 Presence of Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177

8.5 Application to Plant Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1788.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181

9 Shape-based Stiction Detection Using Area Calculations . . . . . . . . . . . 183Timothy I. Salsbury and Ashish Singhal9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1839.2 Method Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

9.2.1 Theoretical Basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1879.2.2 Stiction Detection Hypothesis Test . . . . . . . . . . . . . . . . . . . 1909.2.3 Noise Effects and Practical Implementation . . . . . . . . . . . . 191

9.3 Key Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1979.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1989.5 Application to Industrial Loops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200

Contents xxi

9.5.1 Temperature Control Loop with Stictionfrom a Building Automation System . . . . . . . . . . . . . . . . . . 201

9.5.2 Temperature Control Loop with Stiction from a Pulpand Paper Plant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202

9.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

10 Estimation of Valve Stiction Using Separable Least-squares andGlobal Search Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205Mohieddine Jelali10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20510.2 Basic Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

10.2.1 Identification Model Structure: Hammerstein Model . . . . 20810.2.2 Linear Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20810.2.3 Stiction Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

10.3 Identification Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21010.3.1 Separable Least-squares Estimator . . . . . . . . . . . . . . . . . . . 21010.3.2 Global Search Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . 213

10.4 Key Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21510.4.1 Model Structure Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 21510.4.2 Time-delay Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21610.4.3 Determination of Initial Parameters and Incorporation

of Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21810.5 Simulation Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

10.5.1 First-order-plus-time-delay Process . . . . . . . . . . . . . . . . . . 22010.5.2 Integrating Process with Time Delay . . . . . . . . . . . . . . . . . 220

10.6 Industrial Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22110.6.1 Loop CHEM 25: Pressure Control Loop . . . . . . . . . . . . . . 22110.6.2 Loop PAP 2: Flow Control Loop . . . . . . . . . . . . . . . . . . . . . 22410.6.3 Loop CHEM 24: Flow Control Loop with Setpoint

Changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22410.6.4 Loop POW 2: Level Control Loop . . . . . . . . . . . . . . . . . . . 22510.6.5 Loop POW 4: Level Control Loop . . . . . . . . . . . . . . . . . . . 22610.6.6 Loop MIN 1: Temperature Control Loop . . . . . . . . . . . . . . 22610.6.7 Loop CHEM 70: Flow Control Loop with External

Disturbances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22610.7 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

11 Stiction Estimation Using Constrained Optimisation and ContourMap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229Kwan Ho Lee, Zhengyun Ren and Biao Huang11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22911.2 Stiction Model of Control Valve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231

11.2.1 General Conception . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23111.2.2 Physical Model of Valve Sticion . . . . . . . . . . . . . . . . . . . . . 23111.2.3 Kano’s Valve-stiction Model . . . . . . . . . . . . . . . . . . . . . . . . 23211.2.4 Choudhury’s Valve-stiction Model . . . . . . . . . . . . . . . . . . . 232

xxii Contents

11.2.5 He’s Valve-stiction Model . . . . . . . . . . . . . . . . . . . . . . . . . . 23211.3 Existing Stiction-detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . 233

11.3.1 Open-loop Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23311.3.2 Closed-loop Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23311.3.3 Discussion of Existing Methods . . . . . . . . . . . . . . . . . . . . . 234

11.4 Closed-loop Stiction Detection and Quantification . . . . . . . . . . . . . . 23511.4.1 Basic Principle and Important Steps . . . . . . . . . . . . . . . . . . 23511.4.2 Stiction Detection and Quantification Procedure . . . . . . . . 23611.4.3 Search Space of Stiction-model Parameters . . . . . . . . . . . . 23711.4.4 Constrained Parameter-search Techniques . . . . . . . . . . . . . 23811.4.5 Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 240

11.5 Stiction Detection: Identifiability Analysis . . . . . . . . . . . . . . . . . . . . 24111.5.1 Heuristic Illustration of Closed-loop Identifiability . . . . . . 24111.5.2 Identifiability Analysis for Closed-loop Systems

with Valve Stiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24311.6 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24511.7 Industrial Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

11.7.1 Illustrative Industrial Examples . . . . . . . . . . . . . . . . . . . . . . 25111.7.2 Comparative Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256

11.8 Graphical User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26011.9 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265

12 Oscillation Root-cause Detection and Quantification UnderMultiple Faults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267Srinivas Karra and M. Nazmul Karim12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26712.2 Preliminaries and Brief Review of Model-based Oscillation

Diagnosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26812.2.1 Root-cause for Oscillations and Compensation

Techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26812.2.2 Oscillation Diagnosis and Root-cause Quantification . . . . 26912.2.3 Challenges to be Addressed . . . . . . . . . . . . . . . . . . . . . . . . . 270

12.3 Overview of the Root-cause Detection and QuantificationMethodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27012.3.1 Revisiting Control-valve Characteristics Under Stiction . . 27012.3.2 Oscillation Detection and Diagnosis Methodology . . . . . . 271

12.4 Process-model Identification Under Non-stationary Disturbances . 27112.4.1 Identification of EARMAX Model . . . . . . . . . . . . . . . . . . . 27312.4.2 Illustrative Example: Identification Under

Non-stationary Disturbance . . . . . . . . . . . . . . . . . . . . . . . . . 27512.5 Root-cause Detection and Quantification . . . . . . . . . . . . . . . . . . . . . . 278

12.5.1 OP–PV Model Identification Methodology . . . . . . . . . . . . 27812.5.2 Identification of Controller Transfer Function . . . . . . . . . . 28012.5.3 Oscillation Root-cause Detection and Quantification

Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280

Contents xxiii

12.6 Illustrative Example: Oscillation Diagnosis Under VariousFaulty Situations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28112.6.1 Stiction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28112.6.2 Oscillatory External Disturbance . . . . . . . . . . . . . . . . . . . . . 28312.6.3 Aggressive Controller Tuning . . . . . . . . . . . . . . . . . . . . . . . 28412.6.4 Stiction and Oscillatory External Disturbance . . . . . . . . . . 28512.6.5 Stiction and Aggressive Controller Tuning . . . . . . . . . . . . . 28612.6.6 Oscillatory External Disturbance and Aggressive

Controller Tuning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28712.6.7 Stiction, Aggressive Controller Tuning and Oscillatory

External Disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28812.7 Industrial Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290

12.7.1 Control Loop 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29012.7.2 Control Loop 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29112.7.3 Control Loop 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292

12.8 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293

13 Comparative Study of Valve-stiction-detection Methods . . . . . . . . . . . 295Mohieddine Jelali and Claudio Scali13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29513.2 Selected Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29613.3 Industrial Control Loops Involved in the Study . . . . . . . . . . . . . . . . . 29813.4 Application Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 302

13.4.1 Application Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30213.4.2 Synthesis and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . 30313.4.3 Efficiency of the Techniques, Problems and

Countermeasures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30613.4.4 Comparison on 20 Loops with Known Problems . . . . . . . 31713.4.5 Selected Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31713.4.6 Graphical User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319

13.5 Suggestions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32113.6 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32313.7 Appendix: Tables of Results of the Comparative Study . . . . . . . . . . 324

14 Conclusions and Future Research Challenges . . . . . . . . . . . . . . . . . . . . 359Biao Huang, Mohieddine Jelali and Alexander Horch14.1 Summary of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35914.2 Future Research Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362

14.2.1 Stiction Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36214.2.2 Oscillation Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36314.2.3 Stiction Detection and Estimation . . . . . . . . . . . . . . . . . . . . 36414.2.4 Stiction Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365

xxiv Contents

Appendix A Evaluated Industrial Control Loops . . . . . . . . . . . . . . . . . . . . . 367

Appendix B Review of Some Non-linearity and Stiction-detectionTechniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371B.1 Bicoherence Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371

B.1.1 Non-Gaussianity Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372B.1.2 Non-linearity Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373B.1.3 Total Non-linearity Index . . . . . . . . . . . . . . . . . . . . . . . . . . . 374B.1.4 Ellipse Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374

B.2 Surrogates Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374B.2.1 Surrogate Data Generation . . . . . . . . . . . . . . . . . . . . . . . . . . 375B.2.2 Non-linear Predictability Index . . . . . . . . . . . . . . . . . . . . . . 376

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377

Contributor Biographies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389

List of Contributors

M.A.A. Shoukat ChoudhuryDepartment of Chemical Engineering, Bangladesh University of Engineering andTechnology, Dhaka-1000, Bangladesh, e-mail: shoukat@che.buet.ac.bd

Q. Peter HeDepartment of Chemical Engineering, Tuskegee University, 522 B Luther H. FosterHall, AL 36088, USA, e-mail: qhe@tuskegee.edu

Alexander HorchGroup Process and Production Optimization, ABB Corporate ResearchGermany, Wallstadter Str. 59, 68526 Ladenburg, Germany, e-mail:alexander.horch@de.abb.com

Biao HuangDepartment of Chemical and Materials Engineering, University of Alberta,Edmonton, Alberta, T6G 2G6, Canada, e-mail: biao.huang@ualberta.ca

Mohieddine JelaliDepartment of Plant and System Technology, VDEh-BetriebsforschungsinstitutGmbH (BFI), Sohnstraße 65, 40237 Dusseldorf, Germany, e-mail:mohieddine.jelali@bfi.de

Manabu KanoDepartment of Chemical Engineering, Kyoto University, Nishikyo-ku, Kyoto615-8510, Japan, e-mail: manabu@cheme.kyoto-u.ac.jp

M. Nazmul KarimDepartment of Chemical Engineering, Texas Tech University, Lubbock, TX 79409,USA, e-mail: naz.karim@ttu.edu

xxv

xxvi List of Contributors

Srinivas KarraDepartment of Chemical Engineering, Texas Tech University, Lubbock, TX 79409,USA, e-mail: srini.karra@ttu.edu

Hidekazu KugemotoSumitomo Chemical Co., Ltd., 5-l, Sobiraki-cho, Niihama City, Ehime 792-8521,Japan, e-mail: kugemoto@sc.sumitomo-chem.co.jp

Kwan Ho LeeDepartment of Chemical and Materials Engineering, University of Alberta,Edmonton, Alberta, T6G 2G6, Canada, e-mail: kwanho.lee@ualberta.ca

S. Joe QinDepartment of Chemical Engineering and Materials Science, Electrical Engineer-ing, and Industrial and Systems Engineering, University of Southern California,925 Bloom Walk, HED 211, Los Angeles, CA 90089-1211, USA, e-mail:sqin@usc.edu

Zhengyun RenDepartment of Automation, Donghua University, Shanghai, China, e-mail:renzhengyun@163.com

Maurizio RossiAspenTech - Srl, Pisa, Italy, e-mail: maurizio.rossi@aspentech.com

Timothy I. SalsburyControls Research Department, Johnson Controls, Inc, 507 E Michigan Street,Milwaukee, WI 53202, USA, e-mail: timothy.i.salsbury@jci.com

Claudio ScaliChemical Process Control Laboratory (CPCLab), Department of ChemicalEngineering (DICCISM), University of Pisa, Via Diotisalvi, n.2 56126 Pisa, Italy,e-mail: scali@ing.unipi.it

Sirish L. ShahDepartment of Chemical and Materials Engineering, University of AlbertaEdmonton, Alberta, T6G 2G6, Canada, e-mail: sirish.shah@ualberta.ca

Ashish SinghalAdvanced Control and Operations Research R&D, Praxair, Inc., 175 East Park Dr.,Tonawanda, NY 14150, USA, e-mail: ashish_singhal@praxair.com

Nina F. ThornhillCentre for Process Systems Engineering, Department of Chemical Engineering,Imperial College London, South Kensington Campus, London SW7 2AZ, UK,e-mail: n.thornhill@imperial.ac.uk

Jin WangDepartment of Chemical Engineering, Auburn University, AL 36849, USA, e-mail:wang@auburn.edu

List of Contributors xxvii

Yoshiyuki YamashitaDepartment of Chemical Engineering, Tokyo University of Agriculture andTechnology, 2-24-16 Naka-cho, Koganei, Tokyo 184-8588, Japan, e-mail:yama_pse@cc.tuat.ac.jp

Abbreviations and Acronyms

AC Analyser controlACF Auto-correlation/covariance functionAMIGO Approximate Ms constrained integral gain optimisationANSI American National Standard InstitutionARMAX Auto-regressive moving average exogenousARX Auto-regressive exogenousBAS Building automation systemBJ Box–JenkinsC ControllerCC Cohen–CoonCCF Cross-correlation functionCHEM ChemicalsCHR Chien–Hrones–ReswickCOC Comprehensive oscillation characterisationCPM Control-performance monitoringCPU Central processing unitDCS Distributed control systemDR Decay ratioEARMAX Extended ARMAXEWMA Exponentially weighted moving averageFC Flow controlFOPTD First-order-plus-time-delayGA Genetic algorithmGC Gauge controlGUI Graphical user interfaceIAE Integral of absolute errorIPTD Integrating plus time delayLTI Linear time invariantMC Monte-CarloMET Metal

xxix

xxx Abbreviations and Acronyms

MSE Mean square errorsMV Manipulated variableNGI Non-Gaussianity indexNLA Non-linearity analysisNLI Non-linearity indexNPI Non-predictability indexODE Ordinary differential equationOE Output errorOP Controller outputOS OvershootP ProcessPAP Pulp and paperPC Pressure controlP ProportionalPI Proportional-integralPID Proportional-integral-derivativePOW PowerPSD Power spectral densityPV Process variableQF Quantisation factorSI Stiction indexSISO Single-input single-outputSOPTD Second-order-plus-time-delaySP SetpointTDE Time-delay estimationTF Transfer functionV ValveVP Valve positionVOC Volatile organic compoundZN Ziegler–NicholsZOH Zero-order hold

Chapter 1Introduction

Mohieddine Jelali and Biao Huang

This introductory chapter explains the role of control valves in the process industryand the motivation for the detection of valve stiction in control loops. It reviewssome definitions of the term “stiction” and similar phenomena (deadband, backlash,hysteresis, etc.) that can occur in control valves. Some basic analysis tools will bedescribed, including the input–output relation of valves under stiction and the de-scribing function analysis. Data from typical industrial control loops are presentedto illustrate the effects of stiction on the shape of loop signals, such as the processvariable and the controller output.

1.1 Motivation

Large-scale, highly integrated processing plants, such as oil refineries, ethyleneplants, power plants, and rolling mills, include some hundreds or even thousandsof control loops. The aim of each control loop is to maintain the process at the de-sired operating conditions, safely and efficiently. A poorly performing control loopcan result in disrupted process operation, degraded product quality, higher materialor energy consumption, and thus decreased plant profitability. Therefore, controlloops have been increasingly recognised as important capital assets that should beroutinely monitored and maintained. The performance of the controllers, as well asof the other loop components, can thus be improved continuously, ensuring productsof consistently high quality.

Mohieddine JelaliDepartment of Plant and System Technology, VDEh-Betriebsforschungsinstitut GmbH (BFI),Sohnstraße 65, 40237 Dusseldorf, Germany, e-mail: mohieddine.jelali@bfi.de

Biao HuangDepartment of Chemical and Materials Engineering, University of Alberta, Edmonton, Alberta,T6G 2G6, Canada, e-mail: biao.huang@ualberta.ca

1

2 M. Jelali and B. Huang

Surveys [10, 25, 26, 92] indicate that about 20–30% of all control loops oscillatedue to valve problems caused by valve non-linearities, such as stiction, hystere-sis, deadband or deadzone. Many control loops in process plants perform poorlydue to valve static friction (stiction), as one of the most common equipment prob-lems. It is well known that valve stiction in control loops causes oscillations inthe form of periodic finite-amplitude instabilities, known as limit cycles. This phe-nomenon increases variability in product quality, accelerates equipment wear, orleads to control-system instability.

1.2 Typical Valve-controlled Loop

Figure 1.1 shows a simple configuration of control loops actuated with a controlvalve. A typical example of such a configuration, i.e. a level control loop, is illus-trated in Fig. 1.2. In most applications in the process industry, pneumatic controlvalves are used. The diagram of a typical pneumatic valve is shown in Fig. 1.3. Thevalve aims to restrict the flow of process fluid through the pipe that can be seen atthe very bottom of the figure. The valve plug is rigidly attached to a stem that isattached to a diaphragm in an air-pressure chamber in the actuator section at the topof the valve. When compressed air is applied, the diaphragm moves up and the valveopens. At the same time, the spring is compressed. In process operation, a controlvalve is subject to the following forces: (i) the valve stem driving force caused bythe air pressure, (ii) the spring force associated with the valve travel, (iii) the sealfriction of the seals sealing the process fluid and the stem thrust originating in theprocess fluid passing through the valve body.

Fig. 1.1 Simple feedback scheme for a valve-controlled process with definition of the variablessetpoint (SP), controller output (OP), manipulating variable or valve position (MV) and processvariable (PV) used throughout the book

Stiction in control valves is thought to occur due to seal degradation, lubricantdepletion, inclusion of foreign matter, activation at metal sliding surfaces at hightemperatures and/or tight packing around the stem. The resistance offered from thestem packing is often considered as the main cause of stiction. Another very com-mon cause of stiction is indirectly due to regulations on volatile organic compound(VOC) emissions. In many plants, a team monitors each valve for VOC emissions,usually between the packing and the stem. If any minute leakage is detected, pack-ing in the valve body is tightened far more than is necessary. This causes the valve

1 Introduction 3

to stick making the process run less efficiently with increased energy consump-tion. Stiction often varies over time and operating regimes. Since wear is also non-uniform along the body, frictional forces are different at different stem positions.When the control loop is at steady state, and if a valve exhibits this behaviour, per-sistent oscillations in PV on either side of the setpoint are observed [115].

Control

valve

LC

u

h

r

p1

Q

dQ

Fig. 1.2 Level control loop

Compressed air

Diaphragm

Spring

pd

Valve

stroke

Plug

Valve seat

Exit flow

Valve

drive

Valve

armature

Stiction

Entry flow

Packing

Q

Fig. 1.3 Diagram of a pneumatic control valve [74]

Control valves should be maintained to have acceptable values for the parametersgiven in Table 1.1. In many processes, stiction of 0.5% is considered too much, as