Data Mining and Knowledge Discovery for Process Monitoring and Control

262 Pages·1999·7.983 MB·English

by  Xue Z. Wang BEng, MSc, PhD (auth.)

Advances in Industrial Control
Springer-Verlag London Ltd.
Xue Z. Wang, BEng, MSc, PhD
Department of Chemical Engineering
Un iversity of Leeds
Leeds LS2 9JT
UK
British Library Cataloguing in Publication Data
Wang,XueZ.
Data mining and knowledge discovery for process monitoring
and control. - (Advances in industrial control)
1.Process control-Data processing 2.Data mining
1.Title II.McGreavy, C. .
629.8'9
ISBN 978-1-4471-1137-5
Library of Congress Cataloging-in-Publication Data
Wang, X. Z. (Xue Zhang), 1963-
Data mining and knowledge discovery for process monitoring and
Control/ Xue Z. Wang.
p. cm. -- (Advances in industrial control)
Inc1udes bibliographical references and index.
ISBN 978-1-4471-1137-5 ISBN 978-1-4471-0421-6 (eBook)
DOI 10.1007/978-1-4471-0421-6
1. Process control--Data processing. 2. Data mining. 3. Knowledge
acquisition (Expert systems) 1. Title. II. Series.
TSI56.8.W37 1999
670.42'7563--dc21
Advances in Industrial Control series ISSN 1430-9491
ISBN 978-1-4471-1137-5
© Springer-Verlag London 1999
Origina11y published by Springer-Verlag London Berlin Heidelberg in 1999
Softcover reprint of the hardcover 1s t edition 1999
Figure 4.3, parts of Sections 4.1.2.1 and 4.1.2.2 and Section 4.2 and reproduced from Kourti T, Process
analysis, monitoring and diagnosis, using projection methods - a tutorial. Chemometrics and
Intelligent Laboratory Systems 28:3-21 (copyright Elsevier 1995).
MATLAB® is the registered trademark ofThe MathW orks, Inc., http://www.mathworks.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 reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the
publishers, or in the case of reprographic reproduction in accordance with the terms oflicences issued by
the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should 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 therefore free
for general use.
The publisher makes no representation, express or implied, with regard to the accuracy of the
information contained in this book and cannot accept any legal responsibility or liability for any errors or
omissions that may be made.
9 8 7 6 543 2 1
springer.com
XueZ. Wang
Data Mining and Knowledge
Discovery for Process
Monitoring and Control
With 121 Figures
, Springer
Xue Z. Wang, BEng, MSc, PhD
Department of Chemical Engineering, University of Leeds, Leeds. LS2 9JT.
ISBN 978-1-4471-1137-5
British Library Cataloguing in Publication Data
Wang,Xuez.
Data mining and knowledge discovery for process monitoring
and control. - (Advances in industrial control)
1.Process control-Data processing 2.Data mining
LTitle II.McGreavy, C. .
629.8'9
ISBN 978-1-4471-1137-5
Library of Congress Cataloging-in-Publication Data
Wang, X. Z. (Xue Zhang), 1963-
Data mining and knowledge discovery for process monitoring and
Control/ Xue Z. Wang.
p. cm. -- (Advances in industrial control)
Includes bibliographical references and index.
ISBN 978-1-4471-1137-5 ISBN 978-1-4471-0421-6 (eBook)
DOI 10.1007/978-1-4471-0421-6
1. Process control--Data processing. 2. Data mining. 3. Knowledg
acquisition (Expert systems) 1. Title. II. Series.
TSI56.8.W37 1999
670.42'7563--dc2I
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 reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the
publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by
the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent
to the publishers.
© Springer-Verlag London 1999
Originally published by Springer-Verlag London Berlin Heidelberg in 1999
Softcover reprint ofthe hardcover Ist edition 1999
MA TLABiI> is the registered trademark ofThe MathWorks,lnc., htţp:llwww.mathworks.com
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 reguIations and therefore free
for general use.
The publisher makes no representation, express or implied, with regard to the accuracy of the
information contained in this book and cannot accept any legal responsibility or liability for anyerrors or
omissions that may be made.
Typesetting: Camera ready by author
69/3830-543210 Printed on acid-free paper SPIN 10703341
Advances in Industrial Control
Series Editors
Professor Michael J. Grimble, Professor ofIndustrial Systems and Director
Professor Michael A. Johnson, Professor of Control Systems and Deputy Director
Industrial Control Centre
Department of Electronic and Electrical Engineering
University of Strathc1yde
Graham Hills Building
50 George Street
GlasgowGllQE
United Kingdom
Series Advisory Board
Professor Dr-Ing J. Ackermann
DLR Institut fur Robotik und Systemdynamik
Postfach 1116
D82230 WeBling
Germany
Professor I.D. Landau
Laboratoire d'Automatique de Grenoble
ENSIEG, BP 46
38402 Saint Martin d'Heres
France
Dr D.C. McFarlane
Department of Engineering
University of Cambridge
Cambridge CB2 lQJ
United Kingdom
Professor B. Wittenmark
Department of Automatic Control
Lund Institute of Technology
PO Box 118
S-221 00 Lund
Sweden
Professor D.W. Clarke
Department of Engineering Science
University of Oxford
Parks Road
Oxford OXI 3PJ
United Kingdom
Professor Dr -Ing M. Thoma
Institut fur Regelungstechnik
Universitat Hannover
Appelstr. 11
30167 Hannover
Germany
Professor H. Kimura
Department of Mathematical Engineering and Information Physics
Faculty of Engineering
The University of Tokyo
7-3-1 Hongo
Bunkyo Ku
Tokyo 113
Japan
Professor A.J. Laub
College of Engineering -Dean's Office
University of California
One Shields Avenue
Davis
California 95616-5294
United States of America
Professor J.B. Moore
Department of Systems Engineering
The Australian National University
Research School of Physical Sciences
GPO Box4
Canberra
ACT 2601
Australia
Dr M.K. Masten
Texas Instruments
2309 Northcrest
Plano
TX 75075
United States of America
Professor Ton Backx
AspenTech Europe B.Y.
DeWaal32
NL-5684 PH Best
The Netherlands
This book is dedicated to
my wife Yanmin for her love
and to
Professor Colin McGreavy, an
influential figure in process control
who died on June 23rd 1999
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 in 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.
How is information technology having an impact on the control,
monitoring and operation of large-scale industrial processes? This monograph
by Xue Wang of Leeds University supplies an in-depth answer to this question.
The text has the starting point that, traditionally, control engineers have
concentrated on system dynamics, measurement selection, control structure
determination and the choice of the controller algorithm to be used. This
ignores the large quantity of data generated by modern plant sensing devices as
a source of information for improved control and enhanced process
monitoring. The difficulty in many cases is the sheer quantity of data arising
and the problem of interrogating, analysing and interpreting this data.
Solutions, some old ... multivariate analysis, and some new ... neural networks
and fuzzy sets, are presented by the author along with illustrative examples
usually based on real industrial processes and data. This very instructive
monograph will be of interest to the practising industrial process engineer for
its insights and to the academic control community for its industrial
perspective.
M.J. Grimble and M.A. Johnson
Industrial Control Centre
Glasgow, Scotland, UK
PREFACE
Being able to collect and display to operators a large amount of information is
regarded as one of the most important advances provided in distributed control
(DCS) over earlier analogue and direct digital control systems. The data are used by
plant operators and supervisors to develop an understanding of plant operations
through interpretation and analysis. It is this understanding which can then be used
to identify problems in current operations and find better operational regions which
result in improved products or in operating efficiency.
It has long been recognised that the information collected by DCS systems tends
to overwhelm operators and so makes it difficult to take quick and correct decisions,
especially in critical occasions. For example, olefin plants typically have more than
5000 measurements to be monitored, with up to 600 trend diagrams. Clearly there is
a need to develop methodologies and tools to automate data interpretation and
analysis, and not simply rely on providing the operators large volumes of
multivariate data. The role of the acquisition system should be to provide the
operators with information, knowledge, assessment of states of the plant and
guidance in how to make adjustments. Operators are more concerned with the
current status of the process and possible future behaviour rather than the current
values of individual variables.
Process monitoring tends to be conducted at two levels. Apart from immediate
safe operation of the plant, there is also the need to deal with the long term
performance which has been the responsibility of supervisors and engineers. The
databases created by automatic data logging provide potentially useful sources of
insight for engineers and supervisors to identify causes of poor performance and
opportunities for improvement. Up to now such data sources have not been
adequately exploited.
XII
The above roles of plant operators and supervisors imply that they are an integral
part of the overall control system. The current approach to designing control
systems has not adequately addressed this point. It is done mainly in terms of
identifying the process dynamics, selecting measurements, defining control
structures and selecting algorithms.
This book introduces development in automatic analysis and interpretation of
process operational data both in real-time and over the operational history, and
describes new concepts and methodologies for developing intelligent, state space
based systems for process monitoring, control and diagnosis. It is known that
processes can have multiple steady and also abnormal states. State space based
monitoring and diagnosis can project multivariate real-time measurements onto a
point in the operational state plane and monitor the trajectory of the point which can
identify previously unknown states and the influence of individual variables. It is
now possible to exploit data mining and knowledge discovery technologies to the
analysis, representation, and feature extraction of real-time and historical
operational data to give deeper insight into the systems behaviour. The emphasis is
on addressing challenges facing interpretation of process plant operational data,
including the multivariate dependencies which determine process dynamics, noise
and uncertainty, diversity of data types, changing conditions, unknown but feasible
conditions, undetected sensor failures and uncalibrated and misplaced sensors,
without being overwhelmed by the volume of data.
To cover the above themes, it is necessary to cover the following topics,
• new ways of approaching process monitoring, control and diagnosis
• specification of a framework for developing intelligent, state space based
monitoring systems
• introduction to data mining and knowledge discovery
• data pre-processing for feature extraction, dimension reduction, noise
removal and concept formation
• multivariate statistical analysis for process monitoring and control
• supervised and unsupervised methods for operational state identification
• variable causal relationship discovery
• software sensor design
The methodologies and concepts are illustrated by considering illustrative examples
and industrial case studies.
Xue Z. Wang
Leeds, England, 1999