Luigi Fortuna, Salvatore Graziani, Alessandro Rizzo and Maria G Xibilia Soft Sensors for Monitoring and Control of Industrial Processes With 179 Figures 123 Luigi Fortuna, Prof., Eng Università degli Studi di Catania Dipartimento di Ingegneria Elettrica Elettronica e dei Sistemi 95125 Catania Italy Salvatore Graziani, Prof., Eng., Ph.D Università degli Studi di Catania Dipartimento di Ingegneria Elettrica Elettronica e dei Sistemi 95125 Catania Italy Alessandro Rizzo, Dr., Eng., Ph.D Politecnico di Bari Dipartimento di Elettrotecnica ed Elettronica 70125 Bari Italy Maria G Xibilia, Dr., Eng., Ph.D Università degli Studi di Messina, Facoltà di Ingegneria Dipartimento di Matematica 98166 Messina Italy British Library Cataloguing in Publication Data Soft sensors for monitoring and control of industrial processes - (Advances in industrial control) 1.Detectors - Design 2.Manufacturing processes Mathematical models 3.Process control 4.Electronic instruments 5.Engineering instruments I.Fortuna, L (Luigi), 1953681.2 ISBN-13: 9781846284793 ISBN-10: 1846284791 Library of Congress Control Number: 2006932285 Advances in Industrial Control series ISSN 1430-9491 ISBN-10: 1-84628-479-1 e-ISBN 1-84628-480-5 ISBN-13: 978-1-84628-479-3 Printed on acid-free paper â Springer-Verlag London Limited 2007 MATLABđ is a registered trademark of The MathWorks, Inc., Apple Hill Drive, Natick, MA 01760-2098, U.S.A 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 of licences 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 987654321 Springer Science+Business Media springer.com 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 Descobrimientos 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 Engineering National University of Singapore Engineering Drive 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 Pennsylvania State University Department of Mechanical Engineering 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 Engineering National University of Singapore Engineering Drive Singapore 117576 Professor Ikuo Yamamoto Kyushu University Graduate School Marine Technology Research and Development Program MARITEC, Headquarters, JAMSTEC 2-15 Natsushima Yokosuka Kanagawa 237-0061 Japan 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 The rapid invasion of industrial and process control applications by low-cost computer hardware, graphical-user-interface technology and high-level software packages has led to the emergence of the virtual instrumentation paradigm In fact, some manufacturers quickly recognised the potential of these different aspects for exploitation in producing virtual instrumentation packages and modules as exemplified by the LabVIEW™ product from National Instruments As this monograph makes clear, virtual instrumentation is a computer-based platform of hardware and software facilities that can be used to create customised instruments for a very wide range of measurement tasks These facilities involve: a user interface to enable the flexible construction, operation and visualisation of the measurement task; computational software to allow advanced processing of the measurement data; and software to integrate hardware units and sensors into the virtual instrument and to orchestrate their operation By way of comparison, Professor Fortuna and his colleagues consider “soft sensors” to be a far narrower concept within the topic of virtual instrumentation, stating that “soft sensors focus on the process of estimation of any system variable or product quality by using mathematical models, substituting some physical sensors and using data acquired from some other available ones.” Thus, the methods in this Advances in Industrial Control monograph have very strong links to the procedures of industrial-process-model identification and validation The monograph opens with three chapters that establish the background to soft sensors; this presentation culminates in Chapter where the complete design process for these sensors is described Chapters 4, and are then sharply viii Series Editors’ Foreword focussed on the key steps in soft sensor design: data selection; model structure selection and model validation, respectively Extensions to the basic steps of softsensor design, namely soft-sensor performance enhancement and the modifications needed to facilitate different industrial process applications follow in Chapter and 8, respectively Widening the applications range and role of soft sensors to fault detection and sensor validation configurations is dealt with in Chapter A great strength of Soft Sensors for Monitoring and Control of Industrial Processes is the use, throughout the text, of a set of industrial case studies to demonstrate the successes and drawbacks of the different methods used to create soft-sensor models A number of different methods may be used in each separate step of the soft-sensor design process and the industrial case studies are often used to provide explicit comparisons of the performance of these methods The industrial control and process engineer will find these comparison exercises invaluable illustrations of the sort of results that might be found in industrial applications The monograph also highlights the importance of using knowledge from industrial experts and from the existing industrial process literature This is an important aspect of industrial control that is not very widely acknowledged or taught in control courses Most industrial processes have already generated a significant experimental knowledge base and the control engineer should develop ways of tapping into this valuable resource when designing industrial control schemes This is a monograph that is full of valuable information about the veracity of different methods and many other little informative asides For example, in Chapter 9, there is a paragraph or two on trends in industrial applications This small section seeks to determine whether and how nonlinear models are used in industrial applications It presents some preliminary data and argument that “the number of nonlinear process applications studied through nonlinear models has been clearly increasing over the years, while nonlinear process applications with linearised models have been decreasing.” A very interesting finding that deserves further in-depth investigation and explanation The industrial flavour of this monograph on soft sensors makes it an apposite volume for the Advances in Industrial Control series It will be appreciated by the industrial control engineer for its practical insights and by the academic control researcher for its case-study applications and performance comparisons of the various theoretical procedures M.J Grimble and M.A Johnson Glasgow, Scotland, U.K Preface This book is about the design procedure of soft sensors and their applications for solving a number of problems in industrial environments Industrial plants are being increasingly required to improve their production efficiency while respecting government laws that enforce tight limits on product specifications and on pollutant emissions, thus leading to ever more efficient measurement and control policies In this context, the importance of monitoring a large set of process variables using adequate measuring devices is clear However, a key obstacle to the implementation of large-scale plant monitoring and control policies is the high cost of on-line measurement devices Mathematical models of processes, designed on the basis of experimental data, via system identification procedures, can greatly help, both to reduce the need for measuring devices and to develop tight control policies Mathematical models, designed with the objectives mentioned above, are known either as virtual sensors, soft sensors, or inferential models In the present book, design procedures for virtual sensors based on data-driven approaches are described from a theoretical point of view, and relevant case studies referring to real industrial applications, are described The purpose of the book is to provide undergraduate and graduate students, researchers, and process technologists from industry, a monograph with basic information on the topic, suggesting step-by-step solutions to problems arising during the design phase A set of industrial applications of soft sensors implemented in the real plants they were designed for, is introduced to highlight their potential Theoretical issues regarding soft sensor design are illustrated in the framework of specific industrial applications This is one of the valuable aspects of the book; in fact, it allows the reader to observe the results of applying different strategies in practical cases Also, the strategies adopted can be adapted to cope with a large number of real industrial problems The book is self-contained and is structured in order to guide the interested reader, even those not closely involved in inferential model design, in the development of their own soft sensors Moreover, a structured bibliography reporting the state of the art of the research into, and the applications of, soft sensors is given x Preface All the case studies reported in the book are the result of collaboration between the authors and a number of industrial partners Some of the soft sensors developed are implemented on-line at industrial plants The book is structured in chapters that reflect the typical steps the designer should follow when developing his own applications The reader can refer to the following scheme as a guide with which to search the book for solutions to particular aspects of a typical soft sensor design Also, soft sensor design procedure is not straightforward and the designer sometimes needs to reconsider part of the design procedure For this reason, in the scheme, a path represented by grey lines overlaps the book structure to represent possible soft sensor design evolution Selection of historical data from plant database, outlier detection, data filtering Chapter Model structure and regressor selection Chapters 5,7 and Model estimation Chapters 5,7 and Model validation Chapter Preface xi The state of the art on research into, and industrial applications of, soft sensors is reported in Chapter Chapters and give some definitions and a short description of theoretical issues concerning soft sensor design procedures Chapter deals with the related topic of model-based fault detection and sensor validation, giving both the state of the art and two applications of sensor validation Technical details of plants used as case studies are reported in the Appendix A As a complement to the bibliography section, where works cited in the book are listed, a structured bibliography is provided, in Appendix B, with the aim of guiding the reader in his or her search for contributions on specific aspects of soft sensor design Readers wishing to apply the techniques for soft sensor design described in the book will find data taken from real industrial applications in the book web site: www.springer.com/1-84628-479-1 Catania, March 2006 Luigi Fortuna Salvatore Graziani Alessandro Rizzo M Gabriella Xibilia Acknowledgments We are most grateful to all those from industry and research laboratories, not forgetting our colleagues, who have been working with us for many years of research in this field In particular, our special thanks go to Bruno Andò, Giuliano Buceti, Paolo Debartolo, Giovanni Di Battista, Vito Marchese, Peppe Mazzitelli, and Mario Sinatra Thanks are also due to Tonino Di Bella and Pietro Giannone, who helped with graphics and simulations Finally, we are indebted to those who helped us in a number of different ways: Doretta and Lina, Giovanna and Gaetano, Michele, Pippo and Meluccia, Francesca, Mario, Sara Eva, and Arturo ... applications range and role of soft sensors to fault detection and sensor validation configurations is dealt with in Chapter A great strength of Soft Sensors for Monitoring and Control of Industrial... through a number of industrial case studies 4 Soft Sensors for Monitoring and Control of Industrial Processes 1.2 State of the Art The literature on soft sensors in industrial applications, concerning... of comparison, Professor Fortuna and his colleagues consider ? ?soft sensors? ?? to be a far narrower concept within the topic of virtual instrumentation, stating that ? ?soft sensors focus on the process