Optimal Automated Process Fault Analysis E-Book

Richard J. Fickelscherer

• Herausgeber: John Wiley & Sons
• Kategorie: Fachliteratur
• Sprache: Englisch

Seitenzahl: 350

Veröffentlichungsjahr: 2012

Beschreibung

Tested and proven strategy to develop optimal automated process fault analyzers Process fault analyzers monitor process operations in order to identify the underlying causes of operational problems. Several diagnostic strategies exist for automating process fault analysis; however, automated fault analysis is still not widely used within the processing industries due to problems of cost and performance as well as the difficulty of modeling process behavior at needed levels of detail. In response, this book presents the method of minimal evidence (MOME), a model-based diagnostic strategy that facilitates the development and implementation of optimal automated process fault analyzers. MOME was created at the University of Delaware by the researchers who developed the FALCON system, a real-time, online process fault analyzer. The authors demonstrate how MOME is used to diagnose single and multiple fault situations, determine the strategic placement of process sensors, and distribute fault analyzers within large processing systems. Optimal Automated Process Fault Analysis begins by exploring the need to automate process fault analysis. Next, the book examines: * Logic of model-based reasoning as used in MOME * MOME logic for performing single and multiple fault diagnoses * Fuzzy logic algorithms for automating MOME * Distributing process fault analyzers throughout large processing systems * Virtual SPC analysis and its use in FALCONEER(TM) IV * Process state transition logic and its use in FALCONEER(TM) IV The book concludes with a summary of the lessons learned by employing FALCONEER(TM) IV in actual process applications, including the benefits of "intelligent supervision" of process operations. With this book as their guide, readers have a powerful new tool for ensuring the safety and reliability of any chemical processing system.

Leseprobe

Contents

Cover

Title Page

Copyright

Dedicatioon

Foreword

Preface

Acknowledgments

CHAPTER 1: MOTIVATIONS FOR AUTOMATING PROCESS FAULT ANALYSIS

1.1 INTRODUCTION

1.2 CPI TRENDS TO DATE

1.3 THE CHANGING ROLE OF PROCESS OPERATORS IN PLANT OPERATIONS

1.4 METHODS CURRENTLY USED TO PERFORM PROCESS FAULT MANAGEMENT

1.5 LIMITATIONS OF HUMAN OPERATORS IN PERFORMING PROCESS FAULT MANAGEMENT

1.6 THE ROLE OF AUTOMATED PROCESS FAULT ANALYSIS

1.7 ANTICIPATED FUTURE CPI TRENDS

1.8 PROCESS FAULT ANALYSIS CONCEPT TERMINOLOGY

REFERENCES

CHAPTER 2: METHOD OF MINIMAL EVIDENCE: MODEL BASED REASONING

2.1 OVERVIEW

2.2 INTRODUCTION

2.3 METHOD OF MINIMAL EVIDENCE OVERVIEW

2.4 VERIFYING THE VALIDITY AND ACCURACY OF THE VARIOUS PRIMARY MODELS

2.5 SUMMARY

REFERENCES

CHAPTER 3: METHOD OF MINIMAL EVIDENCE: DIAGNOSTIC STRATEGY DETAILS

3.1 OVERVIEW

3.2 INTRODUCTION

3.3 MOME DIAGNOSTIC STRATEGY

3.4 GENERAL PROCEDURE FOR DEVELOPING AND VERIFYING COMPETENT MODEL-BASED PROCESS FAULT ANALYZERS

3.5 MOME SV&PFA DIAGNOSTIC RULES' LOGIC COMPILER MOTIVATIONS

3.6 MOME DIAGNOSTIC STRATEGY SUMMARY

REFERENCES

CHAPTER 4: METHOD OF MINIMAL EVIDENCE: FUZZY LOGIC ALGORITHM

4.1 OVERVIEW

4.2 INTRODUCTION

4.3 FUZZY LOGIC OVERVIEW

4.4 MOME FUZZY LOGIC ALGORITHM

4.5 CERTAINTY FACTOR CALCULATION REVIEW

4.6 MOME FUZZY LOGIC ALGORITHM SUMMARY

REFERENCES

CHAPTER 5: METHOD OF MINIMAL EVIDENCE: CRITERIA FOR SHREWDLY DISTRIBUTING FAULT ANALYZERS AND STRATEGIC PROCESS SENSOR PLACEMENT

5.1 OVERVIEW

5.2 CRITERIA FOR SHREWDLY DISTRIBUTING PROCESS FAULT ANALYZERS

5.3 CRITERIA FOR STRATEGIC PROCESS SENSOR PLACEMENT

REFERENCES

CHAPTER 6: VIRTUAL SPC ANALYSIS AND ITS ROUTINE USE IN FALCONEERTM IV

6.1 OVERVIEW

6.2 INTRODUCTION

6.3 EWMA CALCULATIONS AND SPECIFIC VIRTUAL SPC ANALYSIS CONFIGURATIONS

6.4 VIRTUAL SPC ALARM TRIGGER SUMMARY

6.5 VIRTUAL SPC ANALYSIS CONCLUSIONS

REFERENCES

CHAPTER 7: PROCESS STATE TRANSITION LOGIC AND ITS ROUTINE USE IN FALCONEERTM IV

7.1 TEMPORAL REASONING PHILOSOPHY

7.2 INTRODUCTION

7.3 STATE IDENTIFICATION ANALYSIS CURRENTLY USED IN FALCONEERTM IV

7.4 STATE IDENTIFICATION ANALYSIS SUMMARY

REFERENCES

CHAPTER 8: CONCLUSIONS

8.1 OVERVIEW

8.2 SUMMARY OF THE MOME DIAGNOSTIC STRATEGY

8.3 FALCON, FALCONEER, AND FALCONEERTM IV ACTUAL KBS APPLICATION PERFORMANCE RESULTS

8.4 FALCONEERTM IV KBS APPLICATION PROJECT PROCEDURE

8.5 OPTIMAL AUTOMATED PROCESS FAULT ANALYSIS CONCLUSIONS

REFERENCES

APPENDIX A: VARIOUS DIAGNOSTIC STRATEGIES FOR AUTOMATING PROCESS FAULT ANALYSIS

A.1 INTRODUCTION

A.2 FAULT TREE ANALYSIS

A.3 ALARM ANALYSIS

A.4 DECISION TABLES

A.5 SIGN-DIRECTED GRAPHS

A.6 DIAGNOSTIC STRATEGIES BASED ON QUALITATIVE MODELS

A.7 DIAGNOSTIC STRATEGIES BASED ON QUANTITATIVE MODELS

A.8 ARTIFICIAL NEURAL NETWORK STRATEGIES

A.9 KNOWLEDGE-BASED SYSTEM STRATEGIES

A.10 METHODOLOGY CHOICE CONCLUSIONS

REFERENCES

APPENDIX B: THE FALCON PROJECT

B.1 INTRODUCTION

B.2 OVERVIEW

B.3 THE DIAGNOSTIC PHILOSOPHY UNDERLYING THE FALCON SYSTEM

B.4 TARGET PROCESS SYSTEM

B.5 THE FALCON SYSTEM

B.6 DERIVATION OF THE FALCON DIAGNOSTIC KNOWLEDGE BASE

B.7 THE IDEAL FALCON SYSTEM

B.8 USE OF THE KNOWLEDGE BASED SYSTEM PARADIGM IN PROBLEM SOLVING

REFERENCES

APPENDIX C: PROCESS STATE TRANSITION LOGIC USED BY THE ORIGINAL FALCONEER KBS

C.1 INTRODUCTION

C.2 POSSIBLE PROCESS OPERATING STATES

C.3 SIGNIFICANCE OF PROCESS STATE IDENTIFICATION AND TRANSITION DETECTION

C.4 METHODOLOGY FOR DETERMINING PROCESS STATE IDENTIFICATION

C.5 PROCESS STATE IDENTIFICATION AND TRANSITION LOGIC PSEUDOCODE

C.6 SUMMARY

APPENDIX D: FALCONEER™ IV REAL-TIME SUITE PROCESS PERFORMANCE SOLUTIONS DEMOS

D.1 FALCONEER™ IV DEMOS Overview

D.2 FALCONEER™ IV DEMOS

INDEX

Cover design: John Wiley & Sons, Inc. Cover image: Courtesy of Richard J. Fickelscher

Copyright © 2013 by John Wiley & Sons. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, 978-750-8400, fax [phone], or on the web at www.copyright.com. Requests to the Publisher for permission should be addressed to the Permissions Department, John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, 201-748-6011, fax [phone], or online at http://www.wiley.com/go/permission.

Limit of Liability/Disclaimer of Warranty: While the publisher and author have used their best efforts in preparing this book, they make no representations or warranties with respect to the accuracy or completeness of the contents of this book and specifically disclaim any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives or written sales materials. The advice and strategies contained herein may not be suitable for your situation. You should consult with a professional where appropriate. Neither the publisher nor author shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our Customer Care Department within the United States at 877-762-2974, outside the United States at 317-572-3993 or fax [phone].

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. For more information about Wiley products, visit our web site at www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Fickelscherer, Richard J. Optimal automated process fault analysis / by Richard J. Fickelscherer, Ph.D., P.E. co-founder, Falconeer Technologies, LLC & Daniel L. Chester, Ph.D. co-founder, Falconeer Technologies, LLC & associate chair, Department of Computer & Information Sciences, University of Delaware. pages cm Includes index. ISBN 978-1-118-37231-9 (cloth) 1. Chemical process control–Data processing. 2. Fault location (Engineering)–Data processing. I. Chester, Daniel L. II. Title. TP155.75.F527 2013 660′.2815–dc23

2012023637

R.J.F.: I dedicate this book to my loving wife, Pat, for all her encouragement and help, and for always believing in me. I am very fortunate to have found her and to have her at the center of my life.

D.L.C.: I dedicate this book to my parents, Fred and Della, for all the guidance they gave me over the years, and for making it possible for me to get the training and develop the skills that led to my contributions to this book. Their encouragement has always been appreciated.

FOREWORD

It is an honor to be asked to write the foreword to Rich and Dan's book on diagnostic reasoning for process plants. The story of the FALCON diagnostic system goes back to when I first joined MIT as a young faculty member in the early 1980s. In those days computing meant numerical computation in Fortran or C, mainframes and minicomputers ruled, and personal computers were underpowered novelties. Process control had begun a slow transition from pneumatic to digital instrumentation, but the first digital controllers were modeled unimaginatively after the PID loops they were replacing.

But Metcalfe's law was in full exponential ascent, and the world was rapidly changing, not only in terms of faster numerical methods. New ideas from artificial intelligence were flooding across the MIT campus, upending the very foundation of computing by means of an entirely new synthesis of symbolic, object-oriented, neural, and rule-based computing. Touching off a great intellectual ferment in chemical engineering, virtually every aspect of process operations was being transformed. Gauges transformed into graphical operator interfaces, fixed threshold alarms into intelligent monitoring and diagnosis, and steady-state operation into dynamic economic optimization.

The University of Delaware was one of the leaders in this exploration, especially in the area of process monitoring and fault diagnosis. Undertaking a joint project with Foxboro and DuPont in the early 1980s, Delaware spearheaded the first industrial application of an expert system, FALCON (fault analysis consultant), for online fault diagnosis of the DuPont adipic acid plant in Victoria, Texas. A key idea, expounded further in this book, was the synthesis of logical (pattern or rule) analysis and quantitative mathematical modeling.

This period of creative experimentation reached its zenith in 1995, at the First International Conference on Intelligent Systems in Process Engineering in Snowmass, Colorado. By that time, I had moved to an MIT spin-off, Gensym Corporation, developers of the G2 real-time expert system development environment. G2 was that generation's ultimate synthesis of graphical UI, structured natural language, object-oriented programming, and rule-based processing. The conference showcased innovative knowledge-based systems ranging from product design to intelligent control, optimization, and diagnostics. Flush with the success of building and deploying hundreds of expert systems to solve real industrial problems, few of us realized just how quickly another revolution, the Internet, was going to overturn everything, yet again.

Throughout these many changes, some determined individuals have persevered to bring the vision of intelligent operations closer to reality. In this book, Rich and Dan explain how they transformed the art of the diagnostic expert system into a practical and reproducible system, implemented in FALCONEER™ IV to increase the operating safety and reliability of process systems. Their system captures many lessons learned during the rapid and often convulsive change of the past 25 years. I wish them the very best.

Mark A. Kramer, Ph.D.

Winchester, MassachusettsFebruary 2012

PREFACE

Process fault analyzers are computer programs that can monitor process operations to identify the underlying cause(s) of operating problems. A general method for creating process fault analyzers for chemical and nuclear processing plants has been sought ever since the incorporation of computers into process control. The motivation has been the enormous potential for improving process plant operations in terms of safety and productivity. Automated process fault analysis should help process operators (1) prevent catastrophic operating disasters such as explosions, fires, meltdowns, and toxic chemical releases; (2) reduce downtime after emergency process shutdowns; (3) eliminate unnecessary process shutdowns; (4) maintain better quality control of process products; and (5) ultimately, allow both higher process efficiency and higher production levels.

A wide variety of logically viable diagnostic strategies now exist for automating process fault analysis. However, automated fault analysis is currently still not widely used within the processing industries. This is due mainly to the following limitations: (1) the prohibitively large development, verification, implementation, or maintenance costs of these programs; (2) an inability to operate a program based on a given diagnostic strategy continuously online or in realtime; and (3) an inability to model process behavior at the desired level of detail, thus leading to unreliable or highly ambiguous diagnoses. Subsequently, a method for efficient production of automated process fault analyzers is still being actively sought. It is our contention that evaluating engineering models of normal process operation with current process data is the most promising and powerful means of directly identifying underlying process operating problems. Doing so generates an unimpeachable source from which to logically infer the current state of the process being modeled. Performing this inference automatically online enables these programs to perform intelligent supervision of the daily operations of their associated process systems. It makes possible a fundamental understanding of a given process system's design and operation to be utilized in evaluating its current operating conditions.

The method of minimal evidence (MOME) is a model-based diagnostic strategy for developing optimal automated process fault analyzers. It was derived at the University of Delaware while developing the FALCON (fault analysis consultant) system, a real-time online process fault analyzer for a commercial-scale adipic acid plant formerly owned and operated by DuPont in Victoria, Texas. It provides a uniform framework for examining both models of normal process operation and their corresponding associated modeling assumptions that are required to build such fault analyzers. MOME can be used directly to correctly diagnose both single- and multiple-fault situations, to determine the strategic placement of process sensors to facilitate fault analysis, and to determine the shrewd division of a large process system for distributing fault analyzers.

The MOME diagnostic strategy was again demonstrated to be effective in a commercial-scale persulfate plant owned and operated by FMC in Tonawanda, New York. Versions of two knowledge-based systems (KBSs) developed using MOME have been running online at this plant since February 2001. In the current implementation, these KBSs [a.k.a. FALCONEERTM IV (FALCON via engineering equation residuals IV)] diligently perform automated sensor validation and fault analysis of both FMC's electrolytic sodium persulfate and liquid ammonium persulfate processes in realtime. The development effort for these two FALCONEER IV applications was more than two orders of magnitude less than that required for the original FALCON system, with even better performance to date. This impressive improvement in the development and maintenance effort required was possible because FALCONEER IV contains a compiler program that automatically generates the sensor validation and proactive fault analysis (SV&PFA) diagnostic logic required to perform competent fault analysis directly from the underlying engineering models of normal process operation. Since the MOME diagnostic strategy is a systematic procedure, creating an algorithm based on it and then codifying that algorithm proved to be straightforward. This treatment describes both the underlying logic of MOME and the fuzzy logic algorithm based on it. It is meant to be a study guide for those who wish to develop such fault analyzers for their own process systems.

Motivations for automating process fault analysis are described in detail in Chapter 1. Our patented methodology for automating process fault analysis (MOME and its associated fuzzy logic algorithm) is then discussed in detail in Chapters 2 to 5. The logic behind model-based reasoning in general and MOME in particular is described in Chatper 2. The MOME logic for performing single- and multiple-fault diagnosis is described in Chapter 3. Also discussed in Chapter 3 are the motivations behind the creation of process fault analyzers based on MOME automatically via the SV&PFA diagnostic rule logic compiler program contained in FALCONEERTM IV. The fuzzy logic algorithm automating MOME as implemented in this compiler is described in Chapter 4. In Chapter 5 the criteria for shrewdly distributing process fault analyzers throughout a large processing plant are described. Some general guidelines for the strategic placement of process sensors for directly facilitating fault diagnosis are also discussed.

Chapter 6 covers the need to augment process fault analysis with trend analysis of the various process sensor measurements and key performance indicators (KPIs) via the virtual statistical process control (virtual SPC) technique of calculating and analyzing exponentially weighted moving averages (EWMAs).

The need to first determine the current overall operating state of the process undergoing automated fault analysis in discussed in Chapter 7. Such determinations provide the proper context required for the fault analyzer to make legitimate diagnoses.

Chapter 8 summarizes the benefits derived and lessons learned when employing FALCONEERTM IV in actual process applications. A systematic procedure to follow when creating such applications is also described. The chapter concludes by summarizing the advantages of distilling the raw information contained in typical process sensor data continuously into value-added knowledge concerning the current state of process operations and having that knowledge be instantaneously available for intelligent supervision of those operations.

For completeness, four appendixes have been added to this treatment as background information. A number of the other various possible diagnostic strategies also used to automate process fault analysis and their limitations are reviewed briefly in Appendix A. Appendix B describes DuPont's adipic acid plant and the original automated process fault analyzer (i.e., the FALCON system) developed for it. The lessons learned from the development of this real-world fault analyzer are discussed in detail. The advantages of using the knowledge-based system paradigm for solving problems, especially those that led directly to the creation of the MOME diagnostic strategy, are also discussed. As described throughout the book, this strategy has since been codified into FALCONEERTM IV. Appendix C outlines the logic that was used by the original hand-compiled FALCONEER system to determine the current process state in FMC's electrolytic sodium persulfate plant. This logic has since been simplified, generalized, and codified in the current implementation of FALCONEERTM IV. Finally, Appendix D describes two downloadable FALCONEERTM IV demos provided to accompany this treatment.

Richard J. Fickelscherer Daniel L. Chester

ACKNOWLEDGMENTS

First, we would like to thank the other two key chief investigators at the University of Delaware on the original FALCON project, Professors Prasad S. Dhurjati and David E. Lamb, along with many student research assistants, including Oliver J. Smith IV, George M. A'zary, Larry Kramer, Dave Mooney, Lisa Laffend, Kathy Cebulka, Apperson Johnson, and Bob Varrin, Jr. We would also like to thank DuPont and its employees Duncan Rowan, Rick Taylor, John Hale, Robert Wagner, Bob Gardener, and Tim Cole, and especially our domain expert, the late Steve Matusevich. At Foxboro Inc., we would like to thank Dick Shirley, Dave Fortin, and the late Terry Rooney. Further, we thank Oliver J. Smith IV and Duncan Rowan for reviewing earlier versions of this treatment; their comments about ways to improve it were invaluable.

We also thank the FMC Corporation for its support in developing the various FALCONEER and FALCONEERTM IV KBS program applications, especially Doug Lenz, John Rovison, Charlie Lymburner, Weidong An, Don Lapham III, John Helieko, Don Stockhausen, and Jim Kaylor. As a cofounder of FALCONEER Technologies LLC, Doug Lenz strongly advocated including virtual SPC capabilities in FALCONEERTM IV; we would like to acknowledge the programming efforts of Lee Daniels required to accomplish both this and the streamlined state identification capability also currently included in FALCONEERTM IV.

Feel free to contact either of us if you would like to find out more about the FALCONEERTM IV software and our company's services. Following is the contact information:

Dr. Richard J. Fickelscherer, PE Tonawanda, New [email protected]

Prof. Daniel Chester Newark, [email protected]

1

MOTIVATIONS FOR AUTOMATING PROCESS FAULT ANALYSIS

1.1 INTRODUCTION

Economic competition within the chemical process industry (CPI) has led to the construction and operation of larger, highly integrated, and more automated production plants. As a result, the primary functions performed by the process operators in these plants have changed. An unfortunate consequence of such changes is that the operators' ability to perform process fault management has been diminished. The underlying reasons for this problem and the methods currently used to counteract it are discussed here.

1.2 CPI TRENDS TO DATE

The CPI constitutes one of the largest and most important segments of the global economy. While developing into its current, relatively stable position, competition for market share among the various chemical producers has greatly intensified. This competition has, in turn, created continuously downward pressure on the market price, and hence the associated profit margin, of most commodity chemical products. Several major trends within the CPI in the operation of production plants have resulted.