Artificial Intelligence in Process Fault Diagnosis: Methods for Plant Surveillance

Artificial Intelligence in Process Fault Diagnosis A comprehensive guide to the future of process fault diagnosis Automation has revolutionized every aspect of industrial production, from the accumulation of raw materials to quality control inspections. Even process analysis itself has become subject to automated efficiencies, in the form of process fault analyzers, i.e., computer programs capable of analyzing process plant operations to identify faults, improve safety, and enhance productivity. Prohibitive cost and challenges of application have prevented widespread industry adoption of this technology, but recent advances in artificial intelligence promise to place these programs at the center of manufacturing process analysis. Artificial Intelligence in Process Fault Diagnosis brings together insights from data science and machine learning to deliver an effective introduction to these advances and their potential applications. Balancing theory and practice, it walks readers through the process of choosing an ideal diagnostic methodology and the creation of intelligent computer programs. The result promises to place readers at the forefront of this revolution in manufacturing. Artificial Intelligence in Process Fault Diagnosis readers will also find: Coverage of various AI-based diagnostic methodologies elaborated by leading experts Guidance for creating programs that can prevent catastrophic operating disasters, reduce downtime after emergency process shutdowns, and more Comprehensive overview of optimized best practices Artificial Intelligence in Process Fault Diagnosis is ideal for process control engineers, operating engineers working with processing industrial plants, and plant managers and operators throughout the various process industries.

Table of Contents:
List of Contributors xix Foreward xxi Preface xxiii Acknowledgements xxv 1 Motivations for Automating Process Fault Analysis 1 1.1 Introduction 2 1.2 The Changing Role of the Process Operators in Plant Operations 4 1.3 Traditional Methods for Performing Process Fault Management 7 1.4 Limitations of Human Operators in Performing Process Fault Management 8 1.5 The Role of Automated Process Fault Analysis 12 2 Various Process Fault Diagnostic Methodologies 16 2.1 Introduction 17 2.2 Various Alternative Diagnostic Strategies Overview 18 2.3 Diagnostic Methodology Choice Conclusions 35 2.A Failure Modes and Effects Analysis 40 3 Alarm Management and Fault Detection 45 3.1 Introduction 46 3.2 Applicable Definitions and Guidelines 46 3.3 The Alarm Management Life Cycle 49 3.4 Generation of Diagnostic Information 53 3.5 Presentation of the Diagnostic Information 55 3.6 Information Rates 59 4 Operator Performance: Simulation and Automation 63 4.1 Background 63 4.2 Automation 65 4.3 Simulation 68 4.4 Research 69 4.5 AI Integration 73 4.6 Case Study: Turbo Expanders Over-Speed 77 4.7 Human-Centered AI 80 5 AI and Alarm Analytics for Failure Analysis and Prevention 85 5.1 Introduction 86 5.2 Post-Alarm Assessment and Analysis 87 5.3 Real-Time Alarm Activity Database and Operator Action Journal 89 5.4 Pre-Alarm Assessment and Analysis 91 5.5 Utilizing Alarm Assessment Information 92 5.6 Examining the Alarm System to Resolve Failures on a Wider Scale 93 5.7 Emerging Methods of Alarm Analysis 99 5.8 Deep Reinforcement Learning for Alarming and Failure Assessment 103 5.9 Some Typical AI and Machine Learning Examples for Further Study 103 5.10 Wrap-Up 111 5.A Process State Transition Logic Employed by the Original FMC Falconeer KBS 112 5.B Process State Transition Logic and its Routine Use in Falconeer IV 123 6 Process Fault Detection Based on Time-Explicit Kiviat Diagram 131 6.1 Introduction 132 6.2 Time-Explicit Kiviat Diagram 133 6.3 Fault Detection Based on the Time-Explicit Kiviat Diagram 134 6.4 Continuous Processes 136 6.5 Batch Processes 138 6.6 Periodic Processes 140 6.7 Case Studies 141 6.8 Continuous Processes 141 6.9 Batch Processes 144 6.10 Periodic Processes 147 6.11 Conclusions 149 6.A Virtual Statistical Process Control Analysis 151 7 Smart Manufacturing and Real-Time Chemical Process Health Monitoring and Diagnostic Localization 160 7.1 Introduction to Process Operational Health Modeling 163 7.2 Diagnostic Localization – Key Concepts 165 7.3 Time 178 7.4 The Workflow of Diagnostic Localization 184 7.5 DL-CLA Use Case Implementation: Nova Chemical Ethylene Splitter 191 7.6 Analyzing Potential Malfunctions Over Time 198 7.7 Analysis of Various Operational Scenarios 201 7.8 DL-CLA Integration with Smart Manufacturing (SM) 208 7.9 AN FR Model Library 210 7.10 Conclusions 216 8 Optimal Quantitative Model-Based Process Fault Diagnosis 221 8.1 Introduction 222 8.2 Process Fault Analysis Concept Terminology 223 8.3 MOME Quantitative Models Overview 226 8.4 MOME Quantitative Model Diagnostic Strategy 234 8.5 MOME SV&PFA Diagnostic Rules’ Logic Compiler Motivations 248 8.6 MOME Fuzzy Logic Algorithm Overview 250 8.7 Summary of the Mome Diagnostic Strategy 265 8.8 Actual Process System KBS Application Performance Results 266 8.9 Conclusions 267 8.A Falconeer IV Fuzzy Logic Algorithm Pseudo-Code 272 8.B Mome Conclusions 281 9 Fault Detection Using Artificial Intelligence and Machine Learning 286 9.1 Introduction 287 9.2 Artificial Intelligence 287 9.3 Machine Learning 288 9.4 Engineered Features 290 9.5 Machine Learning Algorithms 291 10 Knowledge-Based Systems 300 10.1 Introduction 301 10.2 Knowledge 301 10.3 Information Required for Diagnosis 304 10.4 Knowledge Representation 305 10.5 Maintaining, Updating, and Extending Knowledge 309 10.6 Expert Systems 311 10.7 Digitization, Digitalization, Digital Transformation, and Digital Twins 319 10.8 Fault Diagnosis with Knowledge-Based Systems 322 10.9 Graphical Representation of Fault Diagnosis 325 10.10 Conclusions 337 10.A Compressor Trip Prediction 340 11 The Falcon Project 343 11.1 Introduction 344 11.2 The Diagnostic Philosophy Underlying the Falcon System 345 11.3 Target Process System 346 11.4 The Fielded Falcon System 348 11.5 The Derivation of the FALCON Diagnostic Knowledge Base 355 11.6 The Ideal FALCON System 369 11.7 Use of the Knowledge-Based System Paradigm in Problem 12 Fault Diagnostic Application Implementation and Sustainability 374 12.1 Key Principles of Successfully Implementing New Technology 375 12.2 Expectation of Advanced Technology 376 12.3 Defining Success 379 12.4 Learning from History 379 12.5 Example: Regulatory Control Loop Monitoring 380 12.6 What Success Looks Like 385 12.7 Example: Systematic Stewardship 386 12.8 Conclusions 387 13 Process Operators, Advanced Process Control, and Artificial Intelligence-Based Applications in the Control Room 389 13.1 Introduction 391 13.2 History of Sustainable APC 392 13.3 Operators as Ultimate APC Application End Users 394 13.4 APC Application Design Considerations 395 13.5 APC Development – Internal Versus External Experts 398 13.6 APC Technology 398 13.7 APC Support 400 13.8 Conclusions 402 References 402 Index 404

About the Author :
Richard J. Fickelscherer, PhD, PE has worked on advanced process control and process monitoring programs at DuPont, Exxon, Merck Pharmaceuticals, Koch Industries, and FMC, and has since developed and patented a Fuzzy logic-based compiler program to automate process fault analysis.