Process Control System Fault Diagnosis : A Bayesian Approach /: A Bayesian Approach. (2016)

• Record Type: Book
• Further Information: Note: Ruben Gonzalez, Fei Qi, Biao Huang.
• Authors: Gonzalez, Ruben; Qi, Fei; Huang, Biao
• Contents: Preface iii Part One Fundamentals 1 1 Introduction 3 1.1 Motivational Illustrations 3 1.2 Previous Work 4 1.2.1 Diagnosis techniques 4 1.2.2 Monitoring techniques 8 1.3 Book Outline 13 1.3.1 Problem overview and illustrative example 13 1.3.2 Proposed work 13 References 17 2 Prerequisite fundamentals 21 2.1 Introduction 21 2.2 Bayesian Inference and Parameter Estimation 21 2.2.1 Tutorial on Bayesian Inference 26 2.2.2 Tutorial on Bayesian Inference with Time Dependency 30 2.2.3 Bayesian Inference vs Direct Inference 36 2.2.4 Tutorial on Bayesian Parameter Estimation 37 2.3 The EM Algorithm 43 2.3.1 Tutorial: Solution for General Distributions 43 2.4 Techniques for Ambiguous Modes 50 2.4.1 Tutorial on Ɵ parameters in the presence of ambiguous modes 52 2.4.2 Tutorial on probabilities using Ɵ parameters 52 2.4.3 Dempster-Shafer Theory 54 2.5 Kernel Density Estimation 57 2.5.1 From histograms to kernel density estimates 58 2.5.2 Bandwidth Selection 60 2.5.3 Kernel Density Estimation Tutorial 61 2.6 Bootstrapping 64 2.6.1 Bootstrapping Tutorial 64 2.6.2 Smoothed Bootstrapping Tutorial 65 2.7 Notes and References 68 References 69 3 Bayesian diagnosis 71 3.1 Introduction 71 3.2 Bayesian Approach for Control Loop Diagnosis 71 3.2.1 Mode M 71 3.2.2 Evidence E 72 3.2.3 Historical data set D 73 3.3 Likelihood Estimation 74 3.4 Notes and References 76 References 76 4 Accounting for autodependent modes and evidence 77 4.1 Introduction 77 4.2 Temporally dependent evidence 78 4.2.1 Evidence dependence 78 4.2.2 Estimation of evidence transition probability 79 4.3 Temporally dependent modes 84 4.3.1 Mode dependence 84 4.3.2 Estimating mode transition probabilities 87 4.4 Dependent modes and evidence 92 4.5 Notes and References 93 References 94 5 Accounting for incomplete discrete evidence 95 5.1 Introduction 95 5.2 Incomplete evidence problem 95 5.2.1 The unique underlying complete evidence matrix (UCEM) 96 5.3 Diagnosis with incomplete evidence 97 5.3.1 Single missing pattern problem 98 5.3.2 Multiple missing pattern problem 105 5.3.3 Limitations of the single and multiple missing pattern solutions106 5.4 Notes and References 107 References 107 6 Accounting for ambiguous modes: A Bayesian approach 109 6.1 Introduction 109 6.2 Parametrization of likelihood given ambiguous modes 109 6.2.1 Interpretation of proportion parameters 109 6.2.2 Parametrizing likelihoods 111 6.2.3 Informed estimates of likelihoods 112 6.3 Fagin-Halpern combination 113 6.4 Second-order approximation 114 6.4.1 Consistency of Ɵ parameters 114 6.4.2 Obtaining a second-order approximation 115 6.4.3 The second-order Bayesian combination rule 116 6.5 Brief comparison of combination methods 118 6.6 Applying the second-order rule dynamically 119 6.6.1 Unambiguous dynamic solution 119 6.6.2 The second-order dynamic solution 120 6.7 Making a diagnosis 121 6.7.1 Simple diagnosis 121 6.7.2 Ranged diagnosis 121 6.7.3 Expected value diagnosis 122 6.8 Notes and References 125 References 126 7 Accounting for ambiguous modes: A Dempster-Shafer approach 127 7.1 Introduction 127 7.2 Dempster-Shafer Theory 128 7.2.1 Basic Belief Assignments 128 7.2.2 Probability boundaries 129 7.2.3 Dempster’s rule of combination 130 7.2.4 Short-cut combination for unambiguous priors 131 7.3 Generalizing Dempster-Shafer Theory 132 7.3.1 Motivation: Difficulties with BBAs 132 7.3.2 Generalizing the BBA 135 7.3.3 Generalizing Dempster’s rule 138 7.3.4 Short-cut combination for unambiguous priors 139 7.4 Notes and References 140 References 140 8 Making use of continuous evidence through kernel density estimation 141 8.1 Introduction 141 8.2 Performance: continuous methods vs. discrete 142 8.2.1 Average false negative diagnosis criterion 143 8.2.2 Performance of discrete methods vs continuous methods 144 8.3 Kernel density estimation 147 8.3.1 From histograms to kernel density estimates 147 8.3.2 Defining a kernel density estimate 149 8.3.3 Bandwidth selection criterion 151 8.3.4 Bandwidth selection techniques 152 8.4 Dimension Reduction 153 8.4.1 Independence assumptions 154 8.4.2 Principal and independent component analysis 155 8.5 Missing Values 156 8.5.1 Kernel density regression 156 8.5.2 Applying kernel density regression for a solution 158 8.6 Dynamic Evidence 159 8.7 Notes and References 160 References 160 9 Accounting for sparse data within a mode 161 9.1 Introduction 161 9.2 Analytical estimation of monitor output distribution function 162 9.2.1 Control performance monitor 162 9.2.2 Process model monitor 163 9.2.3 Sensor bias monitor 165 9.3 Bootstrap approach to estimate monitor output distribution function 167 9.3.1 Valve stiction identification 167 9.3.2 The bootstrap method 169 9.3.3 Illustrative example 173 9.3.4 Applications 178 9.4 Experimental example 184 9.4.1 Process description 184 9.4.2 Diagnostic settings and results 186 9.5 Notes and references 188 References 188 10 Accounting for sparse modes within the data 193 10.1 Introduction 193 10.2 Algorithms 193 10.2.1 Algorithm for component diagnosis 194 10.2.2 Algorithm for bootstrapping new modes 197 10.3 Illustration 202 10.3.1 Component-based diagnosis 206 10.3.2 Bootstrapping for additional modes 210 10.4 Application 217 10.4.1 Monitor Selection 217 10.4.2 Component Diagnosis 218 10.5 Notes and references 219 References 222 Part Two Application 223 11 Introduction to testbed systems 225 11.1 Simulated System 225 11.1.1 Monitor design 225 11.2 Bench Scale System 227 11.3 Industrial Scale System 229 References 229 12 Bayesian Diagnosis with discrete data 231 12.1 Introduction 231 12.2 Algorithm 232 12.3 Tutorial 234 12.4 Simulated Case 239 12.5 Bench Scale Case 241 12.6 Industrial Scale Case 241 12.7 Notes and References 243 References 243 13 Accounting for Autodependent Modes and Evidence 245 13.1 Introduction 245 13.2 Algorithms 246 13.2.1 Evidence transition probability 246 13.2.2 Mode transition probability 251 13.3 Tutorial 253 13.4 Example Systems 256 13.5 Notes and References 256 References 257 14 Accounting for incomplete discrete evidence 259 14.1 Introduction 259 14.2 Algorithm 259 14.2.1 Multiple missing pattern problem 264 14.3 Tutorial 266 14.4 Simulated Case 269 14.5 Bench Scale Case 271 14.6 Industrial Scale Case 272 14.7 Notes and References 274 &lt
• Edition: 1st
• Publisher Details: Wiley
• Publication Date: 2016
• Extent: 1 online resource (360 pages)
• Subjects: 670.427
• Languages: English
• ISBNs: 9781118770597
• Access Rights: Legal Deposit; Only available on premises controlled by the deposit library and to one user at any one time; The Legal Deposit Libraries (Non-Print Works) Regulations (UK).
• Access Usage: Restricted: Printing from this resource is governed by The Legal Deposit Libraries (Non-Print Works) Regulations (UK) and UK copyright law currently in force.