Detection and Diagnosis of Stiction in Control Loops

Mohieddine Mali Biao Huang Editors with M.A.A. Shoukat Choudhury, Peter He, Alexander Horch, Manabu Kano, Nazmul Karim, Srinivas Karra, Hidekazu Kugemoto, Kwan-Ho Lee, S. Joe Qin, Claudio Scali, Zhengyun Ren, Maurizio Rossi, Timothy Salsbury, Sirish L. Shah, Ashish Singhal, Nina F. Thornhill, Jin Wang and Yoshiyuki Yamashita Detection and Diagnosis of Stiction in Control Loops State of the Art and Advanced Methods <ÖSp rineer g<

Contents List of Contributors Abbreviations and Acronyms xxv xxix 1 Introduction 1 Mohieddine Mali and Biao Huang 1.1 Motivation 1 1.2 Typical Valve-controlled Loop 2 1.3 Stiction Phenomenon and Related Effects 4 1.4 Input-Output Relation of Valves Under Stiction 6 1.5 Limit Cycles due to Stiction 9 1.6 Typical Observations in Control Loops with Sticky Valves 12 1.7 Industrial Examples of Loops with Stiction 15 1.8 Summary and Conclusions 18 Part I Stiction Modelling and Oscillation Detection 2 Stiction Modelling 21 M.A.A. Shoukat Choudhury, Nina F. Thornhill, Manabu Kano and Sirish L. Shah 2.1 Introduction 22 2.2 Physics-based Stiction Modelling 22 2.3 Data-driven Stiction Modelling 24 2.3.1 One-parameter Stiction Model 24 2.3.2 Two-parameter Stiction Model 24 2.3.3 Choudhury's Stiction Model 25 2.3.4 Simulation of the Stiction Model 28 2.3.5 Kano's Stiction Model 29 2.4 Comparison Between Choudhury's and Kano's Stiction Models.. 32 2.4.1 Similarities 32 2.4.2 Differences 32 xvii

xviii Contents 2.4.3 Comparisons Using an Industrial Example 32 2.5 Summary and Conclusions 35 3 An Alternative Stiction-modelling Approach and Comparison of Different Stiction Models 37 Q. Peter He, Jin Wang and S. Joe Qin 3.1 Introduction 37 3.2 He's Two-parameter Model 38 3.3 Three Data-driven Models 40 3.3.1 Implementation of the First-principles Model 41 3.3.2 Comparison of Data-driven Models 43 3.4 Further Investigation of Valve Stiction 45 3.5 He's Three-parameter Model 49 3.6 Simulation Results 51 3.7 An Industrial Example 56 3.8 Summary and Conclusions 57 3.9 Appendix: Proof of the Equivalence Between He's Two-parameter and Three-parameter Model 58 4 Detection of Oscillating Control Loops 61 Srinivas Karra, Mohieddine Jelali, M. Nazmul Karim and Alexander Horch 4.1 Introduction 62 4.2 Root-causes for Oscillatory Control Loops 62 4.2.1 Poor Process and Control System Design 62 4.2.2 Aggressive Controller Tuning 63 4.2.3 Non-linearities in Control-loop Hardware 63 4.2.4 External Oscillatory Disturbances 64 4.3 Characterisation of Oscillations 64 4.3.1 Auto-covariance Function 64 4.3.2 Power Spectrum 65 4.3.3 Strength of Oscillations 65 4.4 Techniques for Detection of Oscillations in Control Loops 67 4.4.1 Detection of Spectral Peaks 67 4.4.2 Regularity of Large Enough Integral of Absolute Error.. 69 4.4.3 Regularity of Upper and Lower IAEs and Zero-crossings 72 4.4.4 Decay-ratio Approach of the Auto-correlation Function.. 74 4.4.5 Regularity of Zero-crossings of the Auto-correlation Function 76 4.4.6 Spectral Envelope Method 78 4.5 Critical Evaluation of Oscillation-detection Methods 80 4.5.1 Features of Industrial Control-loop-oscillation Detection. 80 4.5.2 Detection Test Examples 81 4.5.3 Signal with Coloured Noise 83 4.5.4 Signal with One Predominant Oscillation 84 4.5.5 Signal with Dampened Oscillation 85

Contents xix 4.5.6 Signal with Multiple Oscillations 87 4.5.7 Signal with Intermittent Oscillations 90 4.6 Comprehensive Oscillation Characterisation 93 4.7 Industrial Case Studies 94 4.7.1 Oscillating Flow Control Loop 95 4.7.2 Unit-wide Oscillation Caused by a Sensor Fault 96 4.7.3 Plant-wide Oscillation Caused by a Valve Fault 97 4.8 Summary and Conclusions 100 Part II Advances in Stiction Detection and Quantification 5 Shape-based Stiction Detection 103 Manabu Kano, Yoshiyuki Yamashita and Hidekazu Kugemoto 5.1 Introduction 103 5.2 Method Description 104 5.2.1 Method A 104 5.2.2 Method В 105 5.2.3 Method С 106 5.3 Key Issues 107 5.4 Simulation Results 108 5.5 Application to Industrial Loops Ill 5.6 Summary and Conclusions 113 6 Stiction Detection Based on Cross-correlation and Signal Shape... 115 Alexander Horch 6.1 Introduction 115 6.2 The Cross-correlation Function 117 6.3 Industrial Examples 120 6.3.1 Loop Interaction I 120 6.3.2 Loop Interaction II 121 6.3.3 Flow Control Loop I 123 6.3.4 Flow Control Loop II 125 6.3.5 Level Control 125 6.4 Theoretical Explanation 126 6.4.1 Correlation for Oscillating External Disturbances 127 6.4.2 Tight Tuning 129 6.4.3 Correlation in the Presence of Stiction 129 6.5 Conclusions (Cross-correlation Method) 131 6.6 Stiction Detection for Integrating Processes 132 6.7 Detection in Integrating Loops - Basic Idea 132 6.7.1 Differentiation and Filtering 133 6.7.2 Sample Histogram 135 6.7.3 Distribution for the Stiction Case 136 6.7.4 Distribution for the Non-stiction Case 138 6.8 Examples 142 6.8.1 Level Control Loop with Stiction 142

xx Contents 6.8.2 Level Control Loop Without Stiction 143 6.8.3 Level Control Loop with Deadband 143 6.9 Self-regulating Processes 143 6.9.1 Flow Control Loop with Stiction 144 6.9.2 Flow Control Loop Without Stiction 145 6.9.3 Loops with Dominant P-control 145 6.10 Summary and Conclusions 147 7 Curve Fitting for Detecting Valve Stiction 149 Q. Peter He and S. Joe Qin 7.1 Introduction 149 7.2 Method Description 151 7.2.1 Sinusoidal Fitting 152 7.2.2 Triangular Fitting 153 7.2.3 Stiction Index 153 7.3 Key Issues 154 7.4 Simulation Results 154 7.5 Application to Industrial Loops 159 7.6 Summary and Conclusions 161 8 A Relay-based Technique for Detection of Stiction 165 Claudio Scali and Maurizio Rossi 8.1 Introduction 166 8.2 Trends of Different Variables 168 8.3 Method Description 170 8.3.1 Basic Idea 170 8.3.2 Stiction Index 171 8.3.3 Fitting Procedure 172 8.3.4 Fitting Algorithm 172 8.4 Simulation Results 175 8.4.1 Nominal Case 175 8.4.2 Presence of Noise 177 8.5 Application to Plant Data 178 8.6 Summary and Conclusions 181 9 Shape-based Stiction Detection Using Area Calculations 183 Timothy I. Salsbury and Ashish Singhal 9.1 Introduction 183 9.2 Method Description 185 9.2.1 Theoretical Basis 187 9.2.2 Stiction Detection Hypothesis Test 190 9.2.3 Noise Effects and Practical Implementation 191 9.3 Key Issues 197 9.4 Simulation Results 198 9.5 Application to Industrial Loops 200

Contents xxi 9.5.1 Temperature Control Loop with Stiction from a Building Automation System 201 9.5.2 Temperature Control Loop with Stiction from a Pulp and Paper Plant 202 9.6 Summary and Conclusions 203 10 Estimation of Valve Stiction Using Separable Least-squares and Global Search Algorithms 205 Mohieddine Jelali 10.1 Introduction 205 10.2 Basic Approach 207 10.2.1 Identification Model Structure: Hammerstein Model...208 10.2.2 Linear Model 208 10.2.3 Stiction Model 209 10.3 Identification Approach 210 10.3.1 Separable Least-squares Estimator 210 10.3.2 Global Search Algorithms 213 10.4 Key Issues 215 10.4.1 Model Structure Selection 215 10.4.2 Time-delay Estimation 216 10.4.3 Determination of Initial Parameters and Incorporation of Constraints 218 10.5 Simulation Studies 219 10.5.1 First-order-plus-time-delay Process 220 10.5.2 Integrating Process with Time Delay 220 10.6 Industrial Case Studies 221 10.6.1 Loop CHEM 25: Pressure Control Loop 221 10.6.2 Loop PAP 2: Flow Control Loop 224 10.6.3 Loop CHEM 24: Flow Control Loop with Setpoint Changes 224 10.6.4 Loop POW 2: Level Control Loop 225 10.6.5 Loop POW 4: Level Control Loop 226 10.6.6 Loop MIN 1: Temperature Control Loop 226 10.6.7 Loop CHEM 70: Flow Control Loop with External Disturbances 226 10.7 Summary and Conclusions 227 11 Stiction Estimation Using Constrained Optimisation and Contour Map 229 Kwan Ho Lee, Zhengyun Ren and Biao Huang 11.1 Introduction 229 11.2 Stiction Model of Control Valve 231 11.2.1 General Conception 231 11.2.2 Physical Model of Valve Sticion 231 11.2.3 Kano's Valve-stiction Model 232 11.2.4 Choudhury's Valve-stiction Model 232

Contents 11.2.5 He's Valve-stiction Model 232 11.3 Existing Stiction-detection Methods 233 11.3.1 Open-loop Methods 233 11.3.2 Closed-loop Methods 233 11.3.3 Discussion of Existing Methods 234 11.4 Closed-loop Stiction Detection and Quantification 235 11.4.1 Basic Principle and Important Steps 235 11.4.2 Stiction Detection and Quantification Procedure 236 11.4.3 Search Space of Stiction-model Parameters 237 11.4.4 Constrained Parameter-search Techniques 238 11.4.5 Advantages 240 11.5 Stiction Detection: Identifiability Analysis 241 11.5.1 Heuristic Illustration of Closed-loop Identifiability 241 11.5.2 Identifiability Analysis for Closed-loop Systems with Valve Stiction 243 11.6 Simulations 245 11.7 Industrial Applications 251 11.7.1 Illustrative Industrial Examples 251 11.7.2 Comparative Study 256 11.8 Graphical User Interface 260 11.9 Summary and Conclusions 265 Oscillation Root-cause Detection and Quantification Under Multiple Faults 267 Srinivas Karra and M. Nazmul Karim 12.1 Introduction 267 12.2 Preliminaries and Brief Review of Model-based Oscillation Diagnosis 268 12.2.1 Root-cause for Oscillations and Compensation Techniques 268 12.2.2 Oscillation Diagnosis and Root-cause Quantification... 269 12.2.3 Challenges to be Addressed 270 12.3 Overview of the Root-cause Detection and Quantification Methodology 270 12.3.1 Revisiting Control-valve Characteristics Under Stiction.. 270 12.3.2 Oscillation Detection and Diagnosis Methodology 271 12.4 Process-model Identification Under Non-stationary Disturbances. 271 12.4.1 Identification of EARMAX Model 273 12.4.2 Illustrative Example: Identification Under Non-stationary Disturbance 275 12.5 Root-cause Detection and Quantification 278 12.5.1 OP-PV Model Identification Methodology 278 12.5.2 Identification of Controller Transfer Function 280 12.5.3 Oscillation Root-cause Detection and Quantification Methodology 280

12.6 Illustrative Example: Oscillation Diagnosis Under Various Faulty Situations 281 12.6.1 Stiction 281 12.6.2 Oscillatory External Disturbance 283 12.6.3 Aggressive Controller Tuning 284 12.6.4 Stiction and Oscillatory External Disturbance 285 12.6.5 Stiction and Aggressive Controller Tuning 286 12.6.6 Oscillatory External Disturbance and Aggressive Controller Tuning 287 12.6.7 Stiction, Aggressive Controller Tuning and Oscillatory External Disturbance 288 12.7 Industrial Case Studies 290 12.7.1 Control Loop 1 290 12.7.2 Control Loop 2 291 12.7.3 Control Loop 3 292 12.8 Summary and Conclusions 293 Comparative Study of Valve-stiction-detection Methods 295 Mohieddine Mali and Claudio Scali 13.1 Introduction 295 13.2 Selected Methods 296 13.3 Industrial Control Loops Involved in the Study 298 13.4 Application Results and Discussion 302 13.4.1 Application Results 302 13.4.2 Synthesis and Discussion 303 13.4.3 Efficiency of the Techniques, Problems and Countermeasures 306 13.4.4 Comparison on 20 Loops with Known Problems 317 13.4.5 Selected Examples 317 13.4.6 Graphical User Interface 319 13.5 Suggestions 321 13.6 Summary and Conclusions 323 13.7 Appendix: Tables of Results of the Comparative Study 324 Conclusions and Future Research Challenges 359 Biao Huang, Mohieddine Jelali and Alexander Horch 14.1 Summary of the Book 359 14.2 Future Research Challenges 362 14.2.1 Stiction Modelling 362 14.2.2 Oscillation Detection 363 14.2.3 Stiction Detection and Estimation 364 14.2.4 Stiction Control 365

xxiv Contents Appendix A Evaluated Industrial Control Loops 367 Appendix В Review of Some Non-linearity and Stiction-detection Techniques 371 B.l Bicoherence Method 371 B.l.l Non-Gaussianity Index 372 B.1.2 Non-linearity Index 373 B.1.3 Total Non-linearity Index 374 B.1.4 Ellipse Fitting 374 B.2 Surrogates Analysis 374 B.2.1 Surrogate Data Generation 375 B.2.2 Non-linear Predictability Index 376 References 377 Contributor Biographies 383 Index 389