Title Details

MICHAEL G. PECHT, PHD, is Chair Professor in Mechanical Engineering and Professor in Applied Mathematics, Statistics and Scientific Computation at the University of Maryland, USA. He is the Founder and Director of the Center for Advanced Life Cycle Engineering (CALCE) at the University of Maryland, USA, which is funded by more than 150 leading electronics companies. Dr. Pecht is an IEEE, ASME, SAE, and IMAPS Fellow and serves as editor-in-chief of IEEE Access. He has written more than 30 books, 700 technical articles, and has 8 patents.

MYEONGSU KANG, PHD, is currently a Research Associate at the Center for Advanced Life Cycle Engineering (CALCE), University of Maryland, USA. His expertise is in data analytics, machine learning, system modeling, and statistics for prognostics and systems health management. He has authored/coauthored more than 60 publications in leading journals and conference proceedings.

List of Contributors xxiii

Preface xxvii

About the Contributors xxxv

Acknowledgment xlvii

List of Abbreviations xlix

1 Introduction to PHM 1
Michael G. Pecht andMyeongsu Kang

1.1 Reliability and Prognostics 1

1.2 PHM for Electronics 3

1.3 PHM Approaches 6

1.3.1 PoF-Based Approach 6

1.3.1.1 Failure Modes, Mechanisms, and Effects Analysis (FMMEA) 7

1.3.1.2 Life-Cycle Load Monitoring 8

1.3.1.3 Data Reduction and Load Feature Extraction 10

1.3.1.4 Data Assessment and Remaining Life Calculation 12

1.3.1.5 Uncertainty Implementation and Assessment 13

1.3.2 Canaries 14

1.3.3 Data-Driven Approach 16

1.3.3.1 Monitoring and Reasoning of Failure Precursors 16

1.3.3.2 Data Analytics and Machine Learning 20

1.3.4 Fusion Approach 23

1.4 Implementation of PHM in a System of Systems 24

1.5 PHM in the Internet ofThings (IoT) Era 26

1.5.1 IoT-Enabled PHM Applications: Manufacturing 27

1.5.2 IoT-Enabled PHM Applications: Energy Generation 27

1.5.3 IoT-Enabled PHM Applications: Transportation and Logistics 28

1.5.4 IoT-Enabled PHM Applications: Automobiles 28

1.5.5 IoT-Enabled PHM Applications: Medical Consumer Products 29

1.5.6 IoT-Enabled PHM Applications:Warranty Services 29

1.5.7 IoT-Enabled PHM Applications: Robotics 30

1.6 Summary 30

References 30

2 Sensor Systems for PHM 39
Hyunseok Oh,Michael H. Azarian, Shunfeng Cheng, andMichael G. Pecht

2.1 Sensor and Sensing Principles 39

2.1.1 Thermal Sensors 40

2.1.2 Electrical Sensors 41

2.1.3 Mechanical Sensors 42

2.1.4 Chemical Sensors 42

2.1.5 Humidity Sensors 44

2.1.6 Biosensors 44

2.1.7 Optical Sensors 45

2.1.8 Magnetic Sensors 45

2.2 Sensor Systems for PHM 46

2.2.1 Parameters to be Monitored 47

2.2.2 Sensor System Performance 48

2.2.3 Physical Attributes of Sensor Systems 48

2.2.4 Functional Attributes of Sensor Systems 49

2.2.4.1 Onboard Power and Power Management 49

2.2.4.2 Onboard Memory and Memory Management 50

2.2.4.3 Programmable SamplingMode and Sampling Rate 51

2.2.4.4 Signal Processing Software 51

2.2.4.5 Fast and Convenient Data Transmission 52

2.2.5 Reliability 53

2.2.6 Availability 53

2.2.7 Cost 54

2.3 Sensor Selection 54

2.4 Examples of Sensor Systems for PHM Implementation 54

2.5 Emerging Trends in Sensor Technology for PHM 59

References 60

3 Physics-of-Failure Approach to PHM 61
Shunfeng Cheng, Nagarajan Raghavan, Jie Gu, Sony Mathew, and Michael G. Pecht

3.1 PoF-Based PHM Methodology 61

3.2 Hardware Configuration 62

3.3 Loads 63

3.4 Failure Modes, Mechanisms, and Effects Analysis (FMMEA) 64

3.4.1 Examples of FMMEA for Electronic Devices 68

3.5 Stress Analysis 71

3.6 Reliability Assessment and Remaining-Life Predictions 73

3.7 Outputs from PoF-Based PHM 77

3.8 Caution and Concerns in the Use of PoF-Based PHM 78

3.9 Combining PoF with Data-Driven Prognosis 80

References 81

4 Machine Learning: Fundamentals 85
Myeongsu Kang and Noel Jordan Jameson

4.1 Types of Machine Learning 85

4.1.1 Supervised, Unsupervised, Semi-Supervised, and Reinforcement Learning 86

4.1.2 Batch and Online Learning 88

4.1.3 Instance-Based and Model-Based Learning 89

4.2 Probability Theory in Machine Learning: Fundamentals 90

4.2.1 Probability Space and Random Variables 91

4.2.2 Distributions, Joint Distributions, and Marginal Distributions 91

4.2.3 Conditional Distributions 91

4.2.4 Independence 92

4.2.5 Chain Rule and Bayes Rule 92

4.3 Probability Mass Function and Probability Density Function 93

4.3.1 Probability Mass Function 93

4.3.2 Probability Density Function 93

4.4 Mean, Variance, and Covariance Estimation 94

4.4.1 Mean 94

4.4.2 Variance 94

4.4.3 Robust Covariance Estimation 95

4.5 Probability Distributions 96

4.5.1 Bernoulli Distribution 96

4.5.2 Normal Distribution 96

4.5.3 Uniform Distribution 97

4.6 Maximum Likelihood and Maximum A Posteriori Estimation 97

4.6.1 Maximum Likelihood Estimation 97

4.6.2 Maximum A Posteriori Estimation 98

4.7 Correlation and Causation 99

4.8 Kernel Trick 100

4.9 Performance Metrics 102

4.9.1 Diagnostic Metrics 102

4.9.2 Prognostic Metrics 105

References 107

5 Machine Learning: Data Pre-processing 111
Myeongsu Kang and Jing Tian

5.1 Data Cleaning 111

5.1.1 Missing Data Handling 111

5.1.1.1 Single-Value Imputation Methods 113

5.1.1.2 Model-Based Methods 113

5.2 Feature Scaling 114

5.3 Feature Engineering 116

5.3.1 Feature Extraction 116

5.3.1.1 PCA and Kernel PCA 116

5.3.1.2 LDA and Kernel LDA 118

5.3.1.3 Isomap 119

5.3.1.4 Self-Organizing Map (SOM) 120

5.3.2 Feature Selection 121

5.3.2.1 Feature Selection: FilterMethods 122

5.3.2.2 Feature Selection:WrapperMethods 124

5.3.2.3 Feature Selection: Embedded Methods 124

5.3.2.4 Advanced Feature Selection 125

5.4 Imbalanced Data Handling 125

5.4.1 SamplingMethods for Imbalanced Learning 126

5.4.1.1 Synthetic Minority Oversampling Technique 126

5.4.1.2 Adaptive Synthetic Sampling 126

5.4.1.3 Effect of SamplingMethods for Diagnosis 127

References 129

6 Machine Learning: Anomaly Detection 131
Myeongsu Kang

6.1 Introduction 131

6.2 Types of Anomalies 133

6.2.1 Point Anomalies 134

6.2.2 Contextual Anomalies 134

6.2.3 Collective Anomalies 135

6.3 Distance-Based Methods 136

6.3.1 MD Calculation Using an Inverse Matrix Method 137

6.3.2 MD Calculation Using a Gram–Schmidt Orthogonalization Method 137

6.3.3 Decision Rules 138

6.3.3.1 Gamma Distribution:Threshold Selection 138

6.3.3.2 Weibull Distribution:Threshold Selection 139

6.3.3.3 Box-Cox Transformation:Threshold Selection 139

6.4 Clustering-Based Methods 140

6.4.1 k-Means Clustering 141

6.4.2 Fuzzy c-Means Clustering 142

6.4.3 Self-Organizing Maps (SOMs) 142

6.5 Classification-Based Methods 144

6.5.1 One-Class Classification 145

6.5.1.1 One-Class Support Vector Machines 145

6.5.1.2 k-Nearest Neighbors 148

6.5.2 Multi-Class Classification 149

6.5.2.1 Multi-Class Support Vector Machines 149

6.5.2.2 Neural Networks 151

6.6 StatisticalMethods 153

6.6.1 Sequential Probability Ratio Test 154

6.6.2 Correlation Analysis 156

6.7 Anomaly Detection with No System Health Profile 156

6.8 Challenges in Anomaly Detection 158

References 159

7 Machine Learning: Diagnostics and Prognostics 163
Myeongsu Kang

7.1 Overview of Diagnosis and Prognosis 163

7.2 Techniques for Diagnostics 165

7.2.1 Supervised Machine Learning Algorithms 165

7.2.1.1 Naïve Bayes 165

7.2.1.2 Decision Trees 167

7.2.2 Ensemble Learning 169

7.2.2.1 Bagging 170

7.2.2.2 Boosting: AdaBoost 171

7.2.3 Deep Learning 172

7.2.3.1 Supervised Learning: Deep Residual Networks 173

7.2.3.2 Effect of Feature Learning-Powered Diagnosis 176

7.3 Techniques for Prognostics 178

7.3.1 Regression Analysis 178

7.3.1.1 Linear Regression 178

7.3.1.2 Polynomial Regression 180

7.3.1.3 Ridge Regression 181

7.3.1.4 LASSO Regression 182

7.3.1.5 Elastic Net Regression 183

7.3.1.6 k-Nearest Neighbors Regression 183

7.3.1.7 Support Vector Regression 184

7.3.2 Particle Filtering 185

7.3.2.1 Fundamentals of Particle Filtering 186

7.3.2.2 Resampling Methods – A Review 187

References 189

8 Uncertainty Representation, Quantification, and Management in Prognostics 193
Shankar Sankararaman

8.1 Introduction 193

8.2 Sources of Uncertainty in PHM 196

8.3 Formal Treatment of Uncertainty in PHM 199

8.3.1 Problem 1: Uncertainty Representation and Interpretation 199

8.3.2 Problem 2: Uncertainty Quantification 199

8.3.3 Problem 3: Uncertainty Propagation 200

8.3.4 Problem 4: Uncertainty Management 200

8.4 Uncertainty Representation and Interpretation 200

8.4.1 Physical Probabilities and Testing-Based Prediction 201

8.4.1.1 Physical Probability 201

8.4.1.2 Testing-Based Life Prediction 201

8.4.1.3 Confidence Intervals 202

8.4.2 Subjective Probabilities and Condition-Based Prognostics 202

8.4.2.1 Subjective Probability 202

8.4.2.2 Subjective Probabilities in Condition-Based Prognostics 203

8.4.3 Why is RUL Prediction Uncertain? 203

8.5 Uncertainty Quantification and Propagation for RUL Prediction 203

8.5.1 Computational Framework for Uncertainty Quantification 204

8.5.1.1 Present State Estimation 204

8.5.1.2 Future State Prediction 205

8.5.1.3 RUL Computation 205

8.5.2 RUL Prediction: An Uncertainty Propagation Problem 206

8.5.3 Uncertainty PropagationMethods 206

8.5.3.1 Sampling-Based Methods 207

8.5.3.2 AnalyticalMethods 209

8.5.3.3 Hybrid Methods 209

8.5.3.4 Summary of Methods 209

8.6 Uncertainty Management 210

8.7 Case Study: Uncertainty Quantification in the Power System of an Unmanned Aerial Vehicle 211

8.7.1 Description of the Model 211

8.7.2 Sources of Uncertainty 212

8.7.3 Results: Constant Amplitude Loading Conditions 213

8.7.4 Results: Variable Amplitude Loading Conditions 214

8.7.5 Discussion 214

8.8 Existing Challenges 215

8.8.1 Timely Predictions 215

8.8.2 Uncertainty Characterization 216

8.8.3 Uncertainty Propagation 216

8.8.4 Capturing Distribution Properties 216

8.8.5 Accuracy 216

8.8.6 Uncertainty Bounds 216

8.8.7 Deterministic Calculations 216

8.9 Summary 217

References 217

9 PHM Cost and Return on Investment 221
Peter Sandborn, ChrisWilkinson, Kiri Lee Sharon, Taoufik Jazouli, and Roozbeh Bakhshi

9.1 Return on Investment 221

9.1.1 PHM ROI Analyses 222

9.1.2 Financial Costs 224

9.2 PHM Cost-Modeling Terminology and Definitions 225

9.3 PHM Implementation Costs 226

9.3.1 Nonrecurring Costs 226

9.3.2 Recurring Costs 227

9.3.3 Infrastructure Costs 228

9.3.4 Nonmonetary Considerations and Maintenance Culture 228

9.4 Cost Avoidance 229

9.4.1 Maintenance Planning Cost Avoidance 231

9.4.2 Discrete-Event Simulation Maintenance PlanningModel 232

9.4.3 Fixed-Schedule Maintenance Interval 233

9.4.4 Data-Driven (Precursor to Failure Monitoring) Methods 233

9.4.5 Model-Based (LRU-Independent)Methods 234

9.4.6 Discrete-Event Simulation Implementation Details 236

9.4.7 Operational Profile 237

9.5 Example PHM Cost Analysis 238

9.5.1 Single-Socket Model Results 239

9.5.2 Multiple-Socket Model Results 241

9.6 Example Business Case Construction: Analysis for ROI 246

9.7 Summary 255

References 255

10 Valuation and Optimization of PHM-Enabled Maintenance Decisions 261
Xin Lei, Amir Reza Kashani-Pour, Peter Sandborn, and Taoufik Jazouli

10.1 Valuation and Optimization of PHM-Enabled Maintenance Decisions for an Individual System 262

10.1.1 A PHM-Enabled Predictive Maintenance OptimizationModel for an Individual System 263

10.1.2 Case Study: Optimization of PHM-Enabled Maintenance Decisions for an Individual System (Wind Turbine) 265

10.2 Availability 268

10.2.1 The Business of Availability: Outcome-Based Contracts 269

10.2.2 Incorporating Contract Terms into Maintenance Decisions 270

10.2.3 Case Study: Optimization of PHM-Enabled Maintenance Decisions for Systems (Wind Farm) 270

10.3 Future Directions 272

10.3.1 Design for Availability 272

10.3.2 Prognostics-BasedWarranties 275

10.3.3 Contract Engineering 276

References 277

11 Health and Remaining Useful Life Estimation of Electronic Circuits 279
Arvind Sai Sarathi Vasan and Michael G. Pecht

11.1 Introduction 279

11.2 RelatedWork 281

11.2.1 Component-Centric Approach 281

11.2.2 Circuit-Centric Approach 282

11.3 Electronic Circuit Health Estimation Through Kernel Learning 285

11.3.1 Kernel-Based Learning 285

11.3.2 Health Estimation Method 286

11.3.2.1 Likelihood-Based Function for Model Selection 288

11.3.2.2 Optimization Approach for Model Selection 289

11.3.3 Implementation Results 292

11.3.3.1 Bandpass Filter Circuit 293

11.3.3.2 DC–DC Buck Converter System 300

11.4 RUL Prediction Using Model-Based Filtering 306

11.4.1 Prognostics Problem Formulation 306

11.4.2 Circuit DegradationModeling 307

11.4.3 Model-Based Prognostic Methodology 310

11.4.4 Implementation Results 313

11.4.4.1 Low-Pass Filter Circuit 313

11.4.4.2 Voltage Feedback Circuit 315

11.4.4.3 Source of RUL Prediction Error 320

11.4.4.4 Effect of First-Principles-Based Modeling 320

11.5 Summary 322

References 324

12 PHM-Based Qualification of Electronics 329
Preeti S. Chauhan

12.1 Why is Product Qualification Important? 329

12.2 Considerations for Product Qualification 331

12.3 Review of Current Qualification Methodologies 334

12.3.1 Standards-Based Qualification 334

12.3.2 Knowledge-Based or PoF-Based Qualification 337

12.3.3 Prognostics and Health Management-Based Qualification 340

12.3.3.1 Data-Driven Techniques 340

12.3.3.2 Fusion Prognostics 343

12.4 Summary 345

References 346

13 PHM of Li-ion Batteries 349
Saurabh Saxena, Yinjiao Xing, andMichael G. Pecht

13.1 Introduction 349

13.2 State of Charge Estimation 351

13.2.1 SOC Estimation Case Study I 352

13.2.1.1 NN Model 353

13.2.1.2 Training and Testing Data 354

13.2.1.3 Determination of the NN Structure 355

13.2.1.4 Training and Testing Results 356

13.2.1.5 Application of Unscented Kalman Filter 357

13.2.2 SOC Estimation Case Study II 357

13.2.2.1 OCV–SOC-T Test 358

13.2.2.2 Battery Modeling and Parameter Identification 359

13.2.2.3 OCV–SOC-T Table for Model Improvement 360

13.2.2.4 Validation of the Proposed Model 362

13.2.2.5 Algorithm Implementation for Online Estimation 362

13.3 State of Health Estimation and Prognostics 365

13.3.1 Case Study for Li-ion Battery Prognostics 366

13.3.1.1 Capacity DegradationModel 366

13.3.1.2 Uncertainties in Battery Prognostics 368

13.3.1.3 Model Updating via Bayesian Monte Carlo 368

13.3.1.4 SOH Prognostics and RUL Estimation 369

13.3.1.5 Prognostic Results 371

13.4 Summary 371

References 372

14 PHM of Light-Emitting Diodes 377
Moon-Hwan Chang, Jiajie Fan, Cheng Qian, and Bo Sun

14.1 Introduction 377

14.2 Review of PHM Methodologies for LEDs 378

14.2.1 Overview of Available Prognostic Methods 378

14.2.2 Data-DrivenMethods 379

14.2.2.1 Statistical Regression 379

14.2.2.2 Static Bayesian Network 381

14.2.2.3 Kalman Filtering 382

14.2.2.4 Particle Filtering 383

14.2.2.5 Artificial Neural Network 384

14.2.3 Physics-Based Methods 385

14.2.4 LED System-Level Prognostics 387

14.3 Simulation-Based Modeling and Failure Analysis for LEDs 388

14.3.1 LED Chip-LevelModeling and Failure Analysis 389

14.3.1.1 Electro-optical Simulation of LED Chip 389

14.3.1.2 LED Chip-Level Failure Analysis 393

14.3.2 LED Package-Level Modeling and Failure Analysis 395

14.3.2.1 Thermal and Optical Simulation for Phosphor-Converted White LED Package 395

14.3.2.2 LED Package-Level Failure Analysis 397

14.3.3 LED System-LevelModeling and Failure Analysis 399

14.4 Return-on-Investment Analysis of Applying Health Monitoring to LED Lighting Systems 401

14.4.1 ROI Methodology 403

14.4.2 ROI Analysis of Applying System Health Monitoring to LED Lighting Systems 406

14.4.2.1 Failure Rates and Distributions for ROI Simulation 407

14.4.2.2 Determination of Prognostics Distance 410

14.4.2.3 IPHM, CPHM, and Cu Evaluation 412

14.4.2.4 ROI Evaluation 417

14.5 Summary 419

References 420

15 PHM in Healthcare 431
Mary Capelli-Schellpfeffer,Myeongsu Kang, andMichael G. Pecht

15.1 Healthcare in the United States 431

15.2 Considerations in Healthcare 432

15.2.1 Clinical Consideration in ImplantableMedical Devices 432

15.2.2 Considerations in Care Bots 433

15.3 Benefits of PHM 438

15.3.1 Safety Increase 439

15.3.2 Operational Reliability Improvement 440

15.3.3 Mission Availability Increase 440

15.3.4 System’s Service Life Extension 441

15.3.5 Maintenance Effectiveness Increase 441

15.4 PHM of ImplantableMedical Devices 442

15.5 PHM of Care Bots 444

15.6 Canary-Based Prognostics of Healthcare Devices 445

15.7 Summary 447

References 447

16 PHM of Subsea Cables 451
David Flynn, Christopher Bailey, Pushpa Rajaguru,Wenshuo Tang, and Chunyan Yin

16.1 Subsea Cable Market 451

16.2 Subsea Cables 452

16.3 Cable Failures 454

16.3.1 Internal Failures 455

16.3.2 Early-Stage Failures 455

16.3.3 External Failures 455

16.3.4 Environmental Conditions 455

16.3.5 Third-Party Damage 456

16.4 State-of-the-Art Monitoring 457

16.5 Qualifying and Maintaining Subsea Cables 458

16.5.1 Qualifying Subsea Cables 458

16.5.2 Mechanical Tests 458

16.5.3 Maintaining Subsea Cables 459

16.6 Data-Gathering Techniques 460

16.7 Measuring theWear Behavior of Cable Materials 461

16.8 Predicting Cable Movement 463

16.8.1 Sliding Distance Derivation 463

16.8.2 Scouring Depth Calculations 465

16.9 Predicting Cable Degradation 466

16.9.1 Volume Loss due to Abrasion 466

16.9.2 Volume Loss due to Corrosion 466

16.10 Predicting Remaining Useful Life 468

16.11 Case Study 471

16.12 Future Challenges 471

16.12.1 Data-Driven Approach for Random Failures 471

16.12.2 Model-Driven Approach for Environmental Failures 473

16.12.2.1 Fusion-Based PHM 473

16.12.2.2 Sensing Techniques 474

16.13 Summary 474

References 475

17 Connected Vehicle Diagnostics and Prognostics 479
Yilu Zhang and Xinyu Du

17.1 Introduction 479

17.2 Design of an Automatic Field Data Analyzer 481

17.2.1 Data Collection Subsystem 482

17.2.2 Information Abstraction Subsystem 482

17.2.3 Root Cause Analysis Subsystem 482

17.2.3.1 Feature-Ranking Module 482

17.2.3.2 Relevant Feature Set Selection 484

17.2.3.3 Results Interpretation 486

17.3 Case Study: CVDP for Vehicle Batteries 486

17.3.1 Brief Background of Vehicle Batteries 486

17.3.2 Applying AFDA for Vehicle Batteries 488

17.3.3 Experimental Results 489

Contents xvii

17.3.3.1 Information Abstraction 490

17.3.3.2 Feature Ranking 490

17.3.3.3 Interpretation of Results 495

17.4 Summary 498

References 499

18 The Role of PHM at Commercial Airlines 503
RhondaWalthall and Ravi Rajamani

18.1 Evolution of Aviation Maintenance 503

18.2 Stakeholder Expectations for PHM 506

18.2.1 Passenger Expectations 506

18.2.2 Airline/Operator/Owner Expectations 507

18.2.3 Airframe Manufacturer Expectations 509

18.2.4 Engine Manufacturer Expectations 510

18.2.5 System and Component Supplier Expectations 511

18.2.6 MRO Organization Expectations 512

18.3 PHM Implementation 513

18.3.1 SATAA 513

18.4 PHM Applications 517

18.4.1 Engine Health Management (EHM) 517

18.4.1.1 History of EHM 518

18.4.1.2 EHM Infrastructure 519

18.4.1.3 Technologies Associated with EHM 520

18.4.1.4 The Future 523

18.4.2 Auxiliary Power Unit (APU) Health Management 524

18.4.3 Environmental Control System (ECS) and Air Distribution Health Monitoring 525

18.4.4 Landing System Health Monitoring 526

18.4.5 Liquid Cooling System Health Monitoring 526

18.4.6 Nitrogen Generation System (NGS) Health Monitoring 527

18.4.7 Fuel Consumption Monitoring 527

18.4.8 Flight Control Actuation Health Monitoring 528

18.4.9 Electric Power System Health Monitoring 529

18.4.10 Structural Health Monitoring (SHM) 529

18.4.11 Battery Health Management 531

18.5 Summary 532

References 533

19 PHM Software for Electronics 535
Noel Jordan Jameson,Myeongsu Kang, and Jing Tian

19.1 PHM Software: CALCE Simulation Assisted Reliability Assessment 535

19.2 PHM Software: Data-Driven 540

19.2.1 Data Flow 541

19.2.2 Master Options 542

19.2.3 Data Pre-processing 543

19.2.4 Feature Discovery 545

19.2.5 Anomaly Detection 546

19.2.6 Diagnostics/Classification 548

19.2.7 Prognostics/Modeling 552

19.2.8 Challenges in Data-Driven PHM Software Development 554

19.3 Summary 557

20 eMaintenance 559
Ramin Karim, Phillip Tretten, and Uday Kumar

20.1 From Reactive to Proactive Maintenance 559

20.2 The Onset of eMaintenance 560

20.3 MaintenanceManagement System 561

20.3.1 Life-cycle Management 562

20.3.2 eMaintenance Architecture 564

20.4 Sensor Systems 564

20.4.1 Sensor Technology for PHM 565

20.5 Data Analysis 565

20.6 Predictive Maintenance 566

20.7 Maintenance Analytics 567

20.7.1 Maintenance Descriptive Analytics 568

20.7.2 Maintenance Analytics and eMaintenance 568

20.7.3 Maintenance Analytics and Big Data 568

20.8 Knowledge Discovery 570

20.9 Integrated Knowledge Discovery 571

20.10 User Interface for Decision Support 572

20.11 Applications of eMaintenance 572

20.11.1 eMaintenance in Railways 572

20.11.1.1 Railway Cloud: Swedish Railway Data 573

20.11.1.2 Railway Cloud: Service Architecture 573

20.11.1.3 Railway Cloud: Usage Scenario 574

20.11.2 eMaintenance in Manufacturing 574

20.11.3 MEMS Sensors for Bearing Vibration Measurement 576

20.11.4 Wireless Sensors for Temperature Measurement 576

20.11.5 Monitoring Systems 576

20.11.6 eMaintenance Cloud and Servers 578

20.11.7 Dashboard Managers 580

20.11.8 Alarm Servers 580

20.11.9 Cloud Services 581

20.11.10 Graphic User Interfaces 583

20.12 Internet Technology and Optimizing Technology 585

References 586

21 Predictive Maintenance in the IoT Era 589
Rashmi B. Shetty

21.1 Background 589

21.1.1 Challenges of a Maintenance Program 590

21.1.2 Evolution of Maintenance Paradigms 590

21.1.3 Preventive Versus Predictive Maintenance 592

21.1.4 P–F Curve 592

21.1.5 Bathtub Curve 594

21.2 Benefits of a Predictive Maintenance Program 595

21.3 Prognostic Model Selection for Predictive Maintenance 596

21.4 Internet ofThings 598

21.4.1 Industrial IoT 598

21.5 Predictive Maintenance Based on IoT 599

21.6 Predictive Maintenance Usage Cases 600

21.7 Machine Learning Techniques for Data-Driven Predictive Maintenance 600

21.7.1 Supervised Learning 602

21.7.2 Unsupervised Learning 602

21.7.3 Anomaly Detection 602

21.7.4 Multi-class and Binary Classification Models 603

21.7.5 Regression Models 604

21.7.6 Survival Models 604

21.8 Best Practices 604

21.8.1 Define Business Problem and QuantitativeMetrics 605

21.8.2 Identify Assets and Data Sources 605

21.8.3 Data Acquisition and Transformation 606

21.8.4 Build Models 607

21.8.5 Model Selection 607

21.8.6 Predict Outcomes and Transform into Process Insights 608

21.8.7 Operationalize and Deploy 609

21.8.8 Continuous Monitoring 609

21.9 Challenges in a Successful Predictive Maintenance Program 610

21.9.1 Predictive Maintenance Management Success Key Performance Indicators (KPIs) 610

21.10 Summary 611

References 611

22 Analysis of PHM Patents for Electronics 613
Zhenbao Liu, Zhen Jia, Chi-Man Vong, Shuhui Bu, andMichael G. Pecht

22.1 Introduction 613

22.2 Analysis of PHM Patents for Electronics 616

22.2.1 Sources of PHM Patents 616

22.2.2 Analysis of PHM Patents 617

22.3 Trend of Electronics PHM 619

22.3.1 Semiconductor Products and Computers 619

22.3.2 Batteries 622

22.3.3 Electric Motors 626

22.3.4 Circuits and Systems 629

22.3.5 Electrical Devices in Automobiles and Airplanes 631

22.3.6 Networks and Communication Facilities 634

22.3.7 Others 636

22.4 Summary 638

References 639

23 A PHM Roadmap for Electronics-Rich Systems 64
Michael G. Pecht

23.1 Introduction 649

23.2 Roadmap Classifications 650

23.2.1 PHM at the Component Level 651

23.2.1.1 PHM for Integrated Circuits 652

23.2.1.2 High-Power Switching Electronics 652

23.2.1.3 Built-In Prognostics for Components and Circuit Boards 653

23.2.1.4 Photo-Electronics Prognostics 654

23.2.1.5 Interconnect andWiring Prognostics 656

23.2.2 PHM at the System Level 657

23.2.2.1 Legacy Systems 657

23.2.2.2 Environmental and OperationalMonitoring 659

23.2.2.3 LRU to Device Level 659

23.2.2.4 Dynamic Reconfiguration 659

23.2.2.5 System Power Management and PHM 660

23.2.2.6 PHM as Knowledge Infrastructure for System Development 660

23.2.2.7 Prognostics for Software 660

23.2.2.8 PHM for Mitigation of Reliability and Safety Risks 661

23.2.2.9 PHM in Supply Chain Management and Product Maintenance 662

23.3 Methodology Development 663

23.3.1 Best Algorithms 664

23.3.1.1 Approaches to Training 667

23.3.1.2 Active Learning for Unlabeled Data 667

23.3.1.3 Sampling Techniques and Cost-Sensitive Learning for Imbalanced Data 668

23.3.1.4 Transfer Learning for Knowledge Transfer 668

23.3.1.5 Internet ofThings and Big Data Analytics 669

23.3.2 Verification and Validation 670

23.3.3 Long-Term PHM Studies 671

23.3.4 PHM for Storage 671

23.3.5 PHM for No-Fault-Found/Intermittent Failures 672

23.3.6 PHM for Products Subjected to Indeterminate Operating Conditions 673

23.4 Nontechnical Barriers 674

23.4.1 Cost, Return on Investment, and Business Case Development 674

23.4.2 Liability and Litigation 676

23.4.2.1 Code Architecture: Proprietary or Open? 676

23.4.2.2 Long-Term Code Maintenance and Upgrades 676

23.4.2.3 False Alarms, Missed Alarms, and Life-Safety Implications 677

23.4.2.4 Warranty Restructuring 677

23.4.3 Maintenance Culture 677

23.4.4 Contract Structure 677

23.4.5 Role of Standards Organizations 678

23.4.5.1 IEEE Reliability Society and PHM Efforts 678

23.4.5.2 SAE PHM Standards 678

23.4.5.3 PHM Society 679

23.4.6 Licensing and Entitlement Management 680

References 680

Appendix A Commercially Available Sensor Systems for PHM 691

A.1 SmartButton – ACR Systems 691

A.2 OWL 400 – ACR Systems 693

A.3 SAVERTM 3X90 – Lansmont Instruments 695

A.4 G-Link^®-LXRS^®– LORD MicroStrain®Sensing Systems 697

A.5 V-Link^®-LXRS^®– LORD MicroStrain Sensing Systems 699

A.6 3DM-GX4–25TM – LORD MicroStrain Sensing Systems 702

A.7 IEPE-LinkTM-LXRS^®– LORD MicroStrain Sensing Systems 704

A.8 ICHM^®20/20 – Oceana Sensor 706

A.9 EnvironmentalMonitoring System 200TM – Upsite Technologies 708

A.10 S2NAP^®– RLWInc. 710

A.11 SR1 Strain Gage Indicator – Advance Instrument Inc. 712

A.12 P3 Strain Indicator and Recorder – Micro-Measurements 714

A.13 Airscale Suspension-BasedWeighing System – VPG Inc. 716

A.14 Radio Microlog – Transmission Dynamics 718

Appendix B Journals and Conference Proceedings Related to PHM 721

B.1 Journals 721

B.2 Conference Proceedings 722

Appendix C Glossary of Terms and Definitions 725

Index 731