+353-1-416-8900REST OF WORLD
+44-20-3973-8888REST OF WORLD
1-917-300-0470EAST COAST U.S
1-800-526-8630U.S. (TOLL FREE)

Condition Monitoring with Vibration Signals. Compressive Sampling and Learning Algorithms for Rotating Machines. Edition No. 1. IEEE Press

  • Book

  • 440 Pages
  • January 2020
  • John Wiley and Sons Ltd
  • ID: 5826104

Provides an extensive, up-to-date treatment of techniques used for machine condition monitoring

Clear and concise throughout, this accessible book is the first to be wholly devoted to the field of condition monitoring for rotating machines using vibration signals. It covers various feature extraction, feature selection, and classification methods as well as their applications to machine vibration datasets. It also presents new methods including machine learning and compressive sampling, which help to improve safety, reliability, and performance. 

Condition Monitoring with Vibration Signals: Compressive Sampling and Learning Algorithms for Rotating Machines starts by introducing readers to Vibration Analysis Techniques and Machine Condition Monitoring (MCM). It then offers readers sections covering: Rotating Machine Condition Monitoring using Learning Algorithms; Classification Algorithms; and New Fault Diagnosis Frameworks designed for MCM. Readers will learn signal processing in the time-frequency domain, methods for linear subspace learning, and the basic principles of the learning method Artificial Neural Network (ANN). They will also discover recent trends of deep learning in the field of machine condition monitoring, new feature learning frameworks based on compressive sampling, subspace learning techniques for machine condition monitoring, and much more.

  • Covers the fundamental as well as the state-of-the-art approaches to machine condition monitoring�guiding readers from the basics of rotating machines to the generation of knowledge using vibration signals
  • Provides new methods, including machine learning and compressive sampling, which offer significant improvements in accuracy with reduced computational costs
  • Features learning algorithms that can be used for fault diagnosis and prognosis
  • Includes previously and recently developed dimensionality reduction techniques and classification algorithms

Condition Monitoring with Vibration Signals: Compressive Sampling and Learning Algorithms for Rotating Machines is an excellent book for research students, postgraduate students, industrial practitioners, and researchers.

Table of Contents

Preface xvii

About the Authors xxi

List of Abbreviations xxiii

Part I Introduction 1

1 Introduction to Machine Condition Monitoring 3

1.1 Background 3

1.2 Maintenance Approaches for Rotating Machines Failures 4

1.2.1 Corrective Maintenance 4

1.2.2 Preventive Maintenance 5

1.2.2.1 Time-Based Maintenance (TBM) 5

1.2.2.2 Condition-Based Maintenance (CBM) 5

1.3 Applications of MCM 5

1.3.1 Wind Turbines 5

1.3.2 Oil and Gas 6

1.3.3 Aerospace and Defence Industry 6

1.3.4 Automotive 7

1.3.5 Marine Engines 7

1.3.6 Locomotives 7

1.4 Condition Monitoring Techniques 7

1.4.1 Vibration Monitoring 7

1.4.2 Acoustic Emission 8

1.4.3 Fusion of Vibration and Acoustic 8

1.4.4 Motor Current Monitoring 8

1.4.5 Oil Analysis and Lubrication Monitoring 8

1.4.6 Thermography 9

1.4.7 Visual Inspection 9

1.4.8 Performance Monitoring 9

1.4.9 Trend Monitoring 10

1.5 Topic Overview and Scope of the Book 10

1.6 Summary 11

References 11

2 Principles of Rotating Machine Vibration Signals 17

2.1 Introduction 17

2.2 Machine Vibration Principles 17

2.3 Sources of Rotating Machines Vibration Signals 20

2.3.1 Rotor Mass Unbalance 21

2.3.2 Misalignment 21

2.3.3 Cracked Shafts 21

2.3.4 Rolling Element Bearings 23

2.3.5 Gears 25

2.4 Types of Vibration Signals 25

2.4.1 Stationary 26

2.4.2 Nonstationary 26

2.5 Vibration Signal Acquisition 26

2.5.1 Displacement Transducers 26

2.5.2 Velocity Transducers 26

2.5.3 Accelerometers 27

2.6 Advantages and Limitations of Vibration Signal Monitoring 27

2.7 Summary 28

References 28

Part II Vibration Signal Analysis Techniques 31

3 Time Domain Analysis 33

3.1 Introduction 33

3.1.1 Visual Inspection 33

3.1.2 Features-Based Inspection 35

3.2 Statistical Functions 35

3.2.1 Peak Amplitude 36

3.2.2 Mean Amplitude 36

3.2.3 Root Mean Square Amplitude 36

3.2.4 Peak-to-Peak Amplitude 36

3.2.5 Crest Factor (CF) 36

3.2.6 Variance and Standard Deviation 37

3.2.7 Standard Error 37

3.2.8 Zero Crossing 38

3.2.9 Wavelength 39

3.2.10 Willison Amplitude 39

3.2.11 Slope Sign Change 39

3.2.12 Impulse Factor 39

3.2.13 Margin Factor 40

3.2.14 Shape Factor 40

3.2.15 Clearance Factor 40

3.2.16 Skewness 40

3.2.17 Kurtosis 40

3.2.18 Higher-Order Cumulants (HOCs) 41

3.2.19 Histograms 42

3.2.20 Normal/Weibull Negative Log-Likelihood Value 42

3.2.21 Entropy 42

3.3 Time Synchronous Averaging 44

3.3.1 TSA Signals 44

3.3.2 Residual Signal (RES) 44

3.3.2.1 NA4 44

3.3.2.2 NA4* 45

3.3.3 Difference Signal (DIFS) 45

3.3.3.1 FM4 46

3.3.3.2 M6A 46

3.3.3.3 M8A 46

3.4 Time Series Regressive Models 46

3.4.1 AR Model 47

3.4.2 MA Model 48

3.4.3 ARMA Model 48

3.4.4 ARIMA Model 48

3.5 Filter-Based Methods 49

3.5.1 Demodulation 49

3.5.2 Prony Model 52

3.5.3 Adaptive Noise Cancellation (ANC) 53

3.6 Stochastic Parameter Techniques 54

3.7 Blind Source Separation (BSS) 54

3.8 Summary 55

References 56

4 Frequency Domain Analysis 63

4.1 Introduction 63

4.2 Fourier Analysis 64

4.2.1 Fourier Series 64

4.2.2 Discrete Fourier Transform 66

4.2.3 Fast Fourier Transform (FFT) 67

4.3 Envelope Analysis 71

4.4 Frequency Spectrum Statistical Features 73

4.4.1 Arithmetic Mean 73

4.4.2 Geometric Mean 73

4.4.3 Matched Filter RMS 73

4.4.4 The RMS of Spectral Difference 74

4.4.5 The Sum of Squares Spectral Difference 74

4.4.6 High-Order Spectra Techniques 74

4.5 Summary 75

References 76

5 Time-Frequency Domain Analysis 79

5.1 Introduction 79

5.2 Short-Time Fourier Transform (STFT) 79

5.3 Wavelet Analysis 82

5.3.1 Wavelet Transform (WT) 82

5.3.1.1 Continuous Wavelet Transform (CWT) 83

5.3.1.2 Discrete Wavelet Transform (DWT) 85

5.3.2 Wavelet Packet Transform (WPT) 89

5.4 Empirical Mode Decomposition (EMD) 91

5.5 Hilbert-Huang Transform (HHT) 94

5.6 Wigner-Ville Distribution 96

5.7 Local Mean Decomposition (LMD) 98

5.8 Kurtosis and Kurtograms 100

5.9 Summary 105

References 106

Part III Rotating Machine Condition Monitoring Using Machine Learning 115

6 Vibration-Based Condition Monitoring Using Machine Learning 117

6.1 Introduction 117

6.2 Overview of the Vibration-Based MCM Process 118

6.2.1 Fault-Detection and -Diagnosis Problem Framework 118

6.3 Learning from Vibration Data 122

6.3.1 Types of Learning 123

6.3.1.1 Batch vs. Online Learning 123

6.3.1.2 Instance-Based vs. Model-Based Learning 123

6.3.1.3 Supervised Learning vs. Unsupervised Learning 123

6.3.1.4 Semi-Supervised Learning 123

6.3.1.5 Reinforcement Learning 124

6.3.1.6 Transfer Learning 124

6.3.2 Main Challenges of Learning from Vibration Data 125

6.3.2.1 The Curse of Dimensionality 125

6.3.2.2 Irrelevant Features 126

6.3.2.3 Environment and Operating Conditions of a Rotating Machine 126

6.3.3 Preparing Vibration Data for Analysis 126

6.3.3.1 Normalisation 126

6.3.3.2 Dimensionality Reduction 127

6.4 Summary 128

References 128

7 Linear Subspace Learning 131

7.1 Introduction 131

7.2 Principal Component Analysis (PCA) 132

7.2.1 PCA Using Eigenvector Decomposition 132

7.2.2 PCA Using SVD 133

7.2.3 Application of PCA in Machine Fault Diagnosis 134

7.3 Independent Component Analysis (ICA) 137

7.3.1 Minimisation of Mutual Information 138

7.3.2 Maximisation of the Likelihood 138

7.3.3 Application of ICA in Machine Fault Diagnosis 139

7.4 Linear Discriminant Analysis (LDA) 141

7.4.1 Application of LDA in Machine Fault Diagnosis 142

7.5 Canonical Correlation Analysis (CCA) 143

7.6 Partial Least Squares (PLS) 145

7.7 Summary 146

References 147

8 Nonlinear Subspace Learning 153

8.1 Introduction 153

8.2 Kernel Principal Component Analysis (KPCA) 153

8.2.1 Application of KPCA in Machine Fault Diagnosis 156

8.3 Isometric Feature Mapping (ISOMAP) 156

8.3.1 Application of ISOMAP in Machine Fault Diagnosis 158

8.4 Diffusion Maps (DMs) and Diffusion Distances 159

8.4.1 Application of DMs in Machine Fault Diagnosis 160

8.5 Laplacian Eigenmap (LE) 161

8.5.1 Application of the LE in Machine Fault Diagnosis 161

8.6 Local Linear Embedding (LLE) 162

8.6.1 Application of LLE in Machine Fault Diagnosis 163

8.7 Hessian-Based LLE 163

8.7.1 Application of HLLE in Machine Fault Diagnosis 164

8.8 Local Tangent Space Alignment Analysis (LTSA) 165

8.8.1 Application of LTSA in Machine Fault Diagnosis 165

8.9 Maximum Variance Unfolding (MVU) 166

8.9.1 Application of MVU in Machine Fault Diagnosis 167

8.10 Stochastic Proximity Embedding (SPE) 168

8.10.1 Application of SPE in Machine Fault Diagnosis 168

8.11 Summary 169

References 170

9 Feature Selection 173

9.1 Introduction 173

9.2 Filter Model-Based Feature Selection 175

9.2.1 Fisher Score (FS) 176

9.2.2 Laplacian Score (LS) 177

9.2.3 Relief and Relief-F Algorithms 178

9.2.3.1 Relief Algorithm 178

9.2.3.2 Relief-F Algorithm 179

9.2.4 Pearson Correlation Coefficient (PCC) 180

9.2.5 Information Gain (IG) and Gain Ratio (GR) 180

9.2.6 Mutual Information (MI) 181

9.2.7 Chi-Squared (Chi-2) 181

9.2.8 Wilcoxon Ranking 181

9.2.9 Application of Feature Ranking in Machine Fault Diagnosis 182

9.3 Wrapper Model-Based Feature Subset Selection 185

9.3.1 Sequential Selection Algorithms 185

9.3.2 Heuristic-Based Selection Algorithms 185

9.3.2.1 Ant Colony Optimisation (ACO) 185

9.3.2.2 Genetic Algorithms (GAs) and Genetic Programming 187

9.3.2.3 Particle Swarm Optimisation (PSO) 188

9.3.3 Application of Wrapper Model-Based Feature Subset Selection in Machine Fault Diagnosis 189

9.4 Embedded Model-Based Feature Selection 192

9.5 Summary 193

References 194

Part IV Classification Algorithms 199

10 Decision Trees and Random Forests 201

10.1 Introduction 201

10.2 Decision Trees 202

10.2.1 Univariate Splitting Criteria 204

10.2.1.1 Gini Index 205

10.2.1.2 Information Gain 206

10.2.1.3 Distance Measure 207

10.2.1.4 Orthogonal Criterion (ORT) 207

10.2.2 Multivariate Splitting Criteria 207

10.2.3 Tree-Pruning Methods 208

10.2.3.1 Error-Complexity Pruning 208

10.2.3.2 Minimum-Error Pruning 209

10.2.3.3 Reduced-Error Pruning 209

10.2.3.4 Critical-Value Pruning 210

10.2.3.5 Pessimistic Pruning 210

10.2.3.6 Minimum Description Length (MDL) Pruning 210

10.2.4 Decision Tree Inducers 211

10.2.4.1 CART 211

10.2.4.2 ID3 211

10.2.4.3 C4.5 211

10.2.4.4 CHAID 212

10.3 Decision Forests 212

10.4 Application of Decision Trees/Forests in Machine Fault Diagnosis 213

10.5 Summary 217

References 217

11 Probabilistic Classification Methods 225

11.1 Introduction 225

11.2 Hidden Markov Model 225

11.2.1 Application of Hidden Markov Models in Machine Fault Diagnosis 228

11.3 Logistic Regression Model 230

11.3.1 Logistic Regression Regularisation 232

11.3.2 Multinomial Logistic Regression Model (MLR) 232

11.3.3 Application of Logistic Regression in Machine Fault Diagnosis 233

11.4 Summary 234

References 235

12 Artificial Neural Networks (ANNs) 239

12.1 Introduction 239

12.2 Neural Network Basic Principles 240

12.2.1 The Multilayer Perceptron 241

12.2.2 The Radial Basis Function Network 243

12.2.3 The Kohonen Network 244

12.3 Application of Artificial Neural Networks in Machine Fault Diagnosis 245

12.4 Summary 253

References 254

13 Support Vector Machines (SVMs) 259

13.1 Introduction 259

13.2 Multiclass SVMs 262

13.3 Selection of Kernel Parameters 263

13.4 Application of SVMs in Machine Fault Diagnosis 263

13.5 Summary 274

References 274

14 Deep Learning 279

14.1 Introduction 279

14.2 Autoencoders 280

14.3 Convolutional Neural Networks (CNNs) 283

14.4 Deep Belief Networks (DBNs) 284

14.5 Recurrent Neural Networks (RNNs) 285

14.6 Overview of Deep Learning in MCM 286

14.6.1 Application of AE-based DNNs in Machine Fault Diagnosis 286

14.6.2 Application of CNNs in Machine Fault Diagnosis 292

14.6.3 Application of DBNs in Machine Fault Diagnosis 296

14.6.4 Application of RNNs in Machine Fault Diagnosis 298

14.7 Summary 299

References 301

15 Classification Algorithm Validation 307

15.1 Introduction 307

15.2 The Hold-Out Technique 308

15.2.1 Three-Way Data Split 309

15.3 Random Subsampling 309

15.4 K-Fold Cross-Validation 310

15.5 Leave-One-Out Cross-Validation 311

15.6 Bootstrapping 311

15.7 Overall Classification Accuracy 312

15.8 Confusion Matrix 313

15.9 Recall and Precision 314

15.10 ROC Graphs 315

15.11 Summary 317

References 318

Part V New Fault Diagnosis Frameworks Designed for MCM 321

16 Compressive Sampling and Subspace Learning (CS-SL) 323

16.1 Introduction 323

16.2 Compressive Sampling for Vibration-Based MCM 325

16.2.1 Compressive Sampling Basics 325

16.2.2 CS for Sparse Frequency Representation 328

16.2.3 CS for Sparse Time-Frequency Representation 329

16.3 Overview of CS in Machine Condition Monitoring 330

16.3.1 Compressed Sensed Data Followed by Complete Data Construction 330

16.3.2 Compressed Sensed Data Followed by Incomplete Data Construction 331

16.3.3 Compressed Sensed Data as the Input of a Classifier 332

16.3.4 Compressed Sensed Data Followed by Feature Learning 333

16.4 Compressive Sampling and Feature Ranking (CS-FR) 333

16.4.1 Implementations 334

16.4.1.1 CS-LS 336

16.4.1.2 CS-FS 336

16.4.1.3 CS-Relief-F 337

16.4.1.4 CS-PCC 338

16.4.1.5 CS-Chi-2 338

16.5 CS and Linear Subspace Learning-Based Framework for Fault Diagnosis 339

16.5.1 Implementations 339

16.5.1.1 CS-PCA 339

16.5.1.2 CS-LDA 340

16.5.1.3 CS-CPDC 341

16.6 CS and Nonlinear Subspace Learning-Based Framework for Fault Diagnosis 343

16.6.1 Implementations 344

16.6.1.1 CS-KPCA 344

16.6.1.2 CS-KLDA 345

16.6.1.3 CS-CMDS 346

16.6.1.4 CS-SPE 346

16.7 Applications 348

16.7.1 Case Study 1 348

16.7.1.1 The Combination of MMV-CS and Several Feature-Ranking Techniques 350

16.7.1.2 The Combination of MMV-CS and Several Linear and Nonlinear Subspace Learning Techniques 352

16.7.2 Case Study 2 354

16.7.2.1 The Combination of MMV-CS and Several Feature-Ranking Techniques 354

16.7.2.2 The Combination of MMV-CS and Several Linear and Nonlinear Subspace Learning Techniques 355

16.8 Discussion 355

References 357

17 Compressive Sampling and Deep Neural Network (CS-DNN) 361

17.1 Introduction 361

17.2 Related Work 361

17.3 CS-SAE-DNN 362

17.3.1 Compressed Measurements Generation 362

17.3.2 CS Model Testing Using the Flip Test 363

17.3.3 DNN-Based Unsupervised Sparse Overcomplete Feature Learning 363

17.3.4 Supervised Fine Tuning 367

17.4 Applications 367

17.4.1 Case Study 1 367

17.4.2 Case Study 2 372

17.5 Discussion 375

References 375

18 Conclusion 379

18.1 Introduction 379

18.2 Summary and Conclusion 380

Appendix Machinery Vibration Data Resources and Analysis Algorithms 389

References 394

Index 395

Authors

Hosameldin Ahmed Asoke K. Nandi Brunel University London, UK.