Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives

Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives

An insightful treatment of present and emerging technologies in fault diagnosis and failure prognosis

In Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives, a team of distinguished researchers delivers a comprehensive exploration of current and emerging approaches to fault diagnosis and failure prognosis of electrical machines and drives. The authors begin with foundational background, describing the physics of failure, the motor and drive designs and components that affect failure and signals, signal processing, and analysis.

The book then moves on to describe the features of these signals and the methods commonly used to extract these features to diagnose the health of a motor or drive, as well as the methods used to identify the state of health and differentiate between possible faults or their severity.

Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives discusses the tools used to recognize trends towards failure and the estimation of remaining useful life. It addresses the relationships between fault diagnosis, failure prognosis, and fault mitigation.

The book also provides:

• A thorough introduction to the modes of failure, how early failure precursors manifest themselves in signals, and how features extracted from these signals are processed

• A comprehensive exploration of the fault diagnosis, the results of characterization, and how they used to predict the time of failure and the confidence interval associated with it

• A focus on medium-sized drives, including induction, permanent magnet AC, reluctance, and new machine and drive types

Perfect for researchers and students who wish to study or practice in the rea of electrical machines and drives, Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives is also an indispensable resource for researchers with a background in signal processing or statistics.

Auteur

Elias G. Strangas, PhD, is a Professor at Michigan State University, where he heads the Electrical Machines and Drives Laboratory.

Guy Clerc is a Professor at the Université de Lyon, Ampère, in Villeurbanne, France. Hubert Razik is a Professor at the Université de Lyon, Ampère, in Villeurbanne, France. Abdenour Soualhi is an Assistant Professor at the LASPI Laboratory in Jean Monnet University Roanne.

Texte du rabat
An insightful treatment of present and emerging technologies in fault diagnosis and failure prognosis

In Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives, a team of distinguished researchers delivers a comprehensive exploration of current and emerging approaches to fault diagnosis and failure prognosis of electrical machines and drives. The authors begin with foundational background, describing the physics of failure, the motor and drive designs and components that affect failure and signals, signal processing, and analysis. The book then moves on to describe the features of these signals and the methods commonly used to extract these features to diagnose the health of a motor or drive, as well as the methods used to identify the state of health and differentiate between possible faults or their severity. Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives discusses the tools used to recognize trends towards failure and the estimation of remaining useful life. It addresses the relationships between fault diagnosis, failure prognosis, and fault mitigation. The book also provides:

• A thorough introduction to the modes of failure, how early failure precursors manifest themselves in signals, and how features extracted from these signals are processed
• A comprehensive exploration of the fault diagnosis, the results of characterization, and how they used to predict the time of failure and the confidence interval associated with it
• A focus on medium-sized drives, including induction, permanent magnet AC, reluctance, and new machine and drive types Perfect for researchers and students who wish to study or practice in the rea of electrical machines and drives, Fault Diagnosis, Prognosis, and Reliability for Electrical Machines and Drives is also an indispensable resource for researchers with a background in signal processing or statistics.

Contenu
Contributors xiii

Acknowledgments xv

Acronyms xvii

Introduction xxi

1 Basic Methods and Tools 1

1.1 General Approach 1

1.2 Feature Extraction: Signal and Preconditioning 2

1.2.1 Raw Signals: What Kind of Signals and Sensors? 2

1.2.1.1 Current Sensors 3

1.2.1.2 Vibration Measurement and Accelerometers 13

1.2.1.3 Temperature Sensors 14

1.2.1.4 Field Sensors 16

1.2.1.5 Acoustic Sensors 16

1.2.1.6 Other Sensors 18

1.2.2 Preconditioning 22

1.2.2.1 Signal Features in the Time Domain 22

1.2.2.2 Symmetric Component, Park Component 22

1.2.2.3 Symmetric Component, Park Component 24

1.2.2.4 Signal Features in the Frequency Domain 26

1.2.2.5 Wavelet Analysis 34

1.2.2.6 Instantaneous Amplitude and Frequency 35

1.2.2.7 Bilinear Time-frequency Distributions or Quadratic

Time-frequency Distributions: Cohen's Class 36

1.2.2.7.a Uncertainty Principle of Heisenberg 37

1.2.2.7.b General Representation 37

1.2.2.7.c Properties 38

1.2.2.7.d Different Representations 39

1.2.2.8 Statistic Features 45

1.2.2.9 Cyclostationarity 46

1.2.3 Model Approach 48

1.2.3.1 Kalman Observer 51

1.2.3.2 Extended Observer 52

1.2.3.3 Unscented Kalman Filter 55

1.2.4 Parity Space 56

1.3 Feature Reduction, Principal Component Analysis 60

1.3.1 Principal Component Analysis: A Space Reduction and an Unsupervised Classification 60

1.3.2 Intercorrelation 62

1.3.2.1 Pearson Coefficient r 62

1.3.2.2 Spearman Coefficient 𝜌 63

1.3.3 Information Content: Shannon Entropy 65

1.3.4 Pattern Sizing Reduction for a Supervised Classification 65

1.3.4.1 Selection Criteria 65

1.3.4.2 Sequential Backward Feature Selection and Sequential Forward Feature Selection 67

1.3.5 Pattern Sizing Reduction for an Unsupervised Classification: Laplacian Score 68

1.3.6 Choice of the Number of Classes for an Unsupervised Classification 69

1.3.6.1 Choice of the Number of Classes with a PCA 69

1.3.6.2 General Case 70

1.3.7 Other Quality Criteria of a Classification 71

1.3.7.1 𝑅2index 71

1.3.7.2 CalinskiHarabasz Index 72

1.3.7.3 DaviesBouldin Index 73

1.3.7.4 Silhouette Index 73

1.3.7.5 Dunn Index 74

1.4 Classification Methods 74

1.4.1 Generalities 74

1.4.1.1 Supervised and Unsupervised Clustering 75

1.4.1.2 Measuring the Similarity: Different Distances 76

1.4.2 Supervised Clustering 77

1.4.2.1 k Nearest Neighbors 78

1.4.2.2 Support Vector Machine 80

1.4.2.3 Recurrent Neural Network 82

1.4.3 Unsupervised Clustering 85

1.4.3.1 Hierarchical Classification 86

1.4.3.2 K-means and Centroid Clustering 89

1.4.3.3 Self-organizing Map 90

1.5 Prognosis Methods 93

1.5.1 Prognosis Process 93

1.5.2 Time Series Extrapolation Methods 95

1.5.3 Bayesian Inference 101

1.5.4 Markov Chain 103

1.5.5 Hidden Markov Models 105

1.5.6 Rainflow 110

1.5.6.1 Hidden Semi-Markov Models 114

References 114

2 Applications and Specifics 125

2.1 General Presentation of Motor Drives 125

2.2 Electrical Machines 126

2.2.1 Basics 128

2.2.2 Magnetic Steel and Magnets 129

2.2.3 Windings and Insulation 133

2.3 Machine Models, Operation, and Control 137

2.3.1 Three-phase Windings 137

2.3.2 Induction Machines 138

2.3.2.1 Induction Machine Rotor Field Orientation 140

2.3.2.2 Direct Torque Control 141

2.3.3 Permanent Magnet AC Machines 144

2.4 Faults in Electrical Machines 146 2.4.1 ...