Signal Processing for Fault Detection and Diagnosis in Electric Machines and Systems

Cover
Contents
About the editors
Introduction
1 Parametric signal processing approach
1.1 Fault effects on intrinsic parameters of electromechanical systems
1.1.1 Main failures and occurrence frequency
1.1.2 Origins and consequences
1.1.3 Condition-based maintenance
1.1.3.1 Fault detection methods
1.1.3.2 Fault effects on stator currents
1.1.4 Motor current signature analysis
1.1.4.1 Fault frequency signatures
1.1.4.2 Stator currentAM/FM modulation
1.2 Fault features extraction techniques
1.2.1 Introduction
1.2.2 Stator current model under fault conditions
1.2.2.1 Model assumptions
1.2.2.2 Stator current modelling
1.2.3 Non-parametric spectral estimation techniques
1.2.4 Subspace spectral estimation techniques
1.2.5 ML-based approach
1.2.5.1 Exact ML estimates
1.2.5.2 Approximate ML estimates
1.2.5.3 Model order selection
1.3 Fault detection and diagnosis
1.3.1 Artificial intelligence techniques briefly
1.3.2 Detection theory-based approach
1.3.2.1 Background on binary hypothesis testing
1.3.2.2 GLRT for fault detection
1.3.3 Simulation results
1.3.3.1 Estimation performance
1.3.3.2 Fault detection performance
1.4 Some experimental results
1.4.1 Experimental set-up description
1.4.2 Eccentricity fault detection
1.4.3 Bearing fault detection
1.4.4 Broken rotor bars fault detection
1.5 Conclusion
References
2 The signal demodulation techniques
2.1 Introduction
2.2 Brief status on demodulation techniques as a fault detector
2.2.1 Mono-component and multicomponent signals
2.2.2 Demodulation techniques
2.2.2.1 Mono-dimensional techniques
2.2.2.2 Multidimensional techniques
2.3 Synchronous demodulation
2.4 Hilbert transform
2.5 Teager–Kaiser energy operator
2.6 Concordia transform
2.7 Fault detector
2.7.1 Fault detector based on HT and TKEO demodulation
2.7.2 Fault detector after CT demodulation
2.7.3 Synthetic signals
2.7.3.1 Balanced system (ψ = 0)
2.7.3.2 Unbalanced system (ψ0 = 0, ψ1 = 2π/3, ψ2 = −2π/3)
2.7.3.3 Unbalanced system (ψ0 = 0, ψ1 = 2π/3, ψ2 = −2π/3) under nonstationary supply frequency
2.8 EMD method
2.9 Ensemble EMD principle
2.10 EEMD-based notch filter
2.10.1 Statistical distance measurement
2.10.2 Dominant-mode cancellation
2.10.3 Fault detector based on EEMD demodulation
2.10.4 Synthetic signals
2.11 Summary and conclusion
References
3 Kullback–Leibler divergence for incipient fault diagnosis
3.1 Introduction
3.2 Fault detection and diagnosis
3.2.1 Methodology
3.2.2 Application example of the methodology
3.3 Incipient fault
3.4 FDD as hidden information paradigm
3.4.1 Introduction
3.4.2 Distance measures
3.4.3 Kullback–Leibler divergence
3.5 Case studies
3.5.1 Incipient crack detection
3.5.2 Incipient fault in power converter
3.5.3 Threshold setting
3.5.4 Fault-level estimation
3.6 Trends for KLD capability improvement
3.7 Conclusion
References
4 Higher-order spectra
4.1 Introduction
4.2 Higher-order statistics analysis: definitions and properties
4.2.1 Higher-order moments
4.2.2 Power spectrum
4.2.3 Bispectrum and bicoherence
4.2.4 Estimation
4.3 Bispectrum use for harmonic signals’ nonlinearity detection
4.3.1 Case 1: a simple harmonic wave at frequency F0
4.3.2 Case 2: sum of two harmonic waves at independent frequencies F0,F1; and with F1 = 2F0
4.3.3 Case 3: sum of three harmonic waves at coupled frequencies, F2 = F0 + F1
4.3.4 The use of bispectrum to detect and characterize nonlinearity
4.3.4.1 QPC detection
4.3.4.2 Robustness against the presence of additive Gaussian noise
4.4 Practical applications of bispectrum-based fault diagnosis
4.4.1 BRB fault detection
4.4.1.1 Simulation and experimental tests for BRB fault
4.4.1.2 Model of the BRB stator current
4.4.1.3 Numerical simulation
4.4.2 Bearing multi-fault diagnosis based on stator current HOS features and SVMs
4.4.2.1 Bearing defect signatures
4.4.2.2 BDs stator current bispectrum: a theoretical approach
4.4.2.3 Features extraction and reduction
4.4.2.4 Bearings *multi-fault classification proposed method
4.4.2.5 BD classification based on SVM
4.4.2.6 Experimental results
4.4.2.7 Training and test vectors
4.4.3 Bispectrum-based EMD applied to the nonstationary vibration signals for bearing fault diagnosis
4.4.3.1 Nonstationary nature of defective REB vibration response
4.4.3.2 Brief description of EMD
4.4.3.3 Experimental results
4.4.4 The use of SK for bearing fault diagnosis
4.4.4.1 SK and its application for bearing fault diagnosis
4.4.4.2 SESK proposed method
4.4.4.3 Experimental results
4.5 Conclusions and perspectives
Appendix A
Appendix B
References
5 Fault detection and diagnosis based on principal component analysis
5.1 Introduction
5.2 PCA and its application
5.2.1 PCA method
5.2.2 The geometrical interpretation of PCA
5.2.3 Hotelling’s T2 statistic, SPE statistic and Q–Q plots
5.2.4 Fault detection based on PCA for TE process
5.2.4.1 Case study on Fault 4
5.2.4.2 Case study on Fault 11
5.2.5 Fault diagnosis based on PCA for multilevel inverter
5.2.5.1 Time–frequency transform based on FFT
5.2.5.2 FDD based on PCA
5.2.5.3 Experimental tests
5.3 RPCA and its application
5.3.1 RPCA method
5.3.1.1 Relative Transform
5.3.1.2 Computing RPCs
5.3.2 The geometrical interpretation of RPCA
5.3.3 Fault detection based on RPCA for assembly
5.3.4 Dynamic data window control limit based on RPCA
5.3.5 Fault diagnosis based on RPCA for multilevel inverter
5.4 NPCA and its application
5.4.1 NPCA method
5.4.2 Fault detection based on NPCA for wind power generation
5.4.3 Fault detection based on NPCA for DC motor
5.4.4 ACL based on NPCA
5.4.5 Fault detection based on NPCA-ACL for DC motor
5.5 Conclusions and future works
References
Conclusion
Index
Back Cover