Research On Motor Fault Diagnosis Based On Sparse Signal Representation

Based on the “Research on the key technology and application of morphological component analysis”(natural science project of science research plan in colleges and universities of Hebei province education department. No. Q2012051), this dissertation aimed to motor bearing fault diagnosis based on sparse signal representation. The signal simulation and experiment are carried out. The main research results of this dissertation provide a new way to fault detection and diagnosis of motor. The main research results are given as follows:1. From the view of theoretical analysis and engineering application, the theories of sparse signal representation are introduced briefly. Then the anti-noise, signal decomposition and time frequency distribution performance are discussesed in detail by simulative signal. This method provides a reliable tool to extract the fault feature of the motor bearing fault signals.2. The fault diagnosis method of motor bearing based on morphological component analysis(MCA) is put forward. This method takes better advantage of both morphological diversity and sparsity, using sparse redundant signal representation. The distinct signal component can be sparse represented and separated in different dictionary. Then the feature information of fault bearing can be distinguished in frequency domain or time-frequency domain. The experimental results show that morphological component analysis can effectively diagnosis the fault of motor bearing.3. The fault diagnosis approach of motor bearing based on rational-dilation wavelet transform(RADWT) and morphological component analysis is put forward. This method is combined rational-dilation wavelet transform and morphological component analysis. The suitable quality factor can be selected according to structure feature of bearing fault vibration signal. The rational-dilation wavelet transform of low quality factor can represent transient components and the sustained oscillations can be represented by high quality factor. Then the transient components can be represented sparsely and extracted from the signal by morphological component analysis. Therefore, the bearing fault feature can be distinguished effectively. This way provides a new tool to fault detection and diagnosis of motor bearing.4. The fault diagnosis method of motor bearing based on rational-dilation wavelet transform and basis pursuit is put forward. This method is combined rational-dilation wavelet transform and basis pursuit(BP). As the rational-dilation wavelet transform is over-sampled, the wavelet coefficients that reconstruct a signal are not unique. It is useful to find a sparse set of wavelet coefficients for bearing fault signal. Therefore, the bearing fault signal can be sparse represented and extracted by basis pursuit and rational-dilation wavelet transform with suitable quality factor and redundancy. The useful transient components can be separated from the vibration signal. The experiment results show the basis pursuit based on rational-dilation wavelet transform are highly efficient to processing the motor bearing fault signal with transient noise components. The sparse signal representation has wide application foreground in motor bearing fault diagnosis domain.