Fault diagnosis method of centrifugal pump based on PARAFAC-SVM

Centered on parallel factor analysis(PARAFAC) and support vector machine(SVM), a method of centrifugal pump(CP) fault diagnosis is proposed. A non-stationary multi-fault mode single-source signal feature extraction method is proposed, based on the theory of parallel factor analysis, i.e. spatial information is introduced in time-frequency analysis, and after parallel factor analysis, the spatial signal, the spectral signal (mode2) and the time-domain signal (mode3) are generated. The latter two can obviously describe the natural or fault state of the mechanical equipment by comparing the experimental analysis, and this function is used to evaluate the mapping relationship between the centrifugal pumps in different states and their respective mode2 and mode3. A total of six components in mode2 and mode3 are used as feature vectors, which are input to the SVM classifier, optimized by the improved particle swarm algorithm for fault classification, by constructing an experimental platform to collect the nonlinear multi-fault mode feature signals of centrifugal pumps. Compared with the wavelet packet energy features, the proposed method is more convenient for feature extraction, the classification accuracy of the proposed classifier is significantly improved, and the complexity of the proposed classifier is not significantly increased.