Sparse Component Analysis Based on Support Vector Machine for Fault Diagnosis of Roller Bearings
Gang Tang, Guozheng Li, Huaqing Wang · 2017 International Conference on Sensing, Diagnostics, Prognostics, and Control (SDPC) · 2017
In order to improve the separation performance of blind source separation, a sparse component analysis method based on support vector machine is proposed. Firstly, the sample points of the observed composite signals are selected by calculating the direction angle of the composite signal as standard to discard the interference points. And the selected sample points are trained using support vector machine method. Then the optimal classification plane is determined to classify the observed signals. In addition, the mixed matrix is estimated by means of weighted summation of the confidence intervals. Finally, the source signals are separated based on shortest path method. Experimental results manifest the proposed method can extract the fault signal of rotating machinery successfully.