Bearing residual life prediction based on principal component feature fusion and SVM

In order to solve the problem that using single feature quantity for bearing residual life prediction had large error and it was difficult to estimate bearing residual life under the condition of limited data samples, a bearing residual life prediction method based on principal component feature fusion and support vector machine(SVM) was proposed. This method collects data samples of vibration acceleration signals and extracts the characteristic indexes such as RMS, peak value and wavelet entropy to characterize the degradation trend of bearings. The principal component analysis is used to fuse multiple feature indexs to eliminate the redundancy and correlation between features, and construct regressive feature quantities with relative multi-feature; the regressive feature quantities are input into SVM model for bearing residual life prediction. The field engineering application results show that the bearing residual life prediction method based on principal component feature fusion and SVM can screen out the principal components which contain most of the information under small sample condition, thus reducing the calculation amount while ensuring the prediction accuracy.