Abstract:

As for the fault diagnosis of rolling bearings for urban rail trains，an integrated intelligent diagnosis method was proposed based on principal component analysis （PCA） and random forest （RF）. Firstly， original feature vectors of 21 dimensions were extracted from time domain and frequency domain. Principal component analysis was then utilized to conduct dimension reduction accounting for the correlations within features. Finally， the size-reduced features act as the inputs of the RF. Consequently， an integrated intelligent diagnostic model called PCA-RF model was established. Experimental results demonstrate that the PCA-RF model is superior to other methods such as support vector machines and RF with original feature vectors as inputs， in terms of efficiency and effectiveness. In addition， the proposed method can obtain satisfied identifications among complicated bearing fault patterns which involve not only various fault locations but fault severity levels. In short， the PCA-RF model had a certain value in the diagnosis of rolling bearing faults of urban rail vehicles.

Key words: vehicle engineering, principal component analysis, random forest, rolling bearing, intelligent diagnosis