On cross-domain feature fusion in gearbox fault diagnosis under various operating conditions based on Transfer Component Analysis

This paper addresses the cross-domain feature extraction and fusion from time-domain and frequency-domain with spectrum envelop preprocessing and time domain synchronization average principle using Transfer Component Analysis (TCA) for gearbox fault diagnosis. Considering TCA is developed based on kernel methods, the effects of different kernels including Gaussian kernel, Linear kernel, Polynomial kernel and PolyPlus kernel on the performance of TCA are investigated and evaluated in comprehensive experiments of gearbox testbed under various operating conditions. The experimental results show that the presented method can extract and fuse the cross-domain features of gearbox conditions by enhancing the reuse of historical data under various operating conditions efficiently, compared with other baseline dimension reduction methods. In addition, TCA with Gaussian kernel presents best performance, especially for low frequency levels of operation.