Rolling Bearing Fault Diagnosis Based on Time-Frequency Transform-Assisted CNN: A Comparison Study
Baoye Song, Yiyan Liu, Peng Lu, Xingzhen Bai
Abstract
This paper is concerned with a comparison study on three time-frequency transform methods for CNN-based rolling bearing fault diagnosis, including short-time Fourier transform (STFT), continuous wavelet transform (CWT), and S-transform. The time-frequency transforms are exploited to transform the bearing fault data from 1D vibration signals to 2D time-frequency images, which are then fed into a dedicatedly designed 2D-CNN for fair performance comparison. To evaluate the performance of the time-frequency transform-assisted CNNs, several experiments are implemented based on the designed CNN and the bearing fault data. The superiority of S-transform assisted CNN is confirmed through the evaluation indicators calculated by the fault diagnostic results.