An Intelligent Fault Diagnosis Method of Rolling Bearing with Wide Convolution Kernel Network

摘要

In this paper, a method based on two wide convolution kernels network (WCNN) is proposed for the fault diagnosis of rolling bearing. Firstly, the vibration signal of bearing was input into WCNN network which acquired and extracted the intrinsic features of vibration signal by itself. The whole training process was optimized by the Adam algorithm. Finally, the bearing fault was diagnosed by softmax function. Compared with other common methods, this method does not require manual extraction of fault features. And at the same time, average diagnostic accuracy can be above 99% in different noisy environment and with less training time.