Fault Diagnosis of Rolling Bearing Based on Welch Power Spectrum with Convolution Neural Network
Aiming at the problem that the fault diagnosis of rolling bearing cannot obtain high-precision recognition under small training samples and strong noise,a diagnosis method based on Welch power spectrum combined with convolutional neural network is proposed.This method takes the original time-domain vibration signal as input,transforms the data form with Welchpower spectrum and suppresses high-intensity noise at the same time,then uses the obtained power spectrum to train the convolutional neural network,and finally uses the trained model for bearing fault diagnosis.Compared with WDCNN[1]and other methods,experiments found that under mixed load,the average recognition rate of this method is 99%correct.Oth-er methods require at least 20times the training sample size to achieve this accuracy,which is significantly better than methods such as WDCNN.The anti-noise experiment results show that the stronger the noise interference to the signal,the better the an-ti-noise performance of this method,and its anti-noise performance is significantly better than WDCNNand other methods.
万方数据
维普
