Bearing Fault Diagnosis Method Based on EMD and Multi-channel Convolutional Neural Network

Aiming at the problems that the traditional fault diagnosis methods are not obvious in extracting fault features and have poor generalization ability, an intelligent bearing diagnosis method based on empirical mode decomposition EMD and multi-channel convolutional neural network (MC-CNN) is proposed. Above all, EMD is performed on the bearing vibration signal, and the intrinsic mode function (IMF) components with obvious characteristics are selected by the weighted value of the kurtosis value and the correlation coefficient, and the IMF time domain map is obtained. Next, the multi-channel convolutional neural network is trained by using the IMF image data set. The experimental results show that comparedwith the data set obtained by the kurtosis value method, theweighted method obtains the IMF data set with a bearing fault diagnosis accuracy rate of 99.30%; the original signal is tested with different signal-to-noise ratio noises, and it is found that MC-CNN is significantly better than other network structures, which proved that the proposed method has good generalization ability.