Improved Convolutional Neural Network Fault Diagnosis Method Based on Dropout

Abstract

Aiming at the over-fitting problem of traditional deep learning method in bearing fault diagnosis model, this paper proposes an improved convolutional neural network fault diagnosis method. This method introduces the Dropout optimization method at the fully connected layer of the neural network model, and temporarily discards some neurons from the neural network, thereby reducing network parameters and achieving data dimensionality reduction. By comparing and analyzing the method described in the paper with the traditional CNN network, the results show that the method described in the paper can effectively alleviate the overfitting phenomenon of the traditional CNN network model in bearing fault diagnosis. The model has a strong generalization ability and diagnosis. The result has a higher accuracy.