Rolling bearing fault diagnosis based on improved convolutional neural networks with attention mechanism under strong noise

Abstract

Aiming at the problem that the vibration signal characteristics of rolling bearings are unstable under the background of strong noise, resulting in poor generalization of the diagnostic algorithm, weak noise resistance, and difficulty in achieving effective fault diagnosis, a new method named improved convolutional neural networks with attention mechanism under high noise is proposed. Based on the deep convolutional neural network, the gated recurrent unit (GRU) is introduced to solve the gradient explosion problem in the neural network, the attention mechanism (attention) is introduced to improve the adaptive ability of the network and reduce the difficulty of hyperparameter selection, and the SVM classifier is used to replace the classification layer of the deep convolutional neural network to improve the accuracy of classification. In order to verify the robustness and generalization of the proposed method in a strong noise environment, the bearing dataset of Western Reserve University was used for verification. The experimental results show that the classification accuracy of the proposed algorithm is 24.7 percentage points higher than that of the WDCNN algorithm, which proves that the proposed method has good noise immunity and generalization under the background of high noise.