Bearing Fault Diagnosis Based on Adaptive Convolutional Neural Network With Nesterov Momentum
Shuzhi Gao, Zhiming Pei, Yimin Zhang, Tianchi Li
Abstract
It is difficult to achieve satisfactory classification results for bearing fault diagnosis methods based on prior knowledge. This paper presents an adaptive convolution neural network based on Nesterov momentum for rolling bearing fault diagnosis. Firstly, the traditional momentum method in the network is replaced by Nesterov momentum. Nesterov momentum can predict the falling position of parameters and adjust the parameters in advance, to avoid the problem that the traditional momentum method is likely to miss the optimal solution. Secondly, in order to improve the generalization ability of the network, an adaptive learning rate rule which dynamically adjusts the learning rate according to the rate of error change is proposed. Finally, the original vibration signals are directly inputted into the proposed network to train the fault diagnosis model, and the test data are used to evaluate the model. The experimental results show that compared with the traditional convolutional neural network, the proposed method improves the convergence of the neural network and effectively improves the accuracy of bearing fault classification.