Bearing Fault Diagnosis Based on Multi-scale Neural Networks

In this paper, a framework using multi-scale neural networks for bearing fault diagnosis is proposed. This framework consists of two stages. The first stage decomposes raw bearing signal into multiple multi-scale signals by signal decomposition and transform. The second stage applies the multi-scale signals to the corresponding input channels designated for the multi-scale neural network and concatenates outputs of all parallel sub-neural networks into a single channel which is further used as the input to a fully connected layer for classification. In comparison with the other bearing fault diagnosis methods, our proposed method can achieve high classification accuracy of 98.7% using one-dimensional convolutional neural networks (1D-CNN) with less computation based on Case Western Reserve University Dataset (CWRU).