Bearing fault diagnosis method based on 1D-CNN

In view of the difficulties in feature extraction and fault identification caused by the strong nonlinearity of bearing vibration signals and the redundancy of irrelevant features, this paper proposes a bearing fault diagnosis method based on one-dimensional convolutional neural network (1D-CNN). This method combines the temporal characteristics of bearing data, performs one-dimensional convolution (Conv1D) processing on the traditional convolutional neural network, pre-processes the original data by Fourier time-frequency domain analysis and simple data normalization, and uses sklearn to design and construct the network structure to achieve real-time prediction and diagnosis of bearing faults. Compared with the traditional CNN model, 1D-CNN has fewer network parameters and powerful feature extraction capabilities. The verification experimental results show that the proposed method can achieve 100% accuracy in the bearing fault diagnosis task, and has certain relative advantages compared with other traditional machine learning methods. This method has high classification accuracy and generalization ability for the diagnosis of rolling bearing faults.