Bearing fault diagnosis based on optimal convolution neural network

Artificial intelligence method does not need artificial operation to extract fault characteristics, and greatly reduces the error of human operation. According to the principle of convolution neural network, a new method of optimal convolution neural network (CNN) model is proposed. Firstly, according to the principle of symmetrized dot pattern (SDP), the vibration signal is transformed into a symmetrical image in polar coordinates, which is also called snowflake image. Then, the SDP images are input into the input layer of the convolutional neural network, and the model can diagnose the fault type automatically. By adjusting the number of convolution layers and the size of convolution kernel, the convolution neural network model is determined according to a new index involving accuracy and time ratio. Finally, the test set is used to validate the robustness of the method under different bearing operating conditions.