Fault Diagnosis Algorithm Based on Lightweight Convolutional Neural Network

With the development of deep learning technology, many intelligent diagnosis algorithms applied to mechanical equipment operation and maintenance have emerged, but most of these models have complex structure and large amount of calculation, which limits their application prospects. This paper proposes a lightweight convolutional neural network model, which compresses network structure on the premise of ensuring the diagnostic accuracy, providing basis of construction of intelligent diagnosis system. Based on LeNet, the classical convolutional neural network, and the lightweight model Mobile Net, this network model is constructed for bearing fault diagnosis by using deep separable convolution and global average pooling methods. A fault diagnosis experiment of the constructed network was carried out to test diagnosis accuracy of this network and compare it with mainstream models. The experimental results show that the fault diagnosis network used in this paper obtains better diagnosis accuracy under the condition of greatly reducing the parameters.