Gear Fault Diagnosis Based on Multi-scale Convolutional Neural Network

Convolutional neural network (CNN) has been widely used in fault diagnosis of mechanical equipment such as gears. However, when using traditional CNN for fault diagnosis, there are still difficulties such as taking a long time to train model and feature extraction interfered by noise. Therefore, while ensuring the accuracy of the model, in response to the above problems, this paper proposes a new CNN model - full vector multi-scale CNN (FVMCNN). Firstly, the full vector spectrum algorithm is used to fuse the multi-channel vibration signals of a single sensor in order to retain the original information to the greatest extent and input it from the data source of simplified model to improve the efficiency of model training; Secondly, FVMCNN adaptively extracts multi-scale features of sensor through branch network and the feature fusion strategy based on attention mechanism and the scaling module (DRM) adaptively fuses the features extracted by the branch network to suppress the influence of background noise. Experimental verification is carried out on the parallel shaft gear dataset, and the results show that FVMCNN can achieve more accurate, fast and robust gear fault diagnosis compared with other traditional gear fault diagnosis algorithms.