A Fault Identification Method of Mechanical Element Action Unit Based on CWT-2DCNN

Aiming at the problems of low recognition rate and human intervention in the traditional fault diagnosis of mechanical equipment, a fault identification method based on continuous wavelet transform (CWT) and two-dimensional convolutional neural network (2DCNN) is proposed. By collecting the vibration signals of four kinds of fault states and normal states of the worm rotation unit of the CNC machine tool, the data are preprocessed and identified. Firstly, the vibration signals of each fault state of the element action unit are CWT transformed into the corresponding two-dimensional time-frequency diagram; then, the 2DCNN fault identification model is established, and the time-frequency diagrams of various faults are input to the network as characteristic diagrams for training and testing. Through the adjustment of network parameters, the network performance is gradually optimized; finally, the hybrid domain attention module CBAM is added to further improve the network structure, and the recognition effect is compared with the initial 2DCNN. The results show that the CWT-2DCNN meta-action unit fault recognition model with an attention module can recognize the different states of meta-action units more effectively, and the fault recognition effect is better. By using this method, the different fault types of mechanical element action units can be accurately identified, which has a certain application in the field of mechanical fault identification and diagnosis.

Automated summary

This study proposes a fault identification method based on continuous wavelet transform (CWT) and two-dimensional convolutional neural network (2DCNN) to address the low recognition rate and human intervention issues in traditional mechanical equipment fault diagnosis. By collecting and preprocessing vibration signals of different fault states of a worm rotation unit, a 2DCNN fault identification model is established and optimized with the addition of a hybrid domain attention module CBAM. The results show that this method can effectively recognize different states of mechanical element action unit faults.