Health Diagnosis of Gear Based on Artificial Intelligence Deep Learning and Rigid-Flexible Coupling Dynamics Model

As the core of transmission system, gear fault could be the cause of a huge train accident. In this paper, a novel diagnosis strategy of gear fault based on the artificial intelligence deep learning method and the rigid-flexible coupling dynamic analysis technology is proposed. Firstly, the train-rail dynamics model is created in SIMPACK and ANSYS. The finite element analysis is carried out to make the rail flexible. The rigid-flexible coupling model is obtained by introducing the flexible rail into the rigid SIMPACK model. Then, parametrically model the healthy and the fault gears in MATLAB and SOLIDWORKS. The gear pair is loaded into SIMULINK to sample vibration signals, which are then imported into SIMPACK as vibration exciter to stimulate the dynamic response vibration signal of the train-rail system. Finally, considering that image is the ideal medium for complex information representation, the vibration signal is converted into the image to generate the training dataset of the deep learning network. The improved Efficient Net with the Coordinate Attention Module is proposed and trained to recognize gear faults. The test experiment verifies the effectiveness and reliability of the proposed method.