Intelligent Health Assessment of Aviation Bearing Based on Deep Transfer Graph Convolutional Networks under Large Speed Fluctuations

Abstract

As a critical support and fixed component of aero engines, electro-hydrostatic actuators, and other equipment, the operation of aviation bearings is often subject to high speed, high-temperature rise, large load, and other continuous complex fluctuation conditions, which makes their health assessment tasks more difficult. To solve this problem, an intelligent health assessment method based on a new Deep Transfer Graph Convolutional Network (DTGCN) is proposed for aviation bearings under large speed fluctuation conditions. First, a new DTGCN algorithm is designed, which mainly uses the domain adaptation mechanism to enhance the performance of Graph Convolutional Network (GCN) and the generalization performance of transfer properties. Specifically, order spectrum analysis is employed to resample the vibration signals of aviation bearings and transform them into order spectral signals. Then, the trained 1dGCN is used as the feature extractor, and the designed Dynamic Multiple Kernel Maximum Mean Discrepancy (DMKMMD) is calculated to match the difference in edge distribution. Finally, the aligned features are fed into the softmax classifier for intelligent health assessment. The effectiveness of the proposed diagnostic algorithm and method are validated by using aviation bearing fault data set under large speed fluctuation conditions.

Keywords

• aviation bearings
• intelligent health assessment
• large speed fluctuations
• graph convolutional network (GCN)
• transfer learning