Graph neural network based aerospace motor bearing failure diagnosis

The motor bearings is one of the critical components to ensure the reliable operation of aircraft. In this paper, we propose a fault diagnosis method based on the Graph Convolutional Network (GCN) model. The proposed method integrates the dependencies among multivariate sensor time-series signals from various nodes and constructs a graph data structure that accurately represents the relationships between the nodes. This approach leverages the strengths of GCN in capturing the spatial dependencies and complex interactions inherent in the data, providing a robust framework for fault diagnosis. Finally, using the HIT public dataset as an example, the training results of different model parameters are compared to improve model accuracy and achieve an ideal fault diagnosis accuracy.