Fault Diagnosis for Aircraft Aileron Actuator Based on Graph Regularization Network

During the operation of aircraft aileron actuators, a large number of monitoring samples will be accumulated. However, most of the monitoring samples are without labels since the lack of effective data management and data analysis. Most traditional fault diagnosis methods simply use sample parameters to analyze the health condition of aircraft aileron actuators. These methods ignore the relationship between sample parameters and fail to effectively use the fault information in unlabeled samples, which results in low accuracy of fault diagnosis under label missing conditions. Graph regularization network can exploit the relationship between sample parameters to improve the fault diagnosis accuracy and performance under label missing conditions. In this paper, a fault diagnosis method based on graph regularization network is proposed. First, for the monitoring samples, a sample association graph is constructed based on the K-nearest neighbor method. Second, for the nodes in the graph with labeled samples, subgraphs are extracted and generated, each containing one labeled node and K unlabeled nodes. Then, a fault diagnosis model based on graph regularization network is constructed by using subgraphs, and the relationship between sample parameters is used to constrain the optimization, forcing the fault diagnosis model to be trained toward a fixed direction. Finally, a simulation dataset is generated by injecting different faults to an aileron actuator simulation model established using Simulink and AMESim software, and some of the fault labels are removed for case validation of the proposed fault diagnosis method. The results show that the proposed fault diagnosis method based on graph regularization network can effectively use the relationship between sample parameters and improve the performance of fault diagnosis under label missing conditions compared with typical supervised learning methods.