An interpretable fault diagnosis model for EHA DC drive circuit based on boosting fusion ensemble strategy under actual operating conditions

The DC drive circuit of aircraft electro-hydrostatic actuator (EHA) is the key link to ensure the safety of flight mission, and its fault will pose a serious threat to flight control performance. Aiming at the problems of data imbalance, degradation of diagnostic performance caused by complex disturbance environment, lack of interpretability of model decision-making process and lack of credibility of results in intelligent fault diagnosis in engineering practice, this paper proposes an interpretable fault diagnosis strategy that integrates boosting ensemble learning and Bayesian estimation (iBoEBayes). In the feature engineering stage, based on the maximum information coefficient (MIC) feature correlation analysis and feature screening method, key features are extracted from the EHA drive circuit data collected in the experiment to improve the diagnostic efficiency and reduce the model complexity. The experimental results show that the strategy achieves excellent performance under both data balance and non-balance conditions. Especially in the environment of insufficient fault data and multi-source strong interference, the diagnostic accuracy and stability are significantly better than the comparison algorithm. In the part of interpretability verification, the decision basis of the model is systematically verified by Bayesian prior and posterior probability analysis. Combined with the feature mean and variance analysis, the MIC feature results are mutually confirmed at the global and local levels, thus effectively enhancing the transparency and credibility of the model diagnosis results. The research results provide a reliable theoretical basis and method support for the data imbalance, disturbance environment and interpretable fault diagnosis requirements of the aircraft EHA drive circuit under complex engineering conditions.