Abstract
Accurately identifying bearing faults in aeroengines is crucial for maintaining their lifespan and cost. However, most current models are black-box models, such as deep learning models such as deep neural networks. The decision-making process of these models is more complex and lacks interpretability, which results in insufficient credibility of the results. Furthermore, data collected in real industrial environments can suffer from unbalanced sample categories. Moreover, the models can suffer from local ignorance in the prediction process. These problems can lead to a decrease in the prediction accuracy of the model. Therefore, a fault diagnosis method based on the interpretable belief rule base with a dynamic power set (D-HBRBP-I) is proposed in this study. First, a diagnostic model based on a belief rule base with a dynamic power set was used to address the problem of sample category imbalance and local ignorance. Second, optimizing the model via the P-CMAES algorithm with interpretability constraints can ensure the interpretability of the model after optimization. Finally, experiments were conducted on an aeroengine-bearing dataset. The results show that the proposed model effectively solves the above problem while achieving 99% accuracy.