科研之友
微信
新浪微博
Facebook
分享链接
摘要
The health of a complex electromechanical system is dynamic and is accompanied by a full life cycle. Due to the complexity and coupling of complex electromechanical systems, the establishment of a dynamic and accurate model for the health state is difficult. A belief rule base (BRB) shows outstanding performance in modeling complex systems because it can combine both quantitative information and expert knowledge. In this paper, a double-layer BRB model is proposed to predict the health state of a complex electromechanical system. The two layers achieve different functions: BRB_layer 1 is used to establish the dynamic change of the time series of features, BRB_layer2 is employed to combine the features for predicting the health state of the complex electromechanical system. During this process, the infinite irrelevance method is utilized for feature selection in reducing the scale of the BRB model. Considering the initial parameters are given by experts, which may have boundedness and may not be appropriate for engineering practice, the projection covariance matrix adaption evolution strategy (P-CMA-ES) is chosen as the optimization algorithm to train the initial parameters. To verify the rationality and effectiveness of the proposed model, the low-frequency vibration fault of a certain aero-engine is taken as an example. The results show that the proposed method can predict the health state of a complex electromechanical system precisely according to current and historical data.
