A two‐phase features extraction approach for BRB based fault diagnosis of electromechanical system

Abstract

SummaryBelief rule base (BRB) is an effective nonlinear relationship modeling approach. It has been widely used in the fault diagnosis of electromechanical systems. To improve the performance of the BRB‐based diagnostic model, a two‐phase features extraction approach called CNPCA based on complex network (CN) and principal component analysis (PCA) is proposed in this paper. In the first phase, the weighted visibility graph method is applied to transform the time series data of monitored variables into complex networks. Then the statistical attributes of the constructed networks are extracted as the initial features. In the second phase, the PCA method is used to process the initial features and the principal component features are obtained. After that, the CNPCA‐BRB diagnostic model for the electromechanical system is constructed. The experimental results of the elevator fault diagnosis show that the constructed diagnostic model outperforms better than the classical ones. It demonstrates that the CNPCA approach can ensure the integrity of fault information in the features and improve the separability of the fault features.