Machinery Fault Classification Method Based on Feature Contribution Rate

In order to improve the accuracy of fault classification of complex machinery such as reciprocating compressor and aeroengine, an analysis method combining Dirichlet process mixture model(DPMM) with Bayesian inference contribution(BIC) is proposed according to the characteristic that the sensitivities of feature parameters are vary from fault to fault. It is used to self-learn the statistical distribution model of high dimensional features of the mechanical vibration signals by DPMM method, and the contribution rate of each feature to the model is calculated according to the BIC theory. The fault classification is realized by analyzing the differences between the feature contribution rates of the observed data and different kinds of fault data. The results indicate that the average classification accuracy of the proposed method increases by 19.29% compared with the fault diagnosis method based on Gaussian mixture model(GMM), and increases by 32.71% compared with the fault diagnosis method based on Relief algorithm. Furthermore, this method has characteristics of high timeliness and strong generalization performance. It can effectively classify the complex mechanical faults. © 2020, Editorial Department of JVMD. All right reserved.