Intra-Inter Feature Ranking based Feature Selection Method for Bearing Fault Classification

Abstract

The intelligent condition monitoring system has been accepted by the modern industry. As a result, recently, fault diagnosis methods are developed by integrating signal processing and artificial intelligence techniques. The various signal processing methods are utilized to create a bearing feature space. As an effect, existing feature space is ample and therefore, in this paper, feature selection routine for bearing fault diagnosis is proposed to identify the dominant feature from existing feature space. The suggested feature selection routine uses an ensemble of inter-intra feature ranking method to determine the fault representative, and then the ensemble classifier has trained to learn the defect patterns. The proposed method testified using two vibration datasets, and decent results are achieved. Also, the effect of ensemble feature ranking and intra feature ranking method has observed in comparison to the proposed feature selection routine. © 2019 IEEE.