Fault Diagnosis of Rotating Machine Using an Indirect Observer and Machine Learning

Abstract

Bearing is one of the important mechanical components to reduce friction in rotating machines. Early fault diagnosis in bearings is an important challenge to the prevention of full failure and avoiding disorder of the machine. In this paper, an indirect observer and machine learning technique are adopted for fault identification in bearing. To develop an indirect observer, in the first step, the autoregressive with uncertainty modeling technique is proposed to modeling the RMS (indirect) normal signal of bearing. After that, the robust (sliding fault detection) proportional multi integral with autoregressive external input modeling (ARPMI) observer was used to solve the unknown signal estimation in bearing. Besides, the support vector machine (SVM) technique for fault classification is proposed. The effectiveness of the proposed scheme is validated using Case Western Reverse University (CWRU) dataset. Experimental results show that, the proposed scheme improves the average performance for various rotational speed fault identification by about 10.5% and 13.5% compared with the proportional multi integral with autoregressive external input modeling (APMI) observer and proportional-integral with autoregressive external input modeling (API) observer, respectively.