Pattern recognition of a sensitive feature set based on the orthogonal neighborhood preserving embedding and adaboost_SVM algorithm for rolling bearing early fault diagnosis

Abstract

Early fault diagnosis is a hotspot and difficulty in the research of mechanical fault diagnosis. An early fault diagnosis method based on the orthogonal neighborhood preserving embedding and Adaboost_SVM algorithm for rolling bearing early fault diagnosis is proposed in this paper. Firstly, the vibration signals of rolling bearings are measured online. The correlation coefficients between the early fault indicators and the performance degradation are deeply analyzed based on the full-lifetime vibration data of rolling bearings so as to select the sensitive fault indicators for further fault diagnosis. Secondly, the orthogonal neighborhood preservation embedding (ONPE) is employed to eliminate the redundant information from the original multi-domain feature set. Finally, the classical SVM is improved to form the Adboost-SVM for the early fault diagnosis of rolling bearings. The feasibility and validity of this method are verified by applying the early fault diagnosis of rolling bearings. The results show that Adaboost_SVM can greatly enhance the diagnosis capability for weak features of early faults.