Performance degradation assessment of rolling bearing based on bispectrum and support vector data description

Abstract

In order to realize equipment’s near-zero downtime and maximum productivity a rolling bearing performance degradation assessment is proposed, which is one of the most important techniques. The traditional feature extraction methods based on stationary and Gauss characteristics were unfit to handle non-stationary and non-Gauss signals when the rolling bearing performance begins to degrade. The higher-order statistics (HOS) are fit for handling the non-stationary and non-Gauss signals. Bispectrum not only possesses all the properties of HOS but also has a higher computational efficiency because of a lower order compared with other types of HOS. The support vector data description (SVDD) is a single value classification method which can overcome the problem of a lack of samples in the degradation process of the rolling bearing. A method is proposed in the paper which is based on the bispectrum and SVDD. Firstly, the rolling bearing data of normal state is collected and handled by bispectrum. Based on feature vectors extracted from normal data an SVDD model fitting a tight hypersphere around them is trained. The general distance of test data handled by the bispectrum to this hypersphere is used as the degradation index. In the end, through simulation and a rolling bearing’s accelerated life test, the effectiveness and feasibility of the proposed method is verified.