Sparse subspace clustering for bearing fault classification

Abstract

Bearing is a critical component that effects operational performance of machine. Fault classification to bearing that aims to identify category of bearing fault is helpful to improve reliability and safety of bearing. In this paper, a classification process is presented based on sparse subspace clustering. A sample corresponds to a specific fault state of the bearing is represented by its neighbourhood. Coefficient for data representation is solved by sparse representation. Spectral clustering is performed on the coefficient to classify the samples into its category. Effectiveness of the presented method is validated by test data of bearing with different degrees of fault. Comparison between sparse subspace clustering and other subspace analysis methods shows its effectiveness for classification further.