A Vector Angle Method of Rolling Bearing Fault Classification by Phase-Space Reconstruction Technique

Abstract

Abstract A fault classification method is developed by the phase-space reconstruction technique for a rotor system with a rolling bearing fault. Based on the nonlinear time series analysis, proper choices on time delay and embedding dimension are firstly discussed to accomplish the phase-space reconstruction from an arbitrarily one-dimensional time series, then the vector angle calculation is derived for each point in the reconstructed trajectory of an illustrating artificial signal, from which the vector angles of the points are composed of 90° and non-90° ones. To perform the fault feature analysis of the rolling bearing, an experimental rig of the rotor-bearing system and dynamical model of the system are established to collect the one-dimensional acceleration signals of three types of rolling bearing faults, and the topology of the reconstructed trajectory in three-dimensional phase space is characterized. It is interesting to find that the effects of various rolling bearing faults on the reconstructed trajectories are different, and the fault features can be extracted successfully by the distribution percentage of 90° and non-90° vector angles of the points in the reconstructed trajectory in three-dimensional phase space, which cannot be identified by the traditional attractor reconstruction method. Moreover, the higher the fault frequency, the lower proportion the healthy points.