Rolling element bearing diagnosis based on singular value decomposition and composite squared envelope spectrum

The diagnosis of early-stage defects of rolling element bearings (REBs) using vibration signals is a very difficult task since bearing fault signals are usually weak and masked by shaft rotating signals, gear meshing signals, and strong background noise. Therefore, there are two important main points for bearing fault diagnosis, namely, extracting weak bearing fault signals and identifying the fault types. In this paper, a new bearing fault diagnosis method combining singular value decomposition (SVD) and the squared envelope spectrum (SES) is proposed. An optimal range for the number of rows in the Hankel matrix used in SVD for bearing fault diagnosis is obtained by means of numerical simulation. Then, the sub-signals obtained by SVD are grouped according to their similarity. Eventually, the composite SES is applied to identify the fault type. The performance of the new method is evaluated by means of the actual vibration signals collected from a full-scale test rig of a high-speed vehicle traction system