Bearing Fault Diagnosis with Varying Speed Conditions Based on Multi-Resolution Singular Value Decomposition

Abstract

Fault diagnosis with varying speed conditions has become a topic of interest in the field of rotating machinery diagnosis. Computed order tracking (COT) can successfully remove the effect of varying speed, however it is hard to directly identify the fault characteristics. In this study, a fault diagnosis method with varying speed bearing on the basis of multi-resolution singular value decomposition (MRSVD) is presented. First, the optimal resonance frequency band of the original vibration signal is selected by Fast-Kurtogram. Then, MRSVD is performed on the filtered signal, and the total decomposition layers of MRSVD is related to the decomposition layer corresponding to the maximum mutual information (MI) of detail signal and filter signal. Taking the mean value of all MIs as the threshold, the detail signal, whose corresponding MI is larger than the threshold, is considered as a sensitive one. All the sensitive detail signals are combined into reconstructed signal for computed order tracking (COT) analysis. The advantages and effectiveness of the new method are verified with simulation and experimental signals under varying speed conditions.