Compound Bearing Fault Detection Under Varying Speed Conditions With Virtual Multichannel Signals in Angle Domain

Abstract

Mechanical fault diagnosis under varying speed conditions has gradually become an important issue in rotating machinery monitoring, especially the research on compound fault diagnosis is still rare. Order analysis is one of the most effective and widely used methods to eliminate the effects of speed fluctuations; however, it is difficult to extract compound fault features directly. Moreover, there are often many restrictions on the number and locations of sensors to be installed in practice, which also brings great challenges to compound fault diagnosis under varying speed conditions. In this article, a novel compound fault detection method with virtual multichannel signals in the angle domain is proposed to solve these problems for rolling bearing monitoring under varying speed conditions. First, we aim to use only a vibration acceleration sensor and a noncontact rotational speed sensor to adapt to limitations on sensor installation in practical applications. However, this may result in an underdetermined problem when trying to detect a compound fault with only a single-channel vibration signal. Then we make it up through constructing virtual multichannel signals based on variational mode decomposition theory, before which the nonstationary vibration signal is converted from the time domain into a stationary signal in the angle domain with computed order tracking to eliminate speed fluctuations. Finally, these virtual signals can be easily taken as input to a normal independent component analysis, and different faults can be detected in the order spectrum of each independent component. Case studies on rolling bearing experiments verified the advantages of the proposed method in compound fault detection under varying speed conditions.