A Novel Deconvolution Cascaded Variational Mode Decomposition for Weak Bearing Fault Detection With Unknown Signal Transmission Path

Abstract

Rolling element bearing is one of the most widely used component in mechanical equipment, whose reliability requirements are also increasing. In complex working conditions, the measured signals often contain so strong noisy interferences that the weak bearing fault features are difficult to be detected. Variational mode decomposition (VMD) is a widely used method for bearing fault features extraction. However, it also encountered some problems when dealing with weak fault signals, especially the number determination of decomposition modes, the influence of strong noise and unknown transmission path. In order to solve these problems, a novel deconvolution cascaded variational mode decomposition is proposed for weak bearing fault detection in this article. Considering the influence of transmission path and strong noise, a monitored signal is enhanced by compensation transfer function and optimized resonance component to reduce the impact of noise. Firstly, to compensate the transfer function of a complex unknown transmission path, a deconvolution method with multi-point kurtosis is used, which adjusts the filter adaptively to enhance the periodic impact component in the signal. Then, to determine the decomposition modes number and furtherly remove interferences noise, a cascaded variational mode decomposition is proposed for multi-layer noise suppression. Subsequently, a standardized square envelope spectrum is employed to detect fault characteristics of the resonant mode. Experiments verified the effectiveness of the proposed method, and comparisons show that the envelope factor is increased from 1.0762 to 34.8781, which mean that the fault signal is enhanced by the proposed method.