An Adaptive Stochastic Resonance Modelling Quantified by Gini-Index for the Detection of Unknown Bearing Faults

Bearings are one of the important parts of rotating machinery, and the collected vibration signals containing bearing fault information have strong background noise. Stochastic resonance (SR) is a signal processing method that uses noise to enhance weak fault signatures. However, it requires prior fault knowledge such as bearing fault frequencies which are often difficult to obtain. In this paper, an adaptive stochastic resonance modelling quantified by Gini-index is proposed for identifying unknown faults of rolling bearings under strong background noise. The Gini index (GI) is introduced to describe the behaviour of periodic shock signal with different intensity of added noise. Based on GI and signal to noise ratio (SNR), the influence of system parameters on SR performances is investigated by simulation studies. GI is then used to guide the system parameters to find the optimal resonance response, in which prior knowledge about bearing frequencies are not needed. Two types of bearing failure experimental datasets from the gearbox and motor are processed using the proposed method respectively. The outstanding results demonstrate the effectiveness of the proposed method for unknown fault detection and diagnosis.