Feature Extraction of Bearing Fault Based on Grey Wolf Optimization and Variable Scale Stochastic Resonance

Early fault signals of rolling bearings embody strong nonlinearity and nonstationarity, resulting in the difficulty in extracting weak fault features. Therefore, a method based on grey wolf optimization algorithm and variable-scale stochastic resonance is proposed in this paper. First, the varied-scale stochastic resonance method is employed to overcome the constraints of the small-parameter condition in traditional stochastic resonance, which converts the high-frequency vibration signals of actual bearing faults into low-frequency signals through a secondary transformation, thereby meeting the small-parameter requirement of stochastic resonance. Meanwhile, the grey wolf optimization algorithm is applied to optimize the potential function parameters of the stochastic resonance system, further improving the adaptability of VSAR model. Analysis of simulated signals and vibration signals from outer ring and cage faults of rolling bearings demonstrate that the proposed method can effectively extract fault signal features. Compared with traditional parameter-adjusted stochastic resonance, the GWO and varied scale stochastic resonance exhibits superior performance in fault features extraction, with their amplitudes being doubled. Therefore, this method is suitable to extract weak bearing faults of rolling bearings.