Study on Feature Extraction Method of Fault Signal Based on Reinforcement Cascaded Multi-stable Stochastic Resonance System

Abstract

The paper presents a new method of fault signal feature extraction based on reinforcement cascaded multi-stable stochastic resonance (SR) system and it can extract signal fault feature from heavy background noise. The multi-stable model can further increase the noise utilization to achieve better detection effects of SR for weak signals than the bistable stochastic resonance system. On the basis of cascaded multi-stable stochastic resonance theory, the reinforcement characteristic of stochastic resonance induced by adding the second driven periodic signal is studied. Adding the second driving signals to the first and the second cascaded system, when the time scale is matched with the stochastic fluctuations, the phenomenon of frequency absorption happens and the effect of stochastic resonance in multi-stable system is reinforced. At last, an example of rolling bearing fault signal confirms that this method can efficiently extract weak fault signal feature, and has a good prospect in signal detection fields