Bearing Fault Diagnosis Based on Unsaturated Piecewise Non-linear Bistable Stochastic Resonance under Trichotomous Noise

The classical bistable stochastic resonance (CBSR) has the disadvantage of output saturation, which limits the enhancement capability for weak signal detection. To break the limitation of output saturation, a novel unsaturated piecewise non-linear bistable stochastic resonance (PNBSR) method is proposed. Because the trichotomous noise exists in practical engineering, the PNBSR under trichotomous noise is explored in this paper. The performance of PNBSR is evaluated by the index, i.e., the mean of M times signal-to-noise ratio increase (MSNRI). The double-peak phenomenon of SR is observed under trichotomous noise. Experiments reveal that the proposed PNBSR method performs best on extracting characteristic components from a strong noise background, compared with the CBSR method and the traditional digital filter. Then, the PNBSR is applied to the fault diagnosis of rolling element bearings. The paper focuses on solving practical engineering problems with mathematical methods.