A Novel Exponential Unsaturated Bistable Stochastic Resonance-Boosted Incipient Fault Identification in Rotating Machineries

Abstract Stochastic resonance (SR) for weak fault detection stands as a significant constructive methodology leveraging noise in nonlinear information systems processing. In virtue of the SR technique in conjunction with coupled non-saturated nonlinear systems, an exponential unsaturated bistable stochastic resonance (EUBSR) model is developed to enhance output levels. By integrating the exponential monostable stochastic resonance system (ESR) and the unsaturated bistable stochastic resonance (UBSR) system through coupling coefficients, this model offers a broader spectrum of resonance characteristics. The performance of the EUBSR is evaluated based on the relevant indicators signal-to-noise ratio (SNR) and residence time distribution ratio. These indicators are treated as multi-objective functions, with the coati optimization algorithm employed to optimize both the parameters and coupling coefficients of the EUBSR model simultaneously. Moreover, the paper takes into account the interdependence of nonlinear systems and their interactions by considering both cascade and parallel models of the ESR and UBSR systems. Fault diagnosis is carried out on simulation signals and bearings to validate the effectiveness of the proposed EUBSR model. The results demonstrate that the EUBSR model surpasses not only its individual component models but also cascade and parallel models.