Abstract
In the field of bearing fault diagnosis, the phenomenon of stochastic resonance (SR) has been proven to effectively utilize noise to enhance weak features of early faults. The classical bistable stochastic resonance (CBSR) model, as one of the most widely applied SR methods, faces limitations in feature enhancement due to the complexity of parameter tuning and the issue of output saturation. To address these issues, this paper proposes an improved piecewise unsaturated bistable stochastic resonance (PUBSR) method, which employs an asymmetric potential function to effectively mitigate the output saturation problem of CBSR. Additionally, the cuckoo search (CS) algorithm is used to optimize the potential function parameters, enhancing fault diagnosis performance. Finally, the proposed method is applied to both simulated signals and early bearing fault engineering data. The results demonstrate that compared to the CBSR method, the proposed approach more than doubles the spectral peak value when extracting characteristic frequencies, significantly improving the identifiability of fault features and diagnostic accuracy.