Adaptive frequency-shifted and re-scaling stochastic resonance with applications to fault diagnosis

The difficulty for selecting the best system parameters restricts engineering applications to frequency-shifted and re-scaling stochastic resonance (FRSR). An adaptive FRSR method with time-domain and frequency-domain indexes is developed, where the station of the highest spectral peak and the variance of the zero-crossing distance are chosen as the objective functions, and the optimal parameters are obtained adaptively without predicting the exact frequency of the target signal. Once FRSR is involved, the proposed method is able to detect high frequency, while the traditional stochastic resonance is only for low frequency. The simulation and engineering application on fault diagnosis indicate the ability for weak periodic signals buried in heavy noise.