Adaptive stochastic resonance based on genetic algorithm with applications in weak signal detection

One deficiency of the traditional adaptive stochastic resonance is that only a single parameter can be optimized while the other parameters in the system being fixed. A new adaptive stochastic resonance based on genetic algorithm, which realizes multi-parameter synchronous optimization, is proposed. The signal-noise-ratio of the output of the bi-stable system is determined as the fitness function of genetic algorithm and multi-parameters in stochastic resonance system are selected adaptively. As a result, weak periodical components in original signals are sufficiently amplified. Simultaneously, the optimization algorithm, combined with frequency-shifted and re-scaling stochastic resonance, enables to achieve the stochastic resonance under the conditions of great parameters. The proposed method is evaluated by simulation data and vibration signals measured on defective bearings with outer race fault. The results show that weak periodical components with high frequency buried in strong noise are well extracted in case of small number of sample points.