Analysis of a class of Lorenz-like stochastic resonance model and its application in induction motor fault diagnosis

Stochastic resonance (SR) is a method to enhance weak characteristic signals, but the classical model like Duffing oscillation is limited by fewer parameters when it is applied to detect different signals. Consider chaotic system with the characteristics of adjustable parameters and wide dimensions, from the perspective of chaos to induce SR, which is studied to discover the conversion conditions between chaos and SR system. Firstly, the mechanism of a class of Lorenz-like SR system for SR generation is studied, whose parameters relationship meeting the moving point transition and parameters range satisfying the maximum Lyapunov exponent (MLE) are analyzed. Secondly, within the determined parameters range, the optimal parameters are searched by particle swarm optimization (PSO) to output signal with optimal signal-to-noise ratio (SNR). Finally, the output fault signal after highlighting the characteristics is diagnosed by utilizing extreme learning machine (ELM). The experimental results show that the fault signal is detected and diagnosed efficiently and accurately.