Research On Weak Signal Detection Of Rolling Bearings Based On Stochastic Resonance

Stochastic resonance (SR) is a nonlinear phenomenon, and it occurs only whenthe three of noise, bistable system and input signal, achieve certain synergies. Specificfor a certain noisy signal, it is very necessary to joint adjustment of multipleparameters of stochastic resonance to produce SR phenomenon. However, there is lessresearch about the coupling of system parameters, signal frequency and noiseintensity, the joint parameter adjustment of stochastic resonance lack of guidance, andhighly rely on the experience. Therefore, this paper focuses on the problem of hard torealize multi-parameter joint adjustment of stochastic resonance, and adaptivestochastic resonance is researched.The traditional adaptive stochastic resonance can only achieve one-parameteroptimization, and it is very difficult to select the calculation step of step-changedstochastic resonance (SCSR), a new adaptive SCSR based on particle swarmoptimization (PSO), which can realize the adaptive solving of optimal output of SCSR,is proposed in this paper. The output signal to noise ratio of bi-stable system isdetermined as the fitness function of PSO algorithm, and the structure parameters andcalculation step of SCSR are selected adaptively, as a result, the weak signal under thecondition of large parameters can be detected optimally. The proposed method isapplied to simulation data and vibration signals measured on defective bearings withinner race fault. The results show that the proposed method has advantages ofsimplicity, fast convergence speed and wide range of applications, and possesses agood prospect of engineering application.Adaptive step-changed stochastic resonance using particle swarm optimization isintroduced to cascaded stochastic resonance, and adaptive de-noising method basedon cascaded stochastic resonance is proposed in this paper. By adaptively solvingparameters of each stochastic resonance system, each stochastic resonance system canadaptively output the de-noised signal, and eventually reach the purposes of noisereduction for engineering signal under large parameters condition. This method isapplied to the simulation data and engineering data, analysis results show that themethod can effectively eliminate the high frequency noise of noisy signals with large parameters, highlight the low-frequency useful signal components, and the effect ofnoise reduction is remarkable.Adaptive de-noising method based on cascaded stochastic resonance and EMDare combined together in this paper, and the EMD method based on adaptivede-noising by cascaded stochastic resonance is proposed. In experimental studies, theoriginal input signal and all levels of output signals of bistable system with EMD areprocessed, respectively. After comparative analysis of the results of EMD, we candiscover that, this method can not only improve the signal-to-noise ratio, improve thequality of the EMD, but also can reduce the number of IMFs, improve the operationefficiency.