The Research And Application Of Weak Signal Detection Based On Stochastic Resonance

In a nonlinear system, the periodic or useful weak signal of system output can be enhanced by utilizing the background noise. This phenomenon of converting the nosie energy into the periodic signal is called stochastic resonance. In this thesis, bistable system which was researched widely is considered as the research object, and the following tasks are completed mainly:Firstly, the theoretical basis of stochastic resonance is studied deeply and several existing measurement indexs of stochastic resonance are analyzed in detail. Meanwhile, an adaptive stochastic resonance weak signal detection method based on symbol sequence entropy is proposed in this thesis.Secondly, the traditional stochastic resonance detection method is restricted by the adiabatic approximation theory and can not detect the large parameter signal. So several large frequencies signal detection method used commonly are analyzed in this thesis and the analog signals are detected effectively in simulation experiments.Finally, for the strong background noise and modulation characteristics of rolling bearing vibration signal, the traditional extraction methods based on linear theory are useless, and the modulation signal can be decomposed into different scale frequency bands by using EEMD method based on nonlinear theory. At the same time, the de-noising methods based on linear theory can lead to damage the feature signals more or less, so the stochastic resonance method is used in this thesis. However, in a number of IMFs of EEMD decomposition there are many IMFs to detect the rolling bearings features useless. Meanwhile, the traditional singleparameter optimization method of adaptive stochastic resonance can have the ineffective problem. Bearing fault diagnosis method based on adaptive stochastic resonance of EEMD sensitive IMFs is proposed. The weighted kurtosis index can not only maintain the similarity between the output signal and original signal, but also be sensitive to shocks feature and overcome the missed or false detection of the traditional kurtosis index, and it is treated as the food consistence function of AFSA to optimize the system parameters. Then this proposed method is demonstrated effectively by using the experimental data of rolling bearings data center at Case Western Reserve University in America.