Rolling Bearing Fault Diagnosis Based On Stochastic Resonance And Fourier Decomposition

Rolling bearing is a key component of mechanical equipment.It is of great significance to ensure the rotating mechanical equipment operating safely and stably.When the bearing is damaged,which will lead to equipment break and then reduce the production efficiency of the enterprise,and even threaten the safety of operators.Therefore,it is vital to find an effective method of bearing fault diagnosis.Firstly,this article studies the classical stochastic resonance(SR)theory,and uses the method of frequency-shifted and re-scaling(FR)to eliminate the limitation that the stochastic resonance can only analyze the small signal(signal amplitude,frequency,and noise intensity are all less than 1),at the same time introduces the Beetle Antennae Search(BAS)algorithm to adaptively optimize the parameters of the stochastic resonance system,and applies them in the process of extracting the simulation signal and the bearing fault characteristic signal.Secondly,it is found that the output of stochastic resonance largely depends on the selection of the potential well force which is concluded by researching the classical stochastic resonance potential function model.A suitable potential well force will enhance the resonance effectively,on the contrary,will destroy the resonance.It is found that the potential well force is the first derivative of the potential function.Therefore,this paper strengthens the random resonance by the method of adjusting the potential function.Specifically,the two mono-stable potential functions,the Power Function and the WoodsSaxon function,are combined to construct a new piecewise nonlinear function which is called PWS stochastic resonance model,and PWS potential model provides rich potential well.The comparison shows that: the output signal-to-noise ratio(SNR)and the characteristic frequency amplitude obtained by the PWS stochastic resonance model are better than the classic stochastic resonance model.The PWS stochastic Resonance provides an effective way to enhance target signal for bearing fault diagnosis.Then,this paper analyzes the difference between Fourier Decomposition Method(FDM)and Empirical Mode Decomposition(EMD)in processing non-stationary nonlinear signals.The results show that FDM reduces the effects of end effects and modal mixing,and the obtained signal is more accurate which provides subsequent signal analysis and processing.Finally,aiming at solving the problem that it is difficult to extract weak bearing signal in the strong noise environment,this paper proposes a method that combines FDM and PWS stochastic resonance to extract the weak signal.It is takes the true vibration signal of the wind turbine main bearing as the input signal,and the signal is decomposed into a series of component signals,then several component signals are selected for reconstruction by the principle of maximum correlation.At last the reconstructed signal is subjected to resonance processing by using the PWS stochastic resonance system and obtained the main bearing rotation frequency which is hidden in the decomposed signal.By comparing the obtained results with actual values of the rotation frequency,it is proved that the combination of FDM and PWS stochastic resonance can effectively detect the weak signal hidden in the noise,which provides an effective method for detecting the signal of rotating machinery under the background of strong noise.