Research On The Early Fault Diagnosis Method Of Rolling Bearing Based On Vibration Signal

Driven by the national modernization and improvement of industry technology,rotating machinery had been widely used in every industrial field,such as aerospace,marine,railway,automotive,machine tools,etc.Bearing is a key component in rotating machinery,whose health condition is critical to the safety and function for the whole machinery system.To improvement the equipment safety and avoid economic loss as well as preventing accident occurring,this paper focused on the rolling bearing to introduce its fault characteristic and summaried the diagnosis method for it.The main contents of this paper were as follows:Since the fault characteristic were usually submerged in the strong background noise from the environment and other working components,stochastic resonance?SR?was introduced to enhance the weak signal.By studying the effect of the parameters on the potential function,we found that one of them had a greater effect on the height of the potential well and the other had a larger effect on the distance of the two stable solution of the potential.Since the output of the system was solved by four-order Runge-Kutta methods,the length of the solution step determine the output directly.In order to enhace the fault characteristic as large as possible,this paper studied the output of the system with different inputs.Then,the potential function parameters and the length of solution step were combined by the Kramers ratio of noise.Then whale optimization algorithm was applied to determine them to obtain the best output.The results further revealed the relationship between the target frequency and the solution step length.Even though the fault characteristic can be discovered by the method proposed in section 2,it may fail when the target fault frequency is unknow or not accurate enough.To overcome this drawback,empirical mode decomposition?EMD?and ensemble empirical mode decomposition?EEMD?were applied in section 3.This paper studied the effect of local extremal in the envelope process and found that the upper limitation frequency of the noise added had a direct effect on the decomposition result.In order to select an optimal upper limitation frequency for the noise added,an index merged by kurtosis,correlation coefficient,maximum of autocorrelation coefficient and relative root mean square error,was used to evaluate to decomposition result with different noise upper limitation frequency.Based on the result,a novel adaptive EEMD was porposed in section 3.The results compared with traditional EEMD heighted the superiority of the proposed method.In section 3,EMD and EEMD were applied to process the vibration signal successfully.However,both the methods lack rigorous meathermatical theory.Variational mode decomposition?VMD?,a novel signal process method supported by variational frame method solidly,was applied to detect rolling bearing fault characteristic in section 4.In this part,fractional noise was used to study the equivalent filter characteristic of VMD.And Gauss white noise was applied to study the effect of penalty factor and mode number on the decomposition performance.The results demonstrated penalty determined the bandwidth of the filter mainly and the mode number effect both the center frequency and the bandwidth simultaneously.Base on these results,a dual VMD method was proposed in section 4 to diagnose bearing fault.In the dual VMD system,the VMD in layer one was applied to make sure the fault characteristic signal was included in the interesting mode and the target of second layer was to find out the best penalty factor for VMD.Thus,all parameters can be determined based on the system-self.The results of simulation and experiment signal displayed the effective of the proposed method and the results compared with some other methods highlighted the superior of the proposed method.As introduction above,three methods were applied to detect rolling bearing fault.However,all of them stayed at the stage of qualitative judgement.Quantitative estimation about the fault size will be much more significance for machinery maintenance in practical.In order to make an estimation about the degree of damage for rolling bearing,sparse representation was introduced in section 5.In this section,some commom atoms,like Morlet wavelt atom,Gabor atom and traditional impact atom,were compared with the vibiration response of the fault rolling bearing.By the result,a double-impact atom was proposed in this paper to describle the fault characteristic more accurately.Then an step atom combined with the proposed atom was proposed to make a estimation about the fault size by orthogonal matching pursuit?OMP?.In order to decrease the estimation error,an optimal function named weighted L₂-norm was proposed in this paper.The result demonstrated the result from the proposed function was more accurate.Based on the introduction above,it is clear to see the main work of this paper focused on the signal processing method in the field of rolling bearing fault diagnosis.Some useful methods were proposed and examined by simulation and test signal.Therefore,the work of this paper is meaningful in theory and engineering.