Research On Prediction Methods For The Remaining Useful Life Of Gearbox Based On Data-driven

With the energy crisis and environmental pollution becoming more and more serious,wind power,as a kind of green energy,has been developed rapidly.As a key component of wind turbine,the working performance of gearbox has a direct impact on the life of wind turbines.Therefore,it is a great significance for wind turbines to predict the remaining useful life of gearbox and then make a reasonable and effective maintenance scheme.Among them,the remaining useful life prediction of gearbox is not only the key,but also the difficulty.The accurate prediction of the remaining useful life of the gearbox can improve the reliability and increase the remaining useful life of the wind turbine.Because of the complex fault mechanism and high manufacturing cost of gearbox,it is difficult to carry out a large number of life tests.Therefore,the prediction method that based on physical model is difficult to apply.However,during the operation of the gearbox,a large number of state degradation data that containing the life information of the gearbox are produced.Therefore,the datadriven prediction method is widely used to predict the remaining useful life of gearbox.In this paper,the degradation state of the gearbox is characterized by the vibration data obtained from the fatigue life test of the gears in the gearbox,and the remaining useful life of the gearbox is predicted according to the fault threshold based on combining kernel density estimation and stochastic filtering theory prediction method.The main work and research results are as follows:(1)A remaining useful life prediction method under single degenerate variable is proposed.Firstly,on the basis of analyzing the affected factors of gearbox life,the vibration acceleration signal and noise signal of gears are selected as the degradation state data of the gearbox,and then the remaining useful life prediction model under single degenerate variable is established by combining kernel density estimation and stochastic filtering theory.The method uses the kernel density estimation method based on the data itself to estimate the probability density function of the continuous degenerate state of gears,and gets the probability density function of the degenerate state of the gears,and then uses the state monitoring data to update the parameters of the random filter recursive model,so as to predict the remaining useful life of the gearbox.The validity of the method is verified by the test of the gear.(2)A method for predicting remaining useful life under multiple degenerate variables based on copula function is proposed.The method uses the Copula function to represent the stochastic correlation between the vibration acceleration and the noise of the gears,and the joint distribution function of the remaining useful life of the gears is obtained by the edge distribution function of the remaining useful life under each single degenerate variable,so as to obtain the forecast value of the remaining useful life of the gearbox.The validity and feasibility of the method are also verified.