Research On Prediction Of Bearing Remaining Life In Stages Based On Wiener Process

In the field of petroleum and other industrial machinery and equipment,bearing is a very important part,its durability and reliability affect the health of mechanical equipment.Therefore,how to accurately predict the remaining service life of bearings is an urgent and challenging problem.If the predicted value of bearing life is too large,the uncontrollable failure of bearing will occur at any time,which will lead to serious failure of mechanical equipment and cause great safety problems;If the predicted value is too small,it will lead to premature replacement and increase maintenance costs.The running environment of bearing is very complex,which makes it difficult to predict the bearing life.Therefore,how to process the bearing data,establish the degradation model fitting the actual degradation trend and improve the accuracy of life prediction has become a hot topic in this field.In this paper,the traditional self encoder data fusion method and the traditional Wiener stochastic process modeling method are improved to achieve higher residual life prediction accuracy.The main research work is as follows:Firstly,in the paper proposes a data fusion method combining Spear man correlation coefficient with an improved stack sparse auto encoder.In this method,the Spear man correlation coefficient method is used to screen the data that is the most representative of the degraded performance trend among the mufti-dimensional data containing redundant information,and the redundant information is eliminated.Aiming at the complex nonlinear relationship between bearing data,a nonlinear fusion method is introduced,that is,the improved stack sparse self coding is used to fuse the selected data to extract one-dimensional Abstract comprehensive characteristic index,which provides data support for bearing life prediction.Correlation entropy(MCC)is introduced as the error cost function in the traditional self encoder network to eliminate the influence of non Gaussian noise of bearing data on data fusion effect.Secondly,in order to improve the accuracy of residual life prediction,a modeling method of adaptive Wiener stochastic process is proposed.First of all,considering that the Wiener process has the nature of a first-order Markov process,that is,the predicted remaining life is only related to the degradation state at the current moment,and does not consider the impact of historical data on the degradation process.Therefore,this article expands the degradation drift parameters in the traditional Wiener process to the degradation drift state,and adapt updates them based on historical data;Subsequently,in view of the possible measurement noise problems caused by the actual bearing’s harsh operating environment or sensor accuracy,a random variable that obeys the normal distribution is introduced to represent the measurement noise,so as to eliminate the influence of measurement noise on the prediction accuracy;Finally,considering the staged characteristics of the actual bearing degradation process,a staged degradation model is established,and the above two improvements are incorporated into each degradation stage to make it more suitable for the actual degradation situation.Finally,the improved method is applied to the prediction of bearing residual life,and the prediction process is simulated.The stability and reliability of the prediction method are illustrated by the results of state and parameter estimation,and the effectiveness of the prediction method is illustrated by comparing with the relative error of its classical prediction method.