Degradation Modeling And Remaining Useful Life Prediction Of Railway Wagon Brake Shoes Based On Sparse Data

Performance degradation process is ubiquitous in parts.Modeling the performance degradation process and predicting the remaining useful life(RUL)is one of the important research topics in the field of maintenance and repair.In the life cycle of brake shoes,due to the low monitoring frequency,the degradation data collected are sparse degradation data.For the degradation process of brake shoes and other parts with sparse degradation data,reliability modeling and RUL prediction are of great theoretical and practical significance.Therefore,the degradation modeling and RUL prediction of sparse degradation data are studied in this thesis.Firstly,a Wiener process model with random effects is established for sparse degradation data.The traditional maximum likelihood estimation method can not get the unbiased estimators of the variance and diffusion coefficient of the offset coefficient,which may be more prominent in the sparse data.In this thesis,the parameters of Wiener process model with random effects are estimated by the restrictive maximum likelihood method,which is widely used in mixed effect models.The results show that in the case of sparse degenerate data,the results of the restricted maximum likelihood estimation method are better,and it has the property of asymptotic unbiased estimator.Finally,the method is applied to solve the parameters of brake shoes degradation model of railway wagon.Secondly,degradation model parameter update and RUL prediction are carried out for sparse degradation data.The model parameters were updated based on Bayesian theory,and the probability density function of component RUL was derived based on the first arrival time,and the dynamic prediction of component RUL was carried out under the Bayesian framework.Through simulation analysis,the RUL prediction results of parts with parameter updating and parts without parameter updating were compared.The results show that the RUL prediction results of parts with parameter updating are better than those without parameter updating.In addition,the monitoring timing of sparse degradation data is optimized,and the results show that setting the detection point at the later stage of component operation can effectively improve the accuracy of RUL prediction.At last.The component RUL prediction method based on Bayesian updating framework is applied to RUL prediction of brake shoes of railway wagon.The results show that the proposed method can predict RUL of brake shoes of railway wagon well.The results of this study are helpful to extend the application research of Wiener process model in sparse degradation data,and to conduct degradation modeling and RUL prediction of brake shoes and other parts with sparse degradation data,thus effectively improving the health management level of these parts.