Uncertainty propagation in rail wear prediction using an analytical method and field observations

Abstract

Rail wear management based on accurate rail wear prediction is essential for railway maintenance. The implementation of rail wear prediction models in maintenance decision tools is not yet available due to detailed modelling and the absence of direct coupling with operational conditions. A method that does not provide any confidence interval on the prediction is not very helpful if one wants to use the results of the prediction for maintenance decision-making and there is variation in the input. Therefore, in this study a wear prediction model that does take into account these limitations is used to predict the amount of rail wear with certain confidence bounds. The uncertainty in the output of the model is quantified. This is realized by considering probability distribution functions for the input parameters and analytical analyses. The results obtained from these analyses are then compared with field measurements and a good agreement is found.