Rail breaks arrival rate prediction: A physics-informed data-driven analysis for railway tracks

Abstract

Rail breaks frequently cause service interruptions and are a leading cause of train derailments. In this study a new method for predicting the arrival rate of rail breaks is developed. The proposed model incorporates Finite Element Modeling, into an Approximate Bayesian Computation (ABC) framework to predict the arrival rate of rail breaks in a US Class I railroads. The developed model is an improved version of the prediction model earlier proposed by Ghofrani et al. in [1]. The results of the proposed model show that the arrival rate of rail breaks is majorly associated with higher speed, weight of rail, geometry defects, inspection and grinding frequency. Moreover, the findings imply that inducing the mechanics of rail defects into a data-driven model outperforms the traditional pure data-driven models by over 25%. The proposed model also outperforms the earlier version introduced in [1] by 5%.