Probabilistic Risk Analysis of Broken Rail-Caused Train Derailments

Abstract

Abstract Broken-rail prevention and risk management have been being a major activity for a long time for the railroad industry. The major objective of this research is to evaluate and analyze the broken rail-caused derailment risk using Artificial Intelligence (AI) approaches. The risk model is primarily built upon 1) broken rail probability; 2) probability of broken-rail derailment given a broken rail; and 3) derailment severity, measured by the number of cars derailed. The train derailment risk accounts for derailment probability and derailment consequences simultaneously. Due to the low frequency of broken-rail derailments, it is desirable to estimate the probability of broken rail-caused derailments through the broken rail occurrence. The estimation of the probability of broken rail-caused derailment includes the conditional probability of derailment given broken rail occurrence and the probability of broken rail occurrence. More specially, the probability of broken-rail derailment given a broken rail can be estimated by the statistical relationship between broken-rail derailment and broken rail, given specific variables (e.g., track curvature, signal condition, and annual traffic). The probability of broken rails can be estimated using machine learning techniques based on railroad big data, including maintenance, track layout, traffic and historical inspection records. In terms of derailment consequence, it is defined as the number of cars (both loaded and empty) derailed per derailment that would be estimated based on potentially affecting factors, such as train length, train speed, and train tonnage. The quantitative estimation and analysis of broken rail-caused derailments are based upon the historical records from one Class I railroad company from 2012 to 2016, covering over 20,000 track miles on mainlines. The developed integrated risk model is able to contribute to the prediction of location-centric broken rail-caused derailment risk. Ultimately, the identification of high-risk locations can ultimately aid the railroads to mitigate broken rail risk in a cost-efficient manner and improve railroad safety.