Statistical Analysis of Seasonal Effect on Freight Train Derailments

Abstract

Freight train accidents can damage infrastructure and rolling stock, disrupt operations, and possibly cause casualties and harm the environment. Understanding accident risks associated with major accident causes is an important step toward developing and prioritizing effective accident prevention strategies. This paper developed a negative binomial regression model to estimate freight-train derailment frequency on Class I railroad mainlines, accounting for derailment accident cause, traffic exposure, railroad, and season. The primary focus is to quantitatively measure the seasonal effect on freight-train derailment frequencies given traffic exposure. For model illustration, the analysis focused on three common derailment causes on freight railroads: broken rails, broken wheels, and track buckling, using the empirical Federal Railroad Administration (FRA)-reportable freight railroad derailment data on mainlines gathered between 2000 and 2016. The modeling results show that it tends to have high derailment rates in winter due to broken rails and broken wheels (double that of summer), whereas summer has the highest likelihood of buckling-caused derailment of all of the seasons (e.g., 6 times that of spring and 10 times that of fall). These analytical results can contribute to the risk-based optimization of rail and wheel inspection frequency. The statistical modeling methodology developed in this paper can be adapted to other types of train accidents or accident causes, ultimately supporting the optimal allocation of train safety improvement resources.