Statistical modeling of the distribution characteristics of high-speed railway disruptions

Abstract

Studies on the spatiotemporal distribution and duration characteristics of railway disruptions are very significant for the advanced prediction of disruption and development of real-time dispatch strategies. In this study, historical disruption records of some Chinese High-Speed Railways (HSRs) lines from 2014–2016 were used to investigate the distribution characteristics of railway disruptions. The spatiotemporal probability distribution of four railway lines were calculated and their hotspots (coordinates with high probabilities) and coldspots (coordinates with low probabilities) were revealed using heatmaps. Furthermore, all the disruptions were classified into seven clusters based on their causes, and statistical analysis was carried out on each cluster. In addition, three right-skewed distribution models, namely Log-normal, Weibull, and Gamma distributions, were used to fit the duration of each cluster to uncover its duration regularities. Finally, goodness-of-fit test was performed on the models using the Kolmogorov-Smirnov method, indicating that the duration of each classified disruption can be estimated using a Log-normal distribution function. The obtained spatiotemporal probabilities and duration time distribution models thus can be further applied into estimating the occurrence and duration of railway disruption in real-time dispatching to help dispatchers make advanced decisions.