Statistical modelling of delays in a rail freight transportation network

This study analyzes the transportation network of a major rail freight operator in order to obtain a model of delay propagation of trains connecting intermodal terminals. Operational management of a rail freight operator needs to take into account deviations due to unexpected events such as unplanned maintenance, strikes, railroad works, traffic congestion. The dispatcher makes train assignment decisions based on a number of performance indicators and also on the expectancy that a given train, currently delayed, could recover or limit the amount of delay in the future. We have developed a Markov-chain based model in order to evaluate the evolution of train delays as a train visits successive terminals. Our model is based on the examination of a large set of historical data and we show how we can classify different terminals according to their ability either to absorb or to amplify delays.