Delayed maintenance modelling with speed restriction for railway section

International audienceOn a deteriorated track, speed restriction decelerates the deterioration evolution, allowing the track to have a greater chance to survive for a longer repair delay time. However, if a slower speed increases the lifetime of the track component and thus authorizes a delayed maintenance, it also reduces the throughput of the system and hence the benefit. On the other hand, a longer preventive maintenance delay time gives more time to better organize and plan the maintenance at a reduced cost, and more opportunities to optimally group the maintenance activities on different track components to save maintenance setup costs. We have thus a maintenance optimization problem to solve, and, in this context, the objective of this work is to develop a maintenance cost model to determine the optimal tuning of the two considered decision variables: maintenance delay time and speed restriction.To this aim, we want to assess the effect of speed restriction on the delayed maintenance cost and system profit for a railway section, in which the deterioration of the track component depends on the train speed and the number of passing trains. The problem is to determine an optimal speed restriction and a preventive repair delay to maximize the system benefit (i.e. system profit minus maintenance costs) and minimize the system downtime.Coloured Petri Nets (CPN) are used to model maintenance and operation of the railway section. The CPN model describes the gradual track deterioration from the good to the failed state using a stochastic process. Two levels of maintenance are implemented according to the observed track component states, which are identified by periodic inspection. Different speed restriction policies and maintenance delaying strategies are considered and implemented. Monte Carlo simulations are carried out to evaluate the performance of the railway section under these different policies, which is measured by maintenance cost, the system benefit and the system downtime. Numerical experiments results illustrate how the maintenance decision variables can be optimally tuned