Heuristic based track maintenance scheduling optimisation

The maintenance of railway track is highly relevant for the service quality and safety. This task demands huge technical and human resources. Therefore an optimal deployment of the available resources is crusial for the cost-effectiveness of railway operation. Furthermore, the track geometry maintenance activities need to be planned up to two years in advance. For this reason the French railway operator SNCF measures periodically the geometrical characteristics of its high speed network since its commissioning, i.e. for more than 30 years now. Maintenance planning is based on forecastings of the development of the geometry irregularities in each track sector. In order to achieve good forecasting quality and reliability, this paper presents a method for making middle term (i.e. up to one year) predictions of the railway geometry ageing process. The method relies on a process model describing the track geometry ageing as an exponential function. The function identification is achieved with the help of the Levenberg-Marquardt procedure for sum of squares minimisation of non-linear functions. One of the most important maintenance activities, which is basically achieved by means of tamping and grinding, is the reduction of the track geometry irregularities. Currently, most the planning of tamping interventions is made manually, basing on expert knowledge and analysis of the geometry degradation. In this paper we present a method for automatic optimal planning and scheduling of tamping activities on railway tracks. Implementing the method would entail a more efficient use of the available maintenance resources, leading to an overall improvement of the track geometry condition. The maintenance scheduling is calculated by means of an in-house developed optimisation algorithm. Its main novel aspect is the implementation of a heuristic for the problem resolution instead of commercial optimization packages. This results in a twofold benefit, as both the need of model simplification and the calculation times are dramatically reduced. The method is finally validated by means of a case study on real data of a French TGV high speed line.