Predictive maintenance planning for railway systems.

The ubiquity of data-gathering sensors and the recent advances in high-performance computing technologies make available and possible the intensive use of data to enhance equipment maintenance strategies. Based on real-time data, these strategies include anomaly detection and equipment health/degradation level assessment (including monitoring and prediction), to better support the planning of maintenance activities. Hence, consider the probabilistic predictions of degradation levels of the components of a railway system available. The current paper addresses a maintenance planning problem that optimizes related costs and satisfies reliability requirements while considering the probabilistic nature of degradation levels of railway components. In this sense, a separate chance-constrained stochastic programming model is introduced, and probabilistic parameters are approximated via a Monte Carlo technique. Numerical experiments are conducted on a set of randomly generated instances and a number of key performance indicators are analyzed.