Discrete Event Simulation and Optimization Approaches for the Predictive Maintenance of Railway Infrastructure

This thesis is carried out within the PhD Course in Logistics and Transport at CIELI - Italian Centre of Excellence on Logistics, Transport and Infrastructures, University of Genoa. In this work, a discrete event simulation and optimization model is created to schedule the predictive maintenance activities. Nowadays, after a severe decrease of transport demand during the pandemic period, rail public transport is resuming a central role for both freight and passenger transport. To cope with this increase in demand, to maintain high safety standards and to avoid unnecessary costs, the idea is to switch to predictive maintenance strategy, intervening before an asset failure and when it has reached a certain state of degradation. The degradation and asset future conditions are predicted according to probabilistic models and maintenance deadlines are defined by applying a risk based approach. The problem is first formulated as a MILP (Mixed Integer Linear Programming) optimization problem and then transformed into a simulation-based optimization problem using the ExtendSim software. Different simulative models are created to take into account the stochastic nature of some variables in real processes. After the formal description of the models, some real-world applications are presented. Finally, considerations on the proposed approach are reported highlighting limits and challenges in predictive maintenance planning, such as lack of data and the stochastic and complex environment.