Abstract:
We introduce the problem of vehicle maintenance scheduling under service level agreement (SLA) and preventive maintenance cycles at a single dead-end track. We show that the problem is NP-Hard. We build MILP model to solve the problem and propose formulation improvements based on problem structure. Besides, we develop a heuristic that generates an initial feasible solution to the MILP solver. As a result of computational experiments, we show that improved model, which is a combination of formulation improvements, CPLEX parameter fine-tuning and the heuristic drastically heightens the solution quality compared to the MILP model under given time limits. We select the improved model as a solution method. We create a discrete-event sim ulation environment to determine effects of problem parameters on key performance indicators (KPI). We build two alternative methodologies, as corrective jobs worsen tardiness related KPIs of preventive jobs. One of them is the buffer method which reschedules the result of the solution method, the other one is anticipation method in which we alter objective function of the solution method to favor earliness in the model. In conclusion, these methods diminish tardiness related KPIs of preventive jobs but they increase preventive earliness compared to the solution method.
