A Comparative Study of Optimization Models for Condition-Based Maintenance Scheduling of an Aircraft Fleet

Condition-based maintenance (CBM) scheduling of an aircraft fleet in a disruptive environment while considering health prognostics for set systems is very complex combinatorial problem, which becoming more challenging light the uncertainty included prognostics. This type problem falls under broad category resource-constrained problems and often solved using mixed integer linear programming (MILP) formulation. While MILP framework promising, size can scale exponentially with number considered tasks, leading to significantly high computational costs. The most recent advances artificial intelligence have demonstrated capability deep reinforcement learning (DRL) algorithms alleviate this curse dimensionality, as once DRL agent trained, it achieve real-time optimization schedule. However, there no guarantee optimality. These comparative merits formulation not been discussed literature. study response research gap. We conduct comparison model, are used derive optimal schedule various scenarios fleets different sizes environment, available resources execution each task. quality solutions evaluated on basis four planning objectives, defined according real airline practice. results show that approach achieves better respect prognostics-driven tasks requires less time, whereas model produces stable schedules induces ground time. Overall, provides valuable insights integration