Robust Maintenance Scheduling of Aircraft Fleet: A Hybrid Simulation-Optimization Approach

Abstract

We study the maintenance task scheduling problem for an aircraft fleet in an uncertain environment from the viewpoint of robust optimization. Given a daily horizon, the maintenance tasks delegated to a shop should be scheduled in such a way that sufficient aircrafts are available on time to meet the demand of planned missions. The tasks are either scheduled maintenance activities or unexpected repair jobs when a major fault is detected during pre-or after-flight check of each mission. The availability of skilled labour in the shop is the main constraint. We propose a robust formulation so that the maintenance tasks duration is subject to unstructured uncertainty due to the environmental and human factors. As a result of the specific structure of the primary model and non-convexity of the feasible space, the classical robust optimization methods cannot be applied. Thus, we propose an ϵ-Conservative model in tandem with Monte-Carlo sampling to extract the set of all feasible solutions corresponding to various disturbance vectors. Since the one-way sampling-thenoptimization approach does not guarantee the probabilistic feasibility, we employ a hybrid simulationoptimization approach to ensure that the solutions provided by the ϵ-Conservative model are robust to all uncertainty scenarios. The experimental results confirm the scalability of the proposed methodology by generating the robust optimal solutions, satisfying all conservatism levels and uncertainty scenarios irrespective of the problem size.

Robust Dry Bulk Fleet Route Optimization Under Navigation Risk Consideration

“…Robust optimization is an appealing approach to address the problem with uncertain parameters that are only known to belong to some uncertain set (Zhao et al [21]; Liang et al [22]; Shahmoradi-Moghadam et al [23]). Its objective is to optimize the worst-case objective function (Ben-Tal and Nemirovski, [24]; [25]; Bertsimas and Sim [26]; [27]; Delage and Ye [28]).…”

Adaptive reinforcement learning for task scheduling in aircraft maintenance

“…While the adaptive algorithm falls slightly behind the static algorithm in terms of Ground Time and Time Slack, both critical Key Performance Indicators (KPIs) in aircraft maintenance, it excels significantly in the Change Score metric. This metric gains paramount importance in current maintenance scenarios, as it measures the impact of plan modifications on various aspects such as maintenance personnel, equipment/shop, space, and parallel capacity 11 – 13 . These findings underscore the adaptive algorithm’s exceptional adaptability to handle unexpected changes, making it highly valuable in dynamic maintenance environments where flexibility is of utmost importance.…”

Aircraft fleet availability optimisation: a reinforcement learning approach

“…), for the operation and maintenance (O&M) of mechanical components. While a significant body of work has focused on simulation/optimisation frameworks to solve the maintenance task scheduling problem of a fleet of aircraft [7][8][9][10], in this work we will mainly focus on previous RL applications. Several recent studies [11,12] proposed novel RL frameworks to optimise the O&M of a single mechanical component (e.g., pump/turbine/turbofan engine) and demonstrated that it can find optimal policies when compared to traditional maintenance strategies.…”