A Genetic Algorithm and Monte Carlo Simulation-Based Framework for Maintenance Policy Optimization of Multi-Component Aircraft Systems

Abstract

This thesis presents a maintenance policy optimization framework for multi-component sys-tems that are subject to stochastic degradation and imperfect maintenance. The objective of the thesis is to determine the optimal inspection interval as well as the repair and replace- ment maintenance thresholds that minimize the expected long-run cost of operation. The optimization framework considers component degradation, maintenance actions, and system reliability requirements. The multi-component system is modeled as a series-parallel architecture that consists of multiple subsystems with k-out-of-n redundancy structures. Component degradation is represented using a few different degradation models to reflect different physical failure mech- anisms. Examples include Paris’ Law for fatigue crack growth and Wiener process models for wear accumulation. The system components are inspected periodically, and maintenance decisions are determined using condition-based thresholds that define repair and replacement actions. Repairs are assumed to be imperfect, restoring components to a condition that is better than old but worse than new. A Monte Carlo simulation framework was developed to evaluate maintenance policies under uncertainty. The simulation tracks component degradation, maintenance events, cor- rective actions, and system failures over a specified operating horizon. A genetic algorithm was also developed to optimize the inspection interval and component-level maintenance thresholds. The overall objective function aims to minimize the expected cost per unit time, including inspection, repair, replacement, corrective maintenance, and system failure costs. The proposed framework is ultimately demonstrated through two aircraft-inspired case studies consisting of a turbofan engine, hydraulic pump, electrical generator, and actuator subsystem. In the simplified single-component-per-subsystem case study, the optimized pol- icy resulted in an inspection interval of 211.9 hours and an expected operating cost of $303.79 per hour. For the complex k-of-n system case study, the optimized inspection interval de- creased to 146.5 hours with an expected operating cost of $494.76 per hour. In addition, a sensitivity analysis was conducted to demonstrate how the optimized policy changes when the cost parameters are individually increased and decreased by ±5%. The final results illus- trate the ability of the optimization framework to identify cost-effective maintenance policies while accounting for stochastic degradation behavior and system-level reliability constraints. The methodology provides a flexible approach for condition-based maintenance optimization that can be extended to a wide range of engineering systems.