Artificial Intelligence-Based Structural Health Monitoring and Predictive Maintenance for Civil Aircraft

Abstract

Structural health monitoring (SHM) of civil aircraft is a critical enabler of aviation safety and operational reliability. Traditional SHM pipelines, which depend on physics-based mathematical models and rigidly scheduled maintenance, encounter two persistent bottlenecks: limited adaptability to complex, time-varying operating conditions, and insufficient throughput for the rapidly growing volume of multisource sensor data. Against this background, this paper presents a systematic review of aircraft SHM technologies organized around three pillars—probabilistic safety coefficient design, nano-modified advanced structural materials, and AI-driven predictive maintenance—with emphasis on the third. Specifically, this study examines how one-dimensional convolutional neural networks (1D-CNNs) automate fault feature extraction from raw vibration signals, how long short-term memory (LSTM) networks model long-range degradation trajectories for remaining useful life (RUL) estimation, and how reinforcement learning (RL) frameworks optimize maintenance scheduling under multi-objective constraints. It also benchmarks the open-source datasets most widely used in this field, including the CWRU bearing dataset and NASA C-MAPSS turbofan dataset. Reported results indicate that AI-based predictive maintenance can reduce unscheduled maintenance events by 35%–40% and lower overall MRO costs by 20%-30% while improving dispatch reliability. This study concludes that the integration of structural mechanics principles with data-driven AI models—particularly through physics-informed neural networks and edge-deployed lightweight models—represents the most promising direction for next-generation onboard intelligent health monitoring systems.