A Deep Learning On-Board Health Monitoring Method for Landing Gear Shock-Absorbing Systems

Abstract

Highlights: This paper proposed a new health monitoring method for landing gear shock-absorbing systems based on the dynamic responses of landing gears during aircraft landing phase. The proposed sensor network consists of vertical fuselage and wheel accelerometers, and a deep learning method LDGNet is proposed to conduct health monitoring for three typical shock absorber faults. The proposed health monitoring method is proved to be with good performance for all considered fault conditions. This paper proposed a deep learning on-board health monitoring method for landing gear shock-absorbing systems based on dynamic responses during landing. A deep learning model is developed to conduct health monitoring for faults in shock absorbers. A certain general aviation aircraft is focused on in this paper, and a multi-body dynamic model of the nose landing gear is developed to simulate dynamic responses during landing under various health states and various landing conditions for developing a database for the proposed LDGNet. The simulated database is used to conduct model training and to test the performance of the proposed method. The feasibility and effectiveness of the proposed method are verified.