Simulation-based fuzzy-rough nearest neighbour fault classification and prediction for aircraft maintenance

Yinling Liu (1) , Tao Wang (1) , Haiqing Zhang (2) , Vincent Cheutet (1)

1 DISP - Décision et Information pour les Systèmes de Production
2 School of Software Engineering [Chengdu]

This paper addresses the problem of Fault Classification and Prediction (FCP) without sufficient data in aircraft maintenance. An innovative approach based on Agent-Based Simulation System (ABSS) and the Fuzzy-Rough Nearest Neighbour (FRNN) method is proposed. To do so, a framework for integrating the FRNN approach into ABSS is firstly provided. The concept and architecture models of FRNN-ABSS are then used to describe the FRNN-ABSS system. Subsequently, random and sequence strategies are designed for the FCP of the engine. An algorithm for integrating the FRNN method into ABSS is also developed to automate FCP. Finally, the experiments analysing the impact of different strategies on maintenance costs and service level have been conducted. The results show that the approach proposed has achieved success for random and sequence strategies: 9.3% and 2.5% of maintenance costs have been saved; 4.17% and 12.5% of delayed flights have been changed into on-time flights.

Mots clés: Fault classification and prediction; fuzzy-rough nearest neighbour; agent-based system and simulation; aircraft maintenance

Domaines: Génie mécanique [physics.class-ph]; Ingénierie assistée par ordinateur

Yinling Liu, Tao Wang, Haiqing Zhang, Vincent Cheutet. Simulation-based fuzzy-rough nearest neighbour fault classification and prediction for aircraft maintenance. Journal of Simulation, 2021, 15 (3), pp.202-216. ⟨10.1080/17477778.2019.1680261⟩. ⟨hal-02344426⟩