Validation of FCOM profiles for aircraft engine flight data using neural networks

摘要

This paper explores the application of artificial neural network approach for aircraft engine health monitoring. The Digital Flight Data Recorder (DFDR) has volumes of data which if mined appropriately can provide valuable information about the aircraft health. The Flight Crew Operating Manual (FCOM) lays down operational profiles, which are recommended to be followed for efficient fuel usage and for minimizing maintenance effort. In the proposed system, the information from FCOM profiles and ‘known’ flight data has been fused to train a back propagation feed-forward neural network. The predictions made by the neural network regarding the expected data of required engine parameters have been used to monitor the flight data and diagnose the health of the aircraft engine in relevance to the FCOM profiles. A Matlab GUI has been developed to simulate the ‘unknown’ flight data through a Simulink model for the neural network. Data from the A320 family of aircrafts has been used for training and simulating the model and preliminary results are detailed in the paper. The simulation results exhibit that the data used is fairly healthy and show a very low level of severity of degradation with respect to the profiles studied.