An Aircraft Fuel Flow Model of Cruise Phase Based on LSTM and QAR Data

The Quick Access Recorder (QAR) is an airborne data recording device used to collect and record a large amount of operational data generated by various systems in the aircraft. Aircraft engine fuel flow is an important data recorded by QAR. Modeling fuel flow is of great significance for airlines to monitor the operating status of aeroengine, guide pilot training and QAR data management. In order to establish the fuel flow model of the aircraft, this paper obtains the aircraft's flight parameters, aircraft attitude, control signals, engines, environment (wind and temperature) and other QAR data to construct feature vectors, and uses the weighted average of the two-engine fuel flow as the fitting target. LSTM is used to build regression relationship model between feature vector and fitting target, and the validity and accuracy of the model are verified by real flight data.

I. Introduction

As a flight data recording device, Quick Access Recorder (QAR) records thousands of parameters data generated during the flight of the aircraft. Such as pilots' control signal, aircraft attitude change, speed change, meteorological information, hydraulic and working status of various mechanical systems, etc. However, these data have not been effectively used [1]. Under the increasingly competition in the global aviation market, the airline department is faced with pressure from safety and economy in flight operation, and is eager to find rules from QAR data analysis to supply better flight management guidance and fuel-saving.