STATISTICAL MODELING OF AIRCRAFT ENGINE FUEL FLOW RATE

摘要

Knowledge of the fuel flow rate is importantfor accurate fuel burn and emissions inventorygeneration and for accurate engine performancemodeling. In this paper, machine learning techniquesare applied to aircraft flight recorder datato model the fuel flow rate as a function of theaircraft altitude, ground speed, vertical speed,and takeoff mass in the airborne phases of flight.Models are built using the Classification and RegressionTrees (CART) and the Least SquaresBoosting algorithms. The models are found to beinterpretable. They are also shown to predict fuelflow rate and its variability more accurately thanother frequently-used methods for fuel flow rateestimation. A sensitivity analysis of the predictedfuel flow rate to takeoff mass is also conducted.The proposed nonparametric models can help seta benchmark for developing more sophisticateddata-driven statistical models of engine fuel flowrate.