Modeling of Aircraft Takeoff Weight Using Gaussian Processes

Abstract

The takeoff weight of an aircraft is an important aspect of aircraft performance. However, the takeoff weight of a particular flight is generally not available to entities outside of the operating airline. The preceding observations motivate the development of accurate takeoff weight estimates that can be used for fuel-burn estimation or trajectory prediction. This paper proposes a statistical approach based on Gaussian process regression to determine both a mean estimate of the takeoff weight and the associated prediction interval, using observed data from the takeoff ground roll. The model development and validation are conducted using flight data recorder archives, which also provide ground-truth data. The models are found to have a mean absolute error in takeoff weight of 3.6%, averaged across nine different aircraft types, resulting in a nearly 35% smaller error than the models in the Aircraft Noise and Performance database. Finally, the developed models are used to predict aircraft fuel flow rate during climb out and approach. For the majority of the aircraft types studied, the statistical models of takeoff weight estimation are shown to result in a similar or better fuel flow rate predictive performance as compared to the Aircraft Noise and Performance models.