From industry-wide parameters to aircraft-centric on-flight inference: improving aeronautics performance prediction with machine learning

  • 2020-09-24 16:42:36
  • Florent Dewez, Benjamin Guedj, Vincent Vandewalle
  • 0

Abstract

Aircraft performance models play a key role in airline operations, especiallyin planning a fuel-efficient flight. In practice, manufacturers provideguidelines which are slightly modified throughout the aircraft life cycle viathe tuning of a single factor, enabling better fuel predictions. However thishas limitations, in particular they do not reflect the evolution of eachfeature impacting the aircraft performance. Our goal here is to overcome thislimitation. The key contribution of the present article is to foster the use ofmachine learning to leverage the massive amounts of data continuously recordedduring flights performed by an aircraft and provide models reflecting itsactual and individual performance. We illustrate our approach by focusing onthe estimation of the drag and lift coefficients from recorded flight data. Asthese coefficients are not directly recorded, we resort to aerodynamicsapproximations. As a safety check, we provide bounds to assess the accuracy ofboth the aerodynamics approximation and the statistical performance of ourapproach. We provide numerical results on a collection of machine learningalgorithms. We report excellent accuracy on real-life data and exhibitempirical evidence to support our modelling, in coherence with aerodynamicsprinciples.

 

Quick Read (beta)

loading the full paper ...