Generalizable Aircraft Takeoff Weight Estimation from Trajectory Data using Machine Learning

Recep Ayzit, Melih Safa Cengiz, Barış Başpınar, Mevlüt Uzun, Mustafa Umut Demirezen, Gökhan Inalhan, Javier López Leonés

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

In the performance analysis of an aircraft, the takeoff weight is an important parameter since it directly influences the force equilibrium in flight conditions. Along with the takeoff weight, if flight trajectory and fuel flow characteristics are also available or accurately estimated for a flight, flight performance can be improved by analyzing the data. Due to the airline policies regarding access to flight data, gross weight, fuel consumption, and other aircraft sensory data are challenging to obtain. For that reason, estimating the aircraft weight from surveillance data such as ADS-B is crucial in analyzing aircraft performance characteristics. In this paper, a generalizable machine-learning approach to estimate aircraft takeoff weight with the extracted features from different phases of the trajectory data for three different types of wide and narrow-body aircraft is proposed. The feature extraction process and the validation of the model with ground truth data are performed using the Quick Access Recorder (QAR) data. The evaluation of the method is carried out by creating four different machine-learning models and comparing the mean absolute percentage errors (MAPE) on the test set. Two case studies are analyzed by training models with single aircraft and three different aircraft having distinct takeoff weights to demonstrate that the model is generalizable. The best result is obtained using Extra Trees Regressor for the multiple aircraft case with a mean absolute percentage error and standard deviation of 1.48% TOW and 2.12% TOW, respectively.

Original languageEnglish
Title of host publicationAIAA Aviation and Aeronautics Forum and Exposition, AIAA AVIATION Forum 2023
PublisherAmerican Institute of Aeronautics and Astronautics Inc, AIAA
ISBN (Print)9781624107047
DOIs
Publication statusPublished - 2023
EventAIAA Aviation and Aeronautics Forum and Exposition, AIAA AVIATION Forum 2023 - San Diego, United States
Duration: 12 Jun 202316 Jun 2023

Publication series

NameAIAA Aviation and Aeronautics Forum and Exposition, AIAA AVIATION Forum 2023

Conference

ConferenceAIAA Aviation and Aeronautics Forum and Exposition, AIAA AVIATION Forum 2023
Country/TerritoryUnited States
CitySan Diego
Period12/06/2316/06/23

Bibliographical note

Publisher Copyright:
© 2023, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.

Fingerprint

Dive into the research topics of 'Generalizable Aircraft Takeoff Weight Estimation from Trajectory Data using Machine Learning'. Together they form a unique fingerprint.

Cite this