CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Tail-Number-Specific Aircraft Performance Modeling Using Flight Data and Machine Learning

Estudiants que han llegit aquest projecte:

Director/a: PRATS MENÉNDEZ, XAVIER

Departament: FIS

Títol: Tail-Number-Specific Aircraft Performance Modeling Using Flight Data and Machine Learning

Data inici oferta: 23-01-2026     Data finalització oferta: 23-09-2026


Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Segon director/a (UPC): DE LA TORRE SANGRÀ, DAVID
 
Paraules clau:
Aircraft performace, IA, BADA, MLP, machine learning, fuel flow, tail-number, flight data, QAR
 
Descripció del contingut i pla d'activitats:
Desenvolupar models de performance "tail-number", és a dir models específics per cada unitat d'avió (per matrícula). El modelat es farà a partir de dades reals de vol provinents de Flight Data Recorder / Quick Access Recorder. L'objectiu és construir un model capaç d'estimar i calibrar paràmetres físics com ara coeficients aerodinàmics i paràmetres de motor. Es busca capturar els efectes de degradació, manteniment i variabilitat entre unitats que els models genèrics per tipus d'avió no reflecteixen. El treball combinarà tècniques de machine learning i intel·ligència artificial i es provaran diferents estratègies de modelatge.
 
Overview (resum en anglès):
Aircraft Performance Monitoring serves as a critical engineering process that allows major industrial manufacturers to track and analyze the efficiency, safety and operational areas. Traditional performance estimation and trajectory prediction frameworks are being recently outperformed by machine learning approaches in Air Traffic Management (ATM), that close the gap with real scenarios.

This bachelor thesis shows the design and implementation of a tail-number specific aircraft performance modeling using QAR flight data and a Multi Layer Perceptron (MLP). The primary objective of this thesis is to estimate the fuel flow along the cruise phase, the aerodynamic and fuel performance coefficients, and conclude with the specific tail-number degradation estimation. To perform the study, a QAR dataset comprising Airbus A320 fleet flights is used. An MLP is designed and fed with the QAR input data to estimate the desired results.

The study shows results for different operational aircraft ages and demonstrates good general estimations. The results are not completely accurate because the input data lacks lifetime information about each tail-number present in the dataset. Nevertheless, despite the input dataset callbacks, the general final estimations are robust and reliable. This thesis can serve as a base for future studies to globalize the envelope of the study, where complete input dataset can be used to increase estimations accuracy.

© CBLTIC Campus del Baix Llobregat - UPC