Naslov
Aerodynamic model development with generalized additive models

Autori
Monstein, Raphael

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, diplomski rad, diplomski

Fakultet
School of Engineering

Mjesto
Zurich

Datum
20.02

Godina
2018

Stranica
115

Mentor
Capone, Pierluigi ; Vrdoljak, Milan

Ključne riječi
generalized additive models, flight test data, unscented Kalman filter, aerodynamic model

Sažetak
Aerodynamic models and flight simulations have many applications, with aircraft simulations in pilot training being the most widely known. Beside the training of pilots, aerodynamic models are used to develop and tune flight control systems or to predict the performance and the handling qualities during the development of new aircraft. This aim of thesis is to evaluate if generalized additive models (GAM) are suitable for the development of aerodynamic models from flight test data. To this end, the flight test data is first processed with an unscented Kalman filter to reduce the noise in the data and the total aerodynamic coefficients are calculated. The Kalman filter parameters are estimated using a gradient based optimization to maximize the likelihood of the estimates. Subsequently, the aerodynamic coefficients are modelled with GAM. The development of aerodynamic models with GAM from flight test data has advantages over other methods: GAM are computationally cheap ; the mathematical structure is intuitive and lends itself well for aerodynamic models ; a number of methods are available to help the developer select an appropriate model structure. But, at the moment it is unclear how sensitive GAM are to noise in the measurements, which is always present in practical applications. This uncertainty should be addressed in the future to increase confidence in this method.

Izvorni jezik
Engleski

Znanstvena područja
Zrakoplovstvo, raketna i svemirska tehnika

POVEZANOST RADA