Kalman based neural network analysis with resampling methods for longitudinal aerodynamic coefficient estimation

Date

2020-01-01

Author

Millidere, Murat
Kurt, Huseyin Burak
Balli, Hakan
Uğur, Ömür

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

46
views

0
downloads

Cite This

© 2020, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.Accurateness of the approximated aerodynamic characteristics of an unstable aircraft is considerably significant in flight control system design or high-fidelity flight simulator development. Classical system identification technique consist of the equation and output error methods which are strongly influenced by data quality. Accordingly, unsatisfactory results may be arisen due to measurement or process noise. Consequently, most of the engineers rely on feed forward neural network in order to cope with those undesired drawbacks. But, noisy data dramatically degrades the performance of neural network as well; thus, the Kalman filter based backpropagation algorithm is proposed. Neural network approach has many parameters, including the hyperparameters, which have to be searched for an optimal result. To determine these optimal parameters, the genetic algorithm is used. In this paper, it is aimed to estimate longitudinal aerodynamic characteristics of highly maneuverable unstable aircraft with the engaged control system for flight simulation data. For this purpose, F-16 aircraft is modelled using the aerodynamic database derived from a low-speed wind tunnel test. Successful results show the effectiveness of the proposed method. To assess neural network approach performance, the most common resampling methods are used and compared in order to choose the best for each longitudinal aerodynamic coefficient.

URI

https://hdl.handle.net/11511/69748

Collections

Graduate School of Applied Mathematics, Conference / Seminar