Perpustakaan Digital ITB

Improving aircraft design relies on accurate aerodynamic modeling for predicting flight characteristics, wherein system identification and parameter estimation play a pivotal role. While the existing method demonstrates adequacy in achieving the desired outcomes, this research explores the potential of applying a deep learning approach for parameter estimation. This research assesses the efficacy of Deep Learning in aircraft aerodynamic model identification, comparing it to the established Least Squares method. Utilizing flight test data from the ATTAS research aircraft, the study demonstrates that Deep Learning effectively captures trends and reduces fluctuations. While all models show good predictive capability, slight discrepancies with Least Square estimates and limitations related to data variability and outliers were observed. Future research should focus on overcoming these challenges to enhance the precision of Deep Learning models for parameter estimation.