Loading...
Search
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
Digital Object List
Bookmark
Review List
13 viewed
بکارگیری شبکه عصبی پویا در مدل سازی و شناسایی رفتار دینامیکی یک هواپیمای بال ثابت
ابراهیمی، محمد مهدی
Modeling and Identification of Fixed-Wing Aircraft Behavior using Dynamic Neural Networks
| 2025
13
Viewed
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 58747 (45)
- University: Sharif University of Technology
- Department: Aerospace Engineering
- Advisor(s): Banazadeh, Afshin
- Abstract:
- This thesis builds and validates a continuous-time neural model of fixed-wing flight using the Neural Ordinary Differential Equation framework. A Simulink-based analytic simulator of a Cessna 172 in cruise provides ground truth under step, doublet, and random excitations. The model preserves six analytical kinematics and learns six dynamic derivatives via channel-wise Neural ODE subnetworks from elevator, aileron, and rudder inputs with fixed throttle. Training follows a two-stage pipeline—derivative warm-up then adjoint-based continuous-time optimization with rollout—supported by Butterworth and Savitzky–Golay label denoising, input standardization, and robust target normalization. On analytic data it attains low mean-squared errors across all dynamic outputs, with the largest discrepancies limited to short-period pitch behavior; single-axis and coupled tests show mode-consistent responses and small trajectory deviations over ninety-second windows. Contributions include a physics-aligned channel-wise architecture, a training pipeline that balances local derivative fidelity with rollout coherence, and a reproducible, interpretable evaluation protocol
- Keywords:
- Dynamic Neural Network ; Flight Dynamics ; System Identification ; Fixed-Wing Aircraft ; Neural Ordinary Differential Equations (ODE) ; Dynamics Identification
-
محتواي کتاب
- view