本論文探討二維機翼及三維翼身在不同雷諾數下流場，流體流過俯仰機翼的空氣動力係數，並分析二維機翼所產生的動態失速現象。本文結合狀態空間模型及具有較高的建模精度和對多變量的可擴展性的神經網絡模型，得到一個兼具兩者優點的非穩定氣動混合模型，並採用此混合模型及商用計算流體力學軟體Ansys Fluent進行模擬計算。數值方法使用RANS (Reynolds-average Navier-Stokes)之SST-k-omega紊流模型進行計算，並運用動網格技術模擬不同實驗，結果分別以升力係數及阻力係數對攻角作圖呈現，亦與相關實驗數據做比較。軟體模擬結果升力係數較接近實驗值，阻力係數則由於在大攻角下持續存在流動分離，導致使用RANS時難以準確模擬機翼表面附近的黏性效應，因此準確度較低。但與靜態俯仰結果可看出俯仰運動推遲失速，會使失速發生到更高的攻角。
本論文首先進行混合模型驗證，分別與風洞俯仰機翼數據及軟體二維機翼模擬結果進行比對，以此增加模擬數據可信度。將混合模型使用於三維機翼翼身之俯仰運動的氣動力與擺動攻角之關係的預測。另外，本文提出基因演算法對翼型的優化以及改進翼型的設計，探討優化後的翼型的氣動力，與原本機翼氣動力係數相比，升力係數提高，阻力係數降低，將模擬結果帶入混合模型中學習，計算給定不同俯仰頻率時氣動力隨攻角之變化。
本文結合狀態空間表示與反向傳播神經網絡的混合模型，模擬結果可以看出在不降低模型精度標準的情況下實現混合模型與風洞數據及模擬結果的高度相似，不同擺動頻率的升力係數與阻力係數的預測值亦與軟體模擬結果相似，證明混合模型對飛機非穩態氣動建模是準確和有效的。

In this paper, the aerodynamic coefficients of the fluid flowing through the pitching motion of the two-dimensional airfoil and the three-dimensional wing body are discussed at different Reynolds numbers. The dynamic stall phenomenon caused by the higher attack angle is analyzed. This paper combines the state space model and the neural network model with high modeling accuracy and multi-variable scalability to obtain an unsteady aerodynamic hybrid model. The hybrid model and the commercial computational fluid dynamics software Ansys Fluent were used for these simulations. The numerical method uses the SST-k-omega turbulence model of RANS (Reynolds-average Navier-Stokes) for calculation and the dynamic grid for simulating different pitching motions. The lift coefficients of the software simulation results are close to the experimental data, and the drag coefficients are less accurate due to the continuous flow separation at large angles of attack, which makes it difficult to accurately simulate the viscous effect near the wing surface when using RANS. The hybrid model is first verified and compared with pitching wing data of the wind tunnel and the simulation results of the software. In addition, this paper proposes a genetic algorithm to optimize the airfoil and to improve the design of the airfoil. Compared with the original airfoil aerodynamic coefficient, the lift coefficient is increased, and the drag coefficient is decreased. The change of aerodynamic force with the angle of attack is calculated for different given pitch frequencies and is also investigated by the neural network model.

[1] Ham, Norman D., Garelick, Melvin S., Dynamic Stall Considerations in Helicopter Rotors, Journal of the American Helicopter Society, Vol.13, No.2, pp.49-55, 1968.
[2] Carr, L. W., Progress in analysis and prediction of dynamic stall, Journal of Aircraft, Vol.25, No.1, 1988.
[3] Carr, L. W., Mcalister, K.W., Mccroskey, W. J., Analysis of the development of dynamic stall based on oscillating airfoil experiments, NASA Techniacl Note D-8382, 1977.
[4] Lee, T., Gerontakos, P., Investigation of flow over an oscillating airfoil, Journal of Fluid Mechanical, Vol.512, pp.313-341, 2004.
[5] Visbal, M. R., Shang, J. S., Investigation of Flow Structure around a Rapidly Pitching Airfoil, AIAA Journal, Vol.27, No.8, pp.1044-1051, 1989.
[6] Visbal, M. R., Dynamic Stall of a Constant-Rate Pitching Airfoil, Journal of Aircraft, Vol.27, No.5, pp.400-407, 1990.
[7] Visbal, M. R., Choudhuri, P. G., Knight, D. D., Two-Dimensional Unsteady Leading-Edge Separation on a Pitching Airfoil, AIAA Journal, Vol.32, No.4, pp.673-681, 1994.
[8] Wang, S., Ingham, D. B., Ma, L., Pourkashanian, M., Tao, Z., Turbulence modeling of deep dynamic stall at relatively low Reynolds number, Journal of Fluids and Structures, Vol.33, pp.191-209, 2012.
[9] Gharali, K., Johnson, D. A., Dynamic Stall Simulation of a Pitching Airfoil Under Unsteady Freestream Velocity, Journal of Fluids and Structures, Vol.42, pp.228-244, 2013.
[10] Mojtaba, H., Mohammad, H. D., Behzad, F. F., Esmaeil, E., Dynamic Stall Simulation of a Pitching Airfoil Under Unsteady Freestream Velocity, Journal of Aerospace Technology and Management, Vol.11, 2019.
[11] Ning, D., Qiulin, Q., Ramesh, K. A., Numerical Study of the Aerodynamics of DLR-F6 Wing-Body in Unbounded Flow Field and in Ground Effect, 55th AIAA Aerospace Sciences Meeting, 2017.
[12] Goman, M., Khrabrov, A., State-space representation of aerodynamic characteristics of an aircraft at high angles of attack, Journal of Aircraft, Vol.31, No.5, pp.1109-1115, 1994.
[13] Zakaria, M. Y., Taha, H. E., Hajj, M. R., Hussein, A. A., Experimental-Based Unified Unsteady Nonlinear Aerodynamic Modeling For Two Dimensional Airfoils, 33rd AIAA Applied Aerodynamics Conference, pp.2015-3167, 2015.
[14] Kumar, R., Ghosh, A. K., Nonlinear Longitudinal Aerodynamic Modeling Using Neural Gauss-Newton Method, Journal of Aircraft, Vol.48, No.5, pp.1809-1813, 2011.
[15] Ouyang, G., Lin, J., Zhang, P., Aircraft Unsteady Aerodynamic Hybrid Modeling based on State-space Representation and Neural Network, Proceedings of the 5th International Conference on Pattern Recognition Applications and Methods, pp.232-239, 2016.
[16] Liu, H., Liu, S., Ye, W., Wang, C., Liu, J., Lv, G., Airfoil Optimization Based on Neural Network and Derivative-free Algorithms.
[17] Paul, D. S., Ruxandra, M. B., Two-dimensional airfoil shape optimization for airfoils at low speeds, AIAA Atmospheric Flight Mechanics, 2012.
[18] Mentor, F. R., Two-Equation Eddy-Viscosity Turbulence Models for Engineering Application, AIAA Journal, Vol.32, No.8, pp.1598-1605, 1994.
[19] Ahmad, K. A., Adbullah, M. Z., Watterson, J. K., Numerical Modeling of a Pitching Airfoil, Journal Mekanikal, Vol.30, pp.37-47, 2010.
[20] Geng, F., Kalkman, I., Suiker, A. S. J., Blocken, B., Sensitivity analysis of airfoil aerodynamics during pitching motion at a Reynolds number of 1.35×10^5, Journal of Wind Engineering & Industrial Aerodynamics, Vol.183, 2018.