基於深度學習之數值及圖像預測是近年來熱門研究之一，深度學習是透過架設所需之神經網路架構，將訓練資料作為輸入給予訓練模型來進行深度學習訓練，調整訓練參數學習率( Learning rate ) 、批量大小 ( Batch size ) 與時期 ( Epoch ) 等超參數，直到模型訓練結束，並且儲存預測模型。本研究在於探討氣動力係數預測及壓力流場圖像預測，目標為使用實驗或計算取得大量資料，利用深度學習得到預測法則，以便快速預測不同翼型攻角之氣動力係數及預測圖像。
本研究是利用XFOIL氣動力模擬軟體得到大量氣動力係數樣本作為訓練集，透過編譯之神經網路架構進行訓練，得到預期之訓練模型，最後將氣動力係數預測模型配合基因演算法，進行外型優化。本研究優化翼型為NACA 0012外型座標，將翼型座標透過PARSEC二維參數化，並且將參數當成基因來演算得到最佳解，達到翼型優化之結果。
在流場預測方面，本研究是利用ANSYS FLUENT商用套裝軟體進行壓力流場模擬，本研究紊流模型為SST k-ω模型，計算格點由ANSYS FLUENT內建網格生成功能，採用非結構網格，樣本圖像後處理由ANSYS ENSIGHT進行，最後將樣本圖像作為輸入套到編譯之生成對抗網路(Generative Adversarial Networks，GAN)模型架構，透過生成器(Generator)及鑑別器(Discriminator)之大小值博弈，訓練後得到生成器模型，並且以WGAN架構優化訓練穩定性，讓最後圖像誤差百分比能獲得比原本GAN架構更好的圖像質量及準確度。

In this study, two neural network model architectures were employed to optimize airfoils and predict flow fields. The first architecture utilizes a convolutional neural network to foresee the aerodynamic coefficient. The XFOIL analysis software is employed to acquire the lift and drag coefficients of the airfoil, with collaboration of the algorithm, an airfoil optimization is achieved. The second architecture involves the use of a generative adversarial network (GAN) to predict images. Throughout the generation process, a discriminator is employed to differentiate between real and generated images. The training of the generator is adjusted by the evaluation of the discriminator to obtain the prediction model. To improve the stability of the training loss, changes were made to the neural network architecture and loss function to achieve the goal of predicting the pressure flow field image from the parameter value. Training samples are obtained by simulating the pressure flow field using ANSYS FLUENT software where k-ω SST turbulence model was utilized.

[1] Xiaowei Jin, Shengze Cai, Hui Li, and George Em Karniadakis," NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations." Journal of Computational Physics. 426: p. 109951, 2021.
[2] Yao Zhang, Woong Je Sung, and Dimitri N Mavris." Application of convolutional neural network to predict airfoil lift coefficient. In 2018 AIAA". ASCE/AHS/ASC structures, structural dynamics, and materials conference,2018.
[3] Wenhui Peng, Yao Zhang, and Michel Desmarais." Spatial convolution neural network for efficient prediction of aerodynamic coefficients". in AIAA Scitech 2021 Forum.2021.
[4] Wenhui Peng, Yao Zhang, Eric Laurendeau, and Michel C Desmarais," Learning aerodynamics with neural network." Scientific Reports. 12(1): p. 6779, 2022.
[5] Junfeng Chen, Jonathan Viquerat, and Elie Hachem," U-net architectures for fast prediction in fluid mechanics." 2019.
[6] Haizhou Wu, Xuejun Liu, Wei An, Songcan Chen, and Hongqiang Lyu," A deep learning approach for efficiently and accurately evaluating the flow field of supercritical airfoils." Computers & Fluids. 198: p. 104393, 2020.
[7] Raymond M Hicks and Preston A Henne," Wing design by numerical optimization." Journal of Aircraft. 15(7): p. 407-412, 1978.
[8] Matt W Gardner and SR Dorling," Artificial neural networks (the multilayer perceptron)—a review of applications in the atmospheric sciences." Atmospheric environment. 32(14-15): p. 2627-2636, 1998.
[9] Yann LeCun and Yoshua Bengio," Convolutional networks for images, speech, and time series." The handbook of brain theory and neural networks. 3361(10): p. 1995, 1995.
[10] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio," Generative adversarial networks." Communications of the ACM. 63(11): p. 139-144, 2020.
[11] Martin Arjovsky, Soumith Chintala, and Léon Bottou." Wasserstein generative adversarial networks". in International conference on machine learning. PMLR,2017.
[12] Helmut Sobieczky," Parametric airfoils and wings." Recent development of aerodynamic design methodologies: inverse design and optimization: p. 71-87, 1999.
[13] Melanie Mitchell," An introduction to genetic algorithms". MIT press, 1998.
[14] Henk Kaarle Versteeg and Weeratunge Malalasekera," An introduction to computational fluid dynamics: the finite volume method". Pearson education, 2007.
[15] Suhas Patankar," Numerical heat transfer and fluid flow". Taylor & Francis, 2018.
[16] Joel H Ferziger, Milovan Perić, and Robert L Street," Computational methods for fluid dynamics". Vol. 3. Springer, 2002.
[17] Roger Temam," Navier-Stokes equations: theory and numerical analysis". Vol. 343. American Mathematical Soc., 2001.
[18] Florianr Menter." Zonal two equation kw turbulence models for aerodynamic flows". in 23rd fluid dynamics, plasmadynamics, and lasers conference.1993.
[19] Diederik P Kingma and Jimmy Ba," Adam: A method for stochastic optimization." arXiv preprint arXiv:1412.6980, 2014.
[20] Paul Werbos," New tools for prediction and analysis in the behavioral science." Ph. D. dissertation, Harvard University, 1974.
[21] David E Rumelhart, Geoffrey E Hinton, and Ronald J Williams," Learning representations by back-propagating errors." nature. 323(6088): p. 533-536, 1986.
[22] Vinod Nair and Geoffrey E Hinton." Rectified linear units improve restricted boltzmann machines". in Proceedings of the 27th international conference on machine learning (ICML-10).2010.
[23] Andrew L Maas, Awni Y Hannun, and Andrew Y Ng." Rectifier nonlinearities improve neural network acoustic models". in Proc. icml. Atlanta, Georgia, USA,2013.
[24] Léon Bottou," Stochastic gradient descent tricks." Neural Networks: Tricks of the Trade: Second Edition: p. 421-436, 2012.
[25] Léon Bottou and Olivier Bousquet," The tradeoffs of large scale learning." Advances in neural information processing systems. 20, 2007.
[26] Boris T Polyak," Some methods of speeding up the convergence of iteration methods." Ussr computational mathematics and mathematical physics. 4(5): p. 1-17, 1964.
[27] John Duchi, Elad Hazan, and Yoram Singer," Adaptive subgradient methods for online learning and stochastic optimization." Journal of machine learning research. 12(7), 2011.
[28] Geoffrey Hinton, Nitsh Srivastava, and Kevin Swersky," Neural networks for machine learning." Coursera, video lectures. 264(1): p. 2146-2153, 2012.
[29] J Morgado, R Vizinho, MAR Silvestre, and JC Páscoa," XFOIL vs CFD performance predictions for high lift low Reynolds number airfoils." Aerospace Science and Technology. 52: p. 207-214, 2016.