Journal of Space Science and Technology

Journal of Space Science and Technology

Deep Recurrent Neural Networks for Predicting Supersonic Pressure Coefficient Distributions on Parametric Aerodynamic Geometries

Document Type : Original Research Paper

Authors
Aerospace Research Institute, Ministry of Science, Research and Technology, Tehran, Iran
Abstract
Accurate prediction of surface pressure distributions is critical for the aerodynamic analysis and design of high-speed axisymmetric bodies, particularly for configurations such as sphere-cone and flares. Traditional computational fluid dynamics (CFD) provides reliable results but is computationally expensive, especially when evaluating numerous geometrical variations. To address this challenge, we propose a Bidirectional Long Short-Term Memory (BiLSTM) network as a data-driven surrogate model for predicting the pressure coefficient (Cp​) along axisymmetric surfaces. The BiLSTM leverages the sequential nature of surface geometry to learn both local and global flow features from a compact dataset of high-fidelity CFD simulations. Eighty geometries, each defined by seven key parameters, were simulated at Mach 2 using two-dimensional axisymmetric CFD under laminar flow assumptions. Comparative studies with feedforward neural networks (FNN) and unidirectional LSTM architectures demonstrate the superiority of the BiLSTM, achieving R² = 0.985 and a mean absolute error of 0.045. The results show that the bidirectional structure effectively captures forward and backward spatial correlations, yielding smooth and physically consistent Cp​ predictions, particularly near stagnation and flare compression zones. This framework provides a computationally efficient, high-fidelity, and generalizable approach for aerodynamic prediction, offering significant potential for rapid conceptual design and parametric studies of complex axisymmetric bodies.
Keywords
Subjects

Article Title Persian

شبکه‌های عصبی بازگشتی عمیق برای پیش‌بینی توزیع ضرایب فشار مافوق صوت بر روی هندسه‌های آیرودینامیکی پارامتری

Author Persian

صالح نعمتی
پژوهشگاه هوافضا، وزارت علوم، تحقیقات و فناوری، تهران، ایران
Abstract Persian

پیش‌بینی دقیق توزیع فشار سطح برای تحلیل آیرودینامیکی و طراحی اجسام با تقارن محوری پرسرعت، به‌ویژه برای پیکربندی‌هایی مانند مخروط‌ها و فلرهای کروی، بسیار مهم است. دینامیک سیالات محاسباتی سنتی (CFD) نتایج قابل اعتمادی ارائه می‌دهد اما از نظر محاسباتی گران است، به‌ویژه هنگام ارزیابی تغییرات هندسی متعدد. برای پرداختن به این چالش، ما یک شبکه حافظه کوتاه‌مدت بلندمدت دوطرفه (BiLSTM) را به عنوان یک مدل جایگزین مبتنی بر داده برای پیش‌بینی ضریب فشار (Cp) در امتداد سطوح متقارن محوری پیشنهاد می‌کنیم. BiLSTM از ماهیت ترتیبی هندسه سطح برای یادگیری ویژگی‌های جریان محلی و جهانی از یک مجموعه داده فشرده از شبیه‌سازی‌های CFD با دقت بالا استفاده می‌کند. هشتاد هندسه، که هر کدام توسط هفت پارامتر کلیدی تعریف شده‌اند، در سرعت ماخ ۲ با استفاده از CFD متقارن محوری دوبعدی تحت فرضیات جریان آرام شبیه‌سازی شدند. مطالعات مقایسه‌ای با شبکه‌های عصبی پیش‌خور (FNN) و معماری‌های LSTM تک‌جهته، برتری BiLSTM را نشان می‌دهد و به R² = 0.985 و میانگین خطای مطلق ۰.۰۴۵ دست یافته است. نتایج نشان می‌دهد که ساختار دو جهته به طور مؤثر همبستگی‌های مکانی رو به جلو و عقب را ثبت می‌کند و پیش‌بینی‌های Cp روان و از نظر فیزیکی سازگار، به ویژه در نزدیکی مناطق تراکم سکون و شعله‌ور، ارائه می‌دهد. این چارچوب یک رویکرد محاسباتی کارآمد، با دقت بالا و قابل تعمیم برای پیش‌بینی آیرودینامیکی ارائه می‌دهد و پتانسیل قابل توجهی برای طراحی مفهومی سریع و مطالعات پارامتری اجسام متقارن محوری پیچیده ارائه می‌دهد.

Keywords Persian

شبکه‌های عصبی
جریان‌های مافوق صوت
BiLSTM
دینامیک سیالات محاسباتی (CFD)
ضریب فشار
[1]    M. M. Rai and N. K. Madavan, “Aerodynamic design using neural networks,” AIAA Journal, vol. 38, no. 1, pp. 173-182, 2000, https://doi.org/10.2514/2.938.
[2]    L. Hu, J. Zhang, Y. Xiang, and W. Wang, “Neural networks-based aerodynamic data modeling: A comprehensive review,” IEEE Access, vol. 8, pp. 90805-90823, 2020, https://doi.org/10.1109/ACCESS.2020.2993562.
[3]    S. L. Brunton, B. R. Noack, and P. Koumoutsakos, “Machine learning for fluid mechanics,” Annual Review of Fluid Mechanics, vol. 52, no. 1, pp. 477-508, 2020, https://doi.org/10.1146/annurev-fluid-010719-060214.
[4]    R. Vinuesa and S. L. Brunton, “Enhancing computational fluid dynamics with machine learning,” Nature Computational Science, vol. 2, no. 6, pp. 358-366, 2022, https://doi.org/10.1038/s43588-022-00264-7.
[5]    H. Salehi, S. Das, S. Chakrabartty, S. Biswas, and R. Burgueño, “Damage identification in aircraft structures with self‐powered sensing technology: A machine learning approach,” Structural Control and Health Monitoring, vol. 25, no. 12, 2018, Art. no. e2262, https://doi.org/10.1002/stc.2262.
[6]    M. De Jong et al., “A statistical learning framework for materials science: application to elastic moduli of k-nary inorganic polycrystalline compounds,” Scientific Reports, vol. 6, no. 1, 2016, Art. no. 34256, https://doi.org/10.1038/srep34256.
[7]    J. Kou and W. Zhang, “Data-driven modeling for unsteady aerodynamics and aeroelasticity,” Aerospace Sciences, vol. 125, 2021, Art. no. 100725, https://doi.org/10.1016/j.paerosci.2021.100725.
[8]    J. Li, X. Du, and J. R. Martins, “Machine learning in aerodynamic shape optimization,” Aerospace Sciences, vol. 134, 2022, Art. no. 100849, https://doi.org/10.1016/j.paerosci.2022.100849.
[9]    H. Y. Fan, G. S. Dulikravich, and Z. X. Han, “Aerodynamic data modeling using support vector machines,” Inverse Problems in Science and Engineering, vol. 13, no. 3, pp. 261-278, 2005, https://doi.org/10.2514/6.2004-280.
[10]  Z. Wang, I. Akhtar, J. Borggaard, and T. Iliescu, “Proper orthogonal decomposition closure models for turbulent flows: a numerical comparison,” Computer Methods in Applied Mechanics and Engineering, vol. 237, pp. 10-26, 2012, https://doi.org/10.1016/j.cma.2012.04.015.
[11]  E. Andrés-Pérez and C. Paulete-Periáñez, “On the application of surrogate regression models for aerodynamic coefficient prediction,” Complex & Intelligent Systems, vol. 7, no. 4, pp. 1991-2021, 2021, https://doi.org/10.1007/s40747-021-00307-y.
[12]  A. Teimourian, D. Rohacs, K. Dimililer, H. Teimourian, M. Yildiz, and U. Kale, “Airfoil aerodynamic performance prediction using machine learning and surrogate modeling,” Heliyon, vol. 10, no. 8, 2024, https://doi.org/10.1016/j.heliyon.2024.e29377.
[13]  J. Zhang and X. Zhao, “Machine-learning-based surrogate modeling of aerodynamic flow around distributed structures,” AIAA Journal, vol. 59, no. 3, pp. 868-879, 2021, https://doi.org/10.2514/1.J059877.
[14]  M. F. Negoita and M. V. Hothazie, “A machine learning-based approach for predicting aerodynamic coefficients using deep neural networks and cfd data,” INCAS Bulletin, vol. 16, no. 4, 2024, https://doi.org/10.13111/2066-8201.2024.16.4.9.
[15]  D. Pereira, F. Afonso, and F. Lau, “End-to-end deep-learning-based surrogate modeling for supersonic airfoil shape optimization,” Aerospace, vol. 12, no. 5, 2025, Art. no. 389, https://doi.org/10.3390/aerospace12050389.
[16]  D. A. Sterpu, D. Măriuța, G. Cican, C. M. Larco, and L.-T. Grigorie, “Machine learning prediction of airfoil aerodynamic performance using neural network ensembles,” Applied Sciences, vol. 15, no. 14, 2025, Art. no. 7720.
[17]  G. Catalani, J. Fesquet, X. Bertrand, F. Tost, M. Bauerheim, and J. Morlier, “Towards scalable surrogate models based on Neural Fields for large scale aerodynamic simulations,” Computers & Fluids, 2025, Art. no. 106929, https://doi.org/10.1016/j.compfluid.2025.106929.
[18]  P. Bo, Y. Zhang, L. Junlin, W. Xudong, M. Chang, and B. Junqiang, “Data-driven surrogate model for aerodynamic design using separable shape tensor method,” Chinese Journal of Aeronautics, vol. 37, no. 9, pp. 41-58, 2024, https://doi.org/10.1016/j.cja.2024.03.014.
[19]  X. Hui, J. Bai, H. Wang, and Y. Zhang, “Fast pressure distribution prediction of airfoils using deep learning,” Aerospace Science and Technology, vol. 105, 2020, Art. no. 105949, https://doi.org/10.1016/j.ast.2020.105949.
[20]  K. Li, J. Kou, and W. Zhang, “Deep neural network for unsteady aerodynamic and aeroelastic modeling across multiple Mach numbers,” Nonlinear Dynamics, vol. 96, pp. 2157-2177, 2019, https://doi.org/10.1007/s11071-019-04915-9.
[21]  A. T. Mohan and D. V. Gaitonde, “A deep learning based approach to reduced order modeling for turbulent flow control using LSTM neural networks,” Arxiv Preprint Arxiv:1804.09269, 2018,
https://doi.org/10.48550/arXiv.1804.09269.
[22]  Z. Deng, Y. Chen, Y. Liu, and K. C. Kim, “Time-resolved turbulent velocity field reconstruction using a long short-term memory (LSTM)-based artificial intelligence framework,” Physics of Fluids, vol. 31, no. 7, 2019, https://doi.org/10.1063/1.5111558.
[23]  S. Pawar, S. M. Rahman, H. Vaddireddy, O. San, A. Rasheed, and P. Vedula, “A deep learning enabler for nonintrusive reduced order modeling of fluid flows,” Physics of Fluids, vol. 31, no. 8, 2019, https://doi.org/10.1063/1.5113494.
[24]  J. Wang, C. He, R. Li, H. Chen, C. Zhai, and M. Zhang, “Flow field prediction of supercritical airfoils via variational autoencoder based deep learning framework,” Physics of Fluids, vol. 33, no. 8, 2021, https://doi.org/10.1063/5.0053979.
[25]  K. Taunk, S. De, S. Verma, and A. Swetapadma, “A brief review of nearest neighbor algorithm for learning and classification,” in International Conference on Intelligent Computing and Control Systems (ICCS), Madurai, India, 2019, Paper 1255-1260, 10.1109/ICCS45141.2019.9065747.
[26]  J. A. Sánchez Muñoz, C. Lagarza-Cortés, and J. Ramírez-Cruz, “Aerodynamic drag coefficient prediction of a spike blunt body based on k-nearest neighbors,” Aerospace, vol. 11, no. 9, 2024, Art. no. 757, 10.3390/aerospace11090757.
[27]  A. Kashefi, D. Rempe, and L. J. Guibas, “A point-cloud deep learning framework for prediction of fluid flow fields on irregular geometries,” Physics of Fluids, vol. 33, no. 2, 2021, https://doi.org/10.1063/5.0033376.
[28]  Y. Shen, W. Huang, Z.-g. Wang, D. f. Xu, and C. Y. Liu, “A deep learning framework for aerodynamic pressure prediction on general three-dimensional configurations,” Physics of Fluids, vol. 35, no. 10, 2023, https://doi.org/10.1063/5.0172437.
[29]  R. Mehta, “Numerical simulation of supersonic flow past reentry capsules,” Shock Waves, vol. 15, no. 1, pp. 31-41, 2006.
[30]  H. Shi, M. Xu, and R. Li, “Deep learning for household load forecasting—A novel pooling deep RNN,” IEEE Transactions on Smart Grid, vol. 9, no. 5, pp. 5271-5280, 2017, https://doi.org/10.1109/TSG.2017.2686012.
[31]  J. S. Zhang and X. C. Xiao, “Predicting chaotic time series using recurrent neural network,” Chinese Physics Letters, vol. 17, no. 2, 2000, Art. no. 88, https://doi.org/10.1109/NNSP.1992.253669.
[32]  A. Graves, Supervised Sequence Labelling. Springer, 2012.
[33]  S. Hochreiter, Y. Bengio, P. Frasconi, and J. Schmidhuber, “Gradient flow in recurrent nets: The difficulty of learning long-term dependencies,” in A Field Guide to Dynamical Recurrent Neural Networks. IEEE Press In, 2001, https://doi.org/10.1109/9780470544037.ch14.
[34] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural Computation, vol. 9, no. 8, pp. 1735-1780, 1997, https://doi.org/10.1162/neco.1997.9.8.1735.
 
Volume 19, Issue 3
2026
Pages 56-70

  • Receive Date 16 August 2025
  • Revise Date 29 November 2025
  • Accept Date 18 February 2026
  • First Publish Date 01 September 2026