Abstract
for flow-related design optimization problems, e.g., aircraft and automobile aerodynamic design, computational fluid dynamics (CFD) simulations are commonly used to predict flow fields and analyze performance. While important, CFD simulations are a resource-demanding and time-consuming iterative process. The expensive simulation overhead limits the opportunities for large design space exploration and prevents interactive design. In this paper, we propose FlowDNN, a novel deep neural network (DNN) to efficiently learn flow representations from CFD results. FlowDNN saves computational time by directly predicting the expected flow fields based on given flow conditions and geometry shapes. FlowDNN is the first DNN that incorporates the underlying physical conservation laws of fluid dynamics with a carefully designed attention mechanism for steady flow prediction. This approach not only improves the prediction accuracy, but also preserves the physical consistency of the predicted flow fields, which is essential for CFD. Various metrics are derived to evaluate FlowDNN with respect to the whole flow fields or regions of interest (RoIs) (e.g., boundary layers where flow quantities change rapidly). Experiments show that FlowDNN significantly outperforms alternative methods with faster inference and more accurate results. It speeds up a graphics processing unit (GPU) accelerated CFD solver by more than 14 000×, while keeping the prediction error under 5%.
摘要
对于与流场相关的设计优化问题, 例如飞机和汽车空气动力学设计, 计算流体力学 (CFD) 模拟通常用于预测流场并分析性能. 虽然CFD模拟十分重要, 但它的迭代计算非常需要计算资源且极其耗时. 昂贵的模拟开销限制了大范围设计空间的探索, 并阻碍了实时的交互式设计. 在本文中, 我们提出FlowDNN模型, 它是一种新颖的深度神经网络, 可从CFD结果中高效地学习流场表示. FlowDNN根据给定的流动条件和几何形状可以直接预测预期的流场结果, 从而极大地节省计算时间. FlowDNN首次结合了流体力学的基本守恒定律和注意力机制进行定常流场预测. 这样做不仅可以提高预测准确性, 而且可以维持预测流场的物理一致性, 这对于CFD模拟至关重要. 本文设计了多种指标以评估FlowDNN预测的整体流场和关键区域的结果 (如流场快速变化的边界层). 实验结果表明, FlowDNN明显优于其他方法且具有更短的推理时间和更准确的结果. 它与最新的GPU并行求解器相比, 生成流场的速度提升14 000倍以上, 同时保持预测误差在5%以内.
Similar content being viewed by others
References
Ahmed MYM, Qin N, 2009. Surrogate-based aerodynamic design optimization: use of surrogates in aerodynamic design optimization. Int Conf on Aerospace Sciences & Aviation Technology, p.1–26. https://doi.org/10.21608/ASAT.2009.23442
Amodio M, Krishnaswamy S, 2019. TraVeLGAN: image-to-image translation by transformation vector learning. IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.8975–8984. https://doi.org/10.1109/CVPR.2019.00919
Balabanov VO, Giunta AA, Golovidov O, et al., 1999. Reasonable design space approach to response surface approximation. J Aircr, 36(1):308–315. https://doi.org/10.2514/2.2438
Bhatnagar S, Afshar Y, Pan S, et al., 2019. Prediction of aerodynamic flow fields using convolutional neural networks. Comput Mech, 64(2):525–545. https://doi.org/10.1007/s00466-019-01740-0
Blazek J, 2015. Computational Fluid Dynamics: Principles and Applications (3rd Ed.). Butterworth-Heinemann, Oxford, UK, p.466.
Constantin P, Foias C, 1988. Navier—Stokes Equations. The University of Chicago Press, Chicago, IL, USA, p.199.
Daberkow DD, Mavris DN, 1998. New approaches to conceptual and preliminary aircraft design: a comparative assessment of a neural network formulation and a response surface methodology. World Aviation Congress & Exposition, Article 15. https://doi.org/10.4271/985509
Ernst MH, 1981. Nonlinear model-Boltzmann equations and exact solutions. Phys Rep, 78(1):1–171. https://doi.org/10.1016/0370-1573(81)90002-8
Farrashkhalvat M, Miles JP, 2003. Basic Structured Grid Generation: with an Introduction to Unstructured Grid Generation. Elsevier, Amsterdam, the Netherlands, p.190–226. https://doi.org/10.1016/B978-075065058-8/50008-3
Frankle J, Carbin M, 2019. The lottery ticket hypothesis: finding sparse, trainable neural networks. https://arxiv.org/abs/1803.03635v5
Geneva N, Zabaras N, 2019. Quantifying model form uncertainty in Reynolds-averaged turbulence models with Bayesian deep neural networks. J Comput Phys, 383:125–147. https://doi.org/10.1016/j.jcp.2019.01.021
Guastoni L, Guemes A, Ianiro A, et al., 2020. Convolutional-network models to predict wall-bounded turbulence from wall quantities. https://arxiv.org/abs/2006.12483
Guo XX, Li W, Iorio F, 2016. Convolutional neural networks for steady flow approximation. Proc 22nd ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining, p.481–490. https://doi.org/10.1145/2939672.2939738
Hamdan MKA, Rover DT, Darr MJ, et al., 2019. Mass estimation from images using deep neural network and sparse ground truth. http://arxiv.org/abs/1908.04387
Hu J, Shen L, Sun G, 2018. Squeeze-and-excitation networks. IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.7132–7141. https://doi.org/10.1109/CVPR.2018.00745
Isola P, Zhu JY, Zhou TH, et al., 2017. Image-to-image translation with conditional adversarial networks. IEEE Conf on Computer Vision and Pattern Recognition, p.5967–5976. https://doi.org/10.1109/CVPR.2017.632
Kim T, Cha M, Kim H, et al., 2017. Learning to discover cross-domain relations with generative adversarial networks. Proc 34th Int Conf on Machine Learning, p.1857–1865.
Lee S, You D, 2019. Data-driven prediction of unsteady flow over a circular cylinder using deep learning. J Fluid Mech, 879:217–254. https://doi.org/10.1017/jfm.2019.700
Li DL, Xu CF, Wang YX, et al., 2016. Parallelizing and optimizing large-scale 3D multi-phase flow simulations on the Tianhe-2 supercomputer. Concurr Comput, 28(5):1678–1692. https://doi.org/10.1002/cpe.3717
Ling JL, Kurzawski A, Templeton J, 2016. Reynolds averaged turbulence modelling using deep neural networks with embedded invariance. J Fluid Mech, 807:155–166. https://doi.org/10.1017/jfm.2016.615
Liu Z, Sun MJ, Zhou TH, et al., 2019. Rethinking the value of network pruning. https://arxiv.org/abs/1810.05270
Long J, Shelhamer E, Darrell T, 2015. Fully convolutional networks for semantic segmentation. IEEE Conf on Computer Vision and Pattern Recognition, p.3431–3440. https://doi.org/10.1109/CVPR.2015.7298965
Molchanov P, Tyree S, Karras T, et al., 2017. Pruning convolutional neural networks for resource efficient inference. Int Conf on Learning Representations.
Odena A, Dumoulin V, Olah C, 2016. Deconvolution and checkerboard artifacts. Distill, 1(10):e3. https://doi.org/10.23915/distill.00003
Park J, Woo S, Lee JY, et al., 2018. BAM: bottleneck attention module. https://arxiv.org/abs/1807.06514v1
Raissi M, Perdikaris P, Karniadakis GE, 2017. Physics informed deep learning (part I): data-driven solutions of nonlinear partial differential equations. https://arxiv.org/abs/1711.10561
Raissi M, Perdikaris P, Karniadakis GE, 2019. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving non-linear partial differential equations. J Comput Phys, 378:686–707. https://doi.org/10.1016/j.j.cp.2018.10.045
Ronneberger O, Fischer P, Brox T, 2015. U-Net: convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention, p.234–241. https://doi.org/10.1007/978-3-319-24574-4_28
Srinivasan PA, Guastoni L, Azizpour H, et al., 2019. Predictions of turbulent shear flows using deep neural networks. Phys Rev Fluids, 4:054603. https://doi.org/10.1103/PhysRevFluids.4.054603
Thuerey N, Weissenow K, Prantl L, et al., 2020. Deep learning methods for Reynolds-averaged Navier-ĺCStokes simulations of airfoil flows. AIAA J, 58(1):25–36. https://doi.org/10.2514/1.J058291
Wang R, Kashinath K, Mustafa M, et al., 2020. Towards physics-informed deep learning for turbulent flow prediction. Proc 26th ACM SIGKDD Int Conf on Knowledge Discovery & Data Mining, p.1457–1466. https://doi.org/10.1145/3394486.3403198
Woo S, Park J, Lee JY, et al., 2018. CBAM: convolutional block attention module. European Conf on Computer Vision, p.3–9. https://doi.org/10.1007/978-3-030-01234-2_1
Zhou ZW, Siddiquee MMR, Tajbakhsh N, et al., 2018. UNet++: a nested U-Net architecture for medical image segmentation. 4th Int Workshop on Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support, p.3–11. https://doi.org/10.1007/978-3-030-00889-5_1
Zhu JY, Park T, Isola P, et al., 2017. Unpaired image-to-image translation using cycle-consistent adversarial networks. IEEE Int Conf on Computer Vision, p.2242–2251. https://doi.org/10.1109/ICCV.2017.244
Author information
Authors and Affiliations
Contributions
Donglin CHEN and Xiang GAO designed the research. Siqi WANG and Shizhao CHEN processed the data. Chuanfu XU drafted the paper. Jianbin FANG and Zheng WANG helped organize the paper. Donglin CHEN and Xiang GAO revised and finalized the paper.
Corresponding authors
Ethics declarations
Donglin CHEN, Xiang GAO, Chuanfu XU, Siqi WANG, Shizhao CHEN, Jianbin FANG, and Zheng WANG declare that they have no conflict of interest.
Additional information
Project supported by the National Natural Science Foundation of China (Nos. 61772542, 61972408, and 12102467) and the Foundation of the State Key Laboratory of High Performance Computing, China (Nos. 201901-11 and 202001-03)
Rights and permissions
About this article
Cite this article
Chen, D., Gao, X., Xu, C. et al. FlowDNN: a physics-informed deep neural network for fast and accurate flow prediction. Front Inform Technol Electron Eng 23, 207–219 (2022). https://doi.org/10.1631/FITEE.2000435
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1631/FITEE.2000435