CLC number: TP391
On-line Access: 2022-02-28
Received: 2020-08-28
Revision Accepted: 2022-04-22
Crosschecked: 2021-05-04
Cited: 0
Clicked: 6184
Citations: Bibtex RefMan EndNote GB/T7714
Donglin CHEN, Xiang GAO, Chuanfu XU, Siqi WANG, Shizhao CHEN, Jianbin FANG, Zheng WANG. FlowDNN: a physics-informed deep neural network for fast and accurate flow prediction[J]. Frontiers of Information Technology & Electronic Engineering,in press.https://doi.org/10.1631/FITEE.2000435 @article{title="FlowDNN: a physics-informed deep neural network for fast and accurate flow prediction", %0 Journal Article TY - JOUR
FlowDNN:一种用于快速精确流场预测的物理启发深度神经网络1国防科技大学计算机学院,中国长沙市,410073 2国防科技大学高性能计算国家重点实验室,中国长沙市,410073 3利兹大学计算学院,英国利兹市,LS29JT 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。 关键词组: Darkslateblue:Affiliate; Royal Blue:Author; Turquoise:Article
Reference[1]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. doi: 10.21608/ASAT.2009.23442 [2]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. doi: 10.1109/CVPR.2019.00919 [3]Balabanov VO, Giunta AA, Golovidov O, et al., 1999. Reasonable design space approach to response surface approximation. J Aircr, 36(1):308-315. doi: 10.2514/2.2438 [4]Bhatnagar S, Afshar Y, Pan S, et al., 2019. Prediction of aerodynamic flow fields using convolutional neural networks. Comput Mech, 64(2):525-545. doi: 10.1007/s00466-019-01740-0 [5]Blazek J, 2015. Computational Fluid Dynamics: Principles and Applications (3rd Ed.). Butterworth-Heinemann, Oxford, UK, p.466. [6]Constantin P, Foias C, 1988. Navier–Stokes Equations. The University of Chicago Press, Chicago, IL, USA, p.199. [7]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. doi: 10.4271/985509 [8]Ernst MH, 1981. Nonlinear model-Boltzmann equations and exact solutions. Phys Rep, 78(1):1-171. doi: 10.1016/0370-1573(81)90002-8 [9]Farrashkhalvat M, Miles JP, 2003. Basic Structured Grid Generation: with an Introduction to Unstructured Grid Generation. Elsevier, Amsterdam, the Netherlands, p.190-226. doi: 10.1016/B978-075065058-8/50008-3 [10]Frankle J, Carbin M, 2019. The lottery ticket hypothesis: finding sparse, trainable neural networks. https://arxiv.org/abs/1803.03635v5 [11]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. doi: 10.1016/j.jcp.2019.01.021 [12]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 [13]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. doi: 10.1145/2939672.2939738 [14]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 [15]Hu J, Shen L, Sun G, 2018. Squeeze-and-excitation networks. IEEE/CVF Conf on Computer Vision and Pattern Recognition, p.7132-7141. doi: 10.1109/CVPR.2018.00745 [16]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. doi: 10.1109/CVPR.2017.632 [17]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. [18]Lee S, You D, 2019. Data-driven prediction of unsteady flow over a circular cylinder using deep learning. J Fluid Mech, 879:217-254. doi: 10.1017/jfm.2019.700 [19]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. doi: 10.1002/cpe.3717 [20]Ling JL, Kurzawski A, Templeton J, 2016. Reynolds averaged turbulence modelling using deep neural networks with embedded invariance. J Fluid Mech, 807:155-166. doi: 10.1017/jfm.2016.615 [21]Liu Z, Sun MJ, Zhou TH, et al., 2019. Rethinking the value of network pruning. https://arxiv.org/abs/1810.05270 [22]Long J, Shelhamer E, Darrell T, 2015. Fully convolutional networks for semantic segmentation. IEEE Conf on Computer Vision and Pattern Recognition, p.3431-3440. doi: 10.1109/CVPR.2015.7298965 [23]Molchanov P, Tyree S, Karras T, et al., 2017. Pruning convolutional neural networks for resource efficient inference. Int Conf on Learning Representations. [24]Odena A, Dumoulin V, Olah C, 2016. Deconvolution and checkerboard artifacts. Distill, 1(10):e3. doi: 10.23915/distill.00003 [25]Park J, Woo S, Lee JY, et al., 2018. BAM: bottleneck attention module. https://arxiv.org/abs/1807.06514v1 [26]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 [27]Raissi M, Perdikaris P, Karniadakis GE, 2019. Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations. J Comput Phys, 378:686-707. doi: 10.1016/j.jcp.2018.10.045 [28]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. doi: 10.1007/978-3-319-24574-4_28 [29]Srinivasan PA, Guastoni L, Azizpour H, et al., 2019. Predictions of turbulent shear flows using deep neural networks. Phys Rev Fluids, 4:054603. doi: 10.1103/PhysRevFluids.4.054603 [30]Thuerey N, Weissenow K, Prantl L, et al., 2020. Deep learning methods for Reynolds-averaged Navier-ÍCStokes simulations of airfoil flows. AIAA J, 58(1):25-36. doi: 10.2514/1.J058291 [31]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. doi: 10.1145/3394486.3403198 [32]Woo S, Park J, Lee JY, et al., 2018. CBAM: convolutional block attention module. European Conf on Computer Vision, p.3-9. doi: 10.1007/978-3-030-01234-2_1 [33]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. doi: 10.1007/978-3-030-00889-5_1 [34]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. doi: 10.1109/ICCV.2017.244 Journal of Zhejiang University-SCIENCE, 38 Zheda Road, Hangzhou
310027, China
Tel: +86-571-87952783; E-mail: cjzhang@zju.edu.cn Copyright © 2000 - 2024 Journal of Zhejiang University-SCIENCE |
Open peer comments: Debate/Discuss/Question/Opinion
<1>