Surrogate Modeling of Fluid Dynamics with a Multigrid Inspired Neural Network Architecture

Algebraic or geometric multigrid methods are commonly used in numerical solvers as they are a multi-resolution method able to handle problems with multiple scales. In this work, we propose a modification to the commonly-used U-Net neural network architecture that is inspired by the principles of multigrid methods, referred to here as U-Net-MG. We then demonstrate that this proposed U-Net-MG architecture can successfully reduce the test prediction errors relative to the conventional U-Net architecture when modeling a set of fluid dynamic problems. In total, we demonstrate an improvement in the prediction of velocity and pressure fields for the canonical fluid dynamics cases of flow past a stationary cylinder, flow past 2 cylinders in out-of-phase motion, and flow past an oscillating airfoil in both the propulsion and energy harvesting modes. In general, while both the U-Net and U-Net-MG models can model the systems well with test RMSEs of less than 1%, the use of the U-Net-MG architecture can further reduce RMSEs by between 20% and 70%.

[1]  Nikhil Muralidhar,et al.  PhyNet: Physics Guided Neural Networks for Particle Drag Force Prediction in Assembly , 2020, SDM.

[2]  Taesam Kang,et al.  Experimental and numerical study of a dual configuration for a flapping tidal current generator , 2015, Bioinspiration & biomimetics.

[3]  Yixing Wang,et al.  A novel spatial-temporal prediction method for unsteady wake flows based on hybrid deep neural network , 2019 .

[4]  Karthik Duraisamy,et al.  Prediction of aerodynamic flow fields using convolutional neural networks , 2019, Computational Mechanics.

[5]  Jennifer A. Franck,et al.  Variable thrust and high efficiency propulsion with oscillating foils at high Reynolds numbers , 2019, 1907.01097.

[6]  Lars Ruthotto,et al.  Learning Across Scales - Multiscale Methods for Convolution Neural Networks , 2018, AAAI.

[7]  Jin Hwan Ko,et al.  Effect of hydrofoil flexibility on the power extraction of a flapping tidal generator via two- and three-dimensional flow simulations , 2015 .

[8]  Louis J. Durlofsky,et al.  Deep-learning-based surrogate flow modeling and geological parameterization for data assimilation in 3D subsurface flow , 2020, Computer Methods in Applied Mechanics and Engineering.

[9]  Michael S. Triantafyllou,et al.  Forces on oscillating foils for propulsion and maneuvering , 2003 .

[10]  Thomas Brox,et al.  U-Net: deep learning for cell counting, detection, and morphometry , 2018, Nature Methods.

[11]  Nobuyuki Umetani,et al.  Learning three-dimensional flow for interactive aerodynamic design , 2018, ACM Trans. Graph..

[12]  Paris Perdikaris,et al.  Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations , 2019, J. Comput. Phys..

[13]  Stella X. Yu,et al.  Multigrid Neural Architectures , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[14]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[15]  Vladlen Koltun,et al.  An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling , 2018, ArXiv.

[16]  Nils Thuerey,et al.  Deep Learning Methods for Reynolds-Averaged Navier–Stokes Simulations of Airfoil Flows , 2018, AIAA Journal.

[17]  Multigrid-in-Channels Architectures for Wide Convolutional Neural Networks , 2020, ArXiv.

[18]  Ruohan Gong,et al.  Investigation of convolutional neural network U-net under small datasets in transformer magneto-thermal coupled analysis , 2020 .

[19]  Yao Zhang,et al.  Application of Convolutional Neural Network to Predict Airfoil Lift Coefficient , 2017, ArXiv.

[20]  R. Webster An algebraic multigrid solver for Navier‐Stokes problems , 1994 .

[21]  Anuj Karpatne,et al.  Physics-Guided Machine Learning for Scientific Discovery: An Application in Simulating Lake Temperature Profiles , 2020, Trans. Data Sci..

[22]  Y. Zhou,et al.  Flow structure behind two staggered circular cylinders. Part 1. Downstream evolution and classification , 2008, Journal of Fluid Mechanics.

[23]  Karthik Duraisamy,et al.  Towards Integrated Field Inversion and Machine Learning With Embedded Neural Networks for RANS Modeling , 2019, AIAA Scitech 2019 Forum.

[24]  Rajeev K. Jaiman,et al.  Interaction dynamics of gap flow with vortex-induced vibration in side-by-side cylinder arrangement , 2016 .

[25]  K. Taira,et al.  Super-resolution reconstruction of turbulent flows with machine learning , 2018, Journal of Fluid Mechanics.

[26]  Jinchao Xu,et al.  MgNet: A unified framework of multigrid and convolutional neural network , 2019, Science China Mathematics.

[27]  Nan Li,et al.  A study on using image based machine learning methods to develop the surrogate models of stamp forming simulations , 2020, Journal of Manufacturing Science and Engineering.

[28]  M. Carini,et al.  A novel approach for reconstructing pressure from PIV velocity measurements , 2015 .

[29]  Nagiza F. Samatova,et al.  Theory-Guided Data Science: A New Paradigm for Scientific Discovery from Data , 2016, IEEE Transactions on Knowledge and Data Engineering.

[30]  Quang Tuyen Le,et al.  Modeling transient fluid simulations with proper orthogonal decomposition and machine learning , 2020 .

[31]  Oddvar O. Bendiksen,et al.  Structures, Structural Dynamics and Materials Conference , 1998 .

[32]  T. Bohr,et al.  Wake structure and thrust generation of a flapping foil in two-dimensional flow , 2017 .

[33]  Wayne A. Smith,et al.  Implicit Solution of Preconditioned Navier- Stokes Equations Using Algebraic Multigrid , 1999 .

[34]  Thomas Brox,et al.  U-Net: Convolutional Networks for Biomedical Image Segmentation , 2015, MICCAI.

[35]  Brian J. German,et al.  A Convolutional Neural Network Approach to Training Predictors for Airfoil Performance , 2017 .

[36]  Junfeng Chen,et al.  U-net architectures for fast prediction of incompressible laminar flows , 2019, 1910.13532.

[37]  Weiwei Zhang,et al.  Deep neural network for unsteady aerodynamic and aeroelastic modeling across multiple Mach numbers , 2019, Nonlinear Dynamics.

[38]  Koji Fukagata,et al.  Machine-learning-based reduced-order modeling for unsteady flows around bluff bodies of various shapes , 2020, Theoretical and Computational Fluid Dynamics.

[39]  Daniël M Pelt,et al.  A mixed-scale dense convolutional neural network for image analysis , 2017, Proceedings of the National Academy of Sciences.

[40]  Z. Ibrahim,et al.  Numerical investigation of the vortex-induced vibration of an elastically mounted circular cylinder at high Reynolds number (Re = 104) and low mass ratio using the RANS code , 2017, PloS one.

[41]  Q. Le,et al.  Projection-based and neural-net reduced order model for nonlinear Navier–Stokes equations , 2021 .

[42]  Seyed-Ahmad Ahmadi,et al.  V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation , 2016, 2016 Fourth International Conference on 3D Vision (3DV).

[43]  Karthik Duraisamy,et al.  A paradigm for data-driven predictive modeling using field inversion and machine learning , 2016, J. Comput. Phys..

[44]  Di Zhang Comparison of Various Turbulence Models for Unsteady Flow around a Finite Circular Cylinder at Re=20000 , 2017 .

[45]  Boo Cheong Khoo,et al.  Fast flow field prediction over airfoils using deep learning approach , 2019, Physics of Fluids.

[46]  F. Huera-Huarte,et al.  Effect of diameter ratio on the flow-induced vibrations of two rigidly coupled circular cylinders in tandem , 2019, Journal of Fluids and Structures.

[47]  Hui Li,et al.  Prediction model of velocity field around circular cylinder over various Reynolds numbers by fusion convolutional neural networks based on pressure on the cylinder , 2018 .

[48]  Paris Perdikaris,et al.  Physics-Constrained Deep Learning for High-dimensional Surrogate Modeling and Uncertainty Quantification without Labeled Data , 2019, J. Comput. Phys..

[49]  D. Rockwell,et al.  VORTEX FORMATION FROM A FORCED SYSTEM OF TWO CYLINDERS. PART II: SIDE-BY-SIDE ARRANGEMENT , 1996 .

[50]  Ray W. Grout,et al.  Data recovery in computational fluid dynamics through deep image priors , 2019 .

[51]  J. Templeton,et al.  Reynolds averaged turbulence modelling using deep neural networks with embedded invariance , 2016, Journal of Fluid Mechanics.

[52]  Xudong Song,et al.  A flow feature detection method for modeling pressure distribution around a cylinder in non-uniform flows by using a convolutional neural network , 2020, Scientific Reports.

[53]  Donghyun You,et al.  Data-driven prediction of unsteady flow over a circular cylinder using deep learning , 2018, Journal of Fluid Mechanics.

[54]  Changhoon Lee,et al.  Deep unsupervised learning of turbulence for inflow generation at various Reynolds numbers , 2020, J. Comput. Phys..

[55]  A. Incecik,et al.  The effect of spacing on the vortex-induced vibrations of two tandem flexible cylinders , 2017 .

[56]  Eric Mjolsness,et al.  Multilevel Artificial Neural Network Training for Spatially Correlated Learning , 2018, SIAM J. Sci. Comput..

[57]  T. Kinsey,et al.  Parametric Study of an Oscillating Airfoil in a Power-Extraction Regime , 2008 .

[58]  B. W. Oudheusden,et al.  PIV-based pressure measurement , 2013 .

[59]  R. D. Lonsdale,et al.  AN ALGEBRAIC MULTIGRID SOLVER FOR THE NAVIER—STOKES EQUATIONS ON UNSTRUCTURED MESHES , 1993 .