Machine learning accelerated computational fluid dynamics

Abstract

Numerical simulation of fluids plays an essential role in modeling many physical phenomena, such as weather, climate, aerodynamics and plasma physics. Fluids are well described by the Navier-Stokes equations, but solving these equations at scale remains daunting, limited by the computational cost of resolving the smallest spatiotemporal features. This leads to unfavorable trade-offs between accuracy and tractability. Here we use end-to-end deep learning to improve approximations inside computational fluid dynamics for modeling two-dimensional turbulent flows. For both direct numerical simulation of turbulence and large eddy simulation, our results are as accurate as baseline solvers with 8-10x finer resolution in each spatial dimension, resulting in 40-80x fold computational speedups. Our method remains stable during long simulations, and generalizes to forcing functions and Reynolds numbers outside of the flows where it is trained, in contrast to black box machine learning approaches. Our approach exemplifies how scientific computing can leverage machine learning and hardware accelerators to improve simulations without sacrificing accuracy or generalization.

13 pages, 9 figures

References in corpus (11)

• Learning to Simulate Complex Physics with Graph Networks
• Kohn-Sham equations as regularizer: building prior knowledge into machine-learned physics
• Neural Operator: Graph Kernel Network for Partial Differential Equations
• Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers
• CFDNet: a deep learning-based accelerator for fluid simulations
• Physics-informed Autoencoders for Lyapunov-stable Fluid Flow Prediction
• Learned discretizations for passive scalar advection in a 2-D turbulent flow
• Using Machine Learning to Augment Coarse-Grid Computational Fluid Dynamics Simulations
• Large-Scale Discrete Fourier Transform on TPUs
• Inundation Modeling in Data Scarce Regions
• MomentumRNN: Integrating Momentum into Recurrent Neural Networks

Cited by in corpus (66)

• Fourier Neural Operator for Parametric Partial Differential Equations
• Enhancing Computational Fluid Dynamics with Machine Learning
• Differentiable modeling to unify machine learning and physical models and advance Geosciences
• Machine Learning in Aerodynamic Shape Optimization
• Transfer learning based physics-informed neural networks for solving inverse problems in engineering structures under different loading scenarios
• A Physics-informed Diffusion Model for High-fidelity Flow Field Reconstruction
• Improving aircraft performance using machine learning: a review
• Super-Resolution Analysis via Machine Learning: A Survey for Fluid Flows
• Bridging observation, theory and numerical simulation of the ocean using Machine Learning
• Physics-Informed Neural Operator for Learning Partial Differential Equations
• A Review of Machine Learning Methods Applied to Structural Dynamics and Vibroacoustic
• SocialVAE: Human Trajectory Prediction using Timewise Latents
• JAX-FEM: A differentiable GPU-accelerated 3D finite element solver for automatic inverse design and mechanistic data science
• JAX-FLUIDS: A fully-differentiable high-order computational fluid dynamics solver for compressible two-phase flows
• Graph neural networks for laminar flow prediction around random 2D shapes
• Physics-constrained, low-dimensional models for MHD: First-principles and data-driven approaches
• 2022 Review of Data-Driven Plasma Science
• Multi-resolution partial differential equations preserved learning framework for spatiotemporal dynamics
• Learned Turbulence Modelling with Differentiable Fluid Solvers: Physics-based Loss-functions and Optimisation Horizons
• Attention-Enhanced Neural Network Models for Turbulence Simulation
• A posteriori learning for quasi-geostrophic turbulence parametrization
• Learning physics-constrained subgrid-scale closures in the small-data regime for stable and accurate LES
• Deep reinforcement learning for large-eddy simulation modeling in wall-bounded turbulence
• MAgNET: A Graph U-Net Architecture for Mesh-Based Simulations
• Fast reactive flow simulations using analytical Jacobian and dynamic load balancing in OpenFOAM
• Explaining the physics of transfer learning a data-driven subgrid-scale closure to a different turbulent flow
• DropTrack -- automatic droplet tracking using deep learning for microfluidic applications
• Physics-enhanced deep surrogates for partial differential equations
• Neural Galerkin Schemes with Active Learning for High-Dimensional Evolution Equations
• Physical Activation Functions (PAFs): An Approach for More Efficient Induction of Physics into Physics-Informed Neural Networks (PINNs)
• Flow Completion Network: Inferring the Fluid Dynamics from Incomplete Flow Information using Graph Neural Networks
• Finite volume method network for acceleration of unsteady computational fluid dynamics: non-reacting and reacting flows
• Reinforcement-learning-based actuator selection method for active flow control
• Learning Large-scale Subsurface Simulations with a Hybrid Graph Network Simulator
• Stability analysis of chaotic systems from data
• Differentiable Programming for Earth System Modeling
• Lettuce: PyTorch-based Lattice Boltzmann Framework
• Accelerating high order discontinuous Galerkin solvers using neural networks: 3D compressible Navier-Stokes equations
• Large Scale Distributed Linear Algebra With Tensor Processing Units
• DSPS: Differentiable Stellar Population Synthesis
• Learned coupled inversion for carbon sequestration monitoring and forecasting with Fourier neural operators
• Deep Learning of Conjugate Mappings
• Hydrodynamic coupling for particle-based solvent-free membrane models
• dPV: An End-to-End Differentiable Solar-Cell Simulator
• Model-assisted deep learning of rare extreme events from partial observations
• A general approach for running Python codes in OpenFOAM using an embedded pybind11 Python interpreter
• Local neural operator for solving transient partial differential equations on varied domains
• Global Optimization via Optimal Decision Trees
• Validation and parameterization of a novel physics-constrained neural dynamics model applied to turbulent fluid flow
• Black Hole Weather Forecasting with Deep Learning: A Pilot Study
• AD-NEGF: An End-to-End Differentiable Quantum Transport Simulator for Sensitivity Analysis and Inverse Problems
• Deep Neural Networks to Correct Sub-Precision Errors in CFD
• Variational Data Assimilation with a Learned Inverse Observation Operator
• Promoting global stability in data-driven models of quadratic nonlinear dynamics
• Exploring Electron Beam Induced Atomic Assembly via Reinforcement Learning in a Molecular Dynamics Environment
• Rapidly Encoding Generalizable Dynamics in a Euclidean Symmetric Neural Network
• Towards high-accuracy deep learning inference of compressible turbulent flows over aerofoils
• Physics-Guided Discovery of Highly Nonlinear Parametric Partial Differential Equations
• A Model-Constrained Tangent Slope Learning Approach for Dynamical Systems
• Physics-based Deep Learning
• Geometry encoding for numerical simulations
• Resistance-Time Co-Modulated PointNet for Temporal Super-Resolution Simulation of Blood Vessel Flows
• Gradients are Not All You Need
• rom.js/cfd.xyz: An open-source framework for generating and visualizing parametric CFD results
• Data-driven detection of drifting system parameters
• Physics-Aware Downsampling with Deep Learning for Scalable Flood Modeling