Accelerating eulerian fluid simulation with convolutional networks

Jonathan Tompson, Kristofer Schlachter, Pablo Sprechmann, Kenneth Perlin

Research output: Contribution to conferencePaper

Abstract

Efficient simulation of the Navier-Stokes equations for fluid flow is a long standing problem in applied mathematics, for which state-of-the-art methods require large compute resources. In this work, we propose a data-driven approach that leverages the approximation power of deep-learning with the precision of standard solvers to obtain fast and highly realistic simulations. Our method solves the incompressible Euler equations using the standard operator splitting method, in which a large linear system with many free-parameters must be solved. We use a Convolutional Network with a highly tailored architecture, trained using a novel unsupervised learning framework to solve the linear system. We present real-time 2D and 3D simulations that outperform recently proposed data-driven methods; the obtained results are realistic and show good generalization properties.

Original languageEnglish (US)
StatePublished - Jan 1 2019
Event5th International Conference on Learning Representations, ICLR 2017 - Toulon, France
Duration: Apr 24 2017Apr 26 2017

Conference

Conference5th International Conference on Learning Representations, ICLR 2017
CountryFrance
CityToulon
Period4/24/174/26/17

Fingerprint

Linear systems
simulation
Unsupervised learning
Fluids
Euler equations
Navier Stokes equations
Flow of fluids
learning
mathematics
Simulation
resources
Deep learning
Equations
Data-driven

ASJC Scopus subject areas

  • Education
  • Computer Science Applications
  • Linguistics and Language
  • Language and Linguistics

Cite this

Tompson, J., Schlachter, K., Sprechmann, P., & Perlin, K. (2019). Accelerating eulerian fluid simulation with convolutional networks. Paper presented at 5th International Conference on Learning Representations, ICLR 2017, Toulon, France.

Accelerating eulerian fluid simulation with convolutional networks. / Tompson, Jonathan; Schlachter, Kristofer; Sprechmann, Pablo; Perlin, Kenneth.

2019. Paper presented at 5th International Conference on Learning Representations, ICLR 2017, Toulon, France.

Research output: Contribution to conferencePaper

Tompson, J, Schlachter, K, Sprechmann, P & Perlin, K 2019, 'Accelerating eulerian fluid simulation with convolutional networks' Paper presented at 5th International Conference on Learning Representations, ICLR 2017, Toulon, France, 4/24/17 - 4/26/17, .
Tompson J, Schlachter K, Sprechmann P, Perlin K. Accelerating eulerian fluid simulation with convolutional networks. 2019. Paper presented at 5th International Conference on Learning Representations, ICLR 2017, Toulon, France.
Tompson, Jonathan ; Schlachter, Kristofer ; Sprechmann, Pablo ; Perlin, Kenneth. / Accelerating eulerian fluid simulation with convolutional networks. Paper presented at 5th International Conference on Learning Representations, ICLR 2017, Toulon, France.
@conference{660f56a224104b7c8ca1348b638cd9fa,
title = "Accelerating eulerian fluid simulation with convolutional networks",
abstract = "Efficient simulation of the Navier-Stokes equations for fluid flow is a long standing problem in applied mathematics, for which state-of-the-art methods require large compute resources. In this work, we propose a data-driven approach that leverages the approximation power of deep-learning with the precision of standard solvers to obtain fast and highly realistic simulations. Our method solves the incompressible Euler equations using the standard operator splitting method, in which a large linear system with many free-parameters must be solved. We use a Convolutional Network with a highly tailored architecture, trained using a novel unsupervised learning framework to solve the linear system. We present real-time 2D and 3D simulations that outperform recently proposed data-driven methods; the obtained results are realistic and show good generalization properties.",
author = "Jonathan Tompson and Kristofer Schlachter and Pablo Sprechmann and Kenneth Perlin",
year = "2019",
month = "1",
day = "1",
language = "English (US)",
note = "5th International Conference on Learning Representations, ICLR 2017 ; Conference date: 24-04-2017 Through 26-04-2017",

}

TY - CONF

T1 - Accelerating eulerian fluid simulation with convolutional networks

AU - Tompson, Jonathan

AU - Schlachter, Kristofer

AU - Sprechmann, Pablo

AU - Perlin, Kenneth

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Efficient simulation of the Navier-Stokes equations for fluid flow is a long standing problem in applied mathematics, for which state-of-the-art methods require large compute resources. In this work, we propose a data-driven approach that leverages the approximation power of deep-learning with the precision of standard solvers to obtain fast and highly realistic simulations. Our method solves the incompressible Euler equations using the standard operator splitting method, in which a large linear system with many free-parameters must be solved. We use a Convolutional Network with a highly tailored architecture, trained using a novel unsupervised learning framework to solve the linear system. We present real-time 2D and 3D simulations that outperform recently proposed data-driven methods; the obtained results are realistic and show good generalization properties.

AB - Efficient simulation of the Navier-Stokes equations for fluid flow is a long standing problem in applied mathematics, for which state-of-the-art methods require large compute resources. In this work, we propose a data-driven approach that leverages the approximation power of deep-learning with the precision of standard solvers to obtain fast and highly realistic simulations. Our method solves the incompressible Euler equations using the standard operator splitting method, in which a large linear system with many free-parameters must be solved. We use a Convolutional Network with a highly tailored architecture, trained using a novel unsupervised learning framework to solve the linear system. We present real-time 2D and 3D simulations that outperform recently proposed data-driven methods; the obtained results are realistic and show good generalization properties.

UR - http://www.scopus.com/inward/record.url?scp=85071075559&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071075559&partnerID=8YFLogxK

M3 - Paper

ER -