Constraint-aware neural networks for Riemann problems
From MaRDI portal
Publication:778316
DOI10.1016/J.JCP.2020.109345zbMATH Open1435.76046arXiv1904.12794OpenAlexW2940929008MaRDI QIDQ778316FDOQ778316
Authors: Jim Magiera, Deep Ray, Ch. Rohde, Jan S. Hesthaven 
Publication date: 2 July 2020
Published in: Journal of Computational Physics (Search for Journal in Brave)
Abstract: Neural networks are increasingly used in complex (data-driven) simulations as surrogates or for accelerating the computation of classical surrogates. In many applications physical constraints, such as mass or energy conservation, must be satisfied to obtain reliable results. However, standard machine learning algorithms are generally not tailored to respect such constraints. We propose two different strategies to generate constraint-aware neural networks. We test their performance in the context of front-capturing schemes for strongly nonlinear wave motion in compressible fluid flow. Precisely, in this context so-called Riemann problems have to be solved as surrogates. Their solution describes the local dynamics of the captured wave front in numerical simulations. Three model problems are considered: a cubic flux model problem, an isothermal two-phase flow model, and the Euler equations. We demonstrate that a decrease in the constraint deviation correlates with low discretization errors for all model problems, in addition to the structural advantage of fulfilling the constraint.
Full work available at URL: https://arxiv.org/abs/1904.12794
Recommendations
- Neural Networks with Local Converging Inputs (NNLCI) for Solving Conservation Laws, Part I: 1D Problems
- Conservative physics-informed neural networks on discrete domains for conservation laws: applications to forward and inverse problems
- The neural particle method - an updated Lagrangian physics informed neural network for computational fluid dynamics
- Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data
- scientific article; zbMATH DE number 5174741
artificial neural networksconstraintstwo-phase flowcomputational fluid dynamicsRiemann solverfront-capturing
Artificial neural networks and deep learning (68T07) Hyperbolic conservation laws (35L65) Finite volume methods applied to problems in fluid mechanics (76M12)
Cites Work
- Deep learning
- Discovering governing equations from data by sparse identification of nonlinear dynamical systems
- Riemann Solvers and Numerical Methods for Fluid Dynamics
- A survey of several finite difference methods for systems of nonlinear hyperbolic conservation laws
- A non-oscillatory Eulerian approach to interfaces in multimaterial flows (the ghost fluid method)
- The Riemann Problem for General 2 × 2 Conservation Laws
- Title not available (Why is that?)
- Hyperbolic Conservation Laws in Continuum Physics
- Divergence-free RBFs on surfaces
- Non-classical shocks and kinetic relations: Scalar conservation laws
- Title not available (Why is that?)
- Computing undercompressive waves with the random choice scheme. Nonclassical shock waves
- Positive constrained approximation via RBF-based partition of unity method
- The Riemann problem for a simple model of phase transition
- A finite volume method for undercompressive shock waves in two space dimensions
- Approximate Riemann solver for compressible liquid vapor flow with phase transition and surface tension
- On Riemann solvers and kinetic relations for isothermal two-phase flows with surface tension
- A particle-based multiscale solver for compressible liquid-vapor flow
- Neural-net-induced Gaussian process regression for function approximation and PDE solution
- An artificial neural network as a troubled-cell indicator
- The computation of nonclassical shock waves in porous media with a heterogeneous multiscale method: the multidimensional case
- Data-driven science and engineering. Machine learning, dynamical systems, and control
- A conservative and convergent scheme for undercompressive shock waves
Cited In (28)
- Physics-informed neural networks for high-speed flows
- Continuous limits of residual neural networks in case of large input data
- Bilevel optimization, deep learning and fractional Laplacian regularization with applications in tomography
- Deep learning of first-order nonlinear hyperbolic conservation law solvers
- Physics-informed neural networks for inverse problems in supersonic flows
- Maximum principle preserving space and time flux limiting for diagonally implicit Runge-Kutta discretizations of scalar convection-diffusion equations
- Three ways to solve partial differential equations with neural networks — A review
- A molecular-continuum multiscale model for inviscid liquid-vapor flow with sharp interfaces
- A probabilistic generative model for semi-supervised training of coarse-grained surrogates and enforcing physical constraints through virtual observables
- A data-driven physics-informed finite-volume scheme for nonclassical undercompressive shocks
- Learning to Discretize: Solving 1D Scalar Conservation Laws via Deep Reinforcement Learning
- A multiscale method for two-component, two-phase flow with a neural network surrogate
- Deep learning in computational mechanics: a review
- Learning an optimised stable Taylor-Galerkin convection scheme based on a local spectral model for the numerical error dynamics
- Neural Networks with Local Converging Inputs (NNLCI) for Solving Conservation Laws, Part I: 1D Problems
- Structure-preserving neural networks
- A multi-stage deep learning based algorithm for multiscale model reduction
- Small collaboration: Advanced numerical methods for nonlinear hyperbolic balance laws and their applications. Abstracts from the small collaboration held August 29 -- September 4, 2021 (hybrid meeting)
- Discontinuity computing using physics-informed neural networks
- Neural Networks with Local Converging Inputs (NNLCI) for Solving Conservation Laws, Part II: 2D Problems
- Learning the flux and diffusion function for degenerate convection-diffusion equations using different types of observations
- FluxNet: a physics-informed learning-based Riemann solver for transcritical flows with non-ideal thermodynamics
- Enforcing imprecise constraints on generative adversarial networks for emulating physical systems
- Unsupervised neural-network solvers for multi-material Riemann problems
- Solution of physics-based Bayesian inverse problems with deep generative priors
- Thermodynamically consistent physics-informed neural networks for hyperbolic systems
- Addressing discontinuous root-finding for subsequent differentiability in machine learning, inverse problems, and control
- An interface treatment for two-material multi-species flows involving thermally perfect gases with chemical reactions
Uses Software
This page was built for publication: Constraint-aware neural networks for Riemann problems
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q778316)
