Learned turbulence modelling with differentiable fluid solvers: physics-based loss functions and optimisation horizons
From MaRDI portal
Publication:5038552
DOI10.1017/JFM.2022.738zbMATH Open1498.76049arXiv2202.06988OpenAlexW4298000212MaRDI QIDQ5038552FDOQ5038552
Authors:
Publication date: 7 October 2022
Published in: Journal of Fluid Mechanics (Search for Journal in Brave)
Abstract: In this paper, we train turbulence models based on convolutional neural networks. These learned turbulence models improve under-resolved low resolution solutions to the incompressible Navier-Stokes equations at simulation time. Our study involves the development of a differentiable numerical solver that supports the propagation of optimisation gradients through multiple solver steps. The significance of this property is demonstrated by the superior stability and accuracy of those models that unroll more solver steps during training. Furthermore, we introduce loss terms based on turbulence physics that further improve the model accuracy. This approach is applied to three two-dimensional turbulence flow scenarios, a homogeneous decaying turbulence case, a temporally evolving mixing layer, and a spatially evolving mixing layer. Our models achieve significant improvements of long-term a-posteriori statistics when compared to no-model simulations, without requiring these statistics to be directly included in the learning targets. At inference time, our proposed method also gains substantial performance improvements over similarly accurate, purely numerical methods.
Full work available at URL: https://arxiv.org/abs/2202.06988
Recommendations
- Physics-guided deep learning for generating turbulent inflow conditions
- Deep learning observables in computational fluid dynamics
- Reynolds averaged turbulence modelling using deep neural networks with embedded invariance
- Learning by neural networks under physical constraints for simulation in fluid mechanics
- Physics-Driven Learning of the Steady Navier-Stokes Equations using Deep Convolutional Neural Networks
- Physics-informed neural networks for the Reynolds-averaged Navier-Stokes modeling of Rayleigh-Taylor turbulent mixing
- Deep reinforcement learning of viscous incompressible flow
- Operator inference and physics-informed learning of low-dimensional models for incompressible flows
- Deep learning of the spanwise-averaged Navier-Stokes equations
machine learningincompressible Navier-Stokes equationsconvolutional neural networktwo-dimensional homogeneous decaying turbulenceunder-resolved low-resolution solution
Learning and adaptive systems in artificial intelligence (68T05) Navier-Stokes equations for incompressible viscous fluids (76D05) Turbulence (76F99) Basic methods in fluid mechanics (76M99) Mathematical modeling or simulation for problems pertaining to fluid mechanics (76-10)
Cites Work
- Turbulent Flows
- High-Re solutions for incompressible flow using the Navier-Stokes equations and a multigrid method
- Solution of the implicitly discretized fluid flow equations by operator- splitting
- Comparison of the SMAC, PISO and iterative time-advancing schemes for unsteady flows
- Direct simulation of a self-similar turbulent mixing layer
- Sensitivity of two-dimensional spatially developing mixing layers with respect to uncertain inflow conditions
- Split energy-helicity cascades in three-dimensional homogeneous and isotropic turbulence
- On the inviscid instability of the hyperbolictangent velocity profile
- Title not available (Why is that?)
- Turbulence modeling in the age of data
- Numerical simulation of developing and decaying two-dimensional turbulence
- Reynolds averaged turbulence modelling using deep neural networks with embedded invariance
- Recent advances on the numerical modelling of turbulent flows
- A dynamic eddy-viscosity closure model for large eddy simulations of two-dimensional decaying turbulence
- High-order methods for decaying two-dimensional homogeneous isotropic turbulence
- Neural networks based subgrid scale modeling in large eddy simulations
- Subgrid modelling for two-dimensional turbulence using neural networks
- Toward neural-network-based large eddy simulation: application to turbulent channel flow
- Prediction of aerodynamic flow fields using convolutional neural networks
- DPM: a deep learning PDE augmentation method with application to large-eddy simulation
Cited In (13)
- A probabilistic, data-driven closure model for RANS simulations with aleatoric, model uncertainty
- A physics-informed learning approach to Bernoulli-type free boundary problems
- Towards high-accuracy deep learning inference of compressible flows over aerofoils
- Dynamic calibration of differential equations using machine learning, with application to turbulence models
- A multifidelity deep operator network approach to closure for multiscale systems
- Differentiable hybrid neural modeling for fluid-structure interaction
- Unsupervised deep learning for super-resolution reconstruction of turbulence
- Comparison of neural closure models for discretised PDEs
- Differentiability in unrolled training of neural physics simulators on transient dynamics
- Neural differentiable modeling with diffusion-based super-resolution for two-dimensional spatiotemporal turbulence
- Super-resolution of turbulence with dynamics in the loss
- Combining direct and indirect sparse data for learning generalizable turbulence models
- Ensemble Gradient for Learning Turbulence Models from Indirect Observations
Uses Software
This page was built for publication: Learned turbulence modelling with differentiable fluid solvers: physics-based loss functions and optimisation horizons
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5038552)
