NSFnets (Navier-Stokes flow nets): Physics-informed neural networks for the incompressible Navier-Stokes equations

•NSFnets involve the VP and VV formulations of the Navier-Stokes equations.•NSFnets can directly simulate and sustain turbulence at Reτ∼1,000.•A study on the weights in the loss function to enhance accuracy is performed.•Transfer learning of NSFnets can reduce computational cost and enhance accuracy...

Full description

Saved in:

Bibliographic Details
Published in:	Journal of computational physics Vol. 426; no. C; p. 109951
Main Authors:	Jin, Xiaowei, Cai, Shengze, Li, Hui, Karniadakis, George Em
Format:	Journal Article
Language:	English
Published:	Cambridge Elsevier Inc 01.02.2021 Elsevier Science Ltd Elsevier
Subjects:	Accuracy Algorithms Artificial neural networks Boundary conditions Channel flow Computational efficiency Computational fluid dynamics Computational physics Computer simulation Computing costs Convergence Cost analysis Direct numerical simulation Domain names Finite element method Fluid flow Ill-posed problems Incompressibility Incompressible flow Inverse problems Laminar flow Machine learning Navier-Stokes equations Neural networks Optimization Physics PINNs Training Transfer learning Turbulence Turbulent flow Velocity Velocity-pressure formulation Vorticity Vorticity-velocity formulation Turbulence Ill-posed problems PINNs Velocity-pressure formulation Transfer learning Vorticity-velocity formulation
ISSN:	0021-9991, 1090-2716
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Be the first to leave a comment!