Good morning everyone ! I hope I posted the question in the right section. I’m a beginner with FEniCS and I’m trying to make this program work which should study the navier stokes equation by applying it to the case of a box with a cylinder inside. When I start the program the speed diverges and I c…

Looking at F1, the boundary integrals seem off to me. Usually, you substitute something in place of those integrals to enforce a condition on the traction (or v = 0 when you prescribe a Dirichlet condition on u). That could be a major source of instability. What happens when you remove those terms? …

I may be wrong, but I think this is Chorin’s projection method. There’s a section dedicated to it in @dokken ’s tutorial. This legacy DOLFIN code likely derives from an old tutorial, e.g., here .

And a better splitting scheme is described here: https://jsdokken.com/dolfinx-tutorial/chapter2/ns_code2.html with proper mathematical notes at: https://computationalphysiology.github.io/oasisx/splitting_schemes.html

I’m actually trying to apply Chorin’s projection method. I had found a 2D old example and i tried to modify it in 3D but with no success. Unfortunately I can’t find that exact tutorial. The things I changed are the mesh (originally created with mshr) and the part where I define inflow, outflow and w…

Yes exactly ! I noticed that more up-to-date tutorials, like the one you recommended, work with fenicsx. Do you recommend working with fenicsx instead of fenics? Sorry if I ask stupid questions but I’m really a beginner.

Yes, I do recommend fenicsx over fenics. See: [image] The new DOLFINx solver is now recommended over DOLFIN dolfinx In April 2022 we released versions of the new FEniCS Project components; DOLFINx (v0.4.1) , FFCx (v0.4.2), Basix (v0.4.1) and UFL (2022.1.0.post0). These pr…

Thank you very much, I will try fenicsx. Thanks again everyone for your replies!

Did you try removing the terms that I mentioned? All the sources given by @dokken and @nate on Chorins method do not have the boundary terms dot(p_n*n, v)*ds - dot(mu*grad(U)*n, v)*ds in the predictor step (your F1). For 99% of cases, those integrals should not be there. Unless you have a reason t…

Yes I tried to do as you say and that piece of F1 code was effectively useless. Even if you eliminate it, however, the error remains the same. Another thing I did was plot this navier_stokes_cylinder/velocity3D.xdmf with pyvista. But this comes out. Could it be useful to understand what’s wrong? Sc…

A screenshot without corresponding code doesn’t help

3D navier-stokes equation with divergent velocity

variational formulation

dokken November 21, 2024, 7:51pm 8

Yes, I do recommend fenicsx over fenics.
See:

for context

Topic		Replies	Views
Navier Stokes equiation 3D Uncategorized	6	1375	October 8, 2020
Boundary Conditions & Convergence for 3D Navier-Stokes Uncategorized	14	4936	June 26, 2020
"Problems with the simulated results" - Navier-Stokes equations in 3D General boundary-conditions	9	202	September 14, 2024
Navier Stokes equiation 3D(cylinder.) I/O	5	602	May 3, 2022
Convergence error for 3D Navier-Stokes after changing mesh mesh	9	1720	December 19, 2020

3D navier-stokes equation with divergent velocity

Related topics