Hello ! I am trying to solve the incompressible and invsicid flow in FreeFEM for a 3D geometry. I have written the code to work in parallel using PETSc. The mesh is imported from GMSH mesh file. I am not sure what is wrong with my code as the solution is not correct. I thought the dt might be a problem but changing dt doesn’t affect the result. The weak form of the GE I used, [GE] Boundary Conditions: Dirchlet B.C. for velocity at inlet. Neumaan B.C. for pressure at outlet Periodic B.C. on the top, bottom, front and back walls of the domain. varf formulation of the problem varf GE([u1,u2,u3,p],[v1,v2,v3,q]) =int3d(Th)(u1*v1 + u2*v2 + u3*v3 +(un1*dx(u1) + un2*dy(u1) + un3*dz(u1))*v1*dt +(un1*dx(u2) + un2*dy(u2) + un3*dz(u2))*v2*dt +(un1*dx(u3) + un2*dy(u3) + un3*dz(u3))*v3*dt -(dx(v1) + dy(v2) + dz(v3))*p*dt -(dx(u1) + dy(u2) + dz(u3))*q*dt) +int3d(Th)(un1*v1 + un2*v2 + un3*v3) +on(1, u1=1, u2=0, u3=0) +int2d(Th, 2)(0*q); The edp file for your reference. cyl_new.edp (2.5 KB) The velocity field produced from the code. It doesn’t look correct. [At t=0.5 s] [At t=1.5 s] Thank you for the help ! Regards, Gaurav.

Thanks a lot @cmd and @fb77 for your time and advices ! I think we can close the discussion now since the issue is resolved. The updated code in on GitHub along with the updated report.

Thanks a lot for looking into. Is it okay if put U•n =0 on all the boundaries except the inlet ?

Assuming you mean all boundaries except inlet AND outlet, yes. but that is not trivial with curved boundaries. What are you trying to do? I thought you wanted periodic boundaries?

Yeah with curved boundary, it is any issue. I have periodic bc on the other boundaries except the cylinder boundary, inlet and outlet, does that prevent the penetration. Any help on implementing the no penetration bc for general boundaries ?

You could just solve Poisson’s equation for the complex potential; then, obtain the inviscid velocity via differentiation.

Stream function method doesn’t work for 3D. I have 3D application.

See the notes of @fb77 here: [image] Boundary condition General Discussion If you want to avoid using a Lagrange multiplier (because it enlarges the matrix to invert), some other ways to penalize the normal boundary conditions are possible. In 3D you can use for example the boundary vertices. You have to define first a continuous approximate normal on the boundary (label 1 is for boundary) fespace Vh1(Th,[P2,P2,P2]); varf bdryn([w1,w2,w3],[v1,v2,v3])=int2d(Th,1)(v1*N.x+v2*N.y+v3*N.z); Vh1 [normalappx,normalappy,normalappz]; normalappx[]=bdryn(0,Vh1); Then you can ad…

Hello, The no penetration boundary conditions is one possible issue. I think that another one is the implicit treatment of convection without viscosity. I would recommend using convect().

I had been using convect() earlier but it is not recommended with periodic boundary conditions. I read one discussion about it in the community. So, I switched to this method.

I understand your reason, but I think that the implicit method is unstable here. You could try with convect and no penetration condition instead of periodic, to see if it works or not.

Sure, I will check once. Thank you for the help !

Inviscid Incompressible Flow in FreeFEM

General Discussion

cmd (Chris) June 4, 2024, 10:56am 15

See the notes of @fb77 here:

Topic		Replies	Views
Streamlines using the Elmer method General Discussion	10	291	July 25, 2024
Get the weak form help General Discussion	25	1256	September 2, 2023
Compressible turbulent Navier–Stokes around a sphere (half-sphere, cylindrical coordinates) — strange field despite BCs implemented General Discussion	6	274	September 26, 2025
Third derivative General Discussion	18	990	June 20, 2024
Boundary condition General Discussion	19	1212	April 25, 2024

Inviscid Incompressible Flow in FreeFEM

Related topics