Hi Shujie,


I am not very familiar with the usage of NekMesh, but one reason that may lead to negative Jacobian is that the mesh near the wall is too coarse. This means one element cut across large radian and easily leads to negative Jacobian. I believe Joaquim has developed some techniques to tackle with this problem. But the easy solution is to use finer mesh where the curvature is large. I think this is also Gicomo's first suggestion.


Best wishes,

Zhenguo



On 08/07/2019 11:34, Jiang S. wrote:

Dear Giacomo,


I tried to use NekMesh to generate the mesh, there are two problems:


  1. 3D deformed Jacobian not positive in some elements;
  2. Periodic composites have a different number of elements;

Is there any way to let the periodic surface have the same element number?

Thanks!
Shujie
发件人: Castiglioni, Giacomo <g.castiglioni@imperial.ac.uk>
发送时间: 2019年7月8日 10:30
收件人: Jiang S.; nektar-users; d.moxey@exeter.ac.uk; Yan, Zhenguo
抄送: Sherwin, Spencer J
主题: Re: About the periodic boundary condition on quasi-3D simulation
 

Dear Shujie,

I have used the peralign module many times with success, so I wonder if the problem comes from somewhere else.


A couple of suggestions:
- I tried to generate the mesh from your .geo file and I think the mesh is too coarse to get the wavy features resolved (I get several warnings of this sort "Warning: degenerate tetrahedron")
- with the step file you can generate the mesh directly with NekMesh without having to use Gmsh.

Cheers,

Giacomo

From: Jiang S. <Shujie.Jiang@soton.ac.uk>
Sent: Monday, July 8, 2019 10:05:24 AM
To: Castiglioni, Giacomo; nektar-users; d.moxey@exeter.ac.uk; Yan, Zhenguo
Cc: Sherwin, Spencer J
Subject: About the periodic boundary condition on quasi-3D simulation
 

Hi All,


I use Gmsh to generate a mesh for quasi-3D simulation. I want to use periodic boundary conditions in the 'left' and 'right' side.


But the Nektar++ comes with the error:

   

Fatal   : Level 0 assertion violation
Unable to determine triangle orientation

So I tried to change the mesh normal direction in Gmsh. Then the same error comes again. 
At last, I tried to use peralign in NekMesh, it generated a new mesh, but the same error comes again.

Actually, I tried to change the periodic boundary condition to wall boundary condition. It runs successfully.

Is there something wrong with my mesh?


Many thanks for your help!
Have a nice day!
Shujie