Grid or computation problem with linear elasticity module

Numerical methods and mathematical models of Elmer
Post Reply
npinhao
Posts: 3
Joined: 12 Feb 2017, 14:59
Antispam: Yes
Location: Lisbon

Grid or computation problem with linear elasticity module

Post by npinhao »

Hi,
I've been simulating the thermal distribution and tensions on a 3D tungsten block with a water pipe inside and a Cu interlayer, subject to a heat pulse on the top surface. Everything is consistent with the heat equation solver but the tensions results depend on grid and are non-physical.
I have prepared two meshes, one with Gmsh (frontal (2D) + delaunay (3D)) [1/4 of the geometry] and another with ElmerGrid [1/2 of the geometry] and computed the results on almost identical problems. I attach the images obtained on the von Mises with and without the grid overlay.
The results should be similar but they depend on grid and on the point of the grid.

I am probably doing something wrong and any comment is appreciated.
Attachments
ElmerGrid result and grid overlay.
ElmerGrid result and grid overlay.
vonMises_Egrid_b.png (87.7 KiB) Viewed 2062 times
Gmsh result with grid overlay.
Gmsh result with grid overlay.
vonMises_gmsh_b.png (253.47 KiB) Viewed 2062 times
result with Gmsh grid.
result with Gmsh grid.
vonMises_gmsh.png (95.77 KiB) Viewed 2062 times
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Grid or computation problem with linear elasticity modul

Post by raback »

Hi

Are you using linear elements? Elasticity equation suffers from locking and particularly for stresses that are behind the derivative the results may be quite poor. Using quadratic elements might give nicer stresses. Just try with "Element = p:2" if you don't care to remesh.

-Peter
npinhao
Posts: 3
Joined: 12 Feb 2017, 14:59
Antispam: Yes
Location: Lisbon

Re: Grid or computation problem with linear elasticity module

Post by npinhao »

Hi Peter

Thank you for your suggestions.
I am still making tests with the quadratic elements and the results improve but not enough. The best strategy seems to be to increase the definition of the grid. I will post my final results when finish.

Nuno
Post Reply