Hemisphere-point loads

Numerical methods and mathematical models of Elmer
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Hemisphere-point loads

Post by raback »

Hi

This is not really my cup of tea but I would have a question: Do Ansys and Code Aster have knowledge of the system beyond its discrete finite element representation? I.e. do they know from the CAD geometry that the normal vector at point of load is exactly (1,0,0)? If you have just access to the discrete mesh this information is not trivially available. If you have access to the CAD model you can always retrieve the exact normal and you're not really comparing the same thing!

Also the ShellSolver of Elmer may use exact normal directionality data. This is unfortunately not something that is easily available since it is usually lost in the process of meshing. You will find numerous consistency tests where "mesh.director" data is given to enable more exact solution:

Code: Select all

Shell_And_Solid2/shell_and_solid/mesh.director
Shell_And_Solid/shell_and_solid/mesh.director
Shell_BenchmarkCase1_Quad/p1mesh3/mesh.director
Shell_BenchmarkCase1_Tria/tria3/mesh.director
Shell_BenchmarkCase2_Quad/p1mesh3/mesh.director
Shell_BenchmarkCase2_Tria/tria3/mesh.director
Shell_Cantilever/strip/mesh.director
Shell_Eigenanalysis_Cylinder/fan_blade/mesh.director
Shell_Eigenanalysis_Spherical/piece_of_sphere/mesh.director
Shell_OpenHemisphere/hemisphere/mesh.director
Shell_PinchedCylinder/Mesh/mesh.director
Shell_PlateCase_Quad/plates/mesh.director
Shell_with_Beam_Sections/strip_with_beams/mesh.director
Shell_with_Solid_BenchmarkCase1b/cylinder/mesh.director
Shell_with_Solid_BenchmarkCase1/cylinder/mesh.director
Shell_with_Solid_BenchmarkCase2/cylinder/mesh.director
Shell_with_Solid_Eigenanalysis/cylinder/mesh.director
-Peter
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Hemisphere-point loads

Post by kevinarden »

My understanding of the Ansys, Aster, Nastran, Abaqus solvers is they do not use the CAD geometry to determine the normal vector of the shell. They calculate the normal vector of the shell from the nodal coordinates of each element. They have an element coordinate system for each element and the normal vector is computing using the X axis and N1 to N2, and the Y axis as N1 to Node 4, they normal is then computed. The element coordinate system i then transformed to the global coordinate system for the solution. Stress data after the solution is available in both the global coordinate system and the local element coordinate system.

The mesh director may be key to the Elmer shell solver for a sphere, or perhaps it is simply a matter of using spherical coordinate system.

However Elmer does produce the correct displacement result if the Z displacement is held at the AC boundary.
mdudek
Posts: 9
Joined: 03 Mar 2021, 02:49
Antispam: Yes

Re: Hemisphere-point loads

Post by mdudek »

Thank you all for ideas, I will try to tweak my model according to them. I will let you know is change of coordinate system or normal vector helps with anything. If not I must be happy with blocking an AC edge :)
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Hemisphere-point loads

Post by kevinarden »

https://abaqus-docs.mit.edu/2017/Englis ... -c-le3.htm

There seems to be some inconsistencies on this website posting. The modulus of elasticity in in the text says 68.25 Gpa, but in the input decks provided it is 6.825E7, which I believe is 68.25 MPa.

Also in the input decks provide in the link the Z direction is restrained at the points where the load is applied.

*NSET,NSET=ZCONSTR
8,64

*NSET,NSET=XSYM
18,20,22,25,26,92,96,100,104,108,121,122,123,124
*NSET,NSET=YSYM
8,11,14,23,25,64,69,74,79,84,109,113,117,121
*NSET,NSET=DISPSET
8,64
*RESTART,WRITE
*STEP,PERTURB
*STATIC
*BOUNDARY,OP=NEW
ZCONSTR,3

XSYM,XSYMM
YSYM,YSYMM
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Hemisphere-point loads

Post by kevinarden »

This sif and mesh produces the benchmark results as written, without restraining the Z on the AC boundary. Same results in ABAQUS and NASTRAN.
I was confusing the no rotation conditions on the edges of the symmetry boundaries. No displacement X on the XZ boundary would be U 1 = 0, but no rotation about Y would be DNU 1 = 0. DNU 1 is the difference between deflection in the 1 direction at the bottom and top of the shell. (i.e. no rotation about Y).
results.png
(273.3 KiB) Not downloaded yet
1case.sif
(1.68 KiB) Downloaded 164 times
mesh.zip
(38.75 KiB) Downloaded 159 times
Abaqus U1 Nastran U1 Elmer U1
0.184 0.186 0.184
Abaqus U3 Nastran U3 Elmer U3
.092 0.093 0.090
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Hemisphere-point loads

Post by raback »

Thank you Kevin for the diligent comparison! Could to know that these are in agreement. -Peter
mdudek
Posts: 9
Joined: 03 Mar 2021, 02:49
Antispam: Yes

Re: Hemisphere-point loads

Post by mdudek »

Thank you all for your responses I really appreciate your help and advice. The model is correct now and I managed to understand how the DNU boundary conditions work. Looking forward to do some more complicated cases with Elmer :D
mdudek
Posts: 9
Joined: 03 Mar 2021, 02:49
Antispam: Yes

Re: Hemisphere-point loads

Post by mdudek »

Sorry guys for disturbing you but I am trying to reproduce this case on my own and I got an incorrect results. I was searching for mistakes for a long time and I cannot find any. Would you be so kind to have a look at my simulation to help me what am I doing wrong?
Attachments
case.7z
(16.62 KiB) Downloaded 161 times
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Hemisphere-point loads

Post by kevinarden »

I don't see anything wrong with the input. I should have pointed out that Elmer is much more sensitive to mesh size than NASTRAN or ABAQUS. I had to use a much finer mesh in Elmer to get the correct result. A course mesh as defined in the benchmark did not produce good results in Elmer.

I have seen this in many of the test cases I have ran. I believe the commercial codes have much improved element formulations. Normally in Elmer i would also use higher order elements, but ShellSolver does not support.
mesh.png
mesh.png (193.73 KiB) Viewed 2530 times
mdudek
Posts: 9
Joined: 03 Mar 2021, 02:49
Antispam: Yes

Re: Hemisphere-point loads

Post by mdudek »

I assume that these results are from your simulation, not mine? Because I got really strange numbers no matter the mesh.
Post Reply