Nonlinear solver computes deformation based on force loads

Numerical methods and mathematical models of Elmer
Post Reply
XHZhu
Posts: 11
Joined: 23 May 2021, 02:03
Antispam: Yes

Nonlinear solver computes deformation based on force loads

Post by XHZhu »

Hi, the great Elmer team.
I'm using Elmer to generate datasets for an elastomer deformation project. The elastomer is a thin cylinder (2cm rad with 0.3cm height). The top and lateral surfaces are fixed as our real-world setting, solid objects will contact the bottom surface and deform the elastomer. It seems explicitly simulate two objects' contact can be time-consuming, I decide to apply nodal forces among the bottom surface to imitate the contact.
To do so, I select a cluster of nodes on the bottom surface and apply force loads on each of them. The force's direction and magnitude are determined based on the solid object's geometry. Specifically, I assume the elastomer's surface aligns well with the solid object, and thus contact force's direction is the solid object's surface normal direction. The magnitude for each contact force is then determined based on normal angles.

However, the simulation results seem weird. The results seem unsmooth and wrong. The image shows the FEM results using a ball indentor and 20N contact force at max. The deformation doesn't seem to be correct. When I increase the contact force, the results crash (surface penetrate the mesh). I'm wondering how can I make this work? Is the contact formulation correct and am I applying nodal forces in the right way? I attached the sif file below FYI.

Many thanks if you could provide any suggestions :)
XHZhu
Posts: 11
Joined: 23 May 2021, 02:03
Antispam: Yes

Re: Nonlinear solver computes deformation based on force loads

Post by XHZhu »

Forgot to attach files...
Attachments
elmer_result.png
(108.89 KiB) Not downloaded yet
case.sif
(13.47 KiB) Downloaded 140 times
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Nonlinear solver computes deformation based on force loads

Post by kevinarden »

The sif file looks fine. The mesh is suspect. Looks to coarse, and using tetrahedrons. Solid mechanics require higher order tetrahedrons. Hexahedron provide better results, although it is always a good idea to use higher order elements in solid mechanics. I would mesh finer and use higher order elements. You could also benefit from using a direct solver for the linear part instead of iterative.
XHZhu
Posts: 11
Joined: 23 May 2021, 02:03
Antispam: Yes

Re: Nonlinear solver computes deformation based on force loads

Post by XHZhu »

Hi Kevin,

Thanks for your suggestions! I applied these modifications as you suggested. I first mesh the elastomer finer to 2x more nodes. Then I use tet10 and hexahedron elements.
However, the results don't seem correct. I applied nodal force loads to surface points. For tet10 elements, some nodes have huge displacements (tet10_large.png). I tried to decrease contact forces' magnitude, then there are too few displacements (tet10_small.png). For hex elements, I guess the resolution is not high enough, so the displacements are still coarse (hex.png).

Based on these observations, do you have any further suggestions to improve the results? Especially for tet10 elements.
Attachments
hex.png
(78.03 KiB) Not downloaded yet
tet10_small.png
(121.6 KiB) Not downloaded yet
tet10_large.png
(73.76 KiB) Not downloaded yet
kevinarden
Posts: 2237
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Nonlinear solver computes deformation based on force loads

Post by kevinarden »

Is the hex mesh 8 noded? If not it may need to be finer. Not sure if you are still using the iterative solver for the linear. Depending on Solver the iterative can complete but not actually be a converged solution. Another option is to increase the load over time by switching to scanning or transient and then apply increasing the load 0.1 to 1.0 seconds apply the loads such as

as Force 2 = Variable Time; Real Matc "Tx *-10.0"

Also noticed you are applying the load as
Name = "Force 3"
Displacement 1 Load = Real 7.525478546118551
Displacement 2 Load = Real -5.236233851669401
Displacement 3 Load = Real 4.960642596318379

I have not seen that before, I have always used
Force 1 = Real 7.52

unless I am enforcing a displacement then it is
Displacement 1 = Real 0.001
annier
Posts: 1168
Joined: 27 Aug 2013, 13:51
Antispam: Yes

Re: Nonlinear solver computes deformation based on force loads

Post by annier »

Hi,
The load with the variable name "Displacement i load ", i = 1,2,3 is the nodal force on a mesh. Basically, a target coordinate has to be specified in the mesh for providing the location of action of nodal force.
Some discussions on this force are provided in:
1. viewtopic.php?f=4&t=589
2. viewtopic.php?f=3&t=659
3. viewtopic.php?f=3&t=6281
4. viewtopic.php?f=2&t=6811&start=20
Anil Kunwar
Faculty of Mechanical Engineering, Silesian University of Technology, Gliwice
Post Reply