Element 'flipping' in a large displacement problem

Numerical methods and mathematical models of Elmer
Post Reply
skoushik
Posts: 20
Joined: 22 Jan 2018, 01:55
Antispam: Yes

Element 'flipping' in a large displacement problem

Post by skoushik »

This is a problem of a 2d disc subjected to a large displacement - nearly 50% of the disc radius.
Displacement was applied at one node (at right) and support conditions at few nodes (on left). For a large value of the applied displacement, the elements containing the node where BC is applied, flipped about their edges. But the log file does not have any error / abnormalities indication.
Is it possible this kind of a distortion or anomaly reflect, probably in log file?
What could be happening in the math that result in this?
Attachments
small_Vs_large_disp.jpg
(104.46 KiB) Not downloaded yet
small_20%rad.txt
(26.31 KiB) Downloaded 252 times
large_40%rad.txt
(44.1 KiB) Downloaded 279 times
raback
Site Admin
Posts: 4812
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Element 'flipping' in a large displacement problem

Post by raback »

Hi

Did you use the "ElasticSolve" capable for large displacements? If you did, was nonlinear system convergence obtained?

You might apply the load gradually with such large displacements.

-Peter
skoushik
Posts: 20
Joined: 22 Jan 2018, 01:55
Antispam: Yes

Re: Element 'flipping' in a large displacement problem

Post by skoushik »

Yes we used ElasticSolve for geometric nonlinearity (as I understand). And the elasticity iterations did converge without errors, as can be seen from the log files attached. Some more relevant details :
We use poisson ratio of 0.49 to study an organic material showing deformations as large as rubbers.
We use linear elastic material in an elastostatic simulation.

May I ask clarification on, why does the displaced element 'flip' instead of all the elements (between displaced node and supportted nodes) getting deformed?
And, does it show up in the log file when the elements distortion is very large...?
mika
Posts: 231
Joined: 15 Sep 2009, 07:44

Re: Element 'flipping' in a large displacement problem

Post by mika »

Hi,

If the Poisson ratio is 0.49, the material is already nearly incompressible and to obtain reasonable results the elasticity equations should be solved with a mixed finite element formulation in order to avoid finite element locking. Did you solve the equations with the mixed formulation (Mixed Formulation = Logical True)? I could imagine that the unphysical solution is a combination of the finite element locking and the use of a pointwise Dirichlet BC. Some details of the mixed formulation are explained in the model documentation.

Best regards,
Mika
mika
Posts: 231
Joined: 15 Sep 2009, 07:44

Re: Element 'flipping' in a large displacement problem

Post by mika »

I also mention that a couple of examples for solving finite elasticity equations via the mixed FE formulation are now under the directory .../fem/tests:

https://github.com/ElmerCSC/elmerfem/tr ... ksMembrane
https://github.com/ElmerCSC/elmerfem/tr ... Membrane3D

The solutions have been verified against reference results found in the literature - see the comments in the sif files of the tests.

-- Mika
Post Reply