[NickolasPl]

Hi everyone,

I apologize if I am not writing this thread at the right place.

Anyway, I am an experienced user with OpenFOAM as I have done numerous simulations regarding non-Newtonian and Newtonian fluids.

I'm interested at the moment on some issues regarding Solid Mechanics simulations and thinking that hopefully due to the vast capabilities of OpenFOAM that this can be done with this software.

I would like to run a simple two dimensional case where a solid circular ring is "plastically" or elastically is deformed under load that actually squeezes the ring from two opposites sides, to deform (if plastic deformation) into some short of elliptical ring. I was thinking of doing so by using Dynamic Mesh but yet I haven't found any suitable solvers. The FSI approach I was also thinking would not work in my case since there is only one solid material.

So, what solver can anybody suggest to approach this problem? Any suggestions are welcomed.

Thanks in advance.

Nickolas

[jherb]

Thank you for your question. I also think about this problem (without any solutions yet). I guess you speak about ring compression tests?

One starting point might be the solid mechanics solvers in the foam extend version:
In the tutorials there is an example in solidMechanics/elasticPlasticSolidFoam/rigidBallPlasticBrick
As the name says it a rigid ball is dropped on a plastic brick. For a ring compression test it has to be a plastic ring compressed by a rigid beam.

I would be really interested in what you find out.

[NickolasPl]

Hi Joachim,

Thanks for the fast response. Just earlier today I came across about this solver in the Extended Foam project as you said. I am worried however that I did not see anywhere any files concerning Dynamic Mesh Handling. And I'm saying this because I would expect that the deformation of any arbitrary solid body to be physically represented by the deformation of it's mesh according to any deformation laws that applies there. Unless, if there is something I am missing about this solver. I cannot think of any other way to represent a body's displacement due to deformations at certain regions, other than deformation of the mesh itself.

Do you agree on this? Did you by any chance solve the tutorial of the ball you are referring to?

[jherb]

I just found this PhD thesis:
http://adhesion.ucd.ie/openfoam/UCD_...lPhdThesis.pdf

If I understand the example correctly, it consists of two meshes which are moved. But you are right: It looks like that the mesh(s) are not deformed. But there is another solver (for fluid structure interaction) where the solid (and probably also the fluid) mesh is deformed. So perhaps you can somehow merge these two examples.

[NickolasPl]

Thank you very much for the file...I will take a look!

My simple example is even simpler...it doesn't have to do with FSI as there is no fluid to cause the deformation. It is actually as if you insert a ring between two plates that move vertically so that they can squeeze the ring and deform it to an ellipsoidal geometry. So in mesh terms, you have only one mesh consisting of the ring itself and the plates. And maybe in a more simple way, ignore the plates and assume that an "equal force" deforming the ring is applied at two opposite points around the ring. I thought that this was a simple example and still surprised that nobody has done it yet in OpenFOAM!

Anyway, I will try to keep this thread alive and maybe find a solution to this issue the soonest.

[jherb]

Yes, it is clear that there is no fluid in your example. But the FSI solver has some code which handles the motion/deformation of the solid mesh.

Actually, I talked about this topic with the author of the PhD thesis I have linked above at the OpenFOAM workshop in Zagreb. His estimation for the effort of such a simulation with OpenFOAM were somewhere between a Master and a PhD thesis

Unfortunately there seems to be not much documentation about the solid mechanics part of OpenFOAM (extended) (or even less then for the fluid part).

Here are two presentations from the OpenFOAM Workshop 2014 in Zagreb:
http://sourceforge.net/projects/open...4.pdf/download
http://sourceforge.net/projects/open...4.pdf/download

[bigphil]

Hi,

I am actually currently working on a project using OpenFOAM for metal forming simulations, and I have looked at modelling such a pipe compression case, I copied the example in the Abaqus benchmarks manual 1.1.12.

I intend to share the my general developments in the future; in the meantime I can certainly help you and steer you in the correct direction.

The elasticPlasticSolidFoam solver in foam extend is for small strains and assumes J2 (Mises) plasticity, and is a good place to start to understand how plasticity is dealt with at the moment in FV and foam.

Effectively, the two developments you would need to simulate this pipe compression case are:

• a plasticity model for large strains;
• and an appropriate large strain solver (mathematical model).

elasticNonLinULSolidFoam is an example of a large strain elastic solver which uses an updated lagrangian approach (it moves the mesh), so something similar can be used.

For the plasticity model, a book on computational plasticity is a good place to start; I have found these two useful:

• Simo & Hughes - Computational Inelasticity
• de Souza Neto, Peric & Owen - Computational methods for plasticity

They both discuss plasticity models for large strain (hypo-elastic and hyper-elastic).

Best regards,
Philip

[NickolasPl]

Hi Philip,

Thank you for your insightful reply. I will start looking into the solvers you suggested and proceed with some preliminary results here. I see that there's not enough things done on this issue and so we can post here some of our findings in the future for the good of the OF solid mechanics field

Nickolas

[jherb]

Hi Philip,

you wrote, ...

Originally Posted by bigphil

elasticNonLinULSolidFoam is an example of a large strain elastic solver which uses an updated lagrangian approach (it moves the mesh), so something similar can be used.

but on http://adhesion.ucd.ie/openfoam/UCD_...-_Solvers.html it is written, that elasticNonLinULSolidFoam is an approach creating a solver appropriate for large rotation but small strains.

Instead the elasticNonLinTLSolidFoam solver is described as "appropriate for large strain and large rotations."

Thank you very much for your help
Joachim

[bigphil]

Hi Joachim,

OK, sorry I am being inconsistent in my descriptions.

I will try to clarify:

The elasticNonLinTLSolidFoam solvers uses a total Lagrangian (TL) approach, whereas the elasticNonLinULSolidFoam solver uses an updated Lagrangian (UL) approach.

Both TL and UL approaches can, in theory, be used for large strain, large rotation and large displacement analyses.
However, the practical suitability of an approach (TL, UL, ALE, etc.) really depends on the correct choice (and implementation) of the constitutive model.

In this regard, it could be argued that neither elasticNonLinTLSolidFoam and elasticNonLinULSolidFoam are strictly valid to describe any real material at large strains because they use the St. Venant-Kirchhoff model (this constitutive model does not describe any real material well at truly large strains).

SO, if you want to implement a solver which accurately captures your specific material at large strains, then you first need to find a constitutive model which accurately describes your material (e.g. Neo-Hookean models might work well for large elastic strains in rubber-like materials, or for elasto-plastic metals you could use a hypo-elastic model (like in Abaqus) or a hyper-elastic-plastic model (like in Simo & Hughes).
When the constitutive model is found, the mathematical model can be implemented as either TL or UL or whatever is most convenient.

Hopefully this is a bit clearly,
Philip

[jherb]

Thank you for the help so far.

Now I have tried to start with a test case using the existing solver elasticNonLinULSolidFoam.

All files of the test cases are on github: https://github.com/jmozmoz/foamRingCompressionTest

Of course, I immediately hit some problems
I tried a setup analogous to some ADINA simulations (see pages 17- in the (German) report http://www.grs.de/sites/default/file...GRS-A-3519.pdf and this paper (behind paywall) http://www.sciencedirect.com/science...2954931400209X)



The left side of the ring and the bottom of the ring and the plunger are on symmetry planes. For the right side of the plunger (or how this is called?) I tried timeVaryingFixedDisplacement and timeVaryingSolidTraction boundary conditions. There is a solidContact boundary conditions between the ring and the plunger.

Both cases crash (immediately, if the two objects have contact). Here is the output of the timeVaryingFixedDisplacement case:

/*---------------------------------------------------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | foam-extend: Open Source CFD                    |
|  \\    /   O peration     | Version:     3.1                                |
|   \\  /    A nd           | Web:         http://www.extend-project.de       |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
Build    : 3.1-7d8e040bf53d
Exec     : elasticNonLinULSolidFoam
Date     : Jan 08 2015
Time     : 01:27:02
Host     : iveragh
PID      : 5487
CtrlDict : /home/joachim/foam/foam-extend-3.1/etc/controlDict
Case     : /data/home/joachim/foam/joachim-3.1/run/elasticNonLinULSolidFoam/tutorials/ringTestDisp
nProcs   : 1
SigFpe   : Enabling floating point exception trapping (FOAM_SIGFPE).

// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create time

Create mesh for time = 0

Reading incremental displacement field DU

Creating solidContact patch
    Normal contact model: standardPenalty
    Friction contact model: frictionless
    Interpolation of traction from slave to master: ggi
Patch blockUp    Traction boundary field: DU
    nonLinear set to updated Lagrangian
Creating solidContact patch
Patch hole    Traction boundary field: DU
    nonLinear set to updated Lagrangian
Reading accumulated displacement field U

Reading accumulated strain field epsilon

Reading accumulated stress field sigma

Reading incremental stress field DSigma

Selecting rheology model multiMaterial
Selecting rheology model linearElastic
Selecting rheology model linearElastic
Selecting rheology model linearElastic
Creating constitutive model
Selecting solidInterface procedure none
divSigmaExp method: surface
divSigmaLargeStrainExp method: surface
Move Mesh method: leastSquares

Starting time loop

Time = 0.01

    Slave contact patch ring grabbing normalContactModel pointer from master
    Slave contact patch ring grabbing frictionContactModel pointer from master
    Time 0.01, Corrector 0, Solving for DU using DICPCG, res = 1, rel res = 1, inner iters 2
    Time 0.01, Corrector 1, Solving for DU using DICPCG, res = 1, rel res = 0.905146, inner iters 7
    Time 0.01, Corrector 2, Solving for DU using DICPCG, res = 0.318771, rel res = 0.346876, inner iters 4
    Time 0.01, Corrector 3, Solving for DU using DICPCG, res = 0.214963, rel res = 0.174096, inner iters 4
    Time 0.01, Corrector 4, Solving for DU using DICPCG, res = 0.135466, rel res = 0.0897763, inner iters 4
    Time 0.01, Corrector 5, Solving for DU using DICPCG, res = 0.0788186, rel res = 0.0453993, inner iters 4
    Time 0.01, Corrector 6, Solving for DU using DICPCG, res = 0.0432484, rel res = 0.0227575, inner iters 4
    Time 0.01, Corrector 7, Solving for DU using DICPCG, res = 0.026917, rel res = 0.0114344, inner iters 7
    Time 0.01, Corrector 8, Solving for DU using DICPCG, res = 0.0262867, rel res = 0.00606456, inner iters 7
    Time 0.01, Corrector 9, Solving for DU using DICPCG, res = 0.0253862, rel res = 0.00454009, inner iters 7
    Time 0.01, Corrector 10, Solving for DU using DICPCG, res = 0.0254976, rel res = 1, inner iters 7
    Time 0.01, Corrector 11, Solving for DU using DICPCG, res = 0.266893, rel res = 0.982733, inner iters 20
    Time 0.01, Corrector 12, Solving for DU using DICPCG, res = 0.0262712, rel res = 0.811591, inner iters 19
    Time 0.01, Corrector 13, Solving for DU using DICPCG, res = 0.00586627, rel res = 0.665482, inner iters 18
    Time 0.01, Corrector 14, Solving for DU using DICPCG, res = 0.00297913, rel res = 0.568317, inner iters 17
    Time 0.01, Corrector 15, Solving for DU using DICPCG, res = 0.0022604, rel res = 0.511747, inner iters 17
    Time 0.01, Corrector 16, Solving for DU using DICPCG, res = 0.00205008, rel res = 0.489955, inner iters 18
    Time 0.01, Corrector 17, Solving for DU using DICPCG, res = 0.00201031, rel res = 0.488407, inner iters 18
    Time 0.01, Corrector 18, Solving for DU using DICPCG, res = 0.00197379, rel res = 0.485961, inner iters 18
    Time 0.01, Corrector 19, Solving for DU using DICPCG, res = 0.00194643, rel res = 0.48402, inner iters 18
    Time 0.01, Corrector 20, Solving for DU using DICPCG, res = 0.00482397, rel res = 1.00953, inner iters 12
    Time 0.01, Corrector 21, Solving for DU using DICPCG, res = 0.00403376, rel res = 2.23768, inner iters 17
    Time 0.01, Corrector 22, Solving for DU using DICPCG, res = 0.0086619, rel res = 0.812756, inner iters 17
    Time 0.01, Corrector 23, Solving for DU using DICPCG, res = 0.00330577, rel res = 0.612119, inner iters 17
    Time 0.01, Corrector 24, Solving for DU using DICPCG, res = 0.00258677, rel res = 0.543335, inner iters 17
    Time 0.01, Corrector 25, Solving for DU using DICPCG, res = 0.00246791, rel res = 0.513968, inner iters 17
    Time 0.01, Corrector 26, Solving for DU using DICPCG, res = 0.00269836, rel res = 0.500082, inner iters 17
    Time 0.01, Corrector 27, Solving for DU using DICPCG, res = 0.00334619, rel res = 0.491255, inner iters 16
    Time 0.01, Corrector 28, Solving for DU using DICPCG, res = 0.00480308, rel res = 0.479822, inner iters 14
    Time 0.01, Corrector 29, Solving for DU using DICPCG, res = 0.00797664, rel res = 0.481758, inner iters 12
    Time 0.01, Corrector 30, Solving for DU using DICPCG, res = 0.0511407, rel res = 2.20945, inner iters 18
    Time 0.01, Corrector 31, Solving for DU using DICPCG, res = 0.878745, rel res = 1.00548, inner iters 16
    Time 0.01, Corrector 32, Solving for DU using DICPCG, res = 0.999997, rel res = 1, inner iters 17
    Time 0.01, Corrector 33, Solving for DU using DICPCG, res = 1, rel res = 1, inner iters 18
    Time 0.01, Corrector 34, Solving for DU using DICPCG, res = 1, rel res = 1, inner iters 18

And here the DU boundary file for time 0:

/*--------------------------------*- C++ -*----------------------------------*\
| =========                 |                                                 |
| \\      /  F ield         | foam-extend: Open Source CFD                    |
|  \\    /   O peration     | Version:     3.1                                |
|   \\  /    A nd           | Web:         http://www.extend-project.de       |
|    \\/     M anipulation  |                                                 |
\*---------------------------------------------------------------------------*/
FoamFile
{
    version     2.0;
    format      ascii;
    class       volVectorField;
    location    "0";
    object      DU;
}
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //

dimensions      [0 1 0 0 0 0 0];

internalField   uniform (0 0 0);

boundaryField
{
    left
    {
      type            symmetryPlane;
    }

    ring
    {
      type                 solidContact;
      shadowPatch          blockLeft;
      master               no;
      contactActive        yes;
      value                uniform (0 0 0);
    }

    down
    {
      type            symmetryPlane;
    }

//    up
//    {
//      type            solidTractionFree;
//      nonLinear       updatedLagrangian;
//    }

    hole
    {
      type            solidTractionFree;
      nonLinear       updatedLagrangian;
    }

    frontAndBack
    {
        type          empty;
    }

    blockFrontAndBack
    {
        type          empty;
    }

    blockDown
    {
        type          symmetryPlane;
    }

    blockUp
    {
        type          solidTractionFree;
        nonLinear     updatedLagrangian;
    }

    blockRight
    {
        type          timeVaryingFixedDisplacement;
    outOfBounds   clamp;
    fileName      "$FOAM_CASE/constant/timeVsRightDisp";
    value         uniform (0 0 0);
    }

    blockLeft
    {
      type                 solidContact;
      master               yes;
      contactActive        yes;
      rigidMaster          no;
      shadowPatch          ring;
      interpolationMethod  ggi;
      //interpolationMethod  patchToPatch;
      projectionAlgo       visible;
      projectionDir        contactSphere;
      //projectionDir        vector;
      correctionFrequency   10;
      normalContactModel     standardPenalty;
      standardPenaltyNormalModelDict
        {
          relaxationFactor               0.1;
          penaltyScale                   1;
          returnScale                    1.5;
          contactGapTol                  1e-7;
          limitPenetration               no;
          penetrationLimit               -0.03;
          limitPressure                  no;
          pressureLimit                  1e12;
          correctMissedVertices          yes;
          distanceMethod                 point;
          //distanceMethod                 face;
          oscillationCorrection          yes;
          smoothingSteps                 1;
          infoFrequency                  10;
          }
      frictionContactModel   frictionless;
      value                uniform (0 0 0);
    }


}

// ************************************************************************* //

This model should be just the staring point, using elastic mechanics. At them moment, all material properties are just dummy values.

Also a (probably very ignorant) question:
Assuming the model will work, how can the force be calculated the plunger exerts on the ring?

Thank you for any help.
Joachim

[jherb]

It looks like there is a fundamental problem, using the elasticNonLinULSolidFoam solver together with the solidContact boundary condition: http://www.cfd-online.com/Forums/ope...lp-please.html

[jherb]

The problem seems to be related to the geometry of the case. I uploaded a case, where the solver is at least running to https://github.com/jmozmoz/foamRingC.../plateHoleDisp

Is there a problem if two solids have only contact at one point using the solidContact boundary condition?

[jherb]

The problem was the parameter correctionFrequency for the solidContact boundary condition. By reducing it from 10 to 1 the simulation now works.

[jherb]

Ok, I will proceed with my monologue:

I have modified the elasticNonLinULSolidFoam so it outputs the product sigma & n (stress tensor dot surface normal). Integrating this quantity over a patch should give the force on this patch. In the attached graph the forces measured in an actual ring compression test are compared with the simulations of ADINA and OpenFOAM. The elastic part of OpenFOAM seems to be not to far off. As there is no consideration of the plasticity in the used solver this clear that the simulation breaks down at some point.

[nlc]

Hi Joachim,

I just did a dynamic case for fun with elasticNonLinIncrTLSolidFoam I change the correction correctionFrequency to be 1 as it help as you told us.

I made a short video for fun ! This is dynamic case and the speed is shuch that we see wiggle !!

https://www.youtube.com/watch?v=l0PKDVWA7qs

[jherb]

Thank you for sharing this. Would you be so kind to also share the input files (boundary conditions, blockMeshDict, fvS*).

Thank you very much

Originally Posted by ubald

Hi Joachim,

I just did a dynamic case for fun with elasticNonLinIncrTLSolidFoam I change the correction correctionFrequency to be 1 as it help as you told us.

I made a short video for fun ! This is dynamic case and the speed is shuch that we see wiggle !!

https://www.youtube.com/watch?v=l0PKDVWA7qs

[nlc]

Yes why not this is a simple theoretic test i did just to see if i can get result and might not be totally consistent with what you try to do.

By changing in 0/DU rigidMaster to yes and changing topBar BC to timeVaryingFixedDisplacement you should have a correct setup case for rigid master. Please let me know if everything work for you.


Hear it is:

[Maimouna]

Dear All who posted this thread,

many thanks for this really useful thread. I'll try to post my planning in the next period of time to get help and advice from your side.

Now, I'm working in vortex-induced vibration of an elastic circular cylinder using OpenFOAM. This is FSI problem including fluid and solid (elastic cylinder). I installed FSI solver in foam-extend-3.1. But I'm still struggling in the steps how to start working with this case in FSI solver. I already finished the VIV case for rigid cylinder. Now, I'm looking to get the simulation of an elastic one. For the elastic cylinder, the cylinder will change its shape because of some fluid forces affect on it. The equation that I have to solve here is the dynamic elasticity equation. Regarding my problem, I have some questions

1. Which solver is the best for my case, icoFsiFoam or icoFsiElasticNonLinULSolidFoam solver in solidMechanics tutorial?

Do you have any idea which case should follow to help me in my case problem? I'm thinking about flappingConsoleSmall case, would that help?


I'm looking forward to get your advice. I'll be really happy to work with you.

What about mesh generation? Shall I generate mesh for the plate and another one for cylinder separately? Note that, I generated mesh for the VIV for the rigid cylinder case by Gmsh? Any idea please.

Any answer would be really appreciated and many thanks in advanced.

[nlc]

== solidMechanics/icoFsiElasticNonLinULSolidFoam:

Application
icoFsiElastciNonLinSolidFoam

Description
Transient solver for fluid-solid interaction for an incompressible
fluid and a large strain solid
solid mesh is moved using U interpolated using least squares method

VS

== solidMechanics/deprecatedSolvers/icoFsiFoam :

Application
icoFoam

Description
Transient solver for incompressible, laminar flow of Newtonian fluids with
mesh motion. Set up as a fake fluid structure interaction solver


You decide !! It depend on your need you can look at the source code ! I need to precise that I have as of today no experience with FSI so you need to do your work and reed what those code do to make your own idea. But just by the description and the fact that that icoFsiFoam is in deprecatedSolvers folder I would go with icoFsiElastciNonLinSolidFoam !!

By the way this post was not relay about FSI so starting an other post would be the god way to ask your question.