Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Solved maximum von mises stress too high during large deformation of elestoplastic model

Please login with a confirmed email address before reporting spam

Hi,

I am fairly new at using Comsol and am trying to simulate the expansion of a stent. My goal is to retrieve the maximum von mises stress and determine if the stent fails or not. I have broken the stent down to look at the expansion of a single link. This has been previously done in Ansys, resulting in a maximum von mises stress of 3.82e8 Pa. Everything I have tried in Comsol, however, results in a von mises stress about a magnitude higher (~5.34e9 Pa). This cannot be the true von mises because it would mean the material fails (ultimate stress is 517 MPa), and experimentally I know this is not the case.

I have looked at the biomedical stent model that is in the model gallery and mimicked it, but the results are still a magnitude too high.

Attached is my file. Does anyone know what I might do to fix my model?

Thanks for your time and help!

Shannon Gott


3 Replies Last Post 7 apr 2011, 01:58 GMT-4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 6 apr 2011, 01:53 GMT-4
Hi

there are two things I believe is not optimum: First you use edge loads and not boundary loads, this easily over stresses locally the model, particularly with the very coarse mesh you are using.
Second the forces appear tangent to the surfaces, which might make it slightly tricky to load and for the solver to start

Another remark, I'm astonished that your model has the load section positioned at the Y direction at different heights, your model is not really symmetric like that, you are combining force and moment loads

My experience so far is that ANSYS and COMSOL (as well as NASTRAN) gives similar results for the same models, its only that often its tricky to ensure the models are really the same

Well I played a little with a model I reconstructed, first I'm using a soft spring at the COG (via "Edge Load 1" which is less strict than the disabled Fixed Constraint 1) to hold it in place an to avoid it runs away, then I load it symmetrically, there are different tests in there,

I do not get any large difference between edge or boundary load for Study 1 enable the "Boundary Load 1" (or the "Edge Load 2" but not both) and check that "Prescribed Displacement 3" is disabled.
For "Study 2" disable both "Body Load 1 and Edge Load 2" and enable "Prescribed Displacement 3"

The "Prescribed Displacement 1" might not be required but it's there to enforce the symmetry (or rather antisymmetry of the two sides.

I agree that in Linear Elastic the load gets high, but I'm also using a coarse mesh 2k nodes only, "Extremely fine" is probably better. I have noticed some convergence issues with Elastoplastic material model, one would need to use much smaller parametric sweep steps to help the solver finds it's way.

The issue is not solved, anyhow you differences could well come from the elastoplastic material model parameters, or the physics behind

--
Good luck
Ivar

--
Good luck
Ivar
Hi there are two things I believe is not optimum: First you use edge loads and not boundary loads, this easily over stresses locally the model, particularly with the very coarse mesh you are using. Second the forces appear tangent to the surfaces, which might make it slightly tricky to load and for the solver to start Another remark, I'm astonished that your model has the load section positioned at the Y direction at different heights, your model is not really symmetric like that, you are combining force and moment loads My experience so far is that ANSYS and COMSOL (as well as NASTRAN) gives similar results for the same models, its only that often its tricky to ensure the models are really the same Well I played a little with a model I reconstructed, first I'm using a soft spring at the COG (via "Edge Load 1" which is less strict than the disabled Fixed Constraint 1) to hold it in place an to avoid it runs away, then I load it symmetrically, there are different tests in there, I do not get any large difference between edge or boundary load for Study 1 enable the "Boundary Load 1" (or the "Edge Load 2" but not both) and check that "Prescribed Displacement 3" is disabled. For "Study 2" disable both "Body Load 1 and Edge Load 2" and enable "Prescribed Displacement 3" The "Prescribed Displacement 1" might not be required but it's there to enforce the symmetry (or rather antisymmetry of the two sides. I agree that in Linear Elastic the load gets high, but I'm also using a coarse mesh 2k nodes only, "Extremely fine" is probably better. I have noticed some convergence issues with Elastoplastic material model, one would need to use much smaller parametric sweep steps to help the solver finds it's way. The issue is not solved, anyhow you differences could well come from the elastoplastic material model parameters, or the physics behind -- Good luck Ivar -- Good luck Ivar


Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 6 apr 2011, 17:46 GMT-4
Ivar,

Thank you for your suggestions and prompt response. I had previously tried switching between edge and boundary loads in case of some local stress spike, but as you found, there was no significant difference. I have not tried an extremely fine mesh with the linear elastic model, but I have with the Elastoplastic model. It took about a week and a half solving and ended in the error: "Failed to compute Elastoplastic strain variables. Returned solution has not converged." Does this mean I need smaller parametric sweep steps?

I was also wondering if you could clarify the last statement that you made about the differences possibly coming from the Elastoplastic material model parameters, or the physics behind. Are the parameters you are referring to the inputs in the Elastoplastic model (i.e. Young's modulus, tangent modulus)? And by the physics, do you mean the governing equations in the Elastoplastic model? Do I have to somehow tweak the equations?

Thanks for your time and help,

Shannon
Ivar, Thank you for your suggestions and prompt response. I had previously tried switching between edge and boundary loads in case of some local stress spike, but as you found, there was no significant difference. I have not tried an extremely fine mesh with the linear elastic model, but I have with the Elastoplastic model. It took about a week and a half solving and ended in the error: "Failed to compute Elastoplastic strain variables. Returned solution has not converged." Does this mean I need smaller parametric sweep steps? I was also wondering if you could clarify the last statement that you made about the differences possibly coming from the Elastoplastic material model parameters, or the physics behind. Are the parameters you are referring to the inputs in the Elastoplastic model (i.e. Young's modulus, tangent modulus)? And by the physics, do you mean the governing equations in the Elastoplastic model? Do I have to somehow tweak the equations? Thanks for your time and help, Shannon

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 7 apr 2011, 01:58 GMT-4
Hi

no I have not "gone behind" on the elastoplastic model, you have quite some info in the doc, and by studying the "equations" used by COMSOL (turn them on in the preferences in V4).
What I suspect is that there might be differences between ANSYS and COMSOL in the way these are introduced/simplified, I tend to trust more COMSOL, as its open.

I'm currently fighting (and loosing quite some time) with large deformation and parametric sweep on torque load cases, still not understanding why the solver has such issues to find it's way, and the default logger info is very sparse.

I can confirm that imposing small increments makes often the solver go faster, as it gets less lost and uses less internal steps per demanded steps, so globally it's quicker. I just do not know how to define the optimum (yet ;)

--
Good luck
Ivar
Hi no I have not "gone behind" on the elastoplastic model, you have quite some info in the doc, and by studying the "equations" used by COMSOL (turn them on in the preferences in V4). What I suspect is that there might be differences between ANSYS and COMSOL in the way these are introduced/simplified, I tend to trust more COMSOL, as its open. I'm currently fighting (and loosing quite some time) with large deformation and parametric sweep on torque load cases, still not understanding why the solver has such issues to find it's way, and the default logger info is very sparse. I can confirm that imposing small increments makes often the solver go faster, as it gets less lost and uses less internal steps per demanded steps, so globally it's quicker. I just do not know how to define the optimum (yet ;) -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.