Convergence issues when going from optimized disp control to force control

[bhendo]

Hello,

I have created a simple 1x1 cube model to represent tensile fatigue creep using a neo-hookean elastic term and simo damage term. I optimized this to experimental data by inputting a load curve of my peak displacement representing my creep and then optimized to my constant force. This gave a great fit. I am now trying to apply these same parameters to the same study but instead inputting a load curve of my experimental max force and getting the displacement as the output. When doing so the model quickly diverges. Is this do to the solution possibly not being unique or the damage evolving too quickly? Are there any other reasons why this might occur?

-Thanks for any help

[maas]

Hi,

If you can send us the two models (the displacement one and the force one), I'd be happy to take a closer look and see if I can spot any issues.

Thanks,

Steve

[bhendo]

Hi Dr. Maas,

Was there update on this issue or any obvious issues with my models?

Thanks,

Bradley

[maas]

Hi Bradley,

I took a look at your models and, after conferring with a colleague, suspect that you are seeing the expected behavior of this damage model under load control. Under force driven conditions, there is a maximum value for the damage at which point the model will no longer be able to support any increased load. At that point, the simulation will fail, which will manifest itself in negative jacobians and failure to converge. I hope that explains what you are seeing. Let us know if you have any further questions.

Best,

Steve

[bhendo]

Hi Dr. Maas,

I appreciate the response.

I do have a follow-up question. A collaborator of mine expressed that in many FEA programs regularization may be used to improve the numerical difficulties relating to damage mechanisms possibly through viscous damage mechanisms.

Is this a process you're familiar with and does it exist within FEBio?

I apologize if any of my terminology is off as I am very new to this idea.

Thanks again for all the help,

Bradley

[maas]

Hi Bradley,

I'm not entirely sure what you mean by regularization, by I do know that a false dynamic response can often solve models that otherwise converge difficultly. In this case, you would run the problem as a dynamic analysis and introduce what is called "numerical damping". Basically, you would choose the time integration parameters such that the model loses energy over time, and eventually settles in the solution of the equivalent static problem. Now, this would only work with elastic materials, i.e. problems that don't have history-dependence, so I'm not sure if this would work for damage models. If this doesn't sound like what your collaborator was thinking of, perhaps try to get some more details and then we can look into how to address this in FEBio.

Cheers,

Steve