Simulation of Orthotic Treatment for Scoliosis

[Akilyon]

Hi all,

I want to simulate an orthotic treatment (brace in our case) for scoliosis. I provided the.fs2 file and its results in the following link, https://drive.google.com/drive/folde..._8?usp=sharing. I am getting an error about on-time 51% on the simulation and I would like to ask, which changes do you suggest to get this problem solved? If you have a suggestion you can edit or upload it in the given link.

Things that come to my mind are, playing with the contact type's properties values, changing the model's design, changing its mesh making it finer at critical regions, spraying new areas for contacts and boundary conditions on the model, etc. Also, I have tried the Broyden Solver and it was better than the BFGS in my case so should I try the other quasi-newton methods and linear_solver types and combinations within them?

I had a similar post to this but that was very simple in the design way, https://forums.febio.org/forum/febio...ntact-analysis.

Thank you,
Atalay

[maas]

Hi Atalay,

Can you please elaborate on the error that you are getting? Without knowing the problem, it's difficult to make any recommendations.

Thanks,

Steve

[Akilyon]

Hi Steve,

In the model, I am using contact physics to try to wear the divided braces to a body in the middle. I first wear the left rigid brace and it goes almost smoothly but when the right brace timing starts, I get the termination error. Time steps lowers even to 1e-7 and still doesn't help. I am learning about this physics and am not sure but the error occurs at the contact interfaces. Maybe I am forcing the primary facets of sliding very much and/or the contact parameters might not be properly written. I put my model's results if you would like to check and if I couldn't elaborate as you want it what would you like to learn? I am not good at reading the error .log file.

Thank you,
Atalay

[maas]

Hi Atalay,

Thanks for adding the results. The error you are running into is a negative jacobian, which means one or more elements are getting inverted. Looking at the plot file, there are indeed quite a few elements that are deforming a lot. Since it's a rather big model, my suspicion is that it might be elements near the neck. The top surface of the neck appears to be fully constrained, and, in that area, strains are reaching up to 200% which seems rather high. So, do you really need that fixed constraint on the neck? Perhaps try running the model without adding that and see how it runs?

Also, setting the max retries to a higher number than the default of 5 (yours is set to 11) usually does not help. The effect of this is that febio will take smaller and smaller time steps. At some point the time steps become so small, that numerical noise can affect the results and febio will not be able to converge anymore. I recommend keeping that value to 5 to prevent this. You can try turning on the aggressiveness parameter, which will half the time step size on each failed attempt to solve the time step. Sometimes this works better. However, if the problem is indeed as I suspect (negative jacobians due to overly constrained elements) then changing the time stepping parameters will not help. In this case, it might be better to change your boundary conditions, especially the one on the neck.

Best,

Steve

[Akilyon]

Hi Steve,

I continued studying on this model. Thank you for your mention of strains, I started noticing them in my analysis and made proper changes. I removed the fixed constraints on the neck and it converged better but didn't pass 60% and it was looking non-real. So I fixed the whole body's and spines z-displacement to reduce strains and his time and I got 80% on the simulation time. I would like to ask what should I consider after this point? I put the final version with named HS1164v2.fs2 and its result to the initial Google link, https://drive.google.com/drive/folde...usp=drive_link if you would like to check to check.

I started thinking that I may be missing very major things about contact mechanics. I have read all the contact mechanics in user, theory, and studio manuals, and watched the webinars from them I have tried quadratic elements, new bc.s, small design changes, and meshes (both fine and coarse mesh), etc. but didn't make a big progress. Could you give me some suggestions?

Thank you,

Atalay