the problem of applying Mesh Geometry to the non-conformity contact pair

[Linjie Wang]

Hi, there.

I want to use the mesh geometries of self designed total knee replacement (non-comformity) to predict knee forces. I have successfully calculated the Inverse Kinematics, Residual Reduaction Algorithm and Computed Muscle Control without consideration of mesh geometries. However, when it came to the Forward Dynamics with mesh geometry, the gait became abnormal, which I think was caused by the newly imported mesh geometries as a pair of contact. Because of this extra contact force, the balance between muscle forces decided by CMC and joint forces was broken and further induce extra accelerations to drive each segment to the wrong location which is not even closer to the CMC motion trajectories.

Therefore, I changed to add mesh geometry in the step of CMC. But the solution principle is to track the known kinematics, plus that the elastic foundation that mesh geometry uses is not based on the penalty method so that the interpenetration would not be regarded as an error to adjust. Therefore, unless the motion between the femor and tibia implants are prescribed or adjusted in advance in order to avoid large interpenetration, otherwise the contact force calculated would be unreasonalbly high in some poses where there should be small contact force just because of the large penetration, and then further lead to the failed muscle solution or too large muscle forces.

I wonder if there is any possibility to modify the algorithm of elastic foundation or introduce the property of Force-Dependent-Kinematics into the tool modules say Inverse Dynamics or Residual Reduaction Algorithm.

Thanks,
Linjie

[Dimitar Stanev]

Hi,

I wonder if there is any possibility to modify the algorithm of elastic foundation or introduce the property of Force-Dependent-Kinematics into the tool modules say Inverse Dynamics or Residual Reduaction Algorithm.

It would be nice if someone implements FDK. From my limited experience it is not very easy to modify the existing algorithms since you need to be familiar how things work as a whole, and familiarize yourself with Simbody. I would go with a custom, separate implementation of the problem and use Simbody/OpenSim API to get the required quantities.

Best