Bug? during FDK analysis

Main Forum
1

Hello

Is this problem known to you?:

During a slow FDK simulation (30 min) the model view window is cleared and the simulation starts speeding up and finishes within a minute. The replay only works until the moment where the model disappears in the model view window.

Greets
MMMPPPMMM

2

Hi MMMPPPMMM,

did you have a look where your model ends up by looking at the position of a segment? I would assume that the FDK solver fails and your model ends up somewhere in space due to a failure in solving the FDK equations. If you share a few more details about your model we can probably give some hints for debugging the model.

Best regards
Daniel

3

Thanks for the fast reply.

I will check that, but if the FDK solver fails the analysis should not finish without any error message, right?

It’s a joint model with multiple bone contacts and wrapping ligaments.

Greets
MMMPPPMMM

4

It depends on the settings. It is possible that only a warning is shown telling that the FDK equations could not be solved properly and continue. This is the default behavior because it is the most handy setting for debugging FDK models and doesn’t make a difference if models run out of the box.

Best regards
Daniel

5

I'm sure there was no warning. The simulation finishes very fast without any warnings or errors.

But I will check it again.

Are these settings accessible for the user. The settings from the Reference Manual in the "InverseDynamics.ForceDepKin" Folder are known to me, even if I don't understand them fully ;).

Greets
MMMPPPMMM

6

Yes you are right, I’m sorry. There is no AnyScript option to stop in case the FDK solver does not converge to a solution. But I hope you can handle that.

If you have saved the output, you can also have a look at the ForceDepKinError of the study if there are any large values. Otherwise you are welcome to post your model on this forum so we can debug the problem.

Best regards
Daniel

7

Yes, I've saved the output. Good to know that the ForceDepKinError is saved in the h5 file, I only have to find where it's saved. :slight_smile:

But I already can say that the centroid data of the bones is "NaN" in Matlab from the moment on the model disappears. I've imported the data with "h5read".