acceleration body kinematics

[Irina Wils]

Hi all

I am currently working on a jaw model. I added a screenshot and the topology view in the first slide of the attachement. I used a simplified model to be able to compare te results with manual calculations.

At the moment I am looking at the body kinematics from a constant circular rotation around the z-axis, but they are not what I would expect. I would expect the acceleration to be the derivative of the velocity, but this is not the case. In the slides in attachement you find the positions, velocities and accelerations calculated by the analyse tool. For the translations in the x and y direction, I also plotted the finite difference approximation (a = (v(t+1)-v(t))/delta_t) and the acceleration I calculated manualy (a =omega^2*r). For the z-direction and the orientations both the finite difference approximation and the manual calculations are 0.

Is there any reason the accelerations calculated by the analyse tool differ from the finite diference approximation and my manual calculations?

Kind regards
Irina

[Carmichael Ong]

Something odd does seem to be happening. Looks like the y-axis may give some clues: possibly an offset of gravity in the results. Do you know if the results make sense if you set the gravity to be 0 in the model?

[Irina Wils]

Thanks for the reply. I set the gravity to 0 and then the -9.81 offset in the y-axis disappears, but I still have accelerations that I can't explain...

[Carmichael Ong]

Given the info so far, it seems weird that gravity should show up in the acceleration plots (since the y-axis position doesn't seem to indicate falling).

You noted that you're trying this first with a simplified model, which is good. How simple is it? Is it just one body with a simple rotation around z?

[Irina Wils]

Thank you for your reply. Yes indeed I would not expect gravity to show up in the acceleration. I am applying a constant rotation, so no falling.

And I included the topology view in the attachment. Here I meant with simplification mainly geometry (I am also looking at dynamics and the differences when weight is seen as a point force or a distributed load). In an earlier phase I already made a model with one body, so I took a look at the body kinematics there. The rotational accelerations are similar to the ones in the graphs of the attachment. The acceleration along the x-axis became 0 and the one along the y-axis -9.81 m/s^2...

[Carmichael Ong]

Sorry, was just trying to get a clear picture of the model being used as it's difficult to tell from a picture and topology view alone what types of joints are in there.

Do you happen to have both models (single body and multibody) that you could share? It could help us debug if there's some underlying issue. Seems like you used BodyKinematics at least, did you use any other analysis tools?

If you're familiar with scripting, you may be able to place a Point object on the model and then use Outputs (https://simtk.org/api_docs/opensim/api_ ... Point.html) to calculate these quantities instead.

[Irina Wils]

To give some more background, the jaw is attached to the skull with 2 tempormandibular joints (TMJ's). So I split the jaw in 2 parts, to avoid problems with the fact that in opensim each body can only have one parent body. In the attachement you find 3 models and their kinematics. only_left has only the left part of the jaw. In with_constraint has a skull with the 2 parts of the mandible as child bodies and a weld constraint between them. In the upside_down model the topology was reversed, as inverse dynamics ignores the constraints and when a estimate of the ground reaction force is added where the skull is attached to the neck, this model's topology is very similar to the one of the gait.

All three models give accelerations that I wouldn't expect and indeed I used BodyKinematics. I didn't use any other analysis tools.

PS: for the expected accelerations see attachement of the first message

[Carmichael Ong]

I was just constructing a simple pendulum model but could not find anything that seems off there with BodyKinematics or PointKinematics. I didn't find any attachment on the forum with the models you described, I wasn't sure if that was in error or I missed an attachment somewhere. Please let me know if there is a model I can take a look at.

A couple other clarification questions to perhaps help debug:
- What tools are you using to constraint the motion to a simple, single rotation?
- You used BodyKinematics to get accelerations. [edit: I'm remembering now that you specified linear accelerations]. You were looking specifically at the center of mass columns then, correct?

If this kinematic information is critical, you could look at using scripting methods with either Body or Point (depending on which you're trying to calculate accelerations for) and look at the corresponding outputs.

[Irina Wils]

Apperently something went wrong with attaching the files, I reattached them now.

- I eddited a .mot file to define the motion. I also added these files in attachement.
- And Indeed I am looking at the center of mass columns of the BodyKinematics, both at the linear and rotational.

Thanks again for the response.

[Carmichael Ong]

I made an even simpler model and motion that seems to replicate the same issue. With a single pendulum (one cylinder body with a single pin joint), I used BodyKinematics and PointKinematics (at the origin of the cylinder) with a fixed position at either 0 degrees (vertical) or 90 degrees (horizontal).

Vertical:

pendulum_0deg.PNG (6.23 KiB) Viewed 1537 times

Horizontal:

pendulum_90deg.PNG (5.45 KiB) Viewed 1537 times

For all cases, BodyKinematics center of mass and PointKinematics (with a point at the cylinder origin) report the same values, and are correct for position and velocity.

Oddly, acceleration depends on the position of the pendulum. For the vertical case, the acceleration in the vertical (y) direction is correctly 0, however for the horizontal case, the acceleration is constant at -8.91505.

This would seem to indicate that it might be good to ignore the acceleration file for now.