About Joint constraint forces

Archive AnyScript Yahoo archive
1

Dear Anybody team,

I recently made a simple 3-segment model to use motion data from sway
study.
Three segments include one foot, one shank and one thigh, and
revolute joints are used to connect each segment. (and no muscle)

When checking constraint reaction forces , I saw 5 components.
The thing is that constraint reaction forces for the joint are
somewhat unreasonable.

For the ankle, a compressive force (Fy), M-L force (Fz), and a joint
torques are reasonable. However, an A-P force (Fx) is way too large
(about 100 N for the ankle although I expected a somewhat similar
value from the forceplate).

I don’t see why only A-P directoin forces look stranges when all the
other reaction forces look fine.
Pleaes, let me know what I’m missing here.

Sincerely,
Sunwook

2

Hello Sunwook,

It sounds like a really simple model that should be easy for us to
decipher. Could you upload the model to the group, so that we can
have a look at it?

Thanks,
John

— In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@…>
wrote:
>
> Dear Anybody team,
>
> I recently made a simple 3-segment model to use motion data from
sway
> study.
> Three segments include one foot, one shank and one thigh, and
> revolute joints are used to connect each segment. (and no muscle)
>
> When checking constraint reaction forces , I saw 5 components.
> The thing is that constraint reaction forces for the joint are
> somewhat unreasonable.
>
> For the ankle, a compressive force (Fy), M-L force (Fz), and a
joint
> torques are reasonable. However, an A-P force (Fx) is way too
large
> (about 100 N for the ankle although I expected a somewhat similar
> value from the forceplate).
>
> I don’t see why only A-P directoin forces look stranges when all
the
> other reaction forces look fine.
> Pleaes, let me know what I’m missing here.
>
>
> Sincerely,
> Sunwook
>

3

The file is uploaded.
Thanks for the help.

Sunwook

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…>
wrote:
>
> Hello Sunwook,
>
> It sounds like a really simple model that should be easy for us to
> decipher. Could you upload the model to the group, so that we can
> have a look at it?
>
> Thanks,
> John
>
>
> — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> wrote:
> >
> > Dear Anybody team,
> >
> > I recently made a simple 3-segment model to use motion data from
> sway
> > study.
> > Three segments include one foot, one shank and one thigh, and
> > revolute joints are used to connect each segment. (and no muscle)
> >
> > When checking constraint reaction forces , I saw 5 components.
> > The thing is that constraint reaction forces for the joint are
> > somewhat unreasonable.
> >
> > For the ankle, a compressive force (Fy), M-L force (Fz), and a
> joint
> > torques are reasonable. However, an A-P force (Fx) is way too
> large
> > (about 100 N for the ankle although I expected a somewhat similar
> > value from the forceplate).
> >
> > I don’t see why only A-P directoin forces look stranges when all
> the
> > other reaction forces look fine.
> > Pleaes, let me know what I’m missing here.
> >
> >
> > Sincerely,
> > Sunwook
> >
>

4

Hi Sunwook,

I had a quick look and can see that the input data is very jagged.
This is a classical problem of mocap input: When you differentiate
it, the noise increases a lot. I noticed that you use an 8th order B-
spline interpolation, but this is frequently not enough.

Unless you have already tried it, my advice is to try to filter the
input movement data, for instance with a Butterworth filter.

Please let us all know how it goes.

Best regards,
John

— In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@…>
wrote:
>
> The file is uploaded.
> Thanks for the help.
>
> Sunwook
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> wrote:
> >
> > Hello Sunwook,
> >
> > It sounds like a really simple model that should be easy for us
to
> > decipher. Could you upload the model to the group, so that we
can
> > have a look at it?
> >
> > Thanks,
> > John
> >
> >
> > — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> > wrote:
> > >
> > > Dear Anybody team,
> > >
> > > I recently made a simple 3-segment model to use motion data
from
> > sway
> > > study.
> > > Three segments include one foot, one shank and one thigh, and
> > > revolute joints are used to connect each segment. (and no
muscle)
> > >
> > > When checking constraint reaction forces , I saw 5 components.
> > > The thing is that constraint reaction forces for the joint are
> > > somewhat unreasonable.
> > >
> > > For the ankle, a compressive force (Fy), M-L force (Fz), and a
> > joint
> > > torques are reasonable. However, an A-P force (Fx) is way too
> > large
> > > (about 100 N for the ankle although I expected a somewhat
similar
> > > value from the forceplate).
> > >
> > > I don’t see why only A-P directoin forces look stranges when
all
> > the
> > > other reaction forces look fine.
> > > Pleaes, let me know what I’m missing here.
> > >
> > >
> > > Sincerely,
> > > Sunwook
> > >
> >
>

5

Dear John,

Thanks for the quick response.

Actually, I did use a 4th order Butterworth filter and tried a couple
of different cut-off frequencies (ranging from 1 to 3.5 Hz given the
motion data was captured at 20 Hz).

In addition, I numerically twice-differentiated motion data to see
what kind of acceleration data I’ve got. The order of magnitude was
about 1e-2 to 1e-4, depending on a cut-off freq.

I expected that anyhow I would have a small A-P direction joint force
for the anckle since the foot mass is very small.

Regardless of a cut-off frequency, I had a small M-L direction joint
force and a large A-P direction joint force for the joint.

So far, I couldn’t think of what I did wrong.

Sunwook

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…>
wrote:
>
> Hi Sunwook,
>
> I had a quick look and can see that the input data is very jagged.
> This is a classical problem of mocap input: When you differentiate
> it, the noise increases a lot. I noticed that you use an 8th order
B-
> spline interpolation, but this is frequently not enough.
>
> Unless you have already tried it, my advice is to try to filter the
> input movement data, for instance with a Butterworth filter.
>
> Please let us all know how it goes.
>
> Best regards,
> John
>
> — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> wrote:
> >
> > The file is uploaded.
> > Thanks for the help.
> >
> > Sunwook
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > wrote:
> > >
> > > Hello Sunwook,
> > >
> > > It sounds like a really simple model that should be easy for us
> to
> > > decipher. Could you upload the model to the group, so that we
> can
> > > have a look at it?
> > >
> > > Thanks,
> > > John
> > >
> > >
> > > — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> > > wrote:
> > > >
> > > > Dear Anybody team,
> > > >
> > > > I recently made a simple 3-segment model to use motion data
> from
> > > sway
> > > > study.
> > > > Three segments include one foot, one shank and one thigh, and
> > > > revolute joints are used to connect each segment. (and no
> muscle)
> > > >
> > > > When checking constraint reaction forces , I saw 5 components.
> > > > The thing is that constraint reaction forces for the joint
are
> > > > somewhat unreasonable.
> > > >
> > > > For the ankle, a compressive force (Fy), M-L force (Fz), and
a
> > > joint
> > > > torques are reasonable. However, an A-P force (Fx) is way too
> > > large
> > > > (about 100 N for the ankle although I expected a somewhat
> similar
> > > > value from the forceplate).
> > > >
> > > > I don’t see why only A-P directoin forces look stranges when
> all
> > > the
> > > > other reaction forces look fine.
> > > > Pleaes, let me know what I’m missing here.
> > > >
> > > >
> > > > Sincerely,
> > > > Sunwook
> > > >
> > >
> >
>

6

Hi Sunwook,

I had another look, and there is a perfectly natural explanation:
The reactions in joints refer to one of the two reference frames
connected by the joint. By default is is the first reference frame
you mention in the joint. So, for instance

AnyRevoluteJoint VirtualAnkle = {
Axis = z;
AnyRefFrame &GroundNode = .MyModel.VirtualShank.VirtualAnkleNode;
AnyRefFrame &KneeNode = .MyModel.VirtualFoot.VirtualAnkleNode;
};

means that the forces gets computed in the VirtualShank coordinate
system. Since the shank is slightly slanted with repect to vertical,
some of the globally vertical force in the joint gets projected to
the local x direction. So the reaction forces you see in the x
direction are in the shank’s coordinate system and not in the global
coordinate system.

Best regards,
John

— In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@…>
wrote:
>
> Dear John,
>
> Thanks for the quick response.
>
> Actually, I did use a 4th order Butterworth filter and tried a
couple
> of different cut-off frequencies (ranging from 1 to 3.5 Hz given
the
> motion data was captured at 20 Hz).
>
> In addition, I numerically twice-differentiated motion data to see
> what kind of acceleration data I’ve got. The order of magnitude
was
> about 1e-2 to 1e-4, depending on a cut-off freq.
>
> I expected that anyhow I would have a small A-P direction joint
force
> for the anckle since the foot mass is very small.
>
> Regardless of a cut-off frequency, I had a small M-L direction
joint
> force and a large A-P direction joint force for the joint.
>
> So far, I couldn’t think of what I did wrong.
>
>
> Sunwook
>
>
>
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> wrote:
> >
> > Hi Sunwook,
> >
> > I had a quick look and can see that the input data is very
jagged.
> > This is a classical problem of mocap input: When you
differentiate
> > it, the noise increases a lot. I noticed that you use an 8th
order
> B-
> > spline interpolation, but this is frequently not enough.
> >
> > Unless you have already tried it, my advice is to try to filter
the
> > input movement data, for instance with a Butterworth filter.
> >
> > Please let us all know how it goes.
> >
> > Best regards,
> > John
> >
> > — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> > wrote:
> > >
> > > The file is uploaded.
> > > Thanks for the help.
> > >
> > > Sunwook
> > >
> > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > > wrote:
> > > >
> > > > Hello Sunwook,
> > > >
> > > > It sounds like a really simple model that should be easy for
us
> > to
> > > > decipher. Could you upload the model to the group, so that
we
> > can
> > > > have a look at it?
> > > >
> > > > Thanks,
> > > > John
> > > >
> > > >
> > > > — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> > > > wrote:
> > > > >
> > > > > Dear Anybody team,
> > > > >
> > > > > I recently made a simple 3-segment model to use motion
data
> > from
> > > > sway
> > > > > study.
> > > > > Three segments include one foot, one shank and one thigh,
and
> > > > > revolute joints are used to connect each segment. (and no
> > muscle)
> > > > >
> > > > > When checking constraint reaction forces , I saw 5
components.
> > > > > The thing is that constraint reaction forces for the joint
> are
> > > > > somewhat unreasonable.
> > > > >
> > > > > For the ankle, a compressive force (Fy), M-L force (Fz),
and
> a
> > > > joint
> > > > > torques are reasonable. However, an A-P force (Fx) is way
too
> > > > large
> > > > > (about 100 N for the ankle although I expected a somewhat
> > similar
> > > > > value from the forceplate).
> > > > >
> > > > > I don’t see why only A-P directoin forces look stranges
when
> > all
> > > > the
> > > > > other reaction forces look fine.
> > > > > Pleaes, let me know what I’m missing here.
> > > > >
> > > > >
> > > > > Sincerely,
> > > > > Sunwook
> > > > >
> > > >
> > >
> >
>

7

Dear John,

I sincerely appreciate your help.
I just ran the model again and had reasonable values.
Somehow, I thought the joint force is by default expressed in a
distal segment cooridnate system.
Thanks, again.

Best regards,
Sunwook

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…>
wrote:
>
> Hi Sunwook,
>
> I had another look, and there is a perfectly natural explanation:
> The reactions in joints refer to one of the two reference frames
> connected by the joint. By default is is the first reference frame
> you mention in the joint. So, for instance
>
> AnyRevoluteJoint VirtualAnkle = {
> Axis = z;
> AnyRefFrame &GroundNode
= .MyModel.VirtualShank.VirtualAnkleNode;
> AnyRefFrame &KneeNode = .MyModel.VirtualFoot.VirtualAnkleNode;
> };
>
> means that the forces gets computed in the VirtualShank coordinate
> system. Since the shank is slightly slanted with repect to
vertical,
> some of the globally vertical force in the joint gets projected to
> the local x direction. So the reaction forces you see in the x
> direction are in the shank’s coordinate system and not in the
global
> coordinate system.
>
> Best regards,
> John
>
>
>
> — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> wrote:
> >
> > Dear John,
> >
> > Thanks for the quick response.
> >
> > Actually, I did use a 4th order Butterworth filter and tried a
> couple
> > of different cut-off frequencies (ranging from 1 to 3.5 Hz given
> the
> > motion data was captured at 20 Hz).
> >
> > In addition, I numerically twice-differentiated motion data to
see
> > what kind of acceleration data I’ve got. The order of magnitude
> was
> > about 1e-2 to 1e-4, depending on a cut-off freq.
> >
> > I expected that anyhow I would have a small A-P direction joint
> force
> > for the anckle since the foot mass is very small.
> >
> > Regardless of a cut-off frequency, I had a small M-L direction
> joint
> > force and a large A-P direction joint force for the joint.
> >
> > So far, I couldn’t think of what I did wrong.
> >
> >
> > Sunwook
> >
> >
> >
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > wrote:
> > >
> > > Hi Sunwook,
> > >
> > > I had a quick look and can see that the input data is very
> jagged.
> > > This is a classical problem of mocap input: When you
> differentiate
> > > it, the noise increases a lot. I noticed that you use an 8th
> order
> > B-
> > > spline interpolation, but this is frequently not enough.
> > >
> > > Unless you have already tried it, my advice is to try to filter
> the
> > > input movement data, for instance with a Butterworth filter.
> > >
> > > Please let us all know how it goes.
> > >
> > > Best regards,
> > > John
> > >
> > > — In anyscript@yahoogroups.com, “ninegu_sun” <ninegu_sun@>
> > > wrote:
> > > >
> > > > The file is uploaded.
> > > > Thanks for the help.
> > > >
> > > > Sunwook
> > > >
> > > > — In anyscript@yahoogroups.com, “AnyBody Support”
<support@>
> > > > wrote:
> > > > >
> > > > > Hello Sunwook,
> > > > >
> > > > > It sounds like a really simple model that should be easy
for
> us
> > > to
> > > > > decipher. Could you upload the model to the group, so that
> we
> > > can
> > > > > have a look at it?
> > > > >
> > > > > Thanks,
> > > > > John
> > > > >
> > > > >
> > > > > — In anyscript@yahoogroups.com, “ninegu_sun”
<ninegu_sun@>
> > > > > wrote:
> > > > > >
> > > > > > Dear Anybody team,
> > > > > >
> > > > > > I recently made a simple 3-segment model to use motion
> data
> > > from
> > > > > sway
> > > > > > study.
> > > > > > Three segments include one foot, one shank and one thigh,
> and
> > > > > > revolute joints are used to connect each segment. (and no
> > > muscle)
> > > > > >
> > > > > > When checking constraint reaction forces , I saw 5
> components.
> > > > > > The thing is that constraint reaction forces for the
joint
> > are
> > > > > > somewhat unreasonable.
> > > > > >
> > > > > > For the ankle, a compressive force (Fy), M-L force (Fz),
> and
> > a
> > > > > joint
> > > > > > torques are reasonable. However, an A-P force (Fx) is way
> too
> > > > > large
> > > > > > (about 100 N for the ankle although I expected a somewhat
> > > similar
> > > > > > value from the forceplate).
> > > > > >
> > > > > > I don’t see why only A-P directoin forces look stranges
> when
> > > all
> > > > > the
> > > > > > other reaction forces look fine.
> > > > > > Pleaes, let me know what I’m missing here.
> > > > > >
> > > > > >
> > > > > > Sincerely,
> > > > > > Sunwook
> > > > > >
> > > > >
> > > >
> > >
> >
>