[Insight-users] stiffness in deformableMesh3D

[Quan Chen]

What is stiff matrix?  Is there any reference?  I googled and check
wikipedia.  I am not sure if the implementation in the deformableMesh3D is
correct.
In ComputeDT(), the following code appears to calculate the internal force
on each node in a triangular element through deformation, and sum them up,
and check the external force and subtract the internal force, if there is
remaining force left, that will drive the node further.  Appear to make
sense.  However,
first of all, why the stiffness matrix is applied in this way???
deformation from node 1 is simply scaled by m_K[i][0][0]???   I know there
are m_K[i][1][0] used, but those code also don't make sense to me.
I am wondering if we can just assume linear elastic model so that each node
is associated with a scalar k, then the internal deformation force will be
the displacement relative to the other two node then multiply with that k.
It just my experience that the current filter relies heavly on the external
force, the stiffness never behaved correctly.  Often, I saw sharp spikes
rise out from my mesh surface, that should never happen.

Another improvement suggestion:  the SetDefaultStiffnessMatrix() only set
the component for m_StiffnessMatrix[0].  I suggest to set the same value for
all 10 components.  The code can allow people to apply different stiffness
matrix on different nodes depending on their cell value.  However, people
may mistakenly set their cell value other than 0.  Then they will find that
the stiffness setting never have effect.  I met that situation before.  I
use a "MeshUtilNew.h" file shared by someone in this maillist.  Its
PolyDataToMesh function set the cell value to 3 in the code.  I am sure
other people may also encountered this issue. And it is hard to notice
because it will just set my stiffness to 0 and it still works.
Also, in the current design SetMeshStiffness function allocates the
stiffness mesh for each cell, not sure that is a good approach, since you
only can have 10 different stiffness mesh. Why can't we put the check of
celldata value in ComputeDT?  And what if the cell value is greater than
10?  Maybe need to use % or some warning to make sure it doesn't cause
crash.


code in the ComputeDT():

.....

    m_Displacements->GetPoint (tp[0], v1_pt);
    m_Displacements->GetPoint (tp[1], v2_pt);
    m_Displacements->GetPoint (tp[2], v3_pt);
    v1[0] *= m_K[i]->get(0, 0)*p;
    v1[1] *= m_K[i]->get(0, 0)*p;
    v1[2] *= m_K[i]->get(0, 0)*p;
    v2[0] *= m_K[i]->get(0, 1)*p;
    v2[1] *= m_K[i]->get(0, 1)*p;
    v2[2] *= m_K[i]->get(0, 1)*p;
    v3[0] *= m_K[i]->get(0, 2)*p;
    v3[1] *= m_K[i]->get(0, 2)*p;
    v3[2] *= m_K[i]->get(0, 2)*p;
    v1[0] += v2[0]+v3[0];
    v1[1] += v2[1]+v3[1];
    v1[2] += v2[2]+v3[2];
    m_Forces->GetPoint (tp[0], v2_pt);
   v2[0] -= v1[0];
    v2[1] -= v1[1];
    v2[2] -= v1[2];
    m_Forces->SetPoint (tp[0], v2);
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20070615/69f5a139/attachment.html