[Insight-users] about itk::fem

Yixun Liu yxliu at fudan.edu.cn
Wed Aug 4 03:10:16 EDT 2004


Hi,
I am used to use the Ofelib(one fem lib) to resolve the fem problem, but when I find that itk also has the fem lib I intend to use it in my app. I create a .fem file and use the itk::fem::solver by reading the .fem file. However, two problems make me very puzzled.

1.if the size of the mesh is large, e.g. the number of the nodes > 1000, the solver will take a very very long time(1-2 hours) to finish the computation. I once use other fem libs such as Ofeli which only take 15 minutes.

2.I fix some Dof when applying BC , However, the solution is not same with the values I set on these DOF.

I look at the source code of itk::fem and I find it use the iterative least square to solve the linear system. I do not know if this is the reason causing the very long time computation. Becaus the Ofelib use the CG to solve the linear system and only need very little time.

I only need solve the static linear elastic problem, use the tetrahedron as the element of the mesh. 

<MaterialLinearElasticity>
 0 %Global material number
 E: 3000 % E
 A: 0  % A
 I: 0  % I
 nu: 0.45  % nu
 h: 1
 RhoC: 1
 
<Element3DC0LinearTetrahedronStrain>


thnk you in advance!

Yixun Liu

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/insight-users/attachments/20040804/584f260e/attachment.html


More information about the Insight-users mailing list