Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Create free Quad mesh with different element size dependent on surface parameter
Posted Apr 29, 2015, 12:15 p.m. EDT Mesh, Structural Mechanics Version 4.4, Version 5.0, Version 5.1 2 Replies
Please login with a confirmed email address before reporting spam
Hi,
I am running structural mechanics simulations to calculate stress from vertical deformations. It is very simple, think about a cuboid of an elastic material and the surface gets deformed. My input data is a matrix - lets say 100 x 100 and the values represent the displacement in z direction.
To solve the problem I prescribe the displacement in COMSOL.
Now my first approach was just to use a mapped mesh, however, this is highly inefficient since I have large areas where no deformations occur.
I am therefore searching to create a better mesh (Quad) of the surface with either COMSOL or another (free?) tool.
Any hints on this would be highly appreciated.
Thank you
Philipp
I am running structural mechanics simulations to calculate stress from vertical deformations. It is very simple, think about a cuboid of an elastic material and the surface gets deformed. My input data is a matrix - lets say 100 x 100 and the values represent the displacement in z direction.
To solve the problem I prescribe the displacement in COMSOL.
Now my first approach was just to use a mapped mesh, however, this is highly inefficient since I have large areas where no deformations occur.
I am therefore searching to create a better mesh (Quad) of the surface with either COMSOL or another (free?) tool.
Any hints on this would be highly appreciated.
Thank you
Philipp
2 Replies Last Post Apr 30, 2015, 4:21 a.m. EDT
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Hi
I have some difficulties to follow you:
1) how do you define the deformations ? is it taken/measured on a fixed grid ? of so is it measured on the deformed boundary (spatial frame and not the material frame) ?
2) why should Quads be better than the default mesh ? OK if its to make the displacements coincide with a node, but again in which frame ?
3) large areas with no deformations ? well then why mesh it ?
COMSOL has a large number of mesh options so you can tailor it extensively.
I usually cut/partition up my Domains with internal "mesh boundaries" to help the mesher to adapt to the easiest and "coarsest" mesh possible, with only details where I want.
By the way there was a nice Blog issued a few days ago on Domain Partitioning for improved mesh quality, take a look
--
Good luck
Ivar
I have some difficulties to follow you:
1) how do you define the deformations ? is it taken/measured on a fixed grid ? of so is it measured on the deformed boundary (spatial frame and not the material frame) ?
2) why should Quads be better than the default mesh ? OK if its to make the displacements coincide with a node, but again in which frame ?
3) large areas with no deformations ? well then why mesh it ?
COMSOL has a large number of mesh options so you can tailor it extensively.
I usually cut/partition up my Domains with internal "mesh boundaries" to help the mesher to adapt to the easiest and "coarsest" mesh possible, with only details where I want.
By the way there was a nice Blog issued a few days ago on Domain Partitioning for improved mesh quality, take a look
--
Good luck
Ivar
Please login with a confirmed email address before reporting spam
Posted:
9 years ago
Thank you Ivar for the quick reply!
I think the main point I forgot to mention is: I want to automise the mesh generation. It is easy to create a suitable mesh by hand, but I need to do it hundreds of times in dependence e.g. of the gradient of the deformation. the deformation basically changes over time, which is what I measure. But it changes so slow that it can be assumed stationary.
1) I measure the deformation and have it on a grid, with higher resolution as I will ever manage to have nodes (but thats ok). Lets say I basically know the deformed state but also the undeformed.
2) I am not too familiar with FEM or element quality theory, but as I understood it the you can tolerate higher aspect rations with Quad elements. Even if I would take the default mesh, I would still need to feed the mesh generator somehow with the parameter to change localised the element size.
3) Well, I don't know in advance where those areas are, and there are also areas with smaller deformations where not as small elements are needed than somewhere else.
Here in short again what I try to do:
a) start with measured displacement in z - direction
b) create simple geometry without deformation
c) mesh it with element size dependent on the data
d) presribe the displacement as boundary condition (prescribed displacement)
e) run stationary study
f) get z component of cauchy stress tensor
I hope that makes it clearer.
Thanks a lot
Philipp
I think the main point I forgot to mention is: I want to automise the mesh generation. It is easy to create a suitable mesh by hand, but I need to do it hundreds of times in dependence e.g. of the gradient of the deformation. the deformation basically changes over time, which is what I measure. But it changes so slow that it can be assumed stationary.
1) I measure the deformation and have it on a grid, with higher resolution as I will ever manage to have nodes (but thats ok). Lets say I basically know the deformed state but also the undeformed.
2) I am not too familiar with FEM or element quality theory, but as I understood it the you can tolerate higher aspect rations with Quad elements. Even if I would take the default mesh, I would still need to feed the mesh generator somehow with the parameter to change localised the element size.
3) Well, I don't know in advance where those areas are, and there are also areas with smaller deformations where not as small elements are needed than somewhere else.
Here in short again what I try to do:
a) start with measured displacement in z - direction
b) create simple geometry without deformation
c) mesh it with element size dependent on the data
d) presribe the displacement as boundary condition (prescribed displacement)
e) run stationary study
f) get z component of cauchy stress tensor
I hope that makes it clearer.
Thanks a lot
Philipp