[Gmsh] Complexity of the model

[Roman Putanowicz]

> Hello,
> 
> I have to make a thermal analysis on a very accurate model, actually the
> one the company uses for the design. I am afraid it will be too complex for
> my analysis.
> 
> I was wondering if there is a way to estimate the gross complexity of a
> model, before meshing it. May be according to the size of the step files,
> the number of parts or something else I do not know.

Dear Adrien, 

I do not know any of such estimates but I would imagine that the complexity
of the model measured in number of nodes depends on the geometric properties
of the mesh, like volume, boundary surface area and first of all curvature
of the geometric elements. In case of a non uniform meshes and very complex
geometries it can be very difficult to obtain the estimate. Thus the first
step to build an estimate would be to assume uniform mesh. Then I would 
just mesh the boundary curves of my geometric model (just create the wireframe
mesh). This 1D mesh would be quick to obtain and  would give me the basis
for estimating volumic mesh complexity by giving necessary edge length 
distribution.  And then we can take different routes. One would be to 
estimate the volume of the model (it can be a bounding box of the model
or some more accurate estimation obtained by approximating the model with
smaller boxes). Then assuming that the boxes are most desirably filled with
regular hexagons or regular tetrahedra and having the edge length distribution
we can calculate the volume of such regular cell and then dividing the total
volume of the model by it we can get the estimate of then number of cells
(elements). In case of non uniform meshes we can use different volumes 
of the cells in each part (box) of the model.
The number of nodes can be estimated by looking at usual
connectivity patterns. There was a paper (I think Shephard et. al)
(I can dig it out if necessary)  where they give the statistical
dependence between number of cells and number of nodes.

A similar thing could be done with model surface.

This is just an vague idea and not direct solution of your problem.
Implementing it would take probably longer than generating the mesh 
and seeing it complexity :-)).
However maybe some meshing tools provide such estimates.

Regards,

Roman
-- 
Roman Putanowicz, PhD  < putanowr at l5.pk.edu.pl  >
Institute for Computational Civil Engng (L-5)
Dept. of Civil Engng, Cracow Univ. of Technology
www.l5.pk.edu.pl, tel. +48 12 628 2569, fax 2034