How to convert a big mesh to solid

[ertan]

Hi all,


Is there any method other than shrinkwrap of Pro/E to convert a big mesh (1,500,00 faces) to solid. I tried meshtonurbs in Rhino, but it limited me with only 20,000 faces. Any help is appreciated.


thanks

 

[jeff4136]

What's the goal? I can't imagine a practical purpose and suspect a
1.5 M planar face solid rep (vs. mesh object) will drag the system to
its knees.

 

[ertan]

The geometry is a full human body with cloths on. I will use this model to simulate flow around it. The model, therefore, has to be very smooth. I can reduce the number of faces to some extent but 20,000 is way too less.

 

[jeff4136]

I'm still lost.
What is "smooth"?
G1 & G2 NURBS surfaces?
(Rhino MeshToNURBS just converts facet elements to degree 1x1 planar faces.)


> simulate
Finite element?
Import your existing mesh?

Sorry, I don't know much about what you're doing or what you know about it.
For some interesting (look for their example files, a head among them) scan
to NURBS model info www.dirdim.com
Smooth is the domain of $20K - $30K software suites and skilled modelers,
I believe.

 

[ertan]

Hi,


The smoothness I am talking about is the smoothness of the surface of the body. I do not want any corner or sharp edges on the surface of the human body. As humanswedo not have sharp parts, all we haveare smoothcurves. Ithink you agree with me that the more points (or facets), the smoother is your model. Because youuse more data to define a curvature.


Anyway, I am Fluid Dynamist, and I don't have much knowledge about NURBS surface, but as far as I knowit isbased on Spline interpolation, which generally yields curves (curvy surfaces)in stead of planar surfaces. I don't oppose toyour knowledge, I am just telling what I was assuming. Because I know that Spline interpoaltion is based on equating the derivatives (first, second or third order derivatives) at the common points. Anyway NURBS maybe completely different. However, even if NURBS yields planar faces, If you have enough faces (I think I have), it is supposed to give me still a nice smooth surface. Because you can represent a curve perfectly with straight linesif you have enouh of them.


I am doing Finite Volume Simulation. I can't import my existing mesh, it causes other problems. I don't want to go into details about it.


As I said ,my expertise is completely different, and I do not have enough knowledge about these things.


Thanks for your interest

 

[jeff4136]

> If you have enough faces ... can represent a curve perfectly with
> straight lines if you have enouh of them.


It's an approximation. Once it's 'perfect' the lines are reduced to zero
length; points. But, as you indicate, that's the essence of finite element
meshing, creating a mesh to 'accurately enough' approximate a 'smooth'
surface without needlessly overloading the system.


Maybe someone else can enlighten the both of us but I don't see any future
in trying to get a 'solid' or NURBS surface representation of the object with
Pro/E (going back to the Direct Dimensions reference ... REXX being, I think,
PTC's quite pricey in it's own right Reverse Engineering offering for fitting
'manufacturable', or 'meshable' in this case, surfaces to a scan set or mesh
model).


Without knowing what problems a straight up analysis front end import of the
existing mesh model presents I can only guess that converting to planar NURBS
(ala Rhino MeshToNURBS, or analytic planes for that matter) will solve nothing
and size of the inefficiently represented data set may introduce new problems.


If 'twer me, in the absence of any better suggestions, I'd get with the target
software support people or related user group and kick around options for using
what you have without the intermediate conversion or a more efficient conversion
to another copious data format or modification with a mesh modeler (ReStyle,
Pro/Concept, ...?) if that's what's needed or ...?