More surface modeling questions

SW 2005

I have been trying to get up to speed on surface modeling the last couple of
days. I'm making good progress thanks to the suggestions this group has
offered.

Let me restate the problem and see if anyone has any ideas:
I am importing a 'point cloud' iges file. It is a 3d scan of a surface of an
automobile. It comes in as more than 500 faces that make up the surface.
There are holes in the surface and jagged edges. I can delete these ragged
faces and fill them in later. When the surface comes in, it is not
associated with any coordinate system. I can move it around using Insert,
Surface, Move/Copy. What I would like to do is take some of these faces and
associate them coincident to the standard planes on which I could then draw
sketches to set fill boundaries. Since this is just a part file, I can not
use mates to associate the faces with planes. Is there an equivalent way to
do this 'mating' in a part file? I realize I can create new refernce planes
to sketch on but I would like to use the standared planes if I can.

 

Mark,

Jeezis, you're doing it the hard way. I'd jump off a bridge before going
through all that. Is there any real requirement or is this just your idea?
I'd recommend you forget about the coordinate system and planes altogether.
Don't make sketches on planes. Complete waste of time. Use 3D sketches to
connect vertices with lines and do fills or lofts. Curve through reference
points would be even faster if SW could use them for fills, but the curve
entities can only be used by lofts.

Good luck.

Matt

 

<sigh>

A) You have far too many data points.
B) Each one was measured to within a tolerance. What
was it? As a result, each data point can *differ* from the
desired surface by that tolerance. This is bound to produce
bumps, lumps, dents & ripples.

C) What you need to do is to think about the surface.
How many control points do you think it needed?
Where? Why?
What are it's "planes" of symmetry? Things like that.

Tried fitting some simple *planar* curves to some of
your selected points? How do they look?
Offset the curves ... look for odd stuff in the offsets ... do
they cross themselves? After how much offsetting?

You can "adjust" the points by the amount of that tolerance ...

D) Can of worms.

 

How difficult would it be for you to redo the scan of the automobile
surface yourself ? This way you could check as you go along and see if
you have the right amount of point data to develop the surfaces you
need. It sounds to me like the person doing the scan of the automobile
surface did not have an idea of what was needed by the person who would
take the point cloud data and create surfaces from it.

jon

 

I posted an image of my progress on this part on my web site. Click on the
Solidworks page:
http://home.bellsouth.net/p/s/community.dll?ep=16&groupid=145864&ck=

I know some of you are going to have a hoot when you see how many steps I
went through to do this. I will take some of the posted advice to try
streamline it.
Mark

 

I don't have access to that scanning equipment or the vehicle as I am still
interviewing for the job. If I get the job, I will be learning to use the 3d
scanning equipment, I think. The manager I am interviewing with told me the
guy who supplied the iges data intentially moved the surfaces from the
original coordinates to test the job applicant (me). It threw me off because
I like to have everything pretty well locked down before I start messing
with it. I'm learning that this surface modeling is more 'arts and crafts'
than the familiar solid modeling. No liner edges to work from.
Mark

 

Mark,

"The manager I am interviewing with told me the guy who supplied the
iges data intentionally moved the surfaces from the original
coordinates to test the job applicant (me)."

I don't see how one can create a proper surface model when the point
cloud data is moved and you have no way to determine where it was moved
from.

"I'm learning that this surface modeling is more 'arts and crafts' than
the familiar solid modeling."

IMO, surface modeling is a poorly documented skill and this is what
makes you and many others believe that it's "arts and craft" or that
one needs to be a math genius to do surface modeling. IMO, neither is
true.

I've been working very hard to try and get the advanced surfacing of
the product I prefer to be better documented. I have not met with
resistance. :>)

jon

 

This from a clueless person that's never actually done any of it.

Here's what a brick would probably look like if jb tried:
http://www.digital-eel.com/organism/images/hangin_out_blob.jpg

By now I'd think that 3dinkies knows you all too well.

 

IGES surfaces or just point clouds?
What was used to produce the data?
Some CMMs & such can output surface data in VDA
(and probably other) formats ... but I don't know how
they are generating te surfaces from the measured probe
hits.

BTW, Why not just do your work where the data is. Why move it?

 

OT comment: Nice RV-4 project, Mark. Looks like you cornered the market
on Cleco clamps

 

Mark,

From this, and your last post, I'm assuming that moving the scanned point
cloud and resulting surfs to a "normal" plane within SW is the "test". But
it sure would be a lot easier for you to download a trial copy of Rhino,
import the surfs, move to the desired location, and export to SW.
I do this sometimes when I get an IGES file and/or a parasolid file that
doesn't orient the way I visualize it.
BTW, keep in mind that SW is a Y up modeler and Rhino is a Z up modeler.
I'd be happy to do it in Rhino for you, but I would assume that that would
not be part of your test.

HTH,
Muggs

 

Always amazing to see how many SaladWorks users have no choice but to
use Rhino because SaladWorks does not have the tools to get the job
done.

www.vx.com


jon

 

VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX
VXVXVXVXVXVXVXVXVXVXVX

 

I missed where you offered the poster *any* suggestions.

Any reason you didn't offer any suggestions ?

Is it hard for you to deal with the FACT that SaladWorks does not have
the proper tools and often fails in many real world situations where VX
does have the needed tools ?

www.vx.com

jon

 

Thanks Muggs, but I was able to accomplish my task by creating reference
planes on a couple of flat surfaces, then deleting the jagged surface
elements, then drawing 3d sketches coincident to the new planes and filling
to those sketches. I then thickened the surface to make the part. I did pull
the imported surface around a bit using Insert, Surface, Move but I still
don't know if there is a way to hook it to the standard planes. One
interesting thing about this exercise was how my computer choked on the
part. The iges file was 19 MB and it created a SW part of 195 MB! My
computer with 500 MB ram could not handle it. I used Rhino to whittle it
down to a small chunk and pulled it back into Solidworks. One odd thing was
that the surface would only thicken to 0.070". I wanted 0.080" but it would
not go. I wonder how to diagnose this problem. The only thing I could figure
was it must have encountered self intersecting geometry when I tried to go
to 0.080". Any ideas on this? And another thing, in Rhino, how would you
relocate the surfaces to a coordinate system? Is it just by eyeball, or
what.
Mark

 

You have a known band of .070" - .080" it seems.
Let N = .080-.070 = .010
IF you can create surfaces in the following way (this is simple
with an API program, probably a chore manually): Start with the
known .070" surface. From there offset a surface by N/2 in the
direction of known failure. Intersect it with the original surface.

A) You have intersection.
B) You do not.

IF B then offset the offset surface by N/4 in the direction of
failure & repeat the intersection check with the first offset
surface. See A & B above. Repeat until intersection found,
N/8, N/16, etc. Then reverse offset direction & try some more.
You limit the number of steps with a lower limit for N/(2**K)

If A then reverse the offset direction from the offset surface by
half of the prior offset distance & try again. See above method <g>.

Simple, eh?

BTW, The last *total offest* distance found, from the *original*
parent surface, is the minimum 3D radius of curvature of the original
parent surface +/- your lower limit distance (or twice that limit,
depending on if you halted under intersect or nonintersect
conditions - it was your choice).

Example Note for any *actual* programmers lurking:

Assume that this distance is small, and that your limit test
distance ended at .001" with no intersection.
The number that you ended up with is also the largest
radius ball endmill (in units of .001") that can 100% machine
the parent surface without gouging or leaving uncut stock.
This is sort of handy for automatically selecting tools for cutting
unknown surfaces IF you do the API programming needed <G>.

I assume that was confusing .....

 

Mark,

I don't know how the iges files were created, but you had said that is was
made from scan data.
Some scanning packages create iges "surfaces" by basically taking an STL and
putting a single surface between the three legs of each and every triangle.
So needless to say that makes for a VERY large file and can typically become
non-manifold.
If it were done in Raindrop Geomagic then the surfaces could be as good or
as bad as the person creating the surfaces in Geo. Which makes me think,
when you say that there are "holes" in the surfaces, that not very much time
was spent in Geo to do a nice job.

Jon,
I have purposely stayed out of the c.c.soildworks VS Jon & Cliff debate, but
if you think that you can use one and only one software package to do all
that is needed to "get the job done" then you have once again shown your
hand.
And before you tag me. Yes I have downloaded VX and found it to be
cumbersome. But maybe it's just me.
One thing I think we all tend to forget is that some software packages just
"fit" some peeps better that others.
I have been using MasterCam since version 3 (It was called Clear-Cut then),
but I have a very good friend that swears by Delcam. Go figure!

Anyway, my 2 cents,
Muggs

 

As I've mentioned before, some CMMs can actually output surfaces
in various formats (I've worked with VDA from them), not just
pointsets or triangles.
I see no reason that other scanners of various sorts cannot do the
same.
How they derive the surfaces from the probe ponts I do not know
<G>.
The results of what I've seen are much better than I'd have
expected.

 

"I have purposely stayed out of the c.c.soildworks VS Jon &
Cliff debate, but if you think that you can use one and only
one software package to do all that is needed to "get the
job done" then you have once again shown your hand."

I have never said one package should do it all. When you
make this implication you show how your really biased
despite the denial you start off with.

"And before you tag me."

You showed your hand / bias in your first sentence. IOW, you
tagged yourself.

What I have said is that the proper hybrid modeler negates
the reason to move back and forth between a surface modeler
and a solid modeler. IMO, a SaladWorks user is forced to
use Rhino because the SaladWorks dressing can't cover a
product that has been poorly conceived and designed. How
many Unigrahpics users have to depend on Rhino to get the
job done ?

"Yes I have downloaded VX and found it to be cumbersome."

It can be but VX is a company who unlike SaladWorks Corp.
started with the proper architecture. It's much easier to
make the kinds of changes that I and others would like to
see VX have then the kinds of changes it will take to make
SaladWorks a truly seamless, unified, hybrid modeler. VX has
the foundation that SaladWorks so badly lacks and will
continue to lack. VX showed how much progress they can make
UI wise in just one release from V10 to V11. V11 goes a long
way to removing the "cumbersome" issue that you raised and I
agree with. Still much work to do. Why waste time with a
product has the wrong architecture when you could be helping
to make VX better and end up with a much more complete
solution than the cobbled together approach of SaladWorks
and Rhino ?

"I have been using MasterCam since version 3 (It was called
Clear-Cut then), but I have a very good friend that swears
by Delcam. Go figure!"

Nothing to figure. Your friend probably needs the higher
quality 3 axis toolpath surfacing (at the expense of 2 1/2
axis toolpath control which MasterCAM is better at) than
MasterCAM can deliver. Not sure why you find this so
amazing.

jon

 

http://www.geomagic.com/products/studio/