Advanced Surface

Hello,

I have scanned a very complex surface of a mold with a CMM. The CMM
outputed a point cloud IGES or Step. Now in ProE I need to take this
point cloud system and create a surface. I started making curves and
blending them together. And it apears to work great. Unfortunatly this
is extremly time comsuming because there is about 50,000 points I need
to click on to make curves the curves to guild the blend, Does anyone
know is there away to create a surface from all these points? David
where you at?

Shaun

 

Yes. Use Raindrop Geomagic. 50,000 points sounds like 3 days rather untaxing
work

 

Wildfire 2.0 has a new auto-surface feature that will take in CMM data
and stitch together a surface for you. You can control the number of
curves and number of surfaces, kind of like a STL file output. The
feature is called "Restyle".

I'd recommend using the default values for the curves and surface
patches first, then adjusting them to get a finer and smoother surface.

This feature is the module of the Raindrop Geomagic software integrated
into Pro/E. It's very cool, but it took me about 3 to 4 hours to
figure out how it works.

Lee

 

Thanks, Lee, that's good to know. But allow me one caveat: Raindrop Geomagic /
Restyle is really Pro/SCANTOOLS "Jr.". Pro/SCANTOOLS and a couple others, like
Pro/CMM, were folded into a package called REX (Reverse Engineering eXtension), a
new module of Wildfire. So, you might be interested in what's available in the
bigdaddy version that the introductory jr version is missing: two things, mainly,
averaging and sparsing.

Averaging: this is data analysis, so take out the obviously errant data. It's
clicking along on a steady trajectory, moving a few tenths at a time, and
suddenly, it spikes, several cms (inches, who cares), then settles back down to
the path it was. Scantools identified this kind of stuff, corrected it with
parsing and averaging, and by extension, so does REX (does 'Restyle'? I don't
think so, but correct me if I'm wrong, please). This data anomoly was always
present in scanned data from plastic parts with shiny surfaces, like those from
die cast or plastic injection molding, less from forged or investment cast parts
which started out with less reflective surfaces (lasers pick up stray reflected
light as different depths of same surface, giving anomolous data).

Sparsing: data analysis, as does mathematics of curves in general, shows that the
fewer the points, the smoother the curves (and thus the surfaces built on them).
Pro/SCANTOOLS had functions built into it that said you have a gazillion points
that seem to roughly follow this trajectory (and from them, they pick the points
that accomodate the creation of the smoothest trajectory). This is also, to some
extent, more "averaging", but, in the end, it will support nice, smooth splines
and NURBS surfaces (does Restyle do this kind of smoothing, this sparsing?).

Last comment: with Pro/e, including Style/ReStyle, you always have numbers.
Personally, I can't live without them. With Raindrop Geomagic, I don't know. Maybe
it comes in like an IGES ~ it just is what it is. Most Pro/e people, though, are
trying to get from Pro/e to Pro/e with some freeform features, some NURBS, some
style stuff, but built on top of basic geometry. Is Raindrop Geomagic, in any way
whatsoever, parametric? If you're in Pro/e, can you ever really abandon
parametrics?

David Janes

 

On top of the problem I raised with using non-parametric NURBS modellers like
Geomagic or Rhino, is the one all of them, including REX, share: they don't
analyze geometry for features and where ever possible, give you engineering
features or even parametric block primitives. So, you don't have features to
modify. How do you change a fillet radius when it doesn't recognize that the part
has one? or a wall thickness? or a cylinder diameter? or the length of what you
can see started out as a rectangular base feature? This would be the best way to
capture para,metrics, the way that Pro/e pioneered. I hope it's working on
incorporating this most difficult 'trick' into REX. Maybe Restyle at least gives
you some style features. But without being able to analyze conventional, feature
based geometry and break it down into features, you're still missing the bridge
between NURBS and solids.

On the other hand, if I'm way off base and it actually shines in this area, I'd
definitely like to hear about it. I think everyone here would like to hear more
about the capabilities of one of PTC's better kept 'secrets' called REX.

David Janes

 

In my experience, the quality of your point cloud [PC] and the type of
geometry in the original form [more primitive or freeform] is the real
decider on the best approach for reverse engineering. However much noise
reduction [taking out 'outlyers' from a specified tolerance] and fine tuning
[thining out your point cloud] of the original PC there is still a tolerance
'thickness' [all points are going to be above or below the original surface]
of your PC which could be reflected in your final model. A surface could
have ripples which reflect this tolerance.

Once you have a facet file from the point cloud you then get the option to
move away from the original data set and 'smooth' out the point cloud. The
facet file can then be made manifold - open [closed boundaries] or closed
volume [no single sided edges], consistant normals.

Note: REX/facet feature - PC to facet feature. REX/Restyle - facet feature
to curves/surfaces.

In REX - Restyle there is a 'best fit' option to create section curves
through the facet file, you then use these to create your surfaces - this is
fundamentally what you are doing by creating curves outside of REX.

There is an option the create a limited range of primitives from the PC
data - rev eng a hole for example. But again, this is only ever going to be
as good as the tolerance of your original data.

In experienced hands REX [and similiar software] can save time but don't
ignore 'manual' methods and keep an eye on the surface tolerance.

Sean

~~~~~~~~~~~~~~~~~~~~~~~~~~
Sean Kerslake
Dept. of Design & Technology
Loughborough University
LE11 3TU

01509 228317

~~~~~~~~~~~~~~~~~~~~~~~~~~


----- Original Message -----
From: "shaun" <>
Newsgroups: comp.cad.pro-engineer
Sent: Friday, February 25, 2005 4:20 PM
Subject: Advanced Surface

--
~~~~~~~~~~~~~~~~~~~~~~~~~~
Sean Kerslake
Dept. of Design & Technology
Loughborough University
LE11 3TU

01509 228317

~~~~~~~~~~~~~~~~~~~~~~~~~~