Biasotti Advanced Modeling 101 thought... fit face to a set of faces

I really regret that I had to present so much at this SWx World, because
with that many smart people together in one space I usually walk away with
some new ways of thinking about the software that change how I do my job
(Robert McDonald, I LOVE that animator tip about manipulating the bounding
box to promote clean rotations around products until cameras are
available!)... I can't wait to see some of the sessions on multi-bodies and
"low-carb part modeling" on the CD to catch up on what I missed, and see if
Phil Sluder had any new stuff to get me through day-to-day stuff faster

One technique that has captured my imagination is from Mark Biasottis
Advanced Modeling 101 - performing a deform using another body as a deform
tool.
What has got me thinking is NOT using it as a modeling tool (I do not care
so much for investing a lot of time in developing deform-tool-geometry that
MAY be allowed into the model IF a fickle feature decides to work and
deliver EXACTLY what I desire.. if I did this and it was close but not
exact, the CAD and its limitations could drive the final product, which is a
poisonous notion)

What has wormed into my brain is the concept of using the deform-with-body
tool as a geometry correction and optimization tool. We all do these lofts
with various success adding curvature continuity and tangency - or if they
are successful, they might not be all that smooth, or have ripples. We also
know that every additional face in a model increases the risk of rebuild
errors and latter feature failure.

I love the notion of taking surface-push via selected body and using that
like vacuum forming a final C2 single-face-skin on top of a close-to-precise
mandrill that represents the best currently available model. Where dozens
of patches might be necessary to get the geometry nailed using current
modeling techniques, we would instead swap in one face, and it would be C2.
The accuracy slider would represent the heat of the plastic - the hotter it
gets, the tighter it fits, which means you can get more control from your
tool body but also more flaws would show through (so your tool better be
REALLY good).

On the plane back I started to run through some experiments, and found some
good news and some bad news.

The good news is that the success of the feature depends on the same UV
direction stuff I have been lecturing about for the last two years - if the
UV lines do not 'flow' in the direction that you would sand or sculpt the
product, you will not get the shape you really want. So, hooray, I already
know what is required of the surface that I apply a tool body to in order to
get what I want.
The down side is that the UV flow DOES matter - I cannot just create any old
surface and hope it works.... it has to have the correct UV direction (you
can use face-curves to preview the UV direction if you want to see it)
The deform tool *appears* to work by leaving the UV lines *as-is* and only
changes them in the direction of the 'push direction' - for instance, if you
use the 'TOP' plane as a reference, only the Y value of the UV lines is
affected by the deform tool. The accuracy, I think, has something to do
with the number of UV lines calculated for the surface, though my evidence
is spotty.

The real problem that I see is that the tool body does not seem to effect
the very edges of a provided face in my initial tests, so if you want
tangency to something nearby, it better already be in the target face. The
other significant disappointment is that the tool body does not effect the
unseen geometry outside of the boundary of a given face - if you untrim
after deform, you will see that the original geometry remains intact, and
that the surface goes through crazy s*** to adapt to the tool body but stay
in contact to the unseen&unaffected parts of the surface.

In the short term, I was wondering if anyone was interested in sharing
experiments via this group to see how we can push this tool to clean up
extra faces and create C2 surfaces from non-c2 targets. In the long term,
if we find a workaround that is reliable, I think we might get SWx or a
third party person to automate the steps that we come up with and provide
that magic tool we talked about on the newsgroup a few years back - select a
set of faces, punch a button, and have it get replaced by a single face that
matches the original (within a tolerance) that is smoother and C2!!!

-Ed

 

Sign me up........I have been looking for a way to get segmented
surfaces to become one and maintain and or achieve C2.

 

Looks like its you and me, Arthur. My wife is out of town for rehearsals on
Sunday and Monday, so I might have the opportunity to run some experiments.
I'll start a new thread if I make any progress/ learn anything new. I would
also love to ehar your thoughts/discoveries

 

for your interest I had the sketches appearing at 90 deg problem today Ed.
have you reported it? also I had more problems with coincident relations and
also could not pick the origin.... another problem I had was a mirror
features made a previous loft they intersected invisible....oh well
cheers
neil

 

Ed,
I was also intrigued with Marks use of deform with body.
Rhino has a very similar feature called "Drape" and I've also likened it to
vacuum forming.
I just tried a couple of deform with body in SW 2005, and it was
disappointing with anything but the most rudimentary of shapes.
Please keep me informed if you make any headway.
If you have anything specific, let me know and I'll try to through some time
at it.

Muggs

 

Aint that the truth... I ran some experiemtns last night and was REALLY
dissapointed. I will have to see if I can figure out how it works, and
perhaps what it is supposed to do. I wonder if it was designed with
specific problems in mind, or if it was added as a me-to type feature and
they just didn't scrutinize the resutls. It is pretty awful, in my limited
experience

That said, I refuse to give up. I like the vacuum forming analogy, and
really want it to work that way. I wonder if that anaology can be pushed to
provide a way to squeeze what I want out of this tool. My next round of
experiments will explore whether, like a sheet of plastic, it is failing to
conform to my plug becuase I am pulling it too far in one shot. I can see
using multiple drapes (I like that term and will be using it for now) ont he
smae surface to progressively get it closer to the geometry I want, like
using a rpe-draw box, then vacuum forming, then a pressure assist if
necessary. (Or like a progressive die for those who don't do vacuum
forming)

The other thing I want to see is if the face or edge ID's change on a deform
face. I am absolutely enraprtured by the notion of draping a single face
(or a few simplified ones) over a model before production details are added
(ribs, lips, etc). This way I could change the plug to anythig gat any time
and not have any rebuild difficulties if I need to swap out features, faces,
etc on the plug because I replaced all the small ones with a nice, simple
large one of constant ID.

 

I'm interested, but have zero time right now for experiments. (I really
shouldn't be reading the newsgroup, but I'm addicted and can't help myself.)

Jerry Steiger
Tripod Data Systems
"take the garbage out, dear"

 

Very dissapointing results last night...

On the plus side, multiple applciations of the deform tool do draw the
target surface closer to the tool body within the basic limitations that I
have found.

Limitation-1? It works great on surfaces perpendiculer to the direction of
the push, but the more parallel the face of the tool body gets, the worse
the results.

Another limitation - it conforms much better to the furthest extents of the
tool body in the direction of the punch (the leading edge) and the trailing
edge is weakly followed. I did get some decent results by mimicking a
'pressure assist' - using the 'toolbody' to make a cavity in a second body
and using that in a second series of deforms on the same face. But the
absolute lack of conformity to walls that get increasignly parallel to the
push direction just sort of kills the use of the tool.

To isolate the problem (I had been using a more complex tool body) I ran
some quick experiments last night with a cone as a tool body, and a planar
surface as the face to deform. You can see that, as the walls of the cone
get perpendicular to the push direction, the fidelity is quite good. At
45°. it is just sort of bad. As it grows parallel, it gets really off. The
really odd thing is that it ALWAYS lacks fidelity to the tool body - the
target surface enar the leading edge of the tool body is always in
intersection, the trailing edge is always too proud, and it only matches the
tool body in one ring about 1/2 way down the cone. Extra applications of
deform will get it to match the leading edge better, but the trailing edge
remains too far off.

Bummer.

 

Yeah Ed, I also did some testing this morning with some poor results.
However I did find some interesting data. Have a look at the following link
to a pgn file.
http://home.comcast.net/~muggs828/solidworks/Deform-Test.png

I have a simple 'Body" to deform with, and four different deform surfaces:
1. 6" square - Orange
2. 6" circle - Yellow
3. 6" triangle - Green
4. 18" circle - Purple

Conclusion:
--Edge proximity definitely affects results - See the triangle
--More material (vac form analogy) doesn't allow for better "fit" actually
the closest that SW would allow me for the 18" circle (purple) is .072", all
the others were at .025".
--Square and circle are virtually identical.

BTW, for those not familiar with vacuum forming, the reason that I tried an
18" "sheet" was, again thinking about vacuum forming, using a bigger sheet
of material allows for more drape and therefore more material to pull
tighter into corners.

Take all this for whatever it's worth. If you have anything specific you
want me to try, give a shout.

Muggs

 

I noticed in the pop-up message that the 'fit' number is relative to body
size, so your results with the 18" circle make sense.
Extension of the tool body AWAY from the push direction also effects
reuslts, which doesn't make sense to me.
One other tip when doing tests... the shape is the same regardless of the
acuracy slider at the bottom - the only difference is that, with low
accuracy, SWx delivers some minor striations int he surface. I would want
to go to production with hight accuracy, but when doing tests low accuracy
provides a good preview.