Curvy Stuff 301

www.dimontegroup.com under tutorials.
Please try to download outside of business hours (8-5 Central time, US)

The thrust of this one is less of a 'how-to' and more of a 'how can I police
myself and use CAD without having it influence my design'. I was nervous
about it, but I got the reviews back from the session evaluation forms and
it looks like the point got communicated and was appreciated.



Example:

I have interacted with Industrial Designers who would not allow their
'vision' or 'creativity' to be compromised by manufacturing constraints...
"I don't see why we can't get away with an undercut here"."well, I guess you
have to redesign the circuit board to run cooler'."I guess you'll just have
to pull zero draft there".and the perennial favorite: "how about a
collapsing core...." These are the guys who make it hard to work in my
field because, after dealing with memorable prima-donnas, folks in industry
assume that all Industrial Designers are like that.



What is really ironic is that, though these Industrial Designers will not
compromise their 'vision' because of manufacturing constraints, they will
change direction on their designs if they can't figure out how to model the
shape in CAD.

Why does this bulge out here? 'Oh, I just couldn't model the leaner shape
without problems so this was the best I could get"

Why does it get thin and elongated? -"oh that is just what happens when you
flex a part."



That is f***ed up.

They won't change for constraints of the real world, but they cave-in to
constraints of the virtual one. The more I think about it, the more I hate
them for it.

I see it all the time - CAD-centric design is compromising design - or more
accurately, CAD-'feature'-centric design. I try to show that
'product-centric' design, literally forgetting the features while working
through how to construct your product, actually makes the modeling go more
smoothly.



After all, product are what pays eveyone's salaries. A product does not
care how it was modeled, nor does it care about the features that were used
to model it.



And that is the core of this session - not allowing CAD to do any designing
for you. You are smart, CAD is stupid, so lets let the smart party get back
in control



Best regards,

Ed

 

Thanks, Ed. You're a prince!


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

 

Nice rant <G>.

 

Couldn't agree more!

 

Amen brotha!

 

I couldn't agree more. Being the guy near the end, I am forced on a
daily basis to take solid models from designers who are designing
machined parts they expect us to manufacture from sheet metal. They
absolutely refuse to learn sheet metal design because "It doesn't work
in sheet metal" or "it's easier to just let you figure it out". There
is a reason for that: It can't be made from sheet metal as designed and
the design needs to be modified.

Unfortunately, by the time I get the order, it has been designed,
checked, approved and fully documented. Frequently the changes that are
required for me to manufacture the part warrant a major revision. Aside
from reflecting badly on the designer, the documentation bottleneck
results in a delay that can turn a decent lead time into a burning
crisis by the time all issues are resolved.

There are a few tricks to sheet metal, but they are really quite
simple. Too bad they can't be troubled to learn a process which would
result in better design, fewer problems requiring revisions and quicker
turns.


Dave

 

The direct answer would be: Use the sheet metal tools in SolidWorks.
Using these tools will insure that you are designing a part that can be
flattened. Aside from that, the only real issue is in using a realistic
inside bend radius when designing the part and leaving that radius
loose enough in tolerance (+/- .030") that the manufacturer can tweak
it to fit what their tooling will actually generate.

The trick of course is in learning the tools. This group is a good
resource and there are users of this group who have some great web
sites which give good info on sheet metal design using the tools
available.

I would say the best way to learn is to see what others have done, use
the tools, seek answers when features fail and ask questions if you
can't find the answers. Chances are better than good that someone in
this group has been down the same road you're traveling and can make
your life easier.

Dave

http://www.speff.com

 

Ed,

Your presentations are always intriguing and provocative, and very
welcome, thanks!

We have been thinking a lot about the role Solidworks should take in
our new product development. We are fortunate that our 'stylists' are
also skilled at surfacing in Alias, so many of the ugly feature-based
approaches you talk about are by-passed.

However, using one package or the other, we have three interesting
problems:

1. Using a surfacing approach like the one shown for the 'check-out
scanner' ultimately results in largely inflexible model. Granted, you
never have to turn 'a Volkswagen into a toothbrush', but you often have
to accommodate a larger LCD or new component. Looking at the
complexity of your check-out scanner, the re-work would be similar
using a straight surface modeller or SolidWorks.

2. We used to believe that generating native SolidWorks surfaces would
make it easier for SolidWorks to shell the resulting solid body. By
analysing our Solidworks surfaces back in Alias we can see that many of
the surfaces Solidworks creates (using 'fill surface' etc) are
unnecessarily heavy, compared to the equivalent surfaces you would
create in a modeller like Alias - so the surfaces are worse quality,
not better.

3. None of us want to start a production moulding (with all its
associated ribs, bosses, partlines etc) with a feature tree full of
surfaces. This is because rebuild takes longer (Solidworks insists on
rebuilding everything occasionally), and may be less stable. The
work-around is to develop the volume, save-out and re-import as a
parasolid, then get to work.

It follows that we should use a dedicated surface modeller up front,
with all its tools, and relegate Solidworks to the engineering detail.

Love to know your thoughts (and anyone else who's interested).

Regards,

Anthony

 

THANK YOU ED .. a *

 

Sorry it took so long to follow up on your post - I think you brought out
some really terrific observations/issues, and I wish I had more time to
continue what I bet would be an interesting dialog.
My responses are below.

Not completely true. I am often surprised with how an ungainly looking
model updates like a dream (I had a 1000 feature model that reacted to
changes really well). But that is why I spent the time on the rebuild error
session - once you learn how to get through errors and issues quickly, you
can find that models aren't as inflexible as they might, at first, seem.

But don't get me wrong - I hate having all of those extra patches, and that
is one of the reasons I haven't shared the model yet. On a model like this,
the patches end up being a way to carve out and understand the shape - and
understanding the shape is one of the largest components of trying to do a
model like this. I like to go through a clean up phase once I understand
the shape to see how much I can get out of there - the extra hour or two
pays itself back over the life of the job. I jsut haven;t had a chance to
go through the cleanup phase because of some recent complexity that has
intruded on my life

Starting out with a clay model as a reference helps sidestep over many of
the needs for flexibility - you have most of the changes already worked out.
Stuff like LCD, button placement, etc are largely predictable changes and
the model can be designed with those in mind (I'm working on something
similar right now, and as part of my 'storyboard' I am analyzing what
changes are reasonably likely so those areas are given more attention and
are more stable)

So we leave flexibility where it is most likely, in the 'target' regions.

The 'transition's have very few dims - maybe a tangency magnitude, a plan,
or a curve, but that's pretty much it, so they tend to take care of
themselves when there are model changes. I think that is why I have
eveolved this taret/trnsition methodology , becasue the models do become
more flexible that if you try to make a single loft that does it all

The big problem that remains with multiple patch models is adding and
editing engineering detail - ribs, shells, junk like that, and mostly
because of issues relating to face/edge IDS. I try to get that exterior as
perfect as possible before diving inside because there is no point in ASKING
for problems.

Since the interior is such a big deal, and since it is largely ignored by
tutorials - even mine - I want to focus next years session on it (I'm
looking for a title right now - I was thinking 'Son of Curvy Stuff', but it
implies that youalready know what these other Curvy Stuff sessions were. I
welcome other suggestions)

It is good to know that the rework would be similar, except the parametrics
are really handy (I don't know if Alias is parametric, but I know Rhino is
not). I would hate to have to remodel things just to drag a side out a
little - I love editing a curve and two seconds later my model is done. I
did a big revision of a model yesterday and it took very little time - sure
there were errors, but they were quick to root out using side-by-side
editing - (see trees of blood). It is so nice to have a good model that
can be worked with cleanly with very few issues

I have no comment on that - I do not know 'heavy surfaces' or 'light
surfaces'. Not to be glib, but I really only know if it looks right for my
product or not - I don't care so much about anything else. Maybe I should
learn more about that because I might be missing out on something?

I get torn on this. The best reason to try to avoid surfaces is because
lots of folks that you might hand the model off to just haven't been exposed
to them and don't know how to work with them. Technically, every model has
a tree full of surfaces, because all solid features are surface features. I
do revert to solids as much as is practical because I get a lot of trimming
and knitting for free, and there does seem to be a point on a surface model
where the re-building just goes through the roof for no good reason.
However, I use solids and surfaces interchangeably because I am not thinking
about the features, I am thinking about the faces that the product needs and
don't want to add faces that don't contribute a lot to the final product.

Please, please tell me that you don't really do this! I tried this on one
job, but never again. Every time I had to modify the original swoopy thing
and reimport the parasolid, every additional piece of work required 30
minutes of re-attaching dangling dims, etc
If you can't handle the rebuild time, and the model has distinct phases, you
should be using a master model approach instead - it handles the face IDS
really well. Build your part (for the outside of the scanner, for instance)
then insert that part into a new part. Behind the scenes this is IDENTICAL
to a parasolid import, except that the parametric link is retained along
with all the face and edge IDS. To get the efficiency, just set yourself up
so you do not load externally referenced parts - you will not rebuild the
original because its not even in memory, and you can work on the itnerior
without the burden of the eralier feaure tree. And if cahnges are made to
the original, you only need one reubild on the second part and youa re back
in business (hoepfully with very few errors). I know that I will cover this
in 'Son of Curvy Stuff' or whatever the hell name it ends up with.
-Ed

 

Ed,

Thanks for your reply, I have to admit that we have done the
'parasolid out - import in' approach, but only after the skin is
essentially settled. Thanks for all your other tips, they are
invaluable.

I can really only address one of the points immediately: heavy surfaces
- will try to post on the others later.

I understand that Alias and Solidworks produce splines differently. In
Alias the most basic curve is made up of an edit point at either end,
and two control vertices. The control vertices lie off the curve. You
can pull or push them to control the curve's shape. This is called a
3 degree curve (imagine one line from edit point 1 to CV1, one from CV1
to CV2, and one from CV2 to edit point 2). If you extrude this curve
you get a beautiful, simple surface.

It is possible for a curve to have more than two edit points. For
example, if you extrude a curve with 3 edit points, you will get a line
down the middle of the surface called an isoparm.

When patching between edges of surfaces where the edit points of each
surface don't match, you get multiple isoparms in the patched
surface. In some cases, a patched surface can be incredibly dense with
isoparms.

What's the problem with that? Sometimes very dense isoparms can
result in surface rippling. What's more, offsetting (or shelling)
dense surfaces can result in the offset surface 'folding over' on
itself.

Poor automotive modellers must model an entire vehicle with A1
surfaces, ie no isoparms at all. Of course, they have expensive
software to help them.

I hope to post an image from Alias to show the isoparms on a patched
model - anyone able to host it?

Anthony