Is SolidWorks really a solid modeler?

I have been teaching Computer Graphics this semester. I am using
Bertollini's book which might be a very handy reference to have around
if you can afford it. He has a listing of all possible graphics types
of which a very small group is called solids. Aside from prismatic
solids the other major group consists of things like tetrahedra,
dodecahedra (based on the pentagon), icosahedra (based on the
triangle), etc. I have never found it easy to model these solids in
SolidWorks or any other solid modeler because they tend to work off of
sketches.

Perhaps we should call the current crop of software Prismatic Solid
Modelers or Bounded Volume Modelers instead.

 

How about "Sheet-Knit one, Pearl two" software. Heh, just
kidding!

Mike

 

Is SW a solid modeler? If it isn't, then nothing is. If the book is
about computer graphics, then the focus would likely be more on
rendering, not on modeling.

This goes to the heart of the question: "What is a solid?". One math
professor told me the difference between a solid and a closed surface
boundary is that the solid has normals defined for all of its faces
that are all in the same direction relative to the enclosed volume.

In previous jobs, I did plenty of CAD modeling that was not "feature
based". It was all knitting and trimming of sculpted surfaces into
solids. Sometimes the surfaces were parametrically controlled,
sometimes not.

 

... heart of the question: "What is a solid?". ...

Volume definition with a density attribute gets my vote. It is sort of
interesting, void lumps and all.

Bet if you look around a bit you can find someone that thinks there's
something inside the "solid". (Wul, 'course there is. Just bore a hole
thru it. See?)

8~)

 

Before I started teaching the class I was pretty much of your opinion.
Funny what a change a little reading can do. I am speaking from
Bertollini's classification system. Different people classify the kind
of stuff we construct in different ways. I have posted the chart of
which I spoke at:

http://www.engtran.com/Bertollini-6-15.gif

What I am saying is that SW is limited to a certain class of solids
called prisms and knit surfaces. Regular solids and pyramids take
special effort to construct and some types of regular solid are
extremely difficult to construct even though they have all flat faces.
The kinds of solids of which I speak are the ones the Greeks wrestled
with thousands of years ago. See:

http://mathworld.wolfram.com/PlatonicSolid.html
http://mathworld.wolfram.com/ArchimedeanSolid.html

Try to make some of these things and see what happens. I have
frequently run into dead ends.

Even some of the warped surfaces in this classification are a bit of a
trick.

Patran 2.5 was a real solid modeler in that it not only defined the
outer surface of a volume, but every point inside. Patran 2.5 was
designed for creating finite element models where it was necessary to
know something about the inside.

Bertollini, BTW is a book on drafting, but he calls it by the fancy
name "Technical Graphic Communication" in order to encompass the
broader range of uses that companies with products like SolidWorks
engender.

 

If you look at an Iges file, there is absolutely no difference between an
enclosed surface and a solid. The only difference is how the modeling
software handles the data when adding new features. You could make an
enclosed surface in "surface only" software, export to Iges, pull it up in
SolidWorks, and it would be a solid .

Surfaces have normals just like solids. On solids, the normals are all
pointing the same way. The same condition can exist on an enclosed
surface. The only difference is how the modeling software handles it.

Extrudes and cuts are just trimming and knitting the new geometry into the
existing body. Even fillets are just creating a new face, trimming the
surrounding faces to make it fit, and then knitting it all back together.

The way I think of it is that surfaces and solids are the same thing, but
solids follow a set of rules which make them act more "intelligently".
Sometimes you can use the intelligence to your benefit, and sometimes it's
too much overhead. That's why solid modeling usually saves time as
compared to surface modeling, but when you need extra control or the
automated "intelligence" gets in the way, then surface modeling offers the
face-by-face/trim/knit functions you may need.

matt

 

created a dodecahedron using surfaces planes and mirrors and one knit. 20
min

Corey

 

Interesting thread and discussion. I like to think of solid model's
definition as: Solid models are a means to an end to actualize a
physical model, therefore it (solids) needs to understand volume.
There is another world that runs parallel to our mechanical
engineering one and that is the graphic post-production industry. They
is not concerned about this goal (physical model) but rather their
means to an end is a graphic representation only. Because this is
their goal, they but aside NURBS, CSG (constructive Solid geometry)
techniques in favor of Polygon mesh representation which is faster and
more economic (need not be burden with addition information like
manifold and non-manifold volumetric data). We also have a form of
this polygonal mesh in order to render OpenGL and PhotoWorks data
called a tessellated or faceted model, but that is hidden from the SWX
user.
I wish the there was not a divide between these two worlds because
both have something valuable to offer the other. We need better
rendering and animation techniques that MAX, Maya and Lightwave have
in our Solids world and they could use the ease-of-use and
straight-forward metaphor of Solids modeling to build in their world.

 

Do you mind posting that dodecahedron.

I tried (again) a tetrahedron in the 3D sketcher. By definition all the
points of a regular polygon lie on a sphere and all the edges are equal. So
I should be able to create a series of equilateral triangles assembled into
a bigger equilateral triangle and then put coincident relation between the
vertices and a sphere. First of all, I can't put equal relations on 3D
sketch lines and second when I try attaching vertices SW will choke.

I attempted a dodecahedron by creating a pentagonal surface and then mating
them edge to edge. After about 5 or six pentagons had been added to the
assembly I started getting mate errors. I didn't use mirror to get my
figure. Not only did I start to get mate errors, but I started to get could
not save document errors (and I wasn't saving the document) when I was in
the mate manager.

One obvious problem with this method of modeling regular solids is that when
they are brought into CosmosWorks the unmated edges will not necessarily
stitch properly. There will be unmated edges because, by their nature, they
are overconstrained. (That is why self supporting, rigid structures can be
made from these shapes.

 

I think Bertollini's definition of a solid has to do with multi-dimensional
constructs that have specific mathematical properties in a sense similar to
those that triangles, rectangles, pentagons, etc. have in a plane. For
example, he classifies a hyperpolic parabaloid as a surface even though it
takes up 3D space, but a tetrahedron, right prism or right pyramid are
solids in that there are special relationships between faces, edges and
vertices.

Like you say, some people think there is something inside a solid made by a
3D modeler, but of course there isn't. The 2005 "Select Other" method makes
this obvious. 3D modelers just enclose a volume because that is relatively
easy to do.

My concern at the moment is modeling regular solids of the tetrahedron,
dodecahedron flavor as parametric parts. Right now it seems the 3D sketcher
is not up to it because of the limited set of relations available.
Assemblies run into other problems and limitations.

The use of the word solid as you have stated can get interesting when
density changes inside the volume as in a bi-metal strip or a composite.

 

I was able to make a tetrahedron in 18 minutes. One 3D sketch. No surfaces,
no knitting.

I made a dodecahedron by assembling some extruded pentagons. Then I made a
incontext part from that assembly. No surfaces, no knitting. I had to use
vertices zero distance mates and some of the pentagons show in the assembly
that they are not fully defined but in all reality they are and can't be
moved. They all have 3 vertices mates per part. When I measure the resultant
edge line lengths and angles from each face they are all equal to 8 decimal
places. I also ran the part in cosmos express just fine.

If you want to see the parts send me your e-mail address.

Dan at bovinich d o t n e t

 

This stuff is way over my head (I'm severely mathematically challenged and
not very swift at geometric conceptualization, either), so this is just for
fun, airing thoughts, I learn from voicing dumb suppositions and being set

Takes up 3D space (is non-planar, requires a "bounding box" to enclose it),
but it doesn't define a volume? (Or can it?)

Tets are mult-surface groups that do define a volume? (Tet meshes for FEA
just satisfy the need for an orderly internal node distribution, connect
the dots for stiffness matrix?)

I think it might just be that surface topology is all that's required to
satisfy the requirements, for most purposes, where a "solid" description is
required?

Interesting. What's the application (if you're free to say)? Do any of
the "Golden" concepts (rectangle, ratio, ... ) apply or help with the
layout? Is using "3D" tools the best way to go or would setting up planar
relationships be more appropriate (dunno, not familiar with SW and don't
really grasp the concepts)?

Yes, there are aspects that require assumptions, "fudge factors". Where
they won't work they must be treated as assemblies, specialized entities
(laminate elements), etc., but that all goes beyond the topology
considerations.

 

I created a surface and assembeled it as you did but I only assembled it to
figure out the angle between faces. Once I had this angle I started
mirroring over planes that were defined by selecting the surface and one of
it's edges and using the angle I found in the assembly. Once I had all the
sides put in I knit the volume and there you have it a solid.

 

Just because it is more difficult to create certain 3D (solid) shapes in
SWX doesn't mean it is not a solid modeler. As others have shown here,
it is possible to create these solid shapes in SWX; it just requires
some creative contruction methods.

SWX really is a 3D solid modeler, but its interface and modeling style
just doesn't lend itself to certain types of shapes as well as others.

 

I tried to assembly all the surfaces which would be a prerequisite to
making a parametric model. If you try this, you may find SW starts to
choke or get confused on mates. Ideally one could mate the vertices to
the surface of a sphere much like an circumscribed pentagon's vertices
mate to a circle. Then the circumscribing sphere diameter could be the
parameter (or inscribing).

It would be tedious to have to do all these intermediate steps
especially when triangles, pentagons and hexagons might be combined to
make the solid (of course then it wouldn't be a regular solid).

 

I agree Arlin. In the real world, where companies like SolidWorks have to
make money, why would they make it easy to model these polygons? Survey SW
users and ask them how often they have a need for modeling these shapes
(probably .0001% of users.) I never before made those shapes and in just a
little time I was able to do it (see my previous post.) I'll bet you if SW
had a need for making these weird solids that they would make a module (like
their sheetmetal or mold modules) that would make every one of them in a few
mouse clicks.

Dan

 

I miss these newsgroup modeling challenges.

I got the basic dodecahedron as a single part in 4 minutes, with four
features (two extrudes and a couple of delete faces). But, of course, it
wasn't initally precise because I didn't know the angle for the sides.

I eventally (20 minutes) worked out how to get a layout sketch to give me
the angles and depths - being lazy, I did it with a couple of lofts whose
sections and planes were driven by the layout sketch because I didn't want
to mess with equations, plus the same delete faces. It would only take a
minute or two now to get the layout to drive the simpler tree. But as it
stands, it is a beatifully parametric (and stable!) dodecahedron, and you
can define it however you want - by the length of an edge, the circle
inscribed or circumscribed around a face, or the sphere it must sit inside
or circumscribe.

Anyone beat four features plus an equation? ( I'm counting the equation
driven version that I didn't get around to finishing - two extrudes, two
delete faces, and one sketch that carries all the magic + a derived version
of that sketch)

-Ed

 

The reality is that SolidWorks is just a shortcut for making individual
faces then knitting them together into a solid. Up unitl know they have
focused on useful shapes (cylinders, cubes, etc) that pop up in real world
products. If they could recapture the investment, they could pretty easily
make code for more involved shapes (my proof of this point - remember how
they added n-sided polygons into the sketcher a while back? Ever use
display/delete relations to see that process they used to make those n-sided
polygons was essentially what regular folks used to do manually when they
needed to make n-sided ploygon sketches before the feature was added?)

Cubes are a pain in the ass to model if you have to model each face
individually, but cubes are useful so a coder at SolidWorks made a shortcut
for us. If fashion shifts and dodecahedrons, tetrahedrons, and other
dungeons and dragon dice shapes become essential to industry, Solidworks (or
a third party vendor) will automate that junk in a jiff. But as long as the
ancient greeks remain ancient, a lot of these 'poetic' shapes will stay from
being a button press away.

<end opinionated blather... gotta sign off to watch 'the daily show'>
-Ed

 

Dan,

Sometimes I get a bit philosophical about things which has its place. Asking
a question like why would SW make it easy to model regular solids is a very
good question. It doesn't really have to do with marketing, more with
capability. The capabilities needed in the software to make regular solids
may be more important than you think. Everyone who has made these shapes
here has not been able to make them parametric in the sense that polygons
are parametric in the sketcher. These kinds of abilities extend past
regular solids and into real world problems that are contstrained in 3D
space instead of 2D space. If you try to make a tetrahedron by making the
vertices coincident to a circumscribing sphere it can possibly be done, but
you cannot make the edges equal by relationship. If you try to make a
tetrahedron or dodecahedron by mating faces in an assembly you will get
into trouble with overconstrained mates (and in my case causing SW to
CTDT). Hmmm. And there is a difference between making something that looks
like a dodecahedron and something that is a dodecahedron (i.e., carries
with it the parametric/mathematical properties of a regular solid). So for
example, are you saying that only .0001% of users could use an equal
relation in the 3D sketcher or the ability to make features in sketches and
sketch plane definitions simultaneously related?

As far as reaping financial benefits from the software, I think the focus
should be on the end user, not the producer of the software. No two users
will use the same methods, produce the same parts or utlize the same set of
functionality. There are things in SW to this day that very few people
utilize and yet they are there. I know, because I have done SW training and
support for a long time. If one user can make a million bucks because of a
feature in SW, should they put it in or enable it? (And, no, I haven't
figured out a million dollar project for regular solids yet.)

 

Nice! I'm jealous... I didn't think about extruding one pentagon and
cutting from the opposite side, with 'flip side to cut' active. Very well
done!

-Ed