March 2006 Modeling Challenge

Well the models are now posted. I'll hold off comment till others post.


http://www.niswug.org and click on Design Contest.

Thanks to NISWUG for hosting the results.

TOP

 

Comments anyone?

 

Sure, I'll start.

I think that this challange has shown once again the amazing differences in
thought process.
One of my favorite things is to see how peeps go about making things in
SolidWorks (and how it differs from how I would do it), and this is a great
example.

Thanks to the four contestants and to you TOP as well; Very enjoyable.

Muggs

 

Ed converged on an answer similar to the one Dale and Heikki did and
then he completely diverged with his second solution.

I hope people also look at rebuild times as this is even more amazing
that the contestants approaches.

 

Fixed Link to Ed Eaton's model. This one is real interesting.

 

Only one comment so far but I'm sure many have had a look. I'll have to
pick a winner by the end of the week but I was hoping for some moral
support in the form of a discussion. So let's start off by discussing
performance. There is a big surprise.

The top performing models had times of .25, .28, .3 and .46 seconds on
my system. Obviously these times will be different on your systems.
Times were taken from TOOLS/FEATURE STATISTICS. TimeLord's model with
31 features was the fastest, Dale Dunn's "Truncated Icosahedron" with
15 features was second, Ed Eatons "buckeyball fastest2" with 5
features was thrid and Heikki's TruncatedIcosahedron2 was fourth with 9
features. The average rebuild time for all models was 1.01 seconds
with a standard deviation of 1.29. Maximum rebuild time was 3.66
seconds for Heikki's TruncatedIcosahedron4 with 5 features. If we
deduct the surface feature this model actually had the lowest feature
count to get the truncated icosahedron. So why was the performance so
much slower? Heikki's TruncatedIcosahedron2 used a very similar sketch
with very similar relations and had a very quick time. The sketch in 2
took .32 seconds and in 4 took .5 seconds. Dale's Layout sketch took
..09 seconds and Ed's sketches took far less time to solve being .06
and .01 seconds. For some reason the circular feature pattern in
Heikki's #4 took a lot more time than in his other models. Heikki's
sketches have more relations and include symmetry and sketch patterns.
Dale's and Ed's sketchs have fewer relations. These may be the culprits
in terms of performance.

The question remains as to why TimeLord's model was so fast even with
so many features. The answer seems to be that his sketches were very
simple and didn't even show up on the Feature Statistics radar. Most
features that had measureable times were in the .02 second realm and
the longest feature was .06 seconds. Clearly many simple sketches and
features can give a speed advantage.

 

Thanks for the breakdown TOP, it's hard to a draw definitive conclusion
from just the one challenge regarding performance. I eagerly await future
challenges and more importantly your breakdown of performance.

John Layne
www.solidengineering.co.nz

 

Just logged back on to your site to look at the images, most of the links
are now dead.

John Layne
www.solidengineering.co.nz

 

I'm not seeing any dead links. Clicking on the screen shot should cause
a zip file to download. The zip file contains the model for that screen
shot.

 

My apologies, Firefox was playing up for some reason -- cleared the cache
and all images returned.

John Layne
www.solidengineering.co.nz

 

Dale,

Run feature stats on each one and you will see what takes time. No need
for theories.

What is worth pursuing is eliminating sketch relations one at a time
till the sketch times come down.

I think I'll wait till after beta for the next challenge. No need to
compete with the mothership.

I'll probably have to have several awards for various aspects and the
original criteria will have to be kept just for consistency. But we'll
all have fun.

 

Dale,

And BTW, that extrude on Ed's buckey ball is to close a tiny hole that
needs must be there to make the sweep work. View the ball in hidden
line while rolled back to before the extrude.

 

Your turn. Why not analyze the differences in the layout sketches?

 

You guys are all winners! I finally got a chance to look at the models
during my lunch break. I am amazed.

It does seem that Time Lord and Dale Dunn in his first part violate the
Euclidian spirit of doing things with Dale's 10.812... degree taper and Time
Lord's 10.812... degree plane tilts and the .979... factor in the equation
for sketch 2. I don't suppose that was really part of the rules, but it
would be nice if everything you need to build the part would be included in
a sketch, as in the rest of the parts.

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

 

Jerry,

Well I am trying to stay within the Euclidean spirit of things, but I
have to hand it to TimeLord that his model was blazing fast and that
was probably one of the reasons.

The models really are amazing in breadth of approach even at the sketch
level. I don't think the number of ways this could be done has been
exhausted either.