Verification on rebuild

SW2005 Sp3.1

I just saww a post with reference to 'Verification on Rebuild'.
I never bothered with that setting and it has always been turned off.
I've been working on a small injection molded part this way. It rebuilds
fine.
So, I thought I'd turn on 'Verification on Rebuild', do a rebuild and
see what happens.

Boy , oh, boy...I get filleting errors that won't work no matter what I do.

My question is: If the model is fine without 'Verification on Rebuild'
turned on, and I get what I want from it, why should I turn it on?
Is it going to bite me to leave it turned off?


Aaron

 

Yes, it can and will bit you. Big time.
Remember, your part has geometry errors that just weren't caught. It's
like those characters in the monster movies that close their eyes when
the monster's coming - hello, the monster is STILL coming.

There are four things that can happen in your case:

1. A random rebuild later on will catch the errors and then you will
HAVE to deal with them (I've seen this happen a few times)
2. You may not be able to export the model. A good check is a round
trip translation - try to save as an IGES then bring it back in and see
if it will be a solid again. Often on a part with a hidden error you
can't even get through the first part (exporting will fail)
3. You will not be able to collaborate on this model with anyone who
does use verification on rebuild.
4. You may have trouble adding later features because the underlying
geometry is bad - and you won't know where the problem arises from to
fix it.

What is the performance hit of using verification on rebuild? I have a
part with lots of tough geometry - surfaces, lofts, knits, delete
faces, etc. With it off, the rebuild time is 53 seconds. With it on,
it rebuilds in 57 seconds. And those 3 seconds save me a world of
downstream pain.

I work with it on because I know the problems of working with it off.
Over one week three of the guys I work with called me over because they
were having weird problems with their model that they couldn't
explain. I turned on verification on rebuild, we got the errors to
tell us where the source of the problems were, fixed them and moved on.
We would not have been able to deliver the parts with it off, and all
three guys now keep it active because they know it saves time in the
long run.
Last anecdote: One of my customers was having a terrible time with a
model similar to the one I just described - everything he tried to add
to a model just failed, he had to compromise on the part, etc. I
decided to take a look at it and I saw a big error in the first major
feature of the model - of course I saw it because I had verification on
rebuild on. I looked at the feature properties for the model and saw
he had been working on it for THREE MONTHS - to save a few seconds on
rebuild time (maybe a couple of minutes a day) he spent three months on
a fundamentally flawed model. His only option now is to do major,
major repair and if it still is too f***ed up he may have to throw away
three months of work and start over

 

The whole point is that your model is not alright. With this setting
off SW does reduced checking. It let's things go that perhaps shouldn't
be. Go through Ed Eaton's "Trees of Blood" or "Curvy Stuff" series if
you are doing complex mold models and you will find a lot of tips on
how to avoid problems, detect problems and fix problems.

Verification slows SW down, but may speed up getting to the end because
you will make a more robust model with it on.

 

I think that everybody should work with verification on rebuild
checked, but
I am sure that the folks who don't have what they think are valid
excuses.
When I get their crappy models, I'll fix them somehow and charge for my
time.

JK

I'm not sure I would make such a sweeping statement. I totally agree
when making anything with swoopy stuff in it. This would include sheet
metal. But I live in the land of prismatics these days. There is
probably little reason to have the box checked for me and one good
reason to have it off, assembly rebuild times.

 

On Fri, 29 Jul 2005 14:07:26 -0700, "jk"

I think a few yearsback, having the verification on used to create
alot of crashes - if my memory serves me still.

I just forgot to turn it back on, and now I'll keep it on because as
people have said , better to have a small delay in rebuild time than
forever trying to fix a model that has a hidden flaw.

I solved the shell problem by off setting a surface and then cuting
away the solid rather than use the shell as it was retricting the wall
thickness that it could work with.

Thanks for everyones comments

Jonathan

 

Believe me, if it were only a question of micro-seconds, no one would
turn it off. If you're making complex parts which take more than a
minute to rebuild, verification can add a very significant chunk of
time. When everyone is hollering for more speed, wouldn't you take a
20-50% increase if you could get it?

Personally, I work with it off. It's silly to work with it on all the
time. If I make some geometry which is in any way questionable, I'll
turn it on, ctrl Q, Tools/Check, fix errors if any, turn it off and keep
working. Of course always check it before putting on the last batch of
cosmetic fillets. Sometimes I'll check it if something that should work
doesn't, or if something looks funny or a face disappears, etc.

I think this might be getting overstated. Even if you make complex
shapes, not every part made with verification off is going to get an
error when it's turned on. It is important to check every part, but not
all the time. After working with it a while, you develop a bit of an
intuition about things it's not going to like.

Matt

 

I suppose it is time for an integrity macro that turns this on, runs
CTRL-Q and then runs check. Afterwards off it goes.

As an aside I always wondered why CTRL-Q was something nobody talked
about or documented.

 

Yeah, maybe the macro should save first.

I don't know why SW is so reluctant to share information about more than
the basic functionality with users. Sometimes I forget that not
everyone knows ctrlQ.

Matt

 

Thanks for everyones input.

The problem area was on in inside cosmetic fillet. I have turned on
verification and removed the fillets (for now).
The parts isn't swoopy, but has some funky intersecting planes.
i hope I figure out how to put the fillets in before I'm finished.
I'm wondering , if the problems are only on finish fillets, that no
features will be built on, can't I get away with it (verification off).
I tried exporting as STEP and IGES and it came in fine with the fillets.

One more thing. It seems confusing to refer to Cntrl-Q as a rebuild,
when it is different from the green/red light icon rebuild (or Edit -->
Rebuild, or default Cntrl-B). For a long time I thought I was doing the
same as Cntrl-Q by hitting the green/red light icon. If you look at the
keyboard customization under 'Others' there it is 'Force Regen' -
Cntrl-Q. So it's a regen, not a rebuild.

Now this is even more confusing, because when you look up 'Regenerate'
under Help, it says 'see Rebuild'...ha! what a mess.

Aaron

 

If you dig around in the API docs you will find this for CTRL-Q alias a
forced rebuild:

Description

This method forces a rebuild of all of the features in the model
whether or not they need a rebuild or not.


On the other hand the Red/Green Rebuild icon does this when captured in
a macro:

Description

This method rebuilds only those features in the model that need a
rebuild.

Give this a try guys. I'm not sure how the forced rebuild will behave
with AddtoDB on so do this on a part already saved. I couldn't find
TOOLS/CHECK in the docs or in a recorded macro so am assuming it isn't
there. The AddtoDB call and the display call make the rebuild happen
without affecting the display which should speed it up if my guess is
right. The macro toggles verification on and off so that a normal
rebuild and other operations won't be slowed down.

'
******************************************************************************
' C:\DOCUME~1\KELLNERP\LOCALS~1\Temp\swx716\Macro1.swb - macro recorded
on 07/29/05 by kellnerp
'
******************************************************************************
Dim swApp As Object
Dim Part As Object
Dim boolstatus As Boolean
Dim longstatus As Long, longwarnings As Long
Dim FeatureData As Object
Dim Feature As Object
Dim Component As Object
Sub main()

Set swApp = Application.SldWorks
Set Part = swApp.ActiveDoc

Part.SetAddToDB True
Part.SetDisplayWhenAdded (False)

swApp.SetUserPreferenceToggle swPerformanceVerifyOnRebuild, True
retval = Part.ForceRebuild3(topOnly)
swApp.SetUserPreferenceToggle swPerformanceVerifyOnRebuild, False

Part.SetAddToDB False
Part.SetDisplayWhenAdded (True)

End Sub

 

Paul,
I was wondering if I could hit you up for a favor. If you think of it,
could you do a 'feature statistics' on some of these prismatic parts
sometime to see what the hit REALLY is on rebuild time? On curvy stuff
I used to think like matt that it was significant but in my latest
tests it has been relatively trivial (3-4 seconds per minute of rebuild
time)so the returns vs costs curves intersect in favor of 'verification
on rebuild'. I wonder if the performance gains are more or less
significant on prismatics?
I need to find out because I will be giving 'trees of blood' at user
conferences in Wisconsin and LA this fall (and likely SWX world next
January though I haven't heard anything back yet on the submission). I
would like to have good data to base my advice on because it is a
subject that needs more exposure - I have seen so much time wasted
because people didn't know about it or think that the cahnge in rebuild
time is significant, but on prismatics it might not be to the users
benefit to enable it.

 

I tried this on one model made mainly from surface lofts and fills, 163
features. 3 ctrlQs with verification on and off gave an average of :

Off: 142 seconds
On 180
27% increase

On another one that had some complex shapes, some surfacing, but mainly
ribs and fillets (574 features)

Off: 223
On: 389
55% increase


A third model which was exclusively solids, mainly extrudes and fillets
with 166 features went this way (on a slower computer):

Off: 93
On: 135
45% increase

Anyway, I wish this setting were not so costly, but then if it weren't
costly, there would be no need for the switch to exist. I definitely
run with it off, and toggle it on at critical points and the end.

Matt

 

On a prismatic part (409.878.sldprt) with 384 features TS Toolbox
gives:

26 s
23 s
23 s

With VOR set to on:

28 s
27 s
27 s

This part has no relations or dimensions in it. Nor does it have
anything approaching a problem or questionable geometry.

Adding a shell feature to the highly prismatic part*

29 s
29 s
28 s

with VOR enabled:

40 s
41 s
40 s

So it appears that the types of features in a part have a great
influence on the effect of VOR and not just the complexity of the
geometry. The thought occurred that I should test with features from
the three or four classes you came up with in one of your
presentations. I don't have that list handy right now.

Equally important in any discussion of rebuild performance is the
effect of errors. I haven't looked into that yet coupled with VOR.

*I am using a level 6 result from the STAR2.1 benchmark.

 

I was running the macro on a model of an electric motor. The creator of
the model used multibodies and didn't bother to combine at the end.
There were 43 solids in the part due in large part to patterning the
fins. What I discovered was that adding a combine more than doubled
rebuild times. So perhaps using multibodies in a part without merging
or combining is a "trick" that could be used to get rebuild times down
significantly.

 

it's also a trick that will make wireframe views and drawings a lot more
.... interesting.

 

OK, this looks to be one of those fascinating topics. matt's numbers
caused me to second guess what I had found, but when I checked, sure
enough, on pretty much all the 'complex' parts I pulled up from the
last couple of years of work there was only around a ten percent
difference, and on some there was less than 2! (sorry I don't have the
numbers handy - they are at work) The only exception was a vacuum
formed part that was all solid features with fillets and a shell - that
came close to 25%.

I would bet anything that Paul's theory about feature types is true,
though I wonder if that might be expanded to basic modeling strategy.
In my unscientific analysis of about 6 parts, it seems that the more
surface work you do to make your solid the less verification on rebuild
spanks you. That might be why I am seeing less of a hit than matt -
not that he doesn't use surfaces, but based on seeing some of his stuff
and how he dissected a modeling problem I think we *might* have a
fundementally different approach that could account for some of this.

Of course SWx version (I tested 2005 sp 3.1) and hardware (I have dual
CPUs) might have something to do with it too. I would have to control
for a lot of variables to tunnel through this one (for instance, in
every test I flipped to check email while the model rebuilt - could
that make a difference?)

I wish I was more at leisure to look into this - these differences that
he and I are seeing are the little kinks in the system that lead to big
truths when you dig a little.

Thanks guys... this is turning into a cool thread.
-Ed

 

Ed,

I think you're right about the solids/surfaces issue. My parts with a
bigger percentage of surfaces didn't take as big a hit. I tend to do
the big overall complex shape in surfaces, then make it solid and
shell/rib/fillet. I think shells and fillets in particular give
verification a harder time.

I might be able to do more investigation with a completely surfaced
part, and some more completely solid parts. My "prismatic" parts are
still fully drafted and radiused plastic parts, not sure how that
compares to kellnerp's prismatic parts (machined?).

Matt

 

FWIW I have updated my benchmarking site with two utilities:

Rebuild.swp which I posted as source a few days ago and TSToolbox which
some people have been looking for. Rebuild has a timer in it now which
reports rebuild time in a message box. TSToolbox might come in handy
checking the effect of older version information in new documents in
that it reports the version history of a document.

 

My understanding as to what Verification on rebuild actually does
(which tallies with why it works hardest on shells, internal fillets and
such):

When you do a forced rebuild WITHOUT Verification, every surface is checked
to ensure it does not interpenetrate with ADJACENT surfaces in the model

When you do a forced rebuild WITH Verification, every surface is checked to
ensure it does not interpenetrate with ALL surfaces in the model

It used to be easy to fool SldWks with it turned off: shell a rectangular
block, then put an exterior fillet around the top face, radius exceeding
wall thickness. In the old days, this would create a dodgy Escher figure
where the interior square corner protruded past the exterior fillet, but the
rebuild would only fail if VoR was turned on.

I think you have to try a little harder these days, but the same principles
seem to apply, which gives you a steer as to when it might be important to
turn it on.

Sorry if everyone already knew this, it's just no-one seems to have
mentioned it here.

Andrew Troup