accuracy and merges...

[PAULKORENKIEWIC]

Guru's,

Occasionally when merging two parts Pro/E will give me an error telling
me to change one of the parts to some new accuracy value OR to use
absolute accuracy. I've somewhat blindly always just followed orders
and changed whatever it told me to. I am wondering if I would be better
off going to the absolute method instead? Would I simply change the two
parts to the same absolute accuracy value? And would it matter what
that value is? Any thoughts? Please respond to the list to maintain
the thread...
thanks in advance...

Paul Korenkiewicz
FEV , Inc.
4554 Glenmeade
Auburn Hills, MI, 48326-1766

[ptc-593619]

I have had a lot of problems solved by changing to absolute accuracy, so
much so that I am wondering why I don't make this the default. Why do we
have accuracy values?

[RodGiles]

Dear Paul,

I'd go along with what Richard said. I'd also point out that PTC
recommends that if you are using geometry from one part in another one,
then they should both have the same accuracy. The only way to ensure
this is to use absolute accuracy. My preferred method is to set absolute
accuracy in the start part. There is a good explanation of accuracy on
the PTC web-site - see

[dgschaefer]

I have used nothing but absolute accuracy (except where required to by
customer constraints) for over 10 years with great success. Accuracy is
a way of limiting how much math Pro|E will do to solve the geometry. It
does so by limiting the smallest edge size.

Using relative accuracy, you are setting a ratio between the longest and
shortest edges in a model. That's why you hear occasionally of folks
designing a long extrusion and then having trouble adding small holes to
it. The length of the edge of the extrusion limits the size of the hole
allowed. The theory there was if you have long edges, you have a large
part and you don't want tiny details on it because the scale of the part
makes tiny details hard to manufacture. However, now very large
injection molded parts are quite common and they can have lots of tiny
details.

Using absolute accuracy, you are setting the absolute shortest edge
allowed. No more large parts that can't have small details. It also
makes it easy to match the accuracy between models. Since relative
accuracy is a ratio, it isn't actually setting the accuracy but
providing a formula to do so. The actual accuracy value is calculated
by Pro|E (and recalculated as the part regens) and is unavailable to the
user. So two parts with the same relative accuracy can have very
different actual or effective accuracy values.

I've found that relative accuracy is limiting in other ways as well.
Because it's a ratio, you really need to establish the size of your
model early in the tree. If not, the drastic change in model size may
cause a drastic swing in accuracy. Suddenly, the rounds you want to add
don't work because the edge is too small. Also, if the design changes,
say two smaller parts are now to be merged into one to save assembly
labor. You pick the more complicated part and roll back the tree to add
in the geometry of the other part. Now that the model is much larger,
the calculated accuracy goes up. Suddenly, rounds that should be
unaffected, simply won't regenerate. The message is 'Failed to
intersect with part' or something equally vague. This is due to a
change in accuracy.

Lastly, the farther along a model is, the harder it will be to change
from relative to absolute accuracy. Because relative accuracy changes
as a model is regenerated, there may not be one absolute value that will
work at every point in the tree. It may be effectively 0.0001 at the
start, 0.001 in the middle and .0005 at the end. Finding an absolute
value that allows the model to regen can be a long trial and error
process and it may change the model as the new accuracy value forces
features to be calculated differently.

Bottom line, I have set my absolute accuracy to 0.0001 for inches and
0.00254 for metric years ago and haven't looked back.

Doug Schaefer

[HugoHermans]

Also related to accuracy settings and merges, when confronted with STEP
or IGES, alligning the accuracy of your model with the accuracy of the
source system greatly enhances the import process.

Regards, Hugo.

[BobFrindt]

For what it is worth, I have taken a different approach due to the
accuracy issues and it has worked very well for me. I prefer to use
published and copied geometry of the part surfaces and then solidify them
to create the merge. I have not found any drawbacks yet (I'd like to hear
them if anyone knows of any or I'm setting myself up for problems down the
road) and the accuracy issue never comes into play.

It is an alternative to changing the accuracy of a part for no other
reason other than to get it to merge.

Bob Frindt
Sr. Designer
Parker Hannifin Corporation
Parker Aerospace
Gas Turbine Fuel Systems Division
9200 Tyler Boulevard
Mentor, OH 44060 USA
direct (440) 954-8159
cell: (216) 990-8711
fax: (440) 954-8111
-
www.parker.com



"Korenkiewicz, Paul" <->
07/07/2009 07:19 AM
Please respond to
"Korenkiewicz, Paul" <->


To
-
cc

Subject
[proecad] - accuracy and merges...






Guru's,

Occasionally when merging two parts Pro/E will give me an error telling me
to change one of the parts to some new accuracy value OR to use absolute
accuracy. I've somewhat blindly always just followed orders and changed
whatever it told me to. I am wondering if I would be better off going to
the absolute method instead? Would I simply change the two parts to the
same absolute accuracy value? And would it matter what that value is? Any
thoughts? Please respond to the list to maintain the thread...
thanks in advance...

Paul Korenkiewicz
FEV , Inc.
4554 Glenmeade
Auburn Hills, MI, 48326-1766

[ptc-1932745]

Doing experiments with both copy/geometry, and merge/cutout some time
ago (WF2), I decided that copy geom. was the only way to go. Besides
the accuracy issues that are non-existent using copy geom., I ran into
problems updating dies using merge and cutout because the molded parts
used for the merge/cutout had to be in session for the cavity to update.




Otherwise, when doing a regen, or even rolling the die model back and
forward again produces no changes to the cavity.



Some more obscure config.pro options need to be set for the copy-geom
method to work as expected (automatic update, bring all referenced
models in session as needed, recursively regen parents of current model
as needed, etc...) but it is worth it.



As an example, if I have a model approx 2x2x4" that I want to put into a
die that is 8x8x8, using merge / cutout the proper accuracy may not be
possible (either because the die relative accuracy had to be less that
1e-6, ) or using absolute accuracy would cause the die model to fail.



Secondly, when trying to create an EDM electrode from the die model, the
accuracy is changed again (and for the worse)



Using the copy-geom method, If I open the EDM electrode model and hit
regen, all of the changes are updated. Using the copy geom. method, I
would have to make sure that all of the relevant models are 'in-session'
first. (Not a problem if you have PDM, but a royal pain otherwise)







Christopher Gosnell

TRIGON INC.
FPD Company
124 Hidden Valley Road
McMurray, PA 15317
PH: 724.941.5540
FX: 724.941.8322
www.fpdinc.com

[merrill.rosenow]

How much extra effort is required for Pro/E to go from relative to the finest resolution of absolute accuracy? I have used relative accuracy without problems except for two parts. I simply changed from relative to absolute and continued working. Now that I know what to look for in regards to accuracy problems, I won't get stumped anymore. (It was already mentioned before, but features failing to intersect or in my case, graphical issues and failures for no apparent reason). In most cases is it so terrible to leave the accuracy as relative and change it if required?

One odd thing I've noticed, is the smallest absolute accuracy of one part I am working on is not the same as another part.

[ptc-593619]

That was my question - if running at absolute accuracy solves a lot of
problems, why run at relative accuracy? What are the pros for relative
accuracy that you would want to leave it as such? Why not make your
start part at absolute accuracy, what do you lose?

[ptc-1932745]

Unless you are referencing features from one part to another, I don't
think that the 'accuracy' settings affect assemblies.



One thing to watch for is problems in Manufacturing, Pro-Man uses
absolute accuracy. We have had surface finish issues on highly sculpted
4-5 axis parts that 'seemed' to be related to the accuracy.



For the parts we do (~10x10x20" max) we use .0001 absolute accuracy in
our start parts / assemblies with good results.



Christopher Gosnell

TRIGON INC.
FPD Company
124 Hidden Valley Road
McMurray, PA 15317
PH: 724.941.5540
FX: 724.941.8322
www.fpdinc.com

[HugoHermans]

AFAIK, from the point of view of PTC, it's difficult to set up a generic
configuration if they don't know in advance the general magnitude of the
parts that will be designed. OOTB, ProE has to be capable to tackle
details below 1 mm to several meters. If the order of magnitude is
mostly the same, what's the case for most of the users, starting from a
template with a specific absolute accuracy is by far the best approach.
But if you can't decide on a common absolute accuracy, a relative
accuracy might be an option.

The disadvantage of an accuracy that's too small for most of the parts
(relative or absolute), is that these parts become too big on disk and
to heavy to process.

We design industrial machines with parts from a few grams upto 250 kg,
our template has an absolute accuracy of 0,01 mm. So, details below
0,01 can't be modelled.

Regards, Hugo.

[tyler.smith]

How is everyone else using accuracy with family table instances?

When trying to use absolute accuracy, we get a lot of "accuracy
failures" in our models (one accuracy value does not work for all our
instances). As far as I know, pro/e doesn't allow you to add accuracy
to a family table, so we are using a workaround: use relative accuracy
on our model, then draw a long curve in our model which makes the model
size bigger, which causes pro/e to refigure its accuracy for that
instance.

Does anyone have a more elegant solution? Our parts frequently have
edges that are on the order of .001" long and the length of our bounding
box diagonal is maybe around 20".

Thanks,
Tyler

[ptc-1932745]

It's always the exception that tests the rule...



For your precise reason, we only use absolute accuracy for our models.
If the smallest edge ever is approx .001" (inches or units(?)--- better
left to another discussion), we would set our absolute accuracy to .0001



You can go finer (tighter) with the absolute accuracy by setting the
absolute accuracy lower bound config.pro setting.



On other words, it seems to me that by changing a length of a reference
curve in your family table, you are causing Pro-E to change the
'relative' accuracy to match the smallest edge feature. (20/.001 = .002)
'absolutely'



Again, I would have set the global absolute accuracy to something like
.0001", or .00001", and went from there.



I would be very interested to see if your parts have any 'Geometry
Checks' before or after the accuracy modifications.





Christopher Gosnell

TRIGON INC.
FPD Company
124 Hidden Valley Road
McMurray, PA 15317
PH: 724.941.5540
FX: 724.941.8322
www.fpdinc.com