knowing absolute accuracy value to use

[2ms1]

I'm wrestling with a model that brings my workstation to
a crawl with regerations taking sometimes as long as 10
minutes or so (like 100x slower than I'm used to).

At the same time, I'm having the issue of needing to
change accuracy values in order to carry out a part merge
that currently fails because the parts are apparently too
different in accuracy (current both on the default
relative setting .0012 but one part is larger than the
other)

So I think I probably need to switch to absolute accuracy
and give both parts the same accuracy value. The problem
is I don't know what value to use. I want to use roughly
the same accuracy as is currently in the main part, but I
don't know what value of absolute accuracy correlates to
my current relative accuracy value.

How do you know what absolute accuracy value is
equivalent to your current relative accuracy value?

I need to have good accuracy (especially since a ton of
rounds and stuff seem to fail whenever I change accuracy
value). However, at the same time, I can't afford to
make my regen times unnecessarily longer than they
already are.

 

[megadeth0]

In my experience, the two parts need to be relatively close but not equal. I almost always use absolute as I have had better luck with regens. The part that is being merged to, needs to be slightly more accurate than the merged part. Most of the time I am using very low accuracy, usually .0004. From the sounds of things you will not want to go that low due to the size of your model. So I don't know what range to tell you to try. But keeping the merged to part higher has worked very well for me. Typically I will use .0005, and .0004. For me, I found I needed the higher accuracy to get any sort of consistency. Good luck.

 

[dgs]

The real trick is to start with absolute accuracy. I haven't used anything but absolute, set at .0001" (teh default absolute value), for over a decade. I have had consistent problems wiht models with relative accuracy, mostly with features failing suddenly because the model size changed.

Of course starting with absolute doesn't help you now. I don't know of any way to tell what your part's accuracy is now, but I'm not sure that would help. See, with relative accuracy, the actual accuracy value changes as the part regens. So the value that works now may not work earlier in the tree.

I think the only thing to do is to pick a value and try it. if the model fails, bump ti higher or lower and try again. it may take many tries to convert from relative to absolute on a large model, especially if the model changed size dramatically from start to end.

BTW, I had a chuckle at your complaint of a 10 minute regen, which is a long time now. Back when I started on rev 16 on a Sun Spark 10 (I think), I had a single part of my first assy I worked on that took 20 minutes just to open it. I think regens were 45 mins or so. Pro|E gave you a nice analog clock back then so you knew how long you had been waiting. Actually, I think if you set the config option "clock yes", it's still there.

 

[2ms1]

Switching to absolute accuracy turned out not work. The
finest accuracy Pro/E would let me set the main (merged-
into) model was .0054. This is obviously way less
accurate than .0004 and I'm guessing maybe is less
accurate than the original .0012 relative accuracy?

I still don't understand what relative accuracy is
exactly. But in any case, when I set it to the .0054 a
ton of features failed and had to be redone, and then
when I got to the point merging the other part in it just
failed again anyway.

Fortunately, I was able to get around the problem by
supressing every feature in the smaller part (like
rounds) that I could do after they were merged together
rather than doing them first and then merging.

I feel like I lucked out on this one though. I guess in
the future when I create a model that I think I'll be
merging other parts into, I should start off with some
absolute accuracy and just hope that it isn't too fine to
make things too slow down the line and not too poor to
keep feature creations from failing?
Edited by: 2ms1

 

[dgs]

I'd switch to absolute and not look back. I've not regretted it in 13+ years on Pro|E, but I have lamented that models I've had to work on were in relative.

Here's the difference. Accuracy defines the smallest allowable edge in your model. I think it may be more complex than that, but that's the practical concept. It defines how much math Pro|E will do to build a feature. Once it reaches the smallest edge and it hasn't solved the feature, it gives up. Tighter accuracy allows smaller edges and more math.

Relative accuracy defines a ratio between teh smallest and largest edge. So, when attempting a feature, Pro|E looks at the largest edge in the part, applies the ratio adn determines the smallest edge allowed. So, if your part is ery large, you cannot create fine detail. Folks creating long extrusions have run into issues punching holes in them because of this. Also, as you can see, if the part changes dramatically through the regen, the smallest edge will change as well.

Absolute accuracy is simply the precise smallest edge that Pro|E will allow. I use 0.0001" which seems to work well in general. That menas that no matter what, Pro|E will do the math until the edge size is smaller than 0.0001", then it'll give up.

The idea behind relative accuracy was that if you're making parts that are 6-8 feet accros, you don't want to create tiny edges as your manufacturing process likely can't support it. That's not really true today as very precise injection molded parts can be very large, for example.

 

[tobbo]

Relative accuracy defines a ratio between teh smallest and largest edge. So, when attempting a feature, Pro|E looks at the largest edge in the part, applies the ratio adn determines the smallest edge allowed. So, if your part is ery large, you cannot create fine detail. Folks creating long extrusions have run into issues punching holes in them because of this. Also, as you can see, if the part changes dramatically through the regen, the smallest edge will change as well.

About relative accuracy. Pro/E makes a "virtual box" around the model (a box that could contain the model) and calcualtes the distance from one corner to anohter of the box. ("longest edge of the box") Then it calculates the smallest vaule accepted, like thesmallest edge.


//Tobias

 

[krow72]

Instead of typing in a number, I use select model and choose the main model for the absolute value for all other models. This seems to work for me.


Krow72

 

[magneplanar]

lowest accuracy can be controlled by the option accuracy_lower_bound, you can set it much lower than the .0012 if you wish

if you want to find out the accuracy setting of an import geometry part that is already made, you can set the accuracy based on the file itself and it will tell you what the accuracy is. ( set the accuracy of the part and choose the imprt gemoetry file as the reference part )you can then go back and set the part accuracy before an import feature to match. then when you reactivate the import feature the part and the import will be of the same accuracy...

i think the trick is not to mix relative and absolute accuracy parts

cheers

M

 

[mgnt8]

Back when I started on rev 16 on a Sun Spark 10 (I think), I had a single part of my first assy I worked on that took 20 minutes just to open it. I think regens were 45 mins or so. Pro|E gave you a nice analog clock back then so you knew how long you had been waiting. Actually, I think if you set the config option "clock yes", it's still there.

We called that Pro/COFFEEBREAK. The clock is still around. Doing a rendering & staring at it right now.