Using Constraint for Perpendicularity

[[Li...]]

I good handful of our parts have Perpendicularity callouts.  Either an OD cylinder or ID cylinder to a face/plane.  

We mostly run to a .0005" tolerance.  What is happening when I use Constraint and use the OD or ID to that face or plane it needs to be held to?

I always see and reduction in the actual deviation.  So if it is slightly out, let's say .00055" when go to evaluation and use Constraint and plug in the features it will reduce by .0002-.0003" making that result in spec and look great.  I'm passing out of spec parts by doing this?  There are certain situation when I can manually verify the part on a surface plate and other equipment. This has been working for me.  Seeing manual check data and CMM, meaning they are pretty close to the same result. 

 

Just wondering what "Constraint" is doing. 

[[Ma...]]

Using constraints is forcing that feature to nominal values instead of actuals.

So depending on alignment and actuals of features you can gain better or worse results.

You should avoid this on simple checks like perpendicularity and paralelism.

[[Pa...]]

Imagine if you will:

If you give me and another programmer 3 points in space, we will both find exactly the same circle.  Same size, same location, perfect form. Give us both 4 points and there's some likelihood we'll come up with different circles of different sizes and containing different variations. Give us thousands of points and a few days to calculate, it's almost certain we won't describe the same feature.  It's the same with Calypso.  Some assumptions must be made.

Using a constraint (for this we'll say you've constrained "normal vector") tells Calypso "Regardless of what you'd typically calculate, trust me. This cylinder is perfectly oriented as it should be (you'll notice the actual A1 and A2 values now perfectly match the nominals in the Feature template).  Whatever error you find must be coming from different sources".

So when you constrain normal vector in a perpendicularity evaluation, you are not allowing Calypso to calculate the extent of the deviation from perfect orientation because you've told Calypso to assume the features have perfect orientation. Any deviation found is now being thrown in the buckets of size, form, and location, as you've assured Calypso the orientation is perfect by constraining normal vector.

There are good times to use this.  When evaluating a small portion of a cone and you want its half-angle relative to a coaxial cylinder turned in the same op, I would probably constrain normal vector.  You have good reason to believe that the cylinder and the cone really do share an axis, and you're not necessarily evaluating that axis.  For small portions of a circle (say a fillet rad in a corner), I might constrain two axes when evaluating for size, and I'd constrain for radius when evaluating location.

[[Ze...]]

Something else to consider in perpendicularity is the order of selection of features. You should select the longest, most stable feature as the Datum feature and the shorter feature as the considered feature for perpendicularity in most cases. Even if the drawing says otherwise, unless there is some sort of logical reason not to. The shorter the Datum feature, the greater the chance of projection error of the created tolerance zone.  If you've ever flipped the order of features and noticed a difference, it's commonly caused by this. For example... asking for perpendicular of a 3" diameter plane to a .5" long boss.  

I realize that not reporting what's exactly on the drawing can be problematic. I usually ask to have the print changed, if possible. If not, I explain myself and show both results to illustrate the issue. It's almost always a better result and people like parts that pass. 😀

[[DW...]]

Please sign in to view this quote.

Please sign in to view this username.

 This. Yes, you are most likely passing out of tolerance parts by constraining the orientation of the cylinder axis. In your example, and to help drive home the understanding of what is happening, constrain the plane too. Your perpendicularity result - 0.0000.