Auto-Cylinder vs Circles Constructing a Cylinder: For Perpendicularity

[cmmGrum]

Which is more accurate, the auto-cylinder to plane OR 3 auto-circles (constructing a best fit, least squared, cylinder) to plane.

Essentially they should have the same result, correct?

When I run it I have a 0.0009 difference between the results.

Which number should I trust?

(I am building off of an IGES file)

Looking at previous posts I made sure that I had it set to Perp to Datum

[Mike Ruff]

Are you taking the same exact hits for each method? i.e. Is your auto-cylinder 3 levels with the same number of hits with the same PA and Z as the 3 auto-circles? If your hits are not identical, then that could be why they are slightly off to each other. You also should check that they use the same best fit algorithm, as this can also tweak the features

[cmmGrum]

They are the same PA and Z. The auto cylinder is (Nominals) (Least_Sqr). The constructed cylinder is (Best fit) (Least_Sqr). I am fairly new to this.

I am trying to compare my CMM data to someone that is using Calypso AND an adaptive scan CMM. They are getting around 0.0005 and I keep getting around 0.0015 (while using the auto-cylinder). But when I use constructed cylinder I get closer to their measurement.

[louisd]

the constructed cylinder can be constructed incorrectly, by only using the centroids of each circular cut. In order to have them judged the same, you'd need to use Cir1.hit[1..numhits] in the constructed cylinder. this would then utilize the individual points of each circle cut to construct your cylinder, and then they should be a perfect match.

[AndersI]

Note that the length of the cylinders will be different, and has an effect on the perpendicularity result. An auto cylinder gets the length from the CAD Surface, while the constructed (or measured) cylinder gets its length from the hits (ie. shorter than the auto cylinder). You can test by adding "projected zone" - (P) <length> - to your two perp dimensions to verify that they give the same result when applied to the same length (assuming the hits are the same).

What version of PC-DMIS? I believe the length calculation for auto cylinder has changed during the life of PC-DMIS.

[cmmGrum]

CAD++ Version 2017 R1


I think the fixing (P) has solved or at least gotten me closer.

[cmmGrum]

SOLVED! I had to change the Projected Zone value to get a more accurate result!

[AndersI]

Just be careful - the definition of "Projected zone" is not exactly the same as "Calculation length" - it works for parallelity, angularity and perpendicularity to set the length with (P), as these properties are unaffected by the position. Don't try this trick on a POSITION, CONCENTRICITY or SYMMETRY unless the print actually has a (P)!

[cmmGrum]

Understood. The perp i am trying to measure is pins coming out of a part and the CAD has the entire pin, so the auto-cylinder was trying to put the data from less than half of the pin onto the entire pin. I have the P set to what is actually coming out of the part now.

[AndersI]

Ahh, that situation can be tricky to detect!

[JEFMAN]

For the constructed cylinder, I would use BFRE instead of BF.
For the circle, the points are compensated paralle to the workplane, for the auto cylinder perp to the axis.
The difference shouldn't be large, but should exist...

[acgarcia]

louisd , how do you use Cir1.hit[1..numhits]? I'm using basic circle scans to construct circles along a cylinder. Then I'm constructing a cylinder using the circle constructions I made to get an overall dimension. Would using Cir1.hit[1..numhits] make the constructed cylinder more accurate?

[louisd]

see jefman's explanation below. Basically I stand corrected, by the master himself

[JEFMAN]

louisd , how do you use Cir1.hit[1..numhits]? I'm using basic circle scans to construct circles along a cylinder. Then I'm constructing a cylinder using the circle constructions I made to get an overall dimension. Would using Cir1.hit[1..numhits] make the constructed cylinder more accurate?

In the case of basc scans, the constructed feature uses the hits directly.
For an autofeature using scanning options, the centroid can be used by default (in this case, you can create an axis)

[acgarcia]

Glad to know that i'm doing this right, what exactly is Cir1.hit[1..numhits]? some sort of array?

[louisd]

its a method to force pcdmis to include all individual points in whatever you are intending to use it within. In this example, if you created the circle feature using autofeatures, then call the circle to construct an axis line or a cylinder, it will only pull the autocircle's centroid point. you can use the cir1.hit[1..numhits] to get away from the centroid, and instead it would use all the point data that is being measured in that autofeature

[AndersI]

(Cylinder Construction)

Constructing a cylinder from circles seems to be a special case in PC-DMIS, as this will always use the hits from the circles, not the center points! Can easily be seen by constructing a cylinder from a circle pattern. A cylinder from the four center points would not have a form error of 14.2701 (and would not have 12 hits)...

CYL1       =FEAT/CYLINDER,CARTESIAN,OUT,LEAST_SQR,NO
            THEO/<123.857,50,0>,<0,0,-1>,86.921,0
            ACTL/<123.857,50,0>,<0,0,-1>,86.921,0
            CONSTR/CYLINDER,BF,CIR1,CIR2,CIR3,CIR4,,

            ASSIGN/V1=CYL1.NUMHITS

CylFromCir.PNG