proper taper measurement method

[zbailey]

This last week we have found a somewhat concerning bit of data and hope to see some clarity from this. On site SMEs are looking into to tomorrow hopefully but I figured I would ask here as well. We are attempting to measure a quite short taper on a part, no CAD I am able to share sorry. Here is a breakdown of the issues we are having.

Taper in question is approx. .27" long and 1.6250" at the smallest point with a 10* angle. Tolerance is +0/-0*8'48". This issue was noticed purely by chance as the original programmer was using 12hitsx3rows, and I had to write a quick program to measure an incomplete part that used 12hitsx6rows. His program showed the angle in excess of 10* and mine was less than 10*, with a deviation between the 2 of almost a full degree with the part having not moved at all between runs. Profile scan up the taper coincides with my data, but through a set of tests using all variations we noticed that 12x5 was the only one that showed in on the specific part tested, and that coincides with the profile scan data. I am inclined to believe the profile scan data but the data makes no sense and contradicts what the senior programmers advise for the 2014 program we are on presently.

This data makes no sense as the data put into a graph shows the deviations of 3, 4, 6, 7, and 8 rows is almost identical vs the 5 row and profile scans matching.

Any insight would be greatly appreciated as myself, and the senior programmer at this facility are at a complete loss. I am going to confirm this taper on a form tracer today and hopefully get some idea of the actual taper to compare it to.

[Douglas]

I assume you are probing it as a cone? Being so short I would be concerned that all probe points are on the cone not on the edge, that is the first thing I would check.
Is it an auto feature cone all from a single probe tip or are multiple tips involved? Could form deviation be causing the difference between the 2 results?

Edit, also how are you hitting a cone with 12 points on 6 rows? without each row having 3 points I would not trust it myself anyway

[zbailey]

auto feature cone, confirmed its adequately away from the edges. form error seems to be consistent between the methods (.0003 to .0006 consistently) tried using different size probes as well as the surface finish isn't super smooth, but didn't get much noticeable variance. I don't recall the exact finish spec but we cant get consistent data.

also its 12x6 as in 12 hits on 6 rows so 72 points

[UKCMM]

With the length of 0.27" and angle limit of 0.14666 degree then the difference in the cone diameter will be 0.00069" or 0.00138" depending if is the full or half angle.

With such a close tolerance I would be concerned with form limit and surface finish having a big influence on the results. Is your machine or probing system capable
of working to those limits.

I would recommend setting the part on a sine table and checking the angle that way.

[Paperback Rocker]

Confirm which result is correct using an alternate means of inspection.

[zbailey]

we have, most all exceed 10* on a form tracer measuring the angles off the machined top surface. Issue is we are being told the way that gives us the closest info is incorrect as the senior programmer says adaptive scans and filters are broken in 2014 and for us not to use them. Until 2019 gets validated on our CMM we have to run 2014 so no matter what we are stuck using it and have to make due. We heard back from on site Hexagon and other SMEs today and they say line scans up the taper is the best way, which coincides with the data off the form tracer. Now we just need to try to get management to approve it against the senior programmers recommendations.

[louisd]

If the feature has an inconsistent taper: as in it's barrel-shaped or hourglass shaped, your cone feature will still construct a bestfit theoretially perfect cone, and produce that angle as the angle of the cone's output.
Your part's specific form variation coincides with 5 cuts, as it is able to capture the angle with the most accurate bestfit using those cuts. it's coincidence.

If I were you, I would measure two lines parallel to the axis of the cone (180° apart), with a bunch of hits, then output the angle between the two lines.
I would also output straightness of each line, to affirm that you have form error to substantiate this coincidental correlation between your measured cut counts.
--Turn GRAPH on to map/plot out the error on the cone, this will reveal where the cone is deformed and introducing the angle bias to your different cuts of the feature.
hope this helps

[Douglas]

I believe cylindricity can be applied to a cone and return the form deviation, Rn in the dimension box is also an option for form deviation

[louisd]

I was referring more towards GD&T interpretation, regardless of what the demon can and can't do.

[zbailey]

taper is confirmed straight on form tracer. Myself and the senior programmer at this facility agree that the surface finish is playing a little into the issue we are having, as scans seem to give more consistent data as per hexagon SMEs recommendation to us. Now we just need to get management to approve this method against the companies senior programmer recommendation as they say scans and filters are broken and don't work correctly in 2014.

[Douglas]

my mistake louisd it is roundness that can be used for a cone just tried a quick one, why it does not quite agree with Rn I have asked before and forget why Capture.JPGCapture.JPG

[louisd]

Least squares is used on the location (RN) and legacy circularity.
Tchebychev algorithm (min/max) is used on FCF dims (Circularity/Cylindricity) in version 4.2 and higher, to comply with Y14.5 standard.

This is per "Information on FCF Dimension Calculations" in the help file.

[Douglas]

same thing by a different standard... I will still forget next time I'm sure

[zbailey]

Ok, so my first shift protégé heard back from SMEs and they recommended doing adaptive line scans up the length of the taper. This was one of the methods we tried but were hesitant to use is as per company senior programmer says the filters and adaptive scans are broken in 2014 and refuses to use them unless absolutely necessary. The data gathered using this method correlates to the data off the form tracer closer than any of the others, so we will likely try to pursue using this method.

I appreciate all the recommendations, hopefully this is the end of this issue now and we just wait for approval from management.