understanding my results/ setting myself up for flaming?

[EmilySue]

Hi all,

I may be about to display a total lack of some fundamentals, so please don't be too harsh...

I have the CMM using autocylinder to measure an OD with a callout 0.636-0.637 diameter. The length of the cylinder I can hit with my probe is pretty short, but I put three levels on it, since it is a primary datum. There is no cylindricity callout, but my hits showed lobing, so I started looking at cylindricity, and PCDMIS reports 0.0008.

1) I know the CMM probe can cause lobing in measurements. I have been under the impression that PCDMIS uses calibration data to correct for this lobing when it does all its calculations, before reporting the measurement. Does it?

2) Basic metrology question: my tolerance is basically 0.001 total. With cylindricity 0.0008, that is using up most of the tolerance, and I think my result would have to be within 0.0001 of the mean for the part to still be good. Is this correct?

3) I have the default "least squares" method on. I did try the "minimum circumscribed" on a part to confirm that method gives me a result that is quite a bit bigger. How do y'all decide which to use? (Part of the reason I did this is to show the operators why the CMM is getting different numbers than they are getting with digital mics at the machine. No matter where you put the mic, you're at a high and a low point, so they were ALWAYS getting lower numbers than the least square result and wanted to discount the CMM results. I need to be smart enough to explain what's going on so we can make the best parts we can.)

I keep reading and learning, but often come up against questions that I don't have the knowledge to figure out yet. Thank you in advance for helping out!


JuneScreen.jpg

[NinjaBadger]

Hi all,

I may be about to display a total lack of some fundamentals, so please don't be too harsh...

I have the CMM using autocylinder to measure an OD with a callout 0.636-0.637 diameter. The length of the cylinder I can hit with my probe is pretty short, but I put three levels on it, since it is a primary datum. There is no cylindricity callout, but my hits showed lobing, so I started looking at cylindricity, and PCDMIS reports 0.0008.

1) I know the CMM probe can cause lobing in measurements. I have been under the impression that PCDMIS uses calibration data to correct for this lobing when it does all its calculations, before reporting the measurement. Does it?

2) Basic metrology question: my tolerance is basically 0.001 total. With cylindricity 0.0008, that is using up most of the tolerance, and I think my result would have to be within 0.0001 of the mean for the part to still be good. Is this correct?

3) I have the default "least squares" method on. I did try the "minimum circumscribed" on a part to confirm that method gives me a result that is quite a bit bigger. How do y'all decide which to use? (Part of the reason I did this is to show the operators why the CMM is getting different numbers than they are getting with digital mics at the machine. No matter where you put the mic, you're at a high and a low point, so they were ALWAYS getting lower numbers than the least square result and wanted to discount the CMM results. I need to be smart enough to explain what's going on so we can make the best parts we can.)

I keep reading and learning, but often come up against questions that I don't have the knowledge to figure out yet. Thank you in advance for helping out!


JuneScreen.jpg

The only stupid question is the one you're afraid to ask.


1) Short answer is No. The lobing you describe is only on TTP (touch Trigger Probes) like the TP20 - scanning probes work a different way. With a TTP the calibration does NOT take it out of the results.

Best thing to do is measure a ring gauge to prove the accuracy. You can also report the cylindricity so you know that with this probe you can expect this amount of roundness error. The longer the probe the more error.

I would get around 0.003mm error on a 20/30mm probe up to 0.010mm with a 70/80mm probe. Touch speed can also impact on this a bit (I use 2mm/sec).

2) It depends. I assume your in the US so probably working to ASME, where Rule 1 is that features must exhibit perfect form at MMC. In your case you can see that it's tri-lobed so your assumption about it having to be mid-limit to be good are probably correct. However see below...

3) You should really be using MinCirc for OD's and MaxInsc for ID's - as should I. However I don't, I'm regularly working to tolerances as tight as you have there (and tighter) - I should really be using a scanning probe but we don't have one. If I use the correct algorithms the tri-lobing of the probe (especially with the longer probes) can put diameters out which aren't. It's not a great situation (but we work largely with turned part which exhibit good circularity) so I mainly use least squares.

First things first measure a ring gauge (with the probe/angle in question) and report the diameter and circularity. This proved to the operators that give a feature with good form, the CMM is accurate (you can then also see how much of your 0.0008 comes from your probem and how much from your part.)

The graphic you have there is also a good way of showing what's going on.


Another thing with a tri-lobed part and a TTP is that the tri-lobing can either constructively or destructively interfere with each other, making the results (circularity) better or worse than is true.

[EmilySue]

So much good stuff here. I would like it 5 times if I could. Thank you!

[QS920]

Great info NB!!

[AndersI]

I don't know about ASME, but ISO explicitly says circular datums should be calculated as max inscribed/min circumscribed (and planar datums should use tangent plane). But for studying the actual form of a feature, the calculation method shouldn't matter, as the hits are the same...

[JEFMAN]

In addition of the very good post of NB,, you could try to measure the cylinder with a probe at 90° of which used here (for example, if you measured with A0B0, remeasure with A0B90, and try to see if the defect turns with the head or not...

I will try some maths on another number of posts

[NinjaBadger]

Excellent suggestion for post 4000! I was going to suggest rotating the part 90° but if it's part of a more complex program then it could be tricky! Rotating the probe 90° on it's own axis achieves the same thing!

[EmilySue]

I tried with a probe at right angles to the probe body, then did A90B0. LOL... Next week I will tackle all of this with a fresh brain and try this suggestion correctly!

[dmis100]

Is this part being ran in a 3 jaw chuck.

[JEFMAN]

So you can say how the part was fixed in the chuck*, and you can also say to the machinist that the clamping was too high
* : usually, the jaw were at the level of the max radius : the tool cutted the constrained material, and when the clamp is off, the part grows up...

[acgarcia]

I have the same problem. The usually respond with "we are working with the least amout of clamping pressure so the part wont fly out". Either way, the parts will come out out of round.

[EmilySue]

acgarcia, I would get the exact same response, & do get it every time I bring it up.

[NinjaBadger]

I don't know about ASME, but ISO explicitly says circular datums should be calculated as max inscribed/min circumscribed (and planar datums should use tangent plane). But for studying the actual form of a feature, the calculation method shouldn't matter, as the hits are the same...

I went back and forth with myself for a long time over this (we had diametrical tols of <0.02 and concentricities of [email protected] on the same features, which needed to be measured with a probe with a tri-lobing error of 0.01.

Doing it as per spec would mean passing poor parts and failing good ones. I know the parts are exceptionally round from checking on talyrond, so least squares is the best (most accurate, if not per spec) way.

[AndersI]

Yeah, the classical theory/practice dichotomy combined with "know what you're doing". I get your point!

[EmilySue]

I am SO glad I asked. Theses answers are exactly the kind of information I've been looking for! Best community ever!

[EmilySue]

OK, I tried building a probe with a tip 90 degrees to the probe body, then taking my ring gage. With the same length & ball size, same ring gage, I got 0.0008 concentricity and then 0.0007 concentricity with the probe rotated.

I'm back to my original probe build, 1.5x30 and using A0B0, and now I'm trying different numbers of hits and different speeds to see if I can improve on 0.0008.

[EmilySue]

Sorry, that would be helpful! It's a TP2-5way. No probe changer, no scanning.

[NinjaBadger]

I'm not sure if you've understood (or are replying with incorrect terminology) but Jeff suggested rotating the probe 90deg on it's own axis (not 90deg to the probe body).
I also assume you mean 0.0008 cylindricity (not concentricity)?

You then need to compare where the lobes are as measured with the two vertical probes. If the lobes remain in the same place it's the part. If the lobes move it's coming from the probe.

Regardless of that though, if you're getting 0.0007-0.0008 on a ring gauge it would suggest your original lobing is coming from the probe.

Different number of hits & speeds is all well and good but make sure you qualify at the same speed as you're testing with .

What's more likely is that I believe the force of the TP2 can be manually adjusted, if yours is set too high that would account for this high cylindricity value on a ring gauge. The TP20 (with removable modules) have different forces for different lebgth/weight probes, on the TP2 this should be adjusted to suit whatever build you have on there.

I'm not sure how you adjust it though, I just know I've heard it mentioned before (Matt is your man for all things TP2).

[EmilySue]

yes, cylindricity, not concentricity.

I did try rotating the whole probe body instead of just the tip... so I will go back and see what happens with my ring gage when I just rotate the tip.

It is fun to be learning, but a little exhausting when I think about how tight these tolerances really are and how much better I want our process to be.

:-P

[louisd]

Emily, do you have a XXX grade cylindrical ring gage or somewhat large (maybe a deltronic) or similar plug gage laying around, which you can put on the machine to measure? Something that is NIST traceable and has certified form of <.0002? deltronics and XXX gages would easily meet this value.

If you measure something of a similar shape, that you know has accurate form with the same # hits etc as the part, and it comes out with much better roundness, you can use that to prove the inspection method isn't contributing to the .0008" of measured roundness (as Jeffman suggested above)..

Also, when it starts getting down to tenths of an inch, everything matters.
You can improve accuracy by doing a few things, the first would be do above (measure ring/plug gage), and see how it comapares to what the ring or plug is calibrated to be.
From there: you can calibrate the qual sphere with the same # of diameter hits you are taking around your diameter.
Calibrating and taking hits about a diameter using prime numbers (some say 23 minimum) & same # hits for cal diameter and measured diameter is supposed to help as well, as it will have a better chance of capturing any lobing pattern.
Increase probe hits of course improves repeatability.
Decreasing probe length will always improve repeatability, as does a larger ruby/probe diameter. a 5x10 probe would be most ideal IMO
prehit/retract distances, alignments to attain super-accurate vectors, touchspeed, and environmental conditions are other variables one can play with to improve repeatability & accuracy.

[EmilySue]

I don't know what we have for NIST gages, but will see what I can find.

Since I don't have a probe changer, I'm using a 1.5 x 30 tip so I can reach in the bore of this little part and still hit the short cylinders and small ledges. So I'm trying to do all my fact-finding with the same probe.

This post has given me lots of things to work on and tweak. :-)

[dph51]

Do you have tri-mics to measure the o.d.?

If so, how do the results compare?

[EmilySue]

I wish. We only have a 1-2 tri mic. :-/