Radial runout

[[Pa...]]

Hi,
When measuring a typical shaft, I get (too) high values for radial runout. The same shaft has been measured by a subcontractor, resulting in an approx. five times lower deviation.

See attached picture for an explanation of the setup:
* Datum A-B (3d line) consists of 2 cylinders measured at both ends of the shaft. (Datum A-B = Y vector)
* Each cylinder is created by recall of 2 circles (e.g. A1 and A2), using Cookbook settings for filtering.
* Each circle has been scanned (180deg X+, 180deg X-) using Cookbook settings for speed, number of points and filter settings.

Value for radial runout on Cylinder 1 to Datum A-B is 0.01 mm, whereas our subcontractor gets 0.0022 mm. The latter is similar to what we see in production and makes sense regarding the production process. A runout value of 0.01 mm does not....
After trying several different setups and settings (calibration, scan speed, different probes, alignment), I have ran out of ideas for a possible solution.

Additional information:
* 3mm probes of 60mm length for both X+ and X- (Tensor calibration, S<0.0003)
* Climate Controlled environment; parts have been in the room for >24 hr
* CMM: Prismo Ultra 7/10/5
* Calypso 5.8.28 service pack 7

Any suggestions?

Shaft example.jpg

[[Ri...]]

How do your results look if you use a Stepped Cylinder instead of a 3d-Line to create the axis?

[[SH...]]

Instead of two cylinder measure two circle.

[[Me...]]

Doubt it will change but do you have the evaluation settings for scanning applied to your recalled Cylinders?

[[Pa...]]

I have tried both suggestions (Stepped Cylinder, 2 circle reference), but they don't make a big difference (max. 0.0002 mm)

[[Pa...]]

Yes, I have applied the evaluation settings on the recalled cylinders as well. I have been playing around with the values (mainly UPR value), but this doesn't really change the result.

PS. I have been trying to contact our subcontractor in order to get their settings and/or program, but he is on holiday ....

[[Cl...]]

Have you verified/corelated the result using a test indictor?

[[SH...]]

Do you have manual facility to check the runout?.
I think run out is more accurate in manual then CM M.


Runout is mathematically concentricity plus roundness.
You can counter verify it by calculating the concentricity of your feature with the 3D line plus cylindricity of the feature.

[[He...]]

What is the reason for the runout? Is it the location, form or coning of the feature?

If it is the position:
Is it possible that the shaft bends when positioned like that in the V-blocks? Is it possible to measure it vertically instead? If not. What happens if you rotate the part 180°?
Also what is the form of the datums?

If it is form:
Make sure that the stylus system is good. Check if there are any transition erros between different styli. (I assume you are measuring 2 circle segment at 180° each with two different styli in the same stylus system.)

If it is coning:
Make sure that the manual inspection and supplier have measured total runout and not single runout.

[[Ma...]]

Try just two independent cylinders (no recalls) for -A- and -B- and using these inside a step cylinder. The circle paths you are recalling into your cylinders how are they created? Is it two separate circle paths recalled into one and then recalled again into a cylinder? Also things already mentioned in this post but check these circle paths in a roundness plot. If the circle paths are measured with two different stylus and recalled into one you can come across a "mismatch" between the two paths which can skew your center point. Also be sure you know exactly how the subcontractor measured it. You don't want to drive yourself crazy if you are doing things correct!

[[Pa...]]

Thanks to all for your comments.

Unfortunately I can't discuss this matter with our subcontractor, since their operator is on holiday at this moment. That would probably have saved me some time and some damage to my mental health... 😉

I quite sure that the current measured value can't be true. Partially because of the production method of this master part, but mainly since the inspection tool we have build (and where this master is used for) is showing values as expected. This is confirmed by the results of our subcontractor.

Today I have done several tests based on your comments:
1) I have rotated and flipped the parts without significant difference in the results. Due to the size it is not possible to measure it vertically.
2) Lowered scan speeds
3) Changed the 3D line for a Stepped Cylinder
4) Used cylinders based on recalled circles, but also cylinders without recalls.
5) Recalibrated probes (indeed, I am measuring 2 circle segments at 180° each with two different styli in the same stylus system.)
6) Grouped opposite circle segments within a cylinder

Next I have been looking at the form of the cylinders again. The roundness is a bit higher than expected and shows some kind of vibration. Therefore I have been playing with the filter settings, especially with the UPR.

I'm still running some test but to this point, the following approach is giving the best results:
- cylinders withoud recalls
- stepped cylinder as a reference
- 15 UPR instead of 50/150 (dia between 22 and 33 mm) as recommended
- Scan speed: 3 mm/s, instead of the initial 10 mm/s

With these settings the radial runout improves significantly: difference with subcontractor 0.001 mm instead of 0.008 mm.

[[Me...]]

Please sign in to view this quote.

Do you have an Active or Passive head?

[[Da...]]

By the picture it looks like an XT or XT Gold

[[Ri...]]

It's a Prismo Ultra, so it should be a VAST Gold.

[[Ow...]]

I'll quote Henrik's question: What happens when you rotate it 180°? if the direction of the run-out doesn't follow to the other side (like it should) then it's probably something with the two probes not matching. As mentioned, what's the form of the circle or cylinder of cylinder 1?

What's the blue features on the shaft? Are they installed after the subcontractor has measured it? Totally going off the wall here but, the blue features aren't magnetic are they?

A test indicator should be used to double check who's right or wrong.

[[Ri...]]

Yeah. You don't want to see this.

This is due to probe offset. It's common, and should be addressed.

[[Da...]]

0.5 + L/500 μm = 300mm - - - - - (0.5 + (300/500) = 1.1 Deviation (.0011)
You should be ok with the CMM. I really want to see the roundness of the feature.
Or is it the Position? How long has it been setting on this Fixture? You could be getting bending from gravity?
because of the length it could be a problem. When checking step gage it will start bend after setting on blocks for around an hour or so.

It would only need to bend .00015 inch to see that.

[[Pa...]]

Latest result on a different master (same dimensions)

- Roundness datum A = 0.0018 mm
- Roundness datum B = 0.0015 mm
- Roundness Cylinder 1 = 0.0019 mm
- Radial Runout Cylinder 1= 0.0099 mm

It's not the best quality picture, but it gives an impression of the roundness.

Additional information:
- The part has been rotated many times during yesterdays tests.
- The blue features are press-fitted collars (not magnetic) for identification. They were also installed when measured by subcontractor.
- The head is a Vast gold (active)

[[Ri...]]

This is great information it appears to be a location issue then.

Can you report the A1 & A2 error for the Cylinder that is used for the Radial Runout? I think a question got raised about sag.

[[Da...]]

You don't have probe mismatch problem roundness looks great.
Was it setting in this fixture for a long time after the contractor measured it? If it bent you will not fix it by just turning the part.
You will need to stand it up on one end and hope that it will remove the bend.

Using indicator could help to find the bend then rotate to the high side up. That could also help?
If that is the change that your are seeing just a Guess.

[[Ma...]]

Are you evaluating radial runout of the cylinder or of the individual circular scans that make up the cylinder? If you evaluate radial runout of a cylinder the answer provided by Calypso is actually the same as the cumulative radial runout value. You have to evaluate radial runout of the individual circles and find the greatest deviation to get the correct answer.

[[Da...]]

Please sign in to view this quote.

That is a really good point.

[[Pa...]]

I have done some new tests, after the part has been off the cmm for 2 days. See attached PDF.
When looking at the data of individual circles, the outcome seems logical to me.

Next I have done a similar measurement on a different master part, which has not been on a cmm for more than a week. The results are comparable.

Since the results are repeatable and logical I tend to believe they are right. But I still can't explain why our subcontractor get's better results on radial runout. Next week their operator is back and I will try to get his program.

Master part.pdf