Inconsistant Results

**\v/inston** · 03-27-2006, 08:47 AM

do a gage R&R for each machine.

**Roberto** · 03-27-2006, 09:06 AM

Skew points mean vector points?

I don't understand why the 3rd machine can not run on DCC something not right with the alignment. Is the probe configuration the same on the 3 machines? How about probe calibration results?.

Are you using fixture for holding or open set up? Is the part solid (metal) or plastic?.

Roberto.

**RussL** · 03-27-2006, 09:25 AM

Skew points = vector points used for both manual and DCC iterative alignments. manual has tol of .010, DCC has tol of .0005.
Probe configuration is same on all three machines, calibration results acceptable on all three machines.
We are inspecting/calibrating a metal fixture used to drill holes in part.
Running on the Alpha, the machine is off so far in X/Y plane, it can not find Z banks.

**Paul Sarrach** · 03-27-2006, 09:34 AM

I would also check you styli and you tp2, tp20, (what ever it is) Change them from machine to machine. see if the results follow the probe

**T Miller** · 03-27-2006, 09:36 AM

you may also want to check that your probe comp is on/off the same as the others

**Matthew D. Hoedeman** · 03-27-2006, 09:38 AM

l'd look at the calibration papers you were given when the machines were calibrated. Not sure how much they will help, but you DO need to look at them. Did they do the 'ball-bar' calibration of the machines? If so, hopefully they have written down the measured values the machines got for the distance between the balls of the ball-bar. I would guess that your machines are NOT linear (meaning 12" does NOT measure as 12" on each machine). I would get any cubical piece of stock and check it on each of the 3 machines, level to it, set XYZ zero to a corner, then check the dimensions of all 6 sides the best you can. Also, use the biggest piece you can get that you can fit on all 3 machines.

The ball-bar calibration DOES NOT do anything for the linear of the machine, all it does is make it electronically square.

**RussL** · 03-27-2006, 09:45 AM

Thanks for the responses.
I believe the two older machines are calibrated ball-bar.
The Alpha is calibrated per ISO 10360-2

**george frick** · 03-27-2006, 10:11 AM

try using a standard 3-2-1 alignment or similar instead of the iterative alignment and see if your results are consistant. seems like there's something dysfunctional in the alignment.

**Hilton Roberts** · 03-27-2006, 11:20 AM

Variable results

What you report on your results is a classic case of why it is important to measure something you know, chart the measurements and make the determination of which machine is not telling the truth.

I know that Matt and I have some difference of opinion on the value of ball bar calibration but I will pass on my two cents worth.

An un-calibrated ball bar is ONLY good for squaring up geometry. Period.

A CALIBRATED ball bar CAN be used to verify accuracy IF....IF the assumption is accepted that scale values HAVE NOT CHANGED.

A test part measured both in the short term and over time can be used to determine if a change has occurred.

If you have changed a scale, or adjusted a reader head or replaced a reader, then in my opinion, you need to relaser the machine.

In any event, the laser is the single most accurate method to determine scale accuracy that I am familiar with. I know Zeiss uses a step gage and their machines are very good but as a practical point the lasers are here to stay. As long as the sensors for humidty, pressure and temperature have been calibrated by a REPUTABLE lab, they are as good as it gets.

Normally ball bars are used to set geometry and the machine is lasered to establish scale accuracy. Scales on today's machines are much better than the old scales used 20 years ago in which the machines were checked every 4 inches or so and the scales pushed and pulled to set the distances.

Most machines today have very extensive compensation maps and if those values are not determined correctly or if the values get the sign changed when the compensation map is updated, then your results are going to be suspect.

A test part measured many times over a short period BEFORE calibration and again AFTER calibration can tell you if the machine is improved or not....or if the compensation map has bad numbers in it......

My experience on the Xcel machines has been the scale accuracy needed to be tweaked one time after the machines were installed here in Mesa. There was some settling in after the machines were shipped from Rhode Island.

What I think you are seeing is a perfect example of a machine being "calibrated" and incapable of measuring a real world part.

If you really want to get to the bottom of the problem ( considering what you are reporting, I bet you do ), then you really need a part that has the features you measure on the parts you produce and you need a program to measure that part that runs on all of your PCDMIS machines and you need to chart the results.

There are three scenarios that will most likely show up in your results.

1) Each machine is consistent in the reported values but the values differ from machine to machine. Another way of saying there is a bias between machines that may be due to environment, vibration, temperature or other anomaly that would be detectable on a control chart.

2) One or more machines inconsistent with results and the cause is to be determined.

3) None of the machines consistently report the same values for the same measurand and the cause for the variation has not been determined.

Another place to look for problems is the calibration house. If it is just someone who went out and purchased a laser interferometer and is now in the business of "calibrating" machines, you need to be very cautious.

It is one thing to laser the machine and quite another to use the results to update your compensation map and still another to make sure you can measure a part.

Periodic measurement of something not off the product stream can be very valuable in making the determination if the problem is in production or in the measurement of the part. You are the referee on what is good and what is bad and you must make sure the numbers you provide are accurate and repeatable.

Please feel free to call me at 480-891-7263 if you have any questions.

Just as an add on. If you have the capability in your plant to do the calibration and compensation map updates, that is the way I would go.

My company does not want to do that here. I have used Brown and Sharpe since day one on our machines and I am not disappointed with the quality of service.

I am satisfied that our machines are accurate and repeatable and that is what it is all about.

HTH

Hilton Roberts

**Matthew D. Hoedeman** · 03-27-2006, 11:52 AM

A little clarification, I never said that a ball-bar was useless, only that it can ONLY square the machine electronically. If the user does NOT watch the calibration and the calibration tech closely, you may or may not get a good calibration. If the calibrated ball-bar distance of 12" checks 12.010", but the tech ONLY get the machine square and does NOT tell you that your linear is wrong and gives you a calibration paper saying you are good, then you just got ripped off. The calibration tech from B&S who re-set up my machine after it came back from retro-fitting, told me, after I asked him WHY the measured dimension didn't match what his ball-bar calibration sticker said it should be, told me it didn't matter what it checked at, just as long as the opposite diagonals checked the same (ie sqaure). Now, this IS a different situation than 99% of you will be going through when they come in to calibrate your machine, however, this was after my machine had been sent to B&S for retrofitting by B&S at the B&S factory and calibrated by a B&S tech at the plant, the table map made and the machine shipped back to us. They were then (by contract) to re-calibrate the machine once it was set up in our plant. Then tech came in and that is what happened. Now, maybe they just didn't want to go through the extra cost (to themselves) by doing it right, but they did NOT do it right. There was a sizable amount of error between the measure and the value on the ball bar itself, and the B&S tech told me that it didn't matter, "as long as the diagonal measurements are the same", but that is NOT true and anyone who knows anything about QC knows this.

And Hilton, you did hit the 2 main points right off, an UN-CALIBRATED ball bar can only square the machine and a calibrated ball-bar can be used to CHECK the linear accuracy, but that is where it ends. The ball-bar can NOT correct the linear accuracy and, based on the ONLY time I have used B&S and their ball-bar for a calibration, it was NOT done correctly and the tech told me that it was correct. Square means nothing if the linear is not right, the linear did not match, he said that it didn't matter. Run that by me again? It is OK if 12" measures 12.010" as long as it is square? This is my main beef with the ball bar. A square machine is not calibrated and if the tech is only looking at the cross-measurement-values, and not the actual values, then you probably are not calibrated.

**Hilton Roberts** · 03-27-2006, 12:22 PM

Twin brothers born of different mothers

Matt,

I think we are connected........ I go along with everything you say. I do not think I ever said you could SET accuracy with a ball bar, but I would have said you could verify a change. I agree 100 percent that an uncalibrated ball bar is only good for setting geometry. You will probably remember the ORIGINAL NIST artifact ball bar did NOT have a calibrated length. Only the sphericity of the balls was reported. That makes the original intent of the NIST artifact to do machine geometry and not linear accuracy. As for the calibrated ball bar being used to SET linears.......I agree with you 100 percent........bad idea. I try to never use a single point as a value for accuracy. I would much rather have more data to view trends.

We think a lot alike.......now THAT is scary

H

**RussL** · 03-27-2006, 12:32 PM

Again, thanks for the replies.
As far as measuring a sample part we know periodically, I had made the suggestion numerous times in the three years I have been placed in "QE".
As far as the capability to check the machines with the laser, yes, we have that capability here, but don't do it. The calibration tables are encripted on the Alpha (another problem).
B & S does the calibrations here.
We inspected a part we use to check out the machine tools here, on each CMM, the results are much better (all readings are within .002 from machine to machine - I thought they should be within .001 at least).
To me this implies the (original) problem may more likely be software related somehow (Perhaps how different controllers handle (the same) iterative alignment ?)

**STANLEY** · 03-27-2006, 02:55 PM

In the future make up a master part, and check it on each machine and track the results. Do this after you resolve your current issues.

**ag162** · 03-27-2006, 03:22 PM

I totally agree with the fact that when a tech calibrates your CMM, that you have to be a be aware that it may or may not be measuring true amounts. They check for repeatability and squareness. Always measure a known object to verify the results and be able to ask the important questions without having to make a call.

A.Gore

Inconsistant Results

Inconsistant Results

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Related Topics