CMM program correlation.

[Schrocknroll]

How many of you are required to do a correlation study on your program before accepting final product? This is our procedure, but I question the reasonableness of it.

It basically consists of running the program on a given part. Then either check it on the plate using layout methods or check it manually on a CMM or check it against another CMM program written seperately on another CMM. If the difference between the 2 results are within 10% of the tolerance then all is well. If the difference is between 10% to 20% then you need an explaination. If it is over 20% then you have to figure out why and do another study.

My problem with this is if you are checking a straight diameter, the diffence between the results has more to do with the CMMs than the program. Most of our programs consist of measuring bolt circles, diameters and drop dimensions. There is some compound angles and gage points, but it's not the majority of the dimensions.

We have to do it so I do, but to me it just seems like a big joke, for the most part. I can see double checking the complex alignments and gage points, but not the straight forward stuff.

Does anyone else have an opinion on this?

[Paul Sarrach]

That was long ago, when they first stated getting CMM's into places that always did Plate Layout. At the end we just do a study if they had SPC Ports, and would do a correlation to that.

The CMM is not a good tool for reporting Diameters Period.

And I believe our Industry accepted standard between CMM is .05mm, where we wont complain if our numbers are within that to theres.

[djayne]

We check new parts on shop floor cmm PCDMIS and check again on lab CMM LKDMIS and if they both show a dimension out then we have to find a manual method to confirm and sometimes even have an expensive hard gage made in a rush by a local tooling company. I have never seen the two CMMs not be right after verifying with a manual method.

[Underspec]

There are two best fit methods i know of. Gauss and Chebychev. If both softwares for the CMMs are the same I would say it would be ok. I would probably use two scanning enabled CMMs and dimension max/min to get a good comparison. Attached is the info on the methods. I believe PCDMIS uses the Gauss method, but I am not 100% sure.

[JCcmm]

We do correlation studies with 2 other CMMs. We generally allow .050mm. Sometimes more depending on the feature. Most linear dimesions are within .025, but diameters (as Paul and Underspec stated) are always out to lunch. We rarely use the CMM as a reliable gage for diameters. We have various hand helds and air gages to verify those. We have a few diameters that are created from linear intersects that reliably report diameters or have high enough tolerances that "hide" the variation. Some of our diameters have tolerances of ±.04mm, those can't effectively be gaged by our CMM. But yeah, we do 'em too! (Our Machine Shop is always crying about "machine variation"!

[Schrocknroll]

I've never noticed a problem checking diameters with a CMM. I've checked 32 inch diameters that were .005 out of round on our DEA CMM running PCDMIS with 70 hits and then checked the same diameter on our ZEISS with a VAST scanning head running UMESS using .005 increments between points and usually come within .0001 of the same dimension.

BTW I can check .250 holes on the ZEISS and be within .0001 of the air gage.

[Paul Sarrach]

Zeiss, the Rolls Royce Of CMM's

[Tested]

We do correlation studies often. We do them by run the same parts 3 times on 2 different machines, and then the results must fall with 10%.

We also check diameters with our CMM "DEA" and hold .0001" with no problem. The only time I have problems with a CMM checking a diameter is on a drafted hole; and then only if I can not create the necessary features to create a theo. diameter.

[bw_bob]

The only time we re-check is if a dimension is out of tolerance and it can be checked quickly by caliper,height gage,comparitor or plug gage.

[djayne]

What makes it so difficult for the CMM to check diameters accurately? We've always just taken the CMM data for its word. Is there something we should consider about using CMM data on bearing bores with only 13 microns tolerance?

[Matthew D. Hoedeman]

Well, the main reason I use for explaining this problem is that you are taking at least 3 hits to measure the circle. There IS probe error, you calibrate to help compensate for this, but it will not get rid of ALL the error, even after calibration. AND, the more hits you take in DIFFERENT directions add up this small amount of error and add it all into the feature. A plane, for example, has ALL the hits coming from the same direction, only 1 direction of error. Same with a line. BUT, a circle has 4 directions of error and a sphere has 5 directions. Remember, nothing is perfect. AND, using a TP2 will increase the amount of error due to it's lobbing characteristics. It will come up with a good value for the centers since the direction of error IS all either towards the center point (external feature) or away from the center point (internal feature) but the actual size is what can easily come into question, much more so than the location.

[Tested]

If your holding a calibration of 0.003 standard deviation which is as high as we allow. Your average distance from the mean will be less than half STDDEV. and then if you run the program after creating it the program does further corrections to the feature. Granted nothing is perfect but a CMM is a very good tool for checking a diameter.

[Matthew D. Hoedeman]

The calibration will give you good results AS LONG AS THE TOUCHES are at the same point on the tip from the cal-tool to the feature. This is especially true with TP2's. I have seen as much as 0.0006" of difference depending on where the contact is (with less than 0.0002" positional change), feature-to-tip-to-calibration sphere. The number of hits, start/end angle, lots of things will effect this. 5 hits on calibration sphere with 3 for a circle will never have you hitting ALL the hits where it hit when calibrating. 4 hits for a circle can be done (or at least close). Lots of variables come into play here.

[Mike N.]

I've never noticed a problem checking diameters with a CMM. I've checked 32 inch diameters that were .005 out of round on our DEA CMM running PCDMIS with 70 hits and then checked the same diameter on our ZEISS with a VAST scanning head running UMESS using .005 increments between points and usually come within .0001 of the same dimension.

BTW I can check .250 holes on the ZEISS and be within .0001 of the air gage.

I dont have dia problems either, but I am using an analog SP600 probe and it seems to be very reliable when used in conjunction with "inscribed, circumscribed". The reported size does simulate the functional pin on 90% of the calculations.

[Cumofo]

I can get a hole to read within .000030 of air gauge dimensions. The trick is to do a tip diameter edit using a ring gage. Here is copy of my program it is very basic:

PART NAME : shut down CMM
REV NUMBER :
SER NUMBER :
STATS COUNT : 1

STARTUP =ALIGNMENT/START,RECALL:, LIST= YES
ALIGNMENT/END
MODE/MANUAL
CHECK/ 0.25,1
MOVESPEED/ 25
MANRETRACT/2.54
FLY/ON,0.1181
LOADPROBE/MICRO PECK
TIP/T1A0B0, SHANKIJK=0, 0, 1, ANGLE=0
FORMAT/TEXT,OPTIONS,ID,HEADINGS,SYMBOLS, ;NOM,MEAS,TOL,DEV,OUTTOL, ,
MODE/MANUAL
MOVE/POINT,NORMAL,0.0005,33.7092,-10.5906
NEW =LABEL/
PLN1 =FEAT/PLANE,RECT,TRIANGLE
THEO/10.8484,13.9598,-29.6315,0.000049,-0.0003984,0.9999999
ACTL/11.6651,4.3049,-26.6827,0.004964,-0.0044631,0.9999777
MEAS/PLANE,4
HIT/BASIC,NORMAL,11.2873,14.0888,-29.6314,0.000049,-0.0003984,0.9999999,12.295,4.4747,-26.685,USE THEO = YES
HIT/BASIC,NORMAL,10.3,13.4754,-29.6316,0.000049,-0.0003984,0.9999999,11.5449,4.8657,-26.6797,USE THEO = YES
HIT/BASIC,NORMAL,10.6336,14.6131,-29.6312,0.000049,-0.0003984,0.9999999,11.1272,4.2368,-26.6803,USE THEO = YES
HIT/BASIC,NORMAL,11.1729,13.6619,-29.6317,0.000049,-0.0003984,0.9999999,11.6935,3.6425,-26.6859,USE THEO = YES
ENDMEAS/
CIR1 =FEAT/CIRCLE,RECT,IN,LEAST_SQR
THEO/10.6696,14.0823,-29.7704,0,0,1,0.6497
ACTL/11.8328,4.2272,-26.8052,0,0,1,0.65
MEAS/CIRCLE,4,WORKPLANE
HIT/BASIC,NORMAL,10.9945,14.0823,-29.7704,-1,0,0,12.1562,4.2589,-26.8052,USE THEO = YES
MOVE/CIRCULAR
HIT/BASIC,NORMAL,10.6696,14.4072,-29.7704,0,-1,0,11.5093,4.259,-26.8052,USE THEO = YES
MOVE/CIRCULAR
HIT/BASIC,NORMAL,10.3447,14.0823,-29.7704,1,0,0,11.8601,4.5511,-26.8052,USE THEO = YES
MOVE/CIRCULAR
HIT/BASIC,NORMAL,10.6696,13.7575,-29.7704,0,1,0,11.8601,3.9033,-26.8053,USE THEO = YES
ENDMEAS/
A0 =ALIGNMENT/START,RECALL:STARTUP, LIST= YES
ALIGNMENT/TRANS,XAXIS,CIR1
ALIGNMENT/TRANS,YAXIS,CIR1
ALIGNMENT/TRANS,ZAXIS,PLN1
ALIGNMENT/END
MODE/DCC
CLEARP/ZPLUS,0.5,ZPLUS,0.5
MOVE/CLEARPLANE
CIR2 =AUTO/CIRCLE,SHOWALLPARAMS = YES,SHOWHITS = YES
THEO/0,0,0,0,0,1,0.6497
ACTL/0.0001,-0.0002,0,0,0,1,0.6498
TARG/0,0,0,0,0,1
THEO_THICKNESS = 0,RECT,IN,CIRCULAR,LEAST_SQR,ONERROR = NO,$
AUTO MOVE = BOTH,DISTANCE = 0.5,RMEAS = None,None,None,$
READ POS = NO,FIND HOLE = NO,REMEASURE = NO,$
NUMHITS = 21,INIT = 0,PERM = 0,SPACER = 0,PITCH = 0,$
START ANG = 0,END ANG = 0,DEPTH = 0.1,$
ANGLE VEC = 1,0,0
MEAS/CIRCLE
HIT/BASIC,0.3249,0,-0.1,-1,0,0,0.325,0.0001,-0.1
HIT/BASIC,0.3104,0.0958,-0.1,-0.9556,-0.2948,0,0.3105,0.0958,-0.1
HIT/BASIC,0.2684,0.183,-0.1,-0.8262,-0.5633,0,0.2685,0.1828,-0.1
HIT/BASIC,0.2026,0.254,-0.1,-0.6235,-0.7818,0,0.2025,0.2539,-0.1
HIT/BASIC,0.1187,0.3024,-0.1,-0.3653,-0.9309,0,0.1186,0.3023,-0.1
HIT/BASIC,0.0243,0.324,-0.1,-0.0747,-0.9972,0,0.0243,0.3238,-0.1
HIT/BASIC,-0.0723,0.3167,-0.1,0.2225,-0.9749,0,-0.0724,0.3166,-0.1
HIT/BASIC,-0.1624,0.2814,-0.1,0.5,-0.866,0,-0.1625,0.2812,-0.1
HIT/BASIC,-0.2382,0.221,-0.1,0.7331,-0.6802,0,-0.2383,0.2208,-0.1
HIT/BASIC,-0.2927,0.141,-0.1,0.901,-0.4339,0,-0.2928,0.1406,-0.1
HIT/BASIC,-0.3212,0.0484,-0.1,0.9888,-0.149,0,-0.3212,0.0479,-0.1
HIT/BASIC,-0.3212,-0.0484,-0.1,0.9888,0.149,0,-0.3212,-0.0482,-0.1
HIT/BASIC,-0.2927,-0.141,-0.1,0.901,0.4339,0,-0.2928,-0.1409,-0.1
HIT/BASIC,-0.2382,-0.221,-0.1,0.7331,0.6802,0,-0.2384,-0.2209,-0.1
HIT/BASIC,-0.1624,-0.2814,-0.1,0.5,0.866,0,-0.1625,-0.2815,-0.1
HIT/BASIC,-0.0723,-0.3167,-0.1,0.2225,0.9749,0,-0.0723,-0.3169,-0.1
HIT/BASIC,0.0243,-0.324,-0.1,-0.0747,0.9972,0,0.0243,-0.3242,-0.1
HIT/BASIC,0.1187,-0.3024,-0.1,-0.3653,0.9309,0,0.1187,-0.3026,-0.1
HIT/BASIC,0.2026,-0.254,-0.1,-0.6235,0.7818,0,0.203,-0.2539,-0.1
HIT/BASIC,0.2684,-0.183,-0.1,-0.8262,0.5633,0,0.2686,-0.1831,-0.1
HIT/BASIC,0.3104,-0.0958,-0.1,-0.9556,0.2948,0,0.3107,-0.0958,-0.1
ENDMEAS/
DISPLAYPRECISION/5
DIM LOC1= LOCATION OF CIRCLE CIR2 UNITS=IN ,$
GRAPH=OFF TEXT=OFF MULT=10.00 OUTPUT=BOTH
AX NOMINAL MEAS +TOL -TOL DEV OUTTOL
D 0.64975 0.64980 0.00000 0.00000 0.00005 0.00005 -------->
END OF DIMENSION LOC1