You are on page 1of 16

Coordinate Measuring Machine Statistical Process Control

PRESENTER
Hilton.L. Roberts
Precision Inspection Group
Boeing Helicopters
5000 East Mcdowell Road
Mesa,Arizona
85205
602-891-7263
hihon.lrobertS@boeing.com

PAPER AUTHOR
Hilton.L. Roberts
Precision Inspection Group
Boeing Helicopters
5000 East Mcdowell Road
Mesa,Arizona
85205
602-891-7263
hilton.l.robertS@boeing.com

ABSTRACf

The quality of the measurements from a Coordinate Measuring Machine ( CMM ) is in


large part dependent on the stability of the environment surrounding the CMM and the
stability of the processes that produce output from the collected data points.
This paper will discuss how an artifact was chosen and programmed to allow for CMM
confidence checking and how Tool Qualification has been identified as a significant source
of measurement uncertainty.

1998 NCSL Workshop & Symposium

673

Session 68

INTRODUCTION
A desirable outcome is that Coordinate Measuring Machines be capable of producing
identical outputs from identical inputs. The measurement processes should produce results
so that we may be confident that we are not accepting bad product nor are we rejecting
acceptable product. Ifwe are to use Coordinate Measuring Machines efficiently, then we
must be able to provide reliable and consistent results to Manufacturing.
A common method of validating the suitability of a CMM for use in measuring work
pieces is the manufacturer's statement of accuracy coupled with either a laser mapping
and/or a Ball Bar check. A completed ''Certificate of Accuracy" was all that was required
to begin measuring and reporting values descn"bing the geometrical characteristics of
']>roduct". This approach allowed the CMM to operate in a "state of grace" until the
calibration due date was reached and did not consider that the machine might not be stable
for the duration. One of the reasons that CMM results (Especially those that require
manufacturing to rework pieces) have been received with some skepticism in the shop is
that the accuracy of the measurements is suspect. All too often when Coordinate
Measuring Machines are tasked with verifying a discrepant feature, the readings do not
repeat.
It seemed logical then to try to choose an artifact that could be used to test whether or not
a CMM could consistently report the same values for the same features and whether or
not the reported values for those features could be contained under a normal curve.

BACKGROUND
Some of the information used to conduct these tests was provided by the CMM
manufacturer and some was provided by CMM Instructor Programmers. While this
information was very helpful, it was considered incomplete by those with a background in
Dimensional Metrology. What was missing was a lack of knowledge of how to prepare for
and conduct tests to prove or disprove the accuracy claims.
The manufacturer's accuracy statements are not usually based on the configuration of the
machine that is used to measure parts. The manufacturer does not know what kind of
probe configuration will hang on the end of the ram to measure the parts produced.
Therefore we cannot realistically expect the same accuracy on machines in their
manufacturing/inspection settings.
We need to know what kind of accuracy CAN be realized. We must develop test
parts/programs so that baselines can be established on the CMM for use in determining
when the machine begins to change.

1998 NCSL Workshop & Symposium

674

Session 68

PROCEDURE
One of the statements we heard was that "once tools are qualified in a particular
configuration, they are used until either the tool configuration is changed or the machine
suffers a probe crash".
Chart IB shows a baseline check of the qualification ball from point 65 to point 80. That
tool file was saved and used to measure the qualification ball one day later. That chart,
which starts at point 81, shows a downward shift of approximately .0002 inches. A slight
upward shift one day later can be seen at point 101.
The data suggests that the CMM changed relative to the tool qualification and that change
in tool offsets was reflected in the change on the charts. The data further suggests that
tool qualification should take place before measuring critical parts and that "old" tool
qualifications should not be kept.
The next question that begged to be answered was tool qualification itself Tool
qualification is a twofold process on this CMM. First, we find the center of the
qualification sphere in order to establish the coordinates of the sphere from the X, Y and Z
axis resolver null points. We do this by defining to the computer a negative Z axis tool
without X and Y offsets and a Z length measured to the nearest .1 inch from the center of
the stylus tip to the center of the ram collett. Once this value is determined, other tools in
other attitudes may be qualified. This is done by measuring 5 points on the qualification
sphere by each tool The computer will determine the X,Y and Z lengths of the tools and
use these "offsets" to compute geometry from measured points on parts.
It seemed reasonable to assume that if the manufacturer expressed the uncertainty of the

CMM, determined by a ball bar as .00075 inch bandwidth and that the process of
qualifying tools was "normal", it should be possible to qualifY several tools automatically
and use one tool to establish X,Y and Z zero in the center of the sphere. The rest of the
tools should then be able to measure the size and location of the sphere within that
bandwidth and that those measurements would be normally distributed.
The CMM that was used in these tests was dedicated to wing measurement for a small,
pilotless aircraft. The part program for measuring the wing collected several hundred data
points along the upper and lower mold line as well as the leading and trailing edges and
pivot hole. These data points were reduced to yield aero characteristics related to surface
area as well as lift coefficients and rigging parameters of the wing. Erroneous data used to
rig a wing could cause a failure in flight.
To minimize the possibility of that occurring, a part was chosen as a confidence check that
had some of the characteristics that were of interest on the wings themselves. An angle in
the XV plane was needed as well as a feature that required measurement from both sides
of the part.

1998 NCSL Workshop & Symposium

675

Session 68

The part chosen was the Geowidgit part that is used as a training aid for programming on
our manual machines. This part is aluminum and has several features similar to those of
interest on the production wing.. The part is smal1 enough that it can be measured in more
than one place on the CMM surface plate and can be used to monitor CMM performance
in the four quadrants: X+Y+,X+Y-,X-Y- and X-Y+.
Tools in five different attitudes are used to measure the wing. A program was written to
qualify those tools under numerical control and then to use the negative Z tool to establish
X,Y and Z zero in the center of the qualification sphere. All five tools are used to measure
the sphere 20 times to get a baseline on the machine and X Bar and R charts were
constructed. These charts have been maintained since October 1989. An additional
program was written to automatically align and set zeroes on the GEOWIDGET part.
To date, the charts show that this CMM does operate in a state of statistical control.
Nevertheless, it is interesting to note that although the machine is not in control. certain
things, when they happen, can show a sudden and dramatic shift in the charts.
The program for the GEOWIDGET is run every day. Running the program in the four
quadrants on the machine surface plate yields four measured values for each feature
measured. For example, the counterbore on the Y+ side of the GEOWIDGET is measured
in each of the four quadrants. This provides four data values for that feature that will yield
one data point for the charts on the X,Z location and Diameter of the counterbore.
A glance at chart 6A shows several interesting things. Most striking is the shift in the X
location of the bore. The important thing to consider is whether or not the bore actually
shifted or if something happened to the CMM to cause an APPARENT shift in the bore.
The first thing that was done to determine what happened was to requalify the tools and
run the program again to remeasure the Geowidget. The chart shows that requalifying the
tools and rerunning the program did not change the reported location of the bore. The
operator was questioned and reported that the CMM air supply had been interrupted by a
high temperature alarm that triggered a compressor shutdown. The operator then reported
that when the air supply fell below the pressure to keep the ram up, the ram fell and the
probe hit the surface plate. An interlock that was supposed to keep that from happening
was found to be set at 13 psi. The pressure required to keep the ram from falling is 60 psi.
A right angle iron was set up so that it was parallel to the Y axis of the machine and was
adjusted until it was parallel within .0002 inch as measured by a negative Z axis tool. The
probe was then pointed Y+ and was brought to the front face of the angle plate ( on the Y
- side) and an X axis zero was set. The probe was then pointed Y - and moved to the
other side (Y+ ). The face of the angle plate was measured and was found to be .036 inch
out of square. It was adjusted until the out of squareness condition was less than .0005
inch. A glance at the charts shows that the process was restored. The error was caused
when the tool entered the bore to measure the location. By being out of square in the XV
plane, the probe shanked out and showed a shift in the X axis of the bore.

1998 NCSL Workshop & Symposium

676

Session 68

What is really important here is that the GEOWIDGET part was being used as a
confidence check in a MEASURING process. Had a production part been charted as part
of measuring or adjusting MANUFACTURING process capability, it is likely that
extensive time and resources would have been expended looking in the wrong area to fix a
problem that was showing up on a control chart.
Chart SA shows an angle in the XV plane on the GEOWIDGET. This angle is important
in the measurement of the wings because of a critical bore that runs through the wing
pivot plate and intersects the manufacturing chord plane. To bore this hole at the proper
angle and to achieve the true position requirement of the bore, the program that measures
the bore does a planar alignment on the face of the pivot plate on the Y+ side of the wing.
The CMM was also used to establish the location of a seal hole on the edge of the wing
when no tool was available to drill the hole. This was done by driving the CMM to the
location of the hole as close as posSlole manually in X,Y and Z. The machine was then
backed out ofposition in the Y axis and a dab of white lacquer was applied to the tool tip.
The probe was then moved in the Y axis alone to touch the wing plank and transfer the
lacquer to the wing. The hole could then be drilled manually. The procedure was to make
sure the tool tip was cleaned after this operation to make sure that no lacquer remained in
the tool tip. In chart SA, the probe was not cleaned after use and a flake ofpaint remained.
This caused the angle on the GEOWIDGET to measure. S degree different than previous
charts. An angle of this magnitude amounts to .014 per inch. Considering that the true
position of the pivot bore is at a position over 2 inches away from the face of the planar
alignment, it is easy to see that significant errors in the reported bore location can be
produced.
The fix for this type of problem is to bore the hole oversize, tum a bushing and
interference fit the bushing to the bore and reb ore the pivot hole. The fix is very expensive.
Once again the charts pointed to a problem The charts did not tell us how to fix the
problem nor did the chart specifically indicate what the problem was.
The drive system on this CMM has a split gear type of arrangement with a torsion bar
applying torque to reduce gear backlash. Chart IB shows a typical pattern of the charted
values of the X location of the Qualification sphere. At the right hand of the chart, a
sudden shift in both the average and range charts indicate that something happened to the
machine. Due to the noise that developed in the X axis it was determined that loss of the
X axis anti-backlash adjustment was the probable cause. The machine was in warranty at
the time and the manufacturer sent a technician in to fix the problem When the cover was
removed from the X axis drive motor, it was discovered that the anti-backlash adjustment
had slipped.
Chart IC shows that when the anti-backlash adjustment was reset, the charts resumed
their previous pattern on the range chart but the average chart showed a shift. This was
most likely caused by the technician using a slightly different value for the torque
adjustment.

1998 NeSL Workshop & Symposium

677

Session 68

Now eight years later, we have two new Coordinate Measuring Machines. Utilizing
lessons learned from the past, we decided to incorporate the qualification sphere
measurement as part of the machine run off and to provide a baseline for the machine at
the factory before delivery and at our location upon delivery. We wanted to know how
well the machines performed here in the location they would be used in compared to what
the machines delivered at the factory. We ran a 25 point baseline at the factory and
compared those numbers with numbers generated here after installation. The four charts
that you see are the X, Y, Z location and Diameter of the sphere. The twenty five point
baselines are clearly delineated on the chart. The spike at point 51 was a wake up call that
something was happening to our SPC check. All of a sudden the machine was not able to
determine the location in of the qualification sphere closer than .0045 inch. What was the
problem? We monitored the temperature overnight, examined the program for corruption
and could not figure out what was happening. We took the existing program and ran it on
the other machine and the numbers dialed in. The problem was then isolated to one
machine. That did not help find the problem We looked at what it would take to make the
machine behave the way it was. I thought that the last thing in the world that could cause
such a problem would be tool qualification.. We had examined the hits taken for the
qualifications and the Standard Deviation was. 0001 inch or less. I was thinking to myself
that the only time I had seen a similar problem before was when a tool was loose..... or a
qualification sphere was loose.... I checked the tool and it was secure. I then touched the
qualification sphere and felt a slight movement. I dismounted the sphere and epoxied it in
place and re-ran the program As you can see by the charts, order has been restored. You
could not see the sphere move. The machine probes at a pressure of approximately 3-5
grams. I would not have expected the repeat on the range chart to be as good as it was
given the fact the sphere moved. Once again, measuring something periodically told me
something was wrong. The information did not tell me what was wrong. Experience
helped find the problem but the charts told me something was wrong before I would have
found it by measuring a part used for production. We have part tolerances approaching +/.0025 inch on the entire profile. We cannot be using part tolerance up in the qualification
process. To sum up what we have learned is that you simply must measure and chart
some features on a part that you know in order to be able to analyze and adjust machining
processes that produce the parts you measure.
CONCLUSION
These examples have shown that it is important to know the state of the measuring
equipment. It is critical to have a test part that has some of the characteristics that are of
importance to the production parts and that collecting and evaluating data on a daily basis
becomes procedure. There is no reason why control charts should not be used to establish
calibration intervals. The system of semi-annual or annual cah'brations can be replaced by
a program of data collection and evaluation that will allow CMM users to evaluate the
quality of their measurements being produced by their CMM and to correct any areas of
deficient performance before there is an adverse effect on product quality.

1998 NCSL Workshop & Symposium

678

Session 68

seTZ. '

IS

DEPT.

4~

PART NAME

SPHERE

PART rtO.
'OPERATIOft

HOrtE
G RlR

I'IACH UtE

.....
x-t!OORDiftATE

Grand Plean = -8.888873


l'Iean Range = 8.888238

VARIABLE
LCLxhar = -8.888248
UCLrange= 8.888524

UCLxbar =

8.888895

::= y ----------------------X BAR OIART

-e.~

UCLx

-8.fIII1
-8.888193
LCLx

8.888524
8.888393

8.88
HMHMMM~~~~~~~~~~~~~~~
M~~~~OO~5~~55~~M~~OO~~~

82/811"98

82/82/98

82/83/98

82/85/98

DEPT.

412ft

PART HArlE
PIAOt UtE
OPEMTIOfI
ITDI 1 X
VARIABLE
Grantl Plean = -8.888&18 LCLxbar = -8.888811 UCLxbar =
I1ean Range = 8.~
UCLrange= 8.888,29

PART ftO.

UIDGET

_
X COORDINATE

-8.888489

X BAR OIARTo
8 1588

-~-.......,z;;::::;;--""

8.1IJ1894
8.lIJI15Bl

8.1111J77
0)

co
o

~.~---------------------~.881J93

R alART
8.88243

8.881823
8.881215
UCLr

8.888687

-8.88

~~~~~~~r-"~~~~~~~~~~~~bE~~~~ "~

86 18 89 12 12 8B 89 86 15 86 86 86 85 86 86 89 13 86 86 86 87
.~~~~M85~5~~~~mm~~~~~~

83114198 83/16/98 831191'98 84/82/98 84/211'98 85/151'98 85/23198

DEPT.
41211
MRt NO.
OPEMTIOft
ITEtI 5 NlGLE
GraNl I'Iea.n = 4.488898 LCLxbar =
ne.n Range

= 8.828821

PART HAI'IE
U(DGET
I1AOtIPIE
.VARIABLE-r----4.467662 UCLxhar =

UCLrange= 8.863888

X BAR

awn

4.69919

4.652884
4.686579
4.568273
4.5139(,8 ___________ - _ _ _ _ _ _ _ _

4.467662

UCLx

~~~~~~~=r~~--~~er~~~~~~~~~~r- ~

La.x

R mART
8.871&7
8.&58253

8.43aJ35

8.Z19418
lJCLI'

8.88

Gm_ _~amaa~~~~__. .ma__~~ma__m&~~~. .ag=="~

~~~UU~~~5~~~~~~~n~~~~
.~$~~~~U5~~~%mm~~m~~~

83/14/98 83/16/98 83/19/98 84/82/98 81/21/98 851'151"98 85123198

18
_D.,.
41211
PART "","E
SPHERE
PM! 1'tO.
ftOHE
I'IAO-I UIE
_
, 0PIlVt! 1011
GAGE RaR
'JAR lADLE
X COORD IMTE
GNM I'Iean= -8.~42 LCLxbar = -tL8881B4 UCLxbar:: 8.888188
....n Ral18e :: 8.888195
UCLrange:: 8.888444
X BAR mART

8.818123
UCLx

LCLx

"
-ft.8B8280

R OIART

B.1BlB7
... 8.BIJII','53

-.,'

8.1IJItt435 - - - - - - - - - - - - - - - - - - - - - - - - 8.~17

8.88

--~

- - - -

t-

__--__~__~____~____-+~__~~_____~~~__~___

'-'
89 8& 85 85

ao

85 86 87 89 fT1 11 8G 86 frl 87 89 88 8G 8& fJ7 86

5~~~tl~~~585~~~~~~~m~.~

83119/98 83/31/98 84/1(11'98 85/14/98 85/22/98 8&/25198 81.11V98

0CLr
"~

4~

DEPT.
P,AItT NO.

OPIMTIOft

HOrIE
GAGE RAR

G.aNt I'Iean
-8.888842
PIPMt.. Range = 8.888195

PART fW1E
MOIlttE
VARIABLE
LCLxbar
-8.868184 UCLxbar
UCLrange= 8.888444

....

Ie..

SPHERE

X COORD IrtATE
8.888UII

X BAR mART

8 ....123
8.1II1II84

--r--------~--------------------

~.

1\

UCLx

~.MBHZ~~~\H.---------.------------------------------"~

-8.888124

.'. \/

-tt.l8I2!6 ~~~-----------------------------8.

LeLx

R OWIT

8.
8.
8.
8.
en
Cb
ti)
ti)

o
::)

UCLr

~~---------------------------------------------------- "~

8.88

...

861288898989898989
88 3S 48 88 88 88 88 88 88
8EV81.198 88187/98

sphere
pig
PART PIME
s/n 8897-2346
none
ftACHINE
X coordinate
Sphere NeasureNent
VARIABLE
Grand Plean = -11.888878 LCLxbar = -8.888445 UCLxbar = 8.888384
ftean Range = 8.888&49
UCLrange= 8.881378
X BAR atART
8.888384 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - DEPT.
PART NO.
OPERATION

UCLx

8.888142

r-__i -_______

-8.888821__~~~~~~~~~~__________________
-11.888183

"EAftx

-11.888346
LCLx
-11.888588
R

CHART

8.88558
8.884185
8.88279
UCLr

8.881395 - - - - - - - - - - - - - - - - - - - - - - - 8.88

..;;;;;~;;;;;;;;;;;;===~;;=:;=:~~t==1::~==

"EA"r

~~~~~~~~~~~~~~~~~~~~

18 18 18 18 18 18 18 18 18 46 46 46 46 46 46 46 46 59 18 18
89/82/97
89/13/97
18/86/97
DEPT.
PART NO.
OPERATIO"

sphere
PART fIAI'IE
8897-2346
"ACHINE
Sphere NeasureNent
UARIABLE
Y coord i nate
Grand Plean = 8.888819 LCLxbar = -11.888172 UCLxbar = 8.888211
"ean Range = 8.888332
UCLrange= 8.888788
X BAR CHART
8.888211
P i9
none

UCLx

8.888134
8.888858
-11.888819

~----~~--~------~~--------------------r---~~4----- ~x

-8.888896
-8.888172 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - R CHART
8.88287
8.882153

LCLx

DEPT.
p iO
PART ftME
PART PIC).
none
I'IAQflrtE
OPERATION
Sphere NeaSUreMent
VARIABLE
Grand ftean = 8.888811 LCLxbar = -8.888141 UCLxbar
"ean Ranoe = 8.888264
UCLrange= 8.888557
X BAR alART

8.888164

Sphere
8897-2346

Z cooriinate
8.888164

--------------------------------

8.888183
8.888M2 ~~~~~~==~~~~~
"t"
-8.888819

__

~~~~~

__ ___
~~~

UCLx

IUANx

-8.88888
-8.888141- - - - - - - - - - - - - - - - - - - - - - - - - _ - _ _ _ _ _

LCLx

R CHART

8.88251
8.881883
8.881255
8.888628
8.88

------------------------

UCLr

~~~~~~~~~~~~~~~~===i~~==mMr
88BB88I!J8I!J888BBI!J8888888BBI!J8BBBBBB88118686
81 81 81 81 81 81 81 81 81 82 82 82 82 82 82 82 82 88 19 18
89/82/97
89/13/97
18/86/97

DEPT.
p iO
PART rtA~
PART ftO.
none
MOl IftE
OPERATIOft
Sphere MeasureMent
VARIABLE
Grand Plean = 8.758826 LCLxbar = 8.749998 UCLxbar =
~an Range = 8.888864
UCLrange= 8.888134
X BAR CHART

Sphere
8897-2346

Di.....,ter
8.758863

8.7'58884

UCLx

8.758859
8.758834

__________________

~~------------------+_--~~L----

8.758888

LCLx
8.749983
8.749958

R CHART
8.888134 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 8._1
a -"'-7

A1\ l\ AJ\

r'\

/V\'V~
~ VV .. r"J~ rv V
8.888833
"

"J\

,., 0

UCLr

"EAftr

D.DDUDD

8.88
88 88 88 8B 88 88 88 8B 88 8B 88 88 8B I!J8 88 88 88 11 86 86
11 11 11 11 11 11 11 11 11 89 89 89 89 89 89 89 89 81 19 19
89/82/97
89/13/97
18/86/97

1998 NeSL Workshop & Symposium

685

Session 68

zc

WIDGET

00

000

1998 NeSL Workshop & Symposium

686

Session 68

PIVOT PLATE
PERSPECTIVE

THIS PAGE INTENTIONALLY LEFT BLANK

1998 NCSL Workshop & Symposium

688

Session 68

You might also like