A manager's guide to gauge R&R.

Perhaps the most important element of your quality program is poorly done, or not done at all. Measurement system analysis (MSA), a.k.a, gauge R&R, could be the nail that lost the shoe, that lost the horse, that lost the rider, that lost the battle, that lost the war.

A rush to new gimmicks has always been the hallmark of the quality business. Of course, the quality manager must keep up with the current ideas and tap dance with both the boss and the customer when necessary.

Because of the mad rush to play the ISO and QS games, many worthwhile techniques (required by these very specs.), have fallen by the wayside, or missed being discovered in the first place. Hand wringing over document format and nitpicking examiners (who must find something to report), have replaced many of the basic tools of quality management, like statistical process control (SPC), MSA and others.

Tools for defect prevention

There are many worthwhile tools and techniques available to the quality manager. Which are important? The whole idea of quality management is preventing defects. SPC is important. Measurement system analysis (MSA) is even more important. There are others, but MSA and SPC can actually prevent defects. If you prevent defects, you don't have to explain them, document them, identify them, meet over them or ship them.

How good is your SPC?

SPC is one of the most powerful tools in the quality manager's kit. The greatest SPC sin of all, worse than lack of training and misguided techniques, is the lack of MSA. Without proper MSA, control charts and Cpks don't mean a thing.

Measurement system analysis (MSA)

As a manager, it's not critical that you understand the math. But, like the math in SPC, don't fiddle around with it to make things come out your way.

There's R, there's R, and there's R&R

And, of course, there's MSA. Think of gauge R&R as a part of MSA, because there are other techniques that produce values for linearity and bias and such. Your measurement system consists of the people and gauges used at all levels to accept or reject your product or control a process. If you don't know the capability (gauge R&R) of your measurement system, you don't know if your measurements, or your products, are good or bad.

Language problems

The quality business has many problems caused by the language. MSA has a language problem too. We call the results of a study R&R. But it's really the reverse of R&R. The gauge R&R value is the measurement system error. This error applies across the range of measurements, and could become more important near the upper and lower specification. For example, if you measure a part and find it exactly on the nominal value, the measurement error might not make any difference. The part will be accepted. However, if you measure a part and the reading is close to the specification limits, the measurement error is critical.

The gauge R&R study

Usually, 5-10 parts are measured two or three times (trials) by two or three appraisers in a random order. Random means that part number one for operator one is the same as part number one for operator two, etc., but the order in which they are measured is changed. Randomness is critical to a study. The minimum study is two appraisers and two trials.

The first R - repeatability

Repeatability is the variation in measurements obtained with one measurement instrument when used several times by one appraiser while measuring the same characteristic on the same part (this is from the Automotive Industry Action Group, MSA Handbook).

A simple calculation produces a number known as EV, or equipment variation. This number is expressed in the same unit of measurement as the individual parts used in the study. Think of repeatability as the capability of the gauge itself. You wouldn't measure in thousandths with a yardstick.

The second R - reproducibility 

Reproducibility is the variation in the average of the measurements obtained by different appraisers using the same measuring instrument when measuring the identical characteristic on the same part (also from the AIAG, MSA Handbook).

Think of reproducibility as appraiser variation, which becomes AV when doing the math. People can vary greatly in the use of the same gauge. The final report and charts from a gauge study can show which employees need training, which are okay and which need a new career. AV is also expressed in the unit of measurement of the parts measured.

That leaves R&R

What's it all mean? You do a study, you have the R (EV) and the R (AV), so what's left? R&R, of course! To use these numbers, R and R must be combined mathematmally to find R&R. To make use of R&R, you must relate it to the specifications of the product you are measuring and consider the process that produces the product.

With a little square root magic, the EV and the AV are combined to produce the winning number -- the gauge R&R! R&R is also expressed in the same unit of measurement as the parts being measured, but can also be expressed as a percentage of the tolerance, the process or the total variation.

There's more

Why are you doing all this? Well, in the fast place, it would be nice to know if your measurement system is capable of making correct decisions about whether to accept or reject a part, or adjust a process. Parts are going to vary, gauges are going to vary, people are going to vary, and the big question comes down to, how much? What kind of numbers could alert you to the probability of shipping bad parts or tossing out good ones?

Putting it to use

Let's look at the R&R numbers. The following is a made-up example. Look at the MSA error test chart. The base line is a scale to show the range of the specification. The [+ or -] 3 vertical lines show the upper and lower specification, LSL and USL. The vertical lines between the LSL. and USL. is a scale of the specifications; in this case 0 [+ or -] 3. The scale extends to [+ or -] 4.

The MSA error test

Take a coin from your pocket or purse. The diameter of the coin represents the gauge R&R -- the measurement error. Sliding the coin along the scale will give you some idea of the effect of measurement error on your product acceptance decisions, and demonstrate your risk of not knowing the capability of your measurement system.

Place the coin on the base line with the widest part splitting the nominal 0.0 value and the base line. The measurement error has little effect here; if a part is within its specification, it will be accepted. Now, move the coin to a position between +1 and +2. The error is there, but does it have an effect? Now place the center of the coin on the +3 USL line. The R&R has a big effect here since measurements could be beyond the +3 limit and you wouldn't know it. Try this with a different size coin. The error range changes.

How good is good?

Many people have no more idea of the meaning of R&R than they do the meaning of Cpk. The current popular value seems to be 0.20 or 20% of the specification. New thinking says that the specification may not be enough! If you're using the gauge for SPC, the process range enters into it too. A better process means a better measurement system.

The AIAG MSA Handbook offers these guidelines for acceptance of gauge R&R as a percentage:

* Under 10% error -- the measurement system is acceptable;

* 10% to 30% error -- may be acceptable based upon importance of application, cost of gauge, cost of repairs, etc.

* Over 30% error - needs improvement. Gauge R&R is not as simple as it appears. Just like SPC, tests must be applied before you can believe the numbers.

What's it all mean?

For one thing, it means that if you don't do gauge R&R studies, every measurement you make is suspect. Control charts are full of unknowns. You don't know if you are rejecting good products and accepting bad.

The rules

There are five rules set down in the AIAG Handbook:

* Determine whether the range chart is in control. Any special causes must be corrected. I assume that you have a range chart.

* Check the range chart for adequate discrimination. There should be at least four possible values for the range within the control limits.

* The apparent resolution should be at most one-tenth of total process six sigma (+/-3), for adequate discrimination.

* One-half or more of the averages should fall outside the average chart control limits for the measurement system to be adequate to detect part-to-part variation and to provide useful data for controlling the process. This means that you must have an average chart. When less than half fall-outside the control limits then the measurement system is inadequate to detect part-to-part variations and cannot be used for process control.

* The number of data categories must be at least five, and preferably more, for the measurement system to be acceptable for the analysis of the process.

A final word

The computer has advanced MSA far beyond the simple "range method" study. QS-9000 requires the use of the AIAG MSA Handbook. The handbook requires graphics and added techniques. These include within part variation, ANOVA, linearity, gauge performance and studies for attribute gauges.

I suggest that you order the AIAG/ASQ "MSA Handbook" and wade through the details. It's hard in spots. Call AIAG at 810/358-3003.

David C. Crosby started his quality career as a tester and has worked about every job in the quality profession. He advanced to Corporate Director of Quality for RCA and held similar positions with General Instruments, Interpace and Portec. He formed The Crosby Co. in 1981.