# Measurement System Analysis - Gage R&R

## Gage R&R Compatible with AIAG MSA 4th Edition

**Measurement System Analysis **(MSA) involves **Gage R&R** (repeatability and reproducibility) studies to evaluate your measurement systems.

When I first got involved with quality, I learned about the "five M's" that constituted most root causes: man, machine, materials, methods, and measurement.

Because I worked in a predominantly service industry, I couldn't quite grasp how measurement could be a common cause of variation. But, if you work in manufacturing, you know that **gages** and how they are used can be a key cause of variation.

**Measurement Systems Analysis (MSA**)

*MSA* is actually quite simple, but even seasoned SPC veterans don't seem to understand it. So I thought I'd simplify it for you.

**First,** *Gage R&R* studies are usually performed on variable data - height, length, width, diameter, weight, viscosity, etc.

**Second,** when you manufacture products, you want to monitor the output of your machines to make sure that they are producing products that meet the customer's specifications. This means that you have to measure samples coming off the line to determine if they are meeting your customer's requirements.

**Third,** when you measure, three factors come into play:

- Part variation (differences between individual pieces manufactured)
- Appraiser variation (aka, reproducibility) -

Can two different people get the same measurement using the same*gage*? - Equipment variation (aka, repeatability) -

Can the same person get the same measurement using the same*gage*on the same part in two or more trials?

You want most of the variation to be between the parts, and less than 10% of the variation to be caused by the appraisers and equipment. Makes sense, doesn't it? If the appraiser can't get the same measurement twice, or two appraisers can't get the same measurement, then your measurement system becomes a key source of error.

- MSA Gage R&R Study Quick Reference Card (PDF, 204 KB)

**Conducting a Gage R&R Study**

#### To conduct a Gage R&R study, you will need:

- five to ten parts (# each part) that span the distance between the upper and lower spec limits. The parts should represent the actual or expected range of process variation. Rule of thumb: if you're measuring to 0.0001, the range of parts should be 10 times the resolution (e.g., 0.4995 to 0.5005).
- two appraisers (people who measure the parts)
- one measurement tool or
*gage* - and a minimum of two measurement trials, on each part, by each appraiser
- a
*Gage R&R*tool like the**Gage R&R excel template**in the QI Macros.

### QI Macros for Excel Gage R&R Template

Here are samples of the **Gage R&R template **input sheet and results sections using sample data from the AIAG Measurement Systems Analysis Third Edition.

**Gage R&R System Acceptability**

- % R&R<10% -
*Gage*System Okay

(Most variation caused by parts, not people or equipment) - % R&R<30% - May be acceptable based on importance of application and cost of
*gage*or repair - % R&R>30% -
*Gage*system needs improvement

(People and equipment cause over 1/3 of variation)

**What To Look For**

Repeatability: Percent Equipment Variation

(%EV - Can the same person using the same *gage* measure the same thing consistently)

If you simply look at the measurements, can each appraiser get the same result on the same part consistently, or is there too much variation?

__Example__ (looking at measurements from one appraiser only):

- No Equipment Variation: (Part 1: 0.65, 0.65; Part 2: 0.66, 0.66)

- Equipment Variation: (Part 1: 0.65, 0.67; Part 2: 0.67, 0.65)

If repeatability (Equipment variation) is larger than reproducibility (appraiser variation), reasons include:

*Gage*needs maintenance (*gages*can get corroded)*Gage*needs to be redesigned to be used more accurately

- Clamping of the part or
*gage*, or where it's measured needs to be improved (imagine measuring a baseball bat at various places along the tapered contour; you'll get different results.)

- Excessive within-part variation (Imagine a steel rod that's bigger at one end than the other. If you measure different ends each time, you'll get widely varying results.)

Reproducibility: Percent Appraiser Variation

(% AV - can two appraisers measure the same thing and get the same answer?)

__Example__ (looking at *measurements* of the same part by two appraisers):

- No Appraiser Variation: (Appraiser 1, Part 1: 0.65, 0.65; Appraiser 2, Part 1: 0.65, 0.65)
- Appraiser Variation: (Appraiser 1, Part 1: 0.65, 0.65; Appraiser 2, Part 1: 0.66, 0.66)

If you look at the line graph of appraiser performance, you'll be able to tell if one person over reads or under reads the measurement.

If reproducibility (appraiser variation) is larger than repeatability (equipment variation), reasons include:

- Operators need to be better trained in a consistent method for using and reading the
*gage*

- Calibrations on
*gage*are unclear

- Fixture required to help the operator use
*gage*more consistently

**Mistakes People Make**

Many people call us because they don't like the answer they get using the **Gage R&R** template. Most of the time, it's because they didn't follow the instructions for conducting the study. Here are some of the common mistakes I've seen:

- Forgetting that the Gage R&R study is evaluating their measurement system and NOT their products. Gage R&R does not care about how good your products are. It only cares about how good you measure your products.
- Using only one part. If you only use one part, THERE CAN'T BE ANY PART VARIATION, so people and equipment are the ONLY source of variation.
- Using the one part measurement for all 10 parts (again, there won't be any part variation, so it all falls on the people and equipment).
- Using too many trials (if you use five trials, you have more opportunity for equipment variation).
- Using too many appraisers (if you use all three, you have more opportunity for appraiser variation).
- Using fake data. Try using the AIAG SPC data the QI Macros loads on your computer at c:\qimacros\testdata.
- Using a gage that measures in too much detail. If your part is 74mm +/- 0.05, then you don’t need a gage that measures to a thousandth of an inch (0.001) you only need one that measures to the hundredth of an inch (0.01).

**Challenges You Will Face**

One customer faced an unusual challenge: they were producing parts so precisely that there was little or no part variation even when measured down to 1/10,000th of an inch. Their existing *gage*s ceased to detect any variation from part to part.

As your process improves and your product approaches the ideal target * measurement*, you'll have less part variation and more chance for your equipment or people to become the major source of variation. As your product and your process improve, your measurement system will need to improve as well.

**Conclusion**

Your goal is to minimize the amount of variation and error introduced by measurement, so that you can focus on part variation. This, of course, leads you back into the other root causes of variation: process, machines, and materials.

If you manufacture anything, measurement system analysis can help you improve the quality of your products, get more business from big customers, and baffle your competition. Enjoy.

### Gage R&R Video Series

### Gage R&R Training Resources

- Gage R&R Template for Excel
- Gage R&R Bias and Linearity
- Gage R&R Destructive Testing
- Gage R&R Type 1 Study

- MSA Gage R&R Study Quick Reference Card (PDF, 204 KB)