# Measurement System Analysis (MSA) and Gage RR Overview

## Measurement System Analysis (MSA) and Gage R&R Overview

### Purpose

To determine if a measurement system can generate accurate data, and if the accuracy is adequate to achieve your objectives

### Why use MSA

• To make sure that the differences in the data are due to actual differences in what is being measured and not to variation in measurement methods

•  Note Experience shows that 30% to 50% of measurement systems are not capable of accurately or precisely measuring the desired metric

### Types of MSA

• Gage R&R ( next page)

• Bias Analysis ( see p. 95)

• Stability Analysis ( see p. 97)

• Discrimination Analysis ( see p. 99)

• Kappa Analysis ( see p.100)

### Components of measurement error

Measurements need to be "precise" and "accurate." Accuracy and precision are different, independent properties:

• Data may be accurate (reflect the true values of the property) but not precise (measurement units do not have enough discriminatory power)

• Vice versa, data can be precise yet inaccurate (they are precisely measuring something that does not reflect the true values)

• Sometimes data can be neither accurate nor precise

• Obviously, the goal is to have data that are both precise and accurate

From a statistical viewpoint, there are four desirable characteristics that relate to precision and accuracy of continuous data:

1. No systematic differences between the measurement values we get and the "true value" (lack of bias, see p. 95)

2. The ability to get the same result if we take the same measurement repeatedly or if different people take the same measurement ( Gage R&R, see p. 87)

3. The ability of the system to produce the same results in the future that it did in the past ( stability, see p. 97)

4. The ability of the system to detect meaningful differences (good discrimination, see p. 99)

(Another desirable characteristic, linearity —the ability to get consistent results from measurement devices and procedures across a wide range of uses—is not as often an issue and is not covered in this book.)

 Note Having uncalibrated measurement devices can affect all of these factors. Calibration is not covered in this book since it varies considerably depending on the device. Be sure to follow established procedures to calibrate any devices used in data collection.

## Gage R&R: Collecting the data

### Highlights

Gage R&R involves evaluating the reliability and repeatability of a measurement system.

• Repeatability refers to the inherent variability of the measurement system. It is the variation that occurs when successive measurements are made under the same conditions:

• Same person

• Same thing being measured

• Same characteristic

• Same instrument

• Same set-up

• Same environmental conditions

• Reproducibility is the variation in the average of the measurements made by different operators using the same measuring instrument and technique when measuring the identical characteristic on the same part or same process.

• Different person

• Same part

• Same characteristic

• Same instrument

• Same setup

• Same environmental conditions

### To use Gage R&R …

1. Identify the elements of your measurement system (equipment, operators or data collectors, parts /materials/process, and other factors).

• Check that any measuring instruments have a discrimination that is equal to or less than 1/10 of the expected process variation/specification range

2. Select the items to include in the Gage R&R test. Be sure to represent the entire range of process variation. (Good and Bad over the entire specification plus slightly out of spec on both the high and low sides).

3. Select 2 or 3 operators to participate in the study.

4. Identify 5 to 10 items to be measured.

• Make sure the items are marked for ease of data collection, but remain "blind" (unidentifiable) to the operators

5. Have each operator measure each item 2 to 3 times in random sequence.

6. Gather data and analyze. See pp. 90 to 95 for interpretation of typical plots generated by statistical software.

 Tips In manufacturing you may want to start with one of the Automotive Industry Action Group ( see http://www.AIAG.org) standards … short form: 2 operators measuring 5 items 2 times (= 20 measurements total) long form: 3 operators measuring 10 items 3 times (= 90 measurements total) Be there for the study—NOT as a participant, but as an observer. Watch for unplanned influences. Randomize the items continuously during the study to prevent operator bias from influencing the test. When checking a given measurement system for the first time, let the process run as it normally would (no pre-training, no adjustment of equipment or instruments, no special items).