Scenario
The auditor asks: "Do you perform MSA?" The quality engineer presents a Gage R&R report. The auditor replies: "This is just Gage R&R. Have you performed Bias? What about Linearity? How do you confirm the Stability of the measurement system?"
Many factories believe that MSA is complete once Gage R&R is done, but a complete MSA is an integration of four types of studies.
The Four Studies of MSA
1. Gage R&R (Repeatability and Reproducibility)
Answers: How large is the measurement error? The variation when different people measure the same item, and the variation when the same person measures the same item multiple times.
Method: 2-3 appraisers, each measuring 10 parts 2-3 times.
Acceptance Criteria: %GRR < 10% Excellent, < 30% Acceptable, NDC ≥ 5.
Most commonly performed; most factories only do this.
2. Bias
Answers: Does the measurement device systematically read high or low?
Method:
- Select a reference standard with a calibration certificate.
- The same operator measures 15-25 times.
- Calculate the mean, and perform a t-test to determine if it is significantly different from the standard value.
Acceptance Criteria: t-test P > 0.05, indicating that the bias is not significant (acceptable).
When to Perform: When new measurement equipment is introduced, after repair, or when a zero-point shift of the measurement device is suspected.
3. Linearity
Answers: Is the bias of the measurement device consistent across its entire measurement range?
Method:
- Select 5 reference standards uniformly distributed across the measurement range.
- Measure each 12 times.
- Perform regression with the measured value as X and the bias as Y.
- Test if the regression slope is 0.
Acceptance Criteria: Slope P > 0.05 = acceptable linearity.
When to Perform: When the measurement range is wide, or when the measurement device is suspected to be inaccurate in certain ranges.
4. Stability
Answers: Does the performance of the measurement device change over time?
Method:
- Select 1-2 reference standards.
- Measure 3-5 times at fixed intervals (daily/weekly/monthly).
- Plot an I-MR control chart to monitor the trends of the mean and variation.
Acceptance Criteria: Control chart shows no out-of-control points or trends = stable.
Note: This is long-term monitoring, not a one-time study.
Integrated Decision Matrix for the Four Studies
| Measurement Device Status | What Needs to be Done |
|---|---|
| New Measurement Device Introduction | Bias + Linearity + Gage R&R |
| Regular Calibration | Bias + Stability Control Chart |
| Customer Requires Full MSA | All four studies |
| After Repair | Bias + Gage R&R |
| Annual Audit | At least Gage R&R + Bias |
MSA Execution Sequence
Correct Sequence:
- Perform Bias first: Confirm the accuracy of the measurement device's zero point.
- Then perform Linearity: Confirm accuracy across the entire range.
- Then perform Gage R&R: Confirm variation due to operator interaction and the measurement device itself.
- Continuously perform Stability: Long-term monitoring of measurement device performance drift.
If Bias fails (the measurement device has a systematic deviation), the Gage R&R results will also be unreliable—the bias will be included in "reproducibility," making the problem appear more severe than it actually is.
Measurement System Improvement Priority
| Problem | Solution |
|---|---|
| Bias Fails | Recalibrate (zero-point/span adjustment) |
| Linearity Fails | Send for repair or replace the measurement device; do not use beyond its designed range |
| Gage R&R > 30% | Improve operating methods, replace the measurement device, or add fixtures to standardize measurement conditions |
| Stability Shows a Trend | Shorten calibration cycle, inspect for measurement device wear |
AIAG MSA Manual
The AIAG (Automotive Industry Action Group) MSA reference manual is an industry standard that defines the execution methods and acceptance criteria for all studies. If your client is an automotive manufacturer, they typically directly require adherence to AIAG MSA Fourth Edition.
Golden Quote
"A measurement system is like eyes—if your eyes aren't accurate, the process data you see is an illusion. The four MSA studies are like an eye exam for your eyes; neglecting any one is like working with one eye closed."