msa（多发性硬化症）

Measurement System Analysis (MSA)

When to Activate This Skill

• "Conduct gage R&R study"
• "Evaluate measurement system for [gage/characteristic]"
• "Calculate %GR&R"
• "Perform attribute agreement analysis"
• "What's the ndc for this gage?"
• "Is this measurement system acceptable?"
• "MSA requirements for [characteristic]"

Purpose of MSA

MSA determines how much of the observed process variation is due to the measurement system rather than the actual process. Before making decisions based on measurement data, we must verify the measurement system is adequate.

Why MSA Matters

Without MSA:

• "Good" parts may be rejected
• "Bad" parts may be accepted
• Process capability may be understated
• SPC decisions may be wrong
• Customer complaints may result

With MSA:

• Measurement confidence established
• Gage selection validated
• Training effectiveness verified
• Calibration adequacy confirmed

Types of MSA Studies

Variable MSA (Gage R&R)

For measurements that produce numerical data (dimensions, weight, temperature, etc.)

Attribute MSA

For measurements that produce pass/fail, good/bad, or categorical results

Variable Gage R&R Study

Study Design (Standard AIAG)

Study Execution

1. Select parts - Cover full range of process variation
2. Number parts - Hidden from operator view
3. Randomize - Operator doesn't know which part
4. Measure - Each operator measures all parts, multiple trials
5. Record - Document all measurements
6. Analyze - Calculate %GR&R, ndc

Acceptance Criteria

Interpretation

%GR&R <10%:

• Measurement system acceptable
• Can distinguish part-to-part variation
• Suitable for SPC

%GR&R 10-30%:

• May be acceptable for non-critical applications
• Requires customer approval for critical characteristics
• Consider improvement actions

%GR&R >30%:

• Measurement system not acceptable
• Must improve before use
• Consider: different gage, training, environment

ndc (Number of Distinct Categories):

• Represents the number of non-overlapping confidence intervals that span the product variation
• ndc ≥5 required for variable data
• ndc <5 means the measurement system cannot adequately distinguish part-to-part variation (acts more like attribute good/bad only)

Gage R&R Calculations

ANOVA Method (Preferred)

Analysis of Variance separates total variation into:

• Part-to-part variation
• Operator variation
• Operator × Part interaction
• Repeatability (equipment)
• Reproducibility (operator)

Range Method (X-bar/R)

Simpler calculation, widely used:

Repeatability (EV) = R̄ × K₁
Where: R̄ = average range across all operators
       K₁ = factor based on number of trials

Reproducibility (AV) = √[(X̄diff × K₂)² - (EV²/nr)]
Where: X̄diff = range of operator averages
       K₂ = factor based on number of operators
       n = number of parts
       r = number of trials

GR&R = √(EV² + AV²)

%GR&R = (GR&R / TV) × 100
Where: TV = Total Variation = √(GR&R² + PV²)
       PV = Part Variation

ndc = 1.41 × (PV / GR&R)

Attribute Agreement Analysis

When to Use

• Go/No-go gages
• Visual inspection
• Pass/fail tests
• Any categorical decision

Study Design

Key Metrics

Kappa Interpretation

Other MSA Studies

Bias Study

Measures systematic error (difference from true value)

Method:

1. Obtain reference standard (known true value)
2. Measure standard multiple times (≥10)
3. Calculate average of measurements
4. Bias = Average - Reference value

Acceptance: Bias ≈ 0 or within calibration tolerance

Linearity Study

Measures bias across the measurement range

Method:

1. Select 5+ reference standards across range
2. Measure each standard multiple times
3. Plot bias vs. reference value
4. Fit regression line

Acceptance: Linearity (slope × Process Variation) <5%

Stability Study

Measures variation over time

Method:

1. Select stable reference part/standard
2. Measure periodically (daily, weekly)
3. Plot on control chart
4. Monitor for trends or out-of-control

Acceptance: Stable control chart, no trends

MSA Requirements by Application

IATF 16949 Requirements (7.1.5.1.1)

• MSA required for all measurement systems in Control Plan
• Shall include study guidance and acceptance criteria
• Alternative methods may be used with customer approval

Application Guidelines

Common MSA Issues and Solutions

Output Format

When generating MSA content:

# MSA Study Report

## Study Information
| Field | Value |
|-------|-------|
| **Study Type** | Gage R&R / Attribute Agreement |
| **Gage ID** | [ID] |
| **Gage Description** | [Type, range, resolution] |
| **Characteristic** | [What is measured] |
| **Specification** | [Tolerance] |
| **Study Date** | [Date] |
| **Conducted By** | [Name] |

## Study Parameters
| Parameter | Value |
|-----------|-------|
| Operators | [Number and names] |
| Parts | [Number] |
| Trials | [Number] |
| Total measurements | [Count] |

## Results
| Metric | Value | Acceptance | Status |
|--------|-------|------------|--------|
| %GR&R | [X]% | <10% / <30% | PASS/FAIL |
| ndc | [X] | ≥5 | PASS/FAIL |
| Repeatability | [X]% | - | - |
| Reproducibility | [X]% | - | - |

## Conclusion
[ACCEPTABLE / MARGINAL / UNACCEPTABLE]

## Actions (if required)
- [Action items]

Integration with Related Skills

ControlPlan

All gages in Control Plan require MSA:

• Variable gages: Gage R&R
• Attribute gages: Attribute agreement
• MSA status verified before production

Load: read ~/.claude/skills/Controlplan/SKILL.md

SPC

SPC validity depends on MSA:

• ndc ≥5 required for variable SPC
• Measurement variation affects control limits
• Poor MSA = poor SPC decisions

Load: read ~/.claude/skills/Spc/SKILL.md

AutomotiveManufacturing

MSA supports work instruction development:

• Measurement methods documented
• Gage identification required
• Operator training verified

Load: read ~/.claude/skills/Automotivemanufacturing/SKILL.md

Supplementary Resources

For detailed guidance: read ~/.claude/skills/Msa/CLAUDE.md

For study templates: ls ~/.claude/skills/Msa/templates/

For acceptance criteria: read ~/.claude/skills/Msa/reference/acceptance-criteria.md

For calculation formulas: read ~/.claude/skills/Msa/reference/calculation-formulas.md