4.6 Metrology, Calibration & Measurement Uncertainty

4.6 Metrology, Calibration & Measurement Uncertainty

Metrology is the science of measurement, encompassing both theoretical and practical aspects across all levels of precision. In quality engineering, robust metrology ensures that inspection decisions are valid, measurement equipment is reliable, and product compliance can be proved globally.

---

Traceability & The NIST Metrology Pyramid

Metrological Traceability (defined in ISO/IEC Guide 99) is the property of a measurement result whereby the result can be related to a reference through a documented, unbroken chain of calibrations, each contributing to the measurement uncertainty.

The NIST Traceability Pyramid

1. SI Base Units: The international standard definitions of measurement units based on fundamental physical constants (e.g., the meter defined by the speed of light in vacuum).
2. National Metrology Institutes (NMIs): Organizations such as NIST (United States), NPL (UK), or PTB (Germany) that realize and maintain primary physical standards.
3. Primary Standards: Highest metrological quality standards located in accredited primary laboratories; verified directly against NMI standards.
4. Secondary / Transfer Standards: Maintained in industrial calibration laboratories to calibrate working standards.
5. Working Standards: Calibrated reference blocks, pin gages, or standard weights used to calibrate plant floor inspection tools.
6. Working Instruments: Shop-floor measurement equipment (calipers, micrometers, CMMs, pressure transducers) used for daily production verification.

Test Uncertainty Ratio (TUR) & Test Accuracy Ratio (TAR)

To ensure calibration validity, the reference standard must be substantially more precise than the instrument being evaluated:

$$\text{TUR} = \frac{\text{Tolerance of Instrument Under Test}}{\text{Expanded Uncertainty of Calibration Standard}} \ge 4:1$$

• Target Rule: A minimum TUR of 4:1 (or TAR of 4:1) is mandated by standards such as ANSI/NCSL Z540.3; a 10:1 ratio is preferred wherever technically feasible.

---

Hierarchy of Calibration Standards

---

Calibration Intervals & Out-of-Calibration (OOC) Risk Protocols

Calibration Interval Determination Methods

Equipment calibration intervals should never be assigned arbitrarily. ISO/IEC 17025 recommends methods such as:

1. Fixed Calendar Time: 6, 12, or 24-month cycles (initial starting baseline).
2. Elapsed Usage / Operational Hours: Based on cycle counts or operating hours (ideal for heavy-use production tools).
3. Drift / Control Charting (A3 Method): Plotting calibration values over time to track parameter drift and extrapolate when the tool will exceed tolerances.
4. Staircase / Reactive Adjustment (A4 Method): Automatically extending intervals when tools remain within spec, or shortening intervals when out-of-spec conditions occur.

Out-of-Calibration (OOC) Action Plan

When a gage is found to be Out-of-Calibration during routine recertification:

[ Gage Found OOC ] --> [ Quarantine Instrument ]
                          |
                          v
[ Reverse Traceability Search: Identify All Lots Inspected Since Last Pass Calibration ]
                          |
                          v
[ Risk Assessment: Calculate Maximum Measurement Error vs Product Tolerance ]
                          |
                          v
[ Action: If Error Exceeds Tolerance Limit -> Issue Product Recall / Containment ]

---

Measurement Uncertainty Propagation & GUM Principles

According to the ISO/IEC Guide to the Expression of Uncertainty in Measurement (GUM), every measurement result consists of two parts: the estimated value $y$ and its associated uncertainty $u(y)$.

• Error: Measured Value minus True Value (a single known number once evaluated, or unknown constant).
• Uncertainty: A non-negative parameter characterizing the dispersion of values that could reasonably be attributed to the measured quantity.

Type A Evaluation of Uncertainty

Type A evaluation uses statistical analysis of series of independent, repeated observations under repeatability conditions.

• Standard uncertainty $u_A$ equals the Standard Error of the Mean: $$u_A = s_{\bar{x}} = \frac{s}{\sqrt{n}} = \sqrt{\frac{\sum_{i=1}^{n} (x_i - \bar{x})^2}{n(n-1)}}$$ where $s$ is the sample standard deviation and $n$ is the number of repeat measurements.

Type B Evaluation of Uncertainty

Type B evaluation uses non-statistical information, such as manufacturer specifications, calibration certificates, handbook values, or tolerance limits.

• Rectangular (Uniform) Distribution: Assumed when limits $\pm a$ are specified with no information on probability distribution within the range: $$u_B = \frac{a}{\sqrt{3}}$$
• Triangular Distribution: Assumed when values are concentrated near the midpoint of limits $\pm a$: $$u_B = \frac{a}{\sqrt{6}}$$
• Normal Distribution (from Calibration Certificate): When an expanded uncertainty $U_{cert}$ is given at a specified confidence level with coverage factor $k_{cert}$: $$u_B = \frac{U_{cert}}{k_{cert}}$$

Combined Standard Uncertainty ($u_c$)

For independent (uncorrelated) input uncertainty components $u_1, u_2, \dots, u_N$, the Combined Standard Uncertainty $u_c$ is calculated using the Root-Sum-of-Squares (RSS) law:

$$\nu_c = \sqrt{u_A^2 + u_{B1}^2 + u_{B2}^2 + \dots + u_{BN}^2}$$

If input variables have functional relationship $y = f(x_1, x_2, \dots, x_N)$, sensitivity coefficients $c_i = \frac{\partial f}{\partial x_i}$ scale each component: $u_c(y) = \sqrt{\sum_{i=1}^N c_i^2 u_i^2}$.

Expanded Uncertainty ($U$)

To define an interval with a high level of confidence (typically ~95%), the combined standard uncertainty is multiplied by a coverage factor $k$:

$$U = k \cdot u_c$$

• For a normal distribution, $k = 2$ provides approximately 95.45% coverage ($k = 1.96$ for exactly 95.00%).
• $k = 3$ provides 99.73% coverage.

---

Worked Numerical Example: Uncertainty Budget

Problem: A digital micrometer measures a shaft diameter. Ten ($n = 10$) repeat measurements yield a sample standard deviation $s = 0.0024\text{ mm}$. Additional uncertainty sources are:

1. Calibration Certificate: States standard error $U_{cert} = \pm 0.0010\text{ mm}$ with $k = 2$.
2. Resolution Limit: Micrometer digital resolution is $0.0010\text{ mm}$ (rectangular semi-range $a = 0.0005\text{ mm}$).

Calculate the Combined Standard Uncertainty ($u_c$) and Expanded Uncertainty ($U$) at $k = 2$.

Step 1: Calculate Type A Uncertainty ($u_A$)

$$\nu_A = \frac{s}{\sqrt{n}} = \frac{0.0024}{\sqrt{10}} = \frac{0.0024}{3.1623} = 0.000759\text{ mm}$$

Step 2: Calculate Type B Uncertainties

• Calibration Standard ($u_{B1}$): $$u_{B1} = \frac{U_{cert}}{k} = \frac{0.0010}{2} = 0.000500\text{ mm}$$
• Digital Resolution ($u_{B2}$ Rectangular): $$u_{B2} = \frac{a}{\sqrt{3}} = \frac{0.0005}{\sqrt{3}} = \frac{0.0005}{1.732} = 0.000289\text{ mm}$$

Step 3: Calculate Combined Standard Uncertainty ($u_c$)

$$\nu_c = \sqrt{u_A^2 + u_{B1}^2 + u_{B2}^2} = \sqrt{(0.000759)^2 + (0.000500)^2 + (0.000289)^2}$$ $$\nu_c = \sqrt{0.000000576 + 0.000000250 + 0.0000000835} = \sqrt{0.0000009095} = 0.000954\text{ mm}$$

Step 4: Calculate Expanded Uncertainty ($U$ for $k = 2$)

$$U = k \cdot u_c = 2 \cdot 0.000954\text{ mm} = 0.00191\text{ mm} \quad (\pm 1.91 \; \mu\text{m})$$