Check Standards and Statistical Process Control

Metrology Subject Focus: Calibration laboratories must maintain ongoing statistical control of their measurement processes through Measurement Assurance Programs (MAP), check standards, interlaboratory proficiency testing, and Gage R&R capability studies.

In high-reliability calibration and manufacturing environments, periodic artifact calibration alone cannot guarantee measurement integrity. An instrument calibrated in January may suffer an undetected component failure, physical shock, or excessive drift in February, remaining in service and producing defective calibrations until its next scheduled verification twelve months later.

A Measurement Assurance Program (MAP) monitors the measurement process through suitable check standards, observations, and statistical controls. It can reveal changes between periodic calibrations and support uncertainty evaluation. It cannot detect every failure instantly; sampling frequency, shared influences, and check-standard stability matter.

---

The Metrological Check Standard

The central pillar of any MAP is the check standard (sometimes termed a surveillance standard or control artifact).

Check Standard Criteria and Operational Rules

1. Functional Independence: Choose a check that can detect the effects of concern. Evaluate shared influences and the check standard’s own drift. Measure it as an unknown using representative procedures, fixtures, and personnel; physical separation alone does not establish independent evidence.
2. Long-Term Stability: The check standard artifact must possess proven physical and temporal stability. Typical check standards include precision Zener voltage references, standard resistors in temperature-stabilized baths, Class 1 stainless steel mass standards, Grade 0 ceramic gauge blocks, and resonant quartz pressure transducers.
3. Frequent Surveillance Cycles: The check standard is calibrated at regular, predefined operational intervals (e.g., daily, at the start of each shift, or immediately before and after calibrating high-value customer standards). Each measurement result is logged in real time.
4. Process indicator: A check standard samples the effects represented by its quantity, range, method, and observation schedule. Its results can reveal environmental, operator, comparator, or reference changes, but cannot prove stability of every range or detect every intermittent failure.

---

Statistical Process Control (SPC) in the Calibration Laboratory

Data collected from check standard measurements are plotted on Shewhart control charts to evaluate whether the measurement system is in a state of statistical control.

$\bar{X}$ and $R$ Control Charts (Subgrouped Data)

When several replicate measurements of the check standard ($n = 3\text{ to }5$) are taken during each surveillance interval, an $\bar{X}$ and $R$ chart is employed:

• $\bar{X}$ Chart (Mean): Tracks the average value across subgroups. Monitors process central tendency, detecting systematic calibration offsets, reference standard drift, and ambient temperature shifts.
• $R$ Chart (Range): Tracks the dispersion within subgroups ($R = X_{\text{max}} - X_{\text{min}}$). Monitors process repeatability, detecting electrical noise, thermal turbulence, comparator looseness, or operator variability.

Individuals and Moving Range ($I\text{-}MR$) Charts

In high-echelon calibration laboratories, performing multiple replicate measurements on a check standard every day is often economically or technically impractical ($n = 1$). Under these conditions, the Individuals ($I$) and Moving Range ($MR$) chart is standard:

• Individual Measurement ($X_i$): The raw reading of the check standard recorded on day $i$.
• Moving Range ($MR_i$): The absolute difference between successive individual readings:

$$MR_i = |X_i - X_{i-1}|$$

Calculating $I\text{-}MR$ Control Limits

Given $k$ individual observations:

1. Calculate the Process Centerline (Grand Mean):

$$\bar{X} = \frac{1}{k}\sum_{i=1}^k X_i$$

1. Calculate the Average Moving Range:

$$\overline{MR} = \frac{1}{k-1}\sum_{i=2}^k MR_i$$

1. Estimate the Process Standard Deviation ($\hat{\sigma}$) using the unbiasing constant $d_2 = 1.128$ (for sample size $n=2$):

$$\hat{\sigma} = \frac{\overline{MR}}{d_2} = \frac{\overline{MR}}{1.128}$$

1. Establish Individuals ($I$) Control Limits:

$$\text{Upper Control Limit (UCL)}_I = \bar{X} + 3\hat{\sigma} = \bar{X} + 3\left(\frac{\overline{MR}}{1.128}\right) = \bar{X} + 2.66\,\overline{MR}$$ $$\text{Lower Control Limit (LCL)}_I = \bar{X} - 3\hat{\sigma} = \bar{X} - 3\left(\frac{\overline{MR}}{1.128}\right) = \bar{X} - 2.66\,\overline{MR}$$

1. Establish Moving Range ($MR$) Control Limits (where $D_4 = 3.267$ and $D_3 = 0$ for $n=2$):

$$\text{UCL}_{MR} = D_4 \cdot \overline{MR} = 3.267\,\overline{MR}$$ $$\text{LCL}_{MR} = D_3 \cdot \overline{MR} = 0$$

Control Limits vs. Specification Limits

---

Control-chart signal rules for Out-of-Control Conditions

Control charts are divided into three statistical zones on each side of the centerline:

• Zone C: Centerline to $\pm 1\sigma$ ($68.27\%$ of normal distribution)
• Zone B: $\pm 1\sigma$ to $\pm 2\sigma$ ($95.45\%$ cumulative)
• Zone A: $\pm 2\sigma$ to $\pm 3\sigma$ ($99.73\%$ cumulative)

The four classic Western Electric rules below identify statistical signals for investigation under the laboratory’s monitoring plan. The trend and stratification rules are additional commonly used Nelson-style rules. Select rules and response criteria deliberately; applying more rules changes the false-alarm rate, and a signal is not proof of a particular physical cause.

1. Rule 1 (Action Limit): A single point falls beyond Zone A ($> 3\sigma$ from the centerline). The measurement process is immediately stopped.
2. Rule 2 (Warning Limit): Two out of three consecutive points fall in Zone A or beyond on the same side of the centerline.
3. Rule 3: Four out of five consecutive points fall in Zone B or beyond on the same side of the centerline.
4. Rule 4 (Run of Eight): Eight consecutive points fall on the same side of the centerline. Indicates a systematic shift or uncorrected bias in the reference standard or instrumentation.
5. Rule 5 (Trend): Six consecutive points continuously increase or continuously decrease. Indicates steady instrument drift, component aging, or thermal gradient change.
6. Rule 6 (Stratification): Fifteen consecutive points fall in Zone C (within $\pm 1\sigma$ of the centerline). Indicates artificially suppressed variance, such as incorrect limit calculations or data manipulation.