Overview
ISO 7870-6:2024 specifies EWMA control charts (Exponentially Weighted Moving Average) for monitoring the process mean as part of statistical process control (SPC). The second edition replaces ISO 7870-6:2016 and documents the EWMA technique originally proposed by Roberts (1959). The standard explains how to construct EWMA charts that give greater weight to recent observations, improving detection speed for small to moderate shifts in the process mean while remaining applicable to both variables and attributes data.
Key topics and technical requirements
- Definition and formula for EWMA: z_i = λ x_i + (1 − λ) z_{i−1}, with guidance on the starting value z_0 (typically the target μ_0).
- Smoothing parameter (λ): selection guidance and impact on sensitivity (example weights given for λ = 0.25).
- Control limits and construction: procedures for calculating control limits (factor L, σ considerations) and plotting z-values; includes construction examples.
- Performance metrics: discussion of run length, Average Run Length (ARL), Maximum Run Length (MAXRL), and how ARL varies with shifts in the mean.
- Parameter selection: recommendations for choosing λ, L and subgroup size n (including n = 1 cases).
- Comparisons with other SPC charts: when to prefer EWMA vs Shewhart and CUSUM charts (advantages for detecting small shifts, combination strategies).
- Implementation procedure: step-by-step guidance for deploying EWMA charts in practice.
- Robustness: sensitivity of EWMA to non-normality and limitations.
- Examples and annexes: Annex A (application examples), Annex B (EWMA for proportion nonconforming), Annex C (EWMA for count of nonconformities).
- Reproducible calculations: tables and figures computed with the R-package SPC (Crowder algorithm); R-file available via the ISO standards site.
Applications and who should use it
EWMA control charts are particularly valuable when:
- production rate is slow or subgroup sizes are small (n = 1),
- detecting small or moderate shifts in the process mean is critical,
- sampling/inspection is time-consuming or expensive,
- testing involves safety or high risk.
Typical users:
- quality engineers and process control specialists,
- industrial statisticians and Six Sigma practitioners,
- regulated industries (pharmaceuticals, medical devices, aerospace, nuclear) where early detection of drift is vital,
- laboratories and manufacturers with costly or slow testing procedures.
Related standards
- ISO 7870 series (other control chart types)
- ISO 7870-4: CUSUM control charts (complementary approach for small shifts)
- ISO 3534-2: Vocabulary and symbols - applied statistics (normative reference)
Keywords: ISO 7870-6, EWMA control chart, exponentially weighted moving average, statistical process control, process mean, small shift detection, ARL, λ smoothing parameter, quality control.
