ISO 25337:2010 Plastics QC Single Measurement Statistical Model Standard Deviation Reproducibility

1. Introduction to ISO 25337 and Single-Measurement SD Models

ISO 25337:2010 provides a statistical model for characterizing the precision of single measurement results in quality control testing of plastics. The standard introduces the concept of the “standard deviation of single measurements under reproducibility conditions” (sRLab) as a key metric for evaluating and comparing the performance of QC testing laboratories.

Unlike traditional interlaboratory studies that require multiple replicate measurements at each laboratory, ISO 25337s single-measurement model allows QC laboratories to participate in precision studies using their routine test results. This approach significantly reduces the burden of participation while still providing statistically valid estimates of laboratory precision and bias. The model is particularly valuable for plastics testing where standard test methods (tensile properties, flexural modulus, melt flow rate, etc.) are routinely performed as single measurements rather than replicates.

2. Statistical Model and Methodology

2.1 The Single-Measurement Precision Model

The model is based on the analysis of variance (ANOVA) framework, adapted for the case where each laboratory provides only a single measurement for each test material. The standard deviation of single measurements under reproducibility conditions (sRLab) captures the combined effects of lab-to-lab variability, test piece variability, and test method variability. The standard defines three key precision parameters: repeatability standard deviation (sr), reproducibility standard deviation (sR), and the single-measurement reproducibility standard deviation (sRLab) which is specific to each participating laboratory.

The model assumes that the measured value y for a given material at laboratory i follows: y_i = mu + B_i + e_i, where mu is the true value (population mean), B_i is the laboratory bias (normally distributed with variance sigma_L2), and e_i is the random error (normally distributed with variance sigma_r2). The single-measurement reproducibility variance at laboratory i is sigma_RLab_i2 = sigma_L2 + sigma_r2, estimated by sRLab_i2.

2.2 The Mandel-h and Mandel-k Statistics

ISO 25337 incorporates Mandel-h (between-laboratory consistency) and Mandel-k (within-laboratory consistency) statistics for outlier identification. The Mandel-h statistic measures the deviation of a laboratorys result from the grand average relative to the variability between laboratories. The Mandel-k statistic measures the variability within a laboratory relative to the average within-laboratory variability across all laboratories. Laboratories with Mandel-h or Mandel-k values exceeding critical thresholds (typically h > 3.0 or k > 2.0 at the 1% significance level) require investigation.

3. Engineering Implementation and Practical Use

3.1 Setting Up a Single-Measurement Precision Study

To implement ISO 25337, a QC laboratory needs to follow a structured protocol. First, select at least one test material representing the typical product range (the standard recommends 2-5 materials for comprehensive characterization). Second, collect routine single-measurement results over a period of time sufficient to capture normal process variation (typically 10-30 data points per material). Third, compute the laboratory-specific precision statistics: mean, sRLab, Mandel-h, and Mandel-k values. Fourth, compare sRLab values against established precision targets from collaborative studies or historical data.

The standard recommends that the data collection period span at least 10 different testing days to ensure the results capture day-to-day variability. For each material, at least 10 single measurements are required for meaningful statistical analysis, with 20 or more being preferable. The measurements must be performed under reproducibility conditions (different operators, different days, recalibration between runs) to ensure the sRLab estimate reflects realistic laboratory performance.

3.2 Establishing Production and Warning Limits

A practical application of ISO 25337 is establishing QC charts with appropriate control limits. Using the laboratory-specific sRLab, the standard deviation of the mean (s_mean = sRLab/sqrt(n)) can be used to set warning limits (mean +/- 2s_mean) and action limits (mean +/- 3s_mean) for ongoing QC monitoring. When a laboratorys sRLab exceeds the reproducibility standard deviation sR from the collaborative study, it indicates that the laboratorys precision is worse than the expected interlaboratory variability, triggering investigation and corrective action.

4. Applications and Limitations

ISO 25337 is applicable to any plastics test method where single measurements are the standard practice, including tensile testing (ISO 527), flexural testing (ISO 178), impact resistance (ISO 179, ISO 180), melt mass-flow rate (ISO 1133), density (ISO 1183), and hardness (ISO 2039-1). The model is not suitable for test methods requiring replicate measurements by their nature (e.g., thermal analysis where multiple runs are standard). Laboratories transitioning from replicate-based QC to single-measurement QC should run parallel studies to validate the comparability of sRLab estimates with traditional precision parameters.

5. Frequently Asked Questions