K.2Procedure

The results of two interlaboratory test programmes (programme A, reference [7] and programme B, reference [8]) are given as examples of the type of uncertainties, which are typically obtained when testing metallic materials.

For each material included in the programme, a fixed number of specimen blanks are randomly selected from the stock. A preliminary study checks the homogeneity of this stock and provides data on the "intrinsic" scatter of the mechanical properties within the stock. The blanks are sent to the participating laboratories, where the test pieces are machined to the drawings they normally use. The only requirement for the test pieces and the testing itself are the compliance with the requirements of the relevant standards. As much as possible, it is recommended that the tests be made in a short period of time, by the same operator using the same machine.

In tables K.1 and K.2, these three kinds of error are expressed in terms of a relative uncertainty coefficient: $UC_{r}=\pm2s_{r}/\overline{x}(\%)$ $UC_{L}=\pm2s_{L}/\overline{x}(\%)$ $UC_{R}=\pm2s_{R}/\overline{x}(\%)$

where

• $\bar{x}$ is the general average;
• $s_{r}$ is the estimated repeatability standard deviation within laboratories;
• $s_{L}$ is the estimated variability between laboratories;
• $s_{R}$ is the estimated precision of the test method: reproducibility standard deviation.

These quantities are close to the 95 % confidence interval of $\bar{x}$. They are calculated for each material tested and each property.