Proficiency testing for the calibration of masses

An interlaboraltory comparison has been organised in 2018 in the field of mass calibration. From april to october, 15 participants from 5 european countries sent five masses to each other. The participants were accredited calibration laboratories, non-accredited calibration laboratories and testing laboratories performing their own calibrations.

To ensure the stability of the masses during the circuit, an accredited reference laboratory, which was not part of the participants, calibrated the masses at the beginning and at the end of the campaign.

The calibration items were:

  • 4 masses (OIML class E2) of nominal values: 200mg, 2g, 20g and 200g
  • 1 mass (OIML class F1) of nominal value 20kg.

The preferred calibration method was the reference method of comparing the mass to be calibrated with a standard mass of equivalent nominal value, using a weighing scale. Each laboratory used its procedure and then were free to choose the number of repetitions and the number of calibration cycles. These characteristics were very disparate from one laboratory to another.

Each laboratory reported their results on an Excel file provided by the CT2M, including:

  • The conventional value of each calibrated mass,
  • The expanded calibration uncertainty (k=2),
  • Influence factors considering for the uncertainty calculation,
  • Environmental conditions during calibration.

The results were processed according to ISO 13528 (reference standard for inter-laboratory comparisons). In accordance with this this standard, the assumption of a normal distribution of the data series has been made. This hypothesis has been verified on all the results by graphical method (normal probability plot) without outliers. To identify incoherent results compared to other participants, the Grubbs test has been performed. Only one outlier was highlighted by the Grubbs test for the mass of 2g.          

However, the performance scores shown isolated and discordant results. The Z-score and number En have been used to respectively highlight a bias compared to the robust average of the participants and a bias compared to the reference value, taking into account the expanded uncertainties of both values.

It is important to note that a good correspondence between the robust means and the reference values for each of the calibrated masses was observed. This was also validated by the number En between both values which were in the range [-1; 1] for all the calibrated masses.

The most interesting results are presented in this article :