An evolutionary algorithm for consensus building in inter-laboratory comparisons

Abstract

One of the cornerstones of the international measurement system is a requirement for national metrology institutes (NMIs) to compare their primary measurement standards in order to harmonize measurements worldwide. The need for regular comparison exercises is captured in a mutual recognition arrangement under the auspices of the International Committee for Weights and Measures (the CIPM MRA). These comparisons demonstrate the equivalence of the measurements from NMIs with an international reference value built on the consensus of all participating laboratories. Many approaches have been developed and are available today to estimate such a reference value. An alternative method based on an algorithm drawn from the science of evolution has been investigated. It displays an interesting behaviour towards extreme values which are a crucial problem for consensus building. The algorithm was compared with various approaches for processing a simulated data set and data from an international comparison in radionuclide metrology.