Distribution identification and information loss in a measurement uncertainty network

Measurement uncertainty is an increasingly important consideration in many applications demanding extreme performance levels. In the era of the internet of things and 5G connectivity we can learn more about device performance by utilising the increasing amount of data produced. These data require appropriate information infrastructure to facilitate continuous updating of device performance knowledge. This paper presents the results of a study which NPL undertook with a leading test and measurement device manufacturer to examine how measurement uncertainty propagates through the data traceability chain from national standards to end devices. A hierarchy of siloed calculations and heuristics did not enable a satisfactory metadata exchange within the dataflow to ensure an internally consistent calculation of measurement uncertainty. We therefore propose a novel measurement uncertainty network which contains a set of internally consistent measurement models, traceable to national standards and connected through common quantities. The network facilitates sharing and programmatic processing of measurement data with due regard to timeliness, privacy preservation and adherence to FAIR principles in measurement data exchange. An illustrative example of this network is presented with techniques to determine the best-fitting standard probability distribution for a given dataset and the resulting change in information content.