Measurement uncertainty, data quality and data-driven modelling

The measurement uncertainty is still the most important quality parameter for specifying the result of a measurement. Its calculation, specification in certificates and further use is standardized and accepted internationally [1]. Measurement uncertainty is of enormous importance both for the assessment of measurement and test systems, for their reliable functionality, for safety, as a basis for system control, conformity assessment, etc.

Newer developments also rely on the specification of measurement uncertainty to assess their trustworthiness. This applies first of all to sensor networks, which place new demands on calibration, but also offer new possibilities in the utilization of redundancy. In addition, there is the leveraging of measurement data quality in accordance with the FAIR Guiding Principles [2] and the use of data for industrial condition monitoring.

The combination of trustworthiness of cognitive systems with classical measurement uncertainty appears to be of absolutely outstanding importance in the future. First publications on this topic already exist [3]. The journal “tm – Technisches Messen” has contributed on this with previous special issues on “Measurement Systems and Sensors with Cognitive Features” [4], [5].

The contributions on measurement uncertainty in this issue are based on presentations at the Conference on Measurement Uncertainty Erfurt 2023. A first issue with contributions primarily in German has already been published [6].

The first paper in this special issue by Olaf Werhahn and Sascha Eichstädt on metrological traceability and measurement uncertainty for the determination of air quality using sensor networks is highly topical in terms of its subject matter and approach and highly relevant as air pollution is a major threat to health and climate worldwide.

With the second paper of this issue Markus Pabst et al. address a highly interesting topic from the field of one-step traceability with new developments of the Kibble balance. They extend the well-known concept, in particular the table-top balance, and thus further develop the redesign and simplification of classic traceability schemes.

The next important article by Matthias Bodenbenner et al. addresses leveraging measurement data quality by applying the international FAIR Guiding Principles and thus provides a glimpse into the future handling of data and requirements for data quality by addressing qualitative aspects like completeness, consistency, and reliability.

Steffen Klein et al. extend the often-discussed data-based fault classification for industrial condition monitoring, which is typically limited to known faults, by combining a generic novelty detection approach with supervised classification pipelines. This allows smart systems to continuously learn in real-world scenarios supporting the application expert.

The article by Tino Hausotte et al. deals with the important topic of conformity assessment of geometric specifications, which is based on the evaluation of measurement uncertainty.

Data-based modeling (so-called black box models) is becoming increasingly important not only for the state-adapted modeling of measurement systems or parts thereof, but cognitive systems also necessarily fall back on it. Linda-Sophie Schneider et al. provide a very comprehensive introduction, an overview of current developments and an outlook to the likely future.

Overall, the articles provide a good overview of the current status of the topics of data quality, measurement uncertainty and modeling in metrology.

We hope all readers will enjoy this special issue and will gain new insights, not only into the fields addressed, but also for other related areas.