EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Measurement as Modeling: The Bayesian Bridge Between Artificial Intelligence and Metrological Traceability

Ignacio Lira

IEEE Open Journal of Instrumentation and Measurement · 2026 · 5 · 1000106-1000106

DOI: 10.1109/ojim.2026.3699771 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Open Journal of Instrumentation and Measurement, 2026. Verfasst von Ignacio Lira. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 18 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid integration of artificial intelligence (AI) into instrumentation has created a paradox: while AI-driven “soft sensors” offer unprecedented predictive power, they frequently fail to meet the fundamental requirements of metrological traceability. Measurement is not merely data acquisition; it is a model-based process that constructs a bridge between empirical reality and information. Current AI methods—often functioning as “opaque models”—emphasize scalability at the expense of metrological rigor. They typically lack the explicit structural models needed to validate this bridge, which risks noncompliance with the Guide to the Expression of Uncertainty in Measurement (GUM). This article proposes that the solution lies not in forcing AI into the linear constraints of classical propagation, but in fully embracing the Bayesian approach of GUM Supplement 1. We argue that the training phase of a machine-learning algorithm is metrologically equivalent to a massive Type B evaluation and must be treated as such across the model lifecycle—including data curation, training, validation, deployment, and recalibration—so that uncertainty statements remain current and defensible. By addressing the errors-in-variables (EiV) imperative and reinterpreting neural-network weights and true inputs as random variables with probability density functions (pdfs), and by using practical Bayesian surrogates (for example, Bayesian neural networks (BNNs), ensembles and Monte Carlo dropout, or diffusion-based input posteriors), we establish a framework that disentangles aleatoric from epistemic uncertainties. In doing so, we restore the epistemic link essential for traceability in the digital age.

Zitieren

APA

Lira, I. (2026). Measurement as Modeling: The Bayesian Bridge Between Artificial Intelligence and Metrological Traceability. IEEE Open Journal of Instrumentation and Measurement, 5, 1000106-1000106. https://doi.org/10.1109/ojim.2026.3699771

Vancouver

Lira I. Measurement as Modeling: The Bayesian Bridge Between Artificial Intelligence and Metrological Traceability. IEEE Open Journal of Instrumentation and Measurement. 2026;5:1000106-1000106. doi:10.1109/ojim.2026.3699771

BibTeX
@article{Lira2026,
  author = {Lira, Ignacio},
  title = {Measurement as Modeling: The Bayesian Bridge Between Artificial Intelligence and Metrological Traceability},
  journal = {IEEE Open Journal of Instrumentation and Measurement},
  year = {2026},
  volume = {5},
  pages = {1000106--1000106},
  doi = {10.1109/ojim.2026.3699771}
}
Kostenlos starten →