Evaluation of different implementations for determining Callendar-Van Dusen equation coefficients using MPE, RMSE, and NSE metrics and their impact on measurement uncertainty
MEASUREMENT: JOURNAL OF THE INTERNATIONAL MEASUREMENT CONFEDERATION, cilt.275, ss.121375, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 275
- Basım Tarihi: 2026
- Dergi Adı: MEASUREMENT: JOURNAL OF THE INTERNATIONAL MEASUREMENT CONFEDERATION
- Derginin Tarandığı İndeksler: Scopus, Science Citation Index Expanded (SCI-EXPANDED)
- Sayfa Sayıları: ss.121375
- Marmara Üniversitesi Adresli: Evet
Özet
This study investigates how different practical implementations of coefficient estimation in the calibration of platinum resistance thermometers affect temperature deviations associated with coefficient-related equation mismatch and uncertainty-related outcomes. Within this scope, two implementations for determining the coefficients of the Callendar-Van Dusen equation were considered for ten industrial platinum resistance thermometers (IPRTs): an explicit matrix formulation and a regression-based implementation using standard numerical tools. The resulting coefficient sets were evaluated in terms of temperature deviations between the fitted equations and the calibration data using MPE, RMSE, and NSE metrics. In addition, the contribution of equation-data mismatch was considered within the measurement uncertainty budget. The results indicate that, for the investigated data sets, the explicit matrix formulation leads to smaller temperature deviations, whereas the alternative implementation produces deviations that remain acceptable for typical industrial applications. The study provides quantitative insight into the conditions under which differences between coefficient estimation implementations become negligible and when a more rigorous numerical formulation may be required.