Virtualizing metrological performance: A mathematics-informed digital twin for predictive calibration of elliptical gear flow meters
FLOW MEASUREMENT AND INSTRUMENTATION, cilt.111, ss.1-9, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 111
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.flowmeasinst.2026.103468
- Dergi Adı: FLOW MEASUREMENT AND INSTRUMENTATION
- Derginin Tarandığı İndeksler: Academic Search Ultimate (EBSCO), Engineering Source (EBSCO), Scopus, Science Citation Index Expanded (SCI-EXPANDED), Compendex, INSPEC
- Sayfa Sayıları: ss.1-9
- Gazi Üniversitesi Adresli: Evet
Özet
In industries that utilize fluids, monitoring and controlling the flow is extremely important. In elliptical gear flow meters, commonly utilized in industry, calibration discrepancies arise during the design process due to the intricate interaction between gear geometry and leakage. This work proposes a digital twin (DT) that can swiftly assess the metrological performance of elliptical gear flow meters. The digital twin uses a machine-learning architecture grounded in physical principles. The research uses the Mathematics-Informed Data Augmentation (MIDA) technique, which enhances a constrained experimental dataset by applying physical principles. Nine of the 45 experimentally obtained data points were excluded from the training dataset for the digital twin model. Thus, the DT model has never seen these nine data points. The remaining 36 data points were enhanced to 225 by the MIDA approach, which employed second-degree polynomial regression and added Gaussian noise. The generated decision tree exhibited exceptional performance on the independent test dataset for leakage (ΔQ), calibration error (ΔK), and volumetric efficiency (εv), with R2 values of 0.997, 0.999, and 0.998, respectively. The MAE values of the DT model were 0.021, 0.011, and 0.040, respectively. Consequently, it was determined that the created decision tree is resistant to overfitting. The feature importance analysis identified the a/b ratio and module value of the elliptical gear as the most critical parameters influencing leakage. The DT model distinctly recognized the theoretical K-factor (Kt) as more significant than the volumetric parameters. The residual plot results indicated no variation in the forecasts. They confirmed the homogeneity of variance throughout the full operational range. The proposed digital twin model is expected to reduce the costs associated with computational fluid dynamics (CFD) simulations and experimental methods. A soft-sensor (SS) has been created to minimize both time and expenses in the fabrication of flow meters.