Calibration and inversion using a data-model fusion-driven method with application to lead‑bismuth cooled reactors

With the deepening understanding of turbulent heat transfer phenomena in liquid metals, neither purely experimental measurements or purely numerical simulations alone can meet the demands of current research. In recent years, the development and application of data assimilation technology can contribute to solve this problem. In this work, a data-model fusion-driven method was created based on the Ensemble Kalman filter (EnKF). Then it was applied to the calibration of turbulent Prandtl number ( Pr _t ) and the prediction of average Nusselt number ( Nu ) for the flow and heat transfer phenomena of lead‑bismuth eutectic (LBE). Using the calibrated Pr _t model, the prediction of the average Nu not only achieved high agreement with Johnson's experimental data, with a relative error within 5%, but also significantly outperformed existing correlation formulas. Although its prediction accuracy decreases when extrapolating at low Peclet number ( Pe ), the error remains within ±15%. Under the higher Pe number condition, a single measurement point suffices for the calibration requirement while under the lower Pe condition, at least two measurement points are required to ensure reliable correction.