High-Fidelity Multiphase Modeling of Critical Heat Flux Phenomena in Pressurized Water Reactor Fuel

Abstract

The prediction of Critical Heat Flux (CHF), the limiting phenomenon of boiling heat transfer, has remained a major challenge since the early development of nuclear reactors. Although several mechanistic models have been proposed, current industry practices still rely heavily on empirical correlations and sub-channel codes, which are limited in scope and accuracy outside of their calibration domain.

This thesis aims to improve CHF prediction methodology, focusing specifically on the Departure from Nucleate Boiling (DNB) under high-pressure conditions representative of Pressurized Water Reactors. This work is conducted using Multiphase Computational Fluid Dynamics in a Reynolds-Averaged Navier-Stokes Eulerian-Eulerian framework. Two sets of boiling closures, CASL-FY19 and MITB, were compared in the commerically available Siemens STAR-CCM+ software. An automated Java-based heat flux incrementation method was developed, and validation was performed against experimental data from both circular and square test sections at pressures between 2 and 13.79 MPa.

The performance of a near-wall void fraction DNB criterion was evaluated and found to consistently underpredict experimental values, with an average error of 11.24 \%. This error was found to be strongly dependent on the mass flux of each case. Different modeling configurations were explored to improve results, revealing that accurate modeling of the void fraction distribution plays a central role in prediction quality.

The technical readiness of a heat partitioning based DNB criterion was also assessed. Two key limitations were identified: first, numerical instabilities due to excessive vapor accumulation near the wall at high heat fluxes; second, a persistently low dry area fraction even under vigorous boiling, attributed to an incomplete modeling of merging dry spots and an inadequate contact angle specification.

The findings underscore the importance of improving near-wall void fraction representation for both DNB detection models. While the heat partitioning based model demonstrates a stronger physical foundation and potential for greater accuracy, the near-wall void fraction model currently offers superior robustness and more consistent results within the framework.