• 3-D CFD analysis
• 3-rod bundle
• 7 rod- bundle
• heat transfer with supercritical water

This thesis deals with phenomena related to heat transfer at supercritical pressure and more specifically it aims to check the prediction capabilities of the available turbulence models when facing the supercritical water cooling in rod bundles.
In supercritical pressure conditions, fluids undergo sudden thermodynamic property changes as soon as the so-called “pseudo-critical temperature”, variable with pressure, is crossed. Exceeding this threshold induces the shifting of the fluid conditions from a “liquid-like” to a “gas-like” state, strongly affecting the heat transfer capabilities. In fact, in that case, either deteriorated or enhanced heat transfer may occur. These phenomena are quite challenging to cope with, and the available turbulence models show many difficulties in following the experimental behaviour, either overestimating or underestimating their extent. The more the operating conditions are heavily loaded, the greater are the discrepancies between calculated trends and experimental data.
This is clearly shown in the CFD analyses performed in this work with the commercial code STAR-CCM+ version 10.06.010 (Cd-Adapco, 2015), which mostly concern the IAEA Benchmark exercise (Razumovskiy et al., 2017) about supercritical heat transfer in 3-rod bundles. The great geometrical complexity of the assembly together with the extreme operating conditions pointed out the difficulties of the most accurate models to deal with such conditions. However, a downgrading of these models, shifting from low Reynolds wall treatment to the more approximate wall functions, allowed to get predictions that could better fit the observations, pointing out the need for improvement of low-Reynolds model and suggesting at the same time a viable methodology to achieve reasonable results at present time.
In order to confirm this outcome, three additional cases concerning the 3-rod bundle, performed using the same test facilities (Razumovskiy et al., 2016), were analysed. The simulations confirmed that wall functions might give more satisfying predictions for these challenging conditions, even if they could sometimes provide anyway an overestimation of wall temperature.
The wall function approach was then adopted for two cases of the GIF Benchmark exercise (2013), concerning a much less loaded 7-rod bundle cooled with supercritical water, already addressed in previous work with low-Reynolds models obtaining similar conclusions.