Due to the complicated design of nuclear reactors, many complex phenomena can cause points of failure. In the case of nuclear fuel rods, the axial flow that cools the rods can induce vibrations due to the turbulent nature of the flow. Turbulence-induced pressure fluctuations create small but significant vibration amplitudes, which in turn can cause structural effects such as material fatigue and fretting wear. For this reason, turbulence-induced vibrations have been the subject of many studies, with a recent focus on the development of computational methods, so called Fluid-Structure Interaction (FSI) simulations. While scale-resolving methods can predict pressure fluctuations directly, they are typically too expensive for industrial nuclear applications in FSI. Instead, an Unsteady Reynolds-Averaged Navier-Stokes (URANS) approach coupled with a pressure fluctuation model can be used, to reduce the computational cost. While showing promising results, this approach generally underestimated the vibration amplitudes.

For this reason, an improved pressure fluctuation model, called AniPFM (Anisotropic Pressure Fluctuation Model), was developed. It models velocity fluctuations based on existing methods for synthetic turbulence. In turn, these velocity fluctuations are used to obtain the pressure fluctuations. AniPFM improves the prediction of the pressure fluctuations in three ways. First, whereas previous iterations could only represent the turbulence as isotropic, in the current model anisotropic Reynolds stresses can be embodied. Second, only the scales that can be resolved on the grid are represented by the velocity fluctuations, causing a more realistic distribution of energy along the different wavelengths. Finally, time correlation is introduced based on the transport and decorrelation of turbulence.

From simulating decaying homogeneous isotropic turbulence, it was found that this time correlation method gives a significant improvement over previous methods. From turbulent channel flow simulations, the results show that for anisotropic turbulence, the pressure fluctuations are overestimated, but they are still within a reasonable range of 10% compared to high-resolution data. AniPFM doubles the cost of simulation, compared to a URANS simulation. Even though that is a steep increase, the cost is still much lower than scale-resolving methods.

Finally, the fluid-structure interaction of a brass beam in turbulent water is simulated, which showed the ability of the AniPFM to predict turbulence-induced vibrations. The AniPFM showed errors w.r.t. the replicated experiment that were in a similar range as LES calculations, while using less computational resources. The AniPFM simulations gave an error range of 15-60% w.r.t. experimental data over the full range of simulated flow velocities, whereas a previously used pressure fluctuation model underestimated the RMS amplitude by a factor of six. While further validation is ongoing, the AniPFM has demonstrated its potential for cheaper simulations of turbulence-induced vibrations in industrial nuclear applications.