Density stratification breakup by a vertical jet: Experimental and numerical investigation on the effect of dynamic change of turbulent schmidt number

The hydrogen behavior in a nuclear containment vessel is one of the significant issues raised when discussing the potential of hydrogen combustion during a severe accident. Computational Fluid Dynamics (CFD) is a powerful tool for better understanding the turbulence transport behavior of a gas mixture, including hydrogen. Reynolds-averaged Navier-Stokes (RANS) is a practical-use approach for simulating the averaged gaseous behavior in a large and complicated geometry, such as a nuclear containment vessel; however, some improvements are required. In this paper, we focused on the turbulent Schmidt number Sc-t for improving the RANS accuracy. Some previous studies on ocean engineering mentioned that the Sc-t value gradually increases with the increasing stratification strength. We implemented the dynamic modeling for Sc-t based on the previous studies into the OpenFOAM ver 2.3.1 package. The experimental data obtained by using a small scale test apparatus at Japan Atomic Energy Agency (JAEA) was used to validate the RANS methodology. In the experiment, we measured the velocity field around the interaction region between vertical jet and stratification by using the Particle Image Velocimetry (PIV) system and time transient of gas concentration by using the Quadrupole Mass Spectrometer (QMS) system. Moreover, Large-Eddy Simulation (LES) was performed to phenomenologically discuss the interaction behavior. The comparison study indicated that the turbulence production ratio by shear stress and buoyancy force predicted by the RANS with the dynamic modeling for Sc-t was a better agreement with the LES result, and the gradual decay of the turbulence fluctuation in the stratification was predicted accurately. The time transient of the helium molar fraction in the case with the dynamic modeling was very closed to the VIMES experimental data. The improvement on the RANS accuracy was produced by the accurate prediction of the turbulent mixing region, which was explained with the turbulent helium mass flux in the interaction region. Moreover, the parametric study on the jet velocity indicates the good performance of the RANS with the dynamic modeling for Sct on the slower erosive process. This study concludes that the dynamic modeling for Sct is a useful and practical approach to improve the prediction accuracy.