CFD Applications to Pressurized Thermal Shock-Related Phenomena

Table 6

Characteristic of representative turbulence models in this paper.

Turbulence model		Characteristic
RANS	Eddy viscosity model	Modeled for turbulence effects (Reynolds stress) by eddy viscosity coefficient (flow situation).; (i) Standard model [69]: modeled by turbulent energy and dissipation rate . and are inversely proportional. The disturbance () increases with increasing . In general, wall functions are used near walls.; (ii) Renormalization group (RNG) model [141]: each model constant of . The model is theoretically derived using renormalization group theory (less diffusive than the standard model). It predicts better than the standard model the flow separation leaving along the wall due to the effect of the adverse pressure gradient.; (iii) Standard model [142]: modeled by and specific dissipation rate (). Good accuracy close to the wall but less sensitivity in free stream conditions. The model excels at predicting separation.; (iv) Shear stress transport (SST) model [60]: hybrid model of model with better stability and model with better flow separation prediction. The model is used near the wall, and the model converted from the model is used outside the wall. In addition, the SST model accounts for the transport effects of turbulent shear stress.; In two-fluid models, eddy viscosity models have the problem of too much turbulence due to high-velocity gradients at the free surface; SST accounts for the transport of turbulent shear stress and appropriately predicts the appearance and amount of flow separation under adverse pressure gradients [5].
	Reynolds stress model (RSM)	Reynolds stress is obtained by solving the Reynolds stress transport equation without using the eddy viscosity model. The turbulent anisotropy (the turbulent appearance (statistics) varies with direction) is considered.; (i) Baseline- (BSL-) RSM [72, 143]: higher accuracy can be obtained by solving the transport equation for , local scale vorticity near the wall, and solving the transport equation for in the mainstream away from the wall.; (ii) Speziale–Sarkar–Gatski (SSG) model [88]: high Reynolds number type (do not calculate near walls). The model constants are different from the SSG model. Implemented in NEPTUNE_CFD.
Large eddy simulation (LES)		A methodology that directly simulates resolvable eddy structure and models the effect of the small eddy.; (i) Smagorinsky subgrid-scale (SGS) model [144]: calculate the eddy viscosity coefficient as proportional to the computational cell size and the magnitude of the velocity gradient (SGS eddy viscosity model). The model constant is constant and does not change depending on the flow field. Since the asymptotic behavior near a wall cannot be captured, applying a wall-damping function to the eddy viscosity coefficient is necessary. There are practical problems in extending the model directly to complex flow fields.; (ii) Wall-adapting local eddy viscosity (WALE) model [78]: using the specific tensor generally called , the WALE model calculates wall asymptotic behavior without a wall-damping function. That is, the weakness of the Smagorinsky model as above is improved. The validation of the sole model coefficient in this model is required.