Page 285 - Thermal Hydraulics Aspects of Liquid Metal Cooled Nuclear Reactors
P. 285

Large-eddy simulation: Application to liquid metal fluid flow and heat transfer  255


                                        3=2
                                    d   d
                                  S S
                                   ij  ij
               ν SGS ¼ C W Δ                   ,                       (6.1.2.31)
                           2
                                             5=4
                                  5=2    d   d

                            S ij S ij  + S S
                                        ij  ij
           where

                    1 ∂~ u i ∂~ u k  ∂~ u j ∂~ u k
                  ¼          +        ¼ S ik S kj + Ω ik Ω kj          (6.1.2.32)
               S ij
                    2 ∂x k ∂x j  ∂x k ∂x i
           gives the correct near-wall behavior. The tensor Ω is the antisymmetric part of the
           velocity gradient tensor. The model coefficient was calibrated by Nicoud and
           Ducros (1999) on homogeneous isotropic turbulence at moderate Reynolds and it
           gives C W ¼ 0.25. Bricteux et al. (2009) calibrated the coefficient to C W ¼ 0.20 for
           their channel flow (at Re τ ¼ 395) and showed that the model is inappropriate for lam-
           inar vortical flows because it is still active in solid body rotation regions. They con-
           cluded that the WALE model could be too dissipative, although less than
           Smagorinsky, when solid rotation takes place. This was the rationale for proposing
           a new multiscale version of the WALE model, which will be presented in
           Section 6.1.2.3.3.2.



           6.1.2.3.2 Dynamic models

           6.1.2.3.2.1 The approach
           The philosophy of the dynamic approach, in general, is to calculate, locally, the coef-
           ficient of the subgrid stress model during a simulation. For example, this procedure
           allows to locally determine C S (Eq. 6.1.2.27) in the Smagorinsky model, but could
           be applied to any model so almost each subgrid-scale model has its dynamic version,
           yet it does not work well for the WALE model. This approach was first introduced by
           Germano et al. (1991) and relies on the application of a second-level filtering having a
           larger filter width than the LES filter. This second-level filter is also called a test filter.
           As we here deal with practical implicitly filtered LES, the second filter can be con-
           structed as the projection on a larger grid size αh with α > 1. In spectral space, it means
           we apply a second sharp Fourier cutoff k 2 < k 1 , k 1 being the first-level cutoff of the
           LES (see Fig. 6.1.2.3).
              Eq. (6.1.2.7), using a projection operator ð…Þ, is considered as the level 1 LES. The
                                              g
           level 2 (only) LES equation could be obtained by replacing the operator ð…Þ by the
                                                                      g
           operator ð…Þ related to the level 2 wave number cutoff k 2 ¼ π/αh. Hence it is possible
                   d
           to obtain a relation between the different stress tensors:
                     SGS,1
                           d
               ^    ¼ σ ^    ^ ^  g,                                   (6.1.2.33)
                                d
                           g
               σ ij  ij  + ~ u i ~ u j   ~ u i ~ u j
   280   281   282   283   284   285   286   287   288   289   290