Anisotropic Linear Forcing (ALF)

The Anisotropic Linear Forcing (ALF) is as an efficient way of generating turbulence fluctuations in a volume for scale-resolving simulations.

The ALF method proposed by De Laage de Meux et al. [369] is a good alternative to other synthetic turbulence generation methods, such as digital-filtering-based methods [372] or synthetic-eddy methods [370], [371]. The principal of the ALF is to force the mean (statistical) properties of a scale-resolving simulation towards target values for mean velocity and mean Reynolds-stresses. The target values can be obtained from RANS simulations of the same flow field.

In the following, the superscript ~ denotes a filtered quantity and the brackets <> denote averaging.

The volumetric forcing source term fb,ALF is added to the LES filtered momentum equation given by Eqn. (1384) as:

Figure 1. EQUATION_DISPLAY
t(ρv~)+∇⋅(ρv~v~)=∇⋅p˜I+∇⋅(T+TSGS)+ρfb,ALF
(1470)

De Laage de Meux et al. showed that all volumetric forcing formulations can be written using a general tensorial linear function of the following form:

Figure 2. EQUATION_DISPLAY
fb,ALF=Av~+b
(1471)

where A and b depend on the volumetric forcing approach.

In order to derive adequate formulation for A and b, De Laage de Meux et al. use the following properties:

  1. The contribution of the forcing term to the averaged momentum equation is equal to:
    Figure 3. EQUATION_DISPLAY
    fb,ALF=Av˜+b
    (1472)
  2. The contribution of the forcing to the budget of the resolved stresses is equal to:
    Figure 4. EQUATION_DISPLAY
    Pf=ATres+TresA
    (1473)

    where Tres=v˜v˜ is the resolved mean Reynolds-stress tensor of the LES solution.

The mean velocity and Reynolds-stresses of the scale resolving simulation can thus be driven towards target values using the following constraints:

  1. The mean restoring force to drive the mean velocity towards the target velocity vtarget can be imposed as:
    Figure 5. EQUATION_DISPLAY
    fb,ALF=1τv(vtargetv˜)
    (1474)

    where τv is a relaxation time scale.

  2. A similar constraint can be imposed for the Reynolds-stresses as:
    Figure 6. EQUATION_DISPLAY
    Pf=1τr(TtargetTres)
    (1475)

    where:

    • τr is a second relaxation time scale.
    • Ttarget=vvtarget is the target Reynolds stress tensor.

De Laage de Meux et al. showed that both properties and both constraints above lead to the following definition for A and b:

Figure 7. EQUATION_DISPLAY
A=12IIIr(IrIIrIIIr)[IrTres2HTres2Ir2(Tres2HTresTresHTres2)+(IrIIrIIIr)(Tres2H+HTres2)+(Ir3+IIIr)TresHTresIr2IIr(TresH+HTres)+(Ir2IIIr+IIr(IrIIrIIIr))H]
(1476)
Figure 8. EQUATION_DISPLAY
b=1τv(vtargetv˜)Av˜
(1477)

where:

  • Ir, IIr, and IIIr are the first, second, and third invariant of Tres.
  • H=Pf, where Pf is given by Eqn. (1475)

The two time-scales (τv related to the velocity forcing term and τr related to the Reynolds-stress term ) are defined as:

Figure 9. EQUATION_DISPLAY
τv=CvStarget
(1478)
Figure 10. EQUATION_DISPLAY
τr=max[2Δt,max(Cr0.09ktargetεtarget,νεtarget)]
(1479)

where:

  • Cv and Cr are Model Coefficients.
  • Starget is the modulus of the strain rate tensor given by Eqn. (1129) and computed from the target velocity field vtarget.
  • ktarget and εtarget are the target turbulent kinetic energy and dissipation rate, respectively.
  • ν is the kinematic viscosity.

The mean velocity v˜ and the mean Reynolds-stress components Tres are computed on the fly using an exponentially-weighted averaging (EWA) method over a specified number of time-steps nEWA.

Simcenter STAR-CCM+ applies a limiter to the resulting forcing term fb,ALF, which ensures that the Courant number associated with the forced momentum equation stays below a specified maximum value CFLmax in each cell.

nEWA and CFLmax are available as Model Coefficients.

Model Coefficients

Cr Cv CFLmax nEWA
1 0.1 0.1 750