A New Approach to Treat the RANS-LES interface in PANS [1] Lars - - PowerPoint PPT Presentation
A New Approach to Treat the RANS-LES interface in PANS [1] Lars - - PowerPoint PPT Presentation
A New Approach to Treat the RANS-LES interface in PANS [1] Lars Davidson Lars Davidson, www.tfd.chalmers.se/lada PANS Low Reynolds Number Model [4] k k t + ( kU j ) = + t + ( P ) x j x
PANS Low Reynolds Number Model [4]
∂k ∂t + ∂(kUj) ∂xj = ∂ ∂xj
- ν + νt
σku ∂k ∂xj
- + (P − ε)
∂ε ∂t + ∂(εUj) ∂xj = ∂ ∂xj
- ν + νt
σεu ∂ε ∂xj
- + Cε1P ε
k − C ∗
ε2
ε2 k νt = Cµfµ k2 ε , C ∗
ε2 = Cε1 + fk
fε (Cε2f2 − Cε1), σku ≡ σk f 2
k
fε , σεu ≡ σε f 2
k
fε LRN Damping functions, f2, fµ as in [4] fε = 1.0 LES region: fk = 0.4 RANS region: fk = 1.0
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 2 / 19
PANS Low Reynolds Number Model [4]
∂k ∂t + ∂(kUj) ∂xj = ∂ ∂xj
- ν + νt
σku ∂k ∂xj
- + (P − ε)
∂ε ∂t + ∂(εUj) ∂xj = ∂ ∂xj
- ν + νt
σεu ∂ε ∂xj
- + Cε1P ε
k − C ∗
ε2
ε2 k νt = Cµfµ k2 ε , C ∗
ε2 = Cε1 + fk
fε (Cε2f2 − Cε1), σku ≡ σk f 2
k
fε , σεu ≡ σε f 2
k
fε LRN Damping functions, f2, fµ as in [4] fε = 1.0 LES region: fk = 0.4 RANS region: fk = 1.0 ◮Zonal RANS-LES: fk has a large gradient at the RANS-LES interface
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 2 / 19
PANS: derivation
The PANS k equation is derived by multiplying the RANS k equation by fk. The left hand reads fk Dktot Dt (1) where D/Dt = ∂/∂t + ¯ ui∂/∂xi, ktot = k + kres (modeled plus resolved). If it is assumed that fk constant, Eq. 1 can be re-written as fk Dktot Dt = Dfkktot Dt = Dk Dt , fk = k ktot (2) If fk is not constant, Eq. 2 must be written as (Girimaji & Wallin [2]) fk Dktot Dt = Dfkktot Dt − ktot Dfk Dt = Dk Dt − ktot Dfk Dt This work presents models for the boxed term at RANS-LES interfaces, i.e.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 3 / 19
PANS: derivation
The PANS k equation is derived by multiplying the RANS k equation by fk. The left hand reads fk Dktot Dt (1) where D/Dt = ∂/∂t + ¯ ui∂/∂xi, ktot = k + kres (modeled plus resolved). If it is assumed that fk constant, Eq. 1 can be re-written as fk Dktot Dt = Dfkktot Dt = Dk Dt , fk = k ktot (2) If fk is not constant, Eq. 2 must be written as (Girimaji & Wallin [2]) fk Dktot Dt = Dfkktot Dt − ktot Dfk Dt = Dk Dt − ktot Dfk Dt This work presents models for the boxed term at RANS-LES interfaces, i.e.
◮ horizontal RANS-LES interface in boundary layer (channel flow) www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 3 / 19
PANS: derivation
The PANS k equation is derived by multiplying the RANS k equation by fk. The left hand reads fk Dktot Dt (1) where D/Dt = ∂/∂t + ¯ ui∂/∂xi, ktot = k + kres (modeled plus resolved). If it is assumed that fk constant, Eq. 1 can be re-written as fk Dktot Dt = Dfkktot Dt = Dk Dt , fk = k ktot (2) If fk is not constant, Eq. 2 must be written as (Girimaji & Wallin [2]) fk Dktot Dt = Dfkktot Dt − ktot Dfk Dt = Dk Dt − ktot Dfk Dt This work presents models for the boxed term at RANS-LES interfaces, i.e.
◮ horizontal RANS-LES interface in boundary layer (channel flow) ◮ vertical RANS-LES interface in embedded LES (channel flow) www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 3 / 19
Interface Model 1
In [2], ktotDfk/Dt is represented by introducing an additional turbulent viscosity, νtr, in the momentum equation ∂ ∂xj (νtr¯ sij) , ¯ sij = 1 2 ∂¯ ui ∂xj + ∂¯ uj ∂xi
- where
νtr = Pktr |¯ s|2 , Pktr = νtr|¯ s|2 = ktot Dfk Dt = k fk Dfk Dt The object of Pktr is to decrease νt and facilitate growth of resolved turbulence on the LES side of an interface Hence, only νtr < 0 is used which corresponds to Dfk/Dt < 0 (from RANS to LES). νt + νtr > 0 in the momentum equation (but not in the k equation)
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 4 / 19
Interface Model 2
This model is identical to Model 1 except that k/fk is replaced by ktot i.e. Pktr = k fk Dfk Dt
Model 1
Pktr = ktot Dfk Dt
Model 2
ktot = k + 1 2¯ u′
i¯
u′
ir.a
where subscript r.a. denotes running average. In PANS, fk is defined as fk = k/ktot In post-processing it is usually found that fk > k/ktot (approx. a factor 4 larger) ⇒ |Pktr |model 2 > |Pktr |model 1
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 5 / 19
Interface Model 3
In Models 1 & 2, νtr = Pktr /|¯ s|2 which may cause numerical problems. Model 3 does not involve νtr. The original term ktotDfk/Dt is used in the k equation Adding the term −0.5¯ u′
i
Dfk Dt − k¯ u′
i
¯ u′
m¯
u′
m
Dfk Dt in the momenum equation corresponds to the time-averaged term − ¯ u′
i¯
u′
i
2 Dfk Dt
- − k¯
u′
i ¯
u′
i
¯ u′
m¯
u′
m
Dfk Dt = −ktotDfk Dt in the kres equation. However, this term causes numerical instability. Hence it is not used. Only the term in the k equation, ktotDfk/Dt < 0, is used.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 6 / 19
Interface Models: summary
Models 1 & 2: additional turbulent viscosity, νtr = Pktr /|¯ s|2 < 0, in Pk and momentum equations
◮ Limit in momentum equations: νt + νtr > 0 ◮ Model 1: Pktr = k
fk Dfk Dt
◮ Model 2: Pktr = ktotr.a
Dfk Dt
Model 3: additional production term, Pktr , in k equation without use
- f νtr
Pktr = ktotr.a Dfk Dt < 0 Models 1-3 correspond to the non-commutivity in DES beteen filtering and spatial derivative at RANS-LES interfaces (Hamba [3]) ∂f ∂xi = ∂¯ f ∂xi − ∂∆ ∂xi ∂¯ f ∂∆
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 7 / 19
Fully Developed Channel Flow
The URANS and the LES regions. x y wall yint
LES, fk = 0.4 URANS, fk = 1
Reτ = uτδ/ν = 2 000, Re = 4 000 and Re = 8 000 xmax = 3.2, ymax = 2 and zmax = 1.6. 32 × 32 cells in the x − z plane Ny = 80 cells (Reτ = 2 000 and 4 000) or Ny = 96 (Reτ = 8 000)
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 8 / 19
Fully Developed Channel Flow: Results
100 1000 10 20 30
Velocity y + U+
500 1000 50 100 150
Turbulent viscosity y + (νt + νtr)/ν : Model k
fk Dfk Dt
: Model ktot
Dfk Dt .
: no interface model. +: U+ = ln(y +)/0.4 + 5.2
- : location of the computational cell centers.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 9 / 19
Fully Developed Channel Flow: Results
500 1000 −20 20 40
Production terms in k eq y + Pk + Pktr
0.5 1 −1 −0.5 0.5 1
modeled & resolved stresses y −τ ν
12 − τ12,
−u′v ′ resolved modeled : Model k
fk Dfk Dt
: Model ktot
Dfk Dt .
: no interface model.
- : the location of the computational cell centers.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 10 / 19
Embedded Channel Flow
x y 2 0.95 5.45
LES, fk = 0.4 RANS fk = 1
Interface Reτ = uτδ/ν = 950 The domain size is 6.4 × 2 × 1.6 (x, y, z) 128 × 80 × 64 cells.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 11 / 19
Synthetic Fluctuations at the Interface
0.5 1 1.5 2 0.5 1 1.5 2
y
0.1 0.2 0.3 0.4 0.2 0.4 0.6 0.8 1
Bww(ˆ z) ˆ z : u2
rms/u2 τ
: v 2
rms/u2 τ
: w2
rms/u2 τ
: u′v ′/u2
τ
Two-point correlation of synthetic in- let fluctuations
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 12 / 19
Velocity and Skin Friction
10 10
1
10
2
5 10 15 20
Velocity at x = 5.5 y + ¯ u/uτ,in
2 4 6 0.8 0.9 1 1.1
Friction velocity x uτ
- ∗
: Model k
fk Dfk Dt
: Model ktot
Dfk Dt
: no interface model
- : U+ = ln(y +)/0.4 + 5.2
∗: uτ, RANS when x → ∞
- : uτ, PANS when x → ∞
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 13 / 19
Resolved and Modeled Turbulence
2 4 6 5 10 15
Streamwise fluctuation x max
y (u2 rms(x, y)) 2 4 6 20 40 60 80 100
Turbulent viscosities x max
y ((νtot(x, y)/ν
: Model k
fk Dfk Dt
: Model ktot
Dfk Dt
: no interface model νtot = ν + νt + νtr
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 14 / 19
Production at Interface
−1000 −500 500 0.2 0.4 0.6 0.8 1
x = 1 Production of k y : Pk + Pktr , Model k
fk Dfk Dt .
: Pk + Pktr , Model ktot
Dfk Dt .
: Pk, no interface model;
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 15 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
◮
k fk Dfk Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 1
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
◮
k fk Dfk Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 1
◮ ktot Dfk
Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 2
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
◮
k fk Dfk Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 1
◮ ktot Dfk
Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 2
◮ ktot Dfk
Dt added no νttr to the k eq (Model 3)
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
◮
k fk Dfk Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 1
◮ ktot Dfk
Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 2
◮ ktot Dfk
Dt added no νttr to the k eq (Model 3)
Model 2 works very well
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Conclusions
Three interface models for horizontal (wall-parallel) and vertical interfaces (embedded LES) have been presented:
◮
k fk Dfk Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 1
◮ ktot Dfk
Dt added via νttr to the k eq and mom eq (νttr + νt > 0): Model 2
◮ ktot Dfk
Dt added no νttr to the k eq (Model 3)
Model 2 works very well Model 3 gives identical results to Model 2 (not shown in this presentation)
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 16 / 19
Three-Day CFD Course at Chalmers
Unsteady Simulations for Industrial Flows: LES, DES, hybrid LES-RANS and URANS 29-31 October 2014 at Chalmers, Gothenburg, Sweden Max 16 participants 50% lectures and 50% workshops in front of a PC Registration deadline: 10 October 2014 For info, see http://www.tfd.chalmers.se/˜lada/cfdkurs/cfdkurs.html
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 17 / 19
References I
[1] Davidson, L. A new approach to treat the RANS-LES interface in PANS. In ETMM10: 10th International ERCOFTAC Symposium on Turbulence Modelling and Measurements (Marbella, Spain, 2014). [2] Girimaji, S. S., and Wallin, S. Closure modeling in bridging regions of variable-resolution (VR) turbulence computations. Journal of Turbulence 14, 1 (2013), 72 – 98. [3] Hamba, F. Analysis of filtered Navier-Stokes equation for hybrid RANS/LES simulation. Physics of Fluids A 23, 015108 (2011).
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 18 / 19
References II
[4] Ma, J., Peng, S.-H., Davidson, L., and Wang, F. A low Reynolds number variant of Partially-Averaged Navier-Stokes model for turbulence. International Journal of Heat and Fluid Flow 32, 3 (2011), 652–669. 10.1016/j.ijheatfluidflow.2011.02.001.
www.tfd.chalmers.se/˜lada ETMM10, Marbella, 2014 19 / 19