Sie sind auf Seite 1von 7

1.

Meangreen

2. Background and Theory

Experiments performed by Cooper et al. (1993) on turbulent jets impinging on an orthog-


onal wall provided benchmark data for analyzing turbulence models (Craft et al., 1993;
Fairweather and Hargrave, 2002). Cooper et al. measured the flow and turbulence prop-
erties at various radial and axial locations close to the plate. Cooper et al. reported data
for two jet Reynolds numbers and experiments with different pipe height to pipe diameter
ratios. The simulations described are for only one set of experimental conditions.

Three different types of turbulence models used to simulate the jet; the Reynolds stress
model, k- model, k-ω model. Two forms of the k- equation were used, the standard model
as well as the renormalization group theory (RNG) model. Two forms of the Reynolds stress
model with two forms of the pressure rate of strain tensor (linear and quadratic) were used
to compare the affect of the different forms on the prediction of the jet impingement. Table
1 provides a list of nomenclature used in this paper.

The k- models and k-ω model are two equation turbulent viscosity models, which means
they are based on the turbulent viscosity hypothesis. The turbulent viscosity hypothesis is
comparable to the stress-rate-of-strain relation for a Newtonian fluid and is very important
to understanding the limitations of turbulent viscosity models. The turbulent viscosity
hypothesis is given in eqn. (1) and the definition of the turbulent viscosity used for the k-
models and Reynold’s Stress models is given by (2). The k- ω model uses a different form
for turbulent viscosity and is expressed in table 4.

Turbulent viscosity hypothesis eqn. (1)


 
2 ∂Ui ∂Uj
(1) ui uj = kδij − νt +
3 ∂xj ∂xi

k2
(2) νt = Cµ

1
C. Q. LaMarche
Table 1. List of nomenclature used.
Term Definition
aij anisotropy aij = ui uj − 23 kδij
k turbulent kinetic energy k= 21 tr(uii )
ui uj Reynolds stress
ui r.m.s. fluctuating velocity of direction i
u1 =u r.m.s. fluctuating velocity in the axial direction
U1 =U mean velocity in the axial diraction
u2 =v r.m.s. fluctuating velocity in the radial direction
U2 =V mean velocity in the radial direction
uv mean turbulent shear stress
q
uτ frictional velocity uτ = τρw
Pk production of turbulent kinetic energy term
r radial coordinate
y axial coordinate
y+ y+= uντ y  
∂U
Sij mean rate of strain Sij = 12 ∂xjj + ∂Ui
∂xj
p
S 2Sij Sij
δij kronecker delta
∂u
 dissipation rate of k  = ν ∂ui j
∂xi ∂xi

ω specific 
dissipationrate of k ω = k
∂U
Ωij Ωij = 12 ∂U i
∂xj
− ∂xij
ν kinematic viscosity
νt turbulent viscosity
νef f effective viscosity νef f = ν + νt
ρ density
Turbulence model contribution terms
dk diffusion of turbulent kinetic energy
d ”diffusion” of turbulent dissipation rate
Pk production of turbulent kinetic energy
 dissipation
σk turbulent Prandtl number for k
σ turbulent Prandtl number for 
φij pressure strain term

On the right hand side of eqn. (1) the deviatoric Reynolds stress, or Reynolds stress
anisotropy aij , (R.H.S) is proportional to the mean rate of stain and where νt is the eddy
viscosity, or turbulent viscosity. The turbulent hypothesis is important to the limitation of
the turbulent viscosity models because the hypothesis assumes that the turbulent Reynolds
stress anisotropy is determined by the mean velocity gradients and the relationship between
them. The importance of the turbulent viscosity hypothesis on the results of the simulations
2
C. Q. LaMarche

Table 2: Model equations for the standard k- model


Dk
Dt
= dk + Pk − 
D  2
Dt 1 k Pk −
= d + C  C2 k
∂ νt ∂k
dk = ∂xj
ν+ σk ∂xj
 
∂ νt ∂
d = ∂xj
ν+ σ ∂xj
∂U 2
Pk = ui uj ∂xij = νt S
2
νt = Cµ k
C1 =1.44 C2 =1.92 Cµ =0.09 σk = 1.0 σ = 1.3

Table 3. Model equations for the RNG k- model


Dk
Dt
= dk + Pk − 
D  2
Dt
= d + C1  P − C2 k− R
k k
dk = ∂x∂ j αk νef f ∂x
∂k
j

Pk is same as k- in table


 2
∂ ∂
d = ∂xj
α νef f ∂x j
3 (1−η/η ) 2
R = Cµ η1+βη 3
o 
k
η = Sk/
Cµ = 0.0845
η0 = 4.38 β = 0.012 C1 = 1.42 C2 = 1.68

will be discussed in detail in the results section. Further discussions of the limitations can
be found elsewhere as well (Pope, 2000; Craft, 1993). The turbulent viscosity hypothesis is
generally not valid, but is reasonable when describing simple shear flow.

The standard k- model of Jones and Launder (1972) and Launder and Spaulding (1973)
was used to model the jet as well as the RNG k- model (Zhang and Orzag, 1998). Table
2 provides the model equation for the standard k- model. Table 3 provides the model
equations for the RNG k- model.

The k equation in the k- model is derived from the dotting the fluctuating momentum
balance with the fluctuating velocity component, then the unclosed terms of the equation
are closed by including many assumptions on the forms of the terms in the equation (Pope,
2000).

The RNG k- model has an extra term in the  equation, R , compared to the standard
model, that improves the accuracy of the RNG model for rapidly strained flows, swirling flows
3
C. Q. LaMarche

and has different model constants. The different prediction of the standard and RNG k-
models is attributed to the extra term in the RNG model and difference in model constants.
The effect of the R term can be better understood if the RNG  equation from table 3 is
rearranged to the following:

D  2
(3) = d + C1 Pk − C ∗2
Dt k k


Where C2 is defined below

∗ Cµ η 3 (1 − η/η0 )
(4) C2 = C2 +
1 + βη 3

Since η, defined in table 3, is a function of the mean strain it is clear that the extra term
in the RNG turbulence model makes the RNG model more sensitive to rapid strain and
streamline curvature. The magnitude of the stain will affect the sign or R and therefore
effects the dissipation of . In regions where the dissipation of  is less then  will be higher
so k will be reduced and the turbulent viscosity will be reduced. Similarly in areas with a
high destruction of  the opposite is true. When η>η0 , in regions of high mean strain, the

R term reduces the magnitude of C2 , which means that there is a smaller dissipation of 
and therefore reduces k and νt .

A very important point to make about the RNG k- model is that in regions of weak or
moderate strain the RNG model gives results close to the standard model. According the

Fluent user’s manual η ≈3 in the logarithmic layer. This gives a values of C2 ≈2 which is
close to the value of C2 from the standard k- model given in table 2.

The k-ω model (Wilcox, 1993) is another 2 equation turbulent viscosity model. Table 4 gives
the model equations for the k-ω model.

The k-ω model is empirical, just like the  equation in table 3.The k-ω model has modifica-
tions to account for shear flow spreading, low Reynolds number flow, and adverse pressure
4
C. Q. LaMarche
νt
Γk = ν + σk
Γω = ν + σνωt
 
αo +Ret /Rk
Table 4. Model equations for the k-ω model νt = α∗ ωk α∗ = α∞

1+Ret /Rk
Dk
Dt
= dk + Pk − Yk Ret = νω k
αo∗ = β3i

Dt = dω + Pω − Yω σk = 2.0 σω = 2.0
∂ ∂k ∂U
dk = ∂xj Γk ∂xj = diffusion of k Pk = −ui uj ∂xij =Production of k
Pk = νt S 2 Pω = α ωk Pk
 
∂k
dω = ∂x∂ j Γω ∂x = diffusion of ω
j
Yk and Yω = dissipation of k and ω
Yk = β ∗ fβ ∗ kω; Yω = βfβ ω 2
( (
1, χk ≤ 0 0 Mt ≤ Mt0
fβ ∗ 1+680χ2k F (Mt ) = 2 2
1+400χ2k
χk >0 Mt − Mt0 Mt >Mt0

Mt2 = 2k a
;  a = γRT 
4/15+(Ret /Rβ )4
β ∗ = βi∗ (1 + ζ ∗ F (Mt )); βi∗ = β∞∗
4
 1+(Re t /Rβ )

1+70χω βi∗ ∗
fβ = 1+80χω ; β = βi 1 − βi ζ F (Mt )

Ωij Ωjk Ski
χk = ω1 ∂x ∂k ∂ω
j ∂xj
χ ω = ∗ ω)3
(β∞
∗ ∗
Rk = 6; βi = 0.072; α∞ = 1; α∞ = 0.52; αo = 19 ; β∞ = 0.09; βi = 0.072;

Rβ = 8; Rk = 6; Rω = 2.95; ζ = 1.5; Mt0 = 0.25; σk = 2.0; σω = 2.0;

gradients. The k-ω model has been modified over many years and improvements have been
made to the model. In the k-ω model α∗ depends νt causing low Reynold’s number correc-
tions. At high Reynold’s number α0∗ = αβ∗ = 1 and βi∗ = β∞

and for incompressible flows
(like the current case) β ∗ = βi∗ . A very through description of the changes made to the k-ω
model can be found in the 2006 book by Wilcox. The models for the dissipation terms are
extremely involved and go way beyond the scope of this report. A cursory break down shows
that the destruction of ω (Yω term) is regulated by the fβ (χω ) which is a function of the
mean strain and rotation. An in-depth description of the model and all constants can be
found elsewhere (Fluent user’s guide; Wilcox, 2006).

The Reynolds stress models are not turbulent viscosity models, which means it avoids the
major limitation of the turbulence viscosity hypothesis. For the Reynolds stress turbulence
models, transport equations are solved for the individual Reynolds stresses. Reynolds aver-
aging the Navier-Stokes equations attains the transport equations for the Reynolds stresses.
The turbulent kinetic energy dissipation is also modeled. The downfall of the Reynolds stress
model is that now there are seven equations for three-dimensional flows and five equations
5
C. Q. LaMarche

for two-dimensional flows being solved in order to resolve the turbulence as opposed to two
equations. More equations means that the Reynolds stress models are much more computa-
tionally expensive, but the Reynold’s stress model account for streamline curvature, swirl,
rotation, rapid distortion, and can handle complex flows. In order to close the Reynolds
stress turbulence model, assumptions need to be made about the form of the dissipation
term, the pressure rate of strain tensor (pressure-strain term) and the Reynolds-stress flux
term (Pope, 2000; Fluent User’s Manual).

∂ui uj
(5) + Cij = Dij − Pij + φij − ij + Fij
∂t

Cij = convection of reynolds stress by mean flow.

∂Uk ui uj
(6) Cij =
∂xk

Dij =DT,ij + DM,ij =Diffusion (turbulent (T) and molecular (M))


Pij = is the turbulent stress production
φij =pressure-rate-of-strain-tensor (pressure strain term)

 
∂ p ∂ui uj
(7) Dij = − ui uj uk + (δkj ui + δik uj ) + ν
∂xk ρ ∂xk

Fij = production of Reynolds stress by system rotation.

(8) Fij = −2Ωk (uj um ikm + ui um jkm )

The turbulent viscosity methods are limited by the turbulent viscosity hypothesis, the
Reynolds stress models are limited by the method of modeling the pressure-rate-of-strain
tensor, φij , viscous dissipation, ij and the turbulent diffusion, Dij (Pope, 2000). Fluent uses
6
C. Q. LaMarche

the following form of DT,ij :


 
∂ νt ∂ui uj
(9) DT,ij = σk = 0.82
∂xk σk ∂xk

 
∂Uj ∂Ui
(10) Pij = ui uk + uj uk
∂xk ∂xk

∂ui ∂uj
(11) ij = 2ν
∂xk ∂xk

 
∂ui ∂uj
(12) φij = p +
∂xj ∂ui

The Reynold’s stress models also solves a transport equation for the dissipation rate, .

2
  
D ∂ νt ∂ 1 
(13) = ν+ C1 Pii − C2
Dt ∂xj σ ∂xj 2 k k

Where σ = 1.0, C1 − 1.44, and C2 = 1.92. The turbulent viscosity, νt is computed the
same was as in the equation 2.

The pressure strain tensor in the RSM does not appear in the k equation because the trace
of the tensor is zero. The RSM model accounts for the pressure-rate-of-strain term with
a linear model while the RSMQ model (described below) uses a quadratic pressure stain
assumption. The pressure-rate-of-strain tensor from the Reynolds stress equations serves to
redistribute turbulent energy among the Reynolds stresses.

Tables 5 and 6 provide the Reynolds stress models used in this paper. Table 5 provides the
Reynolds stress model (RSM) employed with a linear pressure-strain model of Gibson and
Launder (1978). Table 6 describes the Reynolds stress model with employed a quadratic
pressure-strain model (RSMQ) (Speziale et al., 1991).

Das könnte Ihnen auch gefallen