Vous êtes sur la page 1sur 5

School of Mechanical Aerospace and Civil Engineering

TPFE MSc Advanced Turbulence Modelling

Introduction

Scalar Transport
T. J. Craft
George Begg Building, C41

In many CFD applications we may be interested in predicting the


behaviour of scalar quantities such as enthalpy, temperature, mass
fraction of a chemical species, etc.
e transported by the flow. We assume that its
Consider a property
evolution can be described by the equation
!
e
e

 e e

e
+S
+
Uj =

t
xj
xj
xj

(1)

e represents any source or sink term that may be present.


where S

Reading:
S. Pope, Turbulent Flows
D. Wilcox, Turbulence Modelling for CFD
Closure Strategies for Turbulent and Transitional
Flows, (Eds. B.E. Launder, N.D. Sandham)
Notes:
Blackboard and CFD/TM web server:
http://cfd.mace.manchester.ac.uk/tmcfd
- People - T. Craft - Online Teaching Material
Scalar Transport

2011/12

e represents the mass fraction of a species being consumed or


If
e may be of great importance and
produced by chemical reaction, then S
has to be modelled from the details of the reaction process.
e represents temperature, then is simply /Pr where Pr is the
If
molecular Prandtl number of the fluid.

Scalar Transport

(2)

Substituting into equation (1) and averaging leads to an equation of the


form




(3)
Uj =
uj + S
+
t
xj
xj
xj

A linear eddy-viscosity model approximates the turbulent stresses by:




Ui Uj
ui uj = (2/3)k ij t
(4)
+
xj
xi

An obvious extension of this is to introduce an eddy-diffusivity for the


scalar fluxes, related to the turbulent viscosity:
ui =

Comparing with the Reynolds averaged Navier-Stokes equations, in this


case the additional terms involve the turbulent scalar fluxes, uj .

In order to close the equation, we need to devise models for these


turbulent scalar fluxes.
For simplicity, we consider the problem of modelling turbulent heat fluxes,
ui , although the principles are easily extended to other scalars.

Scalar Transport

2011/12

2 / 18

e represents the averaged source terms from the original


where S
instantaneous equation.

2011/12

Eddy-Diffusivity Modelling

Equation (1) can be averaged in the same way as the momentum


e into mean and fluctuating parts:
equations. First, we split
e , t) = (x) + (x, t)
(x

1 / 18

e is enthalpy or temperature, the flow is low speed and the effects of


If
e can often be ignored.
viscous heating are negligible, then S

3 / 18

t
t xi

(5)

The total flux terms in the mean temperature transport equation then
become




t
+
xk
Pr t xk

The turbulent Prandtl number, t , is often assumed to be constant.

In a near-wall flow, t = 0.9 is usually adopted. However, in free flows a


lower value (around 0.7) is more appropriate.

Scalar Transport

2011/12

4 / 18

Homogeneous Shear Flow

Scalar Fluxes in Simple Shear

Heated Plane Channel

U(y)

An exact equation can be derived for ui , which shows it has a


generation term of the form

(y)
y
x

Ui
uk
Pi = ui uk
xk
xk

U(y)
T(y)

y
x

In simple shear flow U = U(y ), = (y ), this gives:


P1
P2

d
dU
v
dy
dy
d
= v 2
dy

DNS: Horiuti (1992)

U(y)

= uv

(y)

Expts: Tavoularis & Corrsin (1981)


y

This suggests both the wall-normal and wall-parallel scalar fluxes will be
non-zero.
Measurements and DNS data show that generally |u | > |v |,
particularly in the near-wall region.

Scalar Transport

Heated Axisymmetric Jet

5 / 18

and

Scalar Transport

2011/12

u = 0

A somewhat better model for the scalar fluxes arises from the Daly &
Harlow (1970) Generalised Gradient Diffusion Hypothesis (GGDH):

ie. zero turbulent scalar flux in the streamwise direction.

ui = c

Making the usual boundary layer approximations, the (steady) mean


scalar transport equation becomes



U
+V
=
v

x
y
y
y

Consequently, if streamwise gradients are small compared to


cross-stream ones, the above error in u may not have serious
consequences.

However, the error may be more serious in buoyancy-influenced or other


complex flows.

2011/12

7 / 18

ui uj

xj

In a simple shear flow this gives


v = c

Scalar Transport

6 / 18

Generalized Gradient Diffusion Modelling

However, the eddy diffusivity model gives:

t d
t dy

Expts:
Chevray & Tutu (1978)
2011/12

v =

k 2 d
v

dy

and

u = c

k
d
uv

dy

This does at least give a non-zero u , and also better reflects the
dependence of v on v 2 , seen in the generation terms.

If a stress transport model or non-linear EVM is used which gives good


stress anisotropy levels, the above form can be used as is (typically with
a coefficient c around 0.3).

Scalar Transport

2011/12

8 / 18

With an EVM (which will not generally provide good normal stress
predictions), the GGDH can still be used with some modification.

Ince & Launder (1989) used this model in conjunction with a linear
eddy-viscosity model for the stresses, taking c = (3/2)(c /t ).

In a shear flow (where the model returns

v2

Buoyancy Effects

One important application where temperature (or concentration)


differences must be properly accounted for is in buoyancy-affected flows.

In that case, there is an additional source term in the momentum


equation:
D( Ui )
= . . . + gi
Dt

= (2/3)k ) this gives

c k 2 d
k
d
=
v = c v 2

dy
t dy

k
d
u = c uv

dy
which gives v as in the eddy-diffusivity model, but should give a better
approximation of u .

Scalar Transport

9 / 18

Scalar Transport

(9)

Gk = gi ui

where

(10)

Note that the buoyant generation of k depends on the turbulent scalar


fluxes.

Depending on the sign of v , buoyancy can either enhance or reduce k


levels:

Scalar Transport

2011/12

10 / 18

The buoyancy-related term appearing in the transport equation for ui


leads to an additional generation term in the stress transport equations:

The buoyancy source term in the k equation then becomes

(8)

Buoyancy in Stress Transport Models

Density fluctuations are often expressed in terms of temperature


fluctuations, by introducing the thermal expansion coefficient:

(7)

Consequently, one gets an additional generation term in the turbulent


kinetic energy equation:
Gk = (1/ ) gi ui

2011/12

1
=

More complex algebraic scalar flux models have been proposed, some
developed along similar routes to the non-linear eddy-viscosity models
examined earlier for the Reynolds stresses.

This results in an additional term in the transport equation for the


fluctuating velocity:
Dui
= . . . + gi /
Dt
where is the fluctuating density.

(6)

In a stably stratified layer (with d/dy > 0) we would typically have


v < 0 and hence (with g2 < 0), Gk is negative.
In an unstable layer, d/dy < 0, and hence v > 0 and Gk is
positive.
2011/12

(11)

Gij = gi uj gj ui

(12)

When modelling the pressure-strain redistribution process, a contribution


due to buoyancy should also be included:

ij = ij1 + ij2 + ij3

(13)

Within the framework of the modelling adopted earlier, ij3 is


approximated in a similar manner to ij2 :

ij3 = c3 (Gij (1/3)Gkk ij )

11 / 18

Dui uj
= Pij + Gij + ij ij + dij
Dt

(14)

ij3 is thus assumed to redistribute the buoyancy generation.

Scalar Transport

2011/12

12 / 18

In some instances it may be sufficient to model the scalar fluxes ui


using a GGDH approach (or similar) as described earlier.

However, in strongly buoyant flows this may not be adequate, since the
scalar fluxes are themselves affected by buoyancy.

Some extended algebraic heat flux models have been proposed,


incorporating some buoyancy effects (eg. Hanjalic et al, 1996).

Another option is to adopt a full second-moment closure, solving


transport equations for the scalar fluxes also.

Other terms in the transport equation have to be modelled.

Similar assumptions are typically made for these as for the corresponding
terms in the stress transport equations. However, the details will not be
covered in this course.

The scalar flux buoyancy generation depends on the scalar variance, 2 .

This can be obtained by solving its transport equation, of the form


D 2
= P 2 + diffusion
Dt

Scalar flux transport equations can be derived in a similar manner to


those for the Reynolds stresses. The result can be written in the form
Dui
= Pi + Gi + i i + di
Dt

where the generation rate is given by


(15)

The generation terms Pi and Gi are exact:


Pi = ui uk

(16)

Gi = gi 2

(17)

Scalar Transport

P = 2 ui

Ui

uk
xk
xk

2011/12

13 / 18

Negatively-Buoyant Jet

xi

The dissipation rate is usually either modelled algebraically by


assuming that thermal and dynamic timescales are related (eg.
R = (k / )(2 / 2 ) being constant), or obtained from its own modelled
transport equation.

Scalar Transport

2011/12

14 / 18

Opposed/Buoyant Wall Jet

Axisymmetric downward
directed buoyant jet.

Experiment of Cresswell et al
(1989).

Velocity and shear stress


profiles one diameter
downstream of jet discharge.

2011/12

Isothermal and buoyant cases studied.

LES data from Addad et al (2004).

Requires a good outer flow and near-wall


modelling for accuracy.

TCL RSM; Basic RSM

Scalar Transport

(18)

15 / 18

Scalar Transport

Vertical
velocity
contours.

2011/12

16 / 18

Stably Stratified Mixing Layer

References
Addad, Y., Benhamadouche, S., Laurence, D., (2004) The negatively buoyant jet: LES

Studied experimentally by
Uittenbogaard (1998).

results, Int. J. Heat and Fluid Flow, vol. 25, pp. 795-808.
Chevray, R., Tutu, N.K., (1978) Intermittency and preferential transport of heat in a round jet,

U1 = 0.5 m/s, 1 = 1015 kg/m3

U2 = 0.3 m/s, 2 = 1030 kg/m3

J. Fluid Mech., vol. 88, p. 133.


Cresswell, R., Haroutunian, V., Ince, N.Z., Launder, B.E., Szczepura, R.T., (1989)

Measurement and modelling of buoyancy-modified elliptic turbulent shear flows, Proc. 7th
Turbulent Shear Flows Symposium, Stanford University.
Daly, B.J., Harlow, F.H., (1970) Transport equations in turbulence, Phys. Fluids, vol. 13, pp.

Scalar Transport

2634-2649.

Linear k - scheme overpredicts


turbulence levels and hence mixing.

Hanjalic, K., Kenjeres, S., Durst, F., (1996) Natural convection in partitioned two-dimensional

enclosures at higher Rayleigh numbers, Int. J. Heat Mass Transfer, vol. 39, pp. 1407-1427.
Horiuti, K., (1992) Assessment of two-equation models of turbulent passive-scalar diffusion

Second-moment closures do better at


capturing buoyancy effects on scalar
fluxes and hence reduce mixing.

in channel flow, J. Fluid Mech., vol. 238, pp. 405-433.


Ince, N.Z., Launder, B.E., (1989) On the computation of buoyancy-driven turbulent flow in

rectangular enclosures, Int. J. Heat Fluid Flow, vol. 10, pp. 110-117.
Tavoularis, S., Corrsin, S., (1981) Experiments in nearly homogeneous turbulent shear flow

In this case, as turbulence levels


decrease, modelling of diffusion
becomes more influential.
2011/12

with a uniform mean temperature gradient. Part 1., J. Fluid Mech., vol. 104, pp. 311-347.
Uittenbogaard, R.E., (1998) Measurement of turbulence fluxes in a steady, stratified, mixing

layer, Proc. 3rd Int. Symposium on Refined Flow Modelling and Turbulence Measurements,
Tokyo.
17 / 18

Scalar Transport

2011/12

18 / 18

Vous aimerez peut-être aussi