You are on page 1of 11

Engineering Applications of Computational Fluid Mechanics Vol. 1, No.4, pp.

314324 (2007)

SIMULATION OF TURBULENT HEAT TRANSFER


IN JET IMPINGEMENT OF AIR FLOW ONTO A FLAT WALL
M. Raisee*+, A. Noursadeghi*, B. Hejazi*, S. Khodaparast* and S. Besharati**
* Department of Mechanical Engineering, Faculty of Engineering, University of Tehran,
P.O. Box: 11155/4563, Tehran, Iran
+
E-Mail: mraisee@ut.ac.ir (Corresponding Author)
** Mechanical Machinery Department, Shahrekord University, Shahrekord, Iran
ABSTRACT: This paper deals with the numerical prediction of a turbulent jet impinging orthogonally onto a large
heated flat wall. The Reynolds number based on the jet diameter and the bulk velocity is 2.3 10 3 , and the jet discharge
is two diameters above the heated plate ( H / D = 2 ). The main objective has been to examine the suitability of recently
developed variants of a cubic nonlinear k- model for the prediction of impinging jet flows. The numerical approach
used in this study is the finite-volume method together with the SIMPLE algorithm. For the modeling of turbulence, the
Launder and Sharma low-Re k- model (1974) and a new version of the nonlinear low-Re two-equation model, that has
also been recently shown to produce reliable thermal predictions in impinging jet flows and also flows through pipe
expansions (Craft et al., 1999), have been employed. The numerical results show that the low-Re k- model returns
unrealistically high levels of turbulence energy around the stagnation point and consequently over-predicts the heat
transfer coefficients in this region. The introduction of a differential form of the turbulent length-scale correction term to
the dissipation rate equation improves the thermal predictions of the low-Re k- model. Moreover, the nonlinear k-
model produces superior flow and heat transfer predictions compared with the linear k- model. The results of the
present investigation agree with those reported in Craft et al. (1999).
Keywords:

turbulent heat transfer, impinging jet, turbulence modeling

not parallel to the wall. In this paper a numerical


investigation has been undertaken to assess how
effectively a recently modified nonlinear k- model
can reproduce flow and heat transfer characteristics
in an impinging jet.
In recent years numerous numerical investigations
have been undertaken in order to assess the
capabilities of the well-known turbulence models in
predicting flow and heat transfer characteristics in
turbulent impinging jet. For example, Craft et al.
(1993) examined the predictive capabilities of one
k- model and three second moment closure models
in the prediction of experimental data reported by
Baughn et al. (1992) and Cooper et al. (1993). Their
numerical predictions indicated that the widely used
low-Re number k- model substantially overpredicts the heat transfer rates around the stagnation
point, whilst the second moment closure models
produced better predictions near the stagnation
point. More recently, Behnia et al. (1999) used an
elliptic relaxation model ( v 2 f model) to simulate
heat transfer in several confined and unconfined
impinging jet configurations. The heat transfer

1. INTRODUCTION
An axi-symmetric jet flow impinging orthogonally
onto a flat heated surface is used in a wide range of
industrial applications. This is because it produces
high heat transfer rates at the stagnation point.
Detailed information about flow and heat transfer in
this flow can be very useful for the optimum design
of the variety of engineering instruments such as the
paper dryers and textile, and the impingement
cooling system of gas turbines and electronic
devices. To obtain such information experimentally
can be very difficult and expensive. The only other
way is to solve the governing equations of the flow
and energy. However, flows in impinging jets are
mostly turbulent and their predictions require the
use of turbulence models. The impinging jet flow is
also important from a turbulence modeling point of
view. This is due to the fact that the existing
turbulence models are mostly validated against
flows parallel to the wall; thus, they may not be
able to produce accurate predictions for the
impinging flows whose streamlines are basically
Received: 20 Apr. 2007; Revised: 4 Jul. 2007; Accepted: 6 Jul. 2007
314

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

predictions of the v 2 f model were in close


agreement with the experimental data. The main
objective of present study is to examine the
capabilities of the modified cubic low-Reynoldsnumber two equation model, proposed by
Craft et al. (1999), in the prediction of flow and
thermal fields in an axi-symmetric turbulent jet
impinging on a constant heat flux flat plate.

Nu =

A schematic diagram of the jet impingement is


shown in Fig. 1. For this geometry, the mean flow
data were reported by Cooper et al. (1993) for a
H / D ratio of 2 and a Reynolds number based on
the bulk velocity ( U b ) and jet diameter (D) of
2.3 10 4 . Corresponding thermal field data have

been reported by Baughn et al. (1992) in the form


of local Nusselt number distribution. Here the
Nusselt number is defined as:

Computational Domain

D
r

2L
Fig. 1

(1)

where D is the diameter of the jet, the thermal


conductivity of air, q w the wall heat flux, w the
wall temperature and ref the bulk temperature at
the exit of the jet.
As shown in Fig. 1, the upper boundary of the
computational domain was placed at 2 jet diameter
above the end of the pipe ( H / D = 2.0 ). Over the
remainder of the upper boundary and at the righthand outflow boundary, a zero gradient boundary
condition was applied. At the right-hand side and
bottom boundaries, symmetric and non-slip
boundary conditions were imposed respectively.
For heat transfer calculations, the temperature at the
jet inlet was assumed to be uniform, while a zero
temperature gradient was imposed on the remainder
upper boundary as well as at the right- and left-hand
sides of the computational domain. A constant heat
flux thermal boundary condition was applied at the
lower boundary of the computational domain.

2. GEOMETRY OF FLOW INVESTIGATED


AND BOUNDARY CONDITIONS

q w D
( w ref )

Configuration and computational domain for the impinging jet on a flat plate Re = 23,000 and H/D = 2.

315

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

3. MATHEMATICAL FORMULATION

where the turbulent viscosity, t, is obtained from:

All the flow equations are presented in Cartesian


tensor notation.

k2
t = C f ~

3.1

and the value of constants C and are given in


Table 1.

Mean flow equations

For a steady incompressible flow, the conservation


laws of mass, momentum and energy may be
written as:
y

x j
y

=0

C
0.09

(2)

x j

1 P

+
x i x j

U i

x u i u j
j

(3)

x j

=
x j

Pr x u j
j

k
+ Pk ~ 2
x i

(4)

Turbulence modeling equations

~ = 2 k
x
j

Pk = u i u j

t
x i

(8)

(9)

(10)

U i
x j

(11)

The damping functions f , f 1 and f 2 are given by:


~
f = exp[ 3.4 (1 + 0.02R t ) 2 ]

In this turbulence model the Reynolds stresses and


heat fluxes are obtained from the eddy-viscosity
and eddy-diffusivity approximations respectively:

ui =

0.9

and Pk is the generation rate of turbulent kinetic


energy obtained from:

3.2.1 Linear low-Reynolds-number k- model


(1974)

1.3

homogeneous
dissipation rate which can be related to the real
dissipation rate through:

The turbulence models employed are the widely


used Launder and Sharma (1974) low-Reynoldsnumber k- model and a new version of the
nonlinear low-Reynolds-number k- (Craft et al.,
1999). Computations with the low-Re k- model
have been carried out with the originally proposed
(algebraic) Yap correction term (Yap, 1987) and
also the new version of differential form (NYP)
which is free from any explicit wall distance
(Iacovides and Raisee, 1999).

U j
U
u i u j = 2 3 ij k t i +
x i
x j

k
1.0


( U i ~ ) =
+ t

x i
x i
x i
~
~ 2
+ C 1f1 Pk C 2 f 2
+ E + S
k
k
where the variable ~ is the

where , and Pr are respectively, the density, the


kinematic viscosity, and the Prandtl number of the
fluid.
3.2

C2
1.92


(U i k ) =
+ t

x i
x i
k x i

Energy
(U j )

C1
1.44

To obtain t, transport equations for the turbulence


kinetic energy, k, and its dissipation rate, ~ , are
solved. The transport equations for turbulent kinetic
energy and dissipation rate are written as:

Momentum
(U j U i )

Table 1 Empirical constants for the k- model.

Continuity
U j

(7)

f1 = 1

~
f 2 = 1 0.3 exp R 2t

(5)

(12)

where R t = k 2 ~ is the local turbulent Reynolds


number.
The model constants are given in Table 1.

(6)

316

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

The extra source term, S , stands for the Yap


correction term which is discussed in the later
sections.

The term E was first introduced by Jones and


Launder (1972) and is expressed as:
2Ui
E = 2 t
x j x k

(13)

3.2.2 Nonlinear low-Reynolds-number k- model (1999)


In this turbulence model, turbulent stresses are obtained via the constitutive relation:
k
k
u i u j = 2 3 k ij t S ij + C1 ~t (S ik S kj 1 3 S kl S kl ij ) + C 2 ~t ( ik S kj + jk S ki )

tk
tk2
+ C 3 ~ ( ik jk 1 3 lk lk ij ) + C 4 2 (S ki lj + S kj li )S kl
~

+ C5

tk2
( il lmS mj + Sil lm mj 2 3 Slm mn nl ij )
~ 2

+ C6

tk2
tk2
S
S
S
S
C
+
ij
kl
kl
7
~ 2 ij kl kl
~ 2

(14)

where Sij and ij are strain and vorticity rate


tensors:
U j
U
S ij = i +
x i
x j

U j

U
, ij = i

j x i

C =

(15)

-0.1

0.1

C3

C4

C5

C6

0.26

10 C 2

5C 2

5C 2

{1 exp[ 0.36 exp(0.75)] }

~ k
~ k
S = ~ 0.5SijSij , = ~ 0.5 ij ij

~ ~
and = max(S, ) .

(16)

(17)

Due to the strong dependence of Eq. (16) on the


strain rate, the use of the above C expression in the
computation of flows over sharp corners results in
instability problems. This problem has been
recognized by both Raisee (1999) and Cooper
(1997) in their computation of similar flows. To
remedy these instability problems, the following
form of C function was proposed by Craft et al.
(1999):

Table 2 Values of coefficients in the nonlinear k-


model.
C2

1 + 0.351.5

with

The turbulent heat fluxes, u i , are modeled using


the simple eddy-diffusivity approximation (Eq. (6)).
The model coefficients, C1 to C7, have been
calibrated by Craft et al. (1996), by reference to
several flows, including homogeneous shear flows,
swirling flows and curved channel flows. The
values of these coefficients are given in Table 2.

C1

0.3

12

C = min 0.09,

+
1
3
.
5
f

RS

The k and ~ transport equations and eddy-viscosity


formulation are similar to those of Lauder and
Sharma low-Re k- model; however, the following
modifications are proposed by Craft et al. (1996 &
1999).

(18)

where
~
f RS = 0.235[max(0, 3.333)]2 exp( R t 400)

(19)

2) Near wall damping


In the nonlinear two equation model the viscous
damping function of t is provided by the function
f:

1) Modeling of C
In the original nonlinear eddy-viscosity model C is
~
a function of the strain and vorticity invariants S
~
and .
317

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)


~ 12
~ 2
R
R t
t


f = 1 exp
90
400

where the quantity is defined in the same way as


but, to enhance stability, the Kolmogorov
timescale is used as a lower limit on the timescale
~
~
k ~ employed in the expressions for S and , in
a manner similar to the proposal of Durbin (1991):

(20)

The near-wall source term E is expressed as:


~

S k 2
0.0022 ~t
E=

2Ui

x x
k l

~
R t 250
~
R t > 250

(21)

~
S = max k ~ ,

1
Sij Sij
2

1
ij ij
2

~
= max k ~ ,

(27)

3) Length-scale correction terms


In separated flows, the near-wall length-scale
becomes too large, resulting in excessively high
levels of near-wall turbulence. To remedy this
behavior, Yap (1987) introduced an extra source
term into the dissipation rate equation which is
based on the wall distance y:

The limited R t dependent damping is included for


numerical stability.

~ 2
max (l l e 1)(l l e ) 2 ,0
S = Yap = 0.83
k

The general form of the governing equations of


mean flow, temperature and turbulence fields may
be written as:

4. NUMERICAL METHOD

(22)

where l is the turbulent length-scale k 3 2 ~ , the


equilibrium length-scale l e = 2.55y and y is the
distance to the wall.
To eliminate the dependence of the above source
term on the wall distance, a differential form of the
length-scale correction was proposed by Iacovides
and Raisee (1999) is adopted:
~ 2

S = NYP = max C F(F + 1) 2


,0
k



(rc U ) + (rc V ) = rc + rc + rc S
x
y
x
x y
y

(28)
where x and y represent the coordinates in the
stream-wise and radial directions respectively, rc is
the radius of curvature. is an effective diffusion
coefficient and S denotes the source term.

(23)

where
F=

{[( l x j )( l x j )] 1 2 dl e dy} C l

(24)

represents the difference between the predicted


length-scale gradient, with l = k 3 2 ~ , and the
equilibrium length-scale gradient, dl e/dy, is
defined by:

(1)

P
W

w (2)

~
~
~
dl e dy = C l [1 exp( B R t )] + B C l R t exp( B R t )

(25)
y

where C l = 2.55 , B = 0.1069 and C = 0.83 .


In order to reduce the amount of correction in the
regions of high , a modified version of the
Iacovides and Raisee differential correction term
was proposed by Craft et al. (1999), where the
coefficient C is taken as:
~
0.83 min(1, R t 5)
C =
~
[0.8 + 0.7( 3.33) 4 exp( R t 12.5)]

Fig. 2

A partially staggered grid arrangement:


1) velocity cell; 2) pressure cell.

In the present study the above transport equation is


solved using Finite-Volume methodology in a semistaggered grid system, which is shown in Fig. 2. In
this arrangement, both velocity components (U, V)
are located at the same nodal position which is

(26)

318

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

side of the discretized transport equation, making a


positive contribution to the diagonal coefficient AP.
The difference between the generation rate of the
turbulent kinetic energy Pk and the dissipation rate
~ is also similarly treated when is negative:

staggered in relative to the pressure nodes. All the


Reynolds stresses and scalars are stored at the
pressures nodes. The mean velocity gradient terms
that appear in the equations for k, and the
Reynolds stresses are thus discretised at the scalar
nodal locations from the four surrounding velocity
nodes. The Hybrid differencing scheme is used for
the approximation of the convective terms. The
pressure field is linked to that of velocity through
the well-known SIMPLE, pressure correction
algorithm. To avoid stability problems associated
with pressure-velocity decoupling, a form of the
Rhie and Chow (1983) interpolation scheme,
suitable for a semi-staggered mesh, is also
employed. The details of the above mentioned
numerical methods are described in Raisee (1999).
The fact that a low-Reynolds-number nonlinear
model of turbulence was used necessitated the
introduction of several stabilization measures. In
the discretization of the mean momentum
equations, the turbulent stresses are decomposed
into two components: the enhanced linear
component which, when possible, includes
contributions from the last two cubic terms, and the

A P = A P + [ 2 k 1 2 x j

)2P min (Pk ~ , 0) P ] Vol

k oP

(33)
S u = max(Pk ~ , 0) P Vol

(34)

where k oP is the existing value of k at node P, and


Vol is the cell volume.
In the discretization of the transport equation for the
dissipation rate, the difference between the
generation and destruction rate of ~ is also
absorbed into the diagonal coefficient AP when is
negative:
Vol
A P = A P min(C 1 Pk C 2 ~ , 0 ) P o
k P
Su = S u + max(C 1Pk C 2 ~ , 0) P

higher order component, u i u j , defined by:

~ o
P
koP

(35)
(36)

Vol

0.24

u i u j = 2 k ij t Sij + u i u j
3

(29)

0.2

with
Nu/(Re0.7Pr0.4)

0.16

t = t t k ~ 2 min[(C 6S klS kl + C 7 kl kl ),0]

(30)

This decomposition was arrived at by noting that


the linear and the final two cubic terms in Eq. (14)
can be written as:

0.08

0.04

2
t t k ~ 2 (C 6 S klS kl + C 7 kl kl ) Sij

(31)
0

The enhanced linear component is then directly


absorbed into the turbulent diffusion, employing an
effective viscosity ( + t ) , and only the gradients
of the remaining nonlinear components then appear
explicitly in source term SU. The mean momentum
equation is thus written as:

(U i U j ) = 1 P + ( + t ) Sij u i u j
x j
x i x j

0.12

Fig. 3

r/D

Predicted Nusselt number distribution using the


linear k- model.
Exp. Data
60 50
120 120
180 150

Computations were performed on a fine mesh


consisting of 180 150 grid nodes in radial and axial
directions respectively. The first grid point is
located at y 0.1 (y = y k ) . To assess the
accuracy of the numerical results, a series of gridindependency tests have been performed using two
additional grids: a coarse mesh with 60 50 nodes

(32)

In the discretization of the k transport equation the


near-wall dissipation term, 2(k 1 2 x j ) 2 , is
always divided by the existing (previous iteration)
value of k at node P and transferred to the left-hand
319

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

and a medium mesh with 120 120 nodes. As can be


observed in Fig. 3, though grid refinements from
the coarse grid to the medium grid caused some
differences in the heat transfer predictions,
computations on the fine mesh (180 150 ) showed
no significant difference in the local Nusselt
number distribution with those obtained on the
medium grid. Consequently, the results obtained on
the 180 150 mesh are regarded as grid
independent. All the flow and heat transfer
computations, presented in the subsequent section,
have been obtained on the 180 150 mesh.

4
3

Y/D 2
1
0

Fig. 5

Figures 4 and 5 respectively show the predicted


velocity vectors and streamlines using the low-Re
k- model with the Yap correction term. The
nonlinear k- model streamlines and velocity
vectors were similar and thus are not shown here.
Also the replacement of the Yap term with the
NYP term in the dissipation rate equation results
in similar streamlines and velocity vectors. From
these figures, one can see that at the jet exit the
streamlines are parallel and air leaves the jet with a
fully-developed velocity profile. As the stagnation
point is approached, flow decelerates in the axial
direction and turns, as exhibited by velocity vectors
and by the sharp curvature in the streamlines.
Downstream of the stagnation point, a radial wall
jet parallel to the plate begins to form with a
developing boundary layer. Ambient fluid outside
the free jet is entrained into the core with a
developing shear layer separating the core and the
ambient fluid. The entrainment is evident by the
curving of the streamlines outside the pipe towards
the core of the jet.
4
3

Y/D 2
1

Fig. 4

r/D

r/D

10

Predicted streamlines using the linear k-


model.

Having discussed the main flow features, attention


is now directed to the flow and thermal field
comparisons. Fig. 6 compares the measured streamwise velocity fluctuation at two radial locations
( r / D = 0.5 and 1. ) with the predictions obtained
from the low-Re k- models. It appears that the
low-Re k- model with either form of length-scale
correction terms produces very similar results,
though close to the wall the NYP term predicts
slightly lower turbulence levels. Both models are
seen to produce excessive levels of turbulence
energy, leading to values of the stream-wise
velocity fluctuations up to three times as large as
those measured. This predictive deficit is due to the
fact that the low-Reynolds-number k- model uses a
simple
linear
eddy-viscosity
stress-strain
relationship, which leads to the generation of large
turbulence energy in the irrotational region close to
the stagnation point. Comparisons of the predicted
cross-stream velocity fluctuation with the measured
data in Fig. 7 indicate that the low-Re k- model
also fails to reproduce the experimentally observed
cross-stream turbulence intensity profiles at both
radial locations.
Comparisons of the predicted velocity magnitude
with measured values are presented in Fig. 8. While
at r / D = 0.5 a good agreement with the
experimental is noted, at r / D = 1.0 the lowReynolds-number k- models under-predicts the
measured values close to the wall but over-predicts
them away from the wall. The corresponding heat
transfer predictions, shown in Fig. 9, indicate that
the low-Re k- model with the Yap term
significantly overestimates the heat transfer levels
in the stagnation region which, as shown in figures
6 and 7, is due to high levels of turbulence
intensities predicted in this region. It is observed
that the reduction of the near-wall turbulence
energy due to the inclusion of the NYP term in
the dissipation rate equation, somewhat improves
the Nusselt number predictions. Note, however, that

5. PRESENTATION AND DISCUSSION OF


RESULTS

10

Predicted velocity vectors using the linear k-


model.

320

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

at

downstream

of

the

stagnation

point

at

number
distribution.
Further
downstream
at r / D > 3 , where flow develops parallel to the wall,
the heat transfer predictions of the low-Re k-
models are close to the measure values.

1 / 2 < r / D < 2 , where flow turns from the vertical

direction to the radial, the NYP correction term


does not capture a local maximal in the Nusselt
0.3

0.3

r/D=0.50

r/D=1.0
0.2

0.1

0.1

u /U b

0.2

0.1

Fig. 6

0.2

0.3

0.4

Y/D

0.1

0.2

0.3

0.4

Comparisons of the predicted and measured stream-stream turbulence intensity.

0.3

0.3

r/D=0.50

r/D=1.0
0.2

0.1

0.1

v//U b

0.2

0.1

Fig. 7

0.2

0.3

0.4

Y/D

0.3

0.4

0.8

0.6

0.6

0.4

0.4

0.2

0.2
0

0.1

0.2

Fig. 8

r/D=1.0

0.8

0.3

0.4

Y/D

0.1

0.2

0.3

Comparisons of the predicted and measured mean velocity.


0.24

0.2

0.16

Nu/(Re0.7Pr0.4)

2 0.5
2

0.2

1.2

r/D=0.5

0.1

Comparisons of the predicted and measured cross-stream turbulence intensity.

1.2

(U +V ) /U b

0.12

0.08

0.04

Fig. 9
Exp. Data

r/D

Comparison of the predicted and measured Nusselt number.


Linear k- model with the Yap

321

Linear k- model with the NYP

0.4

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

0.3

0.3

r/D=0.0

r/D=1.0
0.2

0.1

0.1

u//U b

0.2

0.1

0.2

0.3

0.4

Y/D

0.1

0.2

0.3

0.4

Fig. 10 Comparisons of the predicted and measured stream-wise turbulence intensity.


0.3

0.3

r/D=0.50

r/D=1.0
0.2

0.1

0.1

v //U b

0.2

0.1

0.2

0.3

0.4

Y/D

0.1

0.2

0.3

0.4

Fig. 11 Comparisons of the predicted and measured cross-stream turbulence intensity.


1.2

1.2

r/D=0.0

(U 2+V 2)0.5/U b

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0.1

0.2

r/D=1.0

0.8

0.3

0.4

Y/D

0.1

0.2

0.3

0.4

Fig. 12 Comparisons of the predicted and measured mean velocity.


0.24

0.2

Nu/(Re0.7Pr0.4)

0.16

0.12

0.08

0.04

r/D

Fig. 13 Comparison of the predicted and measured Nusselt number.


Exp. Data

Linear k- model

Nonlinear k- model

worth mentioning that in both set predictions,


computations were performed with the inclusion of
the NYP term in the dissipation rate equation. It
is clearly seen that the use of the nonlinear k-
model significantly improves turbulence field

Attention is now focused on the predictive


capabilities of the low-Re nonlinear k- model. In
Fig. 10, the predicted stream-wise turbulence
intensity returned by the linear and nonlinear k-
models are compared with the measured data. It is
322

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

predictions in comparison to the linear k- model.


This superior predictive ability of the nonlinear k-
model is also seen in Fig. 11 which shows that the
profiles of the normal turbulence intensity retuned
by the nonlinear model are now much closer to the
experimental data than those of the linear k-
model. The profiles of magnitude of velocity vector
computed by the linear and nonlinear k- models
are compared with the experimental data in Fig. 12.
It is noted that at the stagnation streamline
( r / D = 0.0 ), both predictions are consistent with the
experiment. However, within the wall jet boundary
layer at r / D = 1.0 , the predicted velocities with the
nonlinear k- model are closer to the experimental
data.
Finally, Fig. 13 presents comparisons between the
predicted and measured local Nusselt numbers.
While the linear k- model overestimates the heat
transfer levels around the stagnation point and
underestimates them in the wall jet boundary layer,
the nonlinear k- model significantly improves the
heat transfer predictions in these regions. As
mentioned earlier the low-Re k- model fails to
predict a second peak in the local Nusselt number
distribution; however, this feature is somewhat
reproduced by the nonlinear k- model. It is worth
mentioning that the flow and heat transfer
predictions of present investigation using the linear
and nonlinear k- models agree with those reported
by Craft et al. (1999) for the same test case and
using similar turbulence models.

consequently, more accurately reproduces the


variation and the levels of local Nusselt number.
NOMENCLATURE
D
h

uiu j

diameter of the jet


height of the jet pipe within the
computational domain
jet-to-plate distance
dimension of the computational domain in
radial direction
Nusselt number
differential form of the Yap term
molecular Prandtl number
local surface heat flux on the wall
Reynolds number
bulk velocity in the jet
mean velocity components (U, V)
fluctuating velocity components (u, v)
stream-wise rms
cross-stream rms
Reynolds stress tensor

u i

turbulent heat fluxes

NYP
Yap
xi
y

differential form of the Yap term


original, algebraic form of the Yap term
Cartesian coordinates (x, y)
distance from the wall

H
L
Nu
NYP
Pr
qw

Re
Ub
Ui
ui
u
v

Greek Symbols

6. CONCLUSIONS

ij
~

This study has considered the application of lowReynolds-number linear and nonlinear eddyviscosity models to numerical prediction of flow
and heat transfer of an axi-symmetric turbulent jet
impinging on a flat wall under a constant heat flux.
Comparisons of the numerical predictions with the
measured data indicate that the linear low-Re k-
model produces high levels of turbulence energy
around the stagnation point and, as a result,
overestimates the heat transfer rates by about 100%.
Although the use of the NYP term somewhat
improves the thermal prediction of the linear k-
model, this model still overestimates the heat
transfer rates around the stagnation point while
underestimating them in the boundary layer
downstream of the stagnation point. The recently
modified version of the nonlinear k model
produces improved flow field predictions and

ref
w

Kronecker delta
isotropic dissipation rate
temperature
averaged fluid temperature at the jet exit
wall temperature
thermal conductivity
fluid dynamic viscosity
fluid kinematic viscosity
fluid density

REFERENCES
1. Baughn J, Yan X and Mesbah M (1992). The
effect of Reynolds number on the heat transfer
distribution from a flat plate to a turbulent
impinging jet. ASME Winter Annual Meeting.

323

Engineering Applications of Computational Fluid Mechanics Vol. 1, No. 4 (2007)

13. Rhie CM and Chow WL (1983). Numerical


study of the turbulent flow past an airfoil with
trailing edge separation. AIAA J. 21(11):1525
1532.
14. Yap CR (1987). Turbulent heat and momentum
transfer in recirculating and impinging flows.
Ph.D. Thesis, Faculty of Technology,
University of Manchester.

2. Behnia M, Parneix S, Shabany Y and Durbin


PA (1999). Numerical study of turbulent heat
transfer in confined and unconfined impinging
jets. Int. J. Heat and Fluid Flow 20(1):19.
3. Craft T, Graham L and Launder BE (1993).
Impinging jet studies for turbulence model
assessment-II. An examination of the
performance of four turbulence models. Int. J.
Heat Mass Transfer 36(10):26852697.
4. Craft TJ, Iacovides H and Yoon JH (1999).
Progress in the use of non-linear two-equation
models in the computation of convective heat
transfer in impinging and separated flows.
Flow, Turbulence and Combustion 63:5980.
5. Craft TJ, Launder BE and Suga K (1996).
Development and application of a cubic eddy
viscosity model of turbulence. Int. J. Heat Fluid
Flow 17:108115.
6. Cooper D (1997). Computation of momentum
and heat transfer in a separated flow using lowReynolds number linear and non-linear k-
models. MRes Dissertation, Department of
Mechanical Engineering, UMIST.
7. Cooper D, Jackson DC, Launder BE and Liao
GX (1993). Impinging jet studies for turbulence
model assessment-I. Flow-field experiments.
Int. J. Heat Mass Transfer 36(10):26752684.
8. Durbin PA (1991). Near-wall turbulence
closure modeling without damping function.
Theoretical Computational Fluid Dynamics
3(1):113.
9. Iacovides H and Raisee M (1999). Recent
progress in the computation of flow and heat
transfer in internal cooling passages of gas
turbine blades. Int. J. Heat Fluid Flow 20:320
328.
10. Jones WP and Launder BE (1972). The
prediction of laminarization with a twoequation model of turbulence. Int. J. Heat Mass
Transfer 15(2):301314.
11. Launder BE and Sharma BI (1974). Application
of the energy dissipation model of turbulence to
the calculation of flow near a spinning disc.
Letters in Heat Mass Transfer 1(2):131138.
12. Raisee M (1999). Computation of flow and heat
transfer through two- and three-dimensional
rib-roughened
passages.
Ph.D.
Thesis,
Department of Mechanical Engineering,
UMIST.

324

You might also like