You are on page 1of 10

CFD PREDICTION FOR CONFINED IMPINGEMENT JET HEAT TRANSFER

USING DIFFERENT TURBULENT MODELS


Y.Q. Zu
1*
, Y.Y. Yan
1
, J.D. Maltson
2
1.
School of the Built Environment, the University of Nottingham NG7 2RD, UK
2.
Siemens Industrial Turbomachinery, Lincoln LN5 7FD, UK
ABSTRACT
The flow and heat transfer characteristics of a confined circular air jet vertically impinging on a flat plate are
numerical analyzed base on the CFD commercial code FLUENT 6.1.18. The relative performance of seven
versions of turbulent models, including the standard c k model, the renormalization group c k model, the
realizable c k model, the standard e k model, the Shear-Stress Transport e k model, the Reynolds stress
model and the Large Eddy Simulation, for the prediction of this type of flow and heat transfer is investigated by
comparing the numerical results with available benchmark experimental data. It is found that Shear-Stress
Transport e k model and Large Eddy Simulation time-variant model can give the better predictions of fluid
flow and heat transfer properties; especially, SST e k model is recommended as the best compromise between
the computational cost and accuracy. Using Shear-Stress Transport e k model, the effects of jet Reynolds
number (Re), jet plate length-to-jet diameter ratio (L/d), target spacing-to-jet diameter ratio (H/d) and jet plate
width-to-jet diameter ratio (W/d) on local Nusselt number (Nu) of the target plate are examined. A correlation for
the stagnation Nu is presented.
INTRODUCTION
Jet impingement is one of the most efficient solutions of cooling hot objects in industrial processes as it
produces a very high heat transfer rate of forced-convection. There is a large class of industrial processes in
which jet impingement cooling is applied such as the cooling of blades/vanes in a gas turbine, the quench of
products in the steel and glass industries and the enhancement of cooling efficiency in the electronic industry.
Over the past 30 years, experimental and numerical investigations of flow and heat transfer characteristics under
single or multiple impinging jets remain a very dynamic research area. The effects of nozzle geometry, jet-to-
surface spacing, jet-to-jet spacing, cross flow, operating conditions, etc. on flow and heat transfer have been
experimentally studied (Donaldso and Snedeker, 1971; Hollworth and Cole, 1987; Vantreuren et al., 1994; San
et al., 1997; Dano et al., 2005; Wang and Mujumdar, 2005; San and Shiao, 2006). Martin (1977), Jambunathan
et al. (1992), Viskanta (1993), and Zuckerman and Lior (2005) referred to a large panel of fundamental studies
and presented complete reviews on the above important parameters influence on heat transfer.
Most industrial applications of impinging jets are concerned with turbulent flow in whole domain
downstream of a nozzle. Modelling of turbulent flow presents the greatest challenge for rapidly and accurately
predicting impingement heat transfer even under a single round jet. Over the past decades, although no single
model has been universally accepted to be superior to all classes of problems, various turbulent models have
been developed successfully to roughly predict impingement flow and heat transfer. However, there are only a
limited number of studies concerned with comparisons of the reliability, availability and capability of different
turbulent models for impingement flows. Thakre and Joshi (2000) evaluated twelve versions of low Reynolds
number c k models and two low Reynolds number RSM models for heat transfer in turbulent pipe flows.
Their comparative analysis between the c k models and RSM models for the Nusselt number prediction is in
favour of the applicability of the c k models even though the RSM model overcomes the assumption of
isotropy and the constancy of turbulent Prandtl number. Shi et al. (2002) presented simulation results for a single
semi-confined turbulent slot jet impinging normally on a flat plate. The effects of turbulence models, near wall
treatments, turbulent intensity, jet Reynolds number, as well as the type of thermal boundary condition on the
heat transfer were studied using the standard c k and RSM models. Their results indicate that both standard
c k and Reynolds stress model (RSM) models predict the heat transfer rates inadequately, especially for low
nozzle-to-target spacing. For wall-bounded flows, large gradients of velocity, temperature and turbulent scalar
quantities exist in the near wall region and thus to incorporate the viscous effects it is necessary to integrate
equations through the viscous sublayer using finer grids with the aid of turbulence models. In the study of Wang
and Mujumdar (2005), five versions of low Reynolds number c k models for the prediction of the heat
transfer under a two-dimensional turbulent slot jet were analyzed by comparison with the available experimental
data. Effects of the magnitudes of the turbulence model constants were also carried out.
In this paper, a numerical study of a confined circular air jet vertically impinging on a flat plate is performed.
The jet flow after impingement is constrained to exit in two opposite directions. The paper aims to recommend
the most suitable model(s) in predicting this type of flow and heat transfer through an investigation of the
relative performance of different turbulent models. Numerical calculations base on the CFD code FLUENT
6.1.18 are conducted, the justification of the models are carried out by comparing the numerical results with
available benchmark experimental data. Then, using the most suitable model, the effects of jet Reynolds number ,
jet plate length-to-jet diameter ratio, target spacing-to-jet diameter ratio and jet plate width-to-jet diameter ratio
on local Nusselt number of the target plate are examined. Also, a correlation for the stagnation Nu is presented.
PROBLEM DESCRIPTION
Geometry and Boundary Condition
Fig. 1 shows the physical domain and boundary conditions of the modelling. Air flow at high velocity
passes through a round jet with both length and diameter d=6mm, vertically impinging on the confined target
plate with two side walls positioned at spanwise distances of y=W/2. The jet after impingement was restricted
to discharge in two opposite directions parallel to the x-axis, and two channel outlets are placed at x=L/2. The
target plate was kept at constant heat flux of 1000W/m
2
; all other walls are adiabatic.
For Re=10000, H/d=2.0, L/d=41.7, W/d=10.42, the local Nusselt number distribution along the positive x-
axis is numerical simulated in different turbulent models to investigate the relative performance of these models
by comparing the numerical results with available benchmark experimental data (San and Shiao, 2006). Then,
using the most suitable model, the effects of jet Reynolds number, jet plate length-to-jet diameter ratio, target
spacing-to-jet diameter ratio and jet plate width-to-jet diameter ratio on local Nusselt number of the target plate
are examined. A correlation for the stagnation Nu is presented.
Fig. 1: The physical domain and boundary conditions
Turbulent Models
In this study, the standard c k model, the renormalization group (RNG) c k model, the realizable c k
model, the standard e k model, the Shear-Stress Transport (SST) e k model, the Reynolds stress model
(RSM) and the Large Eddy Simulation (LES) time-variant model (FLUENT_6.1_Documentation, 2003) were
used.
All the c k , e k and RSM turbulence models belong to the Reynolds-Averaged approach, in which the
Reynolds-averaged Navier-Stokes (RANS) equations were used as the transport equations for the mean flow
quantity, all the scales of the turbulence are modelled based on certain assumptions. The RANS equation in a
Cartesian tensor form can be written as:
( ) 0 =
c
c
+
c
c
i
i
u
x t

(1)
( ) ( ) ( )
j i
j l
l
ij
i
j
j
i
j i
j i
j
i
u u
x x
u
x
u
x
u
x x
p
u u
x
u
t
' '
c
c
+
(
(

|
|
.
|

\
|
c
c

c
c
+
c
c
c
c
+
c
c
=
c
c
+
c
c
o
3
2
(2)
For steady state flow, the time derivative terms drop out. Comparing the RANS momentum equation Eq. (2) to
the to the Navier-Stokes momentum equation, additional terms appear that represent the effect of turbulence. The
Reynolds stresses,
j i
u u ' ' , must be modelled in order to close Eq. (2).
The c k models and e k models employ the Boussinesq hypothesis (Hinze, 1975) to relate the
Reynolds stresses to the mean flow velocity gradients:
ij
i
i
t
i
j
j
i
t j i
x
u
k
x
u
x
u
u u o
|
|
.
|

\
|
c
c
+
|
|
.
|

\
|
c
c
+
c
c
= ' '
3
2
(3)
The standard, RNG and realizable c k models have similar forms with transport equations for turbulence
kinetic energy k and its dissipation rate c . The main issue is how the turbulent viscosity
t
is computed. In the
standard c k model (Launder and Spalding, 1972) the transport equations for the turbulence kinetic energy k
and its dissipation rate c is solved and
t
is computed as:
c


2
k
C
t
= (4)
where

C is a constant. In the RNG c k model (Yakhot and Orszag, 1986), the scale elimination results in a
differential equation for turbulent viscosity which gives Eq. (4) in the high-Reynolds-number limit and also
enables us to include low-Reynolds-number effects by using the original differential relation. While in the
realizable c k model (Shih et al., 1995), the coefficient

C is a function of the mean strain and rotation rates,
the angular velocity of the system rotation, and the turbulence fields.
In the standard e k model (Wilcox, 1998), the transport equations for turbulence kinetic energy k and the
specific dissipation rate e are solved, and
t
is computed as:
e

o
k
t
* = (5)
Where, * o is a function of Reynolds number and is one for high Re.
In the SST e k model (Menter, 1994), definition of the turbulent viscosity is modified to account for the
transport of the principal turbulent shear stress. Here,
t
is given as:
] / *, / 1 max[
1
1 2
e o o e

F
k
t
O
= (6)
Where, O is a function of the mean rate-of-rotation tensor,
2
F is the blending function, and
1
o is a constant.
This feature gives the SST e k model an advantage in term of performance over both the standard e k
model and the standard c k model.
The RSM model (Launder et al., 1975) solves exact transport equations for the transport of the Reynolds
stresses,
j i
u u ' ' . It also includes an additional scale determining equation for c . Then the turbulent viscosity is
calculated similarly to the c k models. The RSM is superior for situations in which anisotropy of turbulence
has a dominant effect on the mean flow. Such cases include highly swirling flows and stress-driven secondary
flows.
Unlike the c k , e k and RSM models which are all based on the RANS equations, LES provides and
alternative approach in which the large eddies are computed in a time dependent simulation that uses a set of
filtered Navier-Stokes (NS) equations (Galperin and Orszag, 1993). The attraction of LES is that, by modelling
less of the turbulence (and solving more), the error induced by the turbulence model will be reduced. However,
the application of LES to industrial fluid simulations is in its infancy. So far, the model has mainly been used to
simulate the fluid flows in simple geometries because of the large computer resources required to resolve the
energy-containing turbulent eddies.
Numerical Procedure
The numerical simulations were carried out using the commercial flow solver FLUENT 6.1.18 based on the
finite volume method. The momentum and energy equations were discretized using the second-order upwind
scheme and other transport equations were discretized using the power law scheme. The discretized equations
were solved using the SIMPLEC algorithm. Default values in FLUENT for all the parameters in the turbulent
models adopted. The commercial package Gambit 2.0.4 was used to generate the geometry and mesh for the
computational domain.
Fig. 2: Computational domain
,The computational domain is a quarter of the real physical geometry due to the symmetry of the problem as
shown in Fig. 2. Hexahedral elements are used for meshing the geometry. Different types of boundary conditions
were used for different zones of the flow domain. For the jet discharge, a mass flow inlet type of boundary
conditions is used. The temperature of the inlet is at 300 K. At the outlet of the computational domain (right
boundary in Fig. 1), the outflow type of boundary condition is used. At two symmetric sections (x=0 and y=0),
the symmetry type of boundary is specified. The no slip wall condition is used for all the other boundaries.
Hereinto, a constant heat flux of 1000W/m
2
at the bottom wall, and a zero heat flux on the other walls. In the
computation, the mean velocity and temperature were normalized with the velocity and temperature of the jet,
respectively. The material of the inlet air is described by the following parameters, 225 . 1 = kg/m
3
,
n=0.0242W/mK, = 7894 . 1 10
-5
kg/ms. The implicit and segregated solver is used for the solution of the
system of governing equations. During the simulations, the value of y
+
for the wall-adjacent cells was fixed
approximately at 1. The solution is assumed to be converged when the normalized residual of the energy
equation is lower than 10
6
and the normalized residuals of continuity and other variables are less than 10
3
.
RESULTS AND DISCUSSION
Comparison of Turbulent Models
For Re=10000, H/d=2.0, L/d=41.7, W/d=10.42, the local Nusselt number distribution along the positive x-
axis is numerically simulated. Seven different turbulent models are used in the simulation to investigate the
relative performance of these models by comparing the numerical results with available benchmark experimental
data (San and Shiao, 2006).
Fig. 3: Local Nusselt number distribution on the positive x-axis
(Re=10000, H/d=2.0, L/d=41.7, W/d=10.42)
The comparison of numerical results with those of experiments is presented in Fig. 3 and shows that SST e k
model and LES time-variant model can give the better predictions of fluid flow and heat transfer properties,
while the simulations by using standard, RNG and realizable c k models, standard e k model, and RSM
model all give large errors compared to available experiment data even with high resolution grids. Comparing
Nu
sg
obtained by different turbulent models using the same grids with that obtained by experiment of San and
Shiao (2006), the c k models, standard e k model and RSM model have the errors as large as 28-116%.
Moreover, they all give the bad predictions of the global distribution of the Nusselt number.
Table 1: Comparison of different turbulence models for confined impingement heat transfer
Turbulence Models Computational Cost Accuracy Error of Nu
sg
(comparison with experiments)
Standard c k Low Poor 101%
RNG c k Low Poor 28%
Realizable c k Low Poor 116%
Standard e k Moderate Poor 67%
SST e k Moderate Good 7%
RSM Moderate Poor 107%
LES High Excellent 4%
The computational cost and accuracy of the different turbulence models on predicting the flow and heat transfer
characteristics of this case are compared in Table 1. With the implementation of the LES time-variant model, the
cost in terms of the computational resource is very high although it gives excellent results.
Fig. 4: Local Nusselt number distribution on the positive x-axis
(Re=10000, d=6.0mm, H/d=2.0, L/d=41.7, W/d=10.42)
As a further check on the consistency of the experimental and numerical results obtained by SST e k
model, the local Nusselt number distribution for Re=30000, H/d=2.0, L/d=41.7, W/d=10.42 are compared. Fig. 4
shows the distribution of Nu along the positive x-axis. Some representative points taken from the experimental
study (San and Shiao, 2006) are additionally shown to illustrate the degree of the quantitative comparison.
Therefore, SST e k model is recommended as the best compromise between the computational cost and
accuracy and is used for all of the following simulations.
Effects of W/don Nu
sg
Fig. 5 (a), (b) show the variation with W/d of Nu
sg
when H/d=3.0, L/d=50 and H/d=5.0, L/d=50, respectively.
From the figures, it can be seen that with the increase of W/d, Nu
sg
descreases and meanwile the corresponding
decreasing rate grows. Moreover, the figures indicate that the increasing Reynolds number results in a rise of
Nu
sg
.
(a) H/d=3.0, L/d=50 (b) H/d=5.0, L/d=50
Fig. 5: Effects of W/d on Nu
sg
.
Effects of L/don Nu
sg
For L/d increasing from 10 to 150, the Nusselt numbers on stagnation point are calculated when H/d=3.0,
W/d=10 and H/d=3.0, W/d=20 at different Reynolds numbers as shown in Fig.6. It is noted that, the effects of
L/d are quite similar with those of W/d, i.e. the increasing L/d results in the decrease of Nu
sg
and the augment of
decreasing rate. However, it should be pointed out that the effects of L/d are relatively weaker than W/d. The
explanation for this phenomenon was presented by San, J. Y., and Shiao (2006). They believed that for an
increase of W/d, the decrease of Nu
sg
is caused by the enhancement of mixing of the hotter recirculation air from
downstream with the air near the jet orifice; while for increasing L/d, the decrease of Nu
sg
is mainly caused by an
enhancement of the flow-mixing near the stagnation point which has less effect than the enhancement of flow-
mixing near the jet orifice on the impingement heat transfer.
(a) H/d=3.0, W/d=10 (b) H/d=3.0, W/d=20
Fig. 6: Effects of L/d on Nu
sg
.
Effects of H/don Nu
sg
As shown in Fig. 7, Nu
sg
decreases with the increasing H/d because the larger jet-to-target spacting can allow
a longer time of mixing and heat transfer between the hotter jet and the surrounding cooler air and therefore
decrease the heat transfer near the stagnation point.
(a) L/d=50, W/d=10 (b) L/d=50, W/d=20
Fig. 7: Effects of H/d on Nu
sg
.
Correlation for Nu
sg
In the experimental study of San and Shiao (2006) for a confined circular air jet impinging on a flat surface,
Nu
sg
was found to be a function of
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0
) / (
d L d W
e d H
+
, which has also been validated by the present
simulations. Therefore, it is assumed that Nu
sg
can be expressed by the following correlation,
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0
) / (
d L d W B
sg
e d H ARe Nu
+
= (7)
By analyzing the results obtained by the present simulation, it is found that the exponent for the jet Reynolds
number, B=0.642; while the constant A=0.423. Thus, the complete correction for the Nusselt number on the
stagnation point can be written as,
. 30000 10000 ; 6 1 ; 150 10 ; 40 5 for
) / ( 423 . 0
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0 642 . 0
s s s s s s s s
=
+
Re H/d L/d W/d
e d H Re Nu
d L d W
sg
(8)
To show the quantitatively agreement between the numerical results and the correlation, Fig. 8 show the
variation of } ) / /{(
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0 d L d W
sg
e d H Nu
+
with Reynolds number under the different geometry
conditions. It is noted that the values of } ) / /{(
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0 d L d W
sg
e d H Nu
+
can be approximated well by
fitting curve,
642 . 0
423 . 0 Re .
(a) H/d=3.0, L/d=50; (b) H/d=3.0, W/d=20; (c) L/d=50, W/d=20
Fig. 8: Variation of } ) / /{(
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0 d L d W
sg
e d H Nu
+
with Re
CONCLUSIONS
The relative performance of seven versions of turbulent models, including the standard c k model, the
renormalization group c k model, the realizable c k model, the standard e k model, the Shear-Stress
Transport e k model, the Reynolds stress model and the Large Eddy Simulation, for the prediction of this type
of flow and heat transfer is investigated by comparing the numerical results with available benchmark
experimental data. It is found that the Shear-Stress Transport e k model and the Large Eddy Simulation time-
variant model can give better predictions of fluid flow and heat transfer. The SST e k model is recommended
as the best compromise between the computational cost and accuracy. Using the Shear-Stress Transport e k
model, the effects of jet Reynolds number, jet plate length-to-jet diameter ratio, target spacing-to-jet diameter
ratio and jet plate width-to-jet diameter ratio on local Nusselt number of the target plate are examined. It is found
that Nu
sg
increases with the rise of the jet Reynolds number; while the increasing W/d, H/d and H/d can all result
in the decrease of Nu
sg
and the augment of decreasing rate. Moreover, for 30000 10000 s s Re , 40 5 s sW/d ,
150 10 s s L/d and 6 1 s s H/d , a correlation
)] / ( 011 . 0 ) / ( 044 . 0 [ 3 . 0 642 . 0
) / ( 423 . 0
d L d W
sg
e d H Re Nu
+
= , is presented.
NOMENCLATURE
English Symbols
A, B constants
d jet diameter (m)
h convective heat transfer coefficient (W/m
2
K)
H jet plate-to-impingement plate spacing (m)
k turbulence kinetic energy
L jet plate length (m)
n thermal conductivity of air (W/m K)
Nu local Nusselt number, hd/n
Pr Prandtl number, o /
q surface heat flux (W/m
2
)
Q volumetric flow rate (m
3
/s)
Re jet Reynolds number, d Q t / 4
aw
T adiabatic wall temperature (K)
j
T jet total temperature (K)
w
T local wall temperature (K)
U jet mean velocity (m/s)
i
u ,
j
u component of velocity
W jet plate width or heated surface width (m)
x x-coordinate (m)
y y-coordinate (m)
+
y dimensionless distance,
t
/ y u y =
+
z z-coordinate (m)
Greek Symbols
o thermal diffusivity (m
2
/s)
dynamic viscosity (kg/ms)
t
turbulent viscosity
density (kg/m
3
)
c dissipation rate of turbulence kinetic energy k
e specific dissipation rate of turbulence kinetic energy k
Subscripts and Superscripts
aw adiabatic wall
sg stagnation point
w wall
REFERENCES
Dano, B. P. E., Liburdy, J. A., and Kanokjaruvijit, K., 2005, Flow characteristics and heat transfer
performances of a semiconfined impinging array of jets: effect of nozzle geometry, International Journal of
Heat and Mass Transfer, Vol. 48, pp 691-701.
Donaldso, C. D., and Snedeker, R. S., 1971, Study of free jet impingement .1. Mean properties of free and
impinging jets, Journal of Fluid Mechanics, Vol. 45, pp 281-&.
FLUENT_6.1_Documentation, FLUENT Inc., 2003.
Galperin, B. A., and Orszag, S. A., 1993, Large Eddy Simulation of Complex Engineering and Geophysical
Flows, Cambridge University Press.
Hinze, J.O., 1975, Turbulence, McGraw-Hill Publishing Co., New York.
Hollworth, B. R., and Cole, G. H., 1987, Heat-transfer to arrays of impinging jets in a cross-flow, Journal of
Turbomachinery - Transactions of the ASME, Vol. 109, pp 564-571.
Jambunathan, K., Lai, E., Moss, M. A., and Button, B. L., 1992, A review of heat-transfer data for single
circular jet impingement, International Journal of Heat and Fluid Flow, Vol. 13, pp 106-115.
Launder, B. E., Reece, G. J., and Rodi, W., 1975, Progress in development of a Reynolds-stress turbulence
closure, Journal of Fluid Mechanics, Vol. 68, pp 537-566.
Launder, B. E., and Spalding, D. B., 1972, Lectures in Mathematical Models of Turbulence, Academic Press,
London, England.
Martin, H., 1977, Heat and mass transfer between impinging gas jets and solid surfaces, Advances in Heat
Transfer, Vol. 13, pp 1-60.
Menter, F. R., 1994, 2-equation eddy-viscosity turbulence models for engineering applications, AIAA
Journal, Vol. 32, pp 1598-1605.
San, J. Y., Huang, C. H., and Shu, M. H., 1997, Impingement cooling of a confined circular air jet,
International Journal of Heat and Mass Transfer, Vol. 40, pp 1355-1364.
San, J. Y., and Shiao, W. Z., 2006, Effects of jet plate size and plate spacing on the stagnation Nusselt
number for a confined circular air jet impinging on a flat surface, International Journal of Heat and Mass
Transfer, Vol. 49, pp 3477-3486.
Shi, Y. L., Ray, M. B., and Mujumdar, A. S., 2002, Computational study of impingement heat transfer under
a turbulent slot jet, Industrial & Engineering Chemistry Research, Vol. 41, pp 4643-4651.
Shih, T. H., Liou, W. W., Shabbir, A., Yang, Z. G., and Zhu, J., 1995, A new Kappa-Epsilon eddy viscosity
model for high Reynolds-number turbulent flows, Computers & Fluids, Vol. 24, pp 227-238.
Thakre, S. S., and Joshi, J. B., 2000, CFD modeling of heat transfer in turbulent pipe flows, Aiche Journal,
Vol. 46, pp 1798-1812.
Vantreuren, K. W., Wang, Z., Ireland, P. T., and Jones, T. V., 1994, Detailed measurements of local heat-
transfer coefficient and adiabatic wall temperature beneath an array of impinging jets, Journal of
Turbomachinery - Transactions of the ASME, Vol. 116, pp 369-374.
Viskanta, R., 1993, Heat-transfer to impinging isothermal gas and flame jets, Experimental Thermal and
Fluid Science, Vol. 6, pp 111-134.
Wang, S. J., and Mujumdar, A. S., 2005, A comparative study of five low Reynolds number k-epsilon
models for impingement heat transfer, Applied Thermal Engineering, Vol. 25, pp 31-44.
Wilcox, D. C., 1998, Turbulence Modeling for CFD, DCW Industries, Inc., La Canada, California.
Yakhot, V., and Orszag, S. A., 1986, Renormalization group analysis of turbulence: I. Basic theory., Journal
of Scientific Computing, Vol. 1, pp 1-51.
Zuckerman, N., and Lior, N., 2005, Impingement heat transfer: Correlations and numerical modeling,
Journal of Heat Transfer, Vol. 127, pp 544-552.

You might also like