Proceedings of the ASME Turbo Expo 2010

June 14-18, 2010, Glasgow, UK

GT2010-22720
HEAT TRANSFER INVESTIGATIONS IN MULTIPLE IMPINGING JETS AT LOW
REYNOLDS NUMBER

Arvind G. Rao Myra Kitron-Belinkov

Assistant Professor Senior Lecturer
Technische Universiteit Delft Kinneret College on the Sea of Galilee
2629 HS Delft, Netherlands Emek-Hayarden 15132, Israel

Vladimir Krapp, Yeshayahou Levy *

Technion - Israel Institute of Technology
Haifa 32000, Israel

ABSTRACT
Jet impingement is a well established cooling methodology NOMENCLATURE
used for cooling turbine blades in gas turbine engines. Jet A area [m2]
impingement results in high heat transfer coefficients as CD coefficient of discharge [-]
compared to other conventional modes of single phase heat Cp specific heat at constant pressure [J/kg-K]
transfer. D jet diameter [m]
h convective heat transfer coefficient [W/m2-K]
Most of the research in jet impingement has been confined L length [m]
to high Reynolds number regime. In order to increase the m& mass flow rate [kg/s]
applicability of this technique to non conventional applications Nu Nusselt Number [-]
like in a low pressure micro turbine combustors or turbine N number of jets
blades, the behavior of such systems in the low Reynolds P pressure [Pa]
number regime should be understood. The present paper is a Pr Prandtl Number [-]
continuation of earlier investigations on the heat transfer Q heat input [W]
behavior of a large jet impingement array in the low Reynolds Re Reynolds Number [-]
number regime, especially in the laminar and transitional T temperature [K]
region. More experiments have been conducted with different t thickness of the jet plate [mm]
geometrical parameters of the array to analyze the effect of V velocity [m/s]
these parameters on the average heat transfer coefficient. X streamwise distance between two holes [m]
Numerical simulations with existing CFD tools were carried Y span wise distance between two holes [m]
out in order to understand the fluid mechanics inside such a Z distance between the jet impingement plate and the
complex system. The CFD model was validated with the heated plate [m]
experiments. Different turbulence models were used and it was Greek scripts
found that the SST-k- model was the best for modeling jet dynamic viscosity [Pa-s]
impingement phenomena. It is anticipated that the results density [kg/m3]
obtained from the present exercise will give better insights in
optimizing the design of multiple jet impingement cooling Subscripts
systems for high heat density applications. AV average
C crossflow
HP heated plate

*
Corresponding Author
CEO, Hom Ve Tena Ltd, Israel
Tel:- +972-829 4 3807
Email:- Levyy@aerodyne.technion.ac.il

1 Copyright 2009 by ASME

J jet
ABBREVIATIONS
CFD Computational Fluid Dynamics
SyBC Symmetric Boundary Condition
RANS Reynolds Averaged Navier Stokes
INTRODUCTION
Jet impingement is one of the most extensively used
technique for cooling the turbine blades, nozzle guide vanes
and turbine seals in gas turbines, thus enabling these parts to
operate in an environment which is at a higher temperature than
their melting point [1,2]. There are also a variety of process
and thermal control applications where impingement with high
velocity jets has become an established method of cooling or
heating surfaces. Examples include drying of paper and
textiles, food processing [3], refrigeration [4], annealing of
metals, glass tempering operations [5], cooling of concentrated
solar photo-voltaic cells [6], etc. Lately this technique is also
being explored for cooling of high density electronic
processors [7-9].
Many studies were carried out on single jet impingement,
both experimental and numerical. Narayanan et al [10] have
given a comprehensive review of the fluid mechanics and heat
transfer within a single jet impingement system. Studies (for
example, [11,12]) have determined that maximum heat transfer
is achieved when the transitional jet (just before the end of the
jet core) impinges on the surface.
For applications requiring highly localized heating or
cooling, a single circular jet may suffice. However, where large
surface must be continuously heated or cooled, multiple jets or
two-dimensional arrays of jet orifices are required to enhance
the overall heat transfer. In multiple impinging jets, even
though the principles of fluid mechanics and heat transfer are
similar to that of a single impinging jet, the heat transfer is
affected to a large extent by four processes, (a) interaction
between adjacent jets before impinging on the surface, if the
interjet distance is small, (b) the collision of wall jets from
adjacent jets, especially when the distance between jet plate
and the target plate is small, (c) the spent air from upstream jets
(called as cross flow), which alter the flow of downstream jets
and affect the stagnation regions of the downstream jets, and
(d) the manner in which this spent fluid is allowed to leave the
array (only from one side /direction, from opposite directions
or from all directions).
Figure 1 shows an inline jet impingement array that is
confined from two sides and therefore the spent air is allowed
to exit from the remaining two opposite sides. The holes are
arranged in an inline fashion. The important geometrical
parameters and ratios have been indicated in the figure. The
direction along the channel through which the spent air exits
the array is termed as streamwise direction. The direction
along the width of the channel where the flow is restricted is
named as spanwise direction.
Y/D
X/D
Exit air
Z/D
Exit
Figure 1: Schematic of the submerged jet impingement array
Many researchers have investigated the heat transfer
characteristics in small arrays of multiple impinging jets,
mainly in the turbulent regime, for both staggered and inline
holes configurations. The heat transfer characteristics in such
systems are often given in terms of a correlation between the jet
Nusselt Number Vs jet Reynolds Number (based on the jet inlet
hole diameter) and the fluid Prandtl Number. San and Lai [13]
investigated the optimum jet spacing for a staggered array with
5 holes for high Re impinging air jets. They found that an
inter-jet spacing of 8-10 jet diameters provided an optimum
value for the heat transfer coefficient. However in their
staggered configuration, the cross flow was not taken into
account. Garimella and Schroeder [14] investigated the heat
transfer distribution in a confined turbulent air jet impingement
system to study the optimum jet to impingement plate distance
and the difference in heat transfer coefficient for a single jet
and multiple jets. Based on their experimental investigations,
they provide the following correlation with average and
maximum deviation of 9% and 28.2% respectively
Z
0.105
Nu = 0.127 Re J 0.693 Pr 0.4 (1)
D
The equation is valid for 5000 Re 20000, 0.5 Z/D 4.
The above correlation compares with an average and maximum
deviation of 12.2 and 24.3% with the correlations for confined
jets of Huber and Viskanta [15]
One of the most comprehensive works to investigate
parameters affecting multiple jet impingement system was
carried out by Florschuetz et al. [16-19]. They provided
correlations to evaluate heat transfer coefficients for air jets
issuing from an array of circular orifices and impinging on a
flat surface. They also investigated and quantified the effect of
crossflow and spent air emerging in an array. Subsequently,
they also examined the effect of initial crossflow and non
2 Copyright 2009 by ASME
uniform array geometry, as encountered in turbine blade
cooling. The correlations provided by Florschuetz et al. [16] is
shown below
Nu = KRejm {1-B[(Z/D)(Vc/Vj)]n} Pr1/3 (3)
Where coefficients K, B and the exponents m, n are
dependent on the geometric parameters of the array in the form
of simple power functions
K, m, B, and n = C ( X / D )
( Y / D) ( Z / D)
nx ny nz
(4)
Inline Pattern
C nx ny nz
K 1.18 -0.944 -0.642 0.169
M 0.612 0.059 0.032 -0.022
B 0.437 -0.095 -0.219 0.275
N 0.092 -0.005 0.599 1.04
The overall range for their survey was, Re = 2500 70000;
cross flow to jet velocity ratio based on the one dimensional
cross flow estimation (VC/VJ) = 0 - 0.8; stream wise interjet
spacing (X/D) = 5-15; spanwise interjet spacing (Y/D) = 4-8;
jet to target plate distance (Z/D) = 1-3; X/Y = 0.625 3.75.
They found that the orifice coefficient of discharge (CD) was
independent of the Reynolds number. Also they report that
staggered pattern results in reduced overall heat transfer
coefficient as compared to the inline holes. The above
correlation also matches quite well with that of Kercher and
Tabakoff [20], who investigated the effect of streamwise
spacing and crossflow in a jet impingement array.
Thus it can be seen that most of the studies carried out on
multiple impingement jet system are experimental in nature and
have been mainly in the high Reynolds number range, ranging
from 2500 to 70,000. Such Reynolds number regimes are
encountered in large gas turbine engines. However if air jet
impingement systems are to be used for portable applications
where the available air pressure is small or pressure drop is
critical or where there are dimensional constraints, then it is
crucial that air impingement phenomena be characterized for
low Reynolds number flow regime. The following work
focuses on this crucial and unexplored area of characterizing
multiple jet impingement systems in low Reynolds number
flow regime.
THE EXPERIMENTAL SETUP
The main aim of this experimental investigation was to
study the heat transfer characteristics of a multiple jet
impingement array operating in the low Reynolds Number
regime. The holes are arranged in an inline pattern, in
accordance with research carried out by Florschuetz et al [17],
which indicated that inline holes result in a better heat transfer
than staggered holes (due to more interference of cross flow in
the latter). The schematic (cross section) of the multiple air jet
impingement system used in the experiments is shown in Fig.
2.
A three stage air filtration system with two air filters and a
dehumidifier is used to supply dry and clean shop air to the jet
impingement system. The air is metered by using a thermal
mass flow meter. The jet impingement array consists of a jet
impingement plate with holes drilled in an inline fashion. The
various configurations of the jet impingement plate are given in
Table 2. Schematic of the impingement plate is shown in Fig.
3. The streamwise interjet spacing (X/D) and the spanwise
interjet spacing (Y/D) are 5 jet diameters each. The jet plate to
bottom plate distance, (Z/D), can be changed with the help of
insulated spacers. A resistance type heater with dimensions of
29 29 mm is used as the heat source. The power delivered to
the heater is monitored through a high precision DC power
supply. The heat density can be varied within the range of 0-
50KW/m2. A copper plate with a thickness of 6mm is used on
top of the heater to evenly distribute the heat and to rule out
any localization of heating. A thermal paste with a thermal
conductivity of 7W/m-K is used in between the heater and the
copper plate for enhancing the conduction from the heater to
the copper plate. The temperature over the heater is monitored
with a fine T-type thermocouple. The Reynolds number within
the system is varied by changing the mass flow rate through the
system. The pressure and temperature at various junctions are
measured with the help of a multichannel data acquisition
system. The air exits from two opposite sides, as shown in Fig.
2. Temperature of the spent air, which is a measure of the heat
absorbed by the working fluid, is measured by using the
average value two T-Type thermocouples mounted at each exit
channel. Some of the heat generated by the heater is conducted
by the supporting structure. The uncertainty in measuring the
mass flow is 0.5 L/min, in temperature is 0.26 C and in the
heat transfer coefficient is 2.6%. More details about the
experimental setup can be found in a related paper [21].
The heater is insulated from its bottom and sides to reduce
heat transfer to the surroundings. Insulated spacers (to reduce
heat conduction from the heated copper plate to the supporting
structure) are used to vary the spacing between the
impingement plate and the heated surface in order to change
the Z/D ratio as required. The spent air (exhaust air) from the
system is collected at the end in a duct where its temperature is
measured to evaluate the total heat absorbed by the air.
The jet Reynolds Number is calculated as,
In V J DJ where 4m& /N (5)
Re J = VJ =
In In DJ2
hav DJ (6)
Nuav =
kair
The heat extraction rate by impinging jets is obtained from the
change in the temperature of the air
3 Copyright 2009 by ASME
 Q = m& CP , J (TEx TIn ) been kept constant at 5. The obtained results are compared
And the average heat transfer coefficient is evaluated as with the correlations of Florschuetz et al [16], which is found to
match fairly well with the experiments, especially for
hav=(Q/ AHP) / (THP TJ) (7)
hole/nozzle diameters of 0.2 and 0.4mm. The correlation over
predicts the Nusselt number for nozzle diameters of 0.6 &
0.8mm. It should be noted that the correlation of Florschuetz et
Air Inlet al. is valid for Re numbers from 2500 to 70,000, higher than
the range investigated in the present experiments. As shown in
Flow straightner Fig.4, it was found that the Nu number increases initially with
Pressure Probe
increase in the nozzle diameter but later decreases. This is due
Jet Nozzle Plate
to the fact that the crossflow decreases with increase in the hole
Exit air diameter (as the total area is constant), hence increasing the
thermocouple
effectiveness of the jets, however increasing the jet diameter
Spent air further reduce the number of jets, resulting in reduced average
outlet heat transfer. Thus there is an optimum diameter / array
Copper Plate configuration to in terms of heat transfer per unit area.
Thermal Paste Heater D Plate Array No. of X/D Y/D t/D Z/D
Thermocouple
(mm (mm2) Size Holes
)
Figure 2: Schematic of the jet impingement system 0.2 29x29 29x29 841 5 5 5 5
0.2 29x29 29x29 841 5 5 5 6
0.2 29x29 29x29 841 5 5 5 8
0.4 29x29 15x15 225 5 5 5 2.5
0.4 29x29 15x15 225 5 5 5 5
841 holes
0.4 29x29 15x15 225 5 5 5 7.5
0.4 29x29 15x15 225 5 5 7.5 2.5
0.4 29x29 15x15 225 5 5 7.5 5
0.4 29x29 15x15 225 5 5 7.5 7.5
Spent air 29mm 0.6 29x29 10x10 100 5 5 3.3 2.5
Spent air
0.6 29x29 10x10 100 5 5 3.3 5
0.6 29x29 10x10 100 5 5 3.3 7.5
Single Row 0.8 29x29 8x8 64 5 5 2.5 2.5
for simulation 0.8 29x29 8x8 64 5 5 2.5 5
29mm 0.8 29x29 8x8 64 5 5 2.5 7.5

Table 1: List of the experiments carried out

Figure 3: Jet impingement plate with 0.2mm holes
20
0.2mm z/d=5
0.4mm z/d=5
0.6mm z/d=5
EXPERIMENTAL RESULTS 16 0.8mm z/d=5
Flor_0.2mm
Flor_0.4mm
Flor_0.6mm
In the jet impingement phenomena, the flow near 12
Flor_0.8mm

impinging jet has a strong rotationality and streamlines

Nu

curvature, whereas the fluid motion in the vicinity of the 8

stagnation point is irrotational. Most of the heat transfer takes
place in the near impingement region, across the low Reynolds
4
number sub layer, and this is the reason why jet impingement is
known to exhibit higher heat transfer coefficients than other
conventional heat transfer methods, like flow through a 0
0 500 1000 1500 2000 2500
channel, flow over a flat plate, etc. Re

Figure 4: the variation of Nu number with Reynolds number

The variation of average Nusselt Number with Reynolds and comparison with Florschuetz et al [16].
number for various nozzle diameters and array sizes is shown
in Fig. 4. The geometrical ratios like X/D, Y/D & Z/D have

4 Copyright 2009 by ASME

 The effect of Z/D on the Nu number is shown in Fig. 5 for 1000
0.4mm diameter jets. It is interesting to see that the Nusselt 0.2mm z/d=5
900
number is maximum at Z/D of 5, as the Z/D is further increased 0.4mm z/d=5

to 7.5, the Nusselt number decreases. This non monotonic 800 0.6mm z/d=5

behavior of Nusselt number with Z/D has been observed in 700 0.8mm z/d=5

single jets as well [12]. Initially as the Z/D increases, the 600

h (W/K/m2)
stagnation region formed by the jet on the impinged plate 500
increases because of the jet spreading and entrainment, 400
however if the Z/D is increased beyond a point such that the jet 300
core ends before impinging on the bottom plate, then the heat
200
transfer coefficient decreases. It can be seen that the results
obtained from the correlations of Florschuetz et al [18], which 100

have also been plotted in the same figure, are not capable of 0
0 0.5 1 1.5 2 2.5 3 3.5
capturing this effect. This could be due to the fact that mass flow rate [gm/s]
Florschuetz et al investigated Z/D ratio between 1-3, which is
in the monotonic region with respect to Z/D variation. Figure 6: The variation of average heat transfer coefficient
with mass flow rate
16
0.4mm z/d=2.5
.
14 0.4mm z/d=5
800
0.4mm z/d=7.5
12
Floruschitz z/d=2.5 700

10 Floruschitz z/d=5
600
Floruschitz z/d=7.5
Nu

8
P l (0 4 /d 5) 500
h [W/K/m2]

6
400
4
300
0.2mm z/d=5
2
200 0.4mm z/d=5
0
0.6mm z/d=5
0 500 1000 1500 2000 2500 100
Re 0.8mm z/d=5
0 P (0 4
Figure 5: The effect of Z/D variation on Nusselt Number 0 1000 2000 3000 4000 5000 6000
Press. Drop [Pa]

It was observed that the average heat transfer coefficient

is higher for jets with low diameter jets for the same mass flow
rate per unit of surface area, as shown in Fig. 6. This is due to
the increase in the number of jets per unit area with the
decrease in the jet diameter. Therefore in systems where mass
flow rate is critical, like in gas turbine engines, it seems to be
preferable to use jet impingement array with small diameter
jets. However for systems where the pressure drop is critical,
as in micro turbines or in cooling of photo voltaic cells, using
smaller diameter jets would result in excess pressure drop, as
shown in Fig. 7. Thus the geometrical parameters of the jet
array can be optimized for various operating conditions and
objectives
Figure 7: The variation of average heat transfer coefficient
with pressure drop
NUMERICAL MODELING
There are not many publications that deal with numerical
investigation of multiple jet impingement systems. Craft et al
[22] evaluated 4 turbulence models (k-, Reynolds stress model
and two variations of RSM incorporating modifications for
wall reflection) for numerical prediction of turbulent jets
discharged from a circular pipe. Their results indicate that
none of the schemes was entirely successful in predicting the
effects of Reynolds number and they attributed this due to the
two-equation eddy viscosity schemes adapted in all cases to
span the near wall sub layer. Recently Hofmann et al [23] did
an assessment of 13 widely used turbulence models for
modeling single jet impingement at high Reynolds Number
flow (Re = 34,000 and Re = 124,000) by using commercial
CFD software Fluent 6.0. They concluded that only Shear
Stress Transport (SST) k- model was able to predict the

5 Copyright 2009 by ASME

laminar turbulent transition and gave satisfactory results for
small nozzle to plate distances as well as for pulsating jets. Velocity Inlet/ mass flow inlet
Center
Sagot el al [24] investigated the jet impingement heat transfer
Last jet
on a plate with constant wall temperature as a boundary
condition. They also report that SST k- model is the most SyBC
suitable RANS turbulence model for modeling jet impingement
SyBC Press
Pressur
phenomenon. outlet
e Outlet
Nozzles CU plate
Const. heat flux
Geometry and Boundary Conditions: - The 3D numerical
Symmetric
simulation of complete array with large number of holes is not
possible due to the large computational requirements.
However using symmetry, the computational domain can be
simplified to a large extent. Since the spent air exits from the Figure 8: Geometry of the computational domain
two sides of the array, there is symmetry as the flow is divided (D=0.2mm, X/D =Y/D=Z/D=5)
at centerline. For computational purposes, the flow domain
can be further divided because it can be assumed that the flow
in all the rows is more or less the same. Thus for the array
with 0.2mm diameter holes, the flow in one row of holes
consisting of 14.5 holes can be approximated to represents the
flow in the entire array as shown in Fig. 8. For the array with
0.4mm holes, one row of holes with 7.5 jets can represent a
row of holes (refer to table 2 for details about the array
geometry) shown in Fig. 9. Thus, by using symmetric
boundary conditions appropriately, this single row of holes
can be used to predict the flow behavior within the complete
impinging array.

Preliminary simulations proved that the inclusion of the
heater and thermal paste do not affect the final steady state
CFD results and therefore they are not included in the model
of the CFD simulations. The copper base is treated as a wall
with a constant heat flux (as done in the experiments). The air
enters the computational domain from the top (red color in
figures 9-10) and then flows through all the nozzles,
impinging on the bottom heated plate and there after exiting
out through the outlet.
The "velocity inlet" is used as a boundary condition for
the inlet and "pressure outlet" boundary condition is used for
the outlet. Symmetric boundary conditions have been used on
the sides and at the center (as shown in Fig. 8) to discretize the
array of holes/orifices into a single row of jets. The
simulations were carried out using a commercial CFD solver,
STAR CCM+. on a parallel processor SGI Altix XE320 with 4
4-7 processors used at a time. STAR CCM algorithms are
finite volume mostly 2nd order methods based on the body of
work by Demirzic, Peric & Ferziger. The flow model is
compressible although the flow velocities involved in this
investigation are much lower than the compressibility limits.
The segregated solvers option is used for faster computations.
Convergence: All simulations were carried out till at
least a convergence criterion of less than 10-3 was satisfied.
Figure 9: Geometry of the computational domain
(D=0.4mm, X/D =Y/D=Z/D=5)
Sr.No Parameter Value
1 Inlet Velocity 1.0 m/s
2 Pressure at Outlet 101325 Pa
3 Ambient Pressure 101325
4 Heat Flux 14268 W/m2
5 Inlet air Temperature 298 K
Table 3: Physical parameters of the simulation
In most cases the highest residual was 5 10-4 and the other were
10-5 or less. The criteria for stopping a simulation were the
simultaneous occurrence of a minimal residual value (<6 10-
), flattening of residual plot and flattening of the average
heated plate temperature with respect to the iteration number.
Grid: - Unstructured grid is used for all simulations. The grid
is constructed to be very dense around the vicinity of the
nozzles, and the cell size increases with the radial distance.
Several simulations with the same geometry but different grid
sizes (from half million to two million) were carried out to
check the grid independence of the simulation. Most grids
varied in the range of 0.7-1.1 million cells. A cut section of the
unstructured grid with the polyhedral cells and wall layers is
shown in Fig. 10.

6 Copyright 2009 by ASME

Turbulence models: - Turbulence models replace the full time
dependent Navier-Stokes equations by simplified averaged
systems which reduce calculation resources and are still able
to predict the effects of turbulence. Many models have been
developed and implemented within commercial CFD software.
Among these, two equation turbulence models are the most
common types. They include two extra transport equations to
represent the turbulent properties of the flow. Often, one of
the transported variables is the mean turbulent kinetic energy
per unit mass k. The second variable varies depending on what
type of two-equation model it is (average turbulent dissipation
rate, , for k- turbulence model or the specific dissipation,
, which is a measure of the inverse time scale of the large
eddies for k- turbulence model). The crucial differences
between these two models lie in the modeling of the diffusive
terms and near-wall treatments, including boundary
conditions. The numerical diffusion treatment is known to
have a central importance near walls and in particular near
boundary layer/free stream interfaces. Here we focus on the k-
models because of their success in solving jet impingement
problems [23, 24].
Figure 10: Unstructured grid with wall layers
The K-Omega Model:-The advantage of the k- model over
the k- model is due to its improved performance for boundary
layers under adverse pressure gradients, as encountered in the
jet impingement phenomena and it can be applied through out
the boundary layer, including the viscous dominated regions of
the flow without any modifications. However, the boundary
layer computations in k- model are very sensitive to the
values of in the free stream, which may result in extreme
sensitivity to the inlet boundary conditions for the internal
flows. There are variations within the k- model, see for
example [25]. In the present work the SST k- model is used,
Wilcox [26]. Three types of corrections are included in the
model: 1, a correction to account for the sensitivity of the free
stream / inlet conditions (based on the products of k and),
2, a correction to improve the free-shear-flow spreading rates
of the model, and, 3, a low Reynolds Number correction that
allows the Standard k- model to be better applied in the low
Reynolds number or transitional flows.
The BSL k- model was proposed by Menter [27]. The
BSL (baseline) model effectively blends a k- model in the far-
field with a k- model near the wall. The BSL model is
designed to give results similar to those of original k- model
of Wilcox model in the inner 50 % of the boundary layer but
gradually changes to the high Re number Jones-Launder k-
model [28] (in a k- formulation) towards the boundary layer
edge. The transformed equation looks very similar to the one
in the standard k- model, but adds an additional non-
conservative cross-diffusion term containing the dot product
k . Inclusion of this term in the transport equation causes
the k - model to predict identical results as the k- model at
the edge of the boundary layer and outwards.
The Shear Stress Transport (SST) k- model was also
proposed by Menter as well [29, 30]. It is also a combination
of the standard k- model in the inner parts of the boundary
layer and k- in the freestream. The SST model is an extension
of the BSL k- model and has the additional ability to account
for the transport of the principal shear stress in adverse
pressure gradient boundary-layers. The model is based on the
assumption that the principal shear-stress is proportional to the
turbulent kinetic energy, which is introduced into the definition
of eddy-viscosity. Therefore this model can be used as a low
Re model without incorporating any additional damping
functions and can be used in cases with adverse pressure
gradient. The SST k- model is less sensitive to the inlet free-
stream turbulence properties, which is one of the main
drawbacks of the Standard k- model. The y+ values were
between 0.25 and 1.
VALIDATION WITH EXPERIMENTS
It is important that the results obtained about the heat
transfer from CFD simulations be compared with experiments
for establishing their validity. Several available turbulence
models were tested but it was found that the SST k-
turbulence model converged best and compared best with
experiments. This is in accordance with the results obtained
by Hofmann et al [23].
Figures 11 and 12 show comparison of the average heat
transfer coefficient obtained from CFD simulations using SST
k- model and the experiments at various Reynolds number
and Z/D ratios for arrays containing 0.4mm and 0.6mm jets
respectively. It is seen that the simulations for the array with
0.4mm jets are in good agreement with the experimentally
obtained results. However the effect of Z/D variation could
7 Copyright 2009 by ASME
 not be captured by the simulations. In the experiments, it was
observed that the Nusselt number has a maximum for the Z/D
ratio of 5, whereas the simulations predict a monotonic
decrease in the Nusselt number with increase in the Z/D ratio.
For the array with 0.6mm diameter jets, it was found that the
CFD simulations over predict the average heat transfer
coefficient, as evaluated from the experiments, by around 15
%. The error is almost constant for the range of Reynolds
Number investigated. Also, as in the previous case, the CFD
simulations were not able to predict the effect of Z/D on
Nusselt number.
14
12
interferes with the downstream impinging jets. The crossflow
forms a boundary layer over the impinging plate and impedes
the formation of stagnation region by the jet, thereby reducing
the heat transfer. The crossflow also contributes to the heat
transfer mechanism of the system by convecting heat along the
channel (like flow over a flat surface); however the
contribution of crossflow in the overall heat transfer process
depends on the geometrical parameters of the impinging
system. The crossflow velocity increases downstream along
the channel in the streamwise direction and therefore the last
jet in the channel experiences the maximum amount of
crossflow. The velocity distribution within the channel as
obtained from the CFD simulations for an array with 0.4mm
diameter jets is shown in Fig. 13.

10

8
0.4mm z/d=2.5
Nu

6 0.4mm z/d=5
0.4mm z/d=7.5
4
CFD z/d=2.5

2 CFD z/d = 5
CFD z/d=7.5 Exit
0
0 500 1000 1500 2000 2500
Re Cross flow
Figure 11: Validation of the average heat transfer coefficient
with experiments for array with 0.4mm diameter jets

Figure 13: Velocity profile within the channel of the array

18 with 0.4mm diameter jets
16

14
It can be seen from Fig. 13 that the velocity profile
changes significantly from the first nozzle (centre jet) in the
12
channel to the last nozzle in the channel (more than 10% in
10
this case). Since there is no cross flow for the centre jet, the
Nu

0.6mm z/d=2.5
8
0.6mm z/d=5
centre jet hits the bottom plate without any distortion. The last
6 0.6mm z/d=7.5 jet experiences the highest cross flow and therefore hits the
4
CFD z/d=2.5 bottom plate way downstream of the flow. The jet velocity
CFD z/d=5 within the nozzle increases linearly from the centre jet to the
2
CFD z/d=7.5 last jet due to the decrease in the static pressure along the
0 streamwise direction.
0 500 1000 1500 2000 2500 3000 3500
Re
The temperature profile of the impinging plate is shown
Figure 12: Comparison of Nusselt number Vs Reynolds in Fig. 14. The temperature distribution determines the local
number obtained from CFD with experiments for 0.6mm jets Nusselt number (and hence the local heat transfer coefficient).
The temperature is lowest (or the heat transfer coefficient is
highest) in the stagnation region where the jet impinges. For
RESULTS AND DISCUSSIONS the first few jets, it can be seen that the temperature
distribution under each jet is circular, similar to those obtained
As mentioned earlier, the flow within a large array of from a single impinging jet. The effectiveness of jets in
impinging jets is quite complex because of the interaction cooling the impinging plate reduces gradually along the
between the jets and the spent flow. The crossflow is defined channel, despite of the fact that jet velocities are increasing
as the spent air from the upstream jets which flows parallel to from the centre jet to the last jet.
the channel in the streamwise direction and therefore

8 Copyright 2009 by ASME

 Cross flow

Exit

Figure 14: The temperature profile of the bottom plate

within the channel of the array with 0.4mm diameter jet

To get more insight into the heat transfer in this

inline array, the variation of Nusselt number along the channel
in the streamwise direction in the form of a 3D plot is shown Figure 16: Streamwise Nusselt Number Variation at Re=830
in Fig 15 for 0.2mm and Fig 16 for 0.4mm jets. It is seen that for 0.4 mm diameter jets
in contrast with intuition, the Nusselt number variation is not
monotonic along the channel. The peak Nusselt number is Since the streamwise variation in the Nusselt number
highest for the 3rd-4th jets, after which the Nu number was not resolved in the present experiments, this observation
decreases further downstream. could not be confirmed experimentally. The fact that the
variation of local Nu number is not monotonic was also
observed by other researchers; however this observation was
never reported explicitly. The experiments carried out by Van
Treuren et al [31] also show a non-monotonic variation of the
Center Jet stagnation Nu number, as depicted in Fig 17 for the case with
12
Z/D=4 and X/D=Y/D=8. The array they investigated was an
10
inline jet array with 8 jets for a Reynolds Number range of
8 Exit
6
3500 to 40000. It can be seen from Fig. 17 that the effect of
4
cross flow is less significant, so the higher jet velocity
2 towards the exit (due to the increasing pressure difference
0 over the nozzle plate) counter the deteriorating effects of the
15 cross flow, except for the last few jets. In the present
10
5 experiments, the effect of crossflow is dominant, particularly
0
for the 0.2mm case where there are more jets (14.5) in the
Figure 15: Streamwise Nusselt Number Variation at Re=420 row. One more reason could be attributed to the fact that in the
for 0.2 mm diameter jets present experiments, the X/D & Y/D values are lower than
those used by Van Treuren [31, 32].

In a series of papers [33-36], the detailed heat transfer

distributions for arrays of jets were measured for a variety of
parameters, configurations, flow orientations and pressure
gradients. When the number of streamwise inline jets is small,
as it is in the work of Gao [34] and Hebert et al. [36], (five
jets), the graph of the Nu number's peaks is increasing and
then decreasing. Huang et al. [33] studied an array of twelve
inline jets with different exhausts configurations whereas the
inlet air direction was parallel to the impinging plane. In the
exit scheme where the exit flow direction is same as the inlet
flow direction, the overall behaviour of Nu along the channel
is similar to those obtained in the present CFD simulations.
The reason for this non monotonic behaviour could be due to
the increasing jet velocity downstream along the channel and

9 Copyright 2009 by ASME

the non linear interference between the cross flow and the
impinging jets.
Downstrea
Figure 17: Variation of Nusselt number along the channel for
Z/D =4; X/D=Y/D=8 [32]
CONCLUSIONS
Large square submerged array with different diameter
jets is analysed in the low Reynolds Number regime, both
experimentally and numerically.
The experimentally obtained heat transfer characteristics
compares well with results obtained by Florschuetz etal
[16-19]. However the correlation could not predict the
effect of Z/D variation on Nusselt number.
It was observed that array with 0.4mm diameter jets
exhibit higher heat transfer coefficients as compared to
other nozzle diameters investigated. Thus there is an
optimum nozzle diameter for achieving maximum heat
transfer coefficient.
It was observed that the Nusselt number increases initially
with increase in the Z/D ratio and then decreases with
further increase in the Z/D ratio.
The CFD simulations were found to predict the heat
transfer characteristics of the jet impingement array with
reasonable accuracy.
The SST k- turbulence model is suitable for modelling
the heat transfer characteristics within multiple impinging
jets. However the error could be as high as 15 %.
It was observed from the simulations that the jet velocity
increases linearly from the centre jet to the last jet due to
the increasing pressure difference across the nozzle plate
along the channel in the streamwise direction.
In contradiction with previous assumptions, it was
observed that the Nusselt number variation in the
streamwise direction is non monotonic, which is probably
due to the complex interaction of cross flow with the
increasing jet velocity along the streamwise axis.
ACKNOWLEDGMENT
The autors would like to thank Hom Ve Tena Ltd. for its
financial support that assisted to conduct the present research
work.
REFERENCES
[1]. The Jet Engine, Rolls Royce publications
[2]. Han, J. C., Dutta, S. and Ekkad, S.V., Gas Turbine Heat
Transfer and Cooling Technology, Taylor & Francis,
Inc., 2001.
[3]. Sarkar, A., Nitin, N., Karwe, M.V., and Singh, R.P., Fluid
Flow and Heat Transfer in Air Jet Impingement in Food
Processing, J. Food science, Vol. 69 (4), 2004, pp. 113-
122.
[4]. Imraan, M., and Sharma, R. N., Jet Impingement Heat
Transfer in a Frost Free Refrigerator: The Influence of
Confinement, Int. Journal Refrigeration, In Press, 2009.
[5]. Cirillo, F., and Isopi, G.M., Glass Tempering Heat
Transfer Coefficient Evaluation and Air Jets Parameter
Optimization, Applied Thermal Engineering, Vol. 29,
2009, 1173-1179.
[6]. Cooling of PV cells
[7]. Guarino, J.R., and Manno, V.P., Characterization of
Laminar Jet Impingement Cooling in Portable Computer
applications, IEEE Transactions on Components and
Packaging Technologies, Vol. 25. (3), 2002, pp. 337-346.
[8]. Kercher, D.S., Lee, J. B., Brand, O., Allen, M.G. and
Glezer, A., Microjet Cooling Devices for Thermal
Management of Electronics, IEEE Trans. Components
and Packaging Technologies, Vol. 26. (2), 2003, pp. 359-
365.
[9]. Fabri, M., Jiang, S., and Dhir, V.K., Experimental
Investigation of Single Phase Micro Jets Impingement
Cooling for Electronic Applications, Proc. HT 2003
ASME Heat Transfer Conferece, Nevada, USA, 2003.
[10]. Narayanan, V., Sayed-Yagoobi, J., and Page, R.H., An
experimental Study of Fluid Mechanics and Heat Transfer
in an Impinging Slot Jet Flow, Int. J. Heat Mass
Transfer, Vol. 47, 2004, pp.1827-1845.
[11]. Jambunathan, K, E., Lai, M., Moss, A., and Button, B.L.,
A review of heat transfer data for single circular jet
impingement", Int. J. Heat Fluid Flow, Vol. 13, 1992, pp.
106-115.
[12]. Martin, H., Heat and Mass Transfer between Impinging
Jets, Adv. Heat Transfer, Vol. 13, 1977, pp. 1-60.
[13]. San, Jung-Yang., and Lai, Mao-De., Optimum jet-to-jet
spacing of heat transfer for staggered arrays of impinging
air jets, Int. J. Heat Mass Transfer, Vol. 44, 2001, pp.
3997-4007.
10 Copyright 2009 by ASME
[14]. Garimella, S. V., and Schroeder, V. P., Local Heat
Transfer Distributions in Confined Multiple Air Jet
Impingement, J. Electronic Packaging, Vol. 123, 2001,
pp. 165-172.
[15]. Huber, A. M., and Viskanta, R., Effect of Jet-Jet Spacing
on Convective Heat Transfer to Confined, Impinging
Arrays of Axisymmetric Air Jets, Int. J. Heat Mass
Transfer, Vol. 37, 1994, pp. 28592869
[16]. Florschuetz, L.W., Truman, C.R., and Metzger, D.E.,
Streamwise Flow and Heat Transfer Distribution for Jet
Array Impingement with Crossflow, J. Heat Transfer,
Vol. 103, 1981, pp. 337-342.
[17]. Florschuetz, L.W., Metzger, D.E., Takeuchi, D. I., and
Berry, R. A., Multiple Jet Impingement Characteristic
Experimental Investigation of In-Line and Staggered
Arrays with Crossflow, 1980, NASA Contractor Report
3217.
[18]. Florschuetz, L.W., Metzger, D.E., and Truman, C.R., J et
Array Impingement With Crossflow - Correlation of
Streamwise Resolved Flow and Heat Transfer
Distributions, 1981, NASA Contractor Report 3373.
[19]. Florschuetz, L. W.; Metzger, D. E.; Su, C. C.; Isoda, Y.;
Tseng, H. H., Jet array impingement flow distributions
and heat transfer characteristics. Effects of initial
crossflow and non uniform array geometry 1982, NASA
Contractor Report 3630.
[20]. Kercher, D.M., and Tabakoff, W., Heat Transfer by a
Square Array of Round Air Jets Impinging Perpendicular
to a Flat Surface Including the effects if Spent Air, J.
Engg Power, 1970, pp 73-82.
[21]. Rao, G.A., Levy, Y., and Kitron-Belinkov, M., Heat
Transfer Characteristics of a Multiple Jet Impingement
System, 48th Israeli Aerospace Conference, 5-7 March,
Tel-Aviv, Israel, 2009.
[22]. Craft, T.J., Graham, L.J.W. and Launder, B.E., Impinging
Jet Studies for Turbulence Model Assessment-II. An
Examination of the Performance of four Turbulence
Models, Int. J. Heat Mass Transfer, Vol. 36(10), 1993,
pp. 2685-2697.
[23]. Hofmann, H. M., Kaiser, R., Kind, M., and Martin, H.,
Calculations of Steady and Pulsating Impinging Jets- An
Assessment of 13 widely used Turbulence Models,
Numerical Heat Transfer, Part B, Vol. 51, 2007, pp. 565-
583.
[24]. Sagot, B., Antonini, G., Christgen, A., and Buron, F., Jet
Impingement Heat Transfer on a Flat Plate at a Constant
Wall Temperature, Int. J Thermal Sci., Vol. 47, 2008,
pp.1610-1619.
[25]. STARCCM+ user guide.
[26]. Wilcox, D.C., Turbulence Modeling for CFD, 2nd
edition, 1998, DCW Industries, Inc.
[27]. Menter, F.R., Improved Two Equation k- Turbulence
Models for Aerodynamic Flows, 1992, NASA Technical
Memorandum 103975.
[28]. Jones, W. P., and Launder B. E., The Calculation of Low-
Reynolds-Number-Phenomena with a Two-Equation
Model of Turbulence, Int. J. Heat Mass Transfer, Vol.
16, 1973, pp. 1119-1130.
[29]. Menter, F. R., Zonal Two Equation k- Turbulence
Models for Aerodynamic Flows", AIAA Paper 93-2906,
1993.
[30]. Menter, F.R., Two-equation eddy-viscosity turbulence
modeling for engineering applications, AIAA Journal,
Vol. 32(8), 1994, pp. 1598-1605.
[31]. Van Treuren, K.W., Wang. Z., Ireland, P.T., and Jones,
T.V., Detailed Measurements of Local Heat Transfer
Coefficient and Adiabatic Wall Temperature beneath an
Array of Impinging Jets, Journal of Turbomachinery,
Vol. 116, 1994, pp. 369 374.
[32]. Van Treuren, K. W., Impingement Flow Heat Transfer
Measurements of Turbine Blades Using a Jet Array,
Thesis submitted for D.Phil degree at the University of
Oxford, June 1994.
[33]. Huang, Y., Ekkad, S.V., Han, J.C., "Detailed heat transfer
distributions under an array of orthogonal impinging jets",
Journal of Thermophysics and Heat Transfer, Vol. 12(1),
1998, pp.73-79.
[34]. Gao, L., "Effect of jet hole array arrangement on
impingement heat transfer", M.Sc. thesis, University of
Louisiana, 2005.
[35]. Gao, L., Ekkad, S.V., Bunker, R.S., "Impingement heat
transfer, part I: linearly stretched arrays of holes", Journal
of Thermophysics and Heat Transfer, Vol. 19(1), 2005,
pp.57-65.
[36]. Hebert, R., Y., Ekkad, S.V., Gao, L., Bunker, R.S.,
"Impingement heat transfer, part II: effect of streamwise
pressure gradient", Journal of Thermophysics and Heat
Transfer, Vol. 19(1), 2005, pp.66-77.
11 Copyright 2009 by ASME