You are on page 1of 13

International Journal of Heat and Mass Transfer 139 (2019) 700–712

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/ijhmt

Evaluation of a developed SST k-x turbulence model for the prediction of


turbulent slot jet impingement heat transfer
Huakun Huang a, Tiezhi Sun a, Guiyong Zhang a,b,c,⇑, Da Li d, Haipeng Wei e
a
Liaoning Engineering Laboratory for Deep-Sea Floating Structures, School of Naval Architecture, Dalian University of Technology, Dalian 116024, China
b
State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian 116024, China
c
Collaborative Innovation Center for Advanced Ship and Deep-Sea Exploration, Shanghai 200240, China
d
Research Institute Ltd., China National Offshore Oil Corporation, Beijing 100027, China
e
Beijing Institute of Astronautical Systems Engineering, Beijing 100076, China

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, a turbulence model based on OpenFOAM has been developed for the jet impingement heat
Received 19 November 2018 transfer, which is based on the standard SST k-x model and considers the effects of cross-diffusion term
Received in revised form 19 April 2019 and the Kato-Launder model. The effects of the modifications have been studied for different nozzle-plate
Accepted 19 May 2019
spacing of 4 and 9.2 in terms of pressure distributions, mean velocity, skin friction and the local Nusselt
Available online 24 May 2019
number distributions. The results have been compared with both the experimental data and the numer-
ical results in the literatures. Good agreement is obtained in the studied cases and the developed model
Keywords:
overcomes the false secondary Nusselt number peak in high nozzle-plate spacing for slot jets which is
Jet impingement
Heat transfer
predicted by the standard SST k-x model. The results show that the modifications improve the model’s
Cross-diffusion term ability to predict the heat transfer for jet impingement by reinforcing the model’s sensitivity to the pres-
Pressure gradient sure gradient. By way of comparative studies, it shows that the SST k-x with transition model and the
OpenFOAM standard SST k-x model are not sensitive to the pressure gradient. Meanwhile, the cases of nozzle-
plate spacing of 2 and 6 are also investigated. This investigation indicates that the pressure gradient plays
an important role for turbulent slot impinging jet. And it also reveals that the effect of cross-diffusion
term should not be ignored in the boundary layer.
Ó 2019 Elsevier Ltd. All rights reserved.

1. Introduction For a low nozzle-plate spacing for jet impingement, the macro-
scopic description involves three main regions including free jet
Owing to the highly localized mass, momentum and heat trans- region, stagnation region and wall-jet region [2]. Each region is
fer, jet impingement is widely used in various fields, such as cool- governed by some specific generic turbulence mechanisms. For
ing or heating surfaces. In ship movements, the high-speed jet the free jet region, there is a potential core of the jet surrounded
flows are used more commonly to improve the maneuverability by the shear layer, where the velocity is uniform and equal to
of the ship. In the underwater environment, to avoid the cavitation, the velocity of the jet exit. In the stagnation region, an adverse
severe noise and degradation, the jet propulsion is widely utilized pressure gradient occurs owing to the interaction of the vortex
on unmanned underwater vehicles [1]. In the field of military, jets developing from free shear layer with the impermeable wall
also are used to generate the lift in vertical takeoff and landing air- [3,4]. In the wall-jet region, the negative pressure gradient leads
craft, cool high speed vehicles and launch rockets. Considering the to a velocity acceleration region [2] and an unsteady separation
large number of practical applications, jet impingement involving of the boundary [3]. In addition, another interesting features for
flow fields and heat transfer are studied comprehensively in this jet impingement are the inner peak, dip and secondary peak along
work. However, despite the apparent geometric simplicity, the the impinging plate. For the inner peak, Lytle and Webb [5] pointed
flow phenomenon of jet impingement, including strong adverse out that the turbulence plays an important role. However, contrary
pressure gradient, flow separation and vortex formation and break causes found by other researchers [6,7] should be the flow acceler-
down, is complex. ation region and thin boundary layer. And it is a minor role for tur-
bulence. For further insight, Kataoka [8] suggested that the cause is
the impact on the impinging plate by large-scale eddies together
⇑ Corresponding author. with the associated pressure pulsations. Choo et al. [9] carried
E-mail address: gyzhang@dlut.edu.cn (G. Zhang).

https://doi.org/10.1016/j.ijheatmasstransfer.2019.05.058
0017-9310/Ó 2019 Elsevier Ltd. All rights reserved.
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 701

Nomenclature

u velocity Tin inlet jet temperature


p pressure Pmax maximum static pressure
cp specific heat capacity at constant pressure Cf skin friction coefficient
T temperature Vin inlet velocity
k turbulent kinetic energy U mean x-component velocity
B nozzle width
L length of impinging plate Greek symbols
H impinging distance q density
I turbulent intensity l dynamic viscosity
lc turbulent length scale k thermal conductivity
kin inlet Turbulent kinetic energy mt eddy viscosity
Pr Prandt number x specific dissipation
y+ dimensionless distance of node to wall xin inlet specific dissipation rate
Nu local Nusselt number sw wall shear
Tw impinging plate temperature

out experiments to study the influence of nozzle-plate spacing and favorable correction is Yap correction (YC) [23], which modifies
also found the pressure is crucial on inner peak. When the flow the effect of cross-diffusion term affected by the pressure gradient
evolves into downstream, the unsteady features cause the dip of deeply. And many researchers have proved that YC improves the
heat transfer [4]. As a result, the transition from laminar to turbu- prediction of heat transfer and flow structures [11,13,24,25].
lence occurs, which accounts for the second peak of heat transfer Unfortunately, it is found the dip and secondary peak are also
[10]. However, the alternative explanation for the second peak is obtained with difficulty for turbulence models without further
caused by the reattachment of the recirculation of the boundary modifications. To achieve the goal, the low Reynolds turbulence
layer of the unsteady separation [4]. Rohlfs et al. [7] provides the models based on the laminar-transition theory have been used
similar explanation that the interactions of the second vortex with popularly in recent years [11,24,26,27]. However, the low
the large scale vortex is the reason for the local increase of heat Reynolds-turbulence models do not have the proper sensitivity to
transfer rate. And Aillaud et al. [2] agreed with the above idea strong pressure gradients. As a result, it is found that an early
and pointed out that the local minimum of heat transfer is the ini- dip, the excessive velocity profiles and skin friction are obtained
tiation of the unsteady separation. in RANS models [21,22,26–29]. Based on early studies, Zuckerman
With consideration to the highly complex phenomenon, getting and Lior [30] pointed out that the standard SST k-x model can
a better understanding of jet impingement and simulating accu- obtain the fair performances in heat transfer. So in many jet simu-
rately are attractive for researchers. Due to the low computational lations, the standard SST k-x model is used to optimize the param-
cost and time consume, the Reynolds-Averaged Navier-Stokes eters [31] and study the flow structures [32]. However, the
(RANS) method has been widely used. However, for two equation standard SST k-x model predicts a false secondary peak for heat
models, the high production of turbulent kinetic energy k leading transfer at high nozzle-plate spacing [22,26]. In order to choose
to a high inner peak in the stagnation region has been found in an appropriate turbulence model, many other researchers have
early studies [11–15]. These excessively high levels of turbulent been doing many comparative studies. For the recent computa-
kinetic energy k will convect downstream and cause early transi- tional development, Anupam et al. [33] reviewed many turbulence
tion [16]. To overcome this problem, Kato and Launder [17] pro- models for jet impingement. Most importantly, they also recom-
posed the vorticity- and strain-based production term (Pk = ltXS) mended the standard SST k-x model for the jet impingement heat
in k equation. The behind reason is that the vorticity is almost zero transfer. With the development of the turbulence models, the four-
in the stagnation, which leads to low production of turbulent equation transition SST model was studied by the Afroz and Sharif
kinetic energy k. Wienand et al. [15] studied the round impinging for the annular impinging flow [34]. Their comparative results
jet for H/D = 2, 6 and 10 using the Kato-Launder model. They shows that the Nusselt number predicted by the transition SST
observed that the Kato-Launder model not only reduces the high model do not match well with the experimental data for any of
turbulent kinetic energy at the stagnation point, but also helps the radial locations. In addition, the one equation intermittency
the standard SST k-x model regenerate the second peak of heat transition model based on the standard SST k-x model was also
transfer. The good performance of the Kao-Launder model was also studied by Huang et al. [14]. However, the intermittency transition
confirmed by Huang et al. [14]. model overpredicts the distribution of the skin frictions for the slot
In addition, as discussed before, the adverse pressure gradient jet impingement. Further, the v2-f model of Durbin [35,36], Direct
occurs owing to the impermeability constrain, playing an impor- Numerical Simulation (DNS) [37], Large Eddy Simulation (LES)
tant role in terms of heat transfer and velocity field for jet impinge- [4,7,38] and RANS/LES [39] model have been applied in recent
ment. And this is an important feature for two equation turbulence years. These models have shown good performances in predicting
models to get the accurate results. Wilcox [18] developed a stress not only the secondary peak, but also the flow fields for round jets.
limiter in terms of eddy viscosity to recover the Bradshaw’s However, for one hand, these methods are not computationally
assumption. The similar stress limiter can be found in Menter’s efficient enough for industrial design and engineering environment
SST k-x model [19,20]. These features are considered as the reason [40]. For another hand, the good performances for round jet may
that it is correctly to predict the heat transfer and turbulent flow not be found in plane jets [33]. As a result, it may cause somewhat
around the stagnation region [21,22]. As Menter [19] said, above overestimation or underprediction if these model are used in the
corrections are activated in an adverse pressure gradient boundary practical design.
layer including the logarithmic and wake parts of the boundary Thus, under the premise of keeping the advantages of RANS
layer, while the k-x model is active everywhere [21]. Another methods, it is strongly demanded to develop a new turbulence
702 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712


model. To our knowledge, for a new modification for heat transfer @k @k
@ @k
þ uj ¼ min P k ; 10b0 kx  b kx þ ð m þ m t rk Þ ð5Þ
impingement, it should be sensitive to the pressure gradients, @t @xj @xj @xj
fairly easy to be incorporated into existing computational fluid
h i
dynamics (CFD) codes and computationally inexpensive [41]. In c b x
@x
@t
þ uj @@xxj ¼ qa xk min G; 1a10 maxða1 x; b1 F 23 SÞ
this paper, a turbulence model has been developed, which is com- h i ð6Þ
patible with the recommended standard SST k-x model and bx2 þ @x@ ðm þ mt rx Þ @@xx þ 2ð1  F 1 Þ rxx2 @x
@k @ x
@x
j j j j
applies the modification to reinforce the sensitivity for pressure
gradient. To evaluate the properties of the developed model, four where P k ¼ G ¼ mt S2 . The turbulence constants are: b0 ¼ b ¼ 0:09,
typical impinging jets with nozzle-plate spacing of 2, 4, 6 and 9.2 rk1 ¼ 0:85, a1 ¼ 5=9, rx1 ¼ 0:5, b1 ¼ 0:075, b1 ¼ 1:0, c1 ¼ 10:0,
are investigated in terms of heat transfer and flow structures. rk2 ¼ 1:0, a2 ¼ 0:44, rx2 ¼ 0:856, and b2 ¼ 0:0828.
The results are compared to the available experiments and numer-
ical results of others. 2.3. The developed SST k-x model
In this paper, Section 2 describes the numerical methodology,
the modifications to SST k-x model and the grid independence. As we know, the standard k-x model is developed from the
Section 3 shows the results for two nozzle-plate spacing of jets standard k-e model. The transforming procedure is as follows:
in terms of pressure distribution, velocity profiles, the skin friction The standard k-x model:
and surface Nusselt number. Finally, the main findings and conclu-
Dqk @ui @ @k
sions are presented in Section 4. ¼ sij  b qxk þ ðm þ rk1 mt Þ ð7Þ
Dt @xj @xj @xj
2. Numerical methodology
Dqk c1 @ui @ @x
¼ sij  b1 qx2 þ ðm þ rx1 mt Þ ð8Þ
Based on the above findings, the work of modifications has been
Dt lt @xj @xj @xj
done using the open source field operation and manipulation Transformed k-e model:
(OpenFOAM) platform [42] to ensure the codes’ accuracy and
Dqk @ui @ @k
robustness. OpenFOAM is a free-source CFD package that provides ¼ sij  b qxk þ ðm þ rk2 mt Þ ð9Þ
the C++ source codes. In this platform, the existing calculation Dt @xj @xj @xj
modules allow the user to achieve desired purposes in an efficient
Dqk c2 @ui @ @x 1 @k @ x
way. ¼ sij  b2 qx2 þ ðm þ rx2 mt Þ þ 2qrx2
Dt lt @xj @xj @xj x @xj @xj
2.1. Governing equations ð10Þ
As it can be seen in Eq. (10), the standard k-e model possesses
The time-averaged governing equations include the conserva-
the cross-diffusion term, i.e. the last term in the equation, which
tion of mass, momentum and energy which are given as:
is of implicit feature [18]. However, the standard k-x model [43]
@ui does not hold the cross-diffusion term in the boundary layer
¼0 ð1Þ
@xi including the viscous sublayer and logarithmic layer. And it is
the same with the standard SST k-x model, as shown in Eq. (6).
@u @ui @p @   The explanation about the absence of the cross-diffusion term
q þ quj ¼ þ 2mSij  qu0i u0j ð2Þ
@t @xj @xi @xj given by Wilcox [18] is based on the performances of the standard
k-x model and the standard k-e model in predicting the turbulent
 
@T @T @ k @T flows, especially wall-bounded flows, but not a strict sequitur.
q þ quj ¼  qu0i T 0 ð3Þ However, the cross-diffusion term is strongly linked with the pres-
@t @xj @xj cp @xj
sure gradient since the gradient of turbulent kinetic energy k is
where u is the velocity, q is the density, p is the pressure, nonvanishing for a non-constant pressure boundary layer. Mean-
 
@u
@ui
Sij ¼ 12
@x
þ @xj and m, k and cp are dynamic viscosity, thermal con- while, the pressure gradient influences the process of laminar-
j i
turbulence transition, especially in the separated flows. Therefore,
ductivity and specific heat capacity at constant pressure, respec-
taking the effect of cross-diffusion term into consideration, it
tively, T is the temperature, u0i u0j is the Reynolds stress and u0i T 0 is would make sense of simulating jet impingement. And this is the
the turbulent heat flux. behind reason that the new model is proposed.
As mentioned above, the heat transfer and flow structures are
2.2. Standard SST k-x model influenced by the adverse pressure gradients. And Wilcox [44]
attributes the success of the k-x model in modelling adverse pres-
The standard SST k-x model [19] retains the accurate formula- sure gradient boundary layers to the lack of cross-diffusion param-
tion of Wilcox k-x model near walls, and uses the k-e model to eter vk ¼ x13 @x
@k @ x
. To reinforce the model’s ability to predict the
j @xj
avoid the freestream dependence away from walls. In the Open-
adverse pressure gradient, based on Wilcox’s work, b in destruc-
FOAM package [42], the production limiter is added in the standard
tion term (Eq. (5)) is replaced as follows:
SST k-x model. And the eddy viscosity is defined as follows:
bf ¼ b1 F 1 þ ð1  F 1 Þb2 ð11Þ
1
mt ¼ qa1 k ð4Þ
maxða1 x; b1 SF 2 Þ where b1 ¼ b0
1þ680v2k
, and b2 ¼ b0 ¼ 0:09.
1þ80v2k

where a1 is the experimental constant with the value 0.31, k is the It is difference from the standard k-x model [44], in which the
turbulent kinetic energy, x is the specific dissipation rate, b1 is 1.0, S change is only activated in free shear flows. While the proposed
is the strain rate magnitude and F2 is the blending function, which is change varies throughout the logarithmic and wake parts of the
n   pffiffi  o2 
m ; 100 boundary layer and goes to 0.09 in the free shear flows using the
defined as F 2 ¼ tanh min max b2 xky ; q500 .
y x
blending function F1. Based on the performances of standard k-x
2

The transportation equations of k and x are defined by model, to keep the good performance near walls for the new
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 703

model, it is essential to activate the cross-diffusion parameter only 2.4. Description of the cases
above the viscous sublayer. Fig. 1 shows the value of bf versus the
dimensionless wall distance y+. In near walls, the value of bf is Fig. 2 shows the computational domain for studied two cases.
almost equal to 0.09 since x is a very large value and F 1 is unity. The nozzle width B is 0.04 m, which is the same with the experi-
Thus, in the near walls the developed model switch to the standard ment of Zhe and Modi [47]. And the length of impinging plate is
k-x model. In the free shear flows, F1 becomes zero, bf is equal to L. H is the impinging distance. Because of symmetry, one half the
b2 , so the standard k-e model is used for the simulation. domain is studied in this work.
The details of boundary conditions in OpenFOAM environment
For the logarithmic and wake parts of the region, it is found that
are shown in Table 1. For the inlet conditions, the Reynolds number
above two regions are strongly linked with the adverse pressure
is 20,000 based on the nozzle width for two nozzle-plate spacing of
gradients through the mathematical analysis [45]. This idea has
4 and 9.2. The inlet turbulent intensity I is 0.01, which is the same
been confirmed by other researchers [19,43,46]. In addition, the
with the experiments [47,48]. The characteristic turbulent length
parameter vk is almost zero in a constant-pressure boundary layer
scale lc is defined as lc ¼ 0:015B. And then, inlet turbulent kinetic
because the turbulent kinetic energy k is almost a constant. In that
case, bf is equal to b , which means that the proposed model per- energy kin is defined by kin ¼ ðV in IÞ2 . Inlet specific dissipation rate
pffiffiffi
forms like the standard SST k-x model. However, in a non- xin is defined as xin ¼ k=27lc b0:25 , and inlet temperature T is
constant-pressure boundary layer, vk cannot be ignored [18]. 300 K. The Impinging and confined plates are both considered to
Therefore, the cross-diffusion term is activated and control the be isothermal and their temperatures are 310 K and 300 K, respec-
destruction of turbulent kinetic energy k. Thus, it is easily draw tively. The value of Prandt number (Pr) is 0.72. At the outlet, the
from the analysis that behaviors for the developed model should pressure outlet and zero-turbulence condition is imposed. So that
be the same with the standard SST k-x model for zero-pressure the ‘‘pressureInletOutletVelocity” is considered for the velocity
gradient flows. field at the outlet. In addition, in order to simulate the low-
Another important modification is the recommended Kato- Reynolds-number flows, a very small value of k is applied to the
Launder model [15] calibrated with the developed model. Hence, wall and the ‘‘omegaWallFunction” is used, which is designed for
the production term Pk in Eq. (5) is redefined as follows: both low and high-Reynolds-number flows. For the constant
properties, such as density and dynamics viscosity, the value can
be obtained according to the jet exit temperature. For details,
Pk ¼ mt SX ð12Þ
constant fluid property of density is 1.1716 kg/m3. Dynamics
  viscosity is 1.835  105 kg/m∙s. The thermal conductivity is
where Xij ¼ 12 @U
@xj
@U
 @x . 0.0263 W/m∙K and the specific heat capacity at constant pressure
i
is 1005.5 W/kg∙K.
The constant density has been applied for both cases. Since the
maximum difference in temperature is 10 K, the application
0.8 ‘‘buoyantBoussinesqSimpleFoam” is chosen and the semi-implicit
method is used for pressure-velocity coupling. The gradient
0.7 scheme is ‘‘Gauss linear”. For the divergence schemes, the terms
of advection of velocity, temperature, x and k are discretized using
0.6 the ‘‘bounded Gauss Upwind”. The ‘‘Gauss linear corrected” is
applied to the Laplacian terms. For the interpolation schemes, the
0.5 ‘‘linear” scheme is considered. In addition, the surface normal gra-
dient scheme is ‘‘corrected”. For each case, when the sum of resid-
βf

0.4
uals is less than 1E5 for the terms of pressure, velocity,
0.3 temperature, k and x, the solution is assumed to be convergent.

0.2
2.5. Grid independence
0.1
Initially, the ANSYS-ICEM is used to create the quadrilateral
0.0 mesh. And then, use the application ‘‘fluentMeshToFoam” to con-
1 10 100 vert into FOAM mesh format (Hexahedral mesh). Table 2 shows
y+ the number of nodes in X and Y directions for different mesh
schemes for both cases. And Fig. 3 shows the nodes distributions
Fig. 1. bf* versus y+. of M2 meshes. For each mesh scheme, the first node to the wall

inlet Mean velocity profile


Confined plate

B/2
symmetry

outlet

Y
X

Impinging plate L=50B

Fig. 2. The model geometry.


704 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712

Table 1
The boundary conditions for the cases in OpenFOAM environment.

Inlet Outlet Impinging plate Confined plate


k 2
kin ¼ ðuin IÞ 0 0 0
u uin ¼ Relt =B pressureInletOutletVelocity noSlip noSlip
pffiffiffi
x xin ¼ k=27lc b0:25 0 omegaWallFunction omegaWallFunction
T 300 K 300 K 310 K 300 K
@P @P @P
@n ¼ 0 ¼0 ¼0
P fixedValue @n @n

mesh are almost the same, as shown in Fig. 4(b). However, the
Table 2 M1 mesh predicts lower results than other meshes and the exper-
The distributions of nodes for both cases.
imental data in the region of 2  x/B  4. Therefore, the following
Mesh M1 M2 M3 studies for the case of H/B = 9.2 are performed by using M2 mesh.
Nodes (X*Y) for H/B = 4 280  80 310  100 368  126
Nodes (X*Y) for H/B = 9.2 170  50 240  70 339  100 3. Results and discussion

In this section, results of the two cases of nozzle-plate spacing


y+ is located at a distance y+  1, which ensures enough grid nodes
of 4 and 9.2 are discussed in detail, where the cases have been
distributed in the viscous near-wall region. The local Nusselt num-
widely used to study the turbulence models in predicting heat
ber is used to study the grid independence, which is defined as
transfer [26–28]. To validate and evaluate the present work, the
Nu ¼ ð@T=@yÞw =½ðT w  T in Þ=B, where Tw is the impingement plate
results are compared with both available experimental data
temperature, Tin is the inlet jet temperature.
[47,48] as well as the numerical results [21,22,26] for the same
Fig. 4 shows the results of Nusselt number distributions using
configurations in terms of pressure distribution, mean velocity
the developed SST k-x model with three meshes termed as M1–
profiles, skin friction and the Nusselt number distributions. In
M3 presented in Table 2. For the case of H/B = 4, as shown in
addition, different computational conditions for H/B = 2 and 6 are
Fig. 4(a), it is observed that the Nusselt number distribution pre-
also presented to investigate the effect of pressure on heat transfer.
dicted by M1 mesh shows obvious deviations comparing to other
To show the improvements of modifications, results of the stan-
meshes, where the results of M2 mesh and M3 mesh almost over-
dard SST k-x model without any other optimizations are also
lap with each other. In addition, the maximum percentage error
presented.
between M2 mesh and M3 mesh is 2.5% at x/B = 8.2. Most impor-
tantly, the results of M2 mesh and M3 mesh match well with the
3.1. Pressure distribution
experimental data around the local minimum. Downstream, the
results of the M2 and M3 mesh also show a good trend comparing
Due to the importance of the pressure gradient for jet impinge-
with the experimental data. Thus, the M2 mesh is considered to
ment, evaluating the model’s ability in predicting the pressure gra-
meet the requirement of grid convergence. And all the following
dient becomes crucial. Results of pressure distribution will affect
results presented for H/B = 4 are obtained by using M2 mesh. For
the velocity and heat transfer which will be discuss in the latter
the case of H/B = 9.2, it can be also observed that the Nusselt num-
Sections. Fig. 5 shows the static pressure distributions for both
ber distributions along the plate predicted by M2 mesh and M3
cases.

(a) H/B=4

(b) H/B=9.2
Fig. 3. The distribution of M2 meshes used in two cases.
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 705

AF97 [48] 100 AF97 [48]


70
M1 M1
60 M2 80 M2
M3 M3
50
60
40
Nu

Nu
30 40
20
20
10

0 0
0 5 10 15 20 0 5 10 15 20
x/B x/B
(a) H/B = 4 (b) H/B=9.2
Fig. 4. The Nusselt number distributions using the developed SST k-x model with three meshes.

It can be found that the maximum static pressure Pmax occurs at constant [4,10]. However, the behind physics of the round jet
the stagnation point owing to the impermeability constrain. and the slot jet are different. Previous studies including the exper-
Around the stagnation point, jet deflects corresponding to the iments and numerical studies [4,7,49,50] have shown that the sep-
streamline curvature. So that the high shear stress occurs, leading aration zone is formed around the second peak of Nusselt number,
to a decrease of pressure and the flow acceleration zone. To be which indicates the pressure increases at a specific position. And
noted, the steep pressure distributions for the case of H/B = 9.2 the same conclusion can be also drawn from the mean velocity
are border than the case of H/B = 4. The reason is that the plate is profiles for the experiment of Ashforth-Frost et al. [48]. For high
within the potential core at H/B = 4 [48]. When the nozzle-plate nozzle-plate spacing, the developed SST k-x model and the stan-
distance increases, the momentum reduces and it is associated dard SST k-x model are almost overlap with the experimental data
with the increase of turbulence, which can be draw from the veloc- downstream as shown in Fig. 5(b). Most importantly, the results
ity field in the experiment [48]. Downstream, the nearly zero pres- predicted by the developed SST k-x model match well with the
sure gradient has been observed by the experiment [48] for both experimental data. However, in the stagnation region, the standard
cases. And the standard SST k-x model also predicts the same phe- SST k-x model predicts a higher absolute pressure gradient and a
nomenon. While, for the case of H/B = 4, the difference between the shorter pressure decreasing zone than the experimental data. This
experimental data and the numerical results of the developed SST indicates that the standard SST k-x has obtained a shorter acceler-
k-x model occurs, as shown in Fig. 5(a). The developed SST k-x ation zone and higher velocity profiles. And this point will be also
model obtains an adverse pressure gradient, which shows that further discussed in Section 3.2.
the pressure increases slightly in the region of x/B  3. Colucci
and Viskanta argued that the constant pressure distribution 3.2. Mean velocity profiles
obtained in experiment is due to the jet spreading, when they stud-
ied the round jet. The jet spreading is prevail at downstream, and Fig. 6 shows the obtained mean x-component velocity along the
therefore the velocity decreases but the pressure is nearly a y direction profiles by comparing with the experimental data and

1.0 AF97 [48] 1.0 AF97 [48]


Standard SST k-ω
Standard SST k-ω
0.8 Developed SST k-ω
0.8 Developed SST k-ω

0.6 0.6
P/Pmax

P/Pmax

0.4 0.4

0.2 0.2

0.0 0.0
0 2 4 6 8 10 0 2 4 6 8 10
x/B x/B
(a) H/B = 4 (b) H/B = 9.2
Fig. 5. The static pressure distribution along the impinging plate for both cases.
706 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712

1.0
1.0

0.8
0.8
AF97 [48]
0.6 ZM00 [47]
0.6

U/V in
AF97 [48]
U/Vin

ZM00 [47] Standard SSTk-ω


0.4 Standard SST k-ω Developed SST k-ω
0.4
Developed SST k-ω k-ω-v2-f uniform [22]
k-ω-v2-f uniform [22] SST k-ω with transition [26]
0.2 0.2
SST k-ω with transition [26] RANS/LES, M2 [21]
RANS/LES, M2 [21]
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
y/B y/B
(a) x/B = 1 (b) x/B = 2
1.0
1.0

0.8
0.8

0.6
0.6
U/V in

AF97 [48]
U/Vin

AF97 [48] ZM00 [47]


0.4 ZM00 [47] 0.4 Standard SST k-ω
Standard SST k-ω Developed SST k-ω
Developed SST k-ω 0.2 k-ω-v2-f uniform [22]
0.2
SST k-ω with transition [26] RANS/LES, M2 [21]

0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
y/B y/B
(c) x/B = 6 (d) x/B = 7
Fig. 6. Results of mean x-component velocity for the case of H/B = 4.

the numerical results for the case of H/B = 4. However, the figures RANS/LES model also obtains proper results, as shown in Fig. 6(a)
show that the two sets of experimental data are of deviations from and (b).
each other in all positions, where the difference is relatively small As shown in Fig. 6(c) and (d), at x/B = 6 and x/B = 7, the flow
at x/B = 7. Due to large number of influential parameters and exper- decelerates when the jet spreads into downstream [4]. When this
imental conditions investigated, it is no surprise that the two happens, a vortex in the upper part near the confinement plate
experimental data show these differences. Despite of this, both develops [49]. In this region, the results of the developed SST k-
experiments are also taken into account to get objective evalua- x model and the RANS/LES model agree very well with the exper-
tions for models. imental data downstream. However, the results of the SST k-x
In the stagnation region of 1  x/B  2, attributing to the imper- with transition model, the k-x-v2-f model and the standard SST
meability constrains, the jet deflects corresponding to the stream- k-x model strongly depart from the experimental data and obtain
line curvature. So that a strong shear occurs in the stagnation the similar velocity profiles at both positions, which indicates that
region, leading to a decrease of pressure. Thus the velocity increase these models generates a delay deceleration zone. Considering the
away from the stagnation point. So that the maximum velocity at pressure distribution of the standard SST k-x model at x/B = 2 and
x/B = 2 is higher than the position of x/B = 1, as shown in Fig. 6(a) x/B = 4, although we do not have pressure distributions of other
and (b). As mention above, within 1  x/B  2, the flow accelerates models, it can be speculated that the reason for the delay deceler-
due to a reduction of pressure. So, good performance in predicting ation zone is that these models predict an almost zero pressure
pressure distributions for standard SST k-x model and the devel- gradient downstream. It seems that the pressure plays a key role
oped SST k-x model in the stagnation region also reflects in the downstream, although previous studies show that jet spreading
velocity profiles, as shown in Fig. 5(a), where the results obtained is prevail downstream in round jet impingement [4,10]. According
by the two models are close to the experimental data. In addition, to the research of W. Rohlfs [7], the separation zone is found
there is a fact that at the stagnation point, there is a thin boundary around the second peak, implying that there is an adverse pressure
layer because of high momentum along the axis. So around the gradient along the plate. It may indicate that the SST k-x with
stagnation region, the turbulent intensity is very low [21]. For transition model, the k-x-v2-f model and the standard SST k-x
the RANS/LES model, the LES model is activated within the imping- model are insensitive to the pressure gradients. And this well
ing region with nearly zero turbulence intensity. And therefore, the explains that the good performance of the RANS/LES model which
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 707

inherits the advantages of LES model in solving the vortex problem zone as well as deceleration zone. For the two cases with different
at x/B = 7. nozzle-plate spacing values, the good performance of the present
Fig. 7 shows the results of dimensionless x-direction component model is that it can predict accurate pressure gradient, especially
mean velocity distributions along the y direction for the case of in the stagnation region, which affects the results of velocity
H/B = 9.2. In the stagnation regions of x/B = 1 and x/B = 2, all the spreading rates.
numerical results show obvious discrepancies from the experimen-
tal data, where, however, the results of the RANS/LES model and
the developed SST k-x model are relatively closer to the experi- 3.3. Skin friction
mental ones. The standard SST k-x model and the k-x-v2-f model
overestimate velocity profiles, owing to a faster decrease of pres- Fig. 8 shows the results of dimensionless skin friction coefficient
 
sure as shown in Fig. 5(b). The steep negative pressure gradients Cf for both cases, where Cf is defined as C f ¼ sw = 0:5qV 2in . The
are associated with too weak turbulent mixing in the shear layer
error of experimental data tested by Zhe and Modi [47] is 14%
of the jet [22], which are attributed to the high momentum in
and up to 19% in the stagnation region based on an uncertainty
the wall jet region. Downstream, the proposed developed SST k-
of 11% in mean velocity and 9% in measuring the distance from
x model has provided very good results in the two positions of the wall. At the stagnation point, skin friction is of zero due to no
x/B = 3 and x/B = 5 by comparing with the experimental data, and
streamwise flow on the impinging plate. Near the stagnation point,
the RANS/LES model provides good results in the latter position.
strong streamline curvature occurs, which leads to the maximum
In addition, it clearly shows that the position of x/B = 5 locates in
streamwise velocity and the value of skin friction reaches the peak.
the deceleration region. Comparing with the position of x/B = 5, it
For the case of H/B = 4, in recent years, Hadziabdic and Hanjalic
can be found that the maximum velocity at x/B = 3 decreases
[4] found that the local minimum is a consequence of the unsteady
slower than the case of H/B = 4, as shown in Fig. 7(c) and (d). And
separation, linking with the thick thermal boundary layer. And
this also reflects that a nearly constant pressure distribution down-
therefore, the local minimum value at x/B = 3 is very similar to
stream exists, as shown in Fig. 5(b).
the local minimum of the Nusselt number (Fig. 9(a)). Consequently,
From this configuration, it may be concluded that the present
the improvements in predicting the pressure gradients will also
developed SST k-x model could predict correct flow acceleration
reflect in terms of skin friction. For this case, the developed SST

0.8 1.0

0.7
0.8
0.6
0.5 0.6
U/V in

U/Vin

0.4 AF97 [48] AF97 [48]


ZM00 [47] 0.4 ZM00 [47]
0.3 Standard SST k-ω
Standard SST k-ω
0.2 Developed SST k-ω
Developed SST k-ω 0.2 k-ω-v2-f uniform [22]
0.1 k-ω-v2-f uniform [22]
RANS/LES, M1 [21]
RANS/LES, M1 [21]
0.0 0.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
y/B y/B
(a) x/B = 1 (b) x/B = 2
1.0
1.0

0.8 0.8

0.6 0.6
U/Vin
U/Vin

AF97 [48]
AF97 [48]
0.4 ZM00 [47]
0.4 ZM00 [47]
Standard SST k-ω
Standard SST k-ω
0.2 Developed SST k-ω
Developed SST k-ω
0.2 k-ω-v2-f uniform [22]
0.0 RANS/LES, M1 [21]
0.0
0.0 0.1 0.2 0.3 0.4 0.0 0.1 0.2 0.3 0.4
y/B y/B
(c) x/B = 3 (d) x/B = 5
Fig. 7. Results of mean x-component velocity distributions along the y direction for the case of H/B = 9.2.
708 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712

0.012 ZM00 [47] 0.012


Standard SST k-ω ZM00 [47]
0.010 Developed SST k-ω 0.010 Standard SST k-ω
k-ω-v2-f uniform [22] Developed SST k-ω
0.008 RANS/LES, M2 [21] 0.008 RANS/LES, M1 [21]
Cf

Cf
0.006 0.006

0.004 0.004

0.002 0.002

0.000 0.000
0 2 4 6 8 10 0 2 4 6 8 10
x/B x/B
(a) H/B=4 (b) H/B=9.2
Fig. 8. Results of skin friction distribution for different nozzle-plate spacing values.

60 Cadek [53] 100


AF97 [48]
Developed SST k-ω 90 Standard SST k-ω
50 Standard SST k-ω Developed SST k-ω
80 SST k-ω with transition [26]
40 k-ω-v2-f uniform [22]
70
RANS/LES, M2 [21]
Nu

Nu

30 60

50
20
40
10
30

0 20
0 5 10 15 20 0 2 4 6 8 10 12 14
x/B x/B

(a) H/B = 2, Re = 11,400 (b) H/B = 4, Re = 20,000

60 Gardon & Akfirat [6] 120


AF97 [48]
Cadek [53] Standard SST k-ω
50 Developed SST k-ω 100
Developed SST k-ω
Standard SST k-ω SST k-ω with transition [26]
40 80
k-ω-v2-f uniform [22]
RANS/LES, M1 [21]
Nu

Nu

30 60

20 40

10 20

0 0
0 5 10 15 20 0 2 4 6 8 10 12 14
x/B x/B
(c) H/B=6, Re = 11,000 (d) H/B=9.2, Re = 20,000
Fig. 9. The distribution of Nusselt number along the impinging plate for different nozzle-plate spacing.

k-x model has reproduced better values than other numerical experimental data. Also the proposed model predicts an obvious
models. In addition, the trend of Cf as well as the position of dip second peak of skin friction, although the position is a little earlier
(x/B = 3) provided by the proposed model matches well with the than the experimental data. However, as mentioned in Section 3.2,
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 709

because of higher shear stress produced by RANS/LES model which k-x model for both cases, especially for H/B = 2. For the case of
can be seen in Fig. 8(a), it also predicts an earlier local minimum of H/B = 4, it has been found that the standard SST k-x model repro-
skin friction. And the similar performance can be found in the duces higher Nusselt number value at the stagnation point than
k-x–v2-f model. Based on the observations discussed above, the the experimental data. As mentioned above, two equation models
performances of skin friction distributions of the hybrid models usually overpredict turbulent kinetic energy k at the stagnation
of k-x-v2-f and RANS/LES will lead to a high velocity profile in point, which is the reason that high Nusselt number at the stagna-
the stagnation region. Unfortunately, it found that just the tion point and premature dip are predicted by the standard SST k-
RANS/LES model obtains a higher velocity profile, while a lower x model. For the developed SST k-x model, the accurate results
velocity profile is predicted by the k-x-v2-f model, as shown in around the stagnation point are attributed to the Kato-Launder
Fig. 6(a). The possible explanation is that enormous influential model, as discussed previously in Section 2.2. After the dip, the
parameters make it difficult to do good comparisons in all aspects. maximum errors of the Nusselt number value between the exper-
Further, Wilcox [18] argued for the standard k-x model plat- imental and numerical results are 40%, 24%, 15%, 15% and 24% for
form, it is usually predicts an earlier local minimum of skin friction the standard SST k-x model, the developed SST k-x model, the
because of constant turbulence numbers in the viscous sublayer. SST k-x with transition model, the k-x-v2-f model and the
And this also can be seen from the performance of the standard RANS/LES model, respectively. Thanks to the modifications consid-
SST k-x model shown in Fig. 6(a). The standard SST k-x model pre- ering the effect of pressure gradients, the developed SST k-x model
dicts a second peak and a dip. But the value of the peak and the has reproduced an accurate dip and a close second peak than other
position of the dip are higher and earlier than the ones predicted referred models, which is also reflected in the skin friction distribu-
by the developed SST k-x model, respectively. Comparing the tions shown in Fig. 8(a). However, the trend of the developed SST
results of the standard SST k-x model with the developed SST k- k-x model depart from the experimental data because the effect
x model, it reveals that considering the effect of the cross- of cross-diffusion term reduces turbulent length scale
diffusion term in logarithmic and wake parts of the boundary layer Lt ¼ k =ðb xÞ with an increase of b*, leading to an increase of dis-
2=3

also helps to improve the prediction of skin friction for jet sipation. Also, in RANS/LES model, its turbulent length scale is
impingement. replaced by the local grid size, indicating a decrease of length scale
For the case of H/B = 9.2, the skin friction monotonically when LES model activates. According to Pope’s research [51], when
decreases after the first peak along the impinging plate, as shown there is a vortex stretching, there is a reduction in length scale.
in Fig. 8(b). In this case, the standard SST k-x model also provides Most importantly, the second vortex around the second peak is
higher values than the experimental data and other numerical found by other researchers using the DNS or the LES methods
results, which is reflected in terms of mean velocity, as shown in [2,52]. So, this explains that the developed SST k-x model and
Fig. 7. In addition, the RANS/LES model and the developed SST k- the RANS/LES model improve the ability of predicting the second
x model have the similar behaviors, and the results of these two peak and velocity profiles. Also, it may be the cause that the devel-
models agree well with the experimental data. However, consider oped SST k-x model and the RANS/LES model have the similar per-
the position of peak, the developed SST k-x model reproduces formances not only in terms of heat transfer but also flow fields.
more accurate results, while a delay position is predicted by the For the case of H/B = 6 and 9.2, obvious improvements can be
RANS/LES model. found for the results of developed SST k-x model, as shown in
Fig. 9(c) and (d). As mentioned above, the impinging plates are
within the developed jet region. The main difference between the
3.4. Surface Nusselt number
low cases and the high cases is that the impinging plate is out of
the potential core of the jet for the high cases. Usually, the length
In this subsection, the large rage of different nozzle-place spac-
of the potential core is 6B–8B from the jet exit, depending to the
ing of 2, 4, 6 and 9.2 are presented to validate the current method
outlet conditions [48]. Consequently, the distributions of Nusselt
in predicting the jet impingement heat transfer. For H/B = 2, the
number is almost monotonic for above two cases. For H/B = 6, it
Reynolds number is 11,400. Jet exit temperature and the impinging
can be found that there is a slight second peak around x/B = 6 for
plate temperature are 348 K and 310 K, respectively. The jet exit
both experimental data [6,53]. And this feature is captured by
turbulent intensity is considered as 4%. For H/B = 6, the Reynolds
the developed SST k-x model as shown in Fig. 9(c). However, the
number is 11,000. Jet exit temperature and the impinging plate
standard SST k-x model obtains a false trend at downstream. For
temperature are 373 K and 338 K, respectively. And the jet exit tur-
H/B = 9.2, all the models, except the developed SST k-x model
bulent intensity is 2%. Other detailed boundary conditions can be
and the RANS/LES model, have provided a false secondary peak
computed according to Section 2.4. For H/B = 4 and H/B = 9.2, the
obviously. And the results of the developed SST k-x model agree
boundary conditions have been described in Section 2.4. It can be
well with the experimental data. However, the RANS/LES model
found that the second peak of Nusselt number exists for low
provides higher results in the region of 0  x/B  6. In addition,
nozzle-plate spacing and disappears for the case of H/B = 9.2.
downstream, the results of the standard SST k-x model and the
Fig. 9 shows the results of Nu for different nozzle-plate spacing
SST k-x with transition model overlap with each other. The reason
under varied conditions. For H/B = 2 and 4, the second peak is
is that the transition is slight. Owing to the present modifications,
clearly presented. For these low nozzle-plate spacing, the imping-
the developed SST k-x model makes a slight false secondary peak
ing plates are within the potential core of the jet. The maximum
and is sensitive to the pressure gradients, which makes the effects
Nusselt number is obtained at the stagnation point by the experi-
of unsteady separation affected by the pressure gradients are esti-
ment due to the thin boundary layer. Away from the stagnation
mated accurately. To get insight this point, the distributions of
point, as mentioned above, the shear layer boundary grows with
pressure for all cases are plotted in Fig. 10. To convenience, the dis-
the spanwise velocity as well as the turbulent intensity. Thus,
tributions of pressure for H/B = 4 and 9.2 are presented here again.
the heat transfer rate drops down to a dip at x/B = 3 for both cases
As shown in Fig. 10, when the nozzle-plate spacing increases,
until the unsteady separation occurs [4]. After the dip, the Nusselt
the pressure tends to a constant at downstream. For the standard
number increases to a second peak due to the reattachment of the
SST k-x model, it can be found that the adverse pressure gradient
boundary layer of the unsteady separation. It can be found that the
is predicted at H/B = 2. While for other cases, the nearly zero-
developed SST k-x model performs better than the standard SST
710 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712

1.0 1.0
Developed SST k-ω AF97 [48]
Standard SST k-ω
Standard SST k-ω
0.8 Developed SST k-ω
0.8

0.6
P/Pmax

P/P max
0.6
0.4

0.4 0.2

0.0
0.2
0 2 4 6 8 10 0 2 4 6 8 10
x/B x/B

(a) H/B = 2, Re = 11,400 (b) H/B = 4, Re = 20,000

1.0 1.0 AF97 [48]


Developed SST k-ω
Standard SST k-ω
Standard SST k-ω
0.8 Developed SST k-ω
0.8

0.6
P/Pmax
P/Pmax

0.6
0.4

0.4 0.2

0.0
0.2
0 2 4 6 8 10 0 2 4 6 8 10
x/B x/B

(c) H/B = 6, Re = 11,000 (d) H/B = 9.2, Re = 20,000


Fig. 10. The results of distributions of pressure for different cases.

pressure gradient is regenerated by the standard SST k-x model at based on the Wilcox’s cross-diffusion correction and the Kato-
downstream. However, the developed SST k-x model obtains an Launder model, which are calibrated to the recommended stan-
adverse pressure gradient for H/B = 2 and 4 obviously. And in both dard SST k-x model. The proposed model has been carried out to
cases, the Nusselt number exhibits a second peak as shown in Fig. 9 study the turbulent slot impinging jet heat transfer for different
(a) and (b). At H/B = 6, a weak adverse pressure gradient can be also nozzle-plate spacing of 4 and 9.2, respectively. The heat transfer
found at x/B > 3 for the results predicted by the developed SST k-x and the flow structures results including the pressure distribution,
model. At this high nozzle-plate spacing, a slight second peak mean velocity, skin friction as well as local Nusselt number distri-
occurs for both experimental data and the results of the developed butions are studied in detail. Also, more boundary conditions of
model. In addition, the dips of the pressure for H/B = 2, 4 and 6 are slots of H/B = 2 and 6 are also studied in terms of heat transfer
almost the same with the associated local minimum of heat trans- and pressure in order to investigate the effect of pressure on heat
fer. Recalling the explanations for the second peak of heat transfer, transfer. Based on the present study, the modifications including
it is well connected to the unsteady separation. At H/B = 9.2, the the effects of cross-diffusion term can effectively reinforce the
pressure gradient is nearly a constant and the results of the devel- model’s ability to predict the heat transfer and flow structure for
oped model also predicts a monotonic distribution of heat transfer. slot jets. And there are some conclusions can be drawn from this
Based on the accurate results of flow fields predicted by the devel- work.
oped SST k-x model, as a result, second peak is linked with the
adverse pressure gradient for the turbulent slot jets. (1) Consider the effect of the cross-diffusion term, the present
modification has been proposed based on the Wilcox’s work
4. Conclusions and the Kato-Launder model. And it is the cause that of the
proposed model predicts the accurate results of flow
A new turbulence model has been developed for the jet structures and heat transfer of slot impinging jets at
impingement under the pressure gradients. The modifications are 2  H/B  9.2.
H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712 711

(2) The proposed model has predicted accurate results in terms [16] R.B. Langtry, S.A. Sjolander, Prediction of transition for attached and separated
shear layers in turbomachinery, in: 38th AIAA/ASME/SAE/ASEE Joint
of heat transfer and the flow structures for the four nozzle-
Propulsion Conference & Exhibit, AIAA, Indianapolis, Indiana, 2002, pp. 1–13.
plate spacing of 2, 4, 6 and 9.2. Especially in high nozzle- [17] M. Kato, B.E. Launder, The modelling of turbulent flow around stationary and
plate spacing, the proposed model improve the predictions vibrating square cylinders, Nith Symposium on ‘‘Turbulent Shear Flows”,
comparing to the standard SST k-x model but also predicts Kyoto, Japan, 1993.
[18] D.A. Wilcox, Turbulence Modeling for CFD, Birmingham Press, Inc., San Diego,
a weak secondary peak. The results show that the new California, 2006, pp. 125–127.
model has an appropriate performance in predicting the [19] F.R. Menter, Two-equation eddy-viscosity turbulence models for engineering
heat transfer at 2  H/B  9.2. applications, AIAA J. 32 (8) (1994) 1598–1605.
[20] F.R. Menter, Zonal two equation k-w turbulence models for aerodynamic
(3) Through the comparative study in this work, it shows that flows, in: 23rd Fluid Dynamics, Plasmadynamics, and Lasers Conference, Fluid
pressure plays an important role in turbulent slot jet Dynamics and Co-located Conferences, AIAA, Orlando, FL, USA, 1993, pp. 1–21.
impingement heat transfer, which shows pressure gradient [21] S. Kubacki, E. Dick, Simulation of plane impinging jets with k–x based hybrid
RANS/LES models, Int. J. Heat Fluid Flow 31 (5) (2010) 862–878.
affects the results of second peak of Nusselt number, velocity [22] E. Khalaji, M.R. Nazari, Z. Seifi, 2D numerical simulation of impinging jet to the
spreading rates as well as skin friction distributions. flat surface by k-x-v2-f turbulence model, Heat Mass Transf. 52 (1) (2016)
127–140.
[23] C.R. Yap, Turbulent Heat and Momentum Transfer in Recirculating and
Impinging Flows, University of Manchester, Manchester, UK, 1987.
Declaration of Competing Interest [24] S.M. Hosseinalipour, A.S. Mujumdar, Comparative evaluation of different
turbulence models for confined impinging and opposing jet flows, Num. Heat
Transf. Part A: Appl. 28 (6) (1995) 647–666.
The authors declared that there is no conflict of interest. [25] S.J. Wang, A.S. Mujumdar, A comparative study of five low Reynolds number
k–e models for impingement heat transfer, Appl. Therm. Eng. 25 (1) (2005) 31–
44.
Acknowledgement [26] R. Dutta, A. Dewan, B. Srinivasan, Comparison of various integration to wall
(ITW) RANS models for predicting turbulent slot jet impingement heat
transfer, Int. J. Heat Mass Transf. 65 (2013) 750–764.
This work was supported by the National Natural Science Foun- [27] J.E. Jaramillo, C.D. Pérez-Segarra, I. Rodriguez, A. Oliva, Numerical study of
dation of China (51579042, 51709042, 51639003), the Defense plane and round impinging jets using RANS models, Num. Heat Transf. Part B:
Industrial Technology Development Program (JCKY2017604C002), Fundam. 54 (3) (2008) 213–237.
[28] Q. Chen, V. Modi, Mass transfer in turbulent impinging slot jets, Int. J. Heat
the Fundamental Research Funds for the Central Universities Mass Transf. 42 (5) (1999) 873–887.
(DUT18RC(4)018, DUT2017TB05), the Natural Science Foundation [29] Z.U. Ahmed, Y.M. Al-Abdeli, F.G. Guzzomi, Flow field and thermal behaviour in
of Liaoning Province of China (20180550619), the China Postdoc- swirling and non-swirling turbulent impinging jets, Int. J. Therm. Sci. 114
(2017) 241–256.
toral Science Foundation (2018M631791) and the computation
[30] N. Zuckerman, N. Lior, Jet impingement heat transfer: physics, correlations,
support of the Supercomputing Center of Dalian University of and numerical modeling, Adv. Heat Transf. 39 (2006) 565–631.
Technology. [31] J. Ortega-Casanova, S.I. Castillo-Sanchez, On using axisymmetric turbulent
impinging jets swirling as Burger’s vortex for heat transfer applications. Single
and multi-objective vortex parameters optimization, Appl. Therm. Eng. 121
References (2017) 103–114.
[32] M. Fénot, X.T. Trinh, E. Dorignac, Flow and heat transfer of a compressible
impinging jet, Int. J. Therm. Sci. 136 (2019) 357–369.
[1] H. Kang, S. Baek, B. Ahn, Y. Yun, S. Kwon, Conceptual design of high-speed
[33] A. Dewan, R. Dutta, B. Srinivasan, Recent trends in computation of turbulent jet
underwater jet engine using high concentration of hydrogen peroxide, Ocean
impingement heat transfer, Heat Transfer Eng. 33 (4–5) (2012) 447–460.
Eng. 153 (2018) 193–200.
[34] F. Afroz, M.A.R. Sharif, Numerical study of turbulent annular impinging jet flow
[2] P. Aillaud, F. Duchaine, L.Y.M. Gicquel, S. Didorally, Secondary peak in the
and heat transfer from a flat surface, Appl. Therm. Eng. 138 (2018) 154–172.
Nusselt number distribution of impinging jet flows: a phenomenological
[35] P.A. Durbin, Near-wall turbulence closure modeling without ‘‘damping
analysis, Phys. Fluids 28 (095110) (2016) 1–22.
functions”, Theor. Comput. Fluid Dyn. 3 (1) (1991) 1–13.
[3] N. Didden, C.-M. Ho, Unsteady separation in a boundary layer produced by an
[36] T. Zhou, D. Xu, J. Chen, C. Cao, T. Ye, Numerical analysis of turbulent round jet
impinging jet, J. Fluid Mech. 160 (1985) 235–256.
impingement heat transfer at high temperature difference, Appl. Therm. Eng.
[4] M. Hadžiabdić, K. Hanjalić, Vortical structures and heat transfer in a round
100 (2016) 55–61.
impinging jet, J. Fluid Mech. 596 (2008) 221–260.
[37] R. Wilke, J. Sesterhenn, Statistics of fully turbulent impinging jets, J. Fluid
[5] D. Lytle, B.W. Webb, Air jet impingement heat transfer at low nozzle-plate
Mech. 825 (2017) 795–824.
spacings, Int. J. Heat Mass Transf. 37 (12) (1994) 1687–1697.
[38] N. Uddin, S.O. Neumann, B. Weigand, LES simulations of an impinging jet: on
[6] R. Gardon, J.C. Akfirat, The role of turbulence in determining the heat-transfer
the origin of the second peak in the Nusselt number distribution, Int. J. Heat
characteristics of impinging jets, Int. J. Heat Mass Transf. 8 (10) (1965) 1261–
Mass Transf. 57 (1) (2013) 356–368.
1272.
[39] S. Kubacki, E. Dick, Hybrid RANS/LES of flow and heat transfer in round
[7] W. Rohlfs, H.D. Huastein, O. Garbrecht, R. Kneer, Insights into the local heat
impinging jets, Int. J. Heat Fluid Flow 32 (3) (2011) 631–651.
transfer of a submerged impinging jet: Influence of local flow acceleration and
[40] S. Alimohammadi, D.B. Murray, T. Persoons, Experimental validation of a
votex-wall interaction, Int. J. Heat Mass Transf. 55 (2012) 7728–7736.
computational fluid dynamics methodology for transitional flow heat transfer
[8] K. Kataoka, Impingement heat transfer augmentation due to large scale eddies,
characteristics of a steady impinging jet, J. Heat Transfer 136 (9) (2014)
in: Ninth International Heat Transfer Conference, 1990, pp. 255–273.
091703.
[9] K. Choo, B.K. Friedrich, A.W. Glaspell, K.A. Schilling, The influence of nozzle-to-
[41] Y.B. Suzen, P.G. Huang, L.S. Hultgren, D.E. Ashpis, Predictions of separated and
plate spacing on heat transfer and fluid flow of submerged jet impingement,
transitional boundary layers under low-pressure turbine airfoil conditions
Int. J. Heat Mass Transf. 97 (2016) 66–69.
using an intermittency transport equation, AIAA J. (2001) 1–21.
[10] D.W. Colucci, R. Viskanta, Effect of nozzle geometry on local convective heat
[42] OpenFOAM, 2017. <https://openfoam.org/>.
transfer to a confined impinging air jet, Exp. Therm. Fluid Sci. 13 (1) (1996)
[43] D.C. Wilcox, Formulation of the k-x turbulence model revisited, AIAA J. 46 (11)
71–80.
(2008) 2823–2838.
[11] T.J. Craft, L.J.W. Graham, B.E. Launder, Impinging jet studies for turbulence
[44] D.A. Wilcox, Turbulence Modeling for CFD, DCW Industries, La Canada,
model assessment—II. An examination of the performance of four turbulence
California, 2000, pp. 146–147.
models, Int. J. Heat Mass Transf. 36 (10) (1993) 2685–2697.
[45] W. Rodi, G. Scheuerer, Scrutinizing the k-e Turbulent model under adverse
[12] S. Ashforth-Frost, K. Jambunathanet, Numerical prediction of semi-confined jet
pressure gradient conditions, J. Fluids Eng. 108 (2) (1984) 174–179.
impingement and comparison with experimental data, Int. J. Num. Meth.
[46] D.A. Johnson, L.S. King, A mathematically simple turbulence closure model for
Fluids 23 (1996) 295–306.
attached and separated turbulent boundary layers, AIAA J. 23 (11) (1985)
[13] M.A.R. Sharif, K.K. Mothe, Evaluation of turbulence models in the prediction of
1684–1692.
heat transfer due to slot jet impingement on plane and concave surfaces, Num.
[47] J. Zhe, V. Modi, Near wall measurements for a turbulent impinging slot jet, J.
Heat Transf. Part B: Fundam. 55 (4) (2009) 273–294.
Fluids Eng. 123 (1) (2000) 112–120.
[14] H. Huang, T. Sun, G. Zhang, L. Sun, Z. Zong, Modeling and computation of
[48] S. Ashforth-Frost, K. Jambunathan, C.F. Whitney, Velocity and turbulence
turbulent slot jet impingement heat transfer using RANS method with special
characteristics of a semiconfined orthogonally impinging slot jet, Exp. Therm.
emphasis on the developed SST turbulence model, Int. J. Heat Mass Transf. 126
Fluid Sci. 14 (1997) 60–67.
(2018) 589–602.
[49] T.H. Park, H.G. Choi, J.Y. Yoo, S.J. Kim, Streamline upwind numerical simulation
[15] J. Wienand, A. Riedelsheimer, B. Weigand, Numerical study of a turbulent
of two-dimensional confined impinging slot jets, international, J. Heat Mass
impinging jet for different jet-to-plate distances using two-equation
Transfer 46 (2002) 252–262.
turbulence models, Eur. J. Mech. B. Fluids 61 (2017) 210–217.
712 H. Huang et al. / International Journal of Heat and Mass Transfer 139 (2019) 700–712

[50] A.M. Achari, M.K. Das, Application of various RANS based models towards [52] T. Dairay, V. Fortuné, E. Lamballais, L.E. Brizzi, Direct numerical simulation of a
predicting turbulent slot jet impingement, Int. J. Therm. Sci. 98 (2015) 332– turbulent jet impinging on a heated wall, J. Fluid Mech. 764 (2015) 362–394.
351. [53] F.F. Cadek, A Fundamental Investigation of Jet Impingement Heat Transfer,
[51] S.B. Pope, An explanation of the turbulent round-jet/plane-jet anomaly, AIAA J. University of Cincinnati, Ohio, 1968.
16 (1978) 279–281.

You might also like