Yang, Wei, Wang - 2011 - Numerical Study of Turbulent Slot Jet Impingement Cooling On A Semi-Circular Concave Surface
Yang, Wei, Wang - 2011 - Numerical Study of Turbulent Slot Jet Impingement Cooling On A Semi-Circular Concave Surface
Yang, Wei, Wang - 2011 - Numerical Study of Turbulent Slot Jet Impingement Cooling On A Semi-Circular Concave Surface
a r t i c l e i n f o a b s t r a c t
Article history: An investigation of the flow field and heat transfer characteristics of a slot turbulent jet impinging on a
Received 12 February 2010 semi-circular concave surface with uniform heat flux has been carried out numerically in this study. The
Received in revised form 14 July 2010 turbulent governing equations are solved by a control-volume-based finite-difference method with a
Accepted 14 July 2010
power-law scheme and the well-known k–e turbulence model and its associate wall function to describe
Available online 11 October 2010
the turbulent structure. In addition, a body-fitted curvilinear coordinate system is employed to transform
the physical domain into a computational domain.
Keywords:
Numerical computations have been conducted with variations of jet exit Reynolds number Re2B
Concave surface
Slot jet
(5920 6 Re2B 6 23,700), dimensionless jet-to-surface distance H/B (0.5 6 H/B 6 12), dimensionless jet
Impingement cooling width B/D (0.033 6 B/D 6 0.05) and the heat flux q00 (1663 W/m2 6 q00 6 5663 W/m2). The theoretical
Turbulent heat transfer model developed is validated by comparing the numerical predictions with available experimental data
Numerical simulation in the literature. The variations of local Nusselt numbers along the semi-circular concave surface decrease
monotonically from its maximum value at the stagnation point. The numerical results show that the local
Nusselt numbers are reasonably predicted with a maximum discrepancy within 15%. As the Reynolds
number fixes, the effect of the impingement distance (H/B) on the average Nusselt (Nuavg) is not signif-
icant except at low H/B = 0.5. This study provides fundamental insight into turbulent slot jet impinge-
ment cooling on the semi-circular concave surface.
Ó 2010 Elsevier Ltd. All rights reserved.
0017-9310/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2010.09.021
Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489 483
Nomenclature
ing jets usually focused on the prediction of heat transfer rates significant impact both on temperature distribution and the aver-
with one or two turbulent models compared. The flow and heat age Nusselt number.
transfer experimental characteristics of a turbulent air jet imping- Numerical investigations of airflow and temperature field in a
ing on a semi-concave surface was presented by Yang et al. [9]. The room with a convective heat source were described by Lu et al.
experimental study has been carried out for jet impingement cool- [18]. They concluded that the thermal wall jet created by the heat
ing on a semi-circular concave surface when a jet flows were source greatly influenced airflow patterns and temperature fields
ejected from round shaped nozzle, rectangular shaped and 2D con- in a room. Numerical predictions of turbulent plane jets discharged
toured nozzle. The flow field induced by a single circular jet normal to a weak or moderate cross stream were presented by
exhausting perpendicularly from a flat plate into a cross flow has Kalita et al. [19]. The predictions were presented to illustrate the
been investigated numerically by Chochua et al. [10]. An evalua- flow pattern involved and to assess the performance of the stan-
tion of the different convection schemes revealed that the sec- dard k–e model by comparison with available experimental data
ond-order upwind scheme did a noticeably better job than the for three different jets to cross stream velocity ratios and the agree-
first-order scheme to predict the velocity profile at the jet exit ment was found to be satisfactory. A four-equation model was pro-
while predicting less mixing than the experimental measurement posed by Mashayek and Taulbee [20] for prediction of dilute
during the jet and free stream interaction. Rhee and Sung [11] ap- turbulent gas–solid flows where the ratio of the particle and the
plied the unsteady k–e –fu model and the diffusivity tensor heat gas densities were large. These models were used to propose four
transfer model to predict the enhancement of heat transfer in a lo- transport equations for the turbulent kinetic energy of the carrier
cally forced turbulent separated and reattaching flow over a back- phase and its rate of dissipation, the turbulence kinetic energy of
ward-facing step. The validation of the diffusivity tensor heat the dispersed phase, and the velocity covariance of the two phases.
transfer model was made by showing the St predictions with the The flow and heat transfer characteristics of a turbulent submerged
relevant experiment. Furthermore, high and low Reynolds stan- circular air jet impinging on a horizontal flat surface was presented
dard k–e models were extensively checked by Roy [12]. by Siba et al. [21]. An investigation into the predictive performance
Chattopadhyay and Saha [13] investigated turbulent flow field of linear and nonlinear eddy-viscosity turbulence models for a con-
and heat transfer from an array of impinging horizontal knife jets fined swirling coaxial jet was presented by Yang and Ma [22]. The
on a moving surface by using large eddy simulation (LES). It has deficiency in predicting scalar field was perhaps due to the
been indicated that increasing surface motion reduced heat trans- assumption of constant turbulent Prandtl number and unsteady
fer for both types of jets. Abdon and Sunden [14] investigated a sin- large-scale motions. Brescianini and Delichatsios [23] tested four
gle round unconfined impinging air jet under different flow and high-Re k–e turbulence models to assess their suitability in turbu-
geometrical conditions, to assess the performance of linear and lent buoyant jets and plumes. Jet impingement onto a hole with
nonlinear two-equation turbulence models. The constitutive rela- constant wall temperature was considered by Yilbas et al. [24].
tionship of the nonlinear models was shown to be dominated by The flow and temperature fields were simulated for a constant
the linear part for the problem. Concerning the second-order mod- standard-off distance, four holes wall temperatures, and two gas
els, some modeling have been made by Shih et al. [15] and showed jet velocities. It was found that flow impinging onto the hole
that second-order models only gave slightly better results than the spilled from the edges of the hole despite using the same hole
eddy viscosity models. Transient conjugate heat transfer during the and nozzle diameters.
impingement of a free jet of high Prandtl number fluid on a solid An inverse methodology was used to determine the turbulent
disk of finite thickness was considered by Rahman et al. [16]. Com- component of the heat transfer coefficient in the stagnation region
puted results included the velocity, temperature, and pressure dis- and wall-jet region. The effect of nozzle-plate spacing in plane
tribution in the fluid and the local and average heat transfer impinging jets using the direct numerical simulation was covered
coefficients at the solid–fluid interface. The conjugate heat transfer in Hattori and Nagano [25]. It was found that the second peak of
from discrete heat source to a two-dimensional jet of a high Pra- the local heat transfer rate (Nusselt number) in the wall jet devel-
ndtl number fluid discharging from a slot nozzle was performed oping region appeared in the lower nozzle-plate spacing case. Xu
by Bula and Rahman [17]. It was found that in addition to jet Rey- and Niu [26] investigated the influences of inlet boundary condi-
nolds number, plate thickness and its thermal conductivity had a tions on the pre-combustion chamber with internal flow patterns,
484 Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489
and validated by experimental data. For the swirling air flow, inlet concluded that local Nusselt number decreased with increasing
boundary condition specifications, two methods are compared inlet turbulence intensity in a wall jet region when standard k–
employing the standard k–e turbulence model. The simulation e model was used. But there is almost no change in the Nusselt
indicated that proper inlet conditions were essential in the case number distribution for this region when RNG k–e turbulence
of a swirling inlet, although there remained some discrepancies be- model was used. Jaramillo et al. [38] studied the different imping-
tween the simulation results and the experimental data. The de- ing jet configurations by means of time-averaged Navier–Stokes
tailed flow field characteristics of a turbulent air slot jet simulations. Comparisons were performed in terms of mean and
impinging on a semicircular convex surface were investigated fluctuating velocities and local Nusselt number. The performance
numerically by Yang and Hwang [27]. This study was useful in of several turbulence models in prediction of convective heat
helping researchers to understand the flow field characteristics in transfer due to slot jet impingement onto flat and concave cylin-
stagnation regions of impingement surfaces. A new algebraic tur- drical surface was evaluated by Sharif and Mothe [39]. It was
bulent mass flux model has been proposed and formulated by found that when the impingement surface was outside the poten-
Wei et al. [28]. The turbulent model mixing between the swirling tial core of the jet, most of the turbulence models predict reason-
air flow and the helium/air jet in a combustor has been simulated ably accurate thermal data. When the impingement surface was
numerically. The gas axial and tangential velocities, as well as the within the potential core of the jet, the turbulence models grossly
gas axial and tangential fluctuating velocities were predicted by over-predicted the Nusselt number in the impingement region,
the new ASM/k–e model, and the results are agreed with the mea- but in the wall jet region the Nusselt number prediction was
surements. In year 2004, the development of a new subgrid-based fairly accurate. Choo et al. [40] investigated experimentally the
wall function has been presented by Craft et al. [29], the boundary- heat transfer characteristics of a micro-scale slot jet impinging
layer-type transport equations were solved locally across an on a heated flat plate. Based on the experimental results, correla-
embedded grid within the near-wall cells. It has been presented tions on the stagnant and average Nusselt numbers for micro-
that the new wall function can achieve excellent alignment with scale impinging slot jet were developed as a function of the Rey-
the results of using low-Re models, while requiring only a modest nolds number and nozzle-to-plate spacing. The thermal charac-
increased in computing time compared with a standard wall teristics of confined and unconfined impinging jets are
function. compared by Choo and Kim [41]. The results showed that the
Three-dimensional numerical simulations of fluid flow and thermal performance of the confined jet was similar to that of
heat transfer characteristics for an inclined jet with cross-flow the unconfined jet under a fixed pumping power condition, while
impinging on a heating plate was presented by Yang and Wang the thermal performance of the confined jet was 20–30% lower
[30]. The generation of a pair of counter-rotating longitudinal vor- than that of the unconfined jet under a fixed flow rate condition.
tices was clearly observed from the computations. The flow and Generalized correlations for the stagnation and average Nusselt
heat transfer in two planar impinging jets using large-eddy sim- numbers for both confined and unconfined impinging jets were
ulation and experiments were demonstrated by Akiyama et al. presented as a function of the dimensionless pumping power
[31]. For both the unforced and forced flows the large-eddy sim- and the Prandtl number. Numerical investigation of convective
ulation predictions have shown that the jet develops stream wise heat transfer process from concave cylindrical surface due to
vorticity, and the turbulent inflow condition provided distur- slot-jet impingement was performed by Sharif and Mothe [42].
bances that the jets can be efficiently amplified. Chen et al. [32] The results indicated that while the jet-exit Reynolds number
provided detailed theoretical solutions on laminar flow for free- and the surface curvature have a significant effect on the heat
surface slot jet impinging onto horizontal surfaces under arbi- transfer process, it was relatively insensitive to the jet-to-target
trary-heat-flux conditions. The thermal and hydraulic boundary spacing.
layers of laminar flow were divided into four regions of flow Over the last few years, CFD using Reynolds-averaged Navier–
along heat transfer surfaces including a stagnation zone and three Stokes equations, coupled with turbulence modeling has become
wall jet zones. Numerical predictions of heat transfer coefficients a standard practical simulation tool for the design and analysis of
under jet impingement from an array of nozzles have been made engineering systems. Numerical modeling of jet impingement flow
by Salamah and Kaminski [33]. A study of the capabilities using and heat transfer is still challenging because of the complex inter-
several turbulence models to capture the physical characteristics action of the flow entrainment, vortex formation, and flow separa-
of a single round nozzle and a single slot nozzle jets was con- tion along the surface. Therefore, the choice of the turbulence
ducted by Coussirat et al. [34]. The results showed that velocity model is very crucial in the numerical analysis of the impingement
fluctuations were not adequately predicted at the stagnation heat transfer process. According to the results of Isman et al. [37],
zone, and their values depend on which turbulence model was they concluded that while the accuracy of some of the models is
used. Yang and Tsai [35] presented numerical study of transient better in the stagnation region, the standard and the RNG k–e mod-
conjugate heat transfer in a high turbulence air jet impinging over els are better over the entire impingement surface. Based on the
a flat circular disk. The turbulent governing equations are re- above mentioned results, the standard k–e model is adopted in this
solved by the control-volume based finite-difference method with study. The objective of the present study is to provide physical in-
a power-low scheme, and the low-Re k–x turbulence model to sight into heat transfer effects and facilitate the validation of
describe the turbulent structure. It was shown that with the numerical heat transfer models.
inclusion of the conjugate heat transfer, the numerical results
compared favorably with the available experimental results. Hof- 2. Mathematical formulation
mann et al. [36] evaluated the performance of 13 widely spread
Reynolds-averaged Navier–Stokes (RANS) turbulence models in 2.1. Governing equations
the prediction of round jet impingement heat transfer from an
isothermal flat surface. They found that the SST k–x model with The schematic diagram of the geometry and the computational
transitional flow option predicted correctly the laminar-turbulent domain is shown in Fig. 1 and the dimensions of the computational
transition occurring at small nozzle-to-plate distance. Isman et al. domain are based on the experimental study by Yang et al. [9]. The
[37] employed the finite-element methods to evaluate the perfor- turbulent two-dimensional Navier–Stokes and energy equations
mance of five two-equation turbulence models in the prediction are solved numerically by a finite-difference scheme and a conti-
of slot jet impingement cooling of constant heat flux surface. They nuity equation to simulate the thermal and turbulent flow fields.
Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489 485
mation. This yields the relationships where J is the determinant of matrix form. The solution is considered to be converged when
the Jacobian matrix of the inverse transformation the normalized residual of the algebraic equation is less than a pre-
scribed value of 104.
@ 1 @ @ @ 1 @ @
¼ yg yn ¼ xn xg ð10Þ
@x J @n @ g @y J @g @n
4. Results and discussion
xn yn
where J ¼ ¼ xn y xg y .
xg yg g n
The numerical algorithm and validation are evaluated in this
Therefore, the transformed equation can be expressed as: study by comparing numerical model predictions with available
experimental data in the literature. The numerical simulation
1 @ @ 1 @ C/ @/ @/
ðqU/Þ þ ðqV/Þ ¼ ðG1 G2 Þ has been conducted with variable parameters including Reynolds
J @n @g J @n J @n @g
number Re2B (Re2B = 5920–23,700), dimensionless jet-to-surface
1 @ C/ @/ @/
þ ðG3 G2 Þ þ S/ ðn; gÞ distance H/B (H/B = 0.5–12), dimensionless jet width B/D (B/D =
J @g J @n @n
0.033–0.05) and two different heat flux q00 (q00 = 1663–5663 W/m2).
ð11Þ
200
180
S/B
}
309x62
309x62
309x62
160 249x60standard k − ε
249x60
249x60
140 189x58
189x58
189x58
1 249x60
e x pt, of Y a ng e t a l. [ 9]
249x60 standard k − ω
120 expt. of Yang et al.[9]
Nu2B
100
80
60
40
20
0
0 3 6 9 12 15 18
S/B
Fig. 3. Effect of grid refinement on local Nusselt number (H/B = 12, B/D = 0.033,
Fig. 2. Typical grid used in the computations. Re = 23,700).
Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489 487
numerical accuracy and are essentially grid independent. In the lit- the downstream with the increase of axial distance H/B. Fig. 5 pre-
erature, reported experimental data has been validated, and a good sents the turbulent kinetic energy counter for different H/B at B/
agreement has been found between the model predictions and D = 0.033 and Re = 23,700. The stagnation point turbulent kinetic
measurements. It can be seen that the standard k–e model and energy is the highest at the large spacing, H/B = 12 and decreases
standard k–x model grossly under-predict the local Nusselt num- as the distance between the nozzle exit and target surface (H/B)
ber in the impingement region at H/B = 12, but the standard k–e is decreased. The variations of the local Nusselt number along
model predict the Nusselt number distribution better than the k– the heated concave surface are of more interest from the perspec-
x in the wall jet region for the semi-circular concave surface tive of heat transfer effects. The local Nusselt number variations
impingement case. along the concave surface are shown in Fig. 6 for H/B = 4, B/
D = 0.033, and q00 = 5663 W/m2. It shows a significant effect of the
4.2. Flow characteristics and heat transfer jet-exit Reynolds number on the local Nusselt number variation.
This may be contributed to the fact that the convection strength
The velocity vectors of a turbulent slot jet impinging on a semi- and turbulence level increase with the increasing Reynolds number
circular concave surface for Re = 23,700, B/D = 0.033 and H/B = 0.5– which enhances heat transfer. The local Nusselt number is highest
12 are shown in Fig. 4. It is shown that the shear flow region devel- in the impingement region at any Reynolds number. It can be seen
oped from the edge of the nozzle which expands due to the mixing that the variants of the k–e model predict reasonably accurate ther-
of jet flow and ambient air, and at H/B = 12 the effect of entrain- mal data compared to the experiment of Yang et al. [9]. The general
ment penetrates to the center line. The ambient flows are en- trend is such that the local number is somewhat over-predicted in
trained to the potential core region when the jet flow develops to the impingement region and under-predicted in the wall jet region
36.97 m / s 36.97 m / s
(a) (b)
36.97 m / s 36.97 m / s
(c) (d)
Fig. 4. Velocity vector for Re = 23,700, B/D = 0.033: (a) H/B = 0.5; (b) H/B = 4; (c) H/B = 8 and (d) H/B = 12.
488 Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489
200
180
2 2
Re
Re
k (m / s ) 160 23700
17800
33.9401 140 11800
31.6774
5920
29.4147 120 23700
27.152
Nu2B
expt. of Yang et al. [9]
24.8894 100
22.6267
20.364 80
18.1014
15.8387 60
13.576
11.3134 40
9.05068
6.78801 20
4.52534
2.26267 0
0 3 6 9 12 15 18
S/B
Fig. 6. Effect of Reynolds number on the local Nusselt number distribution for H/
B = 4, B/D = 0.033 and q00 ¼ 5663 W=m2 .
(a)
200
k (m 2 / s 2 ) 180
B/D=0.050
B/D=0.040 B/D=0.050
51.4341 160
B/D=0.033
48.0051
140 expt. of Yang et al.[9]
44.5762
41.1472
37.7183 120
Nu2B
34.2894
30.8604 100
27.4315
24.0026 80
20.5736
17.1447 60
13.7157
10.2868 40
6.85787
3.42894 20
0
0 3 6 9 12 15 18
S/B
Fig. 7. Effect of slot jet width on the local Nusselt number distribution for H/B = 4,
Re = 23,700 and q00 ¼ 5663 W=m2 .
(b)
Fig. 5. Turbulent kinetic energy for B / D=0.033, Re=23700 (a) H / B=4 and (b) H /
B=12. variations along the impingement surface with the experimental
results of Yang et al. [9] at the same B/D = 0.033. The turbulence
models are not able to predict the laminar-to-turbulent transition
compared to the experiment. There are no predicted secondary and hence the secondary peak of the local Nusselt number is not
maximum in the variation local Nusselt number even though there captured by any of the turbulence models used. Fig. 8 shows the
are experimental evidences when the impingement surface is comparisons of average Nusselt number distribution with Yang
within the potential core of the jet. To figure out the effect of slot et al. [9] for B/D = 0.033 and q00 = 5663 W/m2. For all cases consid-
jet width on the heat transfer along the concave surface, the local ered, the highest heat transfer rates occur at the smallest spacing,
Nusselt number distributions are plotted in Fig. 7. It is clearly H/B = 0.5. It seems to be due to the effect of fluid acceleration be-
shown that the highest Nusselt number occurs at the larger value tween the nozzle exit and the impingement surface. The boundary
of B/D = 0.05. As the B/D decreases, the local Nusselt number de- layer thickness in the stagnation region for a slot jet impingement
creases as well. The noticeable feature is that the surface curvature varies depending on the nozzle-to-surface distances and nozzle
effect on the local Nusselt number is significant at higher surface exit velocities. Smaller spacing can cause larger axial velocity gra-
curvature (B/D = 0.05). It indicates that present standard k–e turbu- dient and also larger temperature gradients on the concave surface,
lence model predict reasonably accurate local Nusselt number which enhances the heat transfer rates.
Y.-T. Yang et al. / International Journal of Heat and Mass Transfer 54 (2011) 482–489 489