Вы находитесь на странице: 1из 16

Applied Thermal Engineering 90 (2015) 648e663

Contents lists available at ScienceDirect

Applied Thermal Engineering


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

Research paper

Numerical investigation of two-phase ow characteristics of


subcooled boiling in IC engine cooling passages using a new 3D
two-uid model
Shiyang Hua*, Ronghua Huang, Pei Zhou
State Key Laboratory of Coal Combustion, School of Energy and Power Engineering, Huazhong University of Science and Technology, Wuhan, 430074, China

h i g h l i g h t s

 A modied bubble departure diameter correlation was developed.


 Two sets of experimental data were used to examine present model.
 The modied boiling model was adopted in a practical engine cooling passage.
 A great improvement in temperature eld was obtained with two-uid boiling model.
 The detailed information on vapor phase was obtained in practical engine cooling passage.

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

Article history: Modern engine cooling systems use subcooled boiling as an effective heat transfer form. The signicantly
Received 2 March 2015 higher heat transfer rate achievable with boiling can keep the temperature in key thermal areas within
Accepted 15 July 2015 an acceptable level. In this study, a three-dimensional two-uid model coupled with RPI wall boiling
Available online 26 July 2015
model was adopted to simulate the local two-phase characteristics of boiling at different passages under
low-pressure condition. Moreover, a modied bubble departure diameter correlation was developed and
Keywords:
implemented in computational uid dynamics code with user-dened function, and the correlation was
Subcooled boiling
veried by two sets of experimental data under different ow conditions. Results showed an improved
Two-uid boiling model
Void fraction
agreement with the measured local void fractions. Finally, the veried model was applied to a practical
Engine cooling passage cooling passage for a water-cooled diesel engine. The simulation results showed that present two-uid
Computational uid dynamic model could get an accurate temperature eld for cylinder head. The distribution of vapor phase in
Bubble departure diameter cooling passage was obtained, and it could be used as a reference to evaluate the reliability of local
boiling. Moreover, due to the function of interphase forces, the majority of bubbles were quickly removed
from near wall and the void fraction was below 0.149.
2015 Elsevier Ltd. All rights reserved.

1. Introduction due to its efcient heat transfer characteristics, using local sub-
cooled boiling to achieve precision cooling, thereby decreasing the
As the modern engine tends to high combustion pressure and stress level of cylinder head under high load conditions, and
high power density output, the new generation of engine cooling increasing heat exchange capacity of radiator and cooling water
system presents much more challenges such as less volume, lower chamber. Due to the complexity of physical phenomena, many
dissipation and higher efciency. On the contrary, the traditional experimental studies [7e9] have been carried out, while the ow
engine cooling systems designed according to convectional heat boiling in the regular passage still could not be well explained and
transfer fail to meet the cooling requirements at high load condi- studied as conventional heat conduction and single-phase con-
tions [1e3]. Latest studies [4e6] focused on boiling heat transfer vection. Subcooled boiling usually occurs when there is local
boiling at the wall surface superheated by the high heat uxes, even
though the bulk ow has not reached the saturated temperature.
* Corresponding author. When boiling develops beyond nucleate regime, overheating oc-
E-mail addresses: huasy576@163.com (S. Hua), rhhuang@hust.edu.cn (R. Huang). curs and may damage the equipment. Thus, in order to achieve

http://dx.doi.org/10.1016/j.applthermaleng.2015.07.037
1359-4311/ 2015 Elsevier Ltd. All rights reserved.
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 649

controlled boiling over a wide range of operation conditions, the governing equations. Mass, momentum and energy transfer be-
investigation of two-phase ow characteristics of subcooled boiling tween the two phases are explicitly modeled. The interactions
ow in cooling passages is an important step of engine design and between phases are calculated based on the locally estimated
development. bubble size and amount [17]. In 1991, Kurul and Podowski [18]
In recent years, the computational uid dynamics (CFD) simu- proposed a new wall boiling model considering the bubble depar-
lation incorporating the modelling of boiling has become an ture diameter, bubble detachment frequency, wall nucleation site
effective approach for such analysis. Moreover, CFD not only reveals density, bubble diameter, and other physical parameters with clear
the presence of possible hot spots, but also provides useful infor- physical signicance. According to this model, the liquid and vapor
mation to address the hydrodynamic issues specially associated phases were solved separately using two sets of governing equa-
with boiling. In particular, CFD reveals the two-phase ow in- tions. Following studies such as Anglart [19] implemented
stabilities and the optimum match of pump, duct, and cooling Podowski's two-uid model into a CFD code to predict the distri-
jacket under boiling ow conditions [3]. Three different approaches bution of void fraction in two-phase bubbly ows in fuel rod
have been proposed to model the two-phase boiling CFD model in bundles. Yao [20] applied the two-uid model to predict the dis-
engineering application. The single-phase boiling model adopts tribution of two-phase parameters in Freon R12 boiling ow,
empirical correlations to calculate the wall temperature and inte- observing that the turbulent dispersion force plays an important
grate heat ux under boiling condition by introducing revised co- role in determining the void fraction distribution and liquid tem-
efcient to enhance heat transfer. The main drawback of single- perature distribution. Ustinenko [21] adopted the overall validation
phase approach is that energy addition translates directly into a strategy to analyze the two-phase ow and heat transfer phe-
rise in temperature rather than phase change. Thus, the calculated nomena in a boiling water reactor fuel bundle under high-pressure
density, temperature, and ow elds will be not complete. Rob- condition. However, Zeitoun [22] Later, Tu [23,24] and Koncar [25]
inson [10] used the empirical correlations to calculate the wall- found that the detachment bubble size at low pressure was
temperature/heat-ux under the boiling condition and to eval- different from the size at high pressure, and introduced some im-
uate the convective and boiling heat transfer in irregular geome- provements to apply the same model to low pressure. Despite the
tries. Based on a division description method, Li [11] proposed a fact that the accuracy of the prediction relies heavily on the quality
modied model with adjustable parameters to be directly applied of sub-models, and many of these models have not yet deeply
to CFD analyses inside internal combustion (IC) engine cooling investigated, the two-uid method is still the most promising.
galleries under a wide range of operating conditions. Nevertheless, The above-mentioned researches show that the two-uid
these results only reect the relationship between wall- model is a detailed method for simulating two-phase ow and
temperature and heat-ux, and cannot describe the physical pro- offers great potential for engineering applications with the rapid
cess of phase change. The second approach, the homogenous ow development of computer hardware. In this paper, the RPI wall
model, represents both the liquid and vapor phases based on the boiling model is coupled in the three-dimensional two-uid model,
assumption that vapor bubbles are small and perfectly mixed with and the local two-phase characteristics of subcooled boiling in
the liquid phase. The concentration of vapor phase is described by different passages will be numerically investigated under low
an additional variableevoid fraction. Thus, the governing equations pressure condition. In addition, many existing models are not
are still for the single phase, and an additional equation describes suitable to low pressures because they have been developed for
void fraction. In 2004, Bo [12] developed a single-phase boiling high-pressure state. Basing on the Unal's model [36] and the Tol-
model for engine cooling water jacket by adding a void fraction ubinski and Kostanchukis' correlation [35], for a better description
equation. The predicted wall heat ux and vapor concentration of the actual behavior at low pressure, this paper proposes a
distributions were in satisfactory agreement with the experimental modied bubble departure diameter model. The new model is
measurements. Then, Li [13] assumed that the vapor and liquid implemented in ANSYS-CFX CFD code through a user-dened
were homogeneously mixed and regarded the boiling liquid eld as FORTRAN program. In addition, two sets of experimental data
a single-phase ow. Thus, the source term of interphase mass veried the modied model under different mass ow rates,
transfer due to evaporation and condensation was calculated by pressures, subcooling, heat ux, section shapes, and ow di-
introducing the driving forceetime scale. In 2009, a variant of the rections. Finally, the veried model simulated the two-phase
Rensselaer Polytechnic Institute (RPI) mechanistic nucleate boiling boiling ow characteristics in a practical cooling passage for a
model provided closures for the wall thermal conditions and mass four-stoke water-cooled diesel engine.
transfer rates due to phase change in the bulk ow; Przulj and Shala
[14] proposed and applied to engine coolant ows the homogenous 2. Two-uid boiling model
ow-boiling model. In 2011, Mohammadi [15] established a
computational model for boiling heat transfer based on homoge- A multidimensional two-uid Eulerian approach describes
nous ow to analyze a cooling system for diesel engines using the subcooled boiling ow in the cooling passages of an IC engine. The
CFD code OpenFOAM (Open source Field Operation And Manipu- two-uid framework is based on space and time averaging and on
lation). Similar studies based on homogenous ow boiling model the interpenetrating continua approach. According to these as-
were conducted by Shala [16] and Hemant [17]. Because of its sumptions, balance equations for mass, momentum and energy
simplicity and high computational efciency, this method has been have been written separately for each phase; in addition, the liquid
widely used in CFD engine coolant boiling analysis. Although the is treated as the continuous phase while the vapor bubbles are the
effect of density change due to the uid phase change is fully dispersed phase. Then, the closure of two-uid equations is
considered, the detailed interfacial dynamics between the two determined by introducing the interphase transfer. Since the two-
phases are not commonly modeled. The source terms in the void uid model is deeply described in literatures [26,27] this paper
fraction equation model the transfer between the two phases as does not discuss the conservation equations.
driven by the pressure difference. Interfacial dynamics are not
considered. Therefore, the calculated void fraction, temperature, 2.1. Turbulence model
and ow eld distribution will be unreasonable especially near the
wall. The third model is a two-uid boiling model, in which the The ow regime through the engine cooling passage is mainly
liquid and vapor phases are solved separately using two sets of turbulent. However, because of the low-density, the motion of the
650 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

dispersed vapor phase is commonly assumed to be dependent on 2.3. Wall boiling model
the continuous liquid phase [25]. The present work applies the
shear stress transport (SST) turbulence model to the continuous Simulation of two-phase ow requires wall heat ux modeling.
liquid phase, and a disperse phase zero-equation model to the va- This paper considers the mechanical model on subcooled boiling
por phase [28]. Furthermore, due to the large bubbles and uctu- ow of Kurul and Podowski [18] including it in the ANSYS-CFX
ating wakes, the boiling ow generates an additional liquid code. The total wall heat ux for nucleate boiling ow is calculated:
turbulence viscosity term mbl , modeled using Sato's eddy viscosity
model [29]. Thus, the viscosity is expressed as: qw qcon qe qque
  p 
hcon Acon Tw  Tl;y const fNw d3w rg llg
meff ml mturb mbl (1) 6
l l s
tw kl rl cpl  
where meff is the effective viscosity in momentum and energy 2Ab f Tw  Tl;y const (7)
l p
equations, ml is molecular viscosity, mturb is the shear-induced tur-
l
! !
bulence viscosity, mbl is dened as Cmb rl adb  u g  u l , and the co-
where T l,yconst is the local liquid temperature calculated through the
efcient Cmb is set to the default vale 0.6. The near wall region is
logarithmic form of the wall function at a xed value of non-
assumed to follow the logarithmic law in single-phase state.
dimensional distance from the wall yconst 250 [32]. To close
the equations for the above wall heat ux partitioning algorithm, a
2.2. Interphase momentum transfer series of sub-models have to be provided, such as the wall nucle-
ation site density Nw, bubble department diameter dw, bubble area
.
The interfacial drag force F drag is modeled according to the inuence fraction Ab, bubble detachment frequency f, and bubble
IshiieZuber [30] and is calculated as: waiting time tw.
The wall nucleation site density is calculated using Egorov and
. 3CD  ! ! Men's formula [36]:
F drag ar ug  ul  u g  u l (2)
4db l  . p
Nw Nref DTsat DTref (8)
where db denotes the average diameter of bubble, mentioned in the
following section. The drag coefcient CD is ow regime dependent,
where Nref 0.8  9.922e5 m2, DTref 10 K, p 1.805, and
thus related to the Reynolds number of vapor Reb [28].
DTsat TwTsat is the wall superheat. The frequency of the bubble
The lift force on the liquid phase can be modeled as:
detachment from the nucleation site is calculated using the
! ! ! ! following empirical correlation [33]:
F lift CL arl u g  u l  V  u l (3)
s
 
4g rl  rg
Herein, CL is the lift force coefcient, calculated basing on f (9)
Tomiyama's model [31]. The Favre average of the interphase drag 3CD rl dw
model [32] describes the turbulent dispersion force:
The drag coefcient factor CD was assumed equal to unity. The
  waiting time between departures of consecutive bubbles is xed at
! 3C m Vrg Vrl
F td Ctd D l  (4) 80% of the bubble detachment period, thus waiting time is calcu-
4db st rg rl
lated as tw 0.8/f.
The area fraction inuenced by the nucleating bubbles is usually
here, st is the turbulent Schmidt number of liquid phase and it
expressed as [18]:
equals 0.9, CD is the momentum transfer coefcient for the inter-
phase drag force, Ctd denotes the turbulent dispersion coefcient  
! Jasub
and its default value is 1.0. The wall lubrication force has been Nw pD2w  80

modeled according to the Antal et al. model [25]: Ab min 1; K ; K 4:8e ;


4 (10)
! 2   rl Cpl DTsub
! u g  ul d ! Jasub
F wl arl max 0; C1 C2 b n w (5) rg llg
db yw
!
where DTsub TsatTl,y (const), Jasub is the subcooled Jacob number
where yw denotes the clearance with the nearest wall, and n w is
the unit normal pointing away from the wall vertical direction. according to Del Valle and Kenning [34].
According to Eq. (5), the wall lubrication force strongly depends on
the local bubble diameter db. In order to obtain a good agreement 2.4. Bubble departure diameter model
between the predicted and experimental void fraction, the coef-
cient values C1 and C2 are adjusted and set to 0.04 and 0.095, Previous section highlights that many sub-models are closely
respectively. Due to the lowest dispersed phase density than the with bubble department diameter, such as the bubble detachment
continuous phase density, the virtual mass force should be frequency, evaporation heat ux, bubble area inuence fraction,
considered. The virtual mass force is proportional to relative phasic and bubble waiting time. Moreover, the bubble department diam-
accelerations as follows eter is a crucial parameter in the wall heat ux partition model. In
 ! ! the original ANSYS-CFX code, dw is modeled as a function of liquid
! dg u g dl u l
F vm arl Cvm  (6) subcooling (Tolubinski and Kostanchuk [35]):
dt dt
" ! #
here Cvm is the virtual mass coefcient, and is set to 0.5. The terms DTsub
!
dwT min dref $exp  ; dmax (11)
dg u g
! DTref
dt
and dldtu l denote the phase material time derivative of the form.
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 651

The parameters of Eq. (11) are dimensional (dref 0.6 mm, " ! #
2:42  105 P 0:709 a DTsub
dmax 1.4 mm, DTsub 45 K [28]), obtained from high-pressure dwnew min Cw $ p $exp  ; dmax
bF DTref
water boiling experimental data. However, experimental data
collected by Unal [36], Zeitoun and Shoukri [37], Bartel [38] and (15)
Prodanovic et al. [39] showed that the bubble departure diameter
where
dw varies with pressure conditions. In later studies, the Unal's
model [36] was proposed to describe the variation of bubble de- r
Tw  Tsat kw rw cpw DTsub
parture size along the heated wall for the low-pressure system. a ;b 
;
Unal assumed that a bubble is subject to simultaneous microlayer 2llg rg p 2 1  rg rl
evaporation at the bottom and to condensation at the top. The
( ) (16)
ul =u0 0:47 ul  0:61m=s
bubble may slide or collapse on the heated surface during its life- F
time. After reaching the maximum size, the bubble departs from 1 ul < 0:61m=s
the wall and migrates into the subcooled liquid, where the bubble
can collapse. Pressure, material of solid wall, wall superheating, where the coefcient Cw is added to Eq. (15) to describe relatively
liquid subcooling and local velocity inuence the bubble size at large bubbles at low pressure conditions, and its value has been set
departure. The original Unal's relationship is calculated as: as 2.8 which is adjusted from some experimental cases. The upper
limit bubble departure diameter dmax is set to 5 mm. The present
2:42  105 P 0:709 a bubble departure diameter is related to pressure, wall material,
dwU p (12) wall superheat, local liquid subcooling, and two phase material.
bF
Furthermore, the remaining unknown in equations for bubble de-
where parture diameter Eq. (15) is the wall temperature Tw, that can be
calculated from the wall heat ux balance Eq. (7) with an iterative
 1=3 s procedure. Therefore, the computational convergence and appli-
qw  h1f $DTsup rw cpw kw
a q ; cation of RPI boiling model has been greatly improved basing on
2C 1=3 rg Hlg pkl =rl cpl kl rl cpl
the new bubble departure diameter model.
8 Except for Tolubinski's model in ANSYS-CFX, the other three
> DTsub
>
> 
eDTsub =31 DTsub  3K types of bubble departure diameter models (Unal's, Fritz's and
>
< 2 1  rg rl new-modied models), are compiled with an user dened
b FORTRAN program.
>
> DTsub
>
>
: 
 DTsub > 3K
2 1  rg rl
2.5. Interphase mass transfer
h . i3
Hlgml cpl 0:013Hlg Pr1:7  0:47 !
l Ul The phase change of subcooled boiling ow is described by the
C q

  ; F max ;1
s rl  rg g Uo bubble evaporation on the heated wall and the bubble condensa-
tion in the bulk ow. After the heat ux components are calculated,
(13) the evaporation rate at wall per unit volume Gw can be obtained
from evaporation heat ux qe:
where ul is the liquid velocity near the wall and u0 0.61 m/s. P is
 
the local pressure.
qe Ai fNw p6d3w rg llg A
In fact, the above semi-mechanistic model is only applicable to Gw     i (17)
the inlet section according to Unal's paper [36], which could not llg cpl DTsub Vi llg cpl DTsub Vi
directly be applied to predict the bubble departure diameter along
the whole heated section. On the other hand, qw is a xed value where Ai and Vi are the heat area and volume of ith near wall cell.
considered in some experimental channel and the inlet bubble After departure from the heated wall, the vapor bubbles may be
departure diameter can be directly calculated. On the contrary, the surrounded by subcooled liquid and condense. The mass transfer to
heat ux in the practical cooling passage is an unknown variable liquid phase from interface is calculated as
that must be calculated from uidesolid conjugate heat transfer,
.
resulting in a more complex and divergent iterative computation.
Ginter hl Alg Ts  Tl llg (18)
Fritz [40] introduced and presented a second departure diam-
eter model for low-pressure condition. The equation is expressed as
it follows: where hg and hl are the gas phase and liquid phase heat transfer
coefcients respectively. Ts is the interfacial temperature, which is
s assumed to be the same for both phases. Alg denotes the interfacial
s
dwF 0:2084   (14) area density, which can be expressed as Alg 6a/db. a and db are the
g rl  rg
void fraction and the average bubble diameter. The heat transfer
coefcient for the liquid phase across the interface can be dened
where 4 is the contact angle, taken as 80 [40]. as hl Nubkl/db. For a particle in a moving incompressible Newto-
In this work, one assumes that the bubble departure diameter nian uid, the bubble Nusselt number Nub can be calculated using
still obeys the Tolubinski and Kostanchuk's empirical correlation at the RanzeMarshall correlation [25]:
low pressure. The reference bubble departure diameter is calcu-
lated through Unal's correlation, which is no longer correlated with Nub 2 0:6Re0:5
1=3
b Prl (19)
wall heat ux. Reference subcooling is also used to evaluate the
liquid temperature inuence. Moreover, an additional restriction In Eq. (19), Reb is the bubble Reynolds number and Prl is the
must be applied to limit the excessive increase of bubble departure liquid Prandtl number.
diameter. Finally, the new modied sub-model for bubble depart- According to previous discussion, one can conclude that the
ment diameter is calculated as: bubble size determines the interfacial momentum transfer as well
652 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

as interfacial heat and mass transfer. In many studies, the re- accumulated time that the probe sensor was exposed to the vapor
searchers assumed constant bubble diameter in the ow eld due phase by the total sampling time. The total sampling time was set
to lack of experimental data about the size of bubble in the bulk to 60 s, to ensure the repeatability of void fraction generation. In
ow. In this paper the average bubble diameter is described by the addition, the uncertainty associated with local void fraction
Kurul-Podowski and Anglart-Nylund's approaches [18], thus the measurement was estimated to be 3%. The mass ow rate, tem-
bubble diameter is modeled as a linear function of local subcooling perature, and pressure were measured at the inlet. Wall heat ux
that shows the maximum value of db1 starting from the near-wall was calculated from the measured voltage drop across the heated
cell: tube, the current through it, and the heat surface area. The system
pressure was in the range 1e2 bar. Literature [41] describes in
db1  db0 d DT  db1 DTsub0 detail the experimental setup.
db DT b0 sub1 (20)
DTsub1  DTsub0 sub DTsub1  DTsub0 Prior to verify the new model, a grid-independence check has
been performed in order to reduce calculation errors and improve
where db0 and db1 are the minimum bubble diameter and the computational efciency. Fig. 2(a) shows the radial void fraction at
maximum bubble diameter at reference subcooling DTsub0 and test position when the size of radial grid is 15, 20, and 25, respec-
DTsub1 respectively. tively. Figure shows that the void fraction proles look pretty much
the same except small differences near the wall cell. Fig. 2(b) pre-
3. Analysis of results and discussion sents the variation of the radial void fraction along the circumfer-
ential direction on three different grids, and Fig. 2(c) presents the
3.1. Model validation variation of the radial void fraction along the heated duct direction
on the three different grids. The comparison of the results shows no
To validate and assess the accuracy of the modied wall boiling signicant difference between grids, thus the model can be
model, two sets of experimental data are considered. The data refer considered grid independent. Calculation adopted a 20  7  230
to an upward ow in a vertical annulus and a horizontal ow in a medium cell volume.
rectangular channel. Fig. 3 compares the measured and predicted values of void
fraction prole along the radial direction using four different bub-
3.1.1. Upward ow in a vertical annulus ble departure diameter sub-models. The gure shows that the
The rst group of experimental data was obtained in a vertical Tolubinski [35], Unal [36] and Fritz's [40] models consistently
annulus by Lee et al. [41]; Fig. 1 shows the simulation domains. A under-predict void fraction; in particular, the Tolubinski's correla-
heating inner tube was installed in the vertical annular test tion results are far below the experimental data because the cor-
channel. The tube was 1.67 m long and its outer diameter was relation arises from the application of model developed only for
19 mm. On the outer tube were mounted two stainless steel tubes high-pressure conditions. Results clearly show that the pre-
(37.5 mm inner diameter both), connected below the measuring dictions of the new-modied model greatly improved and agreed
plane by a transparent 50 mm long glass tube. The measuring closely with the measuring data.
plane was installed 1610 mm downstream of the beginning of To test whether the new-modied model ts more general
heated section. The local vapor phase parameters were measured conditions at low pressure, 18 experimental cases are simulated
by a two-conductivity probe, pre-calibrated through airewater under different wall heat uxes, mass ow rates, subcooling and
experiments in the transparent tube with two quick closing valves. system pressures, and the details are listed in Table 1. Fig. 4
The local void fraction was determined by dividing the compares the predicted void fractions obtained from the new
model using experimental data. Results show a good agreement
with the experimental data in the predicted void fraction dis-
tribution when inlet subcooling, mass ow and heat ux are
relatively low, such as cases 1e6. However, the discrepancy is
signicant with higher liquid subcooling and higher mass ow,
observing many deviations of predicted values from experi-
mental data at cases number 9, 12, 15, and 18. Figures show that
the thermal non-equilibrium of two phases in subcooled boiling
ow contains two distinguishable ow regions, corresponding to
the bubble boundary layer at the near wall and the subcooled
liquid core. In particular, the peak local void faction is predicted
near the heated surface, condition almost similar to Erfeng's
studies [42]. Commonly one thinks that the vapor bubbles at
near wall region are relatively large at low pressure, while the
corresponding wall lubrication force is large according to the
Antal's correlation [25]. Thus, the wall lubrication can push the
vapor bubbles to a relatively far region and impose the distri-
bution of void faction. The coefcient (C1, C2) of model in wall
lubrication force keeps unchanged, which may result in some
large deviations at certain locations. On the other hand, the main
reason to explain deviations from measurements is that
modeling of the average bubble diameter becomes more com-
plex for high heat ux conditions, because of bubble breakup
and coalescence. As reported by Roy et al. [43], in subcooling
boiling of R-113, the vapor bubble number increases with the
wall heat ux, and the bubble diameter increases due to the
Fig. 1. Simulation domains, Lee et al. [41] test section. bubble coalescence, as they move away from the surface, directly
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 653

Fig. 2. Grid inuence on void fraction.

affecting the local void faction near the bubble boundary layer. surface. Fig. 5 shows the experimental setup. Bubble behavior in
Although some researches consider homogeneous multiple size subcooled ow boiling has been acquired using a high-speed
group approach to describe the bubble breakup and coalescence camera. A three-dimensional matrix (X, Y, T: horizontal posi-
[42], it is difcult to adapt the complex model to engine cooling tion, vertical height and time, respectively) was generated to
passages because of the convergence of calculation and evaluate time-dependent scales of the boiling process. At
computational costs. different heights, each section also contains the information
about an average phase distribution resulting from the projec-
3.1.2. Horizontal rectangular channel ow tion of a two-dimensional boiling area onto a one dimensional
One of the unique characteristics of coolant subcooled boiling string. Then, XeT images could provide qualitative information
in the cylinder head regions of heavy-duty diesel engines is that about the decreasing void fraction with increasing height from
boiling generally occurs only on the cooling passage side facing the surface. Literature [44] describes in detail the experimental
the ame plate because of the one-sided heating condition [45]. setup and tests. Table 2 summarizes the conditions of simulated
Although there have been many investigations into subcooled cases. Besides, the computational mesh refers to previous grid
ow boiling with wall heating, most of the effort was focused on size in vertical annulus.
the uniformity of heat ux like nuclear reactor systems. The Fig. 6 shows a comparison between the experimental and the
present work adopts a second set of experimental data coming simulation values at different measuring position. Under different
from Maurus et al. [44]. The experimental apparatus was con- wall heat uxes, simulated and experimental values well matched.
structed and built up for water at low system pressure. The test In constant mass ux, void fraction increases as heat ux increases.
section was a horizontal channel with rectangular section, This is mainly because at constant ow rate, increasing the heat
500 mm long, with 40  40 mm cross-sectional area. A copper ux, the walls superheat increases and more bubbles generated on
bar generates a constant heat ux from the bottom of duct side, heating surface. At the same time, liquid temperature of near-wall
as with the subcooling boiling in engine cooling passages. The region increases as the heat ux increases, bubble condensation
heated wall was a 15 mm  200 mm (width  length) surface, speed decreases, therefore, the bubble volume will increase, and
starting 250 mm downstream of the entrance of the channel. void fraction increases.
Results related to bubble characteristics and vapor distributions Fig. 7 illustrates the comparison between simulated and
were presented for various heat uxes and different liquid mass experimental values of void fraction under different mass uxes.
uxes at a constant subcooling rate of 20 K. The observation area As gure shows, the simulation results are well consistent with
was 14 mm  32 mm big, 10 mm far from the beginning of heated the experimental values, and void fraction decreases with the
654 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

Fig. 3. Comparison of predicted void fractions using different correlations for bubble departure diameter.

increase of mass ux when the same heat ux remains constant. 3.2. Practical application to engine cooling passages
This is because when the heat ux is constant, the increase of ow
rate shortens the heating time of liquid phase, the liquid tem- 3.2.1. Engine basic parameters and boundary conditions
perature of the near-wall region reduced, thus the bubbles apart In order to clearly understand the boiling state in practical en-
from the wall will accelerate the condensation process and gine cooling passages, the numerical simulation of two-phase ow
disappear quickly. Therefore, the void fraction increases as the characteristics in a four-stoke water-cooled diesel engine based on
mass ux decreases. the veried boiling model is carried out. Experimental measure-
ments and numerical simulations only consider cylinder 1; due to
Table 1 the same cavity structure and the independent cooling condition,
Vertical annulus, experimental conditions (Lee et al. [41]). the inuence of other cylinders on cylinder 1 has been neglected.
Fig. 8 (a) illustrates the geometry model of the engine cooling
Experiment no. qw (kW/m2) G (kg/m2 s) Tin( C) Pin (kpa)
passage. The cylinder head temperature measuring points are
1 152.8 478.0 84.3 114.7
showed in Fig. 8 (d). Six K-type thermocouples are located in two
2 169.8 478.1 83.9 114.8
3 225.6 476.8 83.8 116.2 planes at the heights of 5 mm and 12 mm above the ame plate,
4 114.8 477.0 95.6 130.2 respectively.
5 152.9 474.0 95.6 137.5 Basing on the new modied boiling model and the CFD tech-
6 169.7 475.9 95.8 140.5 nology, this paper adopts a conjugate heat transfer simulation of
7 196.9 718.4 83.8 120.8
engine cylinder head and water jacket to get more precise results
8 232.6 718.2 84.0 121.3
9 320.4 718.8 83.8 121.1 [46,47]. The computational domain is a discretized mesh, con-
10 139.1 715.2 93.9 124.7 taining 125,881 elements in the cooling passage domain and
11 197.2 714.4 93.9 132.6 198,093 elements in the cylinder head domain. The wall boiling
12 232.4 716.4 93.9 137.2
model works at the interface of cooling gallery and cylinder head;
13 220.0 1057.2 90.1 134.4
14 251.5 1059.2 90.1 134.5
the two-uid model is used inside of cooling uid, while the only
15 299.9 1061.4 90 134.7 steady heat conduction equations are utilized to calculate the heat
16 173.9 1050.7 100.4 155.1 transfer inside the solid domain.
17 220.2 1054.9 100.2 159.2 In this practical engine, the coolant is pure water, including
18 251.2 1047.2 100.5 163.1
water liquid and water vapor, whose physical properties [48] are
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 655

Fig. 4. Comparison of measured and predicted void fractions, Cases 1e18. (a), (c), (e): high inlet subcooling; (b), (d), (f): low inlet subcooling.

dened from the local uid temperature and saturation tempera- In the meantime, a multi-eld conjugate heat transfer method
ture, respectively. The local saturation temperature is calculated including the two-phase ow heat transfer inside the cooling gal-
from local pressure in passage. Cylinder head is composed of gray lery, the heat conduction in the cylinder head and the spray com-
cast iron and its density is 7200 kg/m3, and specic heat is 475 J/ bustion process inside the cylinder is adopted to simulate the actual
kg K. Table 3 lists the thermal conductivity under different tem- IC engine situation, as Fig. 8 shows. When simulating the spray
peratures. The ow boundary conditions including the ow rate in combustion process of engine, the port wall, ame plate, liner wall
each coolant holes are derived from the CFD analysis of the whole and piston wall are assumed having empirical temperatures in the
engine cooling passage. The calculated mass ow rates are rst iteration. Besides, the boundary of valve-stem wall is set as
0.862 kg/s and 0.729 kg/s at the inlet_1 and inlet_2 respectively, the thermal isolation, because the little part of heat from combustion
outlet pressure is 1 bar and the inlet temperature is 85  C. A very gas can be neglected. Then, the time averaged heat ux prole is
critical factor for the calculation is to set the thermal boundary obtained and specied as boundary condition to the gas side of
condition on engine cylinder head, whose air contacted surfaces are cylinder head surface to complete the uidesolid conjugate heat
set a xed heat transfer coefcient and environment temperature. transfer. Finally, the surface temperature obtained from
656 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

using the same boundary conditions. Fig. 9 shows the comparisons


between experimental and calculated values with single-phase
force convection, and the modied two-uid boiling model at
different testing points. Calculated results show that if boiling is not
considered just like a single-phase, the deviation between predict
and experimental ones is obvious. At the same time, Fig. 9 also
clearly shows that the calculated temperatures of the two-uid
boiling model are much lower than single phase without consid-
ering boiling. That is obvious especially in the area of high thermal
load, such as the measuring points 5 and 6 at different heights,
when the maximum decline is about 13.9  C. The calculation ac-
curacy signicantly improves by activating the two-uid boiling
model as well. However, a discrepancy with the modied boiling
model is still observed at measuring point 5. Two major factors may
cause the difference between simulated and experimental values:
the rst one is the deviation of averaged thermal boundary at local.
The continuous bake of the cylinder head due to the ames and hot
gas, especially at the local of exhausteexhaust bridge area, transfer
a huge amount of heat to cooling passage side. However, the cur-
Fig. 5. Simulation domains, Maurus et al. [44] test section. rent calculation theories of combustion process and heat transfer in
cylinder chamber are not yet mature, which may result in some
deviation for thermal boundary conditions at exhausteexhaust
Table 2
bridge area. The second factor to explain the difference between
Horizontal channel, experimental conditions (Mauruset et al. [44]).
simulations and tests is the limit of the wall boiling model. Strictly
Experiment no. G (kg/m2 s) Tsub(K) Pin (Mpa) qw (MW/m2) speaking, the RPI wall boiling mechanism model is similar to a
1 250 20 0.11 0.2e1.2 semi-empirical model, which is a function of wall temperature in
2 375 20 0.11 0.2e1.2 the nal form. However, the calculation accuracy of heat transfer
3 500 20 0.11 0.2e1.2 characteristics is greatly limited due to too many immature sub-
4 600 20 0.11 0.2e1.2
5 250e500 20 0.11 0.27
models. Especially with high wall heat ux, it is likely to over-
6 250e750 20 0.11 0.37 predict the boiling contribution and then lead to the low temper-
7 250e1000 20 0.11 0.48 ature eld. In spite of this, the trend of computational temperature
8 250e750 20 0.11 0.63 at other testing points coincides with experimental values when
9 250e1250 20 0.11 0.77
the modied two-uid boiling model is used. The mean deviation
10 250e2000 20 0.11 0.94
drops from 4.6% without boiling consideration to 1.43% with the
modied boiling model. Thus, using the modied two-uid boiling
conjugating heat transfer will be used as a boundary condition for model can help improve the precision of CFD result in engine heat
combustion process simulation in the next iteration. The iterative transfer analysis.
method is applied until the temperature of cylinder head tended to Fig. 10 shows the distribution of temperature on the cooling
stabilization. Fig. 8 (c) shows the nal contour of time-averaged passage surfaces for single-phase forced convection and two-uid
heat ux prole at the gas side surface, which includes the gas boiling model. The high temperature regions are close to ame
side of intake port, gas side of exhaust port, gas side of valve seats plate and exhaust port and the trend of distribution is the same for
and ame plate. both the two calculation methods. The temperature contours show
that some of the calculated cooling passage side wall temperatures
3.2.2. Temperature eld analysis are higher than local saturation temperature, meaning that boiling
In order to estimate the effect of two-uid boiling model for heat transfer took place in most of this region. Therefore, boiling
calculation, a single phase analysis is also carried out in engine head factor in engine cooling passage cannot be neglected. As shown in

Fig. 6. Comparison between measured and predicted void fractions under different heat uxes.
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 657

Fig. 7. Comparison between measured and predicted void fractions under different mass uxes.

Fig. 8. The CFD simulation domains of engine cooling system and measuring points.
658 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

Table 3
Gray cast iron thermal conductivity.

Temperature ( C) 100 200 300 400 500


Thermal conductivity (W/(m K)) 50 49.5 49 48 47

Fig. 11. Comparisons between wall temperature distribution on cylinder head surfaces
with single-phase forced convection and two-uid boiling model.

heat convection regions. However, in some high temperature re-


Fig. 9. Comparisons between experimental data and predicted values at each gions such as bridge passages, the coefcients of heat transfer
measuring points. greatly increase because of boiling. The above-mentioned analysis
and discussion on the two-uid boiling model lead to a great
improvement of calculation accuracy in temperature eld of engine
Fig. 10 (b), the highest temperature reached with two-uid boiling cylinder head. These results represent an important premise for
model is 407.5 K, and the corresponding wall superheat is about improvement of thermal stress analysis and are very useful for
32.2 K. As shown in Fig. 10 (a), the temperature is 22.1 K lower than cooling passage design and optimization.
that of 429.6 K calculated with single-phase forced convection.
Fig. 11 shows the comparisons of wall temperature distribution 3.2.3. Flow eld and vapor phase analysis
on cylinder head surfaces with single-phase forced convection and Due to the complexity, variety and random of boiling phenom-
two-uid boiling model. The high temperature regions are located ena, the two-phase boiling cannot be calculated as conventional
in the ame plate and exhaust port where directly contact the hot heat conduction, single-phase convection and thermal radiation.
gases. In this case, the maximum temperature with single-phase Although nucleate boiling heat transfer has high heat transfer ef-
forced convection is 619.2 K, while the maximum temperature of ciency at low wall superheat level, a further heat ux increase may
the two-uid boiling model is 610.1 K. After considering boiling lead to a transition to the lm boiling mode. By then, the vapor
heat transfer, most of temperatures of gas sidewalls drop, similarly generated on the wall is not removed from the vicinity of the wall
to the trend of changes in cooling passage side. The main reason is fast enough, leading to the termination of macroscopic contact
that the difference of heat transfer efciency between forced con- between liquid and wall. This causes a drastic decrease of the heat
vection and nucleate boiling is obvious. When boiling is considered, transfer efciency and a sharp increase of the temperature in the
the absolute value of heat transfer coefcient is low in most pure heated wall, which may damage the equipment. The development
design of engine cooling systems must prevent such overheating
phenomena. Therefore, the prevention cannot disregard the
detailed analysis of ow characteristics and vapor phase
distribution.
Fig. 12(a) shows a three-dimensional distribution of streamline
path in all passages, where the coolant maximum velocity is
5.06 m/s and the minimum bulk velocity is 0.001 m/s in the single-
phase ow. After cooling water from inlet hole enters into the
outside of water jacket, part of it directly ows to the top of water
jacket and then joins into outlet through the two passages at
exhaust port side. This part of coolant (that represents the 18.8% of
the total ow rate coolant) has no cooling effects on the bottom of
cylinder head. Another fraction of coolant ows to the bottom of
the cylinder head to cool ame plate, and then through four bridges
passage arrives at outlet. This part of coolant represents the 81.2% of
the total ow rate, where the percentage of passing
exhausteexhaust bridge passage is about 27.6%. Fig. 12(b) and (c)
show the distribution of liquid velocity and vapor velocity in two-
uid model respectively. Figures show that the streamlines for
different phases are similar in case of two-phase ow. Moreover,
Fig. 10. Comparisons between wall temperature distribution on cooling passage sur- Fig. 12 (a) and (b) show that the liquid velocity in case of two-phase
faces with single-phase forced convection and two-uid boiling model. ow is still similar to the one of single-phase ow. Even so, the
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 659

Fig. 12. Velocity streamlines global distribution in engine cooling passages.

liquid velocity of two-phase boiling ow generally deviates from approaching the heated surface, like Yeoh's vertical annulus
the velocity of single-phase ow due to the non-uniform void simulation report [49].
fraction distribution in local area. Fig. 13 shows the distribution of The relation of wall heat ux and bubble departure diameter has
velocity vector in the thermal critical area; the section position is been introduced in the previous section. Bubble diameter also af-
located at the exhausteexhaust bridge passage, where there is the fects the exchange of mass and momentum in bulk ow. In this
highest thermal load. In the case of two-phase boiling ow, the paper, the new bubble departure diameter model and bubble
maximum liquid velocity is 3.73 m/s, slightly higher than 3.65 m/s diameter model are adopted. Fig. 14 (a) and (b) depict the distri-
for single-phase ow, and moves toward to the heated wall. bution of the bubble diameter and bubble departure diameter near
As shown in Fig. 13 (b) and (c), due to buoyancy force caused by the engine cooling passage. Although there is not experimental
density difference the vapor velocity is greater than the liquid ve- data in engine water jacket to verify the correctness of the pre-
locity. Lee's measurements [41] demonstrated that the vapor ve- dicted diameter, the tendencies of bubble diameter and bubble
locity was higher at the center than the velocities near the heating departure diameter are reasonable. Compared with the previous
rod. This is probably due to the buoyancy effect enhanced by the models, the new model correlates with local ow parameters and
migration to this area of the large bubbles. Further, the vapor ve- physical property parameter, and it is no longer a xed value. As
locity predicted by the current boiling model shows higher values show in Fig. 14 (a), the maximum bubble diameter is about

Fig. 13. Velocity vector distribution in thermal critical area.


660 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

2.41 mm, located at the bottom of engine water jacket, especially at


high temperature zones such as bridges and exhaust side. Fig. 14 (b)
highlights that the maximum bubble departure diameter is 5.0 mm
because of additional constraint. Compared with the distribution of
bubble diameter, the distribution of bubble departure diameter is
concentrated in bridges where boiling phenomenon often occurs;
otherwise, the bubble departure diameter is zero without boiling.
Fig. 15 shows the distribution of heat ux at the cooling passages
wall. According to the RPI boiling model, the total wall heat ux is
split into three different components, i.e. single-phase convection
heat ux, quenching heat ux and evaporation heat ux. Where the
total of convection heat ux and quenching heat ux are trans-
ferred to the liquid phase and the evaporation heat ux is trans-
ferred to the vapor phase. As showed in Fig. 15(b) and (c), most of
heat ux is transferred to the liquid phase, and the evaporation heat
ux is only a small portion in the wall heat ux, even near the
exhausteexhaust bridge passage. The maximum evaporation heat
ux is 32,000 W/m2. Although the single-phase convection heat
ux gradually decreases with the rise of wall temperature, the
quenching and evaporation heat ux increase, especially the sharp
rise in quenching. Therefore, the heat ux to liquid phase still oc-
cupies the major part.
Fig. 16 presents the distribution of vapor-phase information at
cooling passage near the heated wall. Fig. 16 (a) displays the dis-
tribution of evaporative mass ux, which keeps the same when
varying wall temperature. High superheat will make boiling more
violent and more bubbles will be generated near wall, so the
Fig. 14. Bubble diameter distribution and bubble departure diameter distribution near
the heated wall. maximum evaporative mass ux is 0.00138 kg/m2s located at
exhausteexhaust bridge passage closely to the ame plate side,

Fig. 15. Heat ux distribution at the cooling passages wall.


S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 661

(a) Compared with Tolubinski, Unal and Fritz's models, the new-
modied model predictions greatly improved the simple
channel analysis, and the calculated local void fractions
agreed well with the measuring data related to upward ow
in vertical annulus and horizontal ow in rectangular
channel.
(b) Compared with the single-phase ow calculated in engine
cooling passage, the heat transfer enhancement effect due to
the boiling heat transfer at heated wall was considered using
the present two-uid model. Results showed that the new
model reected the real heat transfer and gave an accurate
temperature eld in engine cylinder head. This result rep-
resents an important premise for thermal stress analysis and
cooling passage optimization.
(c) Based on current results, from a global viewpoint the present
boiling state did not introduce a different ow eld, and the
structure of velocity streamlines for liquid and vapor in both
cases of two-phase and single-phase ow in engine cooling
passage were similar. However, due to buoyancy force caused
by density difference, the vapor velocity was evidently
greater than the liquid velocity in local area.
(d) The new two-uid boiling model gave detailed information
on vapor phase in practical engine cooling passage, helping
the engine developing and providing new basis for cooling
systems use with reliable boiling heat transfer. In the pre-
sented practical application, most of the void fraction was
Fig. 16. Evaporation mass ux distribution and void fraction distribution near the below 0.1. Similar values have been found even in high
heated wall. thermal regions such as bridge passages, due to the function
of interphase forces that the majority of bubbles were quickly
removed from near wall. Furthermore, the boiling heat
where the heat ux and wall temperature are highest. However,
transfer in this engine cooling passage was in a stable state
the maximum void fraction equals to 0.149 near the exhaust port
from the point of view of void fraction.
where the velocity is relatively low; the detailed void fraction
distribution is showed in Fig. 16 (b). It is clear that most of the void
Acknowledgements
fraction is less than 0.1 even at high thermal region such as
exhausteexhaust bridge passage. The main reason is that the
This work was funded by the National High Technology
generated bubbles are quickly removed from heated wall under
Research and Development Program of China [Grant
high velocity conditions, due to the inuence of interphase forces
No.2008AA11A121], and the Specic Project of Important National
such as lift force, turbulent dispersion force and wall lubrication
Science & Technology [Grant No. 2011GB113002].
force. Although the present two-uid model cannot directly pro-
vide the distribution of critical heat ux, the obtained void fraction
Nomenclature
of gas phase will be indirectly used to evaluate the reliable boiling
for cooling system. If the void fraction is too high, the ow
resistance in passages will increase, and the boiling heat transfer
Symbols
will get worse and the lm boiling occurs. Therefore, from the
Ab the fraction of wall area subjected to cooling by
point of view of void fraction, the boiling heat transfer in this
quenching
engine cooling passage is in a stable state, because the level of
Acon the single-phase convection inuenced area
vapor volume is low near the heated wall.
Alg interfacial area density
CFD computational uid dynamics
4. Summary and conclusion CD drag coefcient
CL lift force coefcient
The numerical investigation of two-phase ow characteristics in cp specic heat
subcooled boiling ow at low pressure in different passages has Ctd turbulent dispersion coefcient
been carried out using a three-dimensional two-uid model. In Cmb coefcient for additional viscosity term
addition, a new modied bubble departure diameter model for db bubble diameter
low-pressure condition was proposed, basing on the Unal's model dw bubble departure diameter
and on the Tolubinski and Kostanchuk's correlation. The model was dref reference departure diameter
implemented in an ANSYS-CFX CFD code using a user dened db0 the minimum bubble diameter
FORTRAN program. The modied model was veried using two sets db1 the maximum bubble diameter
of experimental data including different mass ow rates, pressures, f bubble detachment frequency
subcooling, heat uxes, section shapes, and ow directions. The F bubble force
veried model was nally used to simulate two-phase boiling ow g gravitational acceleration
characteristics in a practical cooling passage for a four-stoke water- h heat transfer coefcient
cooled diesel engine. The main results of this study can be sum- H specic enthalpy
marized as follows: IC internal combustion
662 S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663

! [5] M.W. Wambsganss, Thermal Management in Heavy Vehicles: a Review


nw unit normal pointing away from the wall vertical
Identifying Issues and Research Requirements, Argonne National Lab., IL (US),
direction 1999.
Nw wall nucleation site density [6] K. Robinson, N.A.F. Campbell, J.G. Hawley, D.G. Tilley, A Review of Precision
Nu Nusselt number Engine Cooling, 1999. SAE paper 1999-01-0578.
[7] W. Yu, D.M. France, D. Singh, R.K. Smith, J. Ritter, T. Vijlbrief, et al., Subcooled
k thermal conductivity ow boiling of ethylene glycol/water mixtures in a bottom-heated tube, Int. J.
K empirical constant Heat. Mass Transf. 72 (2014) 637e645.
P pressure [8] H. Steiner, A. Kobor, L. Gebhard, A wall heat transfer model for subcooled
boiling ow, Int. J. Heat Mass Transf. 48 (19e20) (2005) 4161e4173.
Pr Prandtl number [9] S. Hua, R. Huang, Z. Li, P. Zhou, Experimental study on the heat transfer
q heat ux characteristics of subcooled ow boiling with cast iron heating surface, Appl.
RPI Rensselaer polytechnic Institute Therm. Eng. 77 (0) (2015) 180e191.
[10] K. Robinson, N.A.F. Campbell, J.G.H. Hawley, M.J. Leathard, Predicting Heat
Reb the Reynolds number of vapor
Transfer in Simulated IC Engine Cooling Galleries, 2003, pp. 787e803. ImechE
!
Rpq interaction force Paper.
[11] Z. Li, H.H. Rong, Subcooled boiling heat transfer modelling for internal com-
T temperature bustion engine applications, Proc. Inst. Mech. Eng. Part J. Automob. Eng. 226
Ts interfacial temperature (2012) 301e311.
! [12] T. Bo, CFD Homogeneous Mixing Flow Modelling to Simulate Subcooled
u velocity vector
Nucleate Boiling Flow, 2004. SAE International Paper.
yw distance to the nearest wall [13] G. Li, S. Fu, Y. Liu, Y. Liu, S. Bai, L. Cheng, A homogeneous ow model for
boiling heat transfer calculation based on single phase ow, Energy Convers.
Manag. 50 (7) (2009) 1862e1868.
Greek symbols [14] V. Przulj, M. Shala, In Multi-phase Mixture Modelling of Nucleate Boiling
meff
l
effective viscosity Applied to Engine Coolant Flows, WITPress, New Forest, United kingdom,
2009, pp. 135e146.
ml molecular viscosity [15] A. Mohammadi, H. Hashemi, A. Jazayeri, M. Ahmadi, In Two Phase Flow
mturb
l
turbulence viscosity Simulation for Nucleate Boiling Heat Transfer Calculation in Waterjacket of
mbl Diesel Engine, American Society of Mechanical Engineers, Hamamatsu, Japan,
bubble-induced turbulence viscosity
2011, pp. 1721e1729.
G mass transfer [16] M. Shala, In Simulation of Nucleate Boiling Flow Using a Multiphase Mixture
a volume fraction Modelling Approach, Oxford University Press, Great Clarendon Street, Oxford,
OX2 6DP, United Kingdom, 2012, pp. 47e58.
r density
[17] H. Punekar, S. Das, Numerical Simulation of Subcooled Nucleate Boiling in
s surface tension Cooling Jacket of IC Engine, 2013. SAE paper: 2013-01-1651.
st turbulent Schmidt number [18] N. Kurul, M.Z. Podowski, On the modeling of multidimensional effects in
llg latent heat boiling channels, in: ANS Proc.27th National Heat Transfer Conference, Min-
neapolis, MN, July 28-31, 1991.
Jasub subcooled Jacob number [19] H. Anglart, O. Nylund, CFD application to prediction of void distribution in
tw bubble waiting time two-phase bubbly ows in rod bundles, Nucl. Eng. Des. 163 (1e2) (1996)
DTsat TwTsat superheat temperature 81e98.
[20] W. Yao, C. Morel, Prediction of parameters distribution of upward boiling
DTsub TsatTl subcooled temperature two-phase ow with two-uid models, Int. Conf. Nucl. Eng. (2002)
801e808.
[21] V. Ustinenko, M. Samigulin, A. Ioilev, S. Lo, A. Tentner, A. Lychagin, A. Razin,
Subscripts V. Girin, Y. Vanyukov, Validation of CFD-BWR, a new two-phase computa-
b bubble tional uid dynamics model for boiling water reactor analysis, Nucl. Eng. Des.
238 (3) (2008) 660e670.
con single-phase convection [22] O. Zeitoun, M. Shoukri, Axial void fraction prole in low pressure subcooled
drag body force ow boiling, Int. J. Heat Mass Transf. 40 (4) (1997) 869e879.
e evaporation [23] J.Y. Tu, G.H. Yeoh, On numerical modelling of low-pressure subcooled boiling
ows, Int. J. Heat Mass Transf. 45 (6) (2002) 1197e1209.
F Fritz [24] J.Y. Tu, The inuence of bubble size on void fraction distribution in subcooled
g vapor phase ow boiling at low pressure, Int. Commun. Heat Mass Transf. 26 (5) (1999)
l liquid phase 607e616.
[25] B. Kon car, I. Kljenak, B. Mavko, Modelling of local two-phase ow parameters
lift lift force
in upward subcooled ow boiling at low pressure, Int. J. Heat Mass Transf. 47
que quenching (6e7) (2004) 1499e1513.
ref reference value [26] M. Ishii, K. Mishima, Two-uid model and hydrodynamic constitutive re-
sat saturation lations, Nucl. Eng. Des. 82 (1984) 107e126.
[27] C. Wang, T. Cong, S. Qiu, W. Tian, Y. Wu, G. Su, Numerical prediction of sub-
T Tolubinski and Kostanchuk cooled wall boiling in the secondary side of SG tubes coupled with primary
td turbulent dispersion force coolant, Ann. Nucl. Energy 63 (2014) 633e645.
U Unal [28] ANSYS-CFX Theory Guide, 2012.
[29] Y. Sato, M. Sadatomi, K. Sekoguchi, Momentum and heat transfer in two-
vm virtual mass force phase bubble oweeI. Theory, Int. J. Multiph. Flow. 7 (1981) 167e177.
w wall [30] M. Ishii, N. Zuber, Drag coefcient and relative velocity in bubbly, droplet or
inter interphase particulate ows, AIChE J. 25 (1979) 843e855.
[31] A. Tomiyama, Struggle with computational bubble dynamics, in: 3rd Inter-
wl wall lubrication force national Conference on Multiphase Flow, Lyon, France, 1998, pp. 1e18.
[32] B. Kon car, E. Krepper, CFD simulation of convective ow boiling of refrigerant
in a vertical annulus, Nucl. Eng. Des. 238 (3) (2008) 693e706.
References [33] H.J. Ivey, Relationships between bubble frequency, departure diameter and
rise velocity in nucleate boiling, Int. J. Heat. Mass Transf. 10 (1967)
[1] H.H. Pang, C.J. Brace, Review of engine cooling technologies for modern en- 1023e1040.
gines, Proc. Inst. Mech. Eng. Part J. Automob. Eng. 218 (2004) 1209e1215. [34] H. Victor, M. Del Valle, D.B.R. Kenning, Subcooled ow boiling at high heat
[2] J.B. Heywood, O.Z. Welling, Trends in Performance Characteristics of Modern ux, Int. J. Heat. Mass Transf. 28 (1985) 1907e1920.
Automobile SI and Diesel Engines, 2009. SAE paper 2009-01-1892. [35] V.I. Tolubinski, D.M. Kostanchuk, Vapour bubbles growth rate and heat
[3] H. Steiner, G. Brenn, F. Ramstorfer, B. Breitsch
adel, Increased cooling power transfer intensity at subcooled water boiling, in: 4th International Heat
with nucleate boiling ow in automotive engine applications, in: M. Chiaberge Transfer Conference, Paris, France, 1970.
(Ed.), New Trends and Developments in Automotive System Engineering, [36] H.C. Unal, Maximum bubble diameter, maximum bubble growth time and
InTech, 2011, pp. 249e272. bubble-growth rate, Int. J. Heat. Mass Transf. 19 (1976) 643e649.
[4] N.A.F. Campbell, S.J. Charlton, L. Wong, Designing towards Nucleate Boiling in [37] O. Zeitoun, M. Shoukri, Bubble behavior and mean diameter in subcooled ow
Combustion Engines, 1995. SAE paper 95407. boiling, 1996, ASME J. Heat. Transf. 118 (1996) 110e116.
S. Hua et al. / Applied Thermal Engineering 90 (2015) 648e663 663

[38] M.D. Bartel, Experimental Investigation of Subcooled Boiling, M.S.N.E. Thesis, [44] R. Maurus, V. Ilchenko, T. Sattelmayer, Study of the bubble characteristics and
Purdue University, West Lafayette, IN, USA, 1999. the local void fraction in subcooled ow boiling using digital imaging and
[39] V. Prodanovic, D. Fraser, M. Salcudean, Bubble behavior in subcooled ow analysing techniques, Exp. Therm. Fluid Sci. 26 (2) (2002) 147e155.
boiling of water at low pressures and low ow rates, Int. J. Multiph. Flow. 28 [45] A.J. Torregrosa, A. Broatch, P. Olmeda, O. Cornejo, Experiments on subcooled
(2002) 1e19. ow boiling in I.C. engine-like conditions at low ow velocities, Exp. Therm.
[40] W. Fritz, Berechnung des Maximalvolumes von Dampfblasen, Phys. Z 36 Fluid Sci. 52 (2014) 347e354.
(1935) 379. [46] S. Fontanesi, M. Giacopini, Multiphase CFDeCHT optimization of the cooling
[41] T.H. Lee, G.C. Park, D.J. Lee, Local ow characteristics of subcooled boiling ow jacket and FEM analysis of the engine head of a V6 diesel engine, Appl. Therm.
of water in a vertical concentric annulus, Int. J. Multiph. Flow. 28 (2002) Eng. 52 (2) (2013) 293e303.
1351e1368. [47] S. Hua, R. Huang, Z. Li, Multi-eld conjugating heat transfer in cylinder head of
[42] Y.L. Erfeng Chen, Modeling of low-pressure subcooled boiling ow of water internal combustion engines based on boiling model, Neiranji Xuebao/Trans.
via the homogeneous MUSIG approach, Nucl. Eng. Des. 239 (10) (2009) CSICE 32 (4) (2014) 364e371.
1733e1743. [48] ASHRAE, Fundamentals Handbook, ASHRAE, Atlanta, GA, USA, 2001.
[43] R.P. Roy, V. Velidandla, S.P. Kalra, P. Peturaud, Local measurements in the two- [49] G.H. Yeoh, J.Y. Tu, A bubble mechanistic model for subcooled boiling ow
phase region of turbulent subcooled boiling ow, ASME J. Heat. Transf. 116 predictions, Numer. Heat. Transf. Part B Fundam. 45 (5) (2004) 475e493.
(1994) 660e669.

Вам также может понравиться