Nuclear Engineering and Design 308 (2016) 30–37

Contents lists available at ScienceDirect

Nuclear Engineering and Design

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

Prediction of subcooled flow boiling characteristics using two-fluid

Eulerian CFD model
Francisco A. Braz Filho, Guilherme B. Ribeiro ⇑, Alexandre D. Caldeira
Nuclear Energy Division, Institute for Advanced Studies, Aerospace Science and Technology Department, Trevo Coronel Av. José Alberto Albano do Amarante 1, 12228-001 São José
dos Campos, Brazil

h i g h l i g h t s

CFD multiphase model is used to predict subcooled flow boiling characteristics.

Better agreement is achieved for higher saturation pressures.

Onset of nucleate boiling and saturated boiling are well predicted.

CFD multiphase model tends to underestimate the void fraction.

Factors were adjusted in order to improve the void fraction results.

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

Article history: The present study concerns a detailed analysis of flow boiling phenomena under high pressure systems
Received 20 November 2015 using a two-fluid Eulerian approach provided by a Computational Fluid Dynamics (CFD) solver. For this
Received in revised form 8 August 2016 purpose, a vertical heated pipe made of stainless steel with an internal diameter of 15.4 mm was consid-
Accepted 12 August 2016
ered as the modeled domain. Two different uniform heat fluxes and three saturation pressures were
Available online 23 August 2016
applied to the channel wall, whereas water mass flux of 900 kg/m2 s was considered for all simulation
cases. The model was validated against a set of experimental data and results have indicated a promising
JEL classification:
use of the CFD technique for estimation of the wall temperature, the liquid bulk temperature and the
K. Thermal Hydraulics
location of the departure of nucleate boiling. Changes in factors applied in the modeling of the interfacial
heat transfer coefficient and bubble departure frequency were suggested, allowing a better prediction of
the void fraction along the heated channel. The commercial CFD solver FLUENT 14.5 was used for the
model implementation.
Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction of liquid by the single-phase convective–conductive mechanisms

(Collier and Thome, 1994). This phenomenon occurs when the sat-
Several industrial processes apply flow boiling as a heat transfer uration temperature is exceeded only in a local vicinity of the
mechanism due to its large convective heat transfer. In nuclear sys- heated wall, whereas the bulk temperature is still below satura-
tems, boiling commonly occurs in the hot fuel assemblies of a Pres- tion. Then, subcooled boiling is initiated when the wall superheat
surized Water Reactor (PWR) and in all assemblies of a Boiling reaches a level that activates bubble nucleation. The position
Water Reactor (BWR). Thus, an accurate prediction of the flow boil- where the bubble nucleation initially appears is called onset of
ing phenomena is crucial for a safe operation in these facilities. nucleate boiling (ONB). Steam bubbles are generated at the heated
Furthermore, the proper prediction of the void fraction and average surface at nucleation sites. Further downstream the attached bub-
bubble size along the nuclear core are important aspects that bles grow and then leave the wall at certain critical size. This crit-
should be accounted for the neutron flux estimation and nuclear ical size may depend on the surface tension and on the flow regime
reactivity prediction. of the surrounding fluid.
Subcooled boiling is observed at heated surfaces, when the heat After nucleation, bubbles grow rapidly and then migrate farther
flux applied to the wall is too high to be transferred to the core flow away from the heated wall due to a combination of drag, lift and
buoyancy forces. The difference of temperature between the vapor
⇑ Corresponding author. Fax: +55 12 39441177. bubbles and the subcooled liquid drives the steam condensation,
E-mail address: [email protected] (G.B. Ribeiro). leading to the collapse of the bubbles. Bubble condensation ceases

http://dx.doi.org/10.1016/j.nucengdes.2016.08.016
0029-5493/Ó 2016 Elsevier B.V. All rights reserved.
 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37 31

Nomenclature

A area, m2 q density, kg m3

C coefficient / generic variable
cp specific heat, J kg1 K1 u area of influence
D
!
diameter, m s relaxation time of particulate, s
F force, N s stress tensor, N m2
f drag factor
g gravity, m s2 Subscripts
H heat transfer coefficient, W m2 K1 b bubble
h specific enthalpy, J kg1 cell cell
Ja jacob number d discharge
K coefficient of Exchange, kg m3 s1 drag drag
k turbulent kinetic energy, J kg1 e evaporation
k thermal conductivity, W m1 K1 ht heat transfer
L latent heat, J kg1 i interfacial
m _ mass transfer rate per volume, kg s1 m3 l liquid
N nucleate site density, sites m2 lift lift
!
n unitary vector N nucleation site
P pressure, Pa p phase p
Pr Prandtl number pq phase p to q
q heat transfer rate, W q phase q
q_ heat flux, W m2 q quenching
Re Reynolds number qp phase q to p
T temperature, K sat saturation
!
v velocity, m s1 sub subcooled
V volume, m3 td turbulent dispersion
vm virtual mass
Greek symbols w wall
D difference wl wall lubrication
p Pi number u area of influence
l dynamic viscosity, kg m1 s1
a volume fraction

when the bulk liquid temperature reaches saturation and saturated flow boiling characteristics inside high a pressure system for
convective boiling takes place. non-uniform heating profiles. It was concluded that the non-
Heat transfer from the wall is then described as being carried by uniformity of the input heat flux distribution resulted in significant
turbulent convection of liquid, by transient conduction due to the changes in the heat transfer coefficient and void fraction distribu-
departing bubbles, and by evaporation. Distribution on the entire tion through the channel position. As well, several other studies
wall of heat flux between these mechanisms (wall heat partition- have adopted CFD as a numerical methodology for flow boiling
ing) can be calculated by modeling each mechanism in terms of analysis. Among these studies, the works developed by Zhang
the nucleation site density, the size of departing bubbles, their et al. (2015), Shaver et al. (2013) Koncar and Tiselj (2010),
detachment frequency, and waiting time until the next bubble Krepper et al. (2007), Erfeng et al. (2009), Koncar et al. (2004),
appears on the same site (mechanistic modeling approach). Warrier et al. (2002), Anglart et al. (1997), Dhir et al. (2003),
In a recent article of Braz Filho et al. (2011), the analysis of Prodanovic et al. (2002) and Troshko and Hassan (2001) can be
upward subcooled flow boiling using CFD methodology was used highlighted.
and compared with the classical experimental data of Bartolomei Although the use of the CFD technique is considered promising,
and Chanturiya (1967). The study showed a good prediction of all CFD boiling modeling approaches are in the stage of validation
the ONB, as well as the beginning of saturated boiling. Moreover, of the main phenomena found in industrial applications. Currently,
the calculated void fraction at the channel outlet presented a good two-phase CFD is still a tool that demands a long effort to reach a
agreement with the experimental data. However, the prediction of satisfactory level of reliability and maturity required for regulatory
void fraction distribution along the channel was not satisfactory. In commissions (Bestion, 2012, 2014). Thus, as an intention to
order to improve the void prediction, Braz Filho et al. (2013) pro- address potential drawbacks, the present study used an Eulerian
posed a sensitivity analysis of the source terms of mass balance Multiphase Model to analyze and validate the subcooled flow boil-
equation of both phases, adjusting the vapor production and con- ing phenomena. Also, the model description and results presented
densation processes during the flow boiling. during this study can serve as guidelines for the implementation of
Moreover, Li et al. (2011) presented a new implementation of more complex simulations that apply the exact geometry of reac-
subcooled flow boiling analysis using water as a working fluid. In tor components.
that study, a saturation pressure of 45 bar was used and several The CFD commercial solver FLUENT 14.5 (ANSYS, 2013a) was
turbulence and interfacial transfer models (drag, lift, heat transfer, used in the analysis in conjunction with the boiling model devel-
mass transfer, etc) were studied and compared to Bartolomei and oped by Kurul and Podowski (1990). Furthermore, a series of
Chanturiya (1967) data. Nemitallah et al. (2015) investigated the sub-models were implemented to model the interfacial momen-
32 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37

tum, mass transfer, heat transfer and turbulence interactions. Two 2.2. Mathematical modeling
factors, which adjust the interfacial heat transfer and the waiting
time between departures of two consecutive bubbles, are applied The proposed CFD modeling uses Eulerian Multiphase Model
in order to improve the prediction of flow boiling characteristics. (EMM) with the addition of heat transfer correlations and source
terms in the conservation equations (ANSYS, 2013b). The conserva-
tion equations are written for each phase, following the EMM
2. Model description approach. The summary of the main equations for a generic phase
q are presented as follows.
2.1. Simulation domain and conditions The description of multiphase flow as interpenetrating continua
incorporates the concept of phase volume fractions, denoted here
The system geometry is shown schematically in Fig. 1 and con- by aq . Volume fractions consist in the amount of space occupied
sists in a 2 m heated stainless steel tube with inner diameter of by the respective phase, while the laws of conservation of mass,
15.4 mm. Water was used as a working fluid with a mass flow of momentum and energy are satisfied by each phase separately.
900 kg/s m2. Subcooled water with a subcooled degree of 60 K The mathematical derivation of the conservation equations can
enters into the bottom side of the channel and flows upward be done by ensemble averaging the local instantaneous balance
against gravity. Saturation pressures of 4.5, 3.0 and 1.5 MPa were for each phase or by using the mixture theory approach. The vol-
applied (i.e., saturation temperatures of 530.44, 507.05 and ume of phase q, V q , is defined as
471.47 K, respectively), while uniform heat fluxes of 380 and
Z
570 kW/m2 were applied at the tube wall. The system geometry
Vq ¼ aq dV ð1Þ
and conditions are the same used in the benchmark experiment V
of Bartolomei and Chanturiya (1967), which is used later for vali-
dation purposes. The comparison with the DEBORA (Garnier where in a two-phase system,
et al., 2001) experiment could also be performed, though R-12 is aq þ ap ¼ 1 ð2Þ
applied as working fluid. The high pressure R-12 data provided
by DEBORA (Garnier et al., 2001) experiment corresponds to a The mass conservation equation for phase q is given by
fluid-vapor density ratio that is equal to that of water-steam at
@ !
15 MPa. Nevertheless, the databank provided by Bartolomei and ðaq qq Þ þ r  ðaq qq v q Þ ¼ m
_ pq  m
_ qp ð3Þ
@t
Chanturiya (1967) is until today one of the most used experiment
!
to validate CFD tools under subcooled flow boiling due to its con- where v q and qq are the velocity vector and density of phase q,
ditions such as the use of water, high saturation pressures, heat _ pq consists in the mass transfer rate per vol-
respectively. The term m
fluxes and flow rates (Li et al., 2011; Nemitallah et al., 2015;
ume from phase p to q, whereas m _ qp consists in the mass transfer
Shaver et al., 2013; Krepper et al., 2007). Considering all these
rate per volume from phases q to p.
aspects, the experiment of Bartolomei and Chanturiya (1967) was
The vapor formation rate per unit of volume is represented by
chosen at the present analysis.
the variable m_ lv and is described as the contribution of the interfa-
Thermophysical properties of the liquid phase are assumed to
cial and wall-vapor mass transfer rates. Thus, m _ lv can be related to
vary with the temperature, whereas the vapor phase properties
m_ pq as
are considered at the saturation temperature. The problem was
treated as steady state condition and a two-dimensional axisym- ðT l  T sat Þ Hht Ai q_ e Acell
m _ lv
_ pq ¼ m þ ð4Þ
metric model was chosen to represent the stainless steel tube. L þ cp;l ðT sat  T l Þ L þ cp;l ðT sat  T l Þ
For this particular study, the use of tri-dimensional geometry
would increase unreasonably the computational cost and the con- Here Ai is the interfacial area density, Acell is the cell face area, T l is
vergence efforts of the simulation. A mesh with uniform and the liquid temperature, T sat is the saturation temperature, L is the
square control volumes was created and a grid independence anal- specific latent heat, q_ e is the evaporative heat flux, cp;l is the liquid
ysis was performed in order to achieve unchangeable accurate specific heat. The interfacial heat transfer coefficient Hht is calcu-
solutions for flow boiling. For this reason, three quadrangular lated via the pioneer experimental study of Ranz and Marshall
meshes with 10, 20 and 60 k volumes were evaluated. (1952), where the evaporation of spherical water droplets in
With the purpose of ensuring conformity to the physical prob- upward flow of dry air were used to develop the following
lem, a fully developed liquid flow was considered as flow input. correlation,
 
kl 
A previous simulation was performed without wall heating and
the field solution (i.e., velocity vector and turbulence variables) Hht ¼ C ht 2 þ 0:6Re0:5
b Pr
0:33
ð5Þ
Db
in steady state was applied as the flow profile at the channel inlet.

Fig. 1. Geometry and corresponding boundary conditions.

 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37 33

and !
The wall lubrication force F wl;q is characterized as the force
6aq responsible to push the bubbles away from the wall to the liquid
Ai ¼ ð6Þ bulk and is defined as
Db
! ! ! 2 !
whereas the interfacial area density Ai is given considering equal- F wl;q ¼ C wl qq ap  v q  v p  n w ð15Þ
size spherical bubbles (Kurul and Podowski, 1990), Reb is the rela-
tive Reynolds number based on the diameter of phase q and the rel- Here C wl is the wall lubricant coefficient calculated according to
! ! Antal et al. (1991) for laminar and fully-developed bubbly flow at
ative velocity  v q  v p , and the Prandtl number Pr is calculated for
!
phase p. The factor C ht is the heat transfer factor used to adjust the an adiabatic air–water flow and n w is the unitary vector perpendic-
interfacial heat transfer coefficient Hht and is set to 1.0 by default. ular to the wall. As pointed out by Lopez-de-Bertodano and
The bubble departure diameter Db is obtained via the Tolubinski Prabhudharwadkar (2010), Antal et al. (1991) presented better pre-
and Kostanchuk correlation (Tolubinski and Kostanchuk, 1970), diction results than the parameters suggested by Krepper and
where the bubble diameter of water is function of the bulk subcool- Prasser (2000).
!
ing as The turbulent dispersion force F td;q represents the effect of the
 DT sub  diffusion of small bubbles caused by whirls encountered in the cap
Db ¼ min 0:0006  e 45 ; 0:0014 ð7Þ bubbly flow. This force can be represented according to Lopez-de-
Bertodano (1992) as
where DT sub is the subcooled degree. The momentum conservation
!
equation for phase q yields F td;q ¼ C td qq kq rap ð16Þ
@ ! !! !
ðaq qq v q Þ þ r  ðaq qq v q v q Þ ¼ aq rP þ r  sq þ aq qq g The turbulent kinetic energy is represented by variable kq ,
@t !  whereas the turbulent dispersion coefficient is represented by C td
! ! ! !
þm _ pq !
v p  m_ pq !
v q þ F drag;q þ F lift;q þ F wl;q þ F td;q þ F v m;q ð8Þ and is set to 1.0. According to Lopez-de-Bertodano and
Prabhudharwadkar (2010), better agreements with the experimen-
where P is the pressure shared by all phases, sq is the q phase tal data has been reached when the turbulent dispersion coeffi-
!
stress–strain tensor and v pq is the inter-phase velocity. Variables cient C td is higher than 0.25, originally obtained for adiabatic air–
! ! ! ! ! water flows. The virtual mass force is generated by the relative
F drag;q , F lift;q , F wl;q , F td;q and F v m;q are the drag force, lift force, wall
acceleration between phases and is described as follows
lubrication force, turbulent dispersion force and virtual mass force,
!
respectively. Each one of these forces is associated with an interfa- ! !
! dq v q dq v p
cial momentum transfer mechanism. The interfacial drag force per F v m;q ¼ 0:5qq ap  ð17Þ
! dt dt
unit of volume F drag;q is calculated as
!  
! ! where the term dq/dt stands to the phase material time derivative as
F drag;q ¼ K pq v p  v q ð9Þ
dq / @/ !
where the coefficient of exchange K pq is usually applied for liquid– ¼ þ ð v q  rÞ/ ð18Þ
dt @t
liquid, bubbly and gas–liquid mixtures. The coefficient of exchange
The energy conservation equation for phase q is given by
can be expressed as function of the relaxation time of particulate sp
@    
and drag factor f as
aq qq hq þ r  aq qq ! v q hq
@t
aq ap qp f  
K pq ¼ ð10Þ @P !
sp ¼ aq  r q_ q þ qpq þ m _ pq hp  m
_ pq hq ð19Þ
@t
with !
where h is the specific enthalpy, q_ q is the heat flux vector and qpq is
q 2
p Db the energy exchange term between the different phases.
sp ¼ ð11Þ
lq The boiling model utilized in this work was the mechanistic
approach named RPI model developed by Kurul and Podowski
and (1990). This approach idealizes the boiling phenomenon into the
C drag Re temporally and spatially average effects such as bubble nucleation,
f ¼ ð12Þ growth, departure frequency and waiting time. The RPI model does
24
not incorporate the influence of sliding and coalescing bubbles on
The drag factor f is obtained based on the relative Reynolds wall heat transfer as Basu et al. (2005) and Gilman and Baglietto
number Re as proposed by Schiller and Naumann (1935), while (2013) models.
the drag coefficient C drag for bubbly flows is calculated based on The RPI model separates the total heat flux from a wall to liquid
the Ishii and Zuber correlation (Ishii and Zuber, 1979) as follows into three components as
(  
24 1 þ 0:1Re0:687 =Re Re 6 1000 q_ w ¼ q_ l þ q_ q þ q_ e ð20Þ
C drag ¼ ð13Þ
0:44 Re P 1000
where q_ l , q_ q and q_ e are liquid convective heat flux, the quenching
The lift force is caused by interactions between the bubble and heat flux, and the evaporative heat flux, respectively. Under bubbly
the shear stress profile of the liquid phase. The general expression two-phase flow, the wall is subdivided into portion u (0 6 u 6 1),
can be presented as covered by nucleating bubbles, and portion (1  u), covered by
!   fluid. Moreover, the convective heat flux q_ l is characterized by heat
! ! !
F lift;q ¼ C lift qq ap v q  v p  ðr  v q Þ ð14Þ
being transferred to the liquid phase outside the zone of influence
where the lift coefficient C lift is obtained via the model proposed by of the bubbles and is expressed as
 
Tomiyama et al. (2002), created for Couette flows where the lift q_ l ¼ Hl T w  T cell ð1  uÞ ð21Þ
l
coefficient is represented as function of the bubble size.
34 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37

Here Hl is the single-phase heat transfer coefficient for the liquid 2.3. Numerical treatment
phase and is obtained according to the flow regime. The area of
influence u can be defined on the basis of the bubble departure Discretization of the partial differential equations presented in
diameter and nucleate site density through the following the model discussion is based on the conservative Finite Volume
expression: Method (FVM). For the spatial discretization, the First Order
 p Upwind scheme was applied for all equations. The momentum
u ¼ min 1; C u Nw D2b ð22Þ conservation equations were solved in a coupled manner, without
4
the use of pressure–velocity numerical schemes. A steady state
The area of influence coefficient C u is obtained empirically via condition was set for all simulation cases and solution was consid-
the Del Valle and Kenning correlation (Del Valle and Kenning, ered converged when equation residuals dropped below 104.
1985), which is defined as

C u ¼ 4:8eð80Þ
Ja
ð23Þ 3. Results and discussion
where Ja is the subcooled Jacob number, which is obtained as
In order to ensure grid-independent solutions, meshes with 10,
q cp;l DT sub 20 and 60 k volumes were created. The chosen conditions for this
Ja ¼ l ð24Þ
qv L analysis were a saturation pressure of 4.5 MPa and a heat flux of
570 kW/m2. Fig. 2 presents the vapor volume fraction as function
The nucleate site density is represented by Lemmert and
of vertical location along the channel, considering all computa-
Chawla (1977) correlation, as function of the wall superheat. Thus,
tional meshes. As can be seen, all cases resulted in almost identical
Nw ¼ C N ðDT sup Þm ð25Þ solutions, and thus a mesh independent solution can be achieved
with coarser meshes. Consequently, for the remaining cases pre-
where by default, C N ¼ 210 and m ¼ 1:805. Many authors such as sented in this study, the mesh with 20 k volumes was applied. It
Koncar et al. (2004) have enhanced the Lemmert–Chawla (Gilman is important to mention that meshes with aspect ratio (width/
and Baglietto, 2013) correlation by tuning the constants C N and m. length) different from 1 caused a great difficulty in obtaining the
However, the implementation of new constants is only possible convergence of the final solution.
via user-defined functions in FLUENT 14.5 (ANSYS, 2013a). The estimate of void fraction along the vertical position of the
The quenching heat flux q_ q represents additional energy trans- channel, represented as the thermodynamic quality is presented
fer related to liquid filling the wall vicinity after the bubble detach- in Figs. 3a–c, for the saturation pressures of 4.5, 3.0 and 1.5 MPa,
ment and is defined as respectively, while the value of 380 kW/m2 was set as the heat flux.
 0:5   According to the results shown in Figs. 3a–c, it can be seen an
q_ q ¼ 2p0:5 u C q f b jl ql cp;l T w  T cell
l ð26Þ
increase of the agreement between the numerical calculation and
The variable f b is defined as the bubble departure frequency, the experimental data of Bartolomei and Chanturiya (1967) with
whereas jl , ql and cp;l are the thermal conductivity, density and the increase of saturation pressure. Thus, the solution where a
pressure of 4.5 MPa was applied to the system resulted in the best
specific heat of liquid phase, respectively. The factor C q is the
void fraction prediction along the channel, including the position
quenching factor used to regulate the bubble departure frequency
of the ONB. Other results with lower saturation pressures tend to
and assumes a default value of 1.0. The bubble departure frequency
underestimate the void fraction and increase the point where the
is calculated based on the inertia controlled growth proposed by
ONB occurs.
Cole (1960). Thus,
Additionally, ONB is positioned at lower thermodynamic quali-
 0:5
4gðql  qv Þ ties when higher saturation pressures are applied. For the evalua-
fb ¼ ð27Þ
3ql Db tion of the vapor void fraction at the channel exit section, it can be
noticed an overestimated prediction of the calculated void fraction
This expression was obtained through the force balance of a for all simulating cases. It is important to mention that the study of
bubble rising upwards at terminal velocity in a liquid pool. Bartolomei and Chanturiya (1967) did not show an uncertainty
Although this situation has little connection with the expected fre- analysis, so it not possible to verify if the numerical data fits exper-
quency at low qualities under forced convection, it has been used
as the standard correlation for the prediction of the departure fre-
quency under pool and flow boiling conditions. The evaporative
heat flux is the net heat to form vapor phase and is given by
p
q_ e ¼ D3b f b Nw qv L ð28Þ
6
Several other interaction sub-models could be considered in the
mathematical model. The sub-models applied in this study were
previously evaluated and recommended by Li et al. (2011) for the
prediction of boiling and critical heat flux.
The mixture turbulence model is the default multiphase turbu-
lence model in FLUENT 14.5 and was applied for all simulation
cases. In this case, the use of mixture properties and mixture veloc-
ity is satisfactory to capture the most important aspects of the tur-
bulent flow. As suggested by Li et al. (2011), the SST k  x model
was chosen as the turbulence model. For the SST k  x model,
the wall estimate method is scaled with near-wall grids, while
the wall treatment approach of all other turbulence models is cho-
sen in advance which may not be suitable for the flow develop-
ment along the channel. Fig. 2. Mesh dependency check (4.5 MPa and 570 kW/m2).
 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37 35

Results have shown that, in order to ensure a good prediction of

subcooled flow boiling characteristics, the present model should be
improved. As a manner to enhance the void fraction agreement, the
factors C ht and C q applied to the interfacial heat transfer coefficient
and the bubble departure frequency, respectively, will be varied,
providing a proper tune of boiling and condensation processes.
Fig. 4 presents the bulk liquid temperature along the channel,
represented as a function of the thermodynamic quality, for a sat-
uration pressure of 4.5 MPa and heat flux of 570 kW/m2. As
expected, a good agreement with experimental data was obtained,
showing that energy conservation law has been respected under
FVM. Hence, the onset of saturated nucleate boiling is well pre-
dicted by the proposed CFD model.
Additionally, the liquid temperature in Fig. 4 presented a very
similar linear shape for negative thermodynamic quality. At the
region close to zero quality, the variation of the liquid temperature
with the quality tends to decrease, proving that saturated boiling is
the heat transfer regime from that point.
Fig. 3a. Vapor volume fraction versus quality along the channel (4.5 MPa and The wall temperature for different thermodynamic qualities is
380 kW/m2). shown in Fig. 5. As shown, the wall temperature abruptly increased
close to channel inlet, reaching the critical superheating for bubble
nucleation. Then, wall temperature starts to decrease due to
amount of generated vapor in contact with the pipe wall, providing
an increase of the flow velocity and, consequently, the heat transfer
coefficient.
The comparison between experimental and numerical data
showed a good agreement, presenting a difference of temperature
of 2 K between experimental and numerical temperature peak.
Also, the position where the peak of wall temperature occurs
shows that numerical modeling tends to present ONB at higher
thermodynamic qualities when compared to Bartolomei and
Chanturiya (1967).
Furthermore, the decrease of the temperature found in the
downstream region of the channel shows that the RPI heat flux
partitioning model enhances the wall temperature behavior when
compared with widespread computational codes such as COBRA-IV
(Stewart, 1977) and RELAP5 (RELAP5, 2015), where this effect is
not revealed.
The vapor volume fraction as a function of the thermodynamic
quality can be seen in Fig. 6, for different C ht . As mentioned, C ht is a
Fig. 3b. Vapor volume fraction versus quality along the channel (3.0 MPa and
380 kW/m2). factor applied in the calculation of the interfacial heat transfer
coefficient. It can be concluded from Fig. 6 that the numerical cal-
culation resulted in better void fraction predictions when the fac-
tor C ht presents lower values. The decrease of C ht reduces directly
the interfacial heat transfer coefficient and, consequently, it ham-

Fig. 3c. Vapor volume fraction versus quality along the channel (1.5 MPa and
380 kW/m2).

imental error bars. Also, it is important to mention that the results

presented in Figs. 3a–c did not reflect the same results presented
by Li et al. (2011), where the void fraction was predicted with min-
Fig. 4. Bulk liquid temperature versus quality along the channel (4.5 MPa and
imal deviations. 570 kW/m2).
36 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37

Fig. 5. Wall temperature versus quality along the channel (4.5 MPa and 570 Fig. 7. Void fraction versus quality along the channel, using different C q factors
kW/m2). (4.5 MPa and 570 kW/m2).

pers the condensation process of detached bubbles that flows to 4. Summary and conclusions
the center of the pipe. Thus, for the same quality, less steam is con-
densed and higher void fraction values are obtained. Since the A two-fluid Eulerian multiphase model was used in order to
numerical model tends to underestimate the void fraction, the verify and validate the subcooled flow boiling phenomenon. The
decrease of C ht enhances the prediction of void fraction, as shown model setup for the present case was described and results were
in Fig. 6. compared with the experimental data of Bartolomei and
However, it can be also concluded by Fig. 6 that lower C ht fac- Chanturiya (1967). According to our results, the use of CFD can
tors resulted in worse estimate of the void fraction at the channel be considered as a promising numerical tool for the characteriza-
outlet. Hence, the default value C ht ¼ 1 is indicated as the best fac- tion of subcooled flow boiling. Results indicated that the onset of
tor value for the void fraction prediction at the exit of the channel. nucleate boiling is well predicted for high saturation pressures.
The influence of the bubble departure factor C q in the void frac- Additionally, the onset of saturated boiling can be calculated with
tion estimate is presented in Fig. 7, along the channel position. As a satisfactory agreement with experimental data. Difference of 2 K
can be seen, the decrease of C q provided a better agreement with was achieved between the peaks of calculated and measured wall
the experimental data. Likewise C ht , a decrease of the quenching temperature.
factor C q allows an increase of void fraction, since it raises the wall Furthermore, the void fraction calculated via the EMM model
temperature due to less heat exchange provided by bubble detach- presented lower values when compared to the experimental data.
ment. Thus, more steam is produced due the increase of wall tem- With the intention to improve the void fraction prediction along
perature superheat. Thus, lower values of C q can be applied in the the heated channel, two factors available at the model setup were
CFD model in order to improve the void fraction prediction. adjusted. These factors consist of coefficients applied in the calcu-
On the other hand, as shown by Fig. 7, the void fraction at the lation of the interfacial heat transfer and bubble departure fre-
channel outlet is better predicted when the default value of the quency. The decrease of both factors provided an enhancement
bubble departure factor C q is applied. of the void fraction prediction. However, the best agreement with
the experimental data at the channel outlet was obtained using the
default factor values.

Acknowledgement

The strong support from the Brazilian Air Force is specially rec-
ognized. The authors are also thankful to Roberto D.M. Garcia for
his suggestions and comments.

References

Anglart, H., Nylund, O., Kurul, N., Podowski, M.Z., 1997. CFD prediction of flow and
phase distribution in fuel assemblies with spacers. Nucl. Eng. Des. 177, 215–
228.
ANSYS, 2013a. ANSYS FLUENT 14.5, Documentation.
ANSYS, 2013b. Multiphase Flow, Theory Guide, Documentation.
Antal, S.P., Lahey, R.T., Flaherty, J.E., 1991. Analysis of phase distribution in fully
developed laminar bubbly two-phase flow. Int. J. Multiph. Flow 7, 635–652.
Bartolomei, G.G., Chanturiya, V.M., 1967. Experimental study of true void fraction
when boiling subcooled water in vertical tubes. Therm. Eng. 14, 123–128.
Basu, N., Warrier, G.R., Dhir, V.K., 2005. Wall heat flux partitioning during subcooled
flow boiling: Part 1 – Model development. J. Heat Transfer 127, 131–140.
Bestion, D., 2012. Applicability of two-phase CFD to nuclear reactor thermal
Fig. 6. Void fraction versus quality along the channel, using different C ht factors hydraulics and elaboration of best practice guidelines. Nucl. Eng. Des. 253, 311–
(4.5 MPa and 570 kW/m2). 321.
 F.A. Braz Filho et al. / Nuclear Engineering and Design 308 (2016) 30–37 37

Bestion, D., 2014. The difficult challenge of a two-phase CFD modelling. Nucl. Eng. Kurul, N., Podowski, M.Z., 1990. Multidimensional effects in forced convection
Des. 279, 116–125. subcooled boiling. In: 9th Int. Heat Trans. Conf., pp. 21–26, Jerusalem.
Braz Filho, F.A., Caldeira, A.D., Borges, E.M., 2011. Validation of a multidimensional Lemmert, M., Chawla, J.M., 1977. Influence of flow velocity on surface boiling heat
computational fluid dynamics model for subcooled flow boiling analysis. In: transfer. In: Heat Transfer in Boiling. Academic Press and Hemisphere, pp. 237–
International Nuclear Atlantic Conference - INAC2011, Belo Horizonte. 247.
Braz Filho, F.A., Caldeira, A.D., Borges, E.M., 2013. A sensitivity analysis of the mass Li, H., Vasquez, S.A., Punekar, H., Muralikrishnan, R., 2011. Prediction of boiling and
balance equation terms in subcooled flow boiling. In: International Nuclear critical heat flux using an Eulerian multiphase boiling model. In: Proceedings of
Atlantic Conference - INAC2013, Recife. the ASME 2011 International Mechanical Engineering Congress & Exposition –
Cole, R., 1960. A photographic study of pool boiling in the region of the critical heat IMECE2011, Denver.
flux. AIChE J. 6, 533–538. Lopez-de-Bertodano, M.A., 1992. Turbulent Bubbly Two-phase Flow in a Triangular
Collier, J.G., Thome, J.R., 1994. Convective Boiling and Condensation, third ed. Duct (Ph.D. thesis). Rensselaer Polytechnic Institute.
Oxford Science Publications, Oxford. Lopez-de-Bertodano, M.A., Prabhudharwadkar, D., 2010. CFD two fluid model for
Del Valle, M.V.H., Kenning, D.B.R., 1985. Subcooled flow boiling at high heat flux. Int. adiabatic and boiling bubbly flow in ducts. In: Woo Oh, Hyoung (Ed.),
J. Heat Mass Transf. 28, 1907–1920. Computational Fluid Dynamics. InTech, ISBN: 978-953-7619-59-6.
Dhir, V.K., Jia, W., Maity, S., Bai, Q., Basu, N., Li, D., Warrier, G.R., 2003. Complete Nemitallah, M.A., Habib, M.A., Mansour, R.B., El Nakla, M., 2015. Numerical
numerical simulation of subcooled flow boiling in the presence of thermal and predictions of flow boiling characteristics: current status, model setup and
chemical interactions. Report DOE Grant No: DE-FG03-99SF-21930, Los CFD modelling for different non-uniform heating profiles. Appl. Therm. Eng. 75,
Angeles. 451–460.
Erfeng, C., Yanzhong, L., Xianghua, C., 2009. CFD simulation of upward subcooled Prodanovic, V., Fraser, D., Salcudean, M., 2002. Bubble behavior in subcooled flow
boiling flow of refrigerant-113 using the two-fluid model. Appl. Therm. Eng. 29, boiling of water at low pressures and low flow rates. Int. J. Multiph. Flow 28, 1–
2508–2517. 19.
Garnier, J., Manon, E., Cubizolles, G., 2001. Local measurements on flow boiling of Ranz, W.E., Marshall, W.R., 1952. Evaporation of droplets: Part I and II. Chem. Eng.
refrigerant 12 in vertical tubes. Multiphase Sci. Tech. 13, 1–111. Progress 48, 141–148.
Gilman, L., Baglietto, E., 2013. Advancements on wall boiling modeling in CFD: RELAP5, 2015. http://www.relap.com/info/.
Leveraging new understanding from MIT flow boiling facility. In: NURETH-15, Schiller, L., Naumann, Z., 1935. A drag coefficient correlation. Ver. Deutsch. Ing., 77–
Pisa, Italy. 318
Ishii, M., Zuber, N., 1979. Drag coefficient and relative velocity in bubbly, droplet or Shaver, D., Antal, S., Podowski, M., 2013. Modeling and analysis of interfacial heat
particulate flows. AIChE J. 25, 843–855. transfer phenomena in subcooled boiling along PWR coolant channels. In:
Koncar, B., Tiselj, I., 2010. Influence of near-wall modelling on boiling flow NURETH-15, Pisa, Italy.
simulation. Nucl. Eng. Des. 240, 275–283. Stewart, C.W., 1977. COBRA-IV: The model and the method: BNWL 2214 NRC-4.
Koncar, B., Kljenak, I., Mavko, B., 2004. Modelling of local two-phase flow Tolubinski, V.I., Kostanchuk, D.M., 1970. Vapor bubbles growth rate and heat
parameters in upward subcooled flow boiling at low pressure. Int. J. Heat transfer intensity at subcooled water boiling. In: 4th International Heat Transfer
Mass Transf. 47, 1499–1513. Conference, Paris.
Koncar, B., Kljenak, I., Mavko, B., 2004. Modelling of local two-phase parameters in Tomiyama, A., Tamai, H., Zun, I., Hosokawa, S., 2002. Transverse migration of single
upward subcooled flow boiling at low pressure. Int. J. Heat Mass Transf. 47, bubbles in simple shear flows. Chem. Eng. Sci. 57, 1849–1858.
1499–1513. Troshko, A.A., Hassan, Y.A., 2001. A two-equation turbulence model of turbulent
Krepper, E., Prasser, H.-M., 2000. Measurements and CFD simulations of bubbly flow bubbly flows. Int. J. Multiph. Flow 27, 1965–2000.
in a vertical pipe. AMIFESF Workshop, Computing Methods in Two-phase Flow, Warrier, G.R., Basu, N., Dhir, V.K., 2002. Interfacial heat transfer during subcooled
pp. 1–8. flow boiling. Int. J. Heat Mass Transf. 45, 3947–3959.
Krepper, E., Koncar, B., Egorov, Y., 2007. CFD modelling of subcooled boiling – Zhang, R., Cong, T., Tian, W., Qiu, S., Su, G., 2015. Prediction of CHF in vertical heated
Concept, validation and application to fuel assembly design. Nucl. Eng. Des. 237, tubes based on CFD methodology. Prog. Nucl. Energy 78, 196–200.
716–731.