International Journal of Heat and Mass Transfer 180 (2021) 121769

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer

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

Drift-flux modeling of void fraction for boiling two-phase flow in a

tight-lattice rod bundle
Takashi Hibiki a,∗, Shuichiro Miwa b,c,∗∗, Kenichi Katono d
a
Department of Mechanical Engineering, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon Tong, Hong Kong
b
Division of Applied Quantum Science and Engineering, Graduate School of Engineering, Hokkaido University, North 13, West 8, Kita-ku, Sapporo 060-8628,
Japan
c
Nuclear Professional School, Faculty of Engineering, The University of Tokyo, 2-22 Shirakata, Tokai-mura 319-1188, Japan
d
Hitachi-GE Nuclear Energy, Ltd. 3-1-1 Saiwai-cho, Hitachi, Ibaraki 317-0073, Japan

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

Article history: The concept of a resource-renewal boiling water reactor (RBWR) that uses spent nuclear fuel as a nuclear
Received 24 June 2021 fuel has been proposed to reduce the period required for the radiotoxicity of spent nuclear fuel decaying
Revised 19 July 2021
to the same level as natural uranium ore from about 10 0,0 0 0 years to about 300 years. The RBWR core
Accepted 20 July 2021
is designed based on the tight-lattice core concept to reduce the ratio of a moderator to fuel. Since the
void fraction is one of the most critical design parameters for the RBWR, this study developed the drift-
Keywords: flux correlation applicable to two-phase flow in the tight-lattice rod bundle. The prediction error of the
Drift-flux model newly developed drift-flux correlation is ±11.7 %. The validated rod bundle geometrical conditions are
Rod bundle the number of rods from 4 to 37, the rod spacing from 1.0 to 2.0 mm, the rod diameter from 9.0 to
Tight lattice core
13.7 mm, the mixture volumetric flux from 0.11 to 211 m/s, and the pressure from 0.1 to 7.2 MPa. The
Void fraction
comparison in the distribution parameter between the tight-lattice rod bundle and the conventional BWR
Two-phase flow
rod bundle indicates that the distribution parameter for the tight-lattice rod bundle is larger than that
for the conventional BWR rod bundle. The increased distribution parameter in the tight-lattice core may
be due to the small rod spacing and the triangular array of the rods. Both of the small rod spacing and
the triangular array of the rods tend to increase the distribution parameter.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction technology, plant systems will follow the proven technologies of

current BWRs adopted in pressure vessels, containment vessels,
In connection with the development of next-generation light reactor systems, and turbine systems. Reactor core components
water reactors that effectively utilize uranium resources and need to be modified to improve core neutronic characteristics. The
improve economic efficiency, high conversion BWRs have been neutron resonance absorption by uranium 238 in the resonance
proposed with the focus on saving natural uranium resources by energy region (1–1 keV) should be enhanced to improve the
improving the conversion efficiency of plutonium. Yamashita et al. conversion ratio in BWR, and the neutron deceleration by the
[1] discussed the importance of the high conversion BWRs as moderator should be reduced to increase the conversion ratio to
follows. The conversion ratio is defined as the ratio between the plutonium. The amount of the moderator to the fuel should be
amount of fissile material at the time of fuel removal after burnup reduced to meet these requirements. It is necessary to consider
and the amount of fissile material at the fuel loading. The current several design changes, such as the use of tight-lattice core to
BWR has a conversion ratio of about 0.5. Because high conversion reduce the fuel rod spacing, change of the fuel rod arrangement
BWRs are positioned as an extension of existing light water reactor from a square lattice to a triangular lattice, and reduction of the
density of core moderator with use of boiling two-phase flow. It
is necessary to make the effective water-to-fuel volume ratio in
∗
Corresponding author at: Department of Mechanical Engineering, City Univer- consideration of voids about 0.53 to obtain a conversion ratio of
sity of Hong Kong, 83 Tat Chee Avenue, Kowloon Tong, Hong Kong
∗∗
about 0.9. To achieve this goal in the absence of voids, the spacing
Corresponding author at: Division of Applied Quantum Science and Engineer-
ing, Graduate School of Engineering, Hokkaido University, North 13, West 8, Kita-ku, between fuel rods must be about 1 mm, but if there are voids,
Sapporo 060-8628, Japan. it can be about 2 mm. The presence of voids means that a high
E-mail addresses: [email protected], [email protected] (T. Hibiki), conversion ratio can be obtained without extremely reducing the
[email protected] (S. Miwa).

https://doi.org/10.1016/j.ijheatmasstransfer.2021.121769
0017-9310/© 2021 Elsevier Ltd. All rights reserved.
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

RBWR is characterized by the use of highly toxic transuranium

Nomenclature elements such as plutonium in spent nuclear fuel.
Accurate prediction of two-phase flow characteristics is cru-
Latin characters cial for nuclear thermal-hydraulic safety analyses [3]. Adopting the
b parameter [-] tight-lattice core concept, nuclear, thermal, and safety engineer-
CD drag coefficient [-] ing issues arise. Due to the narrow gap between the fuel rods,
C0 distribution parameter [-] it is necessary to consider the effect of the tight-lattice core on
C∞ asymptotic value of distribution parameter [-] thermos-fluid parameters, such as void fraction, critical power, and
C∞H high-flux parameter [-] pressure loss. Since the equivalent diameter of the coolant chan-
C∞L low-flux parameter [-] nel is extremely smaller than the conventional one, it is consid-
DH hydraulic equivalent diameter of flow channel [m] ered that the nuclear boiling limit (DNB) tends to be more severe
DSm bubble Sauter mean diameter [m] than the conventional light water reactor. Japan Atomic Energy
D0 rod diameter [m] Agency conducted a series of tests to obtain the data of void frac-
E average prediction error tion, critical power, and pressure loss for Innovative Water Reac-
F function [-] tor for Flexible Fuel Cycle [4–6]. MIT performed a series of tests to
G mass flux [kg/m2 s] demonstrate the feasibility of RBWR [7]. Yan reviewed the thermal-
g gravitational acceleration [m/s2 ] hydraulic phenomenon in tight-lattice bundles, including flow and
j mixture volumetric flux [m/s] heat transfer, critical heat flux, turbulent mixing, and vortex struc-
j∞ max critical mixture volumetric flux [m/s] ture [8].
jg superficial gas velocity [m/s] In view of the importance of understanding the thermal-
m parameter [-] hydraulic phenomena in a tight-lattice core, this study focuses on
P pressure [Pa] a void fraction in the tight-lattice channel. This study will review
rb bubble radius (or drag radius) [m] and collect existing drift-flux correlations and existing void fraction
Nμ f viscous number [-] data taken in the tight-lattice channel, and develop a drift-flux cor-
s gap [m] relation applicable to the tight-lattice channel.
Vg j mean drift velocity [m/s]
vg gas velocity [m/s] 2. Existing drift-flux correlations for rod bundle and data
vg j drift velocity [m/s] taken in tight-lattice channels
vr relative velocity between gas and liquid phases
[m/s] 2.1. Drift-flux model
x thermal-equilibrium quality [-]

Greek symbols The drift-flux model was developed to consider the relative ve-
α void fraction [-] locity between gas and liquid phases, vr , in two-phase flow analy-
αcrit critical void fraction [-] ses [9]. One-dimensional form of the drift-flux model is expressed
ρ density difference between gas and liquid phases as:
[kg/m3 ]  jg   
μf vg  = = C0  j + vg j , (1)
liquid viscosity [Pa•s] α
ρf liquid density [kg/m3 ]
where vg , jg , α , and j are the gas velocity, superficial gas veloc-
ρg gas density [kg/m3 ]
ity, void fraction, and mixture volumetric flux, respectively.  and
σ surface tension [N/m]
 indicate the area-averaged and void fraction-weighted mean
Subscripts values, respectively. The distribution parameter, C0 , and the void
+ non-dimensionalized quantity fraction-weighted mean drift velocity (hereafter, drift velocity for
simplicity), vg j , are defined as Eqs. (2) and (3).
Subscripts
B bubbly flow condition α j
C0 ≡ , (2)
cal. calculated value α j
exp . experimental value
and
P pool condition  
  αvg j
Operators vg j ≡ , (3)
•  area-averaged quantity α
• void fraction weighted mean quantity where

vg j ≡ vg − j = (1 − α )vr . (4)
The physical meaning of the distribution parameter is a "covari-
ance" due to non-uniform distributions of void fraction and mix-
gap, which is advantageous in terms of fuel cooling characteristics,
ture volumetric flux, and a well-adapted form of the distribution
mechanical characteristics, and manufacturing. There are cases
parameter is expressed as:
where the fuel rods are arranged in a triangular lattice shape in a

fuel assembly having a square cross-section, and cases where the ρg
C0 = C∞ − (C∞ − 1 ) , (5)
fuel rods are a hexagonal lattice shape fuel assembly. Using the ρf
tight-lattice core concept, a resource-renewal boiling water reactor
(RBWR) that uses spent nuclear fuel as a nuclear fuel has been where C∞ , ρg , and ρ f are the asymptotic value of the distribution
proposed [2]. It takes about 10 0,0 0 0 years for the radiotoxicity of parameter, gas density, and liquid density, respectively. The form
spent nuclear fuel to decay to the same level as natural uranium of Eq. (5) scales the inertia effect on the distribution parameter,
ore, but it can be reduced to about 300 years using RBWR. The and the asymptotic value of the distribution parameter approaches

2
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

unity when the density ratio approaches unity. The distribution pa- Churn flow [16]:
rameter depends on channel geometry, flow regime, and channel 
ρg
size, and is given empirically. C0 = 1.2 − 0.2 , (11)
The physical meaning of the drift-velocity is the difference be- ρf
tween gas velocity and mixture volumetric flux. The relative veloc-  
ity for a dispersed two-phase flow is expressed by [10]:
  √ ρ gσ 0.25
vg j = 2 . (12)
ρ 2f
8 rb
vr |vr | = ρ g(1 − α ), (6)
3 CD ρ f Annular flow [16]:
  
where rb , ρ , g, and CD are the bubble radius, density difference 1 − α ρ gDH (1 − α )
between gas and liquid phases, gravitational acceleration, and drag Vg j =  j + , (13)
α + 4 ρg /ρ f 0.015ρ f
coefficient. Considering Eqs. (3) and (4), the drift velocity can be
derived from a drag law. where the mean drift velocity is defined by:
The drift-flux parameters, such as the distribution parameter
Vg j ≡ vg  −  j. (14)
and drift velocity, depend on channel geometry, flow regime, chan-
nel size, and channel orientation [11–14]. For example, the drift-
2.2. Recent R & D activities in developing a drift-flux correlation in a
flux parameters applicable to upward two-phase flow in a vertical
rod bundle
medium-sized circular channel are shown as follows.
Bubbly flow [15]:
Extensive efforts have been conducted to develop a drift-flux
    
ρg DSm  correlation in a rod bundle. Coddington and Macian [17] evaluated
C0 = 1.2 − 0.2 1 − exp −22 , (7)
ρf DH the performance of drift-flux correlations developed for a rod bun-
dle. The following review focuses on the recent R & D activities in
  developing a drift-flux correlation in a rod bundle.
  √ ρ gσ 0.25
vg j = 2 (1 − α )1.75 , (8) Drift-flux correlation for prototypic 8 × 8 rod bundle under
ρ 2f prototypic temperature and pressure conditions [18,19]
The distribution parameter and drift velocity are given by
where DSm , DH , and σ are the bubble Sauter mean diameter, hy- Eqs. (15) and (16), respectively.
draulic equivalent parameter, and surface tension, respectively. It 
should be noted here that Eq. (7) is applicable to two-component ρg
C0 = 1.1 − 0.1 , (15)
bubbly flow. ρf
Slug flow [16]:
  +   + 
   
 
ρg vg j = vg j,B exp −1.39 jg+ + v+g j,P 1 − exp −1.39 jg+ ,
C0 = 1.2 − 0.2 , (9)
ρf (16)

  ρ gDH where
 +  √
vg j = 0.35 . (10)
vg j,B = 2(1 − α )1.75 ,
ρf (17)

Fig. 1. Schematic diagram of rod bundle simulating core of Marviken BHWR type reactor [20].

3
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

⎧ Drift-flux correlation for 8 × 8 rod bundle under low pressure

⎪whenNμ f ≤ 2.25 × 10−3 and low flow conditions [14]
⎪
⎪
⎪
0.809  ρg −0.157 −0.562 The distribution parameter is given by Eq. (5) with the follow-
⎪
⎪0.0019 D+ Nμ f forD+ ≤ 30
⎪
⎨ H
 −0.157
ρf H ing asymptotic value of the distribution parameter.
 +  ρg 
vg j,P = 0.030 ρf N −0.562 forD+ ≥ 30 . (18) C∞L for j+  ≤  j+ C∞ max
⎪
⎪
μf H
C∞ = , (23)
⎪
⎪ ≥ . × 10−3 C∞H for j+  >  j+ C∞ max
⎪
⎪
when Nμ 2
 −0.157
f 25
⎪
⎩0.92 ρg

ρf forD+ ≥ 30
H C∞H = 1.1 + 1.84e−0.1 j  · F ,
+
(24)
Here, ⎧
  ⎪ 1
 +  vg j ⎨   
vg j ≡  0.25 , (19) F = min 1.70 − 582 ρg
ρ
, (25)
ρ gσ
ρ 2f
⎪
⎩max
f

0
 +  jg  
 
jg ≡  0.25 , (20) C∞H  j+ C∞ max − 1
 + 
ρ gσ C∞L =   jg + 1, (26)
ρ 2f  j+ C∞ max − j+f
D+
DH  +  
H ≡ σ
, (21)
j = m j+f + b, (27)
ρ g C∞ max

μf
Nμ f ≡
σ 0 . 5 , (22) 1
ρf σ m= , (28)
ρ g 1 − αcrit C0

Fig. 2. Schematic diagram of rod bundle test sections. (a) 4-rod bundle (or sub-channel) test section, (b) 7-rod bundle test section, (c) 14-rod bundle test section, (d) 19-rod
bundle test section for air-water flows, (e) 19-rod test section for steam-water flows (f) 37-rod bundle test section.

4
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

Fig. 2. Continued



  
αcrit
Vg+j αcrit = min 0.0284 j+f + 0.125, 0.52 , (32)
b= , (29)
1 − αcrit C0 The drift velocity is given by Eq. (16).
Drift-flux correlation for rod bundle simulating the core of Mar-
 +  j viken BHWR type reactor [18]
j ≡  0.25 , (30)
ρ gσ The rod bundle geometry is shown in Fig. 1 [20]. The distribu-
ρ 2f tion parameter is given by:
  
ρg
 + jf C0 = 1.03 − 0.03 , (33)
jf ≡  0.25 , (31) ρf
ρ gσ
ρ 2f The drift velocity is given by Eq. (16).

5
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

Fig. 2. Continued

2.3. Brief discussion of channel size effect on the drift-flux size given by D+ H
= 30. The critical size is empirically determined
correlation and may range from 30 to 52 [21]. The increased drift velocity is
due to the cap bubble and secondary flow created in a large-sized
A drift-flux correlation is classified into three categories based channel [22]. In bubbly flow, the distribution parameter may not
on a channel size: a mini channel, a medium-sized channel, and a be represented by a single line in a drift-flux plot or a constant
large-sized channel. The flow characteristics in a mini channel are value. The distribution parameter increases with an increased ra-
summarized as follows. When the channel size decreases, the drift tio of superficial gas velocity to mixture volumetric flux, reaches
velocity tends to decrease due to the enhanced surface tension the maximum value, and decreases towards the value in churn
force. The decrease of the drift velocity starts at the channel di- flow [22].
ameter of 4 mm for an atmospheric air-water flow [11], which cor- In a tight-lattice core, three geometrical length scales co-exist.
responds to the non-dimensionalized hydraulic equivalent diame- They are rod spacing (or rod gap), hydraulic equivalent diameter,
ter being 1.5, D+H
= 1.5. When the channel size decreases, the dis- and channel box size. The rod spacing may be categorized as a
tribution parameter increases due to a more concentrated void in mini channel. The hydraulic equivalent diameter may be catego-
the channel [11, 12]. The flow characteristics in a large-sized chan- rized as medium-sized channel. The channel box size may be cate-
nel are summarized as follows. When the channel size increases, gorized as a large-sized channel. The drift-flux parameters may be
the drift velocity tends to increase and level off beyond a critical affected by the competing effect of the channel size.

6
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

Fig. 3. Comparison between existing drift-flux correlation and data [6]. (a) Compar- Fig. 4. Comparison between existing drift-flux correlation and data. (a) Comparison
ison with 19-rod bundle test data under the pressure of 2.0 MPa, (b) Comparison with 37-rod bundle test data under the pressure of 2.0 MPa, (b) Comparison with
with 37-rod bundle test data under the pressure of 7.2 MPa. 37-rod bundle test data under the pressure of 7.2 MPa.

2.4. Existing data collected in flow channel simulating tight-lattice and void fraction meter. Table 1 summarizes the test condi-
core tions of these databases, including channel geometry and thermal-
hydraulic conditions. Fig. 2 summarizes the channel geometries.
Mishima et al. performed void fraction measurement in a sub-
channel simulating high conversion light water reactor using neu- 3. Evaluation of drift-flux correlations with data collected in
tron radiography [23,24]. Kureta et al. also conducted void frac- tight-lattice channels
tion measurement in 7, 14, 19, and 37-rod bundles under pres-
sures ranging from 0.1 to 7.2 MPa [4–6]. The measurements were Fig. 3 shows a comparison between the calculated and mea-
performed by neutron radiography, quick shut-off valve method, sured void fractions [6]. The datasets used in the comparison

7
 T. Hibiki, S. Miwa and K. Katono
Table 1
Geometrical information of rod bundle test sections and test conditions.

No. of Cross-Sectional P s D0 DH G x  jg   jf   j α Measurement

Rods System View [MPa] [mm] [mm] [mm] [kg/m2 s] [-] [m/s] [m/s] [m/s] [-] Method

4 AW Fig. 2(a) 0.1 2.0 9.0 4.54 NA NA NA NA 0.110-3.43 NA NR

7 AW Fig. 2(b) 0.1 1.0 12.0 3.53 NA NA 2.9-211 0.07-0.79 11.9-211 0.30-0.97 NR
7 SW Fig. 2(b) 0.1 1.0 12.0 3.53 115-461 0-0.42 0.88-78.4 0.11-0.48 4.67-76.9 0.29-0.91 NR
14 SW Fig. 2(c) 0.1 1.3 13.7 1.77 156-388 NA NA NA 0.171-31.2 NA NR
19 AW Fig. 2(d) 0.1 1.3 12.0 4.30 NA NA 0.17-7.42 0.14-3.00 0.370-7.81 0.04-0.55 NR
19 SW Fig. 2(e) 2.0 1.3 12.0 4.30 600-2000 NA NA NA 1.57-33.7 NA Void Meter
37 SW Fig. 2(f) 2.0 1.0 13.0 3.62 400-1000 NA NA NA 2.19-17.2 NA Void Meter
37 SW Fig. 2(f) 7.2 1.0 13.0 3.62 600 NA NA NA 2.87-4.42 NA Void Meter

AW: Air-water system, SW: Steam-water system.

8

nel were considered for the comparison. The distribution parame-

medium-sized circular channel and an equilateral triangular chan-
TRAC-BF1 and drift-flux correlations for slug and annular flows in a
dard 1D nuclear thermal-hydraulic system analysis code such as
2.0 MPa and 37-rod bundle under the pressure of 7.2 MPa. A stan-
were the data collected in a 19-rod bundle under the pressure of

(b) Comparison with 37-rod bundle test data under the pressure of 7.2 MPa.
[26]. (a) Comparison with 19-rod bundle test data under the pressure of 2.0 MPa,
Fig. 5. Comparison between existing drift-flux correlation for rod bundle and data

International Journal of Heat and Mass Transfer 180 (2021) 121769

T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

approaches unity. The drift velocity for the equilateral triangular

channel is the same as that for the medium-sized circular channel.
Kureta et al. [6] summarized the comparisons as follows.

• The TRAC-BF1 code and drift-flux correlations tended to over-

estimate the void fraction in the region of x < 0.1 where x is
the thermal-equilibrium quality.
• The drift-flux correlation for an annular flow showed a rea-
sonable agreement with the data in the region of x > 0.1 and
α > 0.5.

Several additional discussions may be possible.

• Some peculiar data in the region of x < 0.06 under the pres-
sure of 2.0 MPa are found. The two data are α = 0.477 at
x = 0.0576 and α = 0.547 at x = 0.0410. Although the qual-
ity decreases from 0.0576 to 0.0410, the void fraction increases
from 0.477 to 0.547. The measurement error may be expected
to be 20 % or larger.
• The drift-flux correlation for an annular flow, Eq. (13), does not
consider the droplet entrainment. If the droplet entrainment
is considered, the calculated void fraction should increase to-
ward the void fraction calculated by the TRAC-BF1 code. Since
TRAC-BF1 tends to overestimate the void fraction, the effective
droplet entrainment may be suppressed due to the accumu-
lated water layer in the rod gaps (s = 1.3 mm).
• The drift-flux correlation for an annular flow apparently agrees
with the data even in a churn flow regime with the void frac-
tion lower than 0.8. The interfacial structure observed in a mini
circular channel is different from the one in a medium-sized
circular channel [11]. In a mini circular channel, large churn
bubbles are more organized like long slug bubbles with en-
hanced liquid film wave. The flow structure of the churn flow in
the mini circular channel may be apparently similar to the an-
nular flow structure with a liquid slug, including small bubbles.
The flow structure difference between medium sized and mini
circular channels may partly explain the apparent agreement
between the drift-flux correlation for the annular flow and data
in the churn flow regime.
• The overestimation of the TRAC-BF1 drift-flux correlation im-
plies that the momentum coupling between gas and liquid
phases is underestimated. In other words, the distribution pa-
rameter or drift velocity is underestimated.

Fig. 4 also shows a comparison between the calculated and

measured void fractions [26]. The datasets used in the comparison
were the data collected in a 37-rod bundle under the pressure of
2.0 and 7.2 MPa. Twelve drift-flux correlations were considered for
the comparison. Shirvan et al. [26] summarized the comparisons
as follows:

• The test conditions used in the comparison were different from

Fig. 6. Drift-flux plots. (a) 4-rod bundle for air-water under pressure of 0.1 MPa, (b)
the prototypic operating conditions of a conventional BWR.
7-rod bundle for air-water under pressure of 0.1 MPa, (c) 7-rod bundle for steam-
water under pressure of 0.1 MPa, (d) 14-rod bundle for steam-water under pressure The mass flux (600 and 800 kg/m2 s) and equivalent hydraulic
of 0.1 MPa, (e) 19-rod bundle for air-water under pressure of 0.1 MPa, (f) 19-rod equivalent diameter were smaller than those in the conven-
bundle for steam-water under pressure of 2.0 MPa, (g) 37-rod bundle for steam- tional BWR.
water under pressure of 2.0 MPa, (h) 37-rod bundle for steam-water under pressure • Chexal-Lellouche correlation [27] implemented in the RELAP5
of 7.2 MPa.
code tended to overestimate the void fraction. This suggested
that the interfacial momentum transfer in the 37-rod bundle
with the small hydraulic diameter was different from that in a
ter for the equilateral triangular channel is given by [25]:
conventional medium-sized channel.
 • The correlation of Liao et al. [28] given by Eqs. (35) and
ρg (36) agreed with the data. However, Shirvan et al. concluded
C0 = 1.39 − 0.35 , (34)
ρf that the agreement was only a coincidence.

It should be noted here that the distribution parameter calcu- ρg
C0 = 1.2 − 0.2 {1 − exp (−18α )}, (35)
lated by Eq. (34) does not approach unity when the density ratio ρf

9
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

 0.25
  ρ gσ
vg j = 0.33 , (36)
ρ 2f

Most of the drift-flux correlations considered in the compari-

Error
son imply that the momentum coupling between gas and liquid

25.9
30.3
4.05
5.85
4.14

11.1
8.36
6.95
11.7
[%]
phases is overestimated. In other words, the distribution parame-
ter or drift velocity is underestimated.
Fig. 5 shows a comparison between the calculated and mea-
sured void fractions. The datasets used in the comparison are the

No. of Data
data collected in a 19-rod bundle under the pressure of 2.0 MPa
and 37-rod bundle under the pressure of 7.2 MPa. Drift-flux corre-

263
lations for prototypic 8 × 8 rod bundle, Eqs. (15) and (16) (here-

34
28
32
62
25
57
22
3
after, Ozaki-Hibiki correlation), and for rod bundle simulating the
core of Marviken BHWR type reactor, Eq. (33) and (16) (hereafter,
FRIGG correlation), are considered for the comparison. The data are

0.0738

0.0714
vg j 
taken from the original figure [5], and the mass velocities for the

0.231
0.222
0.222

0.605
0.605
0.605
[m/s]
data are given in the range from 600 to 20 0 0 kg/m2 s. Two calcu-
lation lines for the mass velocities of 600 and 2000 kg/m2 s are
presented in Fig. 5. The drift-flux correlations considered in the
comparison implies that the momentum coupling between gas and

vg j  Equation
liquid phases are overestimated. In other words, the distribution
parameter or drift velocity is underestimated. This comparison re-
sult implies that the drift-flux correlation validated for a conven-

Churn
Churn
Churn
tional BWR may not be applicable to the prediction of void frac-

Slug

Slug
LSC
LSC
LSC
tion in a tight-lattice channel. The distinctive characteristics of the

Experimentally determined distribution parameter, selected drift velocity correlation and prediction error for each test.
tight-lattice core are the reduced fuel rod spacing and the change
of the fuel rod arrangement from a square lattice to a triangular
lattice. Since the rod spacing and the fuel rod arrangement may

CalculationC0
significantly affect the drift-flux parameters, the drift-flux correla-
tion should be developed by considering these effects.

1.19
1.19

1.19
1.18
1.18
1.16
1.20
1.20
[-]
4. Drift-flux correlation development and its evaluation

4.1. Modeling strategy

ExperimentalC0

An often utilized method to develop a drift-flux correlation is

to determine the distribution parameter and drift velocity with the
data all together through the least-square method. The distribution
1.16
1.21
1.21
1.13
1.27
1.19
1.15
1.30

parameter and drift velocity determined by this method include a

compensation error, which means that the estimation error of the
distribution parameter propagates to that of the drift velocity, vice
versa. In this study, an alternative method is used in developing a
drift-flux correlation. In the alternative method, the drift velocity is
first calculated by a correlation, and the drift velocity is anchored
[MPa]

in the drift-flux plot, namely,  j vs. vg , and the distribution
7.2
0.1
0.1
0.1
0.1
0.1
2.0
2.0
P

parameter is next determined by the data with the least-square

method. In most cases, calculated drift velocities are of the order
of 0.5 m/s at maximum. If the true value of the drift velocity is
considered to be between 0 and 0.5 m/s, the mixture volumetric
flux is higher than 2.5 m/s, the error due to the prediction error in
System

the drift velocity is 10 %. Since a majority of the mixture volumet-

AW
AW

AW
SW
SW

SW
SW
SW

ric fluxes in the available data is higher than 2.5 m/s, the above al-
LSC: Large-sized channel.

ternative method should be appropriate for developing a drift-flux

correlation.
The drift velocity correlations used in developing a drift-flux
No. of Rods

correlation are summarized in Table 2. The drift velocity correla-

tion for slug flow, Eq. (10) is adopted to calculate the drift velocity
Table 2

in a sub-channel formed by 4 rods under atmospheric pressure,

14
19
19
37
37
4
7
7

because the test was performed in a medium void fraction region.

As shown in Fig. 6(a), the gas velocity tends to approach the drift
velocity for slug flow (=0.0738 m/s), not churn flow (=0.231 m/s).
The drift velocity correlation for churn flow, Eq. (12), is adopted
to calculate the drift velocity in 7- and 14-rod bundles under at-
mospheric pressure, because the test was also performed in a high

10
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

Fig. 7. Dependence of distribution parameter on square root of density ratio.

void fraction region. As shown in Fig. 6(b–d), the mixture volu- turbulent flow regimes. Eq. (37) may not be applicable to separated
metric flux is high enough to ignore the effect of the prediction two-phase flow, such as annular flow.
error of the drift velocity on the distribution parameter. The drift Fig. 6 shows the predictive performance of Eq. (37) with the
velocity correlation for slug flow, Eq. (10) is adopted to calculate drift velocity correlation selected for each case. In Fig. 6, the solid
the drift velocity in a 19-rod bundle under atmospheric pressure, and broken lines indicate the calculated values with Eqs. (37) and
because the test was performed in a medium void fraction region. (15), respectively. The gas velocities calculated by Eq. (37) agree
As shown in Fig. 6(e), the gas velocity tends to approach the drift with the data reasonably well, whereas the gas velocities calcu-
velocity for slug flow (=0.0714 m/s), not churn flow (=0.231 m/s). lated by Eq. (15) developed for a conventional BWR core tend to
The drift velocity correlation for a large size channel, Eq. (16) is underestimate the data.
adopted to calculate the drift velocity in a 19-rod bundle under Comparing between Eqs. (37) and (15) suggests that the dis-
the pressure of 2.0 MPa and 37-rod bundle under the pressures of tribution parameter in a tight-lattice core is higher than that in
2.0 and 7.2 MPa, because the channel box sizes exceed D+ H
= 30. a conventional BWR core. The increased distribution parameter in
the tight-lattice core may be due to the small rod spacing and the
4.2. Drift-flux correlation development triangular array of the rods. Both of the small rod spacing and the
triangular array of the rods tend to increase the distribution pa-
In Section 4.1, an appropriate drift velocity correlation for each rameter. Eq. (37) is developed based on the data collected under
tested sub-channel or rod bundle is selected. A value of the distri- wide test conditions, such as broad rod bundle sizes from sub-
bution parameter for each tested sub-channel or rod bundle is de- channel or 4-rod bundle to 37 rod-bundle and broad pressures
termined by the data with the least square method. The obtained from 0.1 to 7.2 MPa. Eq. (37) is considered to include sufficient
distribution parameters are listed in Table 2. The obtained values scalability of the rod bundle size and pressure.
for each case are plotted in Fig. 7. The asymptotic value of the dis- Table 2 lists the prediction error for each tested rod bundle. The
tribution parameter is determined to be 1.22, resulting in the fol- prediction error, E, is simply calculated by:
lowing distribution parameter correlation for the tight-lattice core.
   
  vg,cal. − vg,exp . 
ρg E [%] = × 100, (38)
C0 = 1.22 − 0.22
ρf
, (37) vg,exp . 

It should be noted here that Eq. (37) is applicable to dis- The average prediction error of gas velocity is estimated to be
persed two-phase flow such as cap-bubbly, cap-churn, and churn- ±11.7 %.

11
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

Fig. 8. Comparison between newly developed drift-flux correlation for tight-lattice Fig. 8. Continued
rod bundle and data. (a) Comparison with 7-rod bundle steam-water test data un-
der the pressure of 0.1 MPa, (b) Comparison with 14-rod bundle steam-water test
data under the pressure of 0.1 MPa, (c) Comparison with 19-rod bundle steam-
but no information on the mass flux for each data point is avail-
water test data under the pressure of 2.0 MPa, (d) Comparison with 37-rod bundle
steam-water test data under the pressure of 2.0 MPa, (e) Comparison with 37-rod able. To calculate the relationship between quality and void frac-
bundle steam-water test data under the pressure of 7.2 MPa. tion from the corresponding data presented in the drift-flux plot,
three values of the mass flux, 145, 236, and 346 kg/m2 s are as-
sumed. In Fig. 8(a), the void fractions obtained by the assumed
4.3. Evaluation of newly developed drift-flux correlation mass flux are represented by the circle, triangular, and square for
the mass flux of 145, 236, and 346 kg/m2 s. The void fractions cal-
Fig. 8 compares the newly developed drift-flux correlation with culated by the drift-flux correlation with Eq. (37) are indicated
the data in quality vs. void fraction plane. Since the data shown by thick solid, broken, and dotted lines for the mass flux of 145,
in Fig. 8 are collected from the original figures [4–6], comprehen- 236, and 346 kg/m2 s. The void fractions calculated by the drift-
sive information for each data is not available. In Fig. 8(a), the data flux correlation for annular flow, Eq. (13), are indicated by thin
were collected in the range of mass flux from 145 to 346 kg/m2 s, solid, broken, and dotted lines for the mass flux of 145, 236, and

12
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

correlations could predict the data satisfactorily. Due to the lack of

reliable drift-flux correlation for the tight-lattice rod bundle, the
drift-flux correlation was developed based on the data collected in
various rod bundle sizes from 4 rods to 37 rods and pressures from
0.1 to 7.2 MPa. The obtained results are summarized as follows.
• The comparisons between existing drift-flux correlations and
data obtained in various tight-lattice rod bundles indicated that
the momentum coupling between gas and liquid phases were
overestimated in the existing drift-flux correlations. In other
words, the distribution parameter or drift velocity was under-
estimated.
• The drift-flux correlation, Eqs. (37) and (16) were newly devel-
oped. The distribution parameter was determined by anchoring
the drift velocity using the selected drift velocity correlation in
the drift-flux plot. The newly developed drift-flux correlation
predicted the gas velocity with a prediction error of ±11.7 %.
The validated rod bundle geometrical conditions were the num-
ber of rods from 4 to 37, the rod spacing from 1.0 to 2.0 mm,
the rod diameter from 9.0 to 13.7 mm, the mixture volumetric
flux from 0.11 to 211 m/s, and the pressure from 0.1 to 7.2 MPa.
• The comparison in the distribution parameter between the
tight-lattice rod bundle and the conventional BWR rod bundle
indicated that the distribution parameter for the tight-lattice
rod bundle was larger than that for the conventional BWR rod
bundle. The increased distribution parameter in the tight-lattice
Fig. 8. Continued core might be due to the small rod spacing and the triangular
array of the rods. Both of the small rod spacing and the trian-
gular array of the rods tended to increase the distribution pa-
346 kg/m2 s. In Fig. 8(b–d), the void fraction data are directly ob-
rameter.
tained from the original figures. Each data does not include the
mass flux information, and the mass flux information is given as The current development procedure was heavily dependent on
a range of mass flux. When the void fraction is calculated by the the data collected in the Japan Atomic Energy Agency. To improve
drift-flux correlations, minimum, middle, and maximum values of the robustness and reliability of the drift-flux correlation, further
the mass flux for the experimental range of the mass flux are as- and extensive experimental researches are recommended. Concern-
sumed to produce three lines for three mass flux cases. ing nuclear accident analyses, the data under low-flow and low-
Fig. 8 indicates that the drift-flux correlation with pressure conditions should be collected in a future study.
Eq. (37) agrees with the data reasonably well in the region of
α ≤ 0.75 ∼ 0.80 where a dispersed two-phase flow appears. Declaration of Competing Interest
When the void fraction exceeds 0.9, the drift-flux correlation
for annular flow predicts the void fraction well. The region of None declared.
0.75 ∼ 0.80 ≤ α ≤ 0.9 is considered a transition region between
dispersed and separated two-phase flows. Some interpolation CRediT authorship contribution statement
scheme needs to be implemented in the transition region to
predict the void fraction. Takashi Hibiki: Formal analysis, Writing – review & editing.
In the above evaluations, the drift velocity is computed based Shuichiro Miwa: Data curation, Formal analysis. Kenichi Katono:
on the selected drift velocity correlations. When the drift-flux cor- Conceptualization, Methodology.
relation applies to a prototypic tight-lattice rod bundle, the drift
velocity correlation for a large-sized channel, Eq. (16), is recom- Acknowledgment
mended. In summary, the drift-flux correlation for a prototypic
tight-lattice rod bundle is given by Eqs. (37) and (16). The work described in this paper was partially supported by
a grant from City University of Hong Kong (Project No. 9380126).
5. Conclusion One of the authors (T. Hibiki) appreciates the support.

References
In the viewpoint of the importance of the drift-flux correlation
for the tight-lattice rod bundle, the systematic research was con- [1] J. Yamashita, S. Uchikawa, T. Mochida, Development of high conversion boiling
ducted on the drift-flux correlation development. In this study, the reactor, Hitachi Rev. 70 (4) (1988) 429–432.
basic concept of the drift-flux model was first discussed, and the [2] T. Hino, M. Ohtsuka, K. Moriya, M. Matsuura, Light water reactor system de-
signed to minimize environmental burden of radioactive waste, Hitachi Rev.
drift-flux correlations developed for upward two-phase flow in a
63 (9) (2014) 602–609.
vertical circular channel were introduced. Three drift-flux corre- [3] G.H. Yeoh, Thermal hydraulic considerations of nuclear reactor systems: past,
lations were also introduced based on the recent R & D of the present and future challenges, Exp. Comput. Multiph. Flow 1 (1) (2019) 3–27.
[4] M. Kureta, H. Yoshida, A. Ohnuki, H. Akimoto, Experimental study on void
drift-flux correlations in prototypic rod bundles of a conventional
fraction in tight-lattice rod bundles, in: Proceedingof the 10th International
BWR and a Marviken BHWR type reactor. Then, the channel size Topical Meeting on Nuclear Reactor Thermal Hydraulics (NURETH-10), 2003,
effect on the drift-flux parameters, such as the distribution param- p. B00205.
eter and drift velocity, was discussed. The extensive comparisons [5] M. KURETA, H. TAMAI, T. SATO, M. SHIBATA, A. OHNUKI, H. AKIMOTO, Devel-
opment of design technology on thermal-hydraulic performance in tight-lat-
between existing drift-flux correlations and data collected in var- tice rod bundles : vi-estimation of void fraction, Proceedingof the 15th Inter-
ious tight-lattice rod bundles identified that no existing drift-flux national Conference on Nuclear Engineering (ICONE-15), 2007 ICONE15-10116.

13
T. Hibiki, S. Miwa and K. Katono International Journal of Heat and Mass Transfer 180 (2021) 121769

[6] M. KURETA, H. TAMAI, H. YOSHIDA, A. OHNUKI, H. AKIMOTO, Development of [19] T. Ozaki, T. Hibiki, Drift-flux model for rod bundle geometry, Prog. Nucl. Energy
design technology on thermal-hydraulic performance in tight-lattice rod bun- 83 (2015) 229–247.
dles: v-estimation of void fraction, J. Power Energy Syst. 2 (1) (2008) 271–282. [20] O. Nylund, Full-scale loop studies of BHWR and BWR fuel assemblies, ASEA
[7] X. Zhao, K. Shirvan, Y. Wu, M.S. Kazimi, Critical power and void fraction pre- Res. 10 (1969) 63–125.
diction of tight bundle designs, Nucl. Technol. 196 (3) (2016) 553–567. [21] C.S. Brooks, S.S. Paranjape, B. Ozar, T. Hibiki, M. Ishii, Two-group drift-flux
[8] B.H. Yan, The thermal hydraulic phenomenon in tight lattice bundles: a review, model for closure of the modified two-fluid model, Int. J. Heat Fluid Flow 37
Ann. Nucl. Energy 126 (2019) 330–349. (2012) 196–208.
[9] N. Zuber, J.A. Findlay, Average volumetric concentration in two-phase flow [22] T. Hibiki, M. Ishii, One-dimensional drift–flux model for two-phase flow in a
sytems, J. Heat Transf. 87 (4) (1965) 453–468. large diameter pipe, Int. J. Heat Mass Transf. 46 (10) (2003) 1773–1790.
[10] M. Ishii, T. Hibiki, Thermo-Fluid Dynamics of Two-Phase Flow, Second Edition, [23] K. Mishima, et al., Measurement of void profile in a simulated subchannel of a
Springer, 2011. tight lattice core with use of neutron radiography, Annu. Rep. Res. React. Inst.
[11] K. Mishima, T. Hibiki, Some characteristics of air-water two-phase flow in Kyoto Univ. 27 (1994) 245–253.
small diameter vertical tubes, Int. J. Multiph. flow 22 (4) (1996) 703–712. [24] K. Mishima, T. Hibiki, Y. Saito, H. Nakamura, M. Matsubayashi, The review of
[12] W. Zhang, T. Hibiki, K. Mishima, Correlations of two-phase frictional pressure the application of neutron radiography to thermal hydraulic research, Nucl. In-
drop and void fraction in mini-channel, Int. J. Heat Mass Transf. 53 (1–3) strum. Methods Phys. Res. Sect. A Accel. Spectrom. Detect. Assoc. Equip. 424
(2010) 453–465. (1) (1999) 66–72.
[13] T. Hibiki, One-dimensional drift-flux correlations for two-phase flow in medi- [25] G. Wölk, M. Dreyer, H.J. Rath, Flow patterns in small diameter vertical non-cir-
um-size channels, Exp. Comput. Multiph. Flow 1 (2) (2019) 85–100. cular channels, Int. J. Multiph. Flow 26 (6) (20 0 0) 1037–1061.
[14] T. Hibiki, et al., Constitutive equations for vertical upward two-phase flow in [26] K. Shirvan, N. Andrews, M.S. Kazimi, Best estimate void fraction and critical
rod bundle, Int. J. Heat Mass Transf. 127 (2018) 1252–1266. power correlations for tight lattice BWR bundles, in: Proceedings of the 2013
[15] T. Hibiki, M. Ishii, Distribution parameter and drift velocity of drift-flux model International Congress on Advances in Nuclear Power Plants (ICAPP 2013),
in bubbly flow, Int. J. Heat Mass Transf. 45 (4) (2002) 707–721. 2013, p. FA159.
[16] M. Ishii, “One-dimensional drift-flux model and constitutive equations for rela- [27] B. Chexal, G. Lellouche, J. Horowitz, J. Healzer, and S. Oh, “The Chexal-Lellouche
tive motion between phases in various two-phase flow regimes,” Argonne Na- void fraction correlation for generalized applications,” Nuclear Safety Analysis
tional Laboratory Report, ANL-77-47, 1977. Center of the Electric Power Research Institute, ISA (1991).
[17] P. Coddington, R. Macian, A study of the performance of void fraction correla- [28] L.H. Liao, A. Parlos, P. Griffith, Heat transfer, carryover and fall back in nuclear
tions used in the context of drift-flux two-phase flow models, Nucl. Eng. Des. steam generators during transients, in: Proceedings of the 8th International
215 (3) (2002) 199–216. Heat Transfer Conference, 1986, pp. 2437–2442.
[18] T. Ozaki, R. Suzuki, H. Mashiko, T. Hibiki, Development of drift-flux model
based on 8$\times$ 8 BWR rod bundle geometry experiments under proto-
typic temperature and pressure conditions, J. Nucl. Sci. Technol. 50 (6) (2013)
563–580.

14