Nuclear Engineering and Technology 52 (2020) 947e955

Contents lists available at ScienceDirect

Nuclear Engineering and Technology

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

Original Article

Simple analytical method for predicting the sloshing motion in a

rectangular pool
Won Man Park a, Dae Kyung Choi a, Kyungsoo Kim b, Sung Man Son a, Se Hong Oh a,
Kang Hee Lee c, Heung Seok Kang c, Choengryul Choi a, *
a
Elsoltec Inc, Yongin, Republic of Korea
b
Kyung Hee University, Yongin, Republic of Korea
c
Korea Atomic Energy Research Institution, Daejeon, Republic of Korea

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

Article history: Predicting the sloshing motion of a coolant during a seismic assessment of a rectangular spent fuel pool
Received 6 March 2019 is of critical concern. Linear theory, which provides a simple analytical method, has been used to predict
Received in revised form the sloshing motion in rectangular pools and tanks. However, this theory is not suitable for the high-
16 October 2019
frequency excitation problem. In this study, the authors developed a simple analytical method for pre-
Accepted 31 October 2019
Available online 4 November 2019
dicting the sloshing motion in a rectangular pool for a wide range of excitation frequencies. The corre-
lation among the linear theory parameters, influencing on excitation and convective waves, and the
excitation frequency is investigated. Sloshing waves in a rectangular pool with several liquid heights are
Keywords:
Sloshing
predicted using the original linear theory, a modified linear theory and computational fluid dynamics
Seismic evaluation analysis. The results demonstrate that the developed method can predict sloshing motion over a wide
Spent fuel pool range of excitation frequencies. However, the developed method has the limitations of linear solutions
Linear theory since it neglects the nonlinear features of sloshing motion. Despite these limitations, the authors believe
Computational fluid dynamics that the developed method can be useful as a simple analytical method for predicting the sloshing
motion in a rectangular pool under various external excitations.
© 2019 Korean Nuclear Society, Published by Elsevier Korea LLC. This is an open access article under the
CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction The level of the fluid, namely, the coolant, in an SFP is deter-
mined by considering the cooling of the spent fuel assemblies and
A spent fuel pool (SFP) is a temporary storage for spent fuel the shielded radiation dose from the spent fuel assemblies [3]. For
assemblies after their unloading from the reactor core. Its shape is example, the level should be 23 feet (approximately 7 m) above the
rectangular, and it contains coolant, in which spent fuel assemblies top of the fuel assemblies that are seated in the storage racks for
are immersed during storage. If the spent fuel assemblies are not Westinghouse Advanced Passive 1000 plants (AP1000) [4]. For the
cooled, the temperature of the spent fuel assemblies increases, Shin Kori 1 & 2 nuclear power plants in S. Korea, the level of the
thereby potentially resulting in a hydrogen explosion that is caused coolant should be maintained at higher than 23 ft from the top of
by the solubility of the Zircaloy that surrounds the spent fuel as- the rack, the temperature of the coolants should be lower than
semblies. We have already experienced an explosion in the accident 120
F (48.9
C) during operation and lower than 120
F (60
C)
at the Fukushima Daiichi Nuclear Power Plant, although the ex- during reloading fuel assembly, and the radiation dose rate around
plosion did not occur in an SFP [1,2]. It has also been reported that the free surface should be maintained at less than 0.025 mSv/hr.
water that contained radioactive materials flowed out during the When an external excitation is applied to the structures, tanks
temporal cooling of spent fuel assemblies with seawater instead of or pools containing fluids, such as SFPs, sloshing motion of the
the broken cooling water circulating pumps [1]. For these reasons, contained fluid occurs, for example, sloshing motion of the liquid
the seismic safety of spent fuel assemblies and the SFPs that contain fuel in an aircraft, sloshing motion of the liquid natural gas in a ship,
them has been recognized as an important issue after the and sloshing motion in large storage tanks or pools due to earth-
Fukushima Daiichi Nuclear Power Plant accident. quakes [5]. This fluid sloshing can result in severe mechanical
problems in liquid storage tanks and other structures because the
sloshing motion that is caused by an external excitation could be
* Corresponding author. ELSOLTEC, #303 U-Tower, 120 Heungduk Jungang-rho,
the source of extra forces on the roof and wall of the tank by
Giheung-gu, Yongin, Gyeonggi-do, 16950, Republic of Korea.
E-mail address: [email protected] (C. Choi). impacting them [5e9]. Therefore, it is highly important to

https://doi.org/10.1016/j.net.2019.10.025
1738-5733/© 2019 Korean Nuclear Society, Published by Elsevier Korea LLC. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).
948 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955

investigate the sloshing motion for ensuring the mechanical safety in this study, the authors developed a simple analytical method for
of a structure that contains a large amount of liquid. predicting sloshing motion in a rectangular pool under high- and
Even though additional forces that are caused by impacting the low-frequency excitation conditions using the linear theory. Then,
roof could not be expected in the SFP because the SFP is not a roofed the sloshing motions were predicted under various excitation fre-
tank, the prediction of sloshing wave motion is also highly impor- quencies and validated by comparisons with the results from a
tant in SFP because it results in overflow and loss of the coolant. The commercial CFD package.
coolant in the SPF is used to cool the discharged spent fuel as-
semblies and to shield their radiation dose [3]. Therefore, loss of
coolant for any reason might result in an increase in the coolant 2. Materials and methods
temperature due to insufficient removal of decay heat from the
spent fuel assembly, which might also lead to an additional One of the SFPs in Korean nuclear power plants was selected.
decrease in coolant via evaporation. Furthermore, loss of coolant, The pool was simplified into a two-dimensional rectangular pool, of
which refers to the depletion of the shield that is caused by the which the height and width are 0.815 m and 0.52 m, respectively,
coolant also acting as a shielding material, might lead to the escape which was scaled down to 1/20 the size of the original pool (Fig. 1).
of radiation around the free surface of the coolant [10]. Therefore, Three water levels, namely, 0.61 m, 0.42 m, and 0.23 m, were used.
prediction of sloshing motion of the coolant in the SFP is important Changes in the sloshing elevation on the left wall were calculated
for designing the SFP to ensure not only its structural safety, which for 10 s under various sine-wave excitations using custom MATLAB
is impacted by sloshing, but also its radiation safety, which is (MATLAB 2018a, MathWorks, Inc., Natick, MA, USA) codes for the
impacted by the loss of coolant. linear theory, the modified linear theory, and the CFD analysis.
Various analytical, numerical, and experimental studies have Excitation frequencies of 1 Hz, 2 Hz, 3 Hz, 4 Hz, 5 Hz, 10 Hz, 20 Hz,
been conducted to analyze fluid sloshing motions. Housner pre- and 30 Hz were used, and the amplitude was set to 5 mm for all
sented a simple approximate method with beam-mass and spring- excitation frequencies.
mass models for predicting the water pressure and sloshing
elevation [11]. In Housner's method, the walls are assumed to be
rigid. Epstein suggested a method for considering the flexibility of 2.1. Linear theory
the structure that is based on Housner's study [12]. Faltinsen has
introduced a linear solution for estimating sloshing in rectangular A simple analytical solution, namely, the linear theory, has been
tanks or pools [13]. Based on these analytical studies, various nu- established based on potential wave theory for predicting the
merical methods have also been developed for investigating the sloshing motion of a partially filled liquid in a pool or tank by Fal-
sloshing motion via the finite element method and the finite dif- tinsen [13]. The linear theory was developed for a two-dimensional
ference method [14e16]. Recently, many studies have utilized sloshing problem in a horizontally excited rectangular pool with a
commercial computational fluid dynamics (CFD) software, such as motion that is described by DðtÞ ¼ A$sinðutÞ (Fig. 1), where D, A, u,
Ansys/Fluent (Ansys, Inc., Canonburg, PA, US), Ansys/CFX, and Star and t are the displacement, amplitude, angular velocity of the
CCMþ (Siemens PLM software, Plano, TX, US). The results of these excitation, and time, respectively. Here, the liquid is assumed to be
studies demonstrated that the use of software is a feasible and inviscid and flowing irrotationally.
effective approach for accurately predicting sloshing motion sub- The sloshing elevation in the linear theory (x, Eq. (1)) can be
ject to the external excitation [17e19]. However, this approach re- divided into two terms: the sloshing elevation that corresponds to
quires a large amount of computing resources. external excitation (x1 , Eq. (2)) and the sloshing elevation that
Linear theory, which was introduced by Faltinsen [13], has been corresponds to liquid convection (x2 , Eq. (3)) [5,8].
used to predict fluid behaviors in response to external excitations,
xðx; tÞ ¼ x1 ðx; tÞ þ x2 ðx; tÞ (1)
such as seismic waves, ship motions, and filling in pools or tanks in
mechanical engineering and ocean engineering fields. Previous
studies have shown that the analytical solutions that are obtained A
2 X ∞ 
x1 ðx; tÞ ¼ xu þ Cn $ u $ sinðkn xÞ sinðutÞ (2)
via the linear theory accord with experimental and numerical re- g n¼0
sults under low-frequency excitations [5,8]. In high-frequency
excitation conditions, the nonlinearity of sloshing motion might
play a stronger role, and the error of the linear solution could in-
crease. Consequently, the theory has not been used to predict
sloshing motion in high-frequency excitation conditions.
Jung et al. have investigated sloshing motion under various
excitation frequencies via a numerical method [20]. The results
demonstrated that sloshing motion under both low- and high-
frequency excitation involves sloshing elevation via convective
(its frequency is similar to the first natural frequency of the fluid)
and impulsive (its frequency is similar to the excitation frequency)
effects. The results also demonstrated that the sloshing motion
occurred in a limited range under high excitation frequencies, while
this range was much larger when the excitation frequency was the
same as the first natural frequency of the contained fluid.
The prediction of sloshing motion via the linear theory consists
of convective and impulsive components. The authors expected
that the theory could be used to predict sloshing motion even
under high excitation frequencies if the amplitudes of the
convective and impulsive components of the linear theory could be
limited to a suitable range under high excitation frequencies. Thus, Fig. 1. Schematic diagram of fluid sloshing in a rectangular pool.
 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955 949

 
AX ∞
H 8 ðwhen u<2u1 Þ
x2 ðx; tÞ ¼ un Cn þ 2n sinðkn xÞsinðun tÞ (3) >
g n¼0 u >
>
>
>
> ! !
>
> X∞  
>
>
> A 4 ð1Þn u4
Here, un is the natural frequency of the fluid and kn is the wave- >
> xu þ
2
sinðkn xÞ sinðutÞ
>
<g L k2n u2n  u2
number. The other two parameters, namely, Hn and Cn , are defined n¼1
as follows: x1 ðx;tÞ¼
>
> ðwhen u 2u1 Þ
>
>
>
>
>
>
u2n ¼ g $ kn $tanhðkn hÞ (4) >
>
>
> ! !
>
> X∞  
:A 0 4 ð1Þn u4
2n þ 1 xu þ2
0 sinðkn xÞ sinðutÞ
kn ¼ p (5) g n¼1
L k2n u2n  u 2
L
(10)
  !
4 ð1Þn In the case of u 2u1 , x1 can be derived as follows using the
Hn ¼ u3 (6) format of Eq. (2):
L k2n
A
02 X ∞
0 0

Hn x1 ðx; tÞ ¼ xu þ C n u sinðkn xÞ sinðutÞ (11)
Cn ¼ 2 (7) g n¼1
un  u2
0
Here, n ranges from 0 to infinity; however, in this study, values from 0 Hn
Cn ¼ (12)
0 to 50 were considered. Only odd wavenumbers were included in u2n e u0 2
the solution.
  !
0 0 4 ð1Þnþ1
Hn ¼ u 3 (13)
L k2n
2.2. Modification of the linear theory
0
Sloshing motion consists of impulsive and convective compo- u ¼ 2u1 (14)
nents in the linear solution. The amplitude of the convective
component decreases as the excitation frequency increases, while Here, n ranges from 1 to infinity (50 in this study).
the amplitude of the impulsive component increases. Thus, the In previously published experimental and numerical studies,
excitation component could be mainly represented in the linear mainly the first-mode shape and natural frequency are observed
solution of the free-surface height under a high-frequency excita- [20,21]. Therefore, in the calculation of x2 , the first term of the
tion, while sloshing motion is affected by the convective compo- formula for x2 (when n is 1), namely,
nent. Therefore, the linear theory should be modified to predict  
sloshing motion in which the first-mode natural frequency is A H1
x2 ðx; tÞ ¼ u1 C1 þ sinðk1 tÞsinðu1 tÞ
mainly exhibited. g u2
  !
For the case in which n ranges from 1 to infinity (50 in this u31 u
A 4 ð  1Þn
study), Eq. (5) can be expressed as follows: ¼ sinðk1 tÞsinðu1 tÞ (15)
g L k21 u21 u2
2n  1
kn ¼ p (8) affects mainly x2.
L
For predicting sloshing motion under high frequency, we
In the calculation of x1 , the amplitude of x1 (Ax1 Þ is expressed as modified the first term. According to the convective effect x2 , the
follows: sloshing evaluation is associated with the natural frequency of the
! ! fluid un . However, we assumed that the amplitude that is associ-
X∞  
A 4 ð  1Þnþ1 u4 ated with the first-mode natural frequency of the fluid, namely,
Ax1 ¼ x u2 þ sinðkn xÞ (9)
g n¼1
L k2n u2n  u2  
A H
The amplitude of x1 increases as the excitation frequency, u1 C1 þ 21 ; (16)
g u
namely, u, increases. Here, since x1 is a linear solution, we expect
the error that is caused by the linear solution to increase as u in- is associated with the amplitude of x1 :
creases. Therefore, it is helpful to identify the frequency range in
which the linear theory is applicable. In this study, the threshold A
02 X ∞
0 0

value for the linear theory was selected as twice the first natural xu þ C n u sinðkn xÞ (17)
g n¼1
frequency, namely, u1 , of the fluid via preliminary computer
calculation. Thus, the formula for x1 was derived as follows by Concretely, we assumed that Eq. (16) is related to Eq. (17) on the
replacing 2u1 with u0 in the case of u  2u1 : left wall (at x ¼  2L ). Thus,
950 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955

   X∞   
H1 L 0 0 0 L
u1 C1 þ ¼k $u2þ C n u sin kn (18)
u2 2 n¼1
2

was assumed. Here, k is a scaling factor that is associated with the

excitation frequency u.
We can express Hu2 as follows:
1

  !
H1 4 ð1Þ2
¼ u (19)
u 2 L k21
This is a linear equation with respect to u.
We used C1 to find the scaling factor k. A linear equation of u was
generated between 3 Hz and 10 Hz, as expressed in Eq. (20):

C1 ðuÞ ¼ au þ b (20)
Similar to the assumption in Eq. (18), the slope of the linearized
approximate expression, namely, C1 ðuÞ, is assumed to be k times
0
the slope of the line that connects ð0; bÞ and ðu;C1 ðu ÞÞ. Thus, k can
be calculated via Eq. (21) and Eq. (22).
 0
C1 u  b
k ¼a (21)
u
Fig. 2. CFD model of water and air for simulating fluid sloshing.

au surfaces between fluids (water and air) and the wall. The opening
k¼ (22)
C1 ðu Þ  b
0
static pressure of the superior plane of the air was set to 0 Pa. A
Therefore, x2 can be expressed as follows: commercial CFD package, namely, Ansys/Fluent, was used for
modeling and analysis.

8 ðwhen u < 2u1 Þ

>
>
>
>
>
>  
>
>
>
> AX ∞
Hn
>
>
>
un Cn þ sinðkn tÞsinðun tÞ
< g n¼1 u2
x2 ðx; tÞ ¼ (23)
>
> ðwhen u  2u1 Þ
>
>
>
>
>
>
>
>  
>
>
> A AX ∞
H
: K u sinðk1 tÞsinðu1 tÞ þ un Cn þ 2n sinðkn tÞsinðun tÞ
g g n¼2 u

3. Results
 X∞   
L 0 0 0 L a The natural frequency of the fluid in the pool could be calculated
K¼ $u2 þ C n u sin kn (24)
2 n¼1
2 C1 ðu 0
Þb via Eq. (11) and Eq. (12). However, those equations are applicable
only for odd wavenumbers, and the natural frequencies of the fluid
for the both odd and even wavenumbers could be calculated using
Eq. (30).

2.3. Numerical method sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

 
npg nph
A three-dimensional CFD model, which consists of water and air, un ¼ tanh (30)
L L
was established based on the volume of fluid method for the
tracking of a free surface (Fig. 2) [8]. The width, height and depth of The calculated first-mode natural frequencies were 1.22 Hz,
the pool were set to 0.52 m, 0.815 m, and 0.15 m, respectively. The 1.22 Hz, and 1.15 Hz for the levels of 0.61 m, 0.42 m, and 0.23 m,
temperatures of the water and air were both set to 25
C. The respectively (Table 1). Therefore, the modified linear theory was
densities of water and air were set to 997 kg/m3 and 1.185 kg/m3, applied in cases in which the excitation frequency exceeded 3 Hz.
and the viscosities of water and air were set to 8.899  104 kg/m-s The sloshing elevation that was measured on the left wall was
and 1.831  105 kg/m-s, respectively. The surface tension of water used for comparison with the elevations that were predicted by the
was set to 0.07199 N/m. The standard k-ε turbulence model was linear theory, the modified linear theory, and the CFD analysis. The
used in this study. Nonslip boundary conditions were applied on all sloshing elevation was predicted on the middle line of the wall in
 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955 951

Table 1
Natural frequencies of the fluid with three levels for modes 1 to 5.

Mode Height of the water level

0.61 m 0.42 m 0.23 m

1 1.22 1.22 1.15

2 1.73 1.73 1.73
3 2.12 2.12 2.12
4 2.45 2.45 2.45
5 2.74 2.74 2.74

the CFD analysis.

When excitation frequency was 1 Hz, similar patterns of
changes in the sloshing elevations that were predicted by the CFD
analysis and the linear theory were observed in all water-level
cases (Figs. 3 and 4).
The maximum sloshing elevations that were predicted by the
CFD analysis were 23 mm, 21 mm, and 36 mm for water levels of
0.61 m, 0.42 m, and 0.23 m, respectively. The elevations that were
predicted by the linear theory were 30 mm, 31 mm, and 37 mm,
and the differences between the results of the CFD analysis and the
linear theory were 7.7 mm (34%), 9.9 mm (48%), and 0.9 mm (2%).
The sloshing motion that was predicted by the linear theory over
10 s with a water level of 0.23 m, which showed higher sloshing
elevation than other cases, accorded with the results from the CFD
analysis (Fig. 5).
When the sinewave excitation frequency of 2 Hz was used, only
the linear theory was used to obtain the analytical results. The
maximum sloshing elevation that was predicted by the CFD anal-
ysis was 50 mm, compared to 33 mm by the linear theory. The
difference between the CFD and analytical results was 17 mm (34%).
For the excitation frequency of 3 Hz, both the linear theory and
the modified linear theory were used to obtain the analytical re-
sults. In the cases with excitation frequencies of 3 Hz, 4 Hz, and
5 Hz, the differences between the CFD results and the analytical
results that were obtained using the linear theory were 6 mm
(21%), 13 mm (30%) and 37 mm (55%), respectively. The differences
were 13 mm (44%), 12 mm (28%) and 3 mm (4%) when the modified
linear theory was used for the analytical calculations.
The oscillation at the same frequency as the applied excitation
frequency was calculated in the analytical results using the linear
theory. The results that were obtained using the modified linear
theory showed a similar pattern of the sloshing elevation (Fig. 6).
This phenomenon was more strongly observed when the excitation
frequency was increased to 20 Hz and 30 Hz (Fig. 7).
The differences between the results from the CFD and the linear Fig. 3. Sloshing elevation for an excitation frequency of 1 Hz.
theory were 141 mm (89%), 121 (87%), and 119 mm (87%) with
water levels of 0.61 m, 0. 42 m, and 0.23 m, respectively, under the
excitation frequency of 10 Hz. The differences decreased to 109 mm level in an SFP of a nuclear power plant that was due to sloshing
(69%), 89 mm (64%), and 60 mm (44%) (Fig. 6). overflow that was induced by earthquakes [22]. Only 320 gallons
(approximately 1.2 cubic meters) of water loss was reported in the
Kashiwazaki-Kariwa site during the Niigataken Chuetsu-Oki
4. Discussion earthquake (July 16, 2007, with a moment magnitude of
Mw ¼ 6.6). A decrease in the water level of the SFP by 8 inches
The prediction of fluid (coolant) sloshing in an SFP of a nuclear (approximately 20 cm) due to sloshing overflow was reported at
power plant is highly important for ensuring the safety of the nu- Onagawa plant after the Tohoku earthquake (March 11, 2011, Japan,
clear power plant because it is associated with overflow and loss of Moment magnitude Mw ¼ 9.0). Although decreases in SFP water
the coolant. In this study, the authors have developed a simple levels at the Fukushima Daiichi plant have not been reported, the
method for predicting sloshing motions in a rectangular pool by decreases have been assumed based on evaluations of the SFP
modifying the previously introduced linear theory [13]. Sloshing performance. A decrease of 1.6 ft (0.5 m) in the water level was
motions in a rectangular pool, which is assumed to be a simplified assumed for Unit 2, and decreases of 5 ft (1.5 m) were assumed for
SFP because the shape of an SFP is rectangular, with various water Units 1 and 3. Those values are substantial from a nuclear safety
levels and under various excitation frequencies were predicted and perspective.
compared to the results from CFD analysis. The linear theory has been used to predict the seismic responses
A previously published report described a decrease in the water of liquid storage tanks. The linear theory predicts sloshing motions,
952 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955

Fig. 4. Sloshing motion for a water level of 0.23 m under a 1 Hz sinewave excitation. The analytical results are represented by the black solid line. The red and blue parts correspond
to air and water, respectively, in the CFD analysis. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Fig. 5. Sloshing elevation of the model with a water level of 0.23 for excitation frequencies of 2 Hz, 3 Hz, 4 Hz and 5 Hz.

and its prediction results accord not only with the numerical behavior of the convective component of the sloshing [8], sloshing
analysis but also with the experimental results under low- motions under excitation frequencies of 1 Hz, in which the fluid
frequency and small-amplitude excitations of a short duration resonates with the excitation frequencies, accorded with the CFD
(Fig. 3) [8]. Because the first mode of the filled fluid is the primary results (Fig. 3). Thus, Wu et al. suggested a linear theory formula
 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955 953

Fig. 8. Fast Fourier transform of the sloshing elevation for a water level of 0.23 m
under an excitation frequency of 10 Hz.

that considers only the first-mode natural frequency of the fluid [5].
However, the role of the higher modes becomes increasingly
important with the increment of the excitation frequency, and the
higher natural frequencies are not negligible, as reported by Chen
et al. [14].
Previously published studies have reported that fluid sloshing
behaves with a frequency that is very close to the first mode of the
fluid [20,21]. According to Chen et al., the maximum sloshing
amplitude could be realized using only the convective term [14].
The Fast Fourier transform of the sloshing elevation, which is pre-
sented in Fig. 6 (c), demonstrates that the first mode of the fluid is
the primary behavior of the fluid sloshing; however, other modes of
fluid natural frequencies and impulsive effects also affect the
sloshing motions under high-frequency excitation conditions
(Fig. 8). The linear theory could be modified to predict the sloshing
elevation even under high-frequency excitation based on these
findings.
Although the differences between the CFD and analytical results
decreased by 20%, 23%, and 43% in cases with water levels of 0.61 m,
0.42 m, and 0.23 m, respectively, the differences remain non-
negligible, even in cases in which the modified linear theory was
used. Fig. 9 presents the sloshing motions under sinewave excita-
tion frequencies of 5 Hz and 10 Hz. Splashed water on the wall or
around the wall was predicted for both excitation frequencies, and
the splashed water might have resulted in the sloshing height that
was predicted via CFD analysis being much higher than that via the
modified linear theory. The delayed sloshing motion that was
Fig. 6. Sloshing elevation for an excitation frequency of 10 Hz. predicted by the modified linear theory compared to the CFD re-
sults could also be due to neglected nonlinear parameters, such as
the viscoelasticity and surface tension, which are fluid properties.

Fig. 7. Sloshing elevations of the model with a water level of 0.23 for excitation frequencies of 20 Hz and 30 Hz.
954 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955

Fig. 9. Sloshing motions for a water level of 0.23 m under sinewave excitations of (a) 5 Hz and (b) 10 Hz.

According to Goudarzi and Sabbagh-Yazdi, nonlinear effects in the predicted by the modified linear theory were also affected by the
sloshing motions are observed in practice [8], while the linear neglect of nonlinear features in the linear theory. Therefore, nu-
theory assumes that the nonlinear hydrodynamic effects are merical analysis, such as CFD analysis, should be conducted when
negligible [8,13]. The differences between the CFD and analytical investigating sloshing motions in detail. Only two-dimensional
results under high excitation frequencies (>5 Hz) may result from sloshing motions under horizontal sinewave excitation conditions
nonlinear effects. have been predicted. Thus, the method should be extended to the
To investigate the differences between the CFD and analytical prediction of sloshing motions under three-dimensional irregular
results with the removal of the nonlinear features, the Fast Fourier excitation frequencies. Moreover, the proposed simple analytical
transform was applied to the sloshing elevation (Fig. 10). The results method (the modified linear theory) yielded less conservative
of the Fast Fourier transform demonstrate the improved prediction values in the estimation of the sloshing level in the high-frequency
results of the modified linear theory. The differences in the range than the CFD analysis (Figs. 6 and 7). Therefore, it is necessary
convective component amplitudes decreased from 98%, 98%, and to investigate the differences between the prediction results that
99% to 10%, 5%, and 2% in cases with water levels of 0.61 m, 0.42 m, were obtained via the modified linear theory and via CFD prior to
and 0.23 m, respectively. The differences in the impulsive compo- using the modified linear theory in the seismic assessment of a
nent amplitudes also decreased from 86%, 54%, and 50% to 24%, 2%, nuclear power plant.
and 50%. Although the differences in the impulsive component
amplitude remained high in cases with water levels of 0.61 m and 5. Conclusion
0.23 m, the magnitudes were only 2 mm and 5 mm, respectively.
A linear solution for predicting the solution of a nonlinear In this study, we developed a simple analytical method, namely,
problem generates a larger error if nonlinear features play a modified linear theory, for predicting the sloshing motions of a
stronger role. Thus, the prediction by the original linear theory rectangular pool under a wide range of excitation frequencies.
differed substantially from the CFD results under high-frequency Although the developed method has several limitations, the results
excitation conditions. To reduce the error between the prediction of this study demonstrate that the modified linear theory could
via linear solution and the physical phenomenon, we've identified a predict sloshing motions under both low- and high-frequency
suitable threshold value and modified the original linear theory. excitation conditions. Thus, the authors believe this method,
Consequently, the modified linear theory predicts the sloshing which is a simple analytical method, could be used to quickly es-
motions of the filled fluid more accurately, even under high- timate the maximum sloshing elevation of a rectangular pool and
frequency excitation conditions. The results of this study demon- to predict the water loss that is caused by sloshing-induced over-
strate that the prediction accuracy of the sloshing height was flow. Since the general shape of all spent fuel pools is rectangular,
substantially higher compared to the results of the original linear the modified linear theory can be used to predict the sloshing be-
theory under high-frequency excitation conditions. Therefore, we haviors of the coolant of the spent fuel pool. Moreover, the pro-
believe that the process of establishing the modified linear theory is posed method could be used to predict the sloshing motions of
reasonable because the modified linear theory accurately predicts fluids in various tanks, for example, liquid natural gas tanks, in
sloshing behaviors for both low- and high-frequency excitation various industrial fields.
frequencies. However, this study could not provide a rationale for
selecting the threshold frequency of the applied excitation or the
Declaration of competing interest
equations for the impulsive effects under high-frequency excitation
conditions.
The authors declare that they have no known competing
The modified linear theory that was developed in this study was
financial interests or personal relationships that could have
also established using the linear theory. Thus, the results that were
appeared to influence the work reported in this paper.
 W.M. Park et al. / Nuclear Engineering and Technology 52 (2020) 947e955 955

Acknowledgments

This work was supported by the Korea Institute of Energy

Technology Evaluation and Planning (KETEP) and the Ministry of
Trade, Industry & Energy (MOTIE) of the Republic of Korea (No.
20171510101920).

References

[1] M. Baba, Fukushima accident: what happened? Radiat. Meas. 55 (2013)

17e21.
[2] K. Shozugawa, N. Nogawa, M. Matsuo, Deposition of fission and activation
products after the Fukushima Dai-ichi nuclear power plant accident, Environ.
Pollut. 163 (2012) 243e247.
[3] H. Wang, L. Ge, J. Shan, J. Gou, B. Zhang, Safety analysis of CPR1000 spent fuel
pool in case of loss of heat sink, in: Int. Conf. Nucl. Eng. Proceedings, ICONE,
2013.
[4] U.S.N.R. Commission, Standard Technical Specifications, Westinghouse
Advanced Passive 1000 (AP1000) Plants, vol. 1, U.S. Nuclear Regulatory
Commission, Washington, 2016.
[5] G.X. Wu, Q.W. Ma, R. Eatock Taylor, Numerical simulation of sloshing waves in
a 3D tank based on a finite element method, Appl. Ocean Res. 20 (1998)
337e355.
[6] S. Zama, H. Nishi, M. Yamada, K. Hatayama, Damage of oil storage tanks
caused by liquid sloshing in the 2003 Tokachi oki earthquake and revision of
design spectra in the long-period range, in: 14 Th World Conf. Earthq. Eng,
2008.
[7] M.A. Goudarzi, S.R. Sabbagh-Yazdi, W. Marx, Seismic analysis of hydrody-
namic sloshing force on storage tank roofs, Earthq. Spectra 26 (2010)
131e152.
[8] M. Ali Goudarzi, S. Reza Sabbagh-Yazdi, Investigation of nonlinear sloshing
effects in seismically excited tanks, Soil Dyn. Earthq. Eng. 43 (2012) 355e365.
[9] P.K. Malhotra, Sloshing loads in liquid-storage tanks with insufficient free-
board, Earthq. Spectra 21 (2005) 1185e1192.
[10] T.M. Shin, Safety review of severe accident senario for wet spent fuel storage
facility, J. Korean Radioact. Waste Soc. 9 (2011), 231e2326.
[11] G.W. Housner, The dynamic behavior of water tanks, Bull. Seismol. Soc. Am.
53 (1963) 381e387.
[12] I. Howard, Epstein, Seismic design of liquid storage tanks, ASCE J. Struct. Div.
102 (1976) 1659e1673.
[13] O.M. Faltinsen, A numerical nonlinear method of sloshing in tanks with two-
dimensional flow, J. Ship Res. 22 (1978) 193e202.
[14] W. Chen, M.A. Haroun, F. Liu, Large amplitude liquid sloshing in seismically
excited tanks, Earthq. Eng. Struct. Dyn. 25 (1996) 653e669.
[15] T. Okamoto, M. Kawahara, Two-dimensional sloshing analysis by Lagrangian
finite element method, Int. J. Numer. Methods Fluids 11 (1990) 453e477.
[16] S. Aus der Wiesche, Computational slosh dynamics: theory and industrial
application, Comput. Mech. 30 (2003) 374e387.
[17] S. Nicolici, R.M. Bilegan, Fluid structure interaction modeling of liquid sloshing
phenomena in flexible tanks, Nucl. Eng. Des. 258 (2013) 51e56.
[18] L. Hou, F. Li, C. Wu, A numerical study of liquid sloshing in a two-dimensional
tank under external excitations, J. Mar. Sci. Appl. 11 (2012) 305e310.
[19] Y. Su, Z.Y. Liu, Numerical model of sloshing in rectangular tank based on
Boussinesq-type equations, Ocean Eng. 121 (2016) 166e173.
[20] J.H. Jung, H.S. Yoon, C.Y. Lee, Effect of natural frequency modes on sloshing
phenomenon in a rectangular tank, Int. J. Nav. Archit. Ocean Eng. 7 (2015)
580e594.
[21] H. Akyildiz, N. Erdem Ünal, Sloshing in a three-dimensional rectangular tank:
numerical simulation and experimental validation, Ocean Eng. 33 (2006)
2135e2149.
Fig. 10. Fast Fourier transform of the sloshing elevation for water levels of (a) 0.61 m, [22] U.S.N.R. Commission, Consequence study of a beyond-design-basis earth-
(b) 0.42 m, and (c) 0.23 m under an excitation frequency of 10 Hz. quake affecting the spent fuel pool for a U.S. Mark I boiling water reactor,
NUREG 2161 (2014).