Annals of Nuclear Energy 101 (2017) 215–225

Contents lists available at ScienceDirect

Annals of Nuclear Energy

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

Modeling of flashing-induced flow instabilities for a natural circulation

driven novel modular reactor
Shanbin Shi a,⇑, Mamoru Ishii b
a
Department of Mechanical and Aerospace Engineering, The Ohio State University, 201 West 19th Avenue, Columbus, OH 43210, United States
b
School of Nuclear Engineering, Purdue University, 400 Central Drive, West Lafayette, IN 47907, United States

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

Article history: An analytical study based on frequency domain analysis is presented on the flashing-induced flow insta-
Received 11 August 2016 bility in a natural circulation test facility, which was designed to investigate the flow instability for a
Received in revised form 29 October 2016 BWR-type novel modular reactor (NMR). To address the flashing phenomena at low pressure conditions,
Accepted 4 November 2016
such as initial startup transients or accidents, the liquid enthalpy change in the P-T diagram due to
Available online 19 November 2016
reduced hydrostatic head in the riser or chimney was treated as an axially uniform heat flux. Based on
the drift flux model, the system transfer function was obtained through small perturbations about the
Keywords:
steady state in the frequency domain. The D-partition method was used to determine the neutral stability
Flashing instability
Frequency domain analysis
boundary in the dimensionless stability plane, which was constituted of the subcooling number and
Natural circulation test facility phase change number. From the frequency domain analysis, the flashing stability boundary and the den-
Stability map sity wave oscillations boundary could be predicted. Some parametric studies had been performed on the
system pressure and the inlet flow resistance coefficients in the stability analysis. The results showed that
the flashing stability boundary was more sensitive to the system pressure than the density wave oscilla-
tions. In addition, the theoretical stability boundaries were benchmarked against the experimental stabil-
ity boundaries from quasi-steady state tests. Although the general stability boundary agreed well with
the experiments, certain discrepancies still existed due to the assumptions of thermal equilibrium in cur-
rent study. In the future, the thermal non-equilibrium conditions including subcooled boiling will be
taken into account in the flashing induced stability analysis.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction section for the reactor cavity cooling system during accidental sce-
narios. The void fraction increases in the chimney section due to
The flashing induced stability problem, simplified as flashing the flashing evaporation. The flow fluctuations caused by the flash-
instability, in two-phase flow system is of great importance to ing instability usually have large magnitude depending on the sys-
the safe operation of naturally driven nuclear reactors and pas- tem design. Therefore, the safe operation or the system
sively safe engineering safety system in nuclear industry. The performance would be largely affected by the flashing instability.
designs of naturally driven nuclear reactors, ranging from conven- Although the flashing instability have been widely investigated
tional nuclear reactors to small modular reactors (Shi et al., 2016), experimentally and numerically in the last decade (Inada et al.,
feature a relative long chimney section above the reactor core to 2000; Furuya et al., 2005; Manera et al., 2005; Hu and Kazimi,
increase the driving force of natural circulation. In addition, the 2011; Lee et al., 2015; Su et al., 2002; Guo et al., 2016), the difficul-
design of the reactor cavity cooling system for the generation IV ties of predicting the flashing instability still exists due to its com-
reactors characterizes a long riser above the cooling panel plexity. The demand to improve and develop the mechanistic
(Lisowski et al., 2011). The flashing instability could easily occur model for the flashing instability becomes much urgent and neces-
in the chimney section due to reduced hydrostatic head during sary especially with the advancement of the next generation
the initial startup transients of the nuclear reactors, or in the riser nuclear plants (NGNP).
A recent research has been performed to investigate the flow
instability in a Novel Modular Reactor (NMR) design for low power
⇑ Corresponding author at: Department of Mechanical and Aerospace Engineer- and low pressure conditions. The NMR developed at Purdue
ing, Nuclear Program, The Ohio State University, 201 West 19th Avenue, Columbus, University is a BWR-type small modular reactor design, which
OH 43210, United States.
relies on natural circulation to provide driving force for both nor-
E-mail address: [email protected] (S. Shi).

http://dx.doi.org/10.1016/j.anucene.2016.11.005
0306-4549/Ó 2016 Elsevier Ltd. All rights reserved.
216 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225

Nomenclature

l dynamic viscosity [kg/m-s]

Latin letters m kinematic viscosity [m2/s]
A flow area [m2] n perimeter [m]
c specific heat [J/kg-K] q density [m3/s]
Ck kinematic wave velocity [m/s] s time scale [–]
cp specific heat at constant pressure [J/kg-K] s01;12;23;34;13 residence time in Regions (A), (B), (C), (D), and in the
D hydraulic diameter [m] heated section [s]
f friction factor or frequency [–] X characteristic frequency [–]
g gravitational acceleration [m/s2] x frequency [–]
ifg latent heat of vaporization [J/kg]
j volumetric flux or center-of-volume velocity [m/s] Superscripts
K K factor (Minor loss coefficient) [–] ⁄ dimensionless
l heated section length [m]
L axial length scale [–] Subscripts
m _ mass flow rate [kg/s] e exit/Region D
N dimensionless number [– f liquid
NFr Froude number [–] fl flashing
Nfl flashing number [–] g gas
Nsub subcooling number [–] h heated
NZu Zuber (phase change) number [–] i, in inlet
p pressure [Pa] k each phase
q power [w] m mixture
q00 heat flux [w/m2] me mixture in Region D
q000 volumetric heat generation rate [w/m3] o reference point/component
s complex number [a + bi] s saturation
t time [s] sub subcooling
T temperature [K or °C] W Wall
v velocity [m/s] w wetted
V volume [m3] 3 bottom of Region D
Vgj drift velocity [m/s] 4 top of Region D
z axial coordinate [m] 12 Region B
23 Region C
Greek 34 Region D
a void fraction [–]
Cg mass generation for the vapor phase [kg/m3-s] Operators
D difference [–] – time average
e infinitesimal [–] Dk
Dt
@
¼ @t þ vk  r
k non-boiling length [m]
Kn various transfer function [–]

mal operations and accidental management (Ishii et al., 2015; Wu sis, including the one-dimensional drift flux model, flashing insta-
et al., 2016; Odeh and Yang, 2016). In previous research, a natural bility model, detailed derivations of the kinematics and dynamics
circulation test facility was scaled and designed from the NMR by of the fluid in the riser section. Section 3 presents the experimental
using the three-level scaling method. The experimental study of stability boundaries from the quasi-steady state tests in a small-
startup transients of the NMR showed that more than one flow scaled natural circulation test facility. Section 4 presents the theo-
instability mechanisms occurred in the natural circulation test retical stability maps and discusses the uncertainties between the
facility during the normal initial startup procedures (Shi et al., theoretical and experimental stability maps. Key conclusions are
2015a,b). From the experimental research, two startup procedures, summarized in Section 5.
i.e., very slow startup procedures and pressurized startup proce-
dures, were proposed for the initial startup procedures for the
NMR (Shi et al., 2014) to eliminate the flow instabilities. In addi- 2. Frequency domain analysis
tion, the experimental stability maps were obtained under low
pressures by performing quasi-steady state tests (Shi et al., The basic method used in this research to study the flashing
2015c). Following the experimental research, the theoretical stabil- instability is called linear frequency domain stability analysis,
ity analysis in the frequency domain has been carried out to predict which can be used to obtain stability boundaries for a flow system.
the observed flow instabilities occurred in the test facility (Shi, This method is a conventional way to analyze other flow instabil-
2015). ities such as the density wave oscillations (DWO) (Shi et al.,
In this paper, a new flashing mechanistic model related to the 2016) and Ledinegg instability (flow excursion) etc. The frequency
flashing number combining with the linear frequency domain domain analysis in this research was based on the drift-flux model
analysis under thermal equilibrium conditions have been devel- and its constitutive equations (Zuber, 1967) for a two-phase flow
oped to predict the flow instability boundary for the NMR. Section 2 system. For example, through applying a small perturbation about
introduces the fundamental basis for the frequency domain analy- the steady state in the inlet flow velocity, the system transfer
 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225 217

functions for the perturbed pressure drop along the channel could @ qm @ qm v m
þ ¼0 ð1Þ
be obtained by linearizing the original non-linear partial differen- @t @z
tial equations. The system characteristic equation was solved to Continuity equation for dispersed phase:
obtain the flow stability boundaries between the stable regions
and unstable regions in the frequency domain. Ishii (Ishii, 1971) @ aqg @ aqg v g
þ ¼ Cg ð2Þ
applied this method to investigate the density wave oscillations @t @z
and performed parametric studies of the heat flux, inlet subcooling, Mixture momentum equation:
inlet velocity, system pressure, inlet and outlet K factor, etc. on the

@v m @v m @pm f m
density wave oscillations. Saha (Saha, 1974) developed the thermal qm þ vm ¼  q v m v m  qm g
non-equilibrium model for the two-phase flow system based on @t @z @z 2D m
" 2
#
Ishii’s work. Since the flashing instability normally occurs in the @ ag qg qf V gj
chimney or riser section which is located downstream of a heated  ð3Þ
@z ð1  ag Þqm
section, the frequency domain analysis is applied to the flow sys-
tem including both the heated section and the chimney or riser Mixture enthalpy-energy equation:
section in current research.


@im @im q00W nh @ ag qg qf Difg V gj
qm þ vm ¼  ð4Þ
2.1. Formulation of the problem @t @z A @z qm
where mixture properties are defined as
In order to predict the flow instability observed in the experi-
ments analytically, the mathematical model is developed following qm ¼ aqg þ ð1  aÞqf ð5Þ
Ishii’s formulation of the DWO using conventional four-equation
drift flux model (Ishii, 1971). The drift flux model with the flashing ag qg ig þ ð1  ag Þqf if
im ¼ ð6Þ
model integrated aims to be capable of predicting the flashing qm
instability and the DWO for the NMR. The system of interest con-
sists of four components of the NMR as shown in Fig. 1: pm ¼ apg þ ð1  aÞpf ð7Þ

A. Single-phase upstream unheated region aqg v g þ ð1  aÞqf v f

B. Single-phase heated region with uniform core heat flux, q00 vm ¼ ð8Þ
qm
C. Two-phase heated region with uniform core heat flux, q00
D. Two-phase unheated region with assumed uniform flashing and the drift velocity between two phases can be expressed as
flux, q00fl V gj ¼ ð1  aÞðv g  v f Þ ð9Þ
and the latent heat can be calculated as
2.2. Field equations
Difg ¼ ig  if ð10Þ
The stability problems were formulated in terms of the time
In the case of thermal and mechanical equilibrium conditions,
smoothed and area averaged field and constitutive equations.
the thermal properties of each phase are assumed saturated condi-
The area averaged one-dimensional drift flux model can be
tions. The kinematics and dynamics of the stability problem had
expressed as:
been solved and obtained by Ishii (1971) for the DWO without con-
Mixture continuity equation:
sidering the flashing effects in the chimney section at high pres-
sures. However, the flashing effect might be dominant at lower
pressure transients for the natural circulation system, especially
during the startup transients. With the flashing model in the chim-
ney or riser section, the analysis method for the heated mixture
Region C can be adopted to derive the kinematics and dynamics
of the downstream un-heated Region D.

2.3. Flashing model

In a natural circulation system, superheated liquid near the top

of the riser or chimney can induce flashing due to the reduced
hydrostatic head under low pressure conditions. It is very difficult
to model the exact amount of the flashing vapor generation rate
during such fast and thermal non-equilibrium transients. In cur-
rent research, the vapor mass generation rate was simply derived
based on the water P-T diagram as shown in Fig. 2 by assuming sat-
urated conditions at the core exit. Thus, the liquid pressure chan-
ged from a higher pressure at the bottom of the chimney to an
upward lower pressure. Correspondingly, the enthalpy change rate
of the flashed liquid during this process could be calculated from
the temperature change from the P-T diagram.
Analogue to the core heating, the enthalpy change rate due to
liquid flashing was assumed as a uniform heat source along the
chimney. The assumed uniform flashing heat flux q00fl could be
Fig. 1. System used for analysis of flow instability. expressed as
218 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225

Normally, the characteristic equation was formulated in a series

of dimensionless numbers as

Qðs ; a1 ; a2 ;    ; an Þ ¼ 0 ð17Þ
where a1 to an were independent dimensionless numbers. For
the reactor stability analysis, the phase change number and inlet
subcooling number were selected to form the dimensionless stabil-
ity plane (Nsub  Npch). If the harmonic oscillations were consid-
ered to simplify the solutions, system response could be obtained
by substituting s ¼ jx into Eq. (17) as

Qðjx ; Npch ; Nsub Þ ¼ Q Re ðx ; Npch ; Nsub Þ þ jQ Im ðx ; Npch ; Nsub Þ

¼0 ð18Þ
Thus, Eq. (18) was reduced to

Q Re ðx ; Npch ; Nsub Þ ¼ 0 ð19Þ

Q Im ðx ; Npch ; Nsub Þ ¼ 0 ð20Þ

Eqs. (19) and (20) gave the harmonic frequency surfaces or neu-
Fig. 2. Flashing process in water P-T diagram. tral stability curves in the dimensionless plane of Nsub  Npch. The
D-partition method (Ishii, 1971; Achard et al., 1981) stated that the
number of roots lying in the right half s plane for each region
q00fl ¼ jf qf cf DT ð11Þ divided by surfaces do not change within a subdivision. Since the
neutral stability curve was determined, the stability of each region
and
could be determined by testing the stability at any point in that
jf ¼ ð1  aÞv f ð12Þ region using certain stability criteria such as Nyquist or Mikhailov
Criterion used by Ishii (1971). The number of roots for the dimen-
Therefore, the flashing mass generation rate could be derived sionless characteristic Eq. (16) should be zero for the stability of
from the energy balance equation for a two-phase flow system as the system.
ð1  aÞqf v f cf DT ¼ Cg;fl Difg l ð13Þ

and thus 2.5. Time lags and space lags

ð1  aÞqf v f cf DT Before the derivation of the transfer functions in the Region (D),
Cg;fl ¼ ð14Þ
Difg l two different methods, i.e. the Lagrangian and Eulerian specifica-
tion of the flow field, to describe the enthalpy of the fluid particle
where, a, qf , jf , v f , cf , l, DT, Difg , and Cg;fl represented the void
were illustrated in Figs. 3 and 4, respectively. In Fig. 3, the particle
fraction at the core exit, liquid density, liquid volumetric flux, liq-
entered region (A), (B), (C), and (D) at s0 , s1 , s2 , and s3 , respectively.
uid velocity, specific heat, height of the riser or chimney, tempera-
The residence time between two time points was also called time
ture difference corresponding to pressure differential based on P-T
lag, which was significantly related to the propagation of the den-
diagram, latent heat, and the vapor mass generation rate due to
sity waves. Accordingly, the time lags were replaced by the space
flashing, respectively. It was worthy of mention that the flashing
lags in Fig. 4, which defined the physical boundaries between
process should include the core section if its elevation was so large
two regions. One specially important space lag was the boundary
that the reduced hydrostatic head could not be neglected. How-
ever, in this study, the flashing effect was neglected in the core sec-
tion due to the relatively short core height in the NMR (Ishii et al.,
2015).

2.4. Transfer function and stability boundary

In order to study the flashing instability, the total pressure drop

perturbations of the system considered were analyzed with the
introduction of an inlet velocity perturbation, dv . To obtain the sta-
bility boundary, the system transfer function, 1=Q ðsÞ, was firstly
obtained between the total pressure drop perturbations and the
core inlet velocity perturbations as
1
dv ¼ dDpex ð15Þ
Q ðsÞ
where Q ðsÞ was system characteristic function in the system
transfer function. According to control theory, the system asymp-
totic stability could be determined by the nature of roots of the
characteristic equation given by
Q ðsÞ ¼ 0 ð16Þ
Fig. 3. Lagrangian description of enthalpy.
 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225 219

Here, void propagation equation in this region could be

expressed as

@a @ a qme Cg;fl
þ C ke ¼ ð29Þ
@t @z qf qg
Under the condition of constant qg and qf , the above equation
could be expressed in terms of the mixture density as

@ qme @q Dqqme
þ C ke me ¼ Cg;fl ð30Þ
@t @z qg qf
where

cf ðT 3  T 4 Þð1  aÞv m;l qf

Cg;fl ¼ ð31Þ
le Difg
C ke ; and V gje was the kinematic wave velocity, drift velocity
Fig. 4. Eulerian description of enthalpy.
between two phases in this region, respectively. Cg;fl was the vapor
generation rate due to flashing in this region. T 3 and T 4 were the
between the single-phase liquid and two-phase mixture and was saturation temperatures corresponding to the pressures at the bot-
denoted by k corresponding to the time lag s12 . tom and top of the chimney region, which were denoted by p3 and
p4 , respectively. a and v m;l were the steady state void fraction and
2.6. Kinematics of the un-heated region D with flashing mixture velocity at the exit of heated section, respectively. le was
the axial length of the downstream un-heated Region D. The char-
2.6.1. Volumetric flux equation and density propagation equation acteristic frequency was expressed as
The derivation of volumetric flux equation started from the con-
Dq
tinuity equations for each phase (vapor or liquid phase). Here the Xe ¼ Cg;fl ð32Þ
variables in this section are denoted by subscript e. Similar to the qg qf
vapor phase in Eq. (2), the continuity equation for the liquid phase
could be written as
2.6.2. Kinematic wave velocity
@ð1  aÞqf @ð1  aÞqf v f The integration of Eq. (24) from the exit of the core to location z
þ ¼ Cf ¼ Cg;fl ð21Þ
@t @z in the chimney gave
By carrying out the differentiations of the left hand side of Eqs. je ðz; tÞ ¼ je ðl; tÞ þ Xe ðz  lÞ ð33Þ
(2) and (21) as
By expanding qme into qme þ dqme ðtÞ at the exit of the heated
@ a @ av g Cg;fl a Dg qg section, the equation for C ke became
þ ¼  ð22Þ
@t @z qg qg Dt
C ke ðz; tÞ ¼ C ke ðzÞ þ dC ke ðtÞ
and
¼ je ðl; tÞ þ V gje þ Xe ðz  lÞ
@ð1  aÞ @ð1  aÞv f Cf ð1  aÞ Df qf h i
¼ v me ðl; tÞ þ q ðl;tÞ
qf
þ ¼  ð23Þ  1 V gje þ V gje þ Xe ðz  lÞ ð34Þ
@t @z qf qf Dt me

¼ v me ðlÞ þ qmefðlÞ V gje þ Xe ðz  lÞ þ dv me ðl; tÞ

q
By adding the above two equations and neglecting the com- qf
pressibility effect, the volumetric flux equations could be obtained  2 V gje dqme ðl; tÞ
½qme ðlÞ
as
where V gje was assumed as a constant in Region D. The steady
@je Dq
¼ Cg;fl ð24Þ state part and the perturbation part of C ke could be written as
@z qg qf
qf
where, the volumetric flux of the mixture was given by C ke ðzÞ ¼ v me ðlÞ þ V gje þ Xe ðz  lÞ ð35Þ
qme ðlÞ


qf
je ¼ v me þ  1 V gje ð25Þ qf
qme dC ke ðtÞ ¼ dv me ðl; tÞ  2
V gje dqme ðl; tÞ ð36Þ
½qme ðlÞ
By considering the isochoric process for each phase, Eq. (22)
could be expressed by je and V gje as And the perturbation of mixture velocity and mixture density at the
@a @ a @ðaV gje Þ qme Cg;fl exit of the heated section was given in Ishii’s thesis (Ishii, 1971) as
þ je þ ¼ ð26Þ

@t @z @z qf qg Ac
dv me ðl; tÞ ¼ eest K5 ðl; sÞ ð37Þ
Furthermore, by considering the vapor drift velocity V gj is a Ae
function of void fraction only, Eq. (26) could be re-written as
  dqme ðl; tÞ ¼ eest qf K6 ðl; sÞ ð38Þ
@a @ðaV gje Þ @ a qme Cg;fl
þ je þ V gje þ a ¼ ð27Þ
@t @a @z qf qg So

If we defined the kinematic wave velocity C ke by dC k;e ðtÞ ¼ eest K20 ðl; sÞ ð39Þ
@ðaV gje Þ and the various transfer functions of Kn ðz; sÞ are given in Appen-
C ke ¼ je þ V gje þ a  je þ V gje ð28Þ
@a dix A.
220 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225

2.6.3. Response of mixture density t ¼ s3 þ Ee ðzÞ  Ee ðlÞ ð53Þ

By now, the kinematic wave velocity was solved and could be
The second equality of Eq. (50) could be written as
substituted into Eq. (30) to obtain the density of the mixture. To
dC ke ðtÞ
simplify the derivation, a new variable was defined in this section dðd/e Þ ¼ 2 dC ke ðzÞ ¼ eest K20 ðl; sÞ 1
2 dC ke ðzÞ
½C ke ðzÞ ½C ke ðzÞ
as
K20 ðl;sÞ
ð54Þ
  ¼ eess3 es½Ee ðzÞEe ðlÞ 2 dC ke ðzÞ
q ðz; tÞ ½C ke ðzÞ
/e ðz; tÞ ¼ ln me ð40Þ
qme ðlÞ By defining He ðz; sÞ as
So the density propagation Eq. (30) became Z
1
He ðz; sÞ ¼ es½Ee ðzÞEe ðlÞ 2
dC ke ðzÞ ð55Þ
@/e ðz; tÞ @/ ðz; tÞ ½C ke ðzÞ
þ C ke ðz; tÞ e ¼ Xe ð41Þ
@t @z
Integrating Eq. (54) from l to z, the following equation could be
In order to apply perturbation method, we defined obtained.
/e ðz; tÞ ¼ /e ðzÞ þ d/e ðz; tÞ ð42Þ d/e ðz; s3 Þ  d/e ðlÞ ¼ eess3 K20 ðl; sÞ½He ðz; sÞ  He ðl; sÞ ð56Þ
and Thus /e ðz; s3 Þ in the Region D was expressed as
qme ðz; tÞ ¼ qme ðzÞ þ dqme ðz; tÞ ð43Þ /e ðz; s3 Þ ¼ /e ðzÞ þ d/e ðz; s3 Þ
h i
ðlÞ
By using the order of magnitude analysis, /e ðzÞ and d/e ðz; tÞ ¼ ln CC keðzÞ þ fd/e ðlÞ þ eess3 K20 ðl; sÞ½He ðz; sÞ  He ðl; sÞg
ke
could be expressed as
  ð57Þ
qme ðzÞ
/e ðzÞ ¼ ln ð44Þ And the boundary condition for the /e ðz; tÞ was given as
qme ðlÞ
/e ðl; s3 Þ  0 ð58Þ
dq ðz; tÞ
d/e ðz; tÞ ¼ me ð45Þ Then
qme ðzÞ
d/e ðlÞ ¼ 0 ð59Þ
Substituting Eqs. (42) and (34) into Eq. (41) and using the per-
turbation method, the zeroth order and first order equations were And d/e ðz; tÞ in Eq. (57) could be written as
expressed as
d/e ðz; tÞ ¼ eest es½Ee ðzÞEe ðlÞ K20 ðl; sÞ½He ðz; sÞ  He ðl; sÞ ð60Þ
d/e ðzÞ
C ke ðzÞ ¼ Xe ð46Þ Finally, the response of mixture density qme ðz; tÞ could be
dz obtained from Eqs. (45), (49) and (60) as
and qme ðz;tÞ ðzÞ ðz;tÞ
qme ðlÞ ¼ qqme
me ðlÞ
þ dqqme
me ðlÞ
@d/e ðz; tÞ @d/e ðz; tÞ Xe ð61Þ
þ C ke ðzÞ ¼ dC ke ðtÞ ð47Þ ðlÞ ðz;tÞ qme ðzÞ ðlÞ ðlÞ
@t @z C ke ðzÞ ¼ CCkeðzÞ þ dqqme
me ðzÞ qme ðlÞ ¼ CCkeðzÞ þ CCkeðzÞ d/e ðz; tÞ
ke ke ke

Integrating Eq. (46) from l to z, then where

Z Z
z
1 z
dC ke ðzÞ dqme ðz; tÞ
/e ðzÞ  /e ðlÞ ¼ Xe dz ¼  ¼ eest K21 ðz; sÞ ð62Þ
l C ke ðzÞ l C ke ðzÞ qme ðlÞ
" #
C ke ðzÞ
¼  ln ð48Þ 2.6.4. Center of mass velocity
C ke ðlÞ
In this section, the solution of the mixture velocity was
Comparing Eqs. (44) and (48), the following equation could be obtained after the solutions for the volumetric flux je and mixture
obtained. density qme in the previous sections.


qme ðzÞ C ke ðlÞ qf
¼ ð49Þ v me ¼ je   1 V gje ð63Þ
qme ðlÞ C ke ðzÞ qme
The solution of the perturbed part for the mixture density could be Because C ke ¼ je þ V gje , then
solved by transforming Eq. (47) to Lagrangian form as h i
v me ðz; tÞ ¼ C ke ðz;tÞ  V gje 
qf
qme ðz;tÞ  1 V gje
dz C ke ðzÞdðd/e Þ qf
dt ¼ ¼ ð50Þ ¼ C ke ðzÞ þ dC ke ðtÞ  q ðz;tÞ V gje
C ke ðzÞ dC ke ðtÞXe h i hme
i
qf q
¼ C ke ðzÞ  qme V gje þ dC ke ðtÞ þ qmef dqqme
me
V gje
If the particle entered the bottom of the chimney at t ¼ s3 , the h i h i
q q
first equality in Eq. (50) could be integrated as ¼ C ke ðzÞ  qmefðlÞ qqmemeðlÞ V gje þ dC ke ðtÞ þ qmefðlÞ qqmemeðlÞ dqqme me
V gje
Z z h i h i
dz ðlÞC ke ðzÞ ðlÞC ke ðzÞ dqme qme ðlÞ
t  s3 ¼ ð51Þ ¼ C ke ðzÞ  CC k ðkÞC V gje þ dC ke ðtÞ þ CC k ðkÞC q ðlÞ q V gje
l C ke ðzÞ h k ke ðlÞ
i h k ke ðlÞ me me
i
ðlÞC ke ðzÞ C k ðlÞC ke ðzÞ dqme C ke ðzÞ
The right hand side of Eq. (51) was redefined as
¼ C ke ðzÞ  CC k ðkÞC ðlÞ
V gje þ dC ke ðtÞ þ
C ðkÞC ðlÞ qme ðlÞ C ðlÞ
V gje
k ke k ke ke
Z ð64Þ
dz
Ee ðzÞ ¼ ð52Þ
C ke ðzÞ By noticing C k ðlÞ ¼ C ke ðlÞ at the boundary between Regions C
Thus Eq. (51) became and D, Eq. (64) could be written as
 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225 221

h i
v me ðz;tÞ C ke ðzÞ dC ke ðtÞ C ke ðzÞ C ke ðzÞ 2.7. Dimensionless characteristic functions for the whole system
v fi ¼ C k ðkÞ
þ v fi þ v1
fi C k ðkÞ C ke ðlÞ
eest K21 ðz; sÞV gje
ð65Þ
ðzÞ
þ eve K22 ðl; sÞ
st
¼ CCkeðkÞ The perturbations for the pressure drops in Regions A, B, and C
k fi
and inlet velocity had been derived by Ishii (1971) from the field
equations and non-dimensionalized as
2.6.5. Pressure drop of the downstream un-heated region D with For the upstream un-heated Region A
flashing

2
dDp01 lo 1  fo  1
The pressure drop response in the un-heated region could be ¼ 2k þ s þ 2l ð76Þ
dv  v fi Ao v fi
 
Ao
 i  o
2Do
obtained by integrating the momentum equation in this region.
Z 
 For the liquid heated Region B
lþle
@ v me @ v me
Dp34 ¼ ke qme ðlÞv 2me ðlÞ þ qme þ v me þ g e qme

v fi f s  
@t @z dDp12 fs
l
!# ¼ s s
þ 2k
þ þ v ð1  K3 Þ ð77Þ
dv  v fi 
NFr 2Do fi
12
f @ qf  qme qf qg 2 2Do
þ me qme v 2me þ V dz ð66Þ
2De @z qme  qg qme gje For the heated mixture Region C
n h  i  o
dDp23
¼ ln C r ð1 þ V gj Þ þ V gj s11 s K3  V gj s11 s C 2
2
In real reactor, such as NMR, other terms could be neglected
dv  v 
fi
except the exit throttling ke and the gravitational term. The gravi- n    h V 
i o
1  K3 þ ln C r ss1 K3  s11  1þVgj  s11 C 2
2
tational pressure drop was much bigger than other terms due to þ  1þV 1

gj gj
long chimney region design in natural circulation BWR.  v  n    o
K3
R lþle R lþle þ NFrfi 1  C1 s 1  s1 s11 C 1  ð1  K3 Þ
n      o
r
Dp34g ¼ g e qme dz ¼g e ðqme þ dqme Þdz f m 
 v fi 1 þ 2V gj s11 s12 C 3
l l K3
R lþle R lþl þ 2D  ð1  k Þ 2s  1þV 1

s 1
C ke ðlÞ
¼ ge l q me ðlÞ
þ g e l e dqme dz
C ke ðzÞ
dz 2
gj

C
R lþl ðlÞ
R lþl ð67Þ fs
 2D  v 
fi ð1

K
3Þ
 V gj
q
g 1þV  ð2C r  1Þ s 1
4

¼ g e qme ðlÞ l e CC keðzÞ dz þ g e l e qme ðlÞeest K21 ðz; sÞdz gj

hR ke
i ð78Þ
lþl ðlÞ
¼ g e qme ðlÞ l e CCkeðzÞ dz þ eest K23 ðz; sÞ
ke Then the total pressure drop perturbations can be expressed as
and dDpex ¼ dDp01 þ dDp12 þ dDp23 þ dDp34 ð79Þ
Z lþle
C ke ðlÞ
Dp34g ¼ g e qme ðlÞ dz ð68Þ
l C ke ðzÞ 3. Experimental stability maps

dDp34g ¼ eest K23 ðz; sÞ ð69Þ The experimental stability boundary had been obtained from
previous quasi-steady state tests using the natural circulation test
and facility as shown in Fig. 5. This test facility was designed and con-
structed to perform both transient and quasi-steady flow instabil-
qme ðlÞ C ke ðl þ le Þ ity tests for the NMR. The detailed scaling analysis and design
C re ¼ ¼ ð70Þ
qme ðl þ le Þ C ke ðlÞ could be referred to our previous work (Shi et al., 2015a,b). For
the quasi-steady tests, the test power and inlet flow conditions
1 were important boundary conditions. The maximum power pro-
s34 ¼ Ee ðl þ le Þ  Ee ðlÞ ¼ ln C re ð71Þ vided by four electric heaters (2  2) is 20 kW, which was powered
Xe
with a three-phase 480VAC source and controlled by Watlow SCR
The dimensionless form of Eq. (69) was given as power controller with an accuracy of ±0.5% of its actual power out-
8 put. A similar tube-in-tube exchanger installed on the top of the
>
>
dDp34g 1 X <1 
1
downcomer worked as a cooler to reduce the temperature of the

¼    K20
e   separated flow. Another three-phase preheater of 18 kW was
vv
d  fi C r s  Xe >
>
:
N Fr  
ðA1 ÞC r þ V gje C r þ Xvele installed in the downcomer to adjust the inlet flow temperatures.
e fi
" #) The designed system pressure was 1 MPa and can be controlled
 by the condenser, which was connected to the steam dome on
g  
þ   ð1  es s34 Þ ð72Þ
s v fi the top and the downcomer at the bottom. Besides, this test facility
was equipped with instruments to obtain various thermal hydrau-
and lic parameters, such as local temperatures, absolute pressures,
( ) pressure differentials, local void fractions, and mass flow rates
 1 V gje 1 1      (Shi et al., 2015a,b).
K ¼  K5  C 2 K3  e½ðs þ1Þs23 þs s12  ð73Þ
20
Ae r
1 þ V gj s  1 C 2
r
The quasi-steady tests had been performed under various con-
ditions by changing system pressure, core inlet subcooling, core
inlet flow resistance coefficient, and core heat power etc. Based
1 V gj n     
o
K5 ¼ K3 þ  K3  e½ðs 1Þs23 þs s12  ð74Þ on certain stability criteria, the flow stability for a certain operating
s  1 1 þ V gj condition could be determined from the time trace profile of the
  loop natural circulation flow rate, which was defined as the
 
1  es s12 single-phase core inlet flow velocity. All those stable or unstable
K3 ¼ 1  ð75Þ experimental conditions were plotted in the dimensionless stabil-
s ity plane, which also forms an experimental stability map.
where the characteristic functions K3 and K5 were derived by A typical experimental stability map at a pressure of 200 kPa
Ishii (1971). Also some dimensionless parameters were listed in can be seen in Fig. 6. A group of experimental conditions connected
Appendix B. by a dash line stands for a fixed core heat flux during the
222 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225

Fig. 5. Schematic of the steady state facility.

change number were largely reduced in this region. In addition,

the stable conditions are marked with the blue color, while the
unstable ones are marked with the red color in Fig. 6. The intermit-
tent flow oscillations were mainly caused by the flashing instabil-
ity occurring during the transfer between the stable single-phase
natural circulation and stable two-phase natural circulation. As
can be seen, the stability boundary in this case were very close
to the zero-quality line at the riser or chimney outlet. Since the
zero-quality line at the core exit was derived based on the assump-
tion of the thermal and mechanical equilibrium conditions, the
shift of stability boundary could be caused by the non-
equilibrium phenomenon of subcooled boiling in the core section.
Similarly, Fig. 7 shows the experimental stability map at 400 kPa.
The general trends were very similar to the stability map at
200 kPa. However, the stability boundary moved to the zero-
quality line at the core exit. The flashing effect was highly sup-
pressed at a higher pressure due to reduced thermal non-
equilibrium. Moreover, the whole unstable region of the transition
phase was very thin in the experimental stability plane at the pres-
sure of 400 kPa.

4. Application of frequency domain analysis

Fig. 6. Experimental stability at 200 kPa.
To verify and validate the flashing model, the frequency domain
analysis was implemented to the natural circulation test facility
quasi-steady state tests. At a fixed core heat flux, the flow condi- designed to investigate the flow instability for the NMR under
tions went through the single-phase natural circulation in high- the low power and low pressure conditions. The geometric param-
subcooling region, flashing instability, and two-phase natural cir- eters and operation conditions of the test facility were used as the
culation in low-subcooling region. Due to the significant increase system input parameters of the numerical stability analysis. The
of the natural circulation rate in the two-phase flow, the phase inlet flow conditions and system power were determined by the
 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225 223

the zero-quality line at the core exit, where two dimensionless

numbers are equal, is also included in the stability plane. In the
case of K in = 600, two boundaries could be obtained, including a
flashing boundary and a density wave oscillations (DWO) bound-
ary, in current ranges of the plane. The flashing boundary was
located closely below the zero-quality line at the core exit, while
the DWO boundary moved to the right of the stability plane with
higher phase change number. In the case of K in = 1200, only a flash-
ing boundary could be obtained in current range of the stability
plane. The DWO boundary for Kin = 1200 moved to the right part
with higher phase change number, which exceed the ranges of cur-
rent dimensionless stability plane. In other words, the stable
region was expanded by increasing the inlet flow resistance coeffi-
cient, which was also verified by Ishii (1971).
The pressure effects on the theoretical stability boundary was
also carried out in the frequency domain analysis. Fig. 9 shows
the stability boundary at the system pressure of 400 kPa with a
flow resistance of K in = 600. As can be seen in Fig. 9, only the
DWO boundary could be numerically predicted in the dimension-
less stability plane. Moreover, the pressure effect on the DWO
boundary was not significant, which has also been confirmed by
other researchers (Ishii, 1971). By comparing with the experimen-
tal stability boundary in Fig. 7, the flashing boundary at high pres-
Fig. 7. Experimental stability at 400 kPa.
sure still existed and could be presented by the zero-quality line at
the core exit by comparing with experimental data at 400 kPa.
subcooling number and phase change number, respectively. By Although the theoretical stability boundaries including the
applying the linear stability analysis, the theoretical stability DWO boundary and flashing boundary can be obtained from the
boundary could be determined numerically by using the D- frequency domain analysis. Uncertainties of the flashing boundary
partition method on the dimensionless characteristic equation still exists due to the complexities of the flow instability itself. The
for the perturbations between the total pressure drop and inlet flashing boundary predicted in this study is similar to the Type-I
velocity. In addition, the parametric stability analysis were per- flow instability observed from other researchers (Van Bragt and
formed on the effects of system pressure and inlet flow resistance Van der Hagen, 1998; Van der Hagen et al., 1997; Wang et al.,
etc. on the flow instability. It should be noted that the drift velocity 1994). In current study, the theoretical flashing stability boundary
was neglected in current analysis since flashing occurs at very low in Fig. 8 does not agree very well with the one in Fig. 6 from the
void fraction conditions, which was almost single-phase natural quasi-steady state tests, where the flashing boundary is above
circulation observed from our previous measurements (Shi et al., the zero quality line at the core exit. One possible reason is that
2015a,b). the thermal non-equilibrium conditions, i.e., subcooled boiling, is
Fig. 8 shows the numerical stability map at two different core not yet considered in current frequency domain analysis. Thus
inlet flow resistance coefficients, i.e., K in = 1200 and 600, at a sys- the non-boiling length k can be overestimated in current frequency
tem pressure of 200 kPa. As can be seen, the stability boundaries domain analysis considering thermal equilibrium conditions. Some
are plotted in the dimensionless plane of Nsub  Npch. In addition, preliminary analysis has been done to justify this claim by

Fig. 8. Stability Map at 200 kPa. Fig. 9. Stability Map at 400 kPa.
224 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225

8 " #Xs 1 9
manually reducing the non-boiling length in the frequency domain
V gj < X C k ðkÞ 
=
analysis code. However, a more detailed analysis by taking account K5 ðz; sÞ ¼ K3 ðsÞ þ K3 ðsÞ  ess12 ðA:3Þ
 :
s  X C k ðkÞ C k ðzÞ ;
of the thermal non-equilibrium will be performed in the future.

5. Conclusions K6 ðl; sÞ ¼ K4 ðl; sÞ ðA:4Þ

  
Flashing instability is normally caused by the reduced hydro- Ac qf 2
K20 ðl; sÞ ¼ K5 ðl; sÞ  V K ðl; sÞ ðA:5Þ
static head along the riser or chimney in the natural circulation Ae qme ðlÞ gje 6
system. In order to predict the flashing stability, the frequency
domain analysis was performed to obtain the stability boundary C ke ðlÞ
in the dimensionless stability plane with the newly developed K21 ðz; sÞ ¼ es½Ee ðzÞEe ðlÞ K20 ðl; sÞ½He ðz; sÞ  He ðl; sÞ ðA:6Þ
C ke ðzÞ
flashing model. Following Ishii’s approach to obtain the density
wave oscillations (DWO) boundary, the coolant enthalpy change 2
½C ke ðzÞ
rate during the flashing was considered as an axially uniform heat K22 ðl; sÞ ¼ K20 ðl; sÞ þ K21 ðz; sÞV gje ðA:7Þ
source based on the water P-T diagram. The kinematic and dynam- C k ðkÞC ke ðlÞ
ics of the downstream unheated mixture region (riser or chimney)
R lþle
were obtained. In addition, the pressure drop response was given K23 ðz; sÞ ¼ g e qme ðlÞ K21 ðz; sÞdz
in dimensionless equations for the stability analysis. The system
l
h i
Xe
¼ g e qme ðlÞK20 ðl; sÞ C ðlÞC
le
 sXe  s ð1  e
1 ss34
Þ sXXe e
transfer functions were obtained for the pressure perturbations ke re

with the inlet flow velocity perturbations. The stability boundary ðA:8Þ
were obtained by determining the nature of roots of the system
characteristic equations with D-partition method.
In the dimensionless stability plane of Nsub  Npch, two kinds of Appendix B. Dimensionless numbers and variables
stability boundaries, i.e., the flashing instability boundary and the
density wave oscillations boundary, were predicted using fre- The dimensionless numbers and variables derived by Ishii
quency domain analysis for the natural circulation test facility. (1971) are given as follows:
Among them, DWO boundary has been widely studied and con- Dp
firmed in current analysis. The theoretical flashing instability Dp ¼ 2
ðB:1Þ
ðXlÞ qf
boundary was closely parallel to the zero-quality line at the core
exit at the pressure of 200 kPa. However, the theoretical flashing
C k ðlÞ
boundary was not able to be numerically predicted at the pressure C r ¼ ðB:2Þ
of 400 kPa, which was considered to match with the zero-quality C k ðkÞ
line under high pressure. In addition, the predicted stability
boundaries were compared with the quasi-steady state experi- Xe
Xe ¼ ðB:3Þ
mental data acquired from the natural circulation test facility. X
The biggest difference was that the experimental flashing bound-
s
ary was located above the zero-quality line at the core exit, while s ¼ ðB:4Þ
the theoretical boundary was parallel below the line. The possible
X
reason was that the thermal equilibrium was assumed in current q00W n Dq
study, whereas the subcooled boiling in the core section could X ¼ Xo ¼ ðB:5Þ
Ac Difg qg qf
occur in real quasi-steady state tests. Although the theoretical
flashing boundary showed some discrepancy with the experimen-
tal data, the effect of system pressure and inlet K factor could be
v 2fi
NFr ¼ ðB:6Þ
correctly simulated by using current frequency domain analysis gl
for the natural circulation test facility. In the future, the frequency
Ac
domain analysis including both the flashing model and thermal Ae ¼ ðB:7Þ
non-equilibrium will be carried out and validated against existing Ae
instability experimental data.
 le
le ¼ ðB:8Þ
l
Acknowledgements
V gje
V gje ¼ ðB:9Þ
This material is based upon work supported under a Depart- v fi
ment of Energy Nuclear Energy University Program.
v fi
v fi ¼ ðB:10Þ
Appendix A. Various transfer functions Xl
g
The various transfer functions during the derivations are given g ¼ ðB:11Þ
X2 l
by Ishii (1971) as follows:
Xð1  ess12 Þ k Di12 Dq
K3 ðsÞ ¼ 1  ðA:1Þ s12 ¼ s12 X ¼ ¼ ðB:12Þ
s v fi Difg qg
8" #2 " #Xs þ1 9
1 X < C k ðkÞ C k ðkÞ 12
ss
= s23 ¼ s23 X ¼ ln C r ðB:13Þ
K4 ðz; sÞ ¼ K3 ðsÞ  e ðA:2Þ

C k ðkÞ s  X : C k ðzÞ C k ðzÞ ;
s34 ¼ s34 X ðB:14Þ
 S. Shi, M. Ishii / Annals of Nuclear Energy 101 (2017) 215–225 225


Xe le cpf DT Dq Shi, S., 2015. Investigation of Natural Circulation Instability and Transients in
Nfl ¼  ð1  aÞ ðB:15Þ
v 
fi Difg qg Passively Safe Novel Modular Reactor Dissertation. Purdue University, West
Lafayette, IN.
Shi, S., Brooks, C., Eoh, J., Ishii, M., 2014. Pressurized startup transient analyses for
the BWR-type NMR-50. In: Transactions-American Nuclear Society, vol. 111, pp.
References 1613–1615.
Shi, S., Schlegel, J.P., Brooks, C.S., Lin, Y.-C., Eoh, J., Liu, Z., Zhu, Q., Liu, Y., Hibiki, T.,
Achard, J.L., Drew, D.A., Lahey, R.T., 1981. The effect of gravity and friction on the Ishii, M., 2015a. Experimental investigation of natural circulation instability in a
stability in a channel. Chem. Eng. Commun. 11 (1-3), 59–79. BWR-type small modular reactor. Prog. Nucl. Energy 85, 96–107.
Furuya, M., Inada, F., van der Hagen, T.H.J.J., 2005. Flashing-induced density wave Shi, S., Wu, Z., Liu, Z., Schlegel, J.P., Brooks, C.S., Eoh, J., Yan, Y., Liu, Y., Yang, W.S.,
oscillations in a natural circulation BWR-mechanism of instability and stability Ishii, M., 2015b. Experimental investigation of natural circulation instability
map. Nucl. Eng. Des. 235 (15), 1557–1569. with void-reactivity feedback during startup transients for a BWR-type SMR.
Guo, X., Sun, Z., Wang, J., Yu, S., 2016. Steady-state performances and scaling Prog. Nucl. Energy 83, 73–81.
analyses for an open flashing-driven natural circulation system. Prog. Nucl. Shi, S., Lin, Y.-C., Yang, W.S., Ishii, M., 2015c. Experimental stability maps for a BWR-
Energy 87, 1–14. type small modular reactor. In the 16th International Topical Meeting on
Hu, R., Kazimi, M.S., 2011. Flashing-induced instability analysis and the start-up of Nuclear Reactor Thermal Hydraulics (NURETH-16), Chicago, IL.
natural circulation boiling water reactors. Nucl. Technol. 176 (1), 57–71. Shi, S., Hibiki, T., Ishii, M., 2016. Startup instability in natural circulation driven
Inada, F., Furuya, M., Yasuo, A., 2000. Thermo-hydraulic instability of boiling natural nuclear reactors. Prog. Nucl. Energy 90, 140–150.
circulation loop induced by flashing (analytical consideration). Nucl. Eng. Des. Su, G., Jia, D., Kenji, F., Guo, Y., 2002. Theoretical and experimental study on density
200 (1–2), 187–199. wave oscillation of two-phase natural circulation of low equilibrium quality.
Ishii, M., 1971. Thermally Induced Flow Instabilities in Two-Phase Mixtures in Nucl. Eng. Des. 215 (3), 187–198.
Thermal Equilibrium Dissertation. Georgia Institute of Technology. Van Bragt, D.D.B., Van der Hagen, T.H.J.J., 1998. Stability of natural circulation
Ishii, M., Shi, S., Yang, W.S., Wu, Z., Rassame, S., Liu, Y., 2015. Novel modular natural boiling water reactors: Part I – description stability model and theoretical
circulation BWR design and safety evaluation. Ann. Nucl. Energy 85, 220–227. analysis in terms of dimensionless groups. Nucl. Technol. 121 (1), 40–51.
Lee, J.D., Pan, C., Chen, S.W., 2015. Nonlinear dynamic analysis of a two-phase Van der Hagen, T.H.J.J., Van Bragt, D.D.B., van der Kaa, F.J., Karuza, J., Killian, D.,
natural circulation loop with multiple nuclear-coupled boiling channels. Ann. Nissen, W.H.M., Stekelenburg, A.J.C., Wouters, J.A.A., 1997. Exploring the
Nucl. Energy 80, 77–94. Dodewaard type-I and type-II stability; from start-up to shut-down, from
Lisowski, D.D., Scherrer, R.M., Haskin, T.C., Anderson, M.H., Corradini, M.L., Albiston, stable to unstable. Ann. Nucl. Energy 24 (8), 659–669.
S.M., Tokuhiro, A., 2011. Experimental studies of NGNP reactor cavity cooling Wang, F.-S., Hu, L.-W., Pan, C., 1994. Thermal and stability analysis of a two-Phase
system with water. In: Proceedings of ICAPP, p. 2851. France. natural circulation loop. Nucl. Sci. Eng. 117 (1), 33–46.
Manera, A., Rohde, U., Prasser, H.-M., van der Hagen, T.H.J.J., 2005. Modeling of Wu, Z., Yang, W.S., Shi, S., Ishii, M., 2016. A core design study for a small modular
flashing-induced instabilities in the start-up phase of natural-circulation BWRs boiling water reactor with long-life core. Nucl. Technol. 193 (3).
using the two-phase flow code FLOCAL. Nucl. Eng. Des. 235 (14), 1517–1535. Zuber, N., 1967. Flow excursions and oscillations in boiling, two-phase flow systems
Odeh, F.Y., Yang, W.S., 2016. Core design optimization and analysis of the Purdue with heat addition. In: Symposium on Two-phase Flow Dynamics. pp. 1071–
Novel Modular Reactor (NMR-50). Ann. Nucl. Energy 94, 288–299. 1089.
Saha, P., 1974. Thermally Induced Two-Phase Flow Instabilities, Including the Effect
of Thermal Non-equilibrium between the Phases Dissertation. Georgia Institute
of Technology.