International Journal of Thermal Sciences 98 (2015) 312e331

Contents lists available at ScienceDirect

International Journal of Thermal Sciences

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

Linear stability analysis of flow instabilities with a nodalized reduced

order model in heated channel
Subhanker Paul, Suneet Singh*
Department of Energy Science and Engineering, Indian Institute of Technology-Bombay, Mumbai 400076, India

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

Article history: The prime objective of the presented work is to develop a Nodalized Reduced Order Model (NROM) to
Received 14 October 2014 carry linear stability analysis of flow instabilities in a two-phase flow system. The model is developed by
Received in revised form dividing the single phase and two-phase region of a uniformly heated channel into N number of nodes
20 June 2015
followed by time dependent spatial linear approximations for single phase enthalpy and two-phase
Accepted 13 July 2015
quality between the consecutive nodes. Moving boundary scheme has been adopted in the model,
Available online 14 August 2015
where all the node boundaries vary with time due to the variation of boiling boundary inside the heated
channel. Using a state space approach, the instability thresholds are delineated by stability maps plotted
Keywords:
Nodalized reduced order model
in parameter planes of phase change number (Npch) and subcooling number (Nsub). The prime feature of
Density wave oscillation the present model is that, though the model equations are simpler due to presence of linearelinear
Forced circulation approximations for single phase enthalpy and two-phase quality, yet the results are in good agreement
Natural circulation loop with the existing models (Karve [33]; Dokhane [34]) where the model equations run for several pages
Phase change number and experimental data (Solberg [41]). Unlike the existing ROMs, different two-phase friction factor
Subcooling number multiplier correlations have been incorporated in the model. The applicability of various two-phase
friction factor multipliers and their effects on stability behaviour have been depicted by carrying a
comparative study. It is also observed that the Friedel model for friction factor calculations produces the
most accurate results with respect to the available experimental data.
© 2015 Elsevier Masson SAS. All rights reserved.

1. Introduction
Over the past several years it is evident that two phase flow
instabilities are of paramount interest in thermal hydraulic studies
[1e6]. Among these instabilities, DWO is a major constituent,
which can arise in a boiling system when a constant pressure drop
[7,8] is imposed across a heated channel. Predominantly in nuclear
reactors, this phenomenon can be triggered both in fuel channels
and in steam generators. However, the DWO phenomenon in nu-
clear reactors is also caused due to the coupled neutronic
feedbacks.
The DWO phenomenon can be interpreted as the origin of
instability associated with density perturbations in a fluid with
delay propagations, causing waves of higher and lower density fluid
flowing across a heated channel [5,8]. Broadly speaking, the dif-
ference between inlet and exit fluid density in a boiling system
triggers delays in the transient distribution of pressure drops across
* Corresponding author. Tel.: þ91 22 2576 7843; fax: þ91 22 2576 4890.
E-mail address: suneet.singh@iitb.ac.in (S. Singh).
the channel which may lead to self-sustained oscillations that are
known as DWOs. In addition to these, the flow instabilities in a
natural circulation system are broadly classified into two cate-
gories, namely Type-I and Type-II [9e20]. The Type-I instability
arises at low power conditions where the steam quality is low.
During a low steam quality situation, a slight change in the quality
due to any disturbance causes a large variation in void fraction and
consequently the driving head changes. As a result the flow fluc-
tuations occur. But at high power conditions, when the void frac-
tion and steam quality is large, the frictional pressure losses
increase. Moreover, a large void fraction generates void propaga-
tion time delay in the two-phase region of the system. Under these
situations, any disturbance in the flow can cause large fluctuations
in the fluid density as well as frictional pressure drops. These
fluctuations eventually induce the coolant flow rate to oscillate,
which is classified as Type-II instability. It is worth noting that the
Type-II instability in natural circulation system is classified as
Density wave oscillation (DWO). These instabilities are strongly
undesirable in aforementioned systems, because continual flow
oscillations can cause mechanical vibration to the constituent
components thereby affecting the system performance.

http://dx.doi.org/10.1016/j.ijthermalsci.2015.07.027
1290-0729/© 2015 Elsevier Masson SAS. All rights reserved.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 313

Nomenclature m* boiling boundary (m)

r* density (kg/m3)
A* cross section area (m2)
B*t loop width (horizontal section, m) Subscripts
C0 void distribution parameter 1F single phase
D*h hydraulic diameter of flow channel (m) 2F double phase
D*doc diameter of downcomer (m) doc downcomer
D*r diameter of riser (m) ex exit
f friction factor f liquid
Fr Froude number ðv*2 * *
0 =g Lch Þ g vapour
g* acceleration due to gravity (m/s2) in inlet
h* enthalpy (kJ/kg) m mixture
Ht* downcomer level (m) r riser
J Jacobian matrix
Kinlet inlet pressure loss coefficient Superscripts
Kexit exit pressure loss coefficient * dimensional quantity
L*ch total length of flow channel (m) ~ steady state value
L*r length of riser (m)
L*h length of standpipe (m) Abbreviations
Nf friction number ðfL*ch =2D*h Þ AHWR advanced heavy water reactor
phase change number ðq * Dr* x*h L*ch =A* Dh*fg v*0 r*g r*f Þ
00
Npch BWR boiling water reactor
Nsub subcooling number ððh*sat  h*inlet ÞDr* =Dh*fg r*g Þ DFM drift flux model
Ns number of nodes in single phase DR decay ratio
Nt number of nodes in two-phase DWO density wave oscillation
00
q* wall heat flux (W/m2) HEM homogeneous equilibrium model
*
t time (s) HPNCL high pressure natural circulation loop
v*0 reference velocity (m/s) NIST national institute of standards and technology
v*inlet inlet velocity of coolant (m/s) NROM nodalized reduced order model
Vgj * average drift velocity (m/s) ODE ordinary differential equation
x steam quality PDE partial differential equation
z* distance along the axis of flow channel (m) ROM reduced order model
x*h heated perimeter (m) SB stability boundary
a void fraction

Extensive studies have been carried out to analyze the discussed
instability phenomenon (DWO). To serve the purpose, two general
approaches have been adopted by several researchers namely fre-
quency domain approach and time domain approach. The frequency
domain approach [21] is associated with control-theory techniques,
where transfer functions are obtained using Laplace transformation
of the linearized model of governing equations. The stability of the
system (linear stability characteristics with small perturbation) is
then determined by the analyses of roots of the obtained transfer
functions [9,10,12,13,15,22]. However, it is pointed out that this
technique is not very useful for analyzing the nonlinear stability
characteristics (for large perturbation) of the system due to the
presence of linearized equations. Moreover, in time domain
approach, the conservation equations are either analytically inte-
grated [23e26] over the competing regions, or by applying suitable
numerical techniques, 1-D analyses [27e29] are carried out. In this
type of analysis, the steady state of the system is perturbed and the
system evolution is observed with stepwise gradual change in
operating parameters. The points where undamped or diverging
oscillations are obtained are noted as the stability thresholds. It is
also pointed out that using these methods, the nonlinear stability
analysis of the system is computationally intensive. Because to gain
knowledge of nonlinear stability characteristics, the governing
equations need to numerically simulate for wide range of parame-
ters. In contrast, the study of nonlinear dynamics being a theme of
interest for last two decades, nonlinear stability analyses of heated
channels have become significantly advanced followed by bifurca-
tion analysis. Achard et al. [30] carried out an analytical bifurcation
study of DWO with homogeneous equilibrium model (HEM).
Rizwan-Uddin and Dorning [31] extended that approach by
replacing HEM by drift flux model which results even more
complicated nonlinear delay integro-differential equations. In their
study they showed the impact of void distribution parameter C0 and
the drift velocity Vgj between phases on the sensitivity of stability
boundary (SB) in various planes.
In addition to the mentioned techniques, Reduced Order Models
(ROMs) are also available where using a state space approach, the
linear as well as nonlinear stability characteristics of the system can
be determined. Clausse and Lahey [32] developed a simple model
using approximations for spatial dependence of the single phase
enthalpy and two-phase density. Followed by these developments
Karve [33], Dokhane [34] and Dokhane et al. [35] developed models
for DWO by introducing quadratic profiles of single phase enthalpy
and two-phase quality using HEM as well as drift flux model (DFM).
However due to the presence of quadratic profiles, the system
equations became highly complicated and even run for several
pages. It should be noted that using linearelinear approximations
of single-phase enthalpy and two-phase quality; a simpler model
compared to the aforementioned models had also been presented
by Karve et al. [36], where though the model equations remain
simpler, but the results obtained showed significant discrepancy
with the experimental data sets [37].
The main objective of the current research is to extend the
concept of the model developed by Paul and Singh [38], while
keeping a new insight to analyze the system stability characteristics
by diving the flow channel into N-number of nodes. To achieve the
314 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

goal, the two discrete regions of the channel, single phase and two- 4. Boussinesq approximation is valid for variation of density in the
phase, which are kept separated by a time dependent boiling two phase region.
boundary are divided into Ns- and Nt-number of nodes indepen- 5. Subcooled boiling is neglected.
dently. However, due to presence of time dependent boiling 6. vp/vt term in energy conservation equation is neglected.
boundary, the N node concept is adopted with moving boundary 7. Heat losses in the riser and downcomer are negligible.
scheme. The single phase enthalpy and two-phase quality are lin- 8. Properties of the fluid or mixture remain constant inside the
early approximated between two consecutive nodes and weighted riser and downcomer.
residual method is applied to the conservations equations (PDEs) to
obtain their set of time dependent coupled nonlinear ODEs. With these assumptions, the conservation equations of mass,
Applying state space approach, these nonlinear ODEs are used to momentum and energy for one-dimensional two-phase flow are
determine the stability of the system. The results of the present given by:
model are compared with various existing models [33,34,36] which Mass conservation
are closer to exact solutions [31]. The SBs obtained show good
v
* v

agreement with the said models. Moreover, due to presence of
*
r þ * r* v* ¼ 0 (1)
linear approximations for single phase enthalpy and two-phase vt vz
quality the equations remain far simpler as compared to these Momentum conservation in single phase
aforementioned models. In addition to these, in the existing models
  *
[31,33e36] the two-phase friction factor is approximated as twice vv* vv* vP1F f * * *2
the single-phase friction factor. However, no proper internalization r* þ v* * ¼  r v  r* g* (2a)
vt * vz vz* 2D*h
of two-phase friction factor calculation has been done. It is evident
form past researches [15] that the accurate prediction of two-phase Momentum conservation in two-phase
friction factor is very important because it significantly contributes
  *
towards the total pressure drop across the channel which helps to vv*m vv* vP2F f*
r*m þ v*m m ¼  m* r*m v*2 * *
m  rm g (2b)
determine the stability of the system. It is also found that the vt * vz* vz* 2Dh
prediction of two-phase friction factor is dependent on several
parameters [39] namely hydraulic diameter of the channel, flow Energy conservation
Reynolds number etc. Hence, in the present study, the perfor-
v
* * *  q * x*h
00
mances of different two-phase friction factor calculations have v
* *
r h þ r v h ¼ (3)
been compared. It is pointed out that though a few literatures [15] vt * vz* A*
are available where performance comparison of different two-
As shown in Fig. 1a and b, the single phase and two-phase region
phase frictions factor multipliers is done for the given system, but
of the channel are divided into Ns & Nt number of nodes indepen-
these studies only provided the comparison to steady state data.
dently. Similar to the previous model [38], along with the time
However, the effects on prediction of stability boundaries are not
dependent linearelinear approximations of single phase enthalpy
shown. Hence, in the present study, different two-phase friction
and two-phase quality between two consecutive nodes, here also
factor multipliers have been incorporated to develop the ROM, and
the density variation of coolant due to heating in the single phase
detailed analysis has been carried out to show their effects on
region has been taken into consideration. Weighted residual
prediction of stability boundaries.
method is applied to the mass and energy conservation equations to
reduce them into corresponding time dependent nonlinear ODEs.
2. Mathematical model
The dynamical system that results from the model is of the
following form,
The schematic of a uniformly heated channel is shown in Fig. 1a,
where the flow is driven by an applied external pressure drop _
XðtÞ ¼ FðXðtÞ; hÞ (4)
(forced circulation). In Fig. 1b, a high pressure natural circulation
loop is shown. The subcooled water comes to the inlet of the heated where X(t) is a vector of phase variables and h is the vector of design
channel through the downcomer. During its passage through the and operating parameters.
channel, it gains heat, boils and leave the channel as a two-phase
mixture through riser. At steam drum, the steam is separated and
 T
X ¼ a1;i1 ðtÞ; s1;j1 ðtÞ; vinlet ðtÞ with i ¼ 2 to Ns ; j ¼ 2 to Nt
the separated saturated water is mixed with feed water. More
detailed design configuration of similar kinds of systems can be  T
found in previously published literatures [10,12,13,15,19,20]. During h ¼ Nr ; Nr ; Npch ; Nsub ; Kinlet ; Kexit ; Fr; Nf ;1F ; Nf ;2F ; Ar
normal operating condition the steam drum is maintained at a
constant pressure (say 70 bar, a typical normal operating condition where Ns & Nt are the number of nodes in single phase and two-
of AHWR). The whole system is analyzed by a 1-D axial flow model phase region respectively.
using three basic conservation PDEs of mass, energy and mo- The dimensionless systems of ODEs that form this dynamical
mentum along the length of the channel. These PDEs are reduced to system are,
corresponding ODEs, using weighted residual procedure
[20,33,34,38]. Though in real situation the system dynamics is da1;i1 ðtÞ 2 h
¼ Nr Nr Npch  rinlet vinlet ðtÞa1;i1 ðtÞ
complex, but some simplifying assumptions are made [20,38] to dt bi1 ðLi ðtÞ  Li1 ðtÞÞ
develop the model that helps to simulate the stated phenomenon. i
þ gi1 ðtÞ ;
Assumptions used in this model are as follows,
(5)
1. Flow in the two phase region is homogeneous (no slip between
phases).
ds1;j1 ðtÞ f1;j1 ðtÞ þ f2;j1 ðtÞ þ f3;j1 ðtÞ
2. The system pressure remains constant. ¼ (6)
3. The input heat flux is uniform along the length of the channel.
dt f4;j1 ðtÞ
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 315

a b

Fig. 1. Schematic of a coolant channel: a) in forced circulation system, b) in natural circulation system.

dvinlet ðtÞ v v
¼ f7 ðtÞ (7) ½rðz; tÞhðz; tÞ þ ½rðz; tÞvðz; tÞhðz; tÞ ¼ Nr Nr Npch ; (11)
dt vt vz
where b, g(t), f1(t), f2(t), f3(t), f4(t) and f7(t) are defined in Appendix In the single phase region the velocity of the coolant satisfies
B. The details of derivation of aforementioned set of ODEs are given [38],
in next section.
rinlet vinlet
vðz; tÞ ¼ (12)
rðz; tÞ
2.1. ODE in single phase
Following the above definitions, Eq. (11) is further simplified to,
The energy balance equation in the single phase is written as,
v vhðz; tÞ
v
*  * *  *  * *  v
      ½rðz; tÞhðz; tÞ þ rinlet vinlet ¼ Nr Nr Npch ; (13)
r z ;t h z ;t þ * r* z* ; t * v* z* ; t * h* z* ; t * vt vz
vt * vz
where rinlet and vinlet are the non-dimensional density and velocity
00
q * x*h
¼ of the liquid at the inlet of the flow channel. The other non-
A*
(8) dimensional parameters Nr, Nr and Npch are defined in Appendix A.
In the above equation, enthalpy h(z,t) between two consecutive
The variation of density with increase in enthalpy is obtained by nodes is linearly approximated [20,38,40] as,
property evaluation using NIST fluid properties data sets followed
by regression analysis at a certain pressure (e.g. 70 bar) as, h ¼ hi1 þ a1;i1 ðtÞðz  Li1 ðtÞÞ (14)
*
r ¼ N1*  N2* h* (9) where h1 ¼ hinlet, L1(t) ¼ 0
*
where h is the enthalpy at any position in the single phase region
and hNs ¼ hsat, LNs(t) ¼ m(t)
and N1* and N2* are dimensional constants.
also hi ¼ hi1 þ 1=Ns  1ðhsat  hinlet Þ for i ¼ 2 to Ns
Eq. (9) can be written in a non-dimensional form [38] as,

r ¼ N1  N2 h (10) The position of each node in the single phase region is given as,

hi  hi1
N1* N2* Dh*fg Li ðtÞ ¼ Li1 ðtÞ þ (15a)
where N1 ¼ ; N2 ¼ a1;i1 ðtÞ
r*f r*f
Eq. (15) denotes that, the position of each node in the single
Following the above definitions, Eq. (8) is written in a non- phase region is not constant. Instead, they vary with time. The
dimensional form as, variation of node positions with time is given as,
316 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

  00
dLi ðtÞ dLi1 ðtÞ hi  hi1 da1;i1 ðtÞ vv*m z* ; t * q * x*h Dr*
¼  2 (15b) ¼ (19)
dt dt a1;i1 ðtÞ dt vz* r*f r*g A* Dh*fg

Eq. (19) can be written in a non-dimensional form as,

dL1 ðtÞ
where ¼0

dt vm ðz; tÞ ¼ vm;j1 ðtÞ þ Npch z  Lm;j1 ðtÞ (20)
In conventional fixed position nodalization scheme, the node
positions are first fixed, and then suitable approximations are where j ¼ 2 to Nt
considered between consecutive nodes. However, in the present
vm;1 ðtÞ ¼ rinlet vinlet
model, since the node positions continuously vary with time, hence
the present scheme is called Moving boundary scheme.
In a similar way, the boiling boundary which separates the Lm;1 ðtÞ ¼ LNs ðtÞ & Lm;Nt ðtÞ ¼ 1
single phase and two-phase region is given as,
The length of each node in the two phase region is given as,
hs  hNs1
mðtÞ ¼ LNs ðtÞ ¼ LNs1 ðtÞ þ (15c) 1  LNs ðtÞ
a1;Ns1 ðtÞ Lm;j ðtÞ ¼ Lm;j1 ðtÞ þ (21)
Nt  1
The mass balance equation can now be written in a non-
dmðtÞ dLNs ðtÞ dLNs1 ðtÞ hs  hNs1 da1;Ns1 ðtÞ
and ¼ ¼  2 dimensional form as
dt dt dt a1;Ns1 ðtÞ dt
(15d) vrm ðz; tÞ vr ðz; tÞ
þ vm ðz; tÞ m ¼ Npch rm ðz; tÞ (22)
vt vz
Now, weighted residual method is applied to the given energy
conservation equation (Eq. (13)) in each node to obtain their time The mixture density rm(z, t) is related to mixture quality x(z, t)
dependent ODEs for the phase variables a1,i1(t). These ODEs as,
determine the variation of enthalpy in each node of the single
Nr Nr
phase region. rm ðz; tÞ ¼ (23)
xðz; tÞ þ Nr Nr
Eq. (13) is re-written as,
Now the quality x(z, t) between two consecutive nodes is line-
Ah  S ¼ 0 (16) arly approximated as,
v v

where A¼ ½rðz; tÞ þ rinlet vinlet
vt vz xðz; tÞ ¼ xj1 ðtÞ þ s1;j1 ðtÞ z  Lm;j1 ðtÞ (24)

and S ¼ Nr Nr Npch where x1 ðtÞ ¼ 0

Putting Eqs. (20), (23) and (24) in Eq. (22), applying weighted
Putting Eq. (14) into Eq. (16), multiplying by weight function residual method followed by simplification yields,
Wk(z) and integrating between the limits z ¼ Li1(t) to z ¼ Li(t), it is
written as, ds1;j1 ðtÞ f1;j1 ðtÞ þ f2;j1 ðtÞ þ f3;j1 ðtÞ
¼ : (25)
dt f4;j1 ðtÞ
Li ðtÞ
Z
ðAh  SÞ*Wk ¼ 0
Li1 ðtÞ 2.3. ODE for vinlet(t)

putting Wk ¼ 1, integrating followed by simplification leads to, The coolant inlet velocity vinlet(t) of the heated channel is a
h further state variable of the model. In order to get the final ODE for
da1;i1 ðtÞ 2
¼ Nr Nr Npch  rinlet vinlet ðtÞa1;i1 ðtÞ inlet velocity, the momentum equations of all considered regions
dt bi1 ðLi ðtÞ  Li1 ðtÞÞ are integrated as follows.
i
þ gi1 ðtÞ Momentum balance equation in the single phase region is
written as,
(17)   *
vv* vv* vP1F f * * *2
r* *
þ v* * ¼ *
 r v  r* g* (26)
vt vz vz 2D*h
2.2. ODE in two-phase
After non-dimensionalizing the above equation and integrating
The ODEs for next time dependent variables s1,j1(t) in the two over z between two consecutive nodes in the single phase region,
phase region are obtained from the mixture density equation using i.e. from z ¼ Li1(t) to z ¼ Li(t), we get the pressure drop between
homogeneous equilibrium model which is averaged across the consecutive nodes as,
cross section of the channel. dvinlet ðtÞ
From mass balance, DP1F;i1 ¼ I1;i1 ðtÞ þ f5;i1 ðtÞ (27)
dt
  Thus the total pressure drop across the single phase region is
vr*m z* ; t * v
    given by,
þ * r*m z* ; t * v*m z* ; t * ¼ 0 (18)
vt * vz X
Ns
DP1F ¼ DP1F;i1 (28)
where r*m ðz* ; t * Þ
and v*m ðz* ; t * Þ
are the mixture density and velocity i¼2
in the two-phase region. In two-phase region v*m ðz* ; t * Þ satisfies Momentum balance equation in the two phase region is written
[33], as,
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 317

Fig. 2. A typical stability map in NpcheNsub plane: a) for forced circulation system, b) for natural circulation system.

After non-dimensionalizing the above equation and integrating

  over z between two consecutive nodes in the two-phase region, i.e.
vv*m vv* vP * f*
r*m þ v*m m ¼  2F  m* r*m v*2 * *
m  rm g (29) from z ¼ Lm,j1(t) to z ¼ Lm,j(t), we get the pressure drop between
vt * vz* vz* 2Dh
consecutive nodes as,
318 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

a b

Fig. 3. Effect on stability map of forced circulation system with increase in number of nodes: a) in single phase region, b) in two-phase region, c) in both single phase and two-phase
region simultaneously.

For a forced circulation system:

dv ðtÞ Equating the sum of aforementioned four pressure drops (Eqs.
DP2F;j1 ¼ Q1;j1 ðtÞ inlet þ f6;j1 ðtÞ (30)
dt (28) and (31)e(33)) with an applied constant external pressure
Thus the total pressure drop across the two-phase region is drop (DPext), we get the ODE for inlet velocity of the coolant (vinlet)
given by, as,

X
Nt dvinlet ðtÞ
¼ f7 ðtÞ (34)
DP2F ¼ DP2F;j1 (31) dt
j¼2
For a natural circulation system:
Pressure drop in the riser and downcomer is given by,
where the intermediate terms I1(t), Q1(t), f5(t) and f6(t) are defined
in Appendix B. Pressure drops at the inlet and exit of the channel
dvinlet ðtÞ
are given by, DPriser ¼ k1 ðtÞ þ P1 ðtÞ (35a)
dt
DPinlet ¼ Kinlet rinlet v2inlet (32)

1
DPdowncomer ¼ rinlet ðL  Ht Þ þ ðHt þ Bt þ Lh Þ
Fr h
DPexit ¼ Kexit rm;exit v2m;exit F2exit (33)  
dv ðtÞ
 A2ol Dol Nf ;1F v2inlet ðtÞ þ Aol inlet (35b)
dt
Here, rm,exit and vm,exit are the mixture density and velocity at the
exit of the channel and are defined in Appendix B. Equating the sum of aforementioned six pressure drops (Eqs.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 319

a b

Fig. 4. Effect on stability map of natural circulation system with increase in number of nodes: a) in single phase region, b) in two-phase region, c) in both single phase and two-
phase region simultaneously.

(28), (31)e(33) and (35)) to zero, we get the ODE for inlet velocity of which separates the stable fixed points to unstable fixed points is
the coolant (vinlet) as, called stability boundary.
To perform the stability analysis, the above set of ODEs are
dvinlet ðtÞ
¼ f7 ðtÞ (36) linearized by applying steady state values of the coefficients a1(t)
dt and s1(t). At steady state (equilibrium solutions) a1(t) and s1(t) are
given as,

3. Stability analysis  
Nr Nr Npch
~1 ¼
a
The term stability can be interpreted as the temporal behaviour rinlet vinlet
exhibited by a system when it experiences internal or external
disturbances (perturbations) in its normal operating condition. A
system is said to be stable, when it returns back to its steady state or  
Nr Nr Npch
normal operating conditions after an exposure to a perturbation. ~s1 ¼
rinlet vinlet
However, if the system moves away from its original operating
condition, it is said to be unstable. In geometrical terms, a stable Linear stability analysis is carried out by applying small
system returns back to its stable fixed point or stable node in its perturbation to these steady state values. The stability boundaries
phase space after the exposure to perturbations. Thus the boundary are generated solving the following systems of equations.
320 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

a b

Fig. 5. Comparison of stability boundaries with density invariant models [33,34] for a) Kin ¼ 4.0. b) Kin ¼ 6.0. c) Nf,2F ¼ 4.0.

Table 1
the system is determined from the eigenvalues of the obtained
Design and operating parameter values used for comparison.
Jacobian. Eigenvalues with negative real part defines the system is
Nr Nr Kinlet Kexit Nf,1F Nf,2F Fr stable. However, presence of at least one eigenvalue with positive
1.05397 0.0512 6.0 2.0 2.8 5.6 0.0333 real part makes the system unstable. Fig. 2a and b represents typical
stability maps obtained with the aforementioned approach for
forced circulation and natural circulation system respectively.

_
XðtÞ ¼0
ReðdetðJ  iuIÞÞ ¼ 0 4. Effect of number of nodes on stability boundary
ImðdetðJ  iuIÞÞ ¼ 0
pffiffiffiffiffiffiffi 4.1. Forced circulation system
where i ¼ 1, u is the oscillation frequency, I is an identity ma-
!
trix, J ¼ D!f ð X ðtÞÞ is the Jacobian of the system and Re & Im stands Set of Fig. 3 represents the effect of increasing number of nodes in
X
for real and imaginary part of the determinant (det). The stability of the prediction of stability boundary in a forced circulation system.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331
a b
Fig. 6. Comparison of stability boundaries with a) model prescribed by Papini et al. [22], b) experimental data of water [41].
Fig. 3a and b shows the effect of increase in number of nodes in single
phase and two-phase region respectively. In Fig. 3c, the effect of
increase in number of nodes in combination of both single phase and
two-phase region is shown. Form the set of Fig. 3 it can be concluded
that due to increase in number of nodes the stability boundaries
shows a left shift. It is worth noting that this left shift in stability
boundaries due to increase in number of nodes indicates that as the
number of nodes increases, the stability boundary become closer to
the exact solutions even in the presence of linear approximations of
single phase enthalpy and two-phase quality. The shift pattern due to
different combinations of approximation techniques and exact so-
lutions can be found in several literatures [31,33,34].
4.2. Natural circulation system
Set of Fig. 4 represents the effect of increasing number of nodes
in the prediction of stability boundary in a natural circulation sys-
tem. Fig. 4a and b delineates the effect of increase in number of
nodes in single phase and two-phase region respectively, whereas,
Fig. 4c shows the variation of stability boundaries in a combination
of increase in number of nodes in single phase and two-phase re-
gion. From Fig. 4a, it is observed that, as the number of nodes in
single phase region increases, the stable region in the corresponding
stability map shrinks for both Type-I and Type-II instabilities.
Similar pattern is also observed in Fig. 4b, where the stable zone
shrinks due to increase in number of nodes in two-phase region.
However, in this case, the decrease in stable zone is higher for Type-
II instability, and almost negligible effect is seen for Type-I insta-
bility. This is due to the fact that, the Type-I instability being trig-
gered in a very low exit quality condition, by fixing the number of
nodes in single phase region (approximated enthalpy, which defines
the position of boiling boundary), there remains a low discrepancy
between approximated quality and the exact solution.
It is apropos to note that, when the number of nodes is
increased, though there remain linear approximations of phase
variables, but during the system evolution, due to presence of ap-
proximations between multiple nodes, the phase variables tend to
capture the exact solutions. And hence, this impacts the position of
stability boundaries. In addition to these, it is evident from previous
studies [31,33,36,38] that, stability boundaries obtained with
321
linearelinear approximations of single phase enthalpy and two-
phase quality; only show the conventional L-fold trend. These SBs
are far away from experimental data and exact solutions that lie on
the left side of the obtained SBs (forced circulation system) in
corresponding parameter plane (NpcheNsub plane). Hence the left
shift in the stability boundaries, as the number of nodes are
increased in both regions are justified. Because the increase in
number of nodes helps to find the exact solutions during the sys-
tem evolution.
5. Comparison of results with other existing models
The results of the present analysis are compared with lineare-
linear [36] and quadraticequadratic [33,34] density invariant models
(Fig. 5aec). The corresponding operating conditions are given in
Table 1. To provide the said comparisons, the corresponding figures
are obtained by changing only one parameter at a time (a. Kin ¼ 4.0, b.
Kin ¼ 6.0, c. Nf,2F ¼ 4.0), the rest of the parameters remain fixed as
given in Table 1. In these models the density of the coolant is assumed
to be saturated throughout the single phase region of the channel for
all inlet subcooling numbers. As mentioned [36], linearelinear ho-
mogeneous model with invariant density assumption only shows the
conventional trend of a typical stability map. The map predicted by
the said model is far away from the quadratic models [33,34] which
are closer to the exact solution [31] and experimental data [37]. It is
seen that the stability map obtained with linearelinear density
invariant model by assuming the single phase and two-phase region
of the channel to be composed of only 2 nodes each, is far from the
quadratic model of Dokhane [34]. However, when the same proce-
dure is applied by assuming the single phase and two-phase region of
the channel to be composed of N-number of nodes, the stability map is
in well agreement with the aforementioned quadratic model. In
addition to these, the stability map obtained with density variant
model and only 2 nodes in single phase and two-phase region also
shows good agreement. However, when the density variant model is
applied with N-number of nodes, the stability map shifts more toward
left side and become away from the aforementioned results. This
phenomenon can be anticipated due to the over prediction of single
phase density, the predicted boiling boundary becomes lower than
compared to density invariant model, which in turn higher the
322 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331
contribution of two-phase pressure drop and thus the area of stable
zone decreases.
In Fig. 6a and b, comparison between the results obtained with
the presented model and models prescribed by Papini et al. [22]
with different friction factor multipliers calculations, Ambrosini
et al. [27] and experimental data [41] have been shown. Here also, in
both these figures it has been observed that, the results are in good
agreement with density variant model (Ns ¼ 2, Nt ¼ 2) and density
invariant model with N-nodes assumption. From all these compar-
isons it can be deciphered that, the density variant model (Ns ¼ 2,
Nt ¼ 2) produces as good result as shown by density invariant model
with N-number of nodes assumption in single phase as well as two-
phase region. It is worth noting that both these models show
equivalent results but the number of equations and consequently
the computational effort is significantly low in density variant
model with only 2 node assumption for single phase and two-phase
region as compared to density invariant models.
Fig. 7. Comparison of steady state flow rates calculated with different two-phase friction factor multipliers with experimental data: a) at different pressures for HPNCL, b) at
different pressures for APSARA loop, c) at different power levels in a test loop.
6. Effect of two-phase friction factor multipliers on stability
characteristics
It is worth noting that, a model's applicability to simulate the
steady state nature of a system plays a dominant role to define the
stability behaviour of the system. During a flow at steady state, the
driving buoyancy force is balanced by the resisting frictional forces.
Hence it raises the requirement of accurate prediction of these
forces which includes the predictions of mixture density, void
fraction and friction factor multipliers. Several studies have been
carried out [15,22,42] to calculate the aforementioned quantities.
However, the present study is done to delineate the effects of
different two-phase friction factor multipliers on the stability
characteristics of the system. Following models have been used to
calculate the two-phase friction factor multipliers,
a) Homogeneous Equilibrium Model (HEM) [43].
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 323
20 1
" !#" !#n where F2l
¼1þ þ
vfg mfg ctt c2tt
F2lo ¼ 1þx 1þx where n ¼ 0:25 for Re
vf mf   !0:25
1  x 1:75 mf rg
 2*104 & 0:2 for Re > 2*104 c2tt ¼
x mg rf

b) Jones (MartinellieNelson) [43,44].

d) Friedel [46]
" ! #
rf
F2lo ¼ UðP; GÞ 1:2  1 x0:824
rg
. 3:23FH
F2lo ¼ E þ
þ 1:0 P is in Psia & G is in lb ft2 h Fr 0:045 We0:035

8      
> G G G
>
> 1:36 þ 0:0005P þ 0:1  0:000714 if  0:7
< 106 106 106
UðP; GÞ ¼  6  6  
>
> 10 10 G
>
: 1:26  0:0004P þ 0:119 þ 0:00028P if > 0:7
G G 106

rf fgo
where E ¼ ð1  xÞ2 þ x2 & F ¼ x0:728 ð1  xÞ0:224
rg flo
c) LockharteMartinelli [43,45]
!0:91 !0:19 !0:7
rf mg mg rf rg
H¼ 1 & rH ¼
rg mf mf xrf þ ð1  xÞrg

F2lo ¼ F2l ð1  xÞ1:75 G2 G2 Dh

Fr ¼ & We ¼
gDh r2H rH s

Fig. 8. Stability boundaries obtained for a natural circulation system with different two-phase friction factor multipliers at 5 bar.
324 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

Fig. 9. Stability boundaries obtained for a natural circulation system with different two-phase friction factor multipliers at 70 bar.

phase friction factor multiplier models show good agreement

Table 2
with experimental data with system pressure above 10 bar, but
Test conditions for natural circulation system (5 bar & 70 bar).
form a comparison point of view, the BaroczyeChisholm and HEM
Heated channel length (m) 1.00 model slightly over-predicts the flow rate whereas the Jones
Heated channel diameter (m) 0.01
(MartinellieNelson) model under-predicts it. It is also pointed out
Downcomer length (m) 4.50
Downcomer diameter (m) 0.01
that the Friedel model and LockharteMartinelli model give best
Riser length (m) 3.00 agreement with test data.
Riser diameter (m) 0.01 In the second case, From Fig. 7b (APSARA test data), it has been
Length of standpipe (m) 1.00 observed that, except Jones (MartinellieNelson) model all other
Width of loop (m) 2.25
models over-predict the flow rate. It is also apropos that, in both
Kinlet 1.00
Kexit 1.00 these comparisons (Fig. 7a and b), almost all the two-phase friction
factor multiplier models loses its degree of agreement with test
data at a pressure below 10 bar. At a pressure below 10 bar, they
over-predicts the flow rate. These discrepancies may be due to the
e) BaroczyeChisholm [47] fact that at this pressure range, the correlation for single phase
friction factor is different to that considered in the present study
(f1F ¼ 0.184/Re0.2) over the whole range. It is also found in litera-
tures [39,42] that, the calculation of single phase friction factor is
 h i
sensitive to several parameters such as system pressure, channel
F2lo ¼ 1 þ G2  1 Bx0:875 ð1  xÞ0:875 þ x1:75
diameter, flow rate, roughness factor etc. However, the explicit
formulation of single phase friction factors in the considered test
!0:2
ng mg loops at different pressure ranges being unavailable in the litera-
2
where G ¼ tures, the Blasius equation has been considered in the study. From
nf mf
the aforementioned comparisons it is seen that, the Friedel model
best predicts the flow rate for the case of HPNCL loop. It is also
worth noting that though the Jones model best predicts the flow
6.1. Model comparisons with steady state data rate for APSARA loop, but in APSARA loop the flow rate being so
low, the flow may not be fully developed and hence the conven-
The influences of different two-phase friction factor multipliers tional applicability of different friction factor multipliers is a
on steady state natural circulation flow rate have been shown in concern. Fig. 7c shows the flow rates at different power levels for
Fig. 7aec. For this purpose, the test data of HPNCL [48] and APSARA the test loop (49.3 mm i.d) considered by Vijayan et al. [50]. Similar
[49] loop have been considered. In the first case, From Fig. 7a to the Fig. 7a, here also best agreements are shown by Lock-
(HPNCL test data) it has been observed that, almost all the two- harteMartinelli as well as Friedel model.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 325

Table 3 Table 4
Fractional contribution to total channel pressure drop (5 bar, Nsub ¼ 1.5, Npch ¼ 1.0 Fractional contribution to total channel pressure drop (70 bar, Nsub ¼ 1.5, Npch ¼ 1.0
(Type-I), 2.5 (Type-II)). (Type-I), 2.5 (Type-II)).

BaroczyeChisholm BaroczyeChisholm

Term (Type-I, DR ¼ 1.134) (Type-II, DR ¼ 0.447) Term (Type-I, DR ¼ 1.342) (Type-II, DR ¼ 0.577)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 0.1175 1.1220 1.0372 8.9952 DPacc 0.1092 0.9897 1.0578 8.5735
DPfric,1F 0.2124 2.0821 0.3488 3.0253 DPfric,1F 0.2176 1.9711 0.3480 2.8208
DPfric,2F 0.0712 0.6796 0.4843 4.1998 DPfric,2F 0.0603 0.5460 0.5147 4.1716
DPin 0.2609 2.4917 0.6691 5.8026 DPin 0.2741 2.4831 0.6905 5.5970
DPexit 0.3210 3.0647 1.1071 9.6010 DPexit 0.3759 3.4055 1.4215 11.5217
DPgrav 9.4897 90.6139 7.8845 68.3762 DPgrav 10.0012 90.6045 8.3050 67.3154
DPtotal 10.4727 100.00 11.5311 100.00 DPtotal 11.0383 100.00 12.3374 100.00

HEM HEM

Term (Type-I, DR ¼ 1.033) (Type-II, DR ¼ 0.545) Term (Type-I, DR ¼ 1.280) (Type-II, DR ¼ 0.622)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 0.1206 1.1484 1.0227 8.9450 DPacc 0.1105 1.0000 1.0543 8.5980
DPfric,1F 0.2053 1.9554 0.2484 2.1723 DPfric,1F 0.2150 1.9461 0.2792 2.2772
DPfric,2F 0.0798 0.7600 0.6169 5.3953 DPfric,2F 0.0618 0.5593 0.5596 4.5636
DPin 0.2551 2.4297 0.5335 4.6661 DPin 0.2720 2.4612 0.5962 4.8625
DPexit 0.3803 3.6219 1.5873 13.8829 DPexit 0.4000 3.6202 1.7805 14.5204
DPgrav 9.4582 90.0847 7.4248 64.9383 DPgrav 9.9904 90.4131 7.9921 65.1783
DPtotal 10.4992 100.00 11.4337 100.00 DPtotal 11.0498 100.00 12.2619 100.00

LockharteMartinelli Friedel

Term (Type-I, DR ¼ 0.942) (Type-II, DR ¼ 0.813) Term (Type-I, DR ¼ 1.100) (Type-II, DR ¼ 1.087)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 0.1252 1.1873 0.9967 8.7644 DPacc 0.1211 1.0793 1.0152 8.0873
DPfric,1F 0.1945 1.8453 0.1873 1.6467 DPfric,1F 0.1933 1.7230 0.1707 1.3602
DPfric,2F 0.0975 0.9253 0.7796 6.8553 DPfric,2F 0.1617 1.4419 1.2852 10.2386
DPin 0.2461 2.3344 0.4420 3.8863 DPin 0.2533 2.2580 0.4295 3.4219
DPexit 0.4717 4.4748 1.9232 16.9115 DPexit 0.5992 5.3425 2.3584 18.7882
DPgrav 9.4066 89.2328 7.0435 61.9357 DPgrav 9.8877 88.1552 7.2934 58.1038
DPtotal 10.5417 100.00 11.3722 100.00 DPtotal 11.2162 100.00 12.5524 100.00

Friedel Jones

Term (Type-I, DR ¼ 0.809) (Type-II, DR ¼ 1.412) Term (Type-I, DR ¼ 1.005) (Type-II, DR ¼ 3.075)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 0.1370 1.2767 0.9250 7.9878 DPacc 0.1267 1.1290 0.9449 7.8311
DPfric,1F 0.1657 1.5445 0.1096 0.9465 DPfric,1F 0.1814 1.6160 0.1022 0.8466
DPfric,2F 0.2466 2.2982 1.5168 13.0988 DPfric,2F 0.1192 1.0625 1.1668 9.6694
DPin 0.2211 2.0612 0.3092 2.6704 DPin 0.2428 2.1632 0.3050 2.5276
DPexit 0.7155 6.6688 2.3835 20.5833 DPexit 0.7296 6.5011 2.9649 24.5716
DPgrav 9.2428 86.1506 6.3357 54.7133 DPgrav 9.8227 87.5282 6.5827 54.5537
DPtotal 10.7287 100.00 11.5799 100.00 DPtotal 11.2223 100.00 12.0665 100.00

Jones LockharteMartinelli

Term (Type-I, DR ¼ 0.717) (Type-II, DR ¼ 3.914) Term (Type-I, DR ¼ 0.856) (Type-II, DR ¼ 5.786)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 0.1451 1.3532 0.7886 7.3110 DPacc 0.1381 1.2181 0.8357 7.1027
DPfric,1F 0.1445 1.3474 0.0483 0.4473 DPfric,1F 0.1563 1.3789 0.0532 0.4523
DPfric,2F 0.2047 1.9090 1.3791 12.7848 DPfric,2F 0.1674 1.4760 1.4390 12.2304
DPin 0.2018 1.8822 0.1790 1.6590 DPin 0.2199 1.9393 0.1974 1.6782
DPexit 0.9348 8.7178 3.0688 28.4494 DPexit 0.9957 8.7818 3.5083 29.8183
DPgrav 9.0920 84.7905 5.3231 49.3485 DPgrav 9.6609 85.2060 5.7320 48.7181
DPtotal 10.7229 100.00 10.7868 100.00 DPtotal 11.3383 100.00 11.7657 100.00

6.2. Effect on stability boundaries
The effects of different two-phase friction factors on the insta-
bility thresholds are discernible from the stability maps shown in
Figs. 8 and 9 for a natural circulation system. The test conditions are
given in Table 2. It is observed from Fig. 7a and b that, in both cases;
the LockharteMartinelli model predicts higher mass flow rate
compared to Friedel model and Jones model below 20 bar. More-
over, at high pressures it predicts least flow rate among the
aforementioned three models. The same phenomenon is also
delineated by stability maps (Figs. 8 and 9) drawn at two system
pressures (5 bar & 70 bar). The contributions of different pressure
drops across the channel at a defined steady state with aforemen-
tioned friction factor models are shown in Tables 3 and 4. The effect
326 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

Fig. 10. Stability boundaries obtained for a forced circulation system with different two-phase friction factor multipliers at 5 bar.

Fig. 11. Stability boundaries obtained for a forced circulation system with different two-phase friction factor multipliers at 70 bar.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331
Table 5
Test conditions for forced circulation system (5 bar & 70 bar).
Heated channel length (m) 3.658
Heated channel diameter (m) 0.0124
Kinlet 11.50
Kexit 2.50
on nature of stability is defined by the corresponding decay ratios.
Denoting the dominant eigenvalue as l ¼ s þ iu, the decay ratio of
the system is denoted by DR ¼ expð2pðs=uÞÞ. The decay ratio is a
measure of degree of stability of the system. The lower the value of
decay ratio, the more stable is the system. For growing oscillation
DR will be greater than 1, whereas, for damping oscillation, it will
be less than 1. DR exactly equals 1 denotes pure oscillation char-
acteristics. In this case the system will be said as neutrally stable.
It is evident from previous researches [10,12,13,15,19,20] that, in
the case of Type-I instability, the gravitational pressure drop across
the channel plays a dominant role. The said scenario has been
depicted by Figs. 8 and 9, where from Tables 3 and 4 it is clear that
as the contribution of gravitational pressure drop decreases the
area of stable operating zone corresponding to Type-I instability,
increases. In case 1 (Fig. 8, 5 bar), the contribution of gravitational
pressure drop is highest in BaroczyeChisholm model, whereas it is
least in Jones model. In case 2 (Fig. 9, 70 bar), the contribution of
gravitational pressure drop is highest in BaroczyeChisholm model,
whereas it is least in LockharteMartinelli model. The point A in
Fig. 8 being situated in the unstable side for BaroczyeChisholm and
HEM model the decay ratio (DR) is greater than 1. However, the
same point being in the stable side for LockharteMartinelli model,
Friedel model and Jones model, the decay ratio is less than 1. In
Fig. 9, point C being in the unstable side for BaroczyeChisholm
model, HEM, Friedel model and Jones model the decay ratio is
greater than 1, and it is less than 1 (stable side) for Lock-
harteMartinelli model.
To compare the effects on Type-II (density wave oscillation)
instability boundary, point B & D have been considered in Figs. 8
and 9 respectively. From previous researches [27,31,32,36,51,52] it
is evident that the Type-II instability (Density wave oscillation) is
influenced by the frictional pressure drop across the channel and in
this case the contribution of frictional pressure drop is significant.
Higher the frictional pressure drop, more unstable will be the
system. It can be anticipated as, in the aforementioned figures, the
area of stable operating zone decreases as the contribution of
frictional pressure drop increases across the channel. In the first
case (Fig. 8, 5 bar), the highest contribution of frictional pressure
drop has been observed in Jones model and lowest for Bar-
oczyeChisholm model, and hence the area of stable operating zone
is least for Jones model and largest for BaroczyeChisholm model. In
the second case (Fig. 9, 70 bar), the highest contribution of frictional
pressure drop has been observed in LockharteMartinelli model and
lowest for BaroczyeChisholm model, and hence the area of stable
operating zone is least for Jones model and largest for Bar-
oczyeChisholm model. In Fig. 8, point B being in the stable side of
BaroczyeChisholm, HEM and LockharteMartinelli model, have
decay ratios less than 1, whereas, due to its position in unstable side
for Friedel model and Jones model, have decay ratios greater than 1.
In Fig. 9, the point D being in the stable side for BaroczyeChisholm
and HEM model, and in the unstable side for Friedel model, Lock-
harteMartinelli model and Jones model have decay ratios less than
1 and greater than 1 respectively.
Figs. 10 and 11 delineate the effect of different frictional factor
multipliers on prediction of stability boundary of a forced circula-
tion system. Similar to the natural circulation system, here also the
stability maps are drawn for two different system pressures (5 bar
& 70 bar). The test conditions are given in Table 5. It is observed
327
from Figs. 10 and 11 that the effects are same as seen for Type-II
instability for a natural circulations system. From Table 6 (point
E), it is clear that, at a low system pressure (5 bar), the Jones model
contributes highest frictional pressure drop across the channel,
whereas, the BaroczyeChisholm model contributes the least.
Hence the area of stable operating zone is least for Jones model and
highest for BaroczyeChisholm model. From Table 7 (point F), it is
seen that at high system pressure (70 bar), the LockharteMartinelli
model contributes highest frictional pressure drop across the
channel, whereas, the BaroczyeChisholm model contributes the
least. Hence the area of stable operating zone is least for Lock-
harteMartinelli model and highest for BaroczyeChisholm model. A
comparative study of the stability thresholds obtained by the
aforementioned friction factor models with experimental data
(Solberg [41]) is also shown by Fig. 12. Similar to the Fig. 7a, it is
seen from Fig. 12 that, the Friedel model for two-phase friction
factor calculation produces best agreement with the experimental
data. The system pressure of the experimental data being 81.1 bar,
the shift pattern of the stability boundaries remain same as shown
in Fig. 11 for 70 bar.
7. Conclusions
Flow instabilities are analyzed with state space approach using a
novel nodalized reduced order model (NROM) in forced as well as
natural circulation system. Over the plethora of researches, these
become the topic of interest for the design and development of
nuclear reactor systems such as BWR, AHWR flow channels etc. The
existing ROMs use single spatial approximation for flow variables
over the whole single phase and two-phase region, which is quite
unrealistic. In NROM, the spatial approximation is taken node-wise.
With increasing number of nodes, variation of flow variables can be
captured quite accurately. This can be seen by the convergence of
stability maps with increasing number nodes.
Following insights can be drawn from the study as,
1) In a forced circulation system, as the number of nodes in the
coolant channel increases, the stability maps shift towards left
side of the parameter plane of phase change number (Npch) and
subcooling number (Nsub) and consequently the area of stable
operating zone decreases. However, this shift gets saturated
after a few numbers of nodes and after that no further shift is
observed.
2) In a natural circulation system, as the number of nodes in the
coolant channel increases, the Type-I instability boundaries
show a right shift and Type-II boundaries show a left shift in the
aforementioned parameter plane and overall the area of stable
operating zone decreases. In this case also the shift gets satu-
rated after a few numbers of nodes and after that no further shift
is observed.
3) From the comparisons shown in Figs. 5aec and 6a, b it is
observed that, the predicted instability thresholds with the
present model having density variation in single phase and only
two nodes each in singe phase and two-phase region shows well
agreement with existing density invariant models [22,34] which
are closer to exact solutions [31] and experimental data [37].
4) It is also observed that considering the single phase and two-
phase region of the coolant channel to be composed of N-
number of nodes without considering the density variation in
single phase as done in previous researches [22,33,34], also
provide well agreement with existing model results.
5) By comparing the performance of different friction factor multi-
pliers to the steady state characteristics of a natural circulation
loop (Fig. 7a and b); it is observed that the Friedel model provides
328 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

Table 6 Table 7
Fractional contribution to total channel pressure drop (5 bar, Nsub ¼ 3.0, Npch ¼ 6.0). Fractional contribution to total channel pressure drop (70 bar, Nsub ¼ 3.0, Npch ¼ 6.0).

Term BaroczyeChisholm HEM Term BaroczyeChisholm HEM

(DR ¼ 0.015) (DR ¼ 0.023) (DR ¼ 0.046) (DR ¼ 0.045)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 2.9983 6.2898 2.9983 5.6757 DPacc 3.0704 5.7105 3.0704 5.3112
DPfric,1F 0.3791 0.7953 0.3791 0.7177 DPfric,1F 0.4152 0.7722 0.4152 0.7182
DPfric,2F 0.5948 1.2477 0.9079 1.7186 DPfric,2F 0.6380 1.1866 0.7534 1.3032
DPin 11.6568 24.4531 11.6568 22.0656 DPin 12.7525 23.7176 12.7525 22.0590
DPexit 5.4186 11.3669 10.2634 19.4280 DPexit 7.4355 13.8287 11.3627 19.6549
DPgrav 26.6223 55.8472 26.6223 59.3945 DPgrav 29.4567 54.7844 29.4567 59.9535
DPtotal 47.6699 100.00 52.8278 100.00 DPtotal 53.7683 100.00 57.8109 100.00

Term LockharteMartinelli Friedel Term LockharteMartinelli Friedel

(DR ¼ 0.044) (DR ¼ 0.093) (DR ¼ 0.627) (DR ¼ 0.099)

DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 2.9983 5.3423 2.9983 4.8430 DPacc 3.0704 3.6364 3.0704 4.8479
DPfric,1F 0.3791 0.6755 0.3791 0.6124 DPfric,1F 0.4152 0.4917 0.4152 0.6555
DPfric,2F 1.2024 2.1425 2.4246 3.9162 DPfric,2F 2.4906 2.9497 1.7795 2.8096
DPin 11.6568 20.7696 11.6568 18.8282 DPin 12.7525 15.1031 12.7525 20.1347
DPexit 13.2652 23.6354 17.8300 28.7994 DPexit 36.2510 42.9328 15.8618 25.0438
DPgrav 26.6223 47.4347 26.6223 43.0008 DPgrav 29.4567 34.8862 29.4567 46.5085
DPtotal 56.1241 100.00 61.9111 100.00 DPtotal 84.4365 100.00 63.3361 100.00

Term Jones Term Jones

(DR ¼ 0.214) (DR ¼ 0.371)

DP (non-dim) % of Total DP (non-dim) % of Total

DPacc 2.9983 3.9086 DPacc 3.0704 4.4213

DPfric,1F 0.3791 0.4942 DPfric,1F 0.4152 0.5979
DPfric,2F 2.7442 3.5772 DPfric,2F 1.5797 2.2747
DPin 11.6568 15.1955 DPin 12.7525 18.3633
DPexit 32.3113 42.1203 DPexit 22.1713 31.9260
DPgrav 26.6223 34.7042 DPgrav 29.4567 42.4168
DPtotal 76.7120 100.00 DPtotal 69.4458 100.00

Fig. 12. Comparison of stability boundaries predicted with different two-phase friction factor multipliers with experimental data.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331 329

most accurate results. In addition to these, the Jones model un-  

der-predicts and rest over-predicts the flow rate calculation. Nr Nr s1;j1 Lm;j  Lm;j1 dL
f2;j1 ¼      m;j1
6) It is found that, below 20 bar system pressure, the Lock- Nr Nr þ xj1 Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1 dt
harteMartinelli model offers lower frictional pressure drop
compared to Friedel model and Jones model. However, it of-
 
fers higher frictional pressure drop at a pressure above 20 bar. Nr Nr Lm;j  Lm;j1 dx
7) The same effect is observed (Figs. 8e11) in stability maps also. f3;j1 ¼      j1
Nr Nr þ xj1 Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1 dt
Below 20 bar, due to a lower contribution of frictional pressure
drop, the area of stable operating zone is larger for Lock-
harteMartinelli model compared to Friedel model and Jones Nr Nr Nr Nr þ xj1
model. However, the area of stable operating zone is least for f4;j1 ¼ 1  
s21;j1 Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1
LockharteMartinelli model at pressure above 20 bar. !!
Nr Nr þ xj1
þ Log  
Appendix A Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1

The dimensional variables and parameters used in the devel- I1;i1 ¼ rinlet ðLi  Li1 Þ
opment of model and analysis are given below. The asterisk in-
dicates the original dimensional quantities. The quantities having 
subscript jF are defined for both phases of the fluid. Quantities rinlet vinlet da1;i1
I2;i1 ¼  a1;i1 ðLi  Li1 ÞN2 þ ðN1
having j ¼ 1 are of liquid phase whereas having j ¼ 2 are of vapour a21;i1 N2 dt
phase.  
N1  hi1 N2
 hi1 N2 ÞLog
r*g r*f N1  ðhi1 þ ai1 ðLi  Li1 ÞÞN2
fjF L*ch t * v*0  
Nf ;jF ¼ Nr ¼ Nr ¼ t¼ N1  ðhi1 þ ai1 ðLi  Li1 ÞÞN2 dLi1
2D*h r*f Dr* L*ch þ rinlet vinlet Log
N1  hi1 N2 dt
   
m* t * h* z* ; t *
mðtÞ ¼ hðz; tÞ ¼ ZLi
L*ch Dh*fg vv
  DP1F;1;i1 ¼ rv dz ¼ rinlet vinlet ðvi  vi1 Þ
r* z* ; t * z* vz
rm ðz; tÞ ¼ m * z¼ * Li1
rf Lch

  ZLi
r*inlet v*inlet t * 4A* v*2
rinlet ¼ vinlet ðtÞ ¼ D*h ¼ * Fr ¼ *0 * DP1F;2;i1 ¼ Nf ;1F rv2 dz
r*f v*0 xh g Lch
Li1
 
Nf ;1F r2inlet v2inlet N1  hi1 N2
Lr* B*t N1* N2* Dh*fg ¼ Log
Lr ¼ Bt ¼ N1 ¼ N2 ¼ a1;i1 N2 N1  ðhi1 þ ai1 ðLi  Li1 ÞÞN2
L*ch L*ch r*f r*f

 ZLi
00
q * Dr* x*h L*ch h*sat  h*inlet Dr* DP1F;3;i1 ¼
1
r dz
Npch ¼ Nsub ¼ f2F ¼ f1F F2lo Fr
A* Dh*fg v*0 r*g r*f Dh*fg r*g Li1
ðLi  Li1 Þ  
¼ 2ðN1  hi1 N2 Þ þ N2 a1;i1 ðLi  Li1 Þ
2Fr
D*h D*h A*c A*
Dol ¼ or Aol ¼ or *c
D*r D*doc A*r Adoc
f5;i1 ¼ DP1F;1;i1 þ DP1F;2;i1 þ DP1F;3;i1 þ I2;i1

The pressure drop across the two-phase region is calculated as,


Appendix B DP2F;1;j1 ¼ Npch vm;j1 Q1;j1 þ Npch M2;j1

The intermediate terms and functions that appeared during the 

model development are given below. For the sake of parity all the DP2F;2;j1 ¼ Nf ;2F v2m;j1 Q1;j1 þ 2Npch vm;j1 M2;j1
time parentheses in the terms like m(t), a1(t), s1(t) and vinlet(t) have
2
been omitted. þ Npch M3;j1

bi1 ¼ N1  N2 hi1  4ðLi  Li1 Þa1;i1 3
Q1;j1
DP2F;3;j1 ¼
Fr
dLi1 ðtÞ  
gi1 ðtÞ ¼ a1;i1 N1  2hi1 N2 þ N2 a1;i1 ðLi1  Li Þ
dt
dLNs
DP2F;4;j1 ¼ Npch Q1;j1
dt
f1;j1 ¼ rm;j vm;j  rm;j1 vm;j1
where Q1, M2 and M3 are given as,
330 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331

[4] G.V. Durga Prasad, M. Pandey, M.S. Kalra, Review of research on flow in-
ZLm;j
stabilities in natural circulation boiling systems, Prog. Nucl. Energy 49 (2007)
Q1;j1 ¼ rm ðz; tÞdz 429e451.
[5] S. Kakac, B. Bon, A review of two-phase flow dynamic instabilities in tube
Lm;j1
boiling systems, Int. J. Heat Mass Transf. 51 (2008) 399e433.
  !
[6] L.C. Ruspini, C.P. Marcel, A. Clausse, Two-phase flow instabilities: a review, Int.
Nr Nr Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1
¼ Log J. Heat Mass Transf. 71 (2014) 521e548.
s1;j1 Nr Nr þ xj1 [7] G. Yadigaroglu, Two-phase flow instabilities and propagation phenomena, in:
Von Karman Inst. for Fluid Dyn. Two-Phase Flows in Nucl. Reactors, 1978, pp.
ZLm;j ( 19e34.

 Nr Nr   [8] G. Yadigaroglu, Two-phase flow instabilities and propagation phenomena, in:
M2;j1 ¼ rm ðz; tÞ z  Lm;j1 dz ¼ 2 Lm;j  Lm;j1 s1;j1 J.M. Delhaye, M. Giot, M.L. Riethmuller (Eds.), Thermohydraulics of Two Phase
s1;j1 Systems for Industrial Design and Nuclear Engineering, Taylor & Francis,
Lm;j1 Washington, 1981, pp. 353e396.
!) [9] S.Y. Lee, D.W. Lee, Linear analysis of flow instabilities in an open two-phase
  Nr Nr þ xj1
þ Nr Nr þ xj1 Log   natural circulation loop, Nucl. Eng. Des. 128 (1991) 317e330.
Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1 [10] A.K. Nayak, P. Vijayan, D. Saha, V. Venkat Raj, Mathematical modelling of the
stability characteristics of a natural circulation loop, Math. Comput. Model. 22
(1995) 77e87.
ZLm;j

2 [11] I.S. Kyung, S.Y. Lee, Periodic flow excursion in an open two-phase natural
M3;j1 ¼ rm ðz;tÞ z  Lm;j1 dz circulation loop, Nucl. Eng. Des. 162 (1996) 233e244.
[12] A.K. Nayak, P.K. Vijayan, D. Saha, V. Venkat Raj, M. Aritomi, Analytical study of
Lm;j1 nuclear-coupled density-wave instability in a natural circulation pressure
"
Nr Nr   tube type boiling water reactor, Nucl. Eng. Des. 195 (2000) 27e44.
¼ 3  Lm;j  Lm;j1 2Nr Nr [13] A.K. Nayak, P.K. Vijayan, D. Saha, V.V. Raj, M. Aritomi, Study on the stability
2s1;j1 behaviour of a natural circulation pressure tube type boiling water reactor,
  Nucl. Eng. Des. 215 (2002) 127e137.
þ 2xj1  s1;j1 Lm;j  Lm;j1 s1;j1 [14] J.D. Lee, C. Pan, Nonlinear analysis for a nuclear-coupled two-phase natural
( )# circulation loop, Nucl. Eng. Des. 235 (2005) 613e626.
 2 Nr Nr þ xj1 [15] A.K. Nayak, P. Dubey, D.N. Chavan, P.K. Vijayan, Study on the stability
 2 Nr Nr þ xj1 Log   behaviour of two-phase natural circulation systems using a four-equation
Nr Nr þ s1;j1 Lm;j  Lm;j1 þ xj1
drift flux model, Nucl. Eng. Des. 237 (2007) 386e398.
[16] G.V. Durga Prasad, M. Pandey, S.K. Pradhan, S.K. Gupta, Study of flow in-
f6;j1 ¼ DP2F;1;j1 þ DP2F;2;j1 þ DP2F;3;j1 þ DP2F;4;j1 stabilities in double-channel natural circulation boiling systems, Nucl. Eng.
Des. 238 (2008) 1750e1761.
[17] G.V. Durga Prasad, M. Pandey, Stability analysis and nonlinear dynamics of
For forced circulation system,
natural circulation boiling water reactors, Nucl. Eng. Des. 238 (2008)
( ) 229e240.
PNs PNt [18] G.V. Durga Prasad, M. Pandey, Parametric effects on reactivity instabilities and
DPext  f
i¼2 5;i1
þ f
j¼2 6;j1
þ DPinlet þ DPexit nonlinear oscillations in a nuclear-coupled double channel natural circulation
boiling system, Nucl. Eng. Des. 240 (2010) 1097e1110.
f7 ¼ PNs PNt [19] N. Goudarzi, S. Talebi, Linear stability analysis of a double-channel two-phase
I
i¼2 1;i1
þ rinlet j¼2
Q1;j1 natural circulation loop, Prog. Nucl. Energy 67 (2013) 114e123.
[20] S. Paul, S. Singh, Analysis of sub- and supercritical Hopf bifurcation with a
For natural circulation system, reduced order model in natural circulation loop, Int. J. Heat Mass Transf. 77
(2014) 344e358.
[21] R.T. Lahey Jr., F.J. Moody, The Thermal-Hydraulics of a Boiling Water Nuclear
1
P1 ðtÞ ¼ r Lr þ Nf ;1F r2m;exit v2m;exit A2ol Dol F2exit ðLr þ Bt Þ Reactor, American Nuclear Society, 1977.
Fr m;exit [22] D. Papini, A. Cammi, M. Colombo, M.E. Ricotti, Time-domain linear and non-
dL linear studies on density wave oscillations, Chem. Eng. Sci. 81 (2012)
 rm;exit Npch Aol ðLr þ Bt Þ Ns 118e139.
dt [23] J.L. Mun ~ oz-Cobo, M.Z. Podowski, S. Chiva, Parallel channel instabilities in
boiling water reactor systems: boundary conditions for out of phase oscilla-
tions, Ann. Nucl. Energy 29 (2002) 1891e1917.
k1 ðtÞ ¼ rm;exit Aol ðLr þ Bt Þ [24] W.R. Schlichting, R.T. Lahey, M.Z. Podowski, An analysis of interacting insta-
bility modes, in a phase change system, Nucl. Eng. Des. 240 (2010)
3178e3201.
n [25] S. Paruya, S. Maiti, A. Karmakar, P. Gupta, J.P. Sarkar, Lumped parameterization
rinlet A2ol Dol Nf ;1F v2inlet ðtÞðHt þBt þLh Þ of boiling channel-bifurcations during density wave oscillations, Chem. Eng.
Sci. 74 (2012) 310e326.
P s P t
1
Fr rinlet ðHt Lh Þ
þ N f
i¼2 5;i1
þ N f
j¼2 6;j1
þDPinlet [26] N. Goswami, S. Paruya, Control of density wave oscillations in boiling channel,
Chem. Eng. Sci. 110 (2014) 119e133.
o [27] W. Ambrosini, P. Di Marco, J.C. Ferreri, Linear and nonlinear analysis of
þDPexit þP1 ðtÞ density-wave instability phenomena, Heat Technol. 18 (2000) 27e36.
f7 ðtÞ ¼ PN P N [28] G. Yun, S.Z. Qiu, G.H. Su, D.N. Jia, Theoretical investigations on two-phase flow
s
I
i¼2 1;i1
þrinlet j¼2t
Q1;j1 þk1 ðtÞþrinlet Aol ðHt þBt þLh Þ instability in parallel multichannel system, Ann. Nucl. Energy 35 (2008) 665e676.
[29] Y.J. Zhang, G.H. Su, X.B. Yang, S.Z. Qiu, Theoretical research on two-phase flow
instability in parallel channels, Nucl. Eng. Des. 239 (2009) 1294e1303.
[30] J.L. Achard, D.A. Drew, R.T. Lahey, The analysis of nonlinear density-wave
oscillations in boiling channels, J. Fluid Mech. 155 (1985) 213e232.
[31] Rizwan-Uddin, J.J. Dorning, Some nonlinear dynamics of a heated channel,
Nuclear 93 (1986) 1e14.
[32] A. Clausse, R.T. Lahey, The analysis of periodic and strange attractors during
density-wave oscillations in boiling flows, Chaos Solit. Fractals 1 (1991) 167e178.
[33] A.A. Karve, Nuclear Coupled Thermal Hydraulic Analysis of Boiling Water
Reactor, University of Virginia, Virginia, United States, 1998.
[34] A. Dokhane, BWR Stability and Bifurcation Analysis Using a Novel Reduced
References Order Model and the System Code Ramona, Ecole  polytechnique fe de
rale de
Lausanne, Lausanne, Switzerland, 2004.
[1] J.A. Boure, A.E. Bergles, L.S. Tong, Review of two-phase flow instability, Nucl. [35] A. Dokhane, D. Hennig, R. Chawla, Rizwan-Uddin, Semi-analytical bifurcation
Eng. Des. 25 (1973) 165e192. analysis of two-phase flow in a heated channel, Int. J. Bifurc. Chaos 15 (8)
[2] S. Kakac, T.N. Vezigoraglu, A review of two-phase flow instabilities, in: Adv. (2005) 2395e2409.
Two-Phase Flow Heat Transf., vol. 64, 1983, pp. 577e667. [36] A.A. Karve, Rizwan-Uddin, J.J. Dorning, Stability analysis of BWR nuclear-
[3] L. Tadrist, Review on two-phase flow instabilities in narrow spaces, Int. J. Heat coupled thermal-hydraulics using a simple model, Nucl. Eng. Des. 177
Fluid Flow. 28 (1) (2007) 54e62 (spec. iss.). (1997) 155e177.
 S. Paul, S. Singh / International Journal of Thermal Sciences 98 (2015) 312e331
[37] P. Saha, Thermally Induced Two-Phase Flow Instabilities Including the Effect
of Thermal Non-Equilibrium between Phases, Georgia Institute of Technology,
Georgia, Atlanta, 1974.
[38] S. Paul, S. Singh, A density variant drift flux model for density wave oscilla-
tions, Int. J. Heat Mass Transf. 69 (2014) 151e163.
[39] G.S.S.P. Rao, P.K. Vijayan, D. Saha, P.S. Sarkar, A. Sinha, B.K. Godwal, Experi-
mental Investigations in Two-Phase Natural Circulation Loops Relevant to the
AHWR, 2006. BARC/2006/E/003.
[40] Q. Zhou, Stability and Bifurcation Analyses of Reduced Order Models of Forced
and Natural Circulation BWRs, University of Illinois, Urbana, Illinois, 2006.
[41] K.O. Solberg, Resultats des Essais d'Instabilites surla Boucle “Caline” et Com-
paraisons avecun Code de Calcul, Centre d'Etudes Nucle aires de Grenoble
(CENG), 1966. Note 225.
[42] X. Fang, Y. Xu, Z. Zhou, New correlations of single-phase friction factor for
turbulent pipe flow and evaluation of existing single-phase friction factor
correlations, Nucl. Eng. Des. 241 (2011) 897e902.
[43] N.E. Todreas, M.S. Kazimi, Nuclear Systems I: Thermal Hydraulic Fundamen-
tals, Taylor & Francis, 1993.
[44] R.C. Martinelli, D.B. Nelson, Prediction of pressure drop during forced circu-
lation boiling of water, J. Heat Transf. Trans. ASME. 70 (1948) 695e702.
[45] R.W. Lockhart, R.C. Martinelli, Proposed correlation of data for isothermal
two-phase, two component flow in pipes, Chem. Eng. Prog. 45 (1) (1949)
39e48.
331
[46] L. Friedel, Improved friction pressure drop correlation for horizontal and
vertical two-phase pipe flow, in: European Two-Phase Flow Group Meeting,
1979.
[47] D. Chisholm, Pressure gradients due to friction during the flow of evaporating
two-phase mixtures in smooth tubes and channels, Int. J. Heat Mass Transf. 16
(2) (1973) 347e358.
[48] N. Kumar, A.K. Nayak, P.K. Vijayan, D. Saha, Experimental investigations on
two phase natural circulation in a closed rectangular loop, in: 2nd
JapaneKorea Seminar on Nuclear Thermal Hydraulics and Safety, 2000.
[49] P. Dubey, G.S.S.P. Rao, D.S. Pilkhwal, P.K. Vijayan, D. Saha, Analysis of the
Experimental Data on Two-Phase Natural Circulation from the Flow Pattern
Transition Instability Studies Facility at APSARA Reactor, 2004. BARC/2004/E/
031.
[50] P.K. Vijayan, A.K. Nayak, D. Saha, M.R. Gartia, Effect of loop diameter on the
steady state and stability behaviour of single-phase and two-phase natural
circulation loops, Sci. Technol. Nucl. Install. 2008 (2008).
[51] Rizwan-Uddin, Effect of double-humped axial heat flux variation on the sta-
bility of two-phase flow in heated channels, Int. J. Multiph. Flow. 20 (6) (1994)
1129e1142.
[52] J.D. Lee, C. Pan, Dynamics of multiple parallel boiling channel systems with
forced flows, Nucl. Eng. Des. 192 (1999) 31e44.