Desalination 420 (2017) 258–272

Contents lists available at ScienceDirect

Desalination
journal homepage: www.elsevier.com/locate/desal

CFD prediction of the flashing processes in a MSF desalination chamber MARK

⁎
T.H. Nigim , J.A. Eaton
Mechanical Engineering, National University of Ireland Galway, Ireland

A R T I C L E I N F O A B S T R A C T

Keywords: Vapourisation inside a desalination flashing chamber arises from the energetic evaporation with self-boiling of
Flashing process liquid brine flow due to a reduction of pressure, and is a complex, multiphase, thermo-fluid phenomenon. A
MSF chamber computational field model for the flashing process inside a flashing chamber is developed around a two-phase
Evaporation zone VOF formulation. Two different phase-change mechanisms are allowed for, based on the saturation temperature
CFD prediction
and on the vapour pressure, respectively, that enable the model to compute the phase change regions, also the
Multiphase
level and the shape of the free surface. The model is applied to solve for steady multiphase flow inside a flashing
chamber without a baffle, and thus to predict the chamber flow details and behaviour. Thermo-fluid perfor-
mance results are presented in terms of flow patterns and distributions of hydrostatic pressure, temperature,
vapour volume fraction and mass transfer inside the flashing chamber. The average outlet temperature and
average vapour temperature agree well with the values of a real MSF plant, as does the estimated vapour
production rate. Two main mechanisms of phase change produced by the flashing process are captured in the
simulation. Vapour bubbles are formed at the entrance to the flashing chamber, and bubble production reduces
along its length. At the free surface of the liquid there is also a phase change and mass transfer. Computational
field modelling is valuable for a better understanding of the combined heat transfer, mass transfer and the fluid
dynamics during the flashing flow evaporation processes. Results can be used to estimate MSF design factors
such as non-equilibrium temperature difference and flashing efficiency.

1. Introduction Fig. 1 shows a schematic diagram of a horizontal flow MSF flashing

The multi-stage flash (MSF) desalination system was among the
earliest widely-used approaches to desalination [1], and continues as
one of the most common and simple methods in current use [2]. Cap-
able of producing fresh water of a very high purity, it is also suitable for
the largest-scale production. Indeed, since MSF relies on thermal input,
it is preferably to be built as a cogeneration plant for concurrent pro-
duction both of power and of fresh water [3]. Notwithstanding the
introduction of many alternative desalination processes, MSF remains
very important: in 2014 it accounted for about 22% of global desali-
nation plant production [4,5].
The fundamental principles in the multi-stage flashing flow process
consist first of producing water vapour from the brine, then of con-
densing the vapour to yield distilled water. The key mechanism is the
changing of phase of the brine, which is attained by causing the brine
pressure to fall below its vapour pressure: once the brine becomes
sufficiently superheated, and in the presence of nuclei, vapour bubbles
are produced as a consequence of phase change of the water. The re-
duction in pressure is arranged stepwise, in a continuous flow process,
to occur in successive series of flashing chambers known as stages.
chamber, together with a general configuration of the evaporation
zone. The evaporation zone is fed with brine through a submerged inlet
orifice, and comprises an open flow channel fitted with a flow baffle
leading to a submerged outlet: the outlet forms the inlet orifice to the
next stage. The stage consists of the following basic elements: inlet
orifice, brine pool, flow baffle, splash baffle, outlet orifice, demister,
condenser/preheater tubes, distillate tray, venting line and partition
walls. In a large-scale desalination plant, a flashing stage typically [6]
has width, height, and length dimensions of approximately
18 m × 4 m × 3 m, respectively.
Thermodynamically, the flashing chamber has the highest exergy in
the MSF system [7], and the evaporation zone may have the highest
exergy within the flashing chamber [8]. In this respect, the evaporation
zone is thus considered as the most important part of the stage [1], and
is the focus of this study.
The flashing chamber maybe treated as an open system, mainly
because the boundaries of the system are permeable both to energy and
to mass. In general, the flow field in the evaporation zone may be
classified as multiphase, complex, turbulent and unsteady, featuring a
submerged orifice flow, a wall jet, flow recirculation and an

⁎
Corresponding author.
E-mail address: t.nigim1@nuigalway.ie (T.H. Nigim).

http://dx.doi.org/10.1016/j.desal.2017.06.026
Received 18 July 2016; Received in revised form 14 June 2017; Accepted 25 June 2017
0011-9164/ © 2017 Elsevier B.V. All rights reserved.
T.H. Nigim, J.A. Eaton Desalination 420 (2017) 258–272

Nomenclature vv velocity vector of the vapour phase [m/s]

vy component of velocity in y-direction [m/s]
coeff rate factor of vapourisation or condensation (inverse of a w width of flashing chamber orifice [m]
relaxation time) in Lee Wen Ho model (Eq. (3)) [s− 1] W width of flashing chamber [m]
d vapour bubble diameter [m] x horizontal coordinate [m]
Fcol collapse coefficient [−] y vertical coordinate [m]
Fvap vapourisation coefficient [−] y+ wall-normal distance ≈ y ρτw μ [−]
g gravitational acceleration [m/s2]
h height of inlet orifice to flashing chamber [m] Greek
H height of flashing chamber [m]
L length of flashing chamber [m] α volume fraction [−]
LH latent heat [J/kg] αl liquid phase fraction [−]
ṁ l → v mass transfer rate from liquid to vapour phase [kg m3/s] αnuc nucleation site volume fraction [−]
ṁ v → l mass transfer rate from vapour phase to liquid [kg m3/s] αv vapour phase fraction [−]
M molecular weight [kg/kmol] β parameter in Lee Wen Ho phase-change model (Eq. (5))
n bubble density number [−] [m/s]
n unit vector, outward-normal to S [−] ε dissipation rate of turbulence kinetic energy [m2/s3]
NETD non-equilibrium temperature difference [K] k kinetic energy of turbulence [m2/s2]
ps′ gauge pressure [Pa] μ viscosity [Pa s]
p absolute pressure [Pa] ρl density of liquid [kg/m3]
pop flashing chamber operating pressure ρv density of vapour [kg/m3]
pv vapour pressure [Pa] σ surface tension [N/m]
q heat flux [W/m2] τw wall shear stress [Pa]
rc critical radius of the bubble [m] Ω control volume [m3]
R universal gas constant [J/kmol K]
R′ rate of mass change of a single bubble [kg/s] Subscripts
Rb bubble radius [m]
Rc mass transfer source term, bubble collapse (Eq. (6)) b bubble
[kg m3/s] i number of the stage (flashing chamber) in MSF plant
Re mass transfer source term, bubble growth (Eq. (6)) in inlet
[kg m3/s] l liquid phase
S control surface [m2] nuc nucleation
t time [s] out outlet
T temperature [K] op flashing chamber operating pressure
Tin inlet temperature [K] v vapour
Tl temperature of the liquid phase [K]
Tout outlet temperature [K] Abbreviations
Tsat saturation temperature [K]
Tv temperature of the vapour phase [K] CFD computational fluid dynamics
vb velocity of control surface [m/s] MSF multi-stage flash
vx component of velocity in x-direction [m/s] VOF volume-of-fluid

unconstrained, open surface. Interactions exist between the liquid and
vapour phases involving mass, momentum and energy exchanges re-
lated to vapourisation and condensation mass transfer, mean flow
gradients and turbulence, buoyancy and heat transfer.
Here, for simplicity, we consider two-dimensional flow in a flashing
chamber without any flow baffles. According to the phase type and to
the fluid behaviour, the chamber space may be regarded as divided into
a number of layers and regions. These are indicated in Fig. 2 (adapted
from Lior [9]). In the horizontal sense, there are two main regions: A –
the bubble nucleation region, near the inlet, which includes the sub-
merged jet and recirculating flow; and B - the channel flow, further
downstream, which is predominantly unidirectional. In the vertical
direction, the chamber space may be considered as comprising three
layers: I – the brine flow layer, near the base; II – the free surface; and
III - the vapour layer.
The flow may be generally classified as an open channel flow driven
by a submerged jet [10]. The submerged orifice functions as a sluice
gate, controlling the flow rate and brine level. Such flows are char-
acterised by Froude number and Reynolds number. Here the Froude
number, based on brine depth and mean inlet speed, lies around the
critical value, which may result in a wide variety of surface waves and
unsteady flow patterns: under these circumstances Froude number
effects may be expected to dominate. Moreover, the degree of sub-
mergence of the inlet, characterised by the ratio of brine level at inlet/
inlet orifice height, say, can also strongly influence the flow field, as
will the geometric aspect ratio of the chamber.
Within the brine flow layer, under a process of heterogeneous
flashing vapour bubbles emerge on nuclei of dissolved gas or of solid
particles entrained in the brine. Vapour formation within the brine
depends on the local pressure and temperature: brine pressure pL de-
pends on the local depth below the free surface. The local difference
between the vapour pressure in a bubble and the brine pressure, the
superheat term (pv − pl), is the key factor in determining whether the
bubble will nucleate and grow, or collapse and disappear.
1.1. Purpose of the present work
Many studies [11–40] — experimental, analytical, numerical si-
mulation — have been undertaken with a view to understanding the
mechanisms behind the evaporation zone and the flashing phenomena,
as well as to propose methods for improving the efficiency and per-
formance of the evaporation zone.
Our aim is to produce and validate a multiphase computational field
approach to predict the flow, heat and mass transfer in the evaporation

259
T.H. Nigim, J.A. Eaton
zone of an MSF stage. Once fully validated, the method is applicable to
assisting in the design procedure for MSF systems.
This paper is organised as follows: we next review (§1.2) some of the
previous computational studies for predicting the evaporation zone.
Our computational model for the flashing chamber is presented in §2,
together with a validation case based on an existing stage. An outline of
the mathematical modelling and computational procedure is also pro-
vided. In §3 we show the field prediction results of the thermo-fluid
performance (hydrostatic pressure, flow patterns, thermal performance,
vapour volume fraction and the mass transfer) inside the flashing
chamber. Results are discussed in §4, and our conclusions are presented
in §5.
1.2. Review of computational predictions for the evaporation zone
Seul et al. [41] presented a computational model for simulating the
thermal-hydrodynamic behaviour of the horizontal stream inside a
multi-stage flash evaporator without a flow baffle. The PSI-CELL
method of Crowe et al. [42] for disperse particle flows was adapted and
used to solve the two-dimensional, incompressible, steady, two-phase,
turbulent flow conservation equations of mass, momentum and energy
in conjunction with ancillary equations for motion and growth of va-
pour bubbles. The liquid (continuous) phase was treated using the
Eulerian field approach, whereas the vapour (disperse) phase was
computed with a Lagrangian model. Bubble nuclei entered the domain
at the same velocity as the surrounding liquid, with the initial size and
distribution of bubble nuclei on the inlet boundary specified as para-
meters. Cell-averaged bubble trajectories and cell-averaged bubble size
were computed as part of the solution. The method allowed for inter-
actions between the liquid phase and the vapour bubbles, in terms of
mass transfer (by evaporation), momentum transfer via interfacial
forces and energy transfer by vapourisation. Calculated results of the
flow patterns, in the form of liquid streamlines, temperature distribu-
tions and bubble trajectories, were presented. Seul et al. [41] employed
a standard k - ε turbulence model equation for the liquid phase only.
Only the brine layer was treated, and the free surface level and profile
was prescribed. The temperature boundary condition at the free surface
was determined from the convective heat transfer rate with surface
evaporation, based on kinetic theory. An unrealistic flat (horizontal)
free surface profile was specified as a boundary condition, and
Fig. 1. Schematic of a flashing chamber shape and evaporation zone elements.
260
Desalination 420 (2017) 258–272
interfacial stresses as well as surface tension effects were ignored in the
calculations. In practice, the flow is unsteady, but even in time-average
[43] the free surface can deviate significantly from being flat. No
consideration was given to the vapour layer, or to interactions between
the brine and vapour layers. Since the model did not have the ability to
predict either the brine level or the free surface shape, the imposition of
unrealistic free surface conditions will have adversely affected the
predicted thermo-fluid behaviour, and the pressure distribution and
flow pattern in particular.
In subsequent work, Seul et al. [44] simplified and refined their
bubble dynamic model, void fraction correlation and boundary condi-
tions, and applied their improved model to investigate the effects of the
liquid level on flow behaviour inside a multi-stage flash evaporator.
They confirmed that bubble motions are very much dependent on the
inlet velocity profile, and showed that the evaporation performance is
improved by employing lower liquid levels. Nonetheless, they retained
the free surface boundary condition as horizontal with zero shear.
Miyatake et al. [43] employed a two-dimensional, incompressible,
steady, single-phase, turbulent, isothermal, stream function-vorticity
numerical model in an attempt to reach a quantitative understanding of
the interaction between fluid mechanics and flash evaporation in a
flashing chamber with and without a flow baffle. Although a simple
one-equation, mixing-length turbulence model was used, the flow field
contained a number of distinctly different flow zones. For accurate flow
computations, each flow zone has a characteristic (mixing) length scale
which needs to be calculated. However, the number, type and shape of
flow zone depends on the configuration of the flashing chamber, and
are not know a priori; thus determination of such length scales is
cumbersome, and requires an iterative process starting, for instance,
with a laminar field solution. The curved free surface shape was ob-
tained from laboratory experiments, and prescribed as a zero shear
stress boundary condition.
Miyatake et al. [45] studied the relationship between the flashing
flow pattern and thermal non-equilibrium in a flashing chamber with
flow baffles of different locations and heights. They post-processed the
velocity and pressure distribution results from their previous study
[43], and took the computed streamlines as paths along which to solve
for the temperature drop between inlet and exit, for which they em-
ployed an empirical correlation for the temperature decrease rate of a
flashing liquid. By averaging over several streamlines they computed
T.H. Nigim, J.A. Eaton
Fig. 2. Flashing chamber without flow baffle – regions and zones (after Lior [9]).
the average temperature of the liquid at the exit. They found that the
non-equilibrium temperature difference, NETD = Tout − Tv, is reduced
by increasing the brine temperature, also when the flow rate of the
brine is increased.
Jin and Low [35] investigated the single-phase flow patterns in a
model flash evaporation chamber both experimentally, by making
particle image velocimetry measurements, and numerically, by per-
forming complementary two-dimensional, incompressible, steady,
single-phase, isothermal, turbulent flow field computations. The mea-
sured free surface profiles and inlet gate velocity distributions were
used as boundary conditions, and a standard k – ε turbulent flow model
was employed. Using streamline plots, for a simple horizontal chamber
without a baffle, they showed that the shape of the free surface as well
as details of the recirculating flow pattern depend strongly on the water
level, flow rate, inlet gate height and inlet velocity distribution.
Mansour et al. [46] used the CFX code to predict two-dimensional,
incompressible, steady, two-phase, adiabatic, turbulent flow of the
flashing process inside a flashing chamber with different numbers and
locations of triangular baffles. They employed a two-fluid (liquid water
and water vapour) Euler–Euler, flow formulation in conjunction with a
k – ε turbulence model. The flow was treated as two interpenetrating
continua, with individual conservation laws for mass and momentum
for each phase. The space occupied by each phase within a computa-
tional cell is defined by its local volume fraction, and is part of the flow
field solution. Measures of steam mass production and pressure drop of
the water were used for comparison purposes between different con-
figurations of baffle, and the computations indicated that the location
of baffles strongly affects the thermo-fluid performance. Results
showing contours of vapour volume fraction and of temperature,
pressure and velocity for the water component were presented. How-
ever, conventional flow patterns, in the form of streamlines or path-
lines, which are valuable is assimilating and assessing field results, were
not provided. The bubble nucleation region did not appear in the re-
sults. It is unclear what boundary conditions were employed for each of
the two fluids, and the physics and mechanism for phase change, if any,
were not explained.
As already observed, flashing chamber type flows are very complex
and difficult to predict. From this review of previous computational
work it is evident that better models are needed that are capable of
treating a number of significant aspects not hitherto allowed for. The
local pressure at a point in the brine layer is influenced by its sub-
mergence below the free surface; the local pressure affects the vapour
saturation pressure, and hence the local saturation temperature. In
particular, therefore, it is desirable to have an approach that enables the
shape, elevation and motion of the free surface to be predicted, rather
than to have to specify it empirically, or to impose it unrealistically as a
flat, horizontal boundary. Prediction of phase change and mass transfer
arising both from thermal effects (i.e. temperature difference) and from
mechanical effects (i.e. pressure difference) is also desirable. We next
present a new computational model which aims to address these as-
pects.
261
Desalination 420 (2017) 258–272
2. Current computational model for the flashing chamber
A numerical procedure for flashing flow is developed in this present
work and is applied to predict the multiphase flow field in the eva-
poration zone of an idealised, horizontal flow flashing chamber. Our
multiphase model is based on the FLUENT volume-of-fluid (VOF) code
implementation, and encapsulates two interaction mechanisms for
phase change — based on saturation temperature and vapour pressure,
respectively — that arise during the flashing process.
In addition to including phase change due to both pressure and
thermal effects, our method is novel in introducing and demonstrating the
facility to predict the profile as well as the elevation of the free surface,
both of which affect the pressure distribution in the brine, and thus are
central to stage performance.
We note that the method is capable of predicting the fully three-
dimensional, time-dependent behaviour of the evaporation zone, which
exhibits oscillating, unsteady flow phenomena characteristic of open
channel flows. However, for such calculations the computational time
and cost is very high: moreover, assimilating and presenting the copious
results of a transient analysis is time-consuming and challenging;
combined with an absence of measurement data for validation, it ap-
pears premature to allocate limited resources to such a simulation. For
pragmatic reasons, therefore, we confine our attention here to running
a steady-state, two-dimensional model of the evaporation zone and to
analysing the results of these predictions.
This model is applied to the evaporation zone of an existing MSF de-
salination system for which some data are available. Processing of the
results of the flow field computation enables a visualisation of the flashing
process, and phase change regions can be seen over the computational
domain. The fluid process parameters of hydrostatic pressure and particle
paths (streamlines in steady flow) of flashing flow inside the evaporation
zone are provided and presented. The thermodynamic process parameters
of temperature distribution, vapour volume fraction, heat transfer rate,
and mass transfer rate inside the evaporation zone are also shown.
2.1. Mathematical modelling
The choice of multiphase model is determined by the application,
and by the flow phenomena of the different phases that it is desired to
capture. In this application we simplify the fluids by assuming the
presence of two fluid phases only: pure water (liquid) and water va-
pour. For the present, therefore, real fluid aspects such brine/seawater
mixture properties, brine concentration, and the presence of entrained
solids and non-condensable gases are ignored. The concentration of
vapour varies widely in the evaporation zone of the flashing chamber
(Fig. 2). We anticipate a dilute, disperse flow of vapour bubbles in the
bubble nucleation region near the brine inlet, and a very high vapour
concentration from the free surface upwards. Apart from predicting the
flow fields in the brine and vapour layers, we wish to capture the lo-
cation of the interface between the layers. The combination of a sub-
merged sluice, open-channel flow interacting with recirculation and
T.H. Nigim, J.A. Eaton
phase change may be strongly unsteady, so the model chosen should be
capable of handling these interactions in a robust manner.
Mathematical modelling of two-phase flashing chamber flow has
been undertaken by an Euler-Euler, two-fluid approach [46], and by an
Euler-Lagrange approach [41,44]. Models for these and several other
multiphase formulations are now available in commercial codes. We
have selected an Euler-Euler, one-fluid approach, the Volume of Fluid
(VOF) method [48–64] as implemented in the FLUENT 14.5 finite vo-
lume code [48,49], which is the most appropriate for the present re-
quirement: in particular, it enables the prediction of free surface flows
of arbitrary interface shape.
We now outline the VOF approach used, together with the thermal
and mechanical models used to predict the flashing processes inside the
evaporation zone.
2.1.1. Volume of fluid (VOF) multiphase model
In the general Euler-Euler multiphase approach the different phases
(which may include bubbles and particles as well as multiple fluid
phases) are treated mathematically as interpenetrating continua on a
fixed Eulerian mesh. The concept of a phasic volume fraction is applied,
where the volume fractions of the different phases are assumed to be
continuous functions of space and time, and the sum of the volume
fractions is unity in each cell. For each phase conservation equations
are derived, with a similar equation structure for all phases.
Computing the free surface of flow is an extremely difficult class of
flows involving moving boundaries [49]. The VOF interface tracking
technique is a well-established [50,52,57,60] and validated [51,52,59]
approach, designed for two (or more) immiscible fluids where it is
desired to locate the position of the interfaces between the fluids.
Fig. 3 illustrates a group of rectangular cells in the vicinity of an
interface. The liquid region is shaded. For the flashing simulation, the
liquid phase fraction is presented as αl while the vapour phase is re-
presented as αv in each cell throughout the domain. Three cell types can
be identified: empty cells (where vapour phase αv = 0), cells full of
vapour phase (αv = 1), and cells containing the interface between va-
pour and liquid phase (0 < αv < 1). Based on the local value of αv and
αl, appropriate properties and variables will be assigned to each control
volume within the domain.
At the initial time of the simulation, the position of the boundary is
known. Its location at later times are to be determined as part of the
Fig. 3. Volume fraction on a discrete mesh.
262
Desalination 420 (2017) 258–272
solution. To reach this, the space conservation law (SCL) and boundary
conditions at the free surface must be used. The SCL is used to apply the
mass conservation equation in the limit of zero fluid velocity. The fol-
lowing expression is used to describe the conservation of space when
the control volume shape and/or position changes with respect to time.
d
dt
∫Ω dΩ − ∫S v b ∙ n dS = 0
(1)
where t is time, Ω represents the control volume, S is the control surface
with unit outward normal vector n , and v b is the control surface ve-
locity vector.
In the VOF model the fluids share a single set of momentum equa-
tions, so that the resulting velocity field is shared among the phases.
The energy equation is also shared among the phases, and energy and
temperature are treated as mass-averaged variables. Likewise, arising
from solution of a single set of turbulent transport equations, the tur-
bulence variables are shared. The volume fraction of each of the fluids,
in each computational cell, is computed throughout the domain and is
used to identify any interfaces that emerge as part of the solution.
The results are computed in terms of field properties of the mixture,
together with the phase (or void) fraction field. This means that the VOF
method does not provide separate information about each individual
phase — as does an Eulerian model that solves individual momentum and
continuity equations for every phase — but rather of the shared properties
of the (single-fluid) mixture, which may be regarded as its main limitation.
2.1.2. Phase change models
The phase change models based on local thermal (saturation tem-
perature) and mechanical (vapour pressure) effects, respectively, which
are used for building the flashing chamber simulation, are presented in
this section.
2.1.2.1. Phase change based on thermal effect
The mechanism of interphase mass transfer between the liquid and
the vapour phases is controlled by the vapour mass transfer equation
[48,63].
∂
(α v ρv ) + ⋅(α v ρv v v ) = ṁ l → v − ṁ v → l
∂t (2)
where subscript υ = vapour phase, α = volume fraction, ρv = vapour
T.H. Nigim, J.A. Eaton
density, v v = vapour phase velocity vector, ṁ l → v = mass transfer rate
from liquid to vapour phase, ṁ v → l = mass transfer rate from vapour
phase to liquid.
The Lee Wen Ho model [64] is a mechanistic model representing
phase change from liquid to vapour phase, as well as from vapour to
liquid phase, with a physical basis. The mechanism of the vapourisa-
tion-condensation model is based on checking the temperature of the
liquid phase. Since evaporation is a surface phenomenon, mass transfer
of the molecules takes place at a liquid-vapour interface. When mole-
cules have enough kinetic energy, they can escape from the free surface
as a vapour phase. This kinetic energy is a function of the saturation
temperature of the liquid. At higher temperatures, molecular kinetic
energy is greater.
The local saturation temperature corresponding to the local pressure
of the system is considered as an indicator for this model. In other
words, the vapourisation process takes place when the local tempera-
ture of the liquid phase is higher than the local saturation temperature,
while the condensation process takes place when the temperature of the
vapour phase is higher than the saturation temperature. Thus
if Tl > Tsat (vapourisation), then
(Tl − Tsat )
ṁ l → v = (coeff ) αl ρl
Tsat (3)
If Tv > Tsat (condensation), then
(Tv − Tsat )
ṁ v → l = (coeff ) α v ρv
Tsat (4)
where the coefficient, coeff, must be adjusted carefully in order to
achieve the best description of performance. It also should be inter-
preted as a relaxation time. The coefficient factor can be described in
the following expression
6 M ρl ⎞
coeff = β LH ⎛⎜ ⎟
d 2πRTsat ⎝ l ρv ⎠
ρ − (5)
where β is the accommodation coefficient, which defines the proportion
of vapour molecules being adsorbed at the liquid side of the interfacial
surface: β is a physical characteristic of the gas, and has a value of 1.0
near equilibrium conditions; d = vapour bubble diameter,
M = molecular weight, R = universal gas constant, LH = latent heat,
υ = vapour phase, α = vapour volume fraction, ρl = liquid density.
The vapourisation-condensation flux is based on the Hertz Knudsen
kinetic theory formula for a flat interface. In calculating the interfacial
area density it is assumed that the flow regime is disperse, also that all
vapour bubbles are spherical with the same diameter.
Here, to enable initial computations in the absence of accurate data
for the diameter of the vapour bubbles, and the value of coeff is set to
0.1, thus obviating the need to define the diameter.
2.1.2.2. Phase change based on mechanical effect
The mass transfer mechanism between liquid-vapour phases are
based on the vapour transport equation [48]:
∂
(α v ρv ) +  . (α v ρv v v ) = R e − R c
∂t (6)
where v = vapour phase, αv = vapour volume fraction, ρv = vapour
density, v v = vapour phase velocity vector, Re = mass transfer source
term connected to the growth of the vapour bubbles, Rc = mass transfer
source term connected to the collapse of the vapour bubbles. In the
FLUENT cavitation evaporation-condensation model these mass-
transfer source terms are modelled based on the Rayleigh-Plesset
equation, which describes the growth of a single vapour bubble in a
263
Desalination 420 (2017) 258–272
liquid. It is assumed that there are plenty of nuclei present for inception
of cavitation, and the primary focus is on accounting for bubble growth
and collapse.
The Zwart et al. model [65], which is derived from the generalised
Rayleigh-Plesset equation [66], is based on the assumption that bubbles
in the system are of equal size. While the generalised Rayleigh-Plesset
equation includes the influences of surface tension, liquid viscosity and
inertia on bubble growth, these terms are absent from Zwart et al.
model. In the present application, therefore, bubble growth is assumed
to be dominated by thermal effects.
In the Zwart et al. model the bubble density numbers (n) and the
mass change rate of a single bubble are used for calculating the total
interphase mass transfer rate per unit volume (R′):
DRb
R′ = n ⎛4πRb2ρv ⎞
⎝ Dt ⎠ (7)
where Rb is the bubble radius. To determine whether the bubble will
nucleate and grow, or collapse and disappear, an indicator formula is
required. A critical radius for the bubble is the key to creating such an
indicator. The following indicator is used in the model for this purpose:
If the local absolute pressure p ≤ pv, then
3 αnuc (1 − α v ) ρv 2 Pv − P
R e = Fvap
Rb 3 ρl (8)
If p ≥ pv, then
3α v ρv 2 P − Pv
R c = Fcol
Rb 3 ρl (9)
−6
Here the generalised bubble radius Rb is set at 10 m, the va-
pourisation coefficient Fvap is set to 50, and the condensation coefficient
Fcol is set to 0.01. By setting a volume fraction value for nucleation sites
as input, in conjunction with the generalised bubble radius the bubble
density number is determined. Here, for initial computations, the vo-
lume fraction of nucleation sites is specified as 0.0005.
2.2. Validation case study [46,47]
The present work is built and validated based on modelling the
evaporation zone in one stage of an existing MSF plant. The data used
are for the Sidi Krir Plant at Alexandria, Egypt. The geometry used to
define the computational domain and the operating conditions are
taken from published papers [46,47]. The MSF plant has 20 stages, with
a total production of desalinated water of 5000 m3/day.
The shape and dimensions of the computational domain can be seen
in Fig. 4. Here we simulate the evaporation zone in the first flashing
chamber of this plant. Operating conditions of the first flashing
chamber are:
1. Average inlet temperature of brine 383.15 K
2. Average outlet temperature of brine 379 K
3. Mass flow rate of the brine 1847 ton/h
4. Flashed vapour 12.5 ton/h
5. Flashing chamber operating pressure 1.023 bar
The mass flow rate and inlet temperature of the brine are used to
define inputs to the model, applied as boundary conditions, while for
validation the vapour production and temperature field are computed
and compared with measured values [46].
2.3. Computational case setup
In this section the simulation geometry with boundary conditions
T.H. Nigim, J.A. Eaton Desalination 420 (2017) 258–272

Fig. 4. Computational domain: evaporation zone dimensions and boundary conditions

(source data [46]).

Fig. 7. Hydrostatic pressure distribution along horizontal traverses at y = 0.03 m,

0.17 m, free surface line, 0.3 m.

and assumptions are presented. To predict the thermofluid flow field

inside the flashing chamber, we treat the flow as incompressible, two-
dimensional, steady, adiabatic, two-phase (water liquid and water va-
pour) and turbulent.

2.3.1. Computational domain and mesh

The dimensions of the flashing chamber are typical of an existing
stage of a MSF desalination plant (Fig. 4), with a streamwise length of
1.25 m and an inlet gate height of 60 mm. The computational domain is
bounded by walls on three sides, with one inlet and one outlet each for
the brine. Note that the upper (‘open’) boundary for vapour is situated
within the vapour zone, 0.35 m above the floor of the chamber: the
location of the free surface, which is unknown ab initio, will be de-
Fig. 5. Hydrostatic (gauge) pressure field.
termined by the operating conditions, and is computed as part of the
solution.
Quadrilateral rectangular shape elements, with six inflation layers
near the walls, are used for mesh generation. The total number of ele-
ments is 201,358.
Smooth walls are assumed for these computations. In this config-
uration we anticipate a wall jet developing along the floor of the
chamber, and wall regions experiencing flow separation, reattachment
and impingement. Thus, since much of the flow may be wall-domi-
nated, near the solid surfaces the mesh is refined and extended into the
viscous sublayer. The mesh is generated such that along the walls the
estimated non-dimensional wall-normal distance to the centre of the
first cell, y+, is < 5: this enables enhanced wall functions [48] to be
used along with the k – ε turbulence model to better resolve the flows in
these near-wall regions.

2.3.2. Boundary conditions

The boundary conditions applied are shown in Fig. 4. A uniform
inlet velocity profile of 1.5 m/s at 383.15 K is set, with a uniform va-
pour fraction of zero and standard default values for the turbulent
parameters. The base and vertical walls have no-slip and are adiabatic.
The brine outlet and open boundary are set as uniform pressure
boundaries with uniform vapour fractions of zero and one, respectively.
Fig. 6. Hydrostatic pressure distribution along vertical traverses at x = 0.31 m, 0.62 m, For this low-speed, incompressible flow it is appropriate to employ a
0.93 m. pressure-based solution algorithm. In this algorithm the pressure is
derived from iterative solution of the continuity and momentum
equations, and is computed as a gauge-pressure field. Thus all boundary
condition pressures (Fig. 4) are specified to FLUENT as gauge pressures

264
T.H. Nigim, J.A. Eaton
Fig. 8. Particle paths (streamlines).
Fig. 9. Mixture speed along vertical traverses at x = 0.31 m, 0.62 m, 0.93 m.
(ps′). To enable absolute pressures to be produced a reference pressure
is set at a reference point in the domain. The absolute pressure is then
calculated as the sum of the gauge pressure and the reference pressure.
Here the reference pressure is set, on the top of the vapour layer at a
point on the open boundary, to the saturation pressure of the flashing
chamber (pop = 1.023 bar, corresponding to the saturation temperature
of 373.15 K). The open boundary gauge pressure is therefore set to zero.
265
Desalination 420 (2017) 258–272
Fig. 10. Mixture speed along horizontal traverses at y = 0.03 m, 0.17 m, free surface line,
0.3 m.
We note that when gravitational acceleration is active in FLUENT, as in
the present computations, the hydrostatic component of pressure is
included in the gauge pressure field. Hereinafter this gauge pressure
combination will be referred to as the hydrostatic pressure.
The values of fluid properties are estimated from [67,68].
T.H. Nigim, J.A. Eaton Desalination 420 (2017) 258–272

2.4. Numerical schemes and convergence criteria

An implicit pressure-based solver is used in the VOF model to obtain

a steady-state solution. While in reality the flow will be time-depen-
dent, for the present, however, to simplify the analysis and processing
of results we confine consideration to the imposition of an artificial
steady state. To resolve the pressure-velocity coupling the FLUENT
implementation of the PISO (Pressure Implicit with Splitting of
Operators) algorithm is employed, and both the neighbour and skew-
ness corrections are used to improve the efficiency of PISO calculations
[62,69]. FLUENT uses a collocated scheme in which all variables, in-
cluding velocity and pressure, are stored at cell centres. Interpolation is
therefore required to compute the corresponding values of variables
and of their gradients at cell face locations. A least-squares cell-based
interpolation method is used for computing the gradients. The PRESTO
Fig. 11. Temperature field. scheme is employed for pressure interpolation. For improved accuracy,
second-order upwind differencing is applied for the spatial discretisa-
tion of the convective terms in the momentum, volume fraction, tur-
bulence variables and energy equations. During each iteration, for each
cell the mass-transfer and energy interaction between the two phases
are calculated, then returned as source terms to the particular transport
equations.
The convergence criteria are as follows: for each of the x-velocity, y-
velocity, k, ε and volume-fraction-vapour equations it is 10− 5, for the
continuity equation it is 5 × 10− 4 and for the energy equation it is
10− 6.
The simulation was run on an Intel Core2 quad-CPU Q6600 of
2.4 GHz. Three of the available four cores were used to run this model.
The simulation takes 15,000 iterations to converge, each iteration re-
quiring about 3 s to complete.

3. Results

The predicted steady-state results of the thermo-fluid performance

(hydrostatic pressure, flow patterns, thermal performance, vapour vo-
lume fraction and the mass transfer) inside the flashing chamber are
presented in Figs. 5 through 18.

3.1. Hydrostatic pressure distribution

Fig. 12. Temperature distribution along vertical traverses at x = 0.31 m, 0.62 m, 0.93 m.

Figs. 5, 6 and 7 show the hydrostatic pressure distribution

throughout the flashing chamber. A discontinuity in pressure gradient
is evident between the brine and the vapour layers, and the lower edge
of this discontinuity is taken to be the free surface. There is an essen-
tially linear increase in pressure with depth below the free surface.
Along a horizontal traverse at mid-orifice level (y = 0.03 m, Fig. 7) the
calculated pressure drop is 2179 Pa, corresponding to the pressure
difference required to drive the flow from the inlet to the outlet.

3.2. Particle paths and velocity distribution

Fig. 8 shows particle paths (streamlines), coloured by the speed of

the fluid mixture, throughout the flashing chamber. A detailed view of
the flow pattern in the inlet region is also provided. As commonly ob-
served in flashing chamber experiments [15–23], a distinct brine re-
circulation flow zone appears in the solution.
The vertical traverses (Fig. 9) indicate a sharp discontinuity in ve-
locity between the brine and vapour along the interface, with minima
occurring at the free surface. Along the mid-orifice level (Fig. 10) the
inlet flow decelerates due to jet spreading and subsequently accelerates
to a maximum of 1.88 m/s at the brine exit.

3.3. Thermal performance

Fig. 13. Temperature distribution along horizontal traverses at y = 0.03 m, 0.17 m, free
surface line, 0.3 m.
The rate of flashing is related to the thermal performance, or brine
temperature field. The flashing process across the flashing chamber

266
T.H. Nigim, J.A. Eaton Desalination 420 (2017) 258–272

Fig. 14. Vapour volume fraction ( αv) fields with two different ranges of scale; (a) upper field range [0–1], and (b) lower field range [0–9 × 10− 5].

Fig. 15. Vapour volume fraction distribution along vertical traverses at x = 0.31 m, Fig. 16. Vapour volume fraction distribution along horizontal traverses at y = 0.03 m,
0.62 m, 0.93 m. 0.17 m, free surface line, 0.3 m.

267
T.H. Nigim, J.A. Eaton
Fig. 17. Vapour volume fraction distribution along horizontal traverses y = 0.03 m.
causes a reduction of brine temperature, as can be seen in the overall
temperature field (Fig. 11), and also in the vertical and horizontal
temperature traverses presented in Figs. 12 and 13.
The decrease in temperature takes place in both the horizontal (x-)
and vertical (y-) directions, but at different rates. The differences in
range and scale between Figs. 12 and 13 should be noted. The tem-
perature gradient in vertical (y-) directions is large compared to the (x-)
horizontal. The calculated flashing down (Tin − Tout) is 4.55 °C.
3.4. Vapour volume fraction and mass transfer rate
Fig. 14 presents the distribution of vapour volume fraction ( αv)
throughout the flashing chamber, and for clarity two field images are
provided. The upper image (a) indicates the full range of vapour vo-
lume fraction scale from 0 to 1, and shows that above the free surface
there is a mixture of vapour and liquid phases. The lower set of images
(b) focuses on a restricted range, between 0 and about 0.0001, to better
illustrate the vapour phase distribution throughout the evaporation
zone. The peak values of vapour volume fraction occur at the free
surface and in the bubble nucleation region. In the bubble nucleation
region three different areas are numbered; the maximum bubble for-
mation takes place in area (1), in the inlet jet just below the top edge of
the inlet orifice.
Along vertical traverses above the free surface (Fig. 15), the vapour
volume fraction increase monotonically. Horizontally, (Fig. 16) the
values of vapour volume fraction decrease. The average vapour volume
fraction along the free surface is 6.84 × 10− 3. Fig. 17 shows the dis-
tribution of vapour volume fraction along y = 0.03 m, which passes
through the bubble nucleation region. This happens because of the
reduction in brine temperature, so the phase change rate is decreased.
Fig. 18 shows the mass transfer rate in the bubble nucleation region
and at the free surface layer inside the flashing chamber. The positive
value means that the phase changes from liquid to vapour phase (va-
pourisation process is occurred) while the negative value means the
phase changes from vapour to liquid phase (condensation process is
occurred).
4. Discussion
4.1. Hydrostatic pressure distribution
Hydrostatic pressure resists the flashing process [44] by increasing
268
Desalination 420 (2017) 258–272
vapour saturation pressure with depth, which affects the saturation
temperature. Thus mass transfer and phase change decrease with depth
due to the increase in pressure.
The hydrostatic pressure has the same effect as the brine level on the
flashing performance. The increase of the brine level leads to higher
thermodynamic non-equilibrium losses in the flashing chamber. Also,
the increase of the brine level leads to an increase in the pressure with
depth.
Generally, hydrostatic pressure [44] affects both the recirculation
zone shape and locations (it increases linearly with the brine level), and
bubble growth and rise locations. However, in our simulation case, the
hydrostatic pressure is not enough to block the bubbles formation and
the reduction in brine temperature.
4.2. Particle paths and the velocity distribution
Using the VOF in our simulations does not allow one to study the
velocity distribution separately for each phase in the mixture case. Thus
the velocity distributions in Figs. 8 to 10 are for the mixture only.
However, the VOF model makes it possible to deal with the vapour
layer as well with the brine layers.
Vertically, the velocity magnitude at the free surface is the
minimum. This happens in order to keep conservation of mass concept
over the free surface valid. For x = 0.93 (Fig. 9) the value of the
mixture speed is higher comparing with x = and 0.31, 0.62 m. This is
due to the higher phase changes rate and thus more vapour phase in-
cluded in the mixture as it can be noticed in Fig. 14 at the lower field.
Horizontally, the maximum speed is found to be 1.88 m/s at the end of
the flashing chamber. This happens to keep the conservation of energy
concept valid, since pressure in this point has reached the minimum
value.
The recirculation zone promotes bubble transport from the bubble
nucleation region to the free surface. Other information about the re-
lationship between fluid flow and its thermal behaviour can be reached
by using such simulations, i.e. rate at which turbulence kinetic energy
converts into thermal internal energy. This can provide information
about vertical heat flow, which is required in order for evaporating li-
quid close to the free surface.
4.3. Thermal performance
Generally, thermodynamics of the brine in the evaporation zone
inside the flashing chamber is difficult to predict [70]. However, our
computational model for the evaporation zone can be used to obtain
more information about the thermodynamics of the brine inside the
flashing chamber. We found that when the horizontal rate of decrease
of temperature is greater than the vertical rate of decrease of tem-
dT dT
perature ( dx > dy ), the non-equilibrium temperature difference (NETD)
dT dT
will be positive, while for dx < dy , NETD will be negative. This will be
discussed in future work.
The reduction in brine temperature in the evaporation zone occurs
due to the change phase. This reduction in temperature affects the mass
transfer rate and vapour volume fraction distribution (see Figs. 14 to
18). Also the mixing process affects the rate of temperature decrease.
In our simulation, the temperature gradient in the vertical (y-) di-
rection is large compared to the (x-) horizontal. The turbulent shear
mixing between the upper and lower brine layers influences the local
temperature gradient. The upper layer is colder than the lower, and
since the horizontal temperature gradient is small compared to the
vertical gradient, the preferred direction of heat transfer is vertical
rather than horizontal.
In our simulation we found that the driving force of the phase
change — the temperature difference between the brine temperature
and the saturation temperature — along the evaporation zone has re-
mained positive. In other words, while the phase change reduces the
T.H. Nigim, J.A. Eaton Desalination 420 (2017) 258–272

Fig. 18. Mass transfer rate field with three different ranges of scale; (a) [− 190, + 34] kg/m3s, (b) [− 26, + 8] kg/m3s, and (c) [− 0.0016, + 2.5] kg/m3s.

Table 1 occurrence of finite flashing? Does this mean that the flashing process
Comparison of predicted results and measured values. can be classified based on finite and infinite process? This is worthy of
investigation, and will be considered in future work.
Measured [46] Predicted

Average outlet temperature [K] 379 378.6

Average vapour temperature [K] 375 377.7 4.4. Vapour volume fraction and mass transfer rate
Total flashed vapour [tonne/h] 12.5 13.6

The vapour volume fraction distribution can be used as an indicator

for tracking the vapour phase and bubble formation.
brine temperature, its values are still higher than the saturation tem-
The maximum value of vapour volume fraction in the evaporation
perature corresponding to the flashing chamber pressure, thus main-
zone is 74 × 10− 6 (Fig. 17) and it is located in the bubble nucleation
taining the phase change along the evaporation zone length. This raises
region. The small value obtained is due to hydrostatic pressure (brine
the question of what if, at some location along the evaporation zone,
level) effects. On the other hand, the average value of vapour volume
the phase change reduces the brine temperature to the point where it
fraction at the free surface is 6.84 × 10− 3, which is higher than found
reaches the saturation temperature or lower? Does this lead to no more
in the bubble nucleation region. Thus phase change and mass transfer
change of phase taking place inside the evaporation zone, and the
occurs predominantly at the free surface (Fig. 18), similar to results

269
T.H. Nigim, J.A. Eaton
Fig. 19. Flashing regions as predicted by computational model. Compare with Fig. 2.
observed in the experiments of Gilbert [15].
The total vapourisation mass flow rate can be calculated by com-
puting the net produced vapour flux through the open boundary. From
the field results along the open boundary, we obtain a mean product of
y-component velocity times average fluid density of 0.433 kg/m2 s. The
flashing chamber length is 1.25 m, so that for the corresponding width
of 7 m [47], the estimated vapour mass flow rate is 3.79 kg/s
(13.6 tonne/h).
4.5. Validation
The main challenges for doing such predictions lie in obtaining
sufficient data to build and to validate the computational model. In
particular, it is difficult to find information pertaining to the interior of
the flashing chamber, i.e. bubble nucleation rate, bubble formation,
recirculation zone size and length, brine level, and also the shapes of
the orifices and their number. Also, for the few data that are available,
there is little or no information about measurement locations, methods
or uncertainties.
The VOF [51] multiphase model is well established, and has been
validated for predicting liquid-gas interfaces, including the free surface
shape and levels, in a variety of other applications [51,52,59]. Tami-
larasan [62] has validated the mechanistic Lee Wen Ho vapourisation-
condensation model for vapour volume fraction prediction, including
its application within the Fluent VOF implementation. We have in-
dependently validated the implementation of the Zwart et al. vapour-
isation-condensation model within the Fluent VOF code using data from
detailed and extensive experiments for isothermal flashing water flow
in a converging-diverging nozzle by Abuaf et al. [71]. Under non-
equilibrium (flashing) conditions, the measured average vapour volume
fraction at the outlet of the nozzle was 0.544, for which case our pre-
dicted value is 0.543 [72].
Our simulation of the evaporation zone of the flashing chamber is
validated by checking values of average temperature at the stage exit,
average vapour temperature above the liquid free surface and the
270
Desalination 420 (2017) 258–272
vapour mass flow rate (as noted above in §4.4). In comparing our re-
sults with the available data [46], we find a good agreement with the
values for the real plant (Table 1).
The predicted flashing mechanism in the simulation agrees with the
mechanism described in §1.2 (see Fig. 2). As seen in Fig. 19 vapour
bubbles are formed at the entrance to the flashing chamber, and bubble
production diminishes along its length. At the free surface of the liquid
there is also a phase change and mass transfer. Thus the two main
mechanisms of phase change produced by the flashing process are
captured in the simulation.
The predicted level of brine inside the flashing chamber is 0.2 m
above the inlet. This agrees with design recommendations by El-
Dessouky et al. [6], who recommended that the brine pool be higher
than the gate height by 0.2 m.
The calculated non-equilibrium temperature difference (NETD) va-
lues range from 4.8 to 5.5 °C, which is found to fall within the range of
design recommendations [73]: the NETD for a flashing chamber
without a baffle lies between 4 and 5 °C.
4.6. Future work
From the present work it is clear that computational field modelling
can be invaluable for a better understanding of the combined heat
transfer, mass transfer and the fluid dynamics during the flashing flow
evaporation processes. We note that our method is capable of predicting
the fully three-dimensional, time-dependent behaviour of the eva-
poration zone, which exhibits oscillating, unsteady flow phenomena
characteristic of open channel flows. While feasible, for such calcula-
tions the computational time and cost is very high: moreover, assim-
ilating and presenting the copious results of a transient analysis is time-
consuming and challenging. Despite this potential, as already noted in
§4.5 the key obstacle nonetheless remains the lack of measurements
and data for validation.
In the interim, the current model, run in a steady, two-dimensional
mode, will be used for studying and comparing the effects of variations
T.H. Nigim, J.A. Eaton
in the brine temperature, velocity and flashing chamber geometry on
the thermo-fluid behaviour inside the evaporation zone.
The flashing process may be classified as a finite or infinite
(flashing) process. Depending on which of the two flashing process
variations occurs — finite or infinite —may determine the thermo-fluid
behaviour inside the evaporation zone inside the flashing chamber, and
will also be investigated in future work.
5. Conclusions
• A VOF two-phase computational model of the evaporation zone
inside a flashing chamber has been developed. The model includes
heat transfer and mass transfer, as well as turbulent flow.
• Two different phase-change mechanisms are allowed for, based on
the saturation temperature and the vapour pressure, respectively.
These enable the model to compute the phase change regions, also
the shape of the free surface.
•
The model has been applied to a typical horizontal-flow flashing
chamber without a flow baffle. In a preliminary, steady-state solu-
tion the flow field features, including the recirculation zone and the
free surface level and shape are predicted. The simulation results
give average outlet temperature and average vapour temperature
values that agree closely with those of a real MSF plant.
• Computational field modelling is valuable for a better under-
standing of the combined heat transfer, mass transfer and the fluid
dynamics during the flashing flow evaporation processes.
• Results can be used to estimate MSF design factors such as the non-
equilibrium temperature difference, NETD, flashing down and
flashing efficiency.
• The model can be applied for further studies such as investigations
of the effects of the brine temperature, velocity and flashing
chamber geometry on the thermo-fluid behaviour for both finite and
infinite flashing process classifications.
Acknowledgments
This work was funded through a postgraduate Research Fellowship
from the College of Engineering and Informatics, National University of
Ireland, Galway, Ireland.
References
[1] A.H. Khan, Desalination Processes and Multi-stage Flash Distillation Practice,
Elsevier, Amsterdam, 1986.
[2] L.K. Wang, J.P. Chen, Y. Hung, N.K. Shammas, Membrane and Desalination
Technologies, Springer, London, 2011.
[3] Fichtner GmbH, MENA Regional Water Outlook Part II: Desalination Using
Renewable Energy, Report, Germany, (2011).
[4] International Desalination Association, IDA Desalination Yearbook 2013–2014,
Media Analytics Ltd., Oxford, United Kingdom, 2014.
[5] F.A. Al-Marzooqi, A.A. Al-Ghaferi, I. Saadat, N. Hilal, Application of capacitive
deionisation in water desalination: a review, Desalination 342 (2014) 3–15.
[6] H.T. El-Dessouky, H.M. Ettouney, Fundamentals of Salt Water Desalination, Elsevier
Science B.V., Amsterdam, the Netherlands, 2002.
[7] D.W. Abueidda, M. Gaddalla, Thermodynamics analysis of integrated solar-based
multi-stage flash desalination system, ASME International Mechanical Engineering
Congress & Exposition, IMECE2012-87729, Nov 2012.
[8] M. Al-Weshahi, Exergy Efficiency Enhancement of MSF Desalination by Heat
Recovery From Hot Distillate Water Stages, Newcastle University, UK, 2011.
[9] N. Lior, Some basic observations on heat transfer and evaporation in the horizontal
flash evaporator, Desalination 33 (1980) 269–286.
[10] R. Rautenbach, S. Schäfer, S. Schleiden, Improved equations for the calculation of
non-equilibrium temperature loss in MSF, Desalination 108 (1996) 325–333.
[11] B.M. Fabuss, A. Korosi, Boiling point elevations of sea water and its concentrates, J.
Chem. Eng. Data 11 (October 1966) 606–609.
[12] AMF, Cuno Engineering Corporation Maximum Division, Research and
Development of Multieffect, Multistage Flash Distillation Process, OSW R & D
Annual Report 15 (October 1966).
[13] Westgrath W. Ltd., Chamber Geometry in Multi-stage Flash Evaporator-separator
Studies, OSW R & D Progress Report No. 271, (July 1967).
[14] W.R. Williamson, F.W. Gilbert, Investigation of Multi-stage Flash Phenomena in a
Three-stage Test System, OSW R & D Progress Report No. 525, (Feb 1968).
[15] F.W. Gilbert, An Experimental Investigation of Steady State Characteristics of
271
Desalination 420 (2017) 258–272
Flashing Salt Solution in Open Channel Flow Between Two Submerged Orifices at
Low Pressure (Ph.D. Thesis), University of Connecticut, Connecticut, 1968.
[16] O. Miyatake, K. Murakami, Y. Kawata, T. Fujii, Fundamental experiments of flash
evaporation, Heat Transf. Jpn. Res. 26 (1972) 189–198.
[17] T. Fujii, O. Miyatake, T. Tanaka, T. Nakaoka, H. Matsunaga, N. Sakaguchi,
Fundamental experiments on flashing phenomena in a multi-stage flash evaporator,
Heat Transf. Jpn. Res. 1 (1975) 207–212.
[18] O. Miyatake, T. Fujii, T. Tanaka, T. Nakaoka, Flash evaporation phenomena of pool
water, Heat Transf. Jpn. Res. 1 (1975) 393–398.
[19] O. Miyatake, T. Fujii, T. Hashimoto, An experimental study of multi-stage flash
evaporation phenomena, Heat Transf. Jpn. Res. 1 (1975) 454–459.
[20] M.A. Darwish, S.G. Serageldin, H.T. El-Dessouky, Flashing in desalination units,
Desalination 19 (1976) 93–101.
[21] O. Miyatake, T. Nakaoka, T. Fujii, The effect of nucleation on flash evaporation,
Heat Transf. Jpn. Res. 29 (1976) 228–231.
[22] T. Sugeta, S. Toyama, An experimental study of flash evaporation phenomena at
high liquid flow rate, Heat Transf. Jpn. Res. 4 (1978) 148–153.
[23] N. Lior, R. Greif, A system for the experimental study of flash evaporation,
Desalination 31 (1979) 87–99.
[24] N. Lior, E. Nishiyama, The effect of gas bubbles on flash evaporation, Desalination
45 (1983) 231–240.
[25] N. Lior, Formulas for calculating the approach to equilibrium in open channel flash
evaporators for saline water, Desalination 60 (1986) 223–249.
[26] S. Gopalakrishna, V.M. Purushothaman, N. Lior, An experimental study of flash
evaporation from liquid pools, Desalination 65 (1987) 139–151.
[27] S. Gopalakrishna, N. Lior, Analysis of bubble translation during transient flash
evaporation, Int. J. Heat Mass Transf. 35 (1992) 1753–1761.
[28] O. Miyatake, Comparative study of flashing evaporation rate, Desalination 96
(1994) 163–171.
[29] N. Lior, E. Nishiyama, Some experiments on flash evaporation enhancement by
[30]
electrolytically generated bubbles, Desalination 100 (1995) 71–76.
O. Miyatake, I. Tanaka, N. Lior, A simple universal equation for bubble growth in
pure liquids and binary solutions with a nonvolatile solute, Int. J. Heat Mass Transf.
40 (1997) 1577–1584.
[31] J.-I. Kim, N. Lior, Some critical transitions in pool flash evaporation, Int. J. Heat
Mass Transf. 40 (1997) 2363–2372.
[32] H.E.S. Fath, The non-equilibrium factor and the flashing evaporation rate inside the
flash chamber of a multi-stage flash desalination plant, Desalination 114 (1997)
[33]
277–287.
W.X. Jin, S.C. Low, S.C.M. Yu, Some experimental observations on the single and
multi-phase flow patterns in a model flash evaporation chamber, Int. Commun.
Heat Mass Transf. 26 (1999) 839–848.
[34] O. Miyatake, Y. Koito, K. Tagawa, Y. Maruta, Transient characteristics and per-
formance of a novel desalination system based on heat storage and spray flashing,
Desalination 137 (2001) 157–166.
[35] W.X. Jin, S.C. Low, Investigation of single-phase flow patterns in a model flash
evaporation chamber using PIV measurement and numerical simulation,
Desalination 150 (2002) 51–63.
[36] X. Shichang, L. Wencai, W. Shichang, Experimental studies on orifice behavior and
two-phase flow in flash chamber, Desalination 150 (2002) 93–98.
[37] N. Lior, R. Chung, O. Miyatake, Correlations (updated) for predicting the flow
through MSF plant interstage orifices, Desalination 151 (2003) 209–216.
[38] D. Saury, S. Harmand, M. Siroux, Flash evaporation from a water pool: influence of
the liquid height and of the depressurization rate, Int. J. Therm. Sci. 44 (2005)
953–965.
[39] A. Cipollina, Experimental Study and Dynamic Modelling of Multi Stage Flash
Desalination Units (Ph.D. Thesis), Università di Palermo, Palermo, 2005.
[40] A. Cipollina, G. Micale, L. Rizzuti, Investigation of flashing phenomena in MSF
chambers, Desalination 216 (2007) 183–195.
[41] K.W. Seul, S.Y. Lee, Numerical predictions of evaporative behaviour of horizontal
stream inside a multi-stage-flash distillator, Desalination 79 (1990) 13–35.
[42] C.T. Crowe, M.P. Sharma, D.E. Stock, The particle-source-in cell (PSI-CELL) method
for gas droplet flow, J. Fluids Eng. 99 (1977) 325–332.
[43] O. Miyatake, T. Hashimoto, N. Lior, The liquid flow in multi-stage flash evapora-
tors, Int. J. Heat Mass Transf. 35 (1992) 3245–3257.
[44] K.W. Seul, S.Y. Lee, Effect of liquid level on flow behaviour inside a multi-stage
flash evaporator - a numerical prediction, Desalination 85 (1992) 161–177.
[45] O. Miyatake, T. Hashimoto, N. Lior, The relationship between flow pattern and
thermal non-equilibrium in the multi-stage flash evaporation process, Desalination
91 (1993) 51–64.
[46] K.M. Mansour, H.E.S. Fath, Numerical simulation of flashing process in MSF flash
chamber, Desalin. Water Treat. 51 (2013) 2231–2243.
[47] K.M. Mansour, H.E.S. Fath, O. El-Samni, Computational fluid dynamics study of
MSF flash chambers sub-components; I – vapor flow through demister, The
Fifteenth International Water Technology Conference, IWTC15. Alexandria, Egypt,
2011.
[48] ANSYS Fluent Theory Guide, Release 14.5, (2016).
[49] J.H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, Springer, New
York, 2002.
[50] W.F. Noh, P. Woodward, SLIC (Simple Line Interface Calculation), in: A.I. van de
Vooren, P.J. Zandbergen (Eds.), Proceedings, Fifth International Conference on
Fluid Dynamics, Lecture Notes in Physics, 59 Springer, Berlin, 1976, pp. 330–340.
[51] B.D. Nichols, C.W. Hirt, Numerical simulation of boiling water reactor vent-clearing
hydrodynamics, Nucl. Sci. Eng. 73 (1980) 196–209.
[52] C.W. Hirt, B.D. Nichols, Volume of fluid (VOF) method for the dynamics of free
boundaries, J. Comp. Phys. 39 (1981) 201–225.
T.H. Nigim, J.A. Eaton
[53] J.M. Hyman, Numerical methods for tracking interfaces, North-Holland Physics
Publishing Division, Physica 12D (1984) 396–407.
[54] O. Ubbink, Numerical Prediction of Two Fluid Systems With Sharp Interfaces (Ph.D.
Thesis), University of London, London, 1997.
[55] W.J. Rider, D.B. Kothe, Reconstructing volume tracking, J. Comp. Phys. 141 (1998)
112–152.
[56] R. Scardovelli, S. Zaleski, Direct numerical simulation of free surface and interfacial
flow, Ann. Rev. Fluid Mech. 31 (1999) 567–603.
[57] S. Kvicinsky, F. Longatte, J.L. Kueny, F. Avellan, Free Surface Flows: Experimental
Validation of Volume of Fluid (VOF) Method in the Plane Wall Case, Proceedings of
the 3rd ASME, JSME Joint Fluids Engineering Conference, San Francisco,
California, (1999).
[58] R. Saurel, R.A. Abrall, Multiphase Godunov method for compressible multifluid and
multiphase flows, J. Comp. Phys. 150 (1999) 425–467.
[59] D.M. Hargreaves, H.P. Morvan, N.G. Wright, Validation of the volume of fluid
method for free surface calculation: the broad-crested weir, Eng. Appl. Comput.
Fluid Mechanics 1 (2007) 136–146.
[60] V. Hernandez-Perez, Gas-liquid Two-phase Flow in Inclined Pipes (PhD thesis),
University of Nottingham, 2008.
[61] A.P. Laleh, W.Y. Svrcek, W.D. Monnery, Design and CFD studies of multiphase se-
parators – a review, Can. J. Chem. Eng. 90 (2012).
[62] A. Tamilarasan, Numerical Modelling of Heat Transfer and Evaporation
Characteristics of Cryogenic Liquid Propellant (Master of Science in Mechanical
Engineering Thesis), University of Washington, 2015.
[63] K.M. Wrobel, Modelling and Numerical Simulation of Flows With Evaporation and
Condensation (Ph.D. Thesis), University of Twente, Enschede, 90-365-2387-7,
2006.
272
Desalination 420 (2017) 258–272
[64] W.H. Lee, A Pressure Iteration Scheme for Two-phase Modeling, Technical Report
LA-UR 79-975, Los Alamos Scientific Laboratory, Los Alamos, New Mexico, 1979.
[65] P.J. Zwart, A.G. Gerber, T.A. Belamri, Two-phase flow model for predicting cavi-
tation dynamics, Fifth International Conference on Multiphase Flow, Yokohama,
Japan, 2004.
[66] C.E. Brennen, Fundamentals of Multiphase Flow, Cambridge University Press, 2005.
[67] N.B. Vargaftik, B.N. Volkov, L.D. Voljak, International tables of the surface tension
of water, J. Phys. Chem. 12 (1983) 817–820.
[68] M.H. Sharqawy, J.H. Lienhard V, S.M. Zubair, Thermophysical properties of sea-
water: a review of existing correlations and data, Desalin. Water Treat. 16 (2010)
354–380.
[69] H. Versteeg, W. Malalasekera, An Introduction to Computational Fluid Dynamics:
The Finite Volume Method, 2nd ed., Prentice Hall, 978-0-13-127498-3, 2007.
[70] R. Rautenbach, S. Schafer, Inter-stage Brine Flow in MSF Chambers, Desalination
and Water Resources: History, Development and Management of Water Resources
vol. I, (DESWARE). United Nations Educational, Scientific and Cultural
Organization (UNESCO), & Encyclopaedia of Life Support Systems (ELSS), 978-1-
84826-419-9, 2010.
[71] N. Abuaf, B.J.C. Wu, G.A. Zimmer, P. Saha, A Study of Non-Equilibrium Flashing of
Water in a Converging-Diverging Nozzle: Volume 1 – Experimental, United States
Nuclear Regulatory Commission Office of Nuclear Regulatory Research, 1981.
[72] T.H. Nigim, Computational Modelling of Thermofluid Flashing in MSF Desalination
(Ph.D. Dissertation), National University of Ireland, Galway, 2017 (submitted for
publication).
[73] Desalting Systems and Services, Inc., Design and Furnishing Flash Enhancers for
MSF Test Module, OSW Res. Devel. Progress Report No. 565, (1971).