See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/349863740

Thermal non-equilibrium in spray flash with vacuum vapor extraction

Article in Desalination · July 2021

DOI: 10.1016/j.desal.2021.115023

CITATIONS READS

8 212

4 authors:

Guangyu Guo Bo Zhang

New Jersey Institute of Technology New Jersey Institute of Technology
14 PUBLICATIONS 114 CITATIONS 12 PUBLICATIONS 146 CITATIONS

SEE PROFILE SEE PROFILE

Chao Zhu Zhiming Ji

Independent Researcher New Jersey Institute of Technology
67 PUBLICATIONS 1,149 CITATIONS 69 PUBLICATIONS 1,395 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Guangyu Guo on 15 March 2021.

The user has requested enhancement of the downloaded file.

 Desalination xxx (xxxx) 115023

Contents lists available at ScienceDirect

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

Thermal non-equilibrium in spray flash with vacuum vapor extraction

Guangyu Guo, Bo Zhang, Chao Zhu ⁎, Zhiming Ji
Department of Me ch an ical an d Indu strial Engineering, New Jersey Institut e of Tech no logy, Newark, NJ 07102, United St at es of America

ARTICLE INFO ABSTRACT

Keywords: The spray flash is the core procedu re in spray flash desalination, wh ich is an intense ph ase-ch anging
Spra y flas h process with great potential in efficient low-temp erature evaporation and high evaporation capacity. In a
Va cuum va por extraction vacuum ing spray flash with active vapor extraction and absenc e of external heating, the latent heat is self-
Process mo deling
supp lied by the evaporating droplets, resulting in a temperature difference be tween continued cooling
CF D si mula tion
droplets and the hotter yield vapor along with the spray transport. Understanding such therma l non-
Poly dispersi on effect
equilibrium be tween vapor and droplets is cu rial and be neficial to desalination indu stry for improving
evaporation efficiency and energy recovery. This paper aims to establish an integrated experime nt-
modeling-CFD me thodology to stud y the spray flash and its therma l non-equilibrium un der active vacuum -
ing. In this work, various effects on the spray flash, includ ing the polydispersion of atomized sprays, the
heat condu ction coupled evaporation of droplets, as well as the non-un iformity in vapor transport, are dis-
cu ssed. The proposed me thodology is based on a Lagrangian-Eulerian modeling approach , wh ich is calcu -
lated by a simplified process model and nu me rically solved via compu tational fluid dynamics (CFD) of AN-
SYS FLUENT with user-defined func tions (UDF), respectively. Two experime ntal systems are developed:
One is for calibrating droplets' velocity and size distribution; the other is a lab-scale vacuu m spray flash sys-
tem for me asuring the evaporative ch aracteristics. A dime nsionless corrective me thod for the lump ed heat
capacity model is proposed in order to improve the accu racy of discrete ph ase modeling in CFD simulation.
Theoretical predictions by the process model and CFD simulation me et good agreeme nts with experime nts.
Case stud ies of the operation parame tric effects on flash ch aracteristics are demonstrated via both modeling
predictions and experime ntal me asureme nts. The results indicate that the temperature difference be tween
the extracted vapor and the disch arged water can be effectively generated du ring the process. Such a gap is
positively impacted by the operating vacuu m level and the initial spray temperature. The reported therma l
non-equilibrium ph enomenon indicates highly efficient evaporation of spray flash in an active-vacu um ing
environme nt, and such potential could be further utilized in desalination.
1. Introduction heated surfa ce, such as heat exchanger tubes with steam superheating,
but also leads to surfa ce scaling problems that may significantly reduce
Spray flash desalination is one of the most promising thermal distil- the heat and mass transfer efficiency [5]. Therefore, in most industrial
lation methods for solving the global wa ter scarcity [1–3], in which the applications of desalination, the flash evaporation is usually generated
vapor separation is realized by the process of spray flash [4]. Therefore, via the second wa y (in an under-saturated pressure environment) to
studying and understanding the mechanisms of spray flash evaporation avoid scaling problems [6–8], with typical modes of pool flash or spray
is crucial for us to improve and optimize the industrial processes of flash [8]. Pool flash, a more traditional distillation method, can be a
spray flash desalination. steady-state evaporation process with continued heating to the evapo-
Flash evaporation is a very fa st evaporation phenomenon that oc- rating liquid and continued extraction of generated vapor [9]. Com-
curs when a liquid medium becomes superheated due to a sudden envi- pared to that, spray flash is typically an unsteady-state evaporation
ronmental change, such as by directly contacting a superheated solid process with continued cooling of droplets since the latent heat is
surfa ce or by being exposed to a space of under-saturated pressure mostly self-supplied from the internal energy of droplets [10]. Ta king
(such as a vacuum environment). However, realizing the flash evapora- advantage of a dramatically increased specific area for evaporation in
tion by direct surfa ce contact in the industry not only requires a super-

⁎ Correspond ing author.

E-mail address: ch ao.zhu @njit.edu (C. Zh u).

https://doi.org/10.1016/j.desal.2021.115023
Received 13 December 2020; Received in revised form 6 February 2021; Accepted 14 February 2021
0011-9164/© 2021
G. Guo et al.
an atomized spray [11], the spray flash can achieve a much higher
evaporation rate than the pool flash [4].
Specifically, spray flash in a vacuuming environment is a special
mechanical process of droplet evaporation, where an atomized liquid
spray is injected into a vacuum chamber whose pressure is maintained
by a vacuum pump by extracting the generated vapor out of the cham-
ber [12]. This simple spray flash has many unique advantages, includ-
ing the low-temperature evaporation, high evaporation rate within a
relatively compact chamber, and no scaling problems [8]. Neverthe-
less, such a simple process involves very complicated phase transfer and
transport mechanisms, with the integrated coupling of mass, momen-
tum, and energy among different phases of droplets and vapor under
highly non-equilibrium and non-uniform transport conditions.
According to previous experimental studies in recent decades, the
evaporation process in a spray flash can be af fected by many spray
characteristics, such as the initial temperature of sprayed droplets [13],
the direction and velocity of injection [14,15], and the spray flow rate
[16]. For example, a higher level of superheating of the injected spray
normally leads to higher levels of evaporating rate, amount, and effi-
ciency. Even a spray flash with a low level of superheating, for in-
stance, 2 °C, may also be effectively completed [15]. Likewise, a spray
upwa rdly injected against gravity may also evaporate better than a
downwa rd one due to a longer residence time of droplets in the trans-
port and evaporation process [15]. While the experimental studies can
result a few semi-empirical correlations of some parametric effects, a
comprehensive understanding and reliable quantitative predictions of
the evaporation and transport characteristics of a spray flash have to
be obtained from the first-principle-based mechanistic modeling.
In theory, droplet evaporation is basically governed by two coupled
mechanisms: the heat transfer to the interfa ce from the inner liquid side,
and the heat and mass transfer on the evaporating interfa ce from the
gaseous ambient [17]. In the case of droplet flash evaporation of pure
substance in a vacuuming environment, the evaporation rate can be
limited by the heat transfer from the liquid side due to the intense heat
consumption of phase change by vacuuming and no avail of heat trans-
fer to the evaporating surfa ce from the ambient side [18]. Although the
relative motion between droplets and extracted vapor flow may cause
asymmetric and enhanced convection [19], such asymmetry and en-
hancement can be ignored when the pressure difference between the
saturated vapor-liquid phase on the evaporating surfa ce of the droplet
and the vacuuming environment is orders-of-magnitude higher than
that induced from the vapor flow over the droplets [11].
So fa r, all evaporation models are based on the phase transfer theo-
ries under idealized conditions such as near phase equilibrium (where
Chapman-Enskog theory can be applied, or Boltzmann probability dis-
tributions become valid) and ideal gas approximation for the vapor
phase [17,20–26]. In such approaches, the evaporation rate models are
represented by the kinetic theory model [21,27] and diffusion-
controlled model [24], whereas the evaporation heat must be provided
from the coupled heat transfer models governed by the heat transfer
from the vapor side (by conduction or convection) and ambient envi-
ronments (such as thermal radiations), heat conduction from droplet
(with or without considerations of internal motions) or a combination
of all heat transfer modes. It must be noted that neither the vacuuming
flash is near phase equilibrium nor the saturated vapor on the evapora-
tion surfa ce follows the ideal gas law. Nevertheless, there has been no
better modeling available for the estimation of droplets evaporation
rate in a vacuuming flash. As a result, both kinetic theory model and
diffusion-controlled model have been extended into the evaporation
modeling of spray flash [28–34].
A very important feature in flash evaporation modeling of an atom-
ized spray is its polydispersion or non-uniform distributions in droplet
size, droplet velocity, injection angle, as well as the number density of
droplets or flow rate [35–37]. The polydispersion, especially in typi-
cally widely-distributed droplet sizes, causes a strong non-uniformity in
2
Desalination xxx (xxxx) 115023
flash evaporation via the significant differences in exposed evapora-
tion area as well as in the transport characteristics (such as residence
time) of droplets of different sizes or injection velocities [11]. Unfortu-
nately, this polydispersion depends not only on the types of nozzles and
physical mechanisms of atomization but also on the operation condi-
tions, which leads to the nonexistence of universal modeling to predict
the atomized spray characteristics at the nozzle tip or injection inlet
[11,38]. Consequently, the details in spray polydispersion have to be
experimentally determined case-by-case. The polydispersion in atom-
ized spray also imposes greater challenges to the associated evapora-
tion and transport modeling. For example, an overall averaged equiva-
lent diameter for hydrodynamic transport (such as the hydraulic equiv-
alent diameter of all droplets) can be quite different from an overall av-
eraged surfa ce-area equivalent diameter for evaporation. That means
no single averaged diameter can reasonably reflect both hydrodynamic
transport and evaporation characteristics in the modeling of an atom-
ized spray flash. So fa r, most models in spray flash ignore such polydis-
persion for simplicity by assuming all droplets have the uniform size,
mass flux, and velocity [28,39,40].
However, spray polydispersion may have a significant impact on
the flash evaporation under certain circumstances. For example, for a
spray flash in an evaporation chamber of limited dimension, some fine
droplets may be able to complete their evaporation within the cham-
ber, while larger ones may not have enough time to complete the va-
porization before colliding onto the chamber boundary or exiting out
of the chamber. Using a uniform size for the droplet evaporation and
transport (even averaged over a droplet size distribution) may misrep-
resent the total surfa ce area of evaporation or phase transfer and the
non-uniformity in phase transport. Hence a proper modeling considera-
tion is needed to account for this polydispersion effect, such as adopting
a set of sub-grouped droplet sizes and mass fluxes to mimic the polydis-
persed spray transport. Consequently, in this study, a combined experi-
ment-modeling methodology is established to introduce the polydisper-
sion effect of atomized sprays and the vacuuming effect on the evapo-
ration capacity in a spray flash, which is realized by adopting the La-
grangian approach for the discrete phase of droplets [12].
In a vacuum spray flash that is absent of external heating, the la-
tent heat is self-supplied by the evaporating droplets, resulting in a con-
tinued decrease in droplet temperature. Moreover, with the vapor ex-
traction by vacuuming, the extracted vapor is expected to possess a
higher temperature than the remaining droplets or liquid residue. Such
a temperature difference will demonstrate such a high thermal non-
equilibrium process is very efficient in evaporation, which could benefit
many industrial processes such as thermal distillation. Moreover, the
thermal energy in the vapor of higher temperatures could also be recov-
ered to further improve energy efficiency [41]. However, in most indus-
trial multi-stage flash desalination plants, the superheat level of the liq-
uid in each evaporator remains relatively low, in which the vapor ex-
traction is passively condensation controlled by the next stage [42–46].
Therefore, in this study, the evaporation chamber is directly connected
with a vacuum source on purpose (named active vapor extraction) to
obtain a high thermal non-equilibrium condition during spray flash.
Meanwhile, this study aims to investigate such interesting thermal non-
equilibrium between the extracted vapor and discharged spray residue,
characterized by the temperature difference between the two, during a
vacuum spray flash.
For vacuum spray flash in a bounded environment with active va-
por extraction, due to the non-uniformed distribution of vapor sources
and velocities caused by spray polydispersion, as well as the thermal
non-equilibrium of vapor exists in the full-field ambient all the time, a
full-scale computational fluid dynamics (CFD) study is desired to inves-
tigate such non-uniformity and thermal non-equilibrium of vapor. Sev-
eral CFD studies on spray flash have been reported based on different
applications [47–49]. The spray flash evaporation in all these CFD
studies is based on the lumped heat capacity method of the droplet,
G. Guo et al.
which is due to the simplified point-treatment of droplets in the discrete
phase modeling. However, direct use of the lumped heat capacity
model without a proper modification in evaporation rate formulation
can be erroneous, especially for vacuum spray flash in which the evapo-
ration intensity is very strong [17,40,50]. Nevertheless, introducing the
effective heat conduction model into droplet trajectory models in a La-
grangian-Eulerian CFD simulation requires a sub-grid description of
temperature distributions within each droplet, which would dramati-
cally increase the computational capacity requirement beyond practi-
cal reality. Therefore, based on the comparative study between the
lumped heat capacity method and the effective heat conduction model
for droplet flash, a dimensionless corrective expression is proposed in
this study to modify the evaporation rate equation so that a modified
lumped heat capacity model can be used into the Lagrangian-Eulerian
modeling and its associated CFD simulation.
In conclusion, to study the thermal non-equilibrium phenomena ex-
isted in a spray flash with active vapor extraction, an integrated exper-
iment-modeling-CFD methodology is aiming to be proposed in this pa-
per, and several sub-tasks need to be achieved following the research
logic: (1) determination of polydispersion characteristics of atomized
nozzle spray; (2) establishment of an experimental system for measure-
ments of some spray flash characteristics, which is also used for model
validations; (3) development of a Lagrangian-Eulerian process model
to investigate multiple parametric effects on the vacuum spray flash;
(4) determination of the evaporation coefficient used in the evapora-
tion model, which represents an empirical correction for the flash non-
equilibrium and various non-uniformities in an atomized spray (such as
droplet shape and size); (5) development of a corrective method for the
lumped heat capacity method used in discrete phase modeling of CFD,
realized by comparing with a derived non-dimensional effective heat
conduction model within the droplet during the evaporation; (6) per-
form a full-scale CFD simulation to study the phase coupling of non-
uniformity and thermal non-equilibrium in vacuum spray flash and va-
por transport inside the evaporation chamber. Specifically, the polydis-
persed size distribution from spray atomization is experimentally deter-
mined via a laser-scanning aided technology. The polydispersed size
Fig. 1. Schematic diagram of vacuu m spray flash system.
3
Desalination xxx (xxxx) 115023
distribution is then implemented into the process model and CFD model
as input conditions of spray. A lab-scale experimental spray flash evap-
oration system has been set up for the determination of the evaporation
coefficient as well as model validation, in which the vacuumed evapo-
rator is connected to a vacuum pump to realize a timely vapor extrac-
tion. The extracted vapor is condensed via external condensers and an-
alyzed. With the evaporation coefficient empirically determined under
spray flash in a vacuum, the evaporation model is then implemented
into the process model and CFD model. The parametric process model is
based on a simplification that ignores the vapor-droplet interactions,
whereas the full-field three-dimensional CFD model (via user-defined
function and size/time-dependent correction) is developed to reflect the
reality in vapor-droplet interactions as well as the non-equilibrium and
non-uniformity in the phase transfer and transport.
2. Experimental study
The main objectives of experiments include the measurements of to-
tal evaporation rate, the temperature of feed-in atomized spray, the
temperature of extracted vapor, the temperature of liquid discharged
from evaporation chamber, the temperature profile of spray flash
along with the chamber height, the pressure of vacuum chamber, the
flow rate of feed-in spray, as well as the atomized spray characteristics
such as injection velocity and droplet size distribution.
Fig. 1 shows a schematic diagram of the vacuum spray flash sys-
tem. The main components of the system consist of a feed circulation
pump (Warren Rupp double diaphragm pumps, manufa cture (Mfr.)
number: WR10PP6XPP9), a heater of feed heating (Vulcan immersion
heater, Mf r. number: AUW250B), a nozzle (Bete Fog Nozzle, Mf r.
number:1/4WL1-1/2-60) for generating atomized spray, a vacuum-
maintained evaporation chamber, a feed mixer of replenished supply
and spray residue, a vapor condenser, a yield chamber, and a vacuum
pump. As shown in Fig. 1, the system has two main flow paths: the
feed-spray path (marked as blue) and the vapor-condensate path
(marked as dotted green). The feed is pressurized and heated-up into a
flowrate-temperature regulated nozzle that is inserted into the evapo-
G. Guo et al.
ration chamber. An atomized spray is injected from the nozzle into the
evaporation chamber, where spray flash is formed due to the signifi-
cant pressure differences between the super-saturated droplets and the
depressurized environment. The evaporation chamber is thermally in-
sulated, with ports of vapor extraction and spray residue discharge.
The discharged spray residue from the evaporation chamber is mixed
with replenished feed in the mixer (with a discharging port for the op-
tional discharge of some or all feed residue) and then recirculated,
through the pump and heater, back to the nozzle. The vapor, gener-
ated from spray flash, is extracted into a condenser where the vapor
condensate is collected. The evaporation chamber, vapor-condenser,
and a vacuum-regulator are connected in series to the vacuum pump.
Coolant flow is used to ensure sufficient heat removal and hence va-
por condensation in the condenser. An insignificant amount of non-
condensable gases (NCG), degassed or co-vaporized from spray flash,
cannot be condensed by the condenser and hence are extracted into
the vacuum pump and discharged into the ambient.
In this study, the temperature measurements are based on thermo-
couples and thermocouple profile probe. The temperature profile probe
(Omega ten sensing type k thermocouple profile probes) measures the
temperature distribution (along the probe direction) during the spray
flash process [4,13]. The pressure measurements are obtained using
vacuum pressure gages (IFM pressure sensor, model number: PG2794).
The flow rates of inlet spray and feed replenishment are regulated by
flow meters (IFM efector magnetic inductive flow meter, model num-
ber: sm6004). The feed heating is maintained by an electric-powered
heater. To reduce the surfa ce scaling and limit the amount of NCG, reg-
ular tap wa ter is used as the feed liquid, which is also helpful for the
model validation since the feed is now a single-component condensable
liquid without the complication in concentration-related concerns.
For a spray flash, the atomization characteristics from a nozzle
spray majorly depend on two curial parameters, Weber number and
Reynolds number, respectively representing the surfa ce tension and the
viscous effects with respect to the inertia effect [51–56]. Specifically, it
will strongly depend on the type of nozzle, the feed liquid, and various
operating conditions, including the nozzle pressure, ambient pressure,
flowrate, and feed temperature in our case [11]. Hence, the atomiza-
tion characteristics such as polydispersion are highly case-sensitive. In
this study, we have developed a spray measurement system, which is
based on combined concepts of laser marked shadowgraphy (LMS )
Fig. 2. Schematic experimental system for characterizing atomized spray.
4
Desalination xxx (xxxx) 115023
[57], paralleled laser sweeping (PLS) [58], and particle image ve-
locimetry (PIV) [38,59].
Fig. 2 shows a schematic diagram of the experimental system for
characterizing atomized spray. An atomized spray is illuminated by a
pseudo-laser-sheet, which is further enhanced by a reflector normal to
the laser beam. This pseudo laser sheet is formed by a vertically sweep-
ing laser beam from a laser scanner with a high sweep frequency [58].
The laser sheet is arranged crossing the medium vertical plane of the
spray. The images of droplets within the illumined volume of a laser
beam are recorded by a digital camera, as shown in more detail in Fig.
3. The feed spray is preheated by a heater and regulated by a pressure
regulator. The pressure drop across the nozzle spray is measured by a
differential pressure gauge, whereas the feed spray temperature and
flowrate are measured, respectively, by a thermocouple and a rotame-
ter.
Similar to the droplet-marking principle in LMS [57,60], a targeted
droplet can be marked with glaring points by a laser beam with reflec-
tion. The image is pre-calibrated by a scale located on the laser sheet,
and once the marked droplet is obtained, the size will be determined
based on pixel analysis. The sizes of these laser-marked droplets can
thus be determined through the recorded high-resolution images. It
should be noted that a string of multiple exposures of the same droplet
in a recorded image can also be used to determine the velocity of the
droplet. The difference between the size photographing and the veloc-
ity photographing of droplets is how to coordinate the expose time of
the camera and the sweep frequency of the laser beam. For size deter-
mination, the exposure time of the camera must be short enough to
capture a sharp image, meanwhile it still needs to be long enough for
the laser beam to finish one sweep of the view field. For velocity deter-
mination, the exposure time becomes non-critical, however the sweep
frequency should be fa ster enough to sweep over multiple times of the
tracked droplet during the exposure.
In our case, the laser beam is generated by Laser Physics Reliant
1000 m argon laser, and the laser sweep is realized by Optical Flow
System laser sweeper (Model: SCN-CN2A). The photography area
(marked as yellow zone in Fig. 3) by the camera (Canon EOS 6D Ma rk
II) is located at 139–259 mm along the vertical direction (defining noz-
zle exit is 0), and 0–179 mm along the horizontal direction (defining
spray central ax is is 0). Fig. 4 illustrates an example of captured images
of laser-marked droplets for helping further explanation. Here the red
G. Guo et al.
Fig. 3. Laser-sweeping-aided particle image system.
Fig. 4. Exam ples of images from spray me asuring system.
(a) A string image of a droplet trajectory.(b ) Images of mu ltiple droplets.
spots are the mark point on the droplet by laser (Helium-Neon in this
case). Fig. 4(a) shows a string of five images of a droplet, with the cam-
era exposure time is five times longer than the sweep period of the laser.
This original image is captured in the dark environment, but the back-
ground color is inverted for a better presentation. During this exposure
time, any droplets moving vertically along the sweeping plane will be
highlighted five times by the laser beam, and generate a string of five
highlight points in the digital picture recorded. Thus, based on the cam-
era shutter speed, laser scanning period, and inter-point distance (or
numbers of pixels covered), the instantaneous velocity of the droplet
can be obtained. The average droplet velocity of the spray is deter-
mined by averaging multiple sets of data from different trajectories.
Fig. 4(b) shows a case with the camera exposure time is about the same
as one period of the laser sweep. In that case, any droplets on the laser
sheet are highlighted only one-time by the sweeping laser beam. Such
5
Desalination xxx (xxxx) 115023
an image is used for droplet size determination. In Fig. 4(b), two red
spots are identified by circled marks. It is noted that many unfocused
droplets represent the ones outside either the laser sheet or the focal
zone of the camera. Assuming the spray is steady-state, we can take
many images for the determination of droplet size distribution.
3. Vacuum spray flash model
The modeling of spray flash aims to quantify various parametric ef-
fects and process characteristics on the spray evaporation, which is cru-
cial to optimize the design and operation of the spray flash system. The
key parameters include the atomized spray characteristics (such as
droplet size/velocity distribution and mass flowrate), geometric system
arrangements (such as chamber shape, size and orientation, spray noz-
zle insertion, ports of vapor extraction, and liquid discharge), as well as
G. Guo et al. Desalination xxx (xxxx) 115023

operating conditions (such as spray temperature and vacuuming pres- Ba sset force are negligibly small compared to the drag force. Hence, the
sure). equation of motion of ith droplet is expressed by:
Spray flash is a droplet-vapor two-phase flow transport, with in-
tense non-equilibrium phase changes (evaporation). In spray flash, the (1)
specific sensible heat change is typically at least two order-of-
magnitude smaller than the specific latent heat [17]. For atomized
where the subscript “i” denotes the ith droplet, m d is the mass of droplet;
spray of high inertia (such as of large droplet size and/or high injection
ud and uv denote, respectively, the droplet and vapor velocities; g is the
velocity), the droplet transport is dominated by the droplet inertia from
gravitational acceleration; CD is the drag coefficient; Ad is the cross-
initial nozzle injection, with minor influences of droplet-vapor interac-
sectional aera normal to the droplet trajectory; and ρv is the vapor den-
tions (especially in a vacuuming environment) and droplet-droplet in-
sity.
teractions (due to fa st dispersion and a tiny volumetric fraction of
At high droplet Reynolds number (e.g., 30 < Re dm < 200), with the
droplet phase in the droplet-vapor flow), as schematically described in
correction of evaporation [17] and the Hadamard-Rybcyznski correc-
Fig. 5. In the vacuum spray flash, the vapor phase is solely formed from
tion of a fluid particle [64], the drag coefficient Cd can be estimated by:
spray-dispersed droplet evaporation. Hence, the vapor flow is governed
by the spatial distribution of vapor generation sources (from both
droplet transport and accompanied evaporation) as well as by the vac-
uuming extraction. (1a)
In the two-phase flow modeling of vacuum spray flash, it is conve-
nient to adopt a hybrid Lagrangian-Eulerian approach, in which the
droplet transport is described by Lagrangian trajectory modeling
(starting from the nozzle tip). Meanwhile, the vapor flow is described in where h fg is the latent heat; Re d,m is the droplet Reynolds number, modi-
an Eulerian framework (bounded by the evaporation chamber). The fied with the vapor film viscosity μ v,m, and is defined by:
Lagrangian model yields the spatial dispersion of sprayed droplets and
results in the spatial distribution of vapor generation rate, which are (1b)
implemented into the Eulerian equations of vapor flow. For simplicity,
in the Lagrangian model, the interactions of droplet-droplet and
The droplet diameter d d is related to the droplet mass m d by:
droplet-wa ll are ignored, and the droplet trajectory is assumed to be
predetermined based on droplet inertia.
(2)
3.1. Lagrangian mo del of individual droplets
The change rate of droplet mass is determined from the evaporation
While the disintegrated elements in an atomized liquid spray can be rate of the droplet:
highly non-spherical at the initial nozzle injection [61], the surfa ce ten-
sion will quickly contract the elements into droplets [62]. For simplic- (3)
ity, we ignore this initial-stage of non-sphericity and assume all
droplets are spherical without breakup or coalescence during their In vacuum spray flash, the evaporation rate is governed by the pres-
transport. sure difference between the saturated vapor pressure at the droplet sur-
In the Lagrangian model, the dynamic motion of an individual fa ce and the ambient pressure controlled by vacuuming. Hence, assum-
droplet is governed by the overall momentum transfer to the droplet. ing no temperature slip over the droplet interfa ce (i.e., surfa ce vapor
Conceptually, these momentum transfer terms include the drag force, temperature is the same as the surfa ce droplet temperature), the evapo-
gravity, buoyancy, and forces due to acceleration such as added mass ration rate could be determined from the modified Hertz-Kundsen rela-
force and Ba sset force [63]. However, due to the large density ratio of tion [21]:
the droplet to vapor (about 103), the buoyancy, added mass force, and

(4)

where α is the evaporation coefficient (which is a case-sensitive empiri-

cal constant to be determined from the experiment), Mv is the molecular
weight of the vapor, Ru is the universal gas constant of ideal gases, Psat
is the saturated vapor pressure at droplet surfa ce, P∞ is the ambient
pressure, and Tds is the droplet surfa ce temperature. With the further ap-
proximation of ideal gas for the saturated vapor and the gas phase only
consists of evaporated vapor, the saturated vapor pressure can be re-
lated to the droplet surfa ce temperature by Clausius-Clapeyron equa-
tion [65]:

(5)

where T∞ is the saturated temperature at P∞. Alternatively, in order to

account for the non-ideal gas behavior of saturated vapor, Eq. (5) can
be replaced with an empirical correlation of the corresponding pressure
and temperature of saturated vapor for better or more realistic model-
ing.

Fig. 5. Schematic droplet-vapor transport in a vacuu m spray flash.

6
G. Guo et al.
The droplet surfa ce temperature is determined from the overall heat
transfer to the droplet surfa ce for the latent heat required by the flash-
induced evaporation. In the absence of external heating, especially un-
der convective cooling due to lower ambient temperature, the latent
heat for the droplet evaporation has to be derived from the thermal en-
ergy of the droplet itself. Assuming the heat conduction within the
droplet is spherically-symmetric, the heat conduction equation within
the droplet can be expressed in the spherical coordinate:
(6)
where r is the radial coordinate centered in the droplet, Tl is the time-
dependent local liquid temperature within the droplet, c l is the specific
heat of the liquid, and k eff is the effective thermal conductivity in flash
evaporation [50]:
(6a)
where k is the molecular conductivity, and Pe is Peclet number of the
droplet. The initial condition is given by spray temperature at nozzle in-
jection:
(6b)
The boundary conditions are given by the symmetric condition at
the droplet center and heat transfer balance at the droplet surfa ce, re-
spectively expressed as:
(6c)
(6d)
where h m is the convective heat transfer coefficient, which may be ap-
proximated using Ranz and Ma rshall correlation with vapor-film prop-
erties and relative droplet velocity [65]. In which the Nusselt number is
determined by:
(6e)
Thus, the Lagrangian model of a droplet flash and transport is es-
tablished. With six governing equations (Eqs. (1)–(6)) for six unknowns
(m d, ud, d d, Tl, Psat and ), the problem is closed.
The hydrodynamic transport and evaporation of different-sized
droplets, predicted by the individual droplet evaporation model, can
behave quite differently. In a vacuum chamber, the smaller droplets are
likely to achieve higher evaporation completeness due to a larger spe-
cific surfa ce and longer time duration, and the larger droplets will have
less evaporation completeness. As a result, using one average droplet
size (such as Sauter mean diameter) to simulate polydispersed spray
evaporation may result in a severely biased estimation of the overall
evaporation rate. Therefore, in the Lagrangian modeling of a polydis-
persed spray, droplets are sub-grouped into N size-based groups (i.e.,
droplets in the same group will have similar sizes) according to the
droplet size distribution, which helps to account for this polydispersion
effect on the spray transport and evaporation.
3.2. Eulerian mo del of vapor flow
One of the major objectives of this study is to estimate the extracted
vapor temperature and other vapor transport properties. The vapor
flow is modeled using the Eulerian approach, with the coupling of
evaporating sprays from the Lagrangian trajectory model. It is as-
7
Desalination xxx (xxxx) 115023
sumed that the droplets are treated as point sources in the vapor flow so
that the volumetric effect of droplets is ignored in the vapor flow.
The mass conservation equation is given by:
(7)
where the S m is the vapor generation rate per volume, which is the sum
of individual vapor generation rates of all droplets within the control
volume Vcell. For a polydispersed spray, S m can be expressed by:
(7a)
where subscripts i and j denote the ith droplet of jth size group (or on jth
trajectory), N j is the total number of droplets of jth size-group within
Vcell, N is the total number of droplet size groups, and is the evapora-
tion rate of an individual droplet which is previously solved from the
Lagrangian model. N j depends on the local number density of droplets
of group j, n j, by:
(7b)
Assuming there is no droplet breakup or coalescence, the mass con-
servation leads to a droplet number density conservation, which is ex-
pressed by:
(7c)
The momentum conservation equation is given by [66]:
(8)
where the Fvd is the droplet-vapor momentum transfer per volume,
which is roughly approximated by the sum of individual drag forces of
all droplets within the control volume Vcell. For a polydispersed spray,
Fvd can be expressed by:
(8a)
where Fij is given by:
(8b)
The energy equation is approximated by [66]:
(9)
The equation of state is given by:
(10)
The turbulent viscosity μ T can be determined by a turbulence model,
such as the k-ε model. The turbulent thermal conductivity KT can be re-
lated to μ T by gas kinetic theory.
Thus, for the Eulerian model, there are four independent Eqs. (7)–
(10) matching four independent variables (ρ, uv, p and Tv), which can be
solved by coupling with the Lagrangian model of droplets.
3.3. ANSYS FLUENT-based numerical simu lation
The Lagrangian-Eulerian model above of spray flash can be numeri-
cally solved using the Discrete Phase Model of the commercial CFD
solver ANSYS FLUENT. As each iteration of the computation, ANSYS
G. Guo et al. Desalination xxx (xxxx) 115023

FLUENT firstly calculates the vapor phase of the Eulerian model, then
(11b)
solves the discrete droplet phase of the Lagrangian model. The results
will be used to feedback and update the source terms to the flow do-
main for the next iteration. Since ANSYS FLUENT does not have the (11c)
specific discrete phase model for self-heat supplied droplets flash evap-
oration [67], the User-defined functions governed by the flash model is
coupled with the simulation.
(11d)
3.3.1. Modified lumped heat capacity method
As we know, due to the point-treatment of droplets in the discrete
phase model of CFD, the droplets can only be expressed as a lumped where β*, L* and Bi stand for, respectively, dimensionless parameters of
heat capacity model [67], in which the heat conduction within droplets flash evaporation and latent heat as well as the effective Biot number
is approximated as infinitely fa st and results in an overestimation of the of heat convection, defined by:
evaporation rate [31,50]. Hence, it would be interesting to check the
possibility of modifying the lumped heat capacity model by introduc-
ing some correction fa ctors that will offset the difference in evapora-
tion rate between the lamped heat capacity model and the effective
heat conduction model. (11e)
To ensure generality, the effective heat conduction model is ex-
pressed in the dimensionless form.
According to Eqs. (4) to (6), we have:
The reduced diameter, d ⁎, is time-dependent with evaporation,
(11) which is given by the mass conservation equation, Eq. (3). Substituting
Eqs. (2) and (4) into Eq. (3), then transform to the dimensionless for-
where dimensionless superheat ratio, radius ratio, Fourier number, and mat, it yields:
reduced diameter are defined by:
(12)

where Ja is the Jakob number, defined by:

(11a)
(12a)

Ba sed on the above dimensionless heat conduction model, a typical

According to Eqs. (6)–(6d), the dimensionless initial and boundary
conditions become:
result is illustrated in Fig. 6 under conditions of β* = 78.8 and
Bi = 0.0069, representing an operating condition of flash evaporation
of 1-mm droplet at the initial temperature of 100 °C under the evapora-
tion pressure of 0.6 atm in our case. It reveals the relationships among
the superheat ratio (θ*), dimensionless radial position (R*), and the di-

Fig. 6. Typical dimensionless distributions of superheat within a droplet vs time.

8
G. Guo et al.
mensionless time (Fo ), clearly showing a transient process from an ini-
tial state of strong thermal non-equilibrium and non-uniformity within
the droplet towa rd an equilibrium state with uniformity.
For cases where both β* and Bi number are small, the temperature
distributions within droplets are kept nearly uniform during the entire
process of flash evaporation, as illustrated in Fig. 7, where β* = 1 and
Bi = 0.0069. This represents a limit case where the lumped heat capac-
ity model becomes valid.
The dimensionless form of lumped heat capacity model can be ap-
proximated, from Eq. (11) under conditions of small β* and Bi number
(so that the surfa ce temperature is approximated by the volume-
averaged droplet temperature), as:
Fig. 7. Exem plified superheat within a droplet vs time at sm all β* and Bi
nu mbe r.
Fig. 8. Evaporation residue vs Fo nu mber of 1 mm droplet (β* = 78 .8 and Bi = 0. 00 69 ).
9
Desalination xxx (xxxx) 115023
(13)
where is the volume-averaged relative temperature of the droplet,
and the initial condition is given by Eq. (11). The difference in droplet
evaporation predicted by the lumped heat capacity model (Eqs. (13)
and (12)) and the effective conduction model (Eqs. (11) and (12)) under
the same operating conditions (i.e., the same values of β* and Bi) can be
calculated. It is noted that, based on Eq. (2), the evaporation residue
can be defined in terms of reduced diameter as:
(14)
where m d0 and m d∞ represent the initial and the equilibrium droplet
mass, respectively. Such a comparison is exemplified by Fig. 8 for a
1 mm droplet evaporated under β* = 78.8 and Bi = 0.0069. The
evaporation residue by both models decrease in a nearly exponential
trend along with Fourier number (or linearly in the semi-log scales),
which further suggests:
(14a)
where λ is the slope in the linear curve fitting in the semi-log scales.
In Fig. 8, the lumped heat capacity model shows a higher evapora-
tion rate than the effective heat conduction model, which is due to the
problematic assumption of uniform temperature distribution in the
lumped heat capacity model that over-estimates the surfa ce tempera-
ture during the droplet evaporation. Ba sed on the near-exponential de-
crease in droplet mass shown in the result, it is plausible to propose a
temperature polarization correction (αL) on the droplet evaporation
rate predicted by the lumped heat capacity model, which would then
represent the droplet evaporation rate predicted by the effective heat
conduction model.
G. Guo et al. Desalination xxx (xxxx) 115023

(15)

where the subscripts E and L stand for the effective conduction model
and lumped heat capacity model, respectively. In other words, with the
temperature polarization correction, the lumped heat model can be
used to replace the effective heat conduction model for the droplet flash
evaporation. And according to Eqs. (14) and (15), the value of coeffi-
cient αL for a spray with determined physical properties evaporated un-
der the same operating condition is only determined by the droplet size
and time step.
The lumped heat capacity model with the evaporation rate correc-
tion is then given by:

(16)

where

(17)

Finally, by implementing Eq. (17) into the User-Defined Functions

(UDF) in ANSYS FLUENT, we can use the Lagrangian-Eulerian model Fig. 9. Geometric structure in CFD simulation.
to simulate the spray flash and dispersion in a vacuum chamber.
Table 1
3.3.2. Case study of ANSYS FLUENT Setup details of CFD simulation.
In this study, the modeling equations are solved via ANSYS FLUENT So lv er descri ption
and its associated UDFs. The geometric structure is shown in Fig. 9.
Steady/unstead y Tran si ent/unsteady pa rticle tracki ng
Once the droplets reach the chamber wa ll or if the droplet temperature
Dimension 3-D
is sufficiently close to the equilibrium temperature corresponding to the Turb ulence Re-Norm al ized Group (RNG) k-ε mo del
vacuum pressure, the droplets will be assumed as immediately escaped mo del
from the chamber. The liquid outlet in CFD simulation is treated as a Discrete phas e Droplets with UDF evap or ating la ws an d coupled heat-
part of “wall” for the continuous phase (vapor) since the liquid outlet mo del ma ss so lution; intera cted with continuous phas e
Injection so li d cone; sp ra y an gel 57°; sp ra y velo city 3. 23 m/s;
pipe in the experimental system does not allow any vapor escape,
Ro si n-Ra mm ler si ze distri bution: κ = 4. 432 an d δ = 1. 099
which is alwa ys filled up with saturated wa ter. Since all droplets are
treated as points in the CFD simulation, there are no volumetric effects
Bounda ry conditions
of droplets on vapor flow nor volumetric accumulation of wa ter
residue. All chamber wa lls are thermally insulated. Spra y inlet Ma ss flow inlet (zero-in); DPM inlet
The CFD case is simulated with the operational pressure of 40 kPa Steam outlet Pressure outlet; DPM escape
and the inlet spray temperature of 367.15 K (94 °C). The simulation is Water outlet Ma ss flow outlet (zero-out); DPM escape
wa ll No heat flux; DPM escape
run through a transient approach, which will be terminated when all
critical variables (such as domain temperatures, velocities, and flow
rate at the vapor outlet) become time-insensitive (after about 300 s in phase, and overall evaporation rate of droplets between 0.58 and 1.33
the case). Ba sed on the size distribution described by Eq. (17) or Fig. 9, million. To reduce the computational cost, the 0.58 million-case is
the atomized spray injection consists of seven groups of different sizes adopted for the simulation.
of droplets in each parcel or spray trajectories. In this case, Re-
Normalized Group (RNG) k-ε model [67] is adopted as the turbulence 4. Results and discussion
model, which is widely reported as one of the best applicable turbulence
models for in-chamber flow simulation with a spray [47,48,68]. The In the following, a case study accompanied with both theoretical
other detailed information of the setup is summarized in Ta ble 1. calculations and experiments is used to illustrate some unique charac-
The mesh sensitivity analysis wa s conducted on three different mesh teristics and effects in a vacuum-assisted spray flash, including the
densities to ensure the results are independent of the grid, corresponding spray polydispersion and its impact on flash evaporation rate, the oper-
to the total cell number are 0.28 million, 0.58 million, and 1.33 million, ational parametric effects (initial temperature and operating pressure)
respectively. A ten-second operating wa s simulated for these mesh on the spray flash, and the non-uniform phase distributions of spray
models and it wa s found the results will not change significantly be- flash inside the evaporation chamber. The basic operational and geo-
tween 0.58 million and 1.33 million with the value differences range metric information is summarized in Ta ble 2, with a basic system lay-
from 0.14% to 1.2% for curial parameters such as average outlet tem- out schematically described in Fig. 5. The heat losses via the chamber to
perature of two phases, the average temperature of full-field vapor the ambient are insignificant (less than 1%), judged by the experimen-

10
G. Guo et al. Desalination xxx (xxxx) 115023

Table 2
Key geome tric and operational parame ters.
Para metric Va lue Unit

Spra y nozzle Full -cone –

Spra y an gl e 57 °
Nozzle lo cation Top center –
Spra y injection velo city 3. 23 m/s
Spra y li quid Water –
Spra y flow ra te 44.1 ml /s
Cham ber heig ht 0. 5 m
Cham ber diam eter 0. 3 m
Va por outlet diam eter 0. 75 inch
Li quid outlet diam eter 0. 5 inch
Spra y initia l temp erature 70–100 °C
Cham ber va cuum pr essure 0. 25–0. 8 atm
Cham ber ma teri al Stai nless steel with insula tion N/A
Therma l resi stan ce of cham ber 2. 16 K· m 2/W

tal estimation of enthalpy imbalance between the inlet feed and outlets
of vapor and liquid.
Fig. 10. Exam ple of the PDF of droplet sizes with the initial spray tempera-
ture of 94 °C.
4.1. Polydispersion effect

3.3, the temperature polarization effect increases with the increase in

The most important polydispersion characteristics of a liquid spray
the droplet diameter. That means the larger the droplet, the more cor-
is the co-existence of various initial droplet sizes, typically by a simulta-
rection is needed for the evaporation rate in the lumped heat capacity
neous liquid breakup from the combined actions of jet-induced shear-
model (which we have to use in the FLUNT-based simulation). Hence,
tearing and turbulent instability [35,54]. Due to the lack of reliable
based on Eq. (15) and the evaporation rates predicted respectively from
modeling and complexity in various nozzle design [61], operation con-
the lumped heat capacity model and the effective heat conduction
ditions, and liquid properties, the droplet size distribution is mostly de-
model, the temperature polarization fa ctor can be correlated as a func-
termined via experimental means, such as using the laser-sweeping
tion of droplet diameter and time. This case-dependent correlation is
aided photographic system in Fig. 3. Because the spray needs to be de-
implemented into Eq. (16), which is then used in the UDF of ANSYS
termined as the initial condition, measurements were performed in a
FLUENT for the evaporation rate formula in the lumped heat capacity
normal room-environment in order to minimize the effect of intense
model of the discrete phase.
evaporation on droplets sizes [69,70]. As mentioned in Ta ble 2, the
In our cases, the two parameters, λE, and λL, in corrective expression
only major variable in our case that influences polydispersion charac-
αL from Eq. (15) are determined and listed in Ta ble 3, corresponding to
teristics between each different operating condition is the different
different droplet sizes. The linear fitting treatment (as shown in Fig. 8)
spray initial temperature, which might noticeably af fect temperature-
fits reasonably well under every droplet size group, with values of the
dependent physical properties of the surfa ce tension and the viscosity
coefficients of determination (R2) range from 0.996 to 1. The error mar-
[35]. Therefore, the droplet size distribution has been calibrated for
gin of the corrected lumped heat capacity model predictions on the
specific operations distinguished by the inlet spray temperature. Ba sed
evaporation rate is within 5% by comparing with the effective heat
on the experimental results and previous work, the experimental data
conduction model.
are most likely to be best-fitted (by using the least-square method) into
Once λE and λL are determined, the coefficient αL only changes with
a Rosin-Rammler size distribution [35,71,72], which is expressed in
time (Fo number). Exemplary tendencies to show such changes with
terms of probability density function(PDF) by
time are illustrated in Fig. 11. It shows with the time increasing (larger
Fo number), the slope of the curve becomes flatter or even positive.
(18) That means with the process going on, the evaporation rate in the
lumped heat capacity model is approaching to then will even become
where κ and δ are two fitting parameters, respectively for shape and smaller than it in the effective heat conduction model. It makes sense
scale. since the effective heat conduction model will produce a relatively
In this work for all cases, a full cone spray nozzle without swirling larger evaporating temperature in the latter part of the evaporation
(Bete 1/4WL1-1/2) is used, with a wa ter flow rate of 44 ml/s under a process.
specific temperature. Fig. 10 illustrates an example of droplet size dis-
tribution corresponding to the operation of 94 °C initial spray tempera- 4.2. Operation parametric effects
ture (used in model validations and CFD), which is obtained from an
analysis of 933 measured droplets (similar to those in Fig. 4). The One of the most important objectives of this study is to investigate
droplet size range is shown to be between 0.4 and 2.0 mm, with the thermal non-equilibrium or temperature difference between the ex-
κ = 4.432 and δ = 1.099. tracted vapor and the liquid residue af ter a vacuum spray flash process.
Another polydispersion effect is on the correction of evaporation Such thermal non-equilibrium is most likely to be af fected by the key
rate in the lumped heat capacity model. As discussed above in Section

Table 3
Corrective expressions corresponding to different droplet sizes.
Droplet si zes 0. 4 mm 0. 6 mm 0. 8 mm 1. 0 mm 1. 2 mm 1. 4 mm 1. 6 mm

λE −3.4 69 −4.4 78 −5.2 53 −5.7 2 −6.0 76 −6.2 71 −6.4 46

λL −5.2 31 −7.8 14 −10. 43 −13. 4 −15. 64 −18. 2 −20. 8

11
G. Guo et al. Desalination xxx (xxxx) 115023

Fig. 11. Temperature polarization factor vs droplet diameter.

operation conditions, including the initial spray temperature and the

ambient vacuum pressure inside the evaporation chamber.
For the parametric analysis, the Lagrangian trajectory model de-
scribed in Section 3.1 may be approximated by simplifying the droplet-
vapor phase interactions, i.e., assuming the vapor is moving up uni-
formly within the chamber. This simplification is justifiable under the
current case study, in which the cross-section averaged vapor velocity
(e.g., less than 0.04 m/s) is much less than the spray velocity (e.g.,
about 3.3 m/s), and the calculation of our model indicates the convec-
tive heat transfer is at least four orders-of-magnitude smaller than the
latent heat transfer during the droplet evaporation. Such a simplifica-
tion enables the spray model to be directly solvable without considering
the cross-sectional non-uniformity of the vapor phase, which is
adopted as the process model in this paper.
The effects of inlet spray temperature and pressure in the vacuum
chamber on the temperatures of extracted vapor and wa ter residue is
shown in Figs. 12 and 13, with the inlet spray temperature varying
from 70 to 100 °C and the operating pressure varying from 25 to
80 kPa, respectively. Each experimental data is averaged from six sets
of measurements, with the error bar indicating the standard deviation
Fig. 13. Superheat level vs temperature difference of two exit phases.
of the group. Since the theoretical results are calculated case by case,
the dash-line interlinked each point is just for showing a tendency in-
stead of a real continued result curve.
Fig. 12 shows at the same inlet spray temperature, a higher operat-
ing pressure (or lower vacuum level) will result in a narrower tempera-
ture difference between vapor and wa ter, which means the thermal
non-equilibrium level (temperature gap) is positively related to the su-
perheat level of the spray. Under the same vacuuming pressure, the va-
por temperature increases linearly with the inlet spray temperature
while the wa ter residue temperature remains close to the equilibrium
temperature, which also indicates the evaporation process is nearly
completed in all cases. However, we can see the temperature difference
can be effectively generated even under a low superheat level of the
spray, which indicates the process of spray flash with active vapor ex-
traction is highly efficient in evaporation (discharged heat from hotter
vapor to cooler droplets during the process is little). The result presents
quantitative relations between superheat level of the spray and ther-
mal non-equilibrium level of the two phases. Such a relation will be a
very critical parameter for desalination system design if the heat is ex-
pected to be further recovered from hot vapor to cool residue via a
heat exchanger, in which the heat recovery efficiency has to depend on
that temperature difference.
Fig. 13 shows the superheat of the droplets (Tin-Tsat) and the tem-
perature difference (Tv,o-Tw,o) follow a linear relationship based on
Fig. 12. Effects of spray inlet temperature and operating pressure on tem- modeling and experimental results. The slopes of different operations
peratures of extracted vapor and water residu e.

12
G. Guo et al.
are quite close to each other, with an average value of about 0.47. The
tiny variations on the slopes of each operation are appeared due to the
non-linear relation between the saturated pressure and temperature of
the wa ter/vapor. We can also notice the deviation between experi-
ments and modeling predictions increases with rising the superheat de-
grees of the spray, in which the experimental measurements are lower
than computational results. Such errors may be caused by enhanced lo-
cal mixing of heat and mass between the two phases, which cannot be
fully considered by this simplified process models.
It is noted that the stronger superheating of spray means a bigger
pressure difference between the saturated vapor from droplet evapora-
tion and the vacuuming environment. Due to such significant non-
equilibrium in pressure, the evaporative coefficient α in Eq. (4) needs to
be calibrated since the evaporate rate equation is derivated from the as-
sumption of pseudo-equilibrium. Ba sed on the experimental results of
the case of 60 kPa, the α is calibrated as 0.01. Assuming the evapora-
tion coefficient remains unchanged within the vacuuming range in this
study, the modeling predictions on the droplet temperature distribu-
tions along with the chamber centerline (where a multi-point thermo-
couple probe is located) can be compared to the measured temperature
profile, as shown in Fig. 14. The good agreement between the modeling
predictions and measurements also partially verifies the assumption of
an unchanged evaporation coefficient. We can also notice, the more in-
tense the evaporation is (the higher superheat level the spray is), the
larger deviation is obtained during the experimental measurements. We
believe such error results from the larger fluctuation of temperature
measurements and the larger deviation to the ideal hypothesis of queso-
thermal-equilibrium in theoretical modeling, due to the increased uni-
formity of local energy and momentum transports as well as the ther-
mal non-equilibrium between two phases. The averaged temperature
profile along the chamber centerline of the spray from the specific CFD
case is also plotted and compared in the figure. In all cases, the droplets
approach the thermal equilibrium about half-wa y (0.25 m) before their
exit from the chamber, indicating a sufficient length in the chamber de-
sign for the spray flash. However, to focus on the evaporating moment
more clearly, the plot range is cut-off at 0.3 m.
4.3. Non-uniform spray dispersion and vapor flow
The interactive effect of vapor on the spray flash wa s investigated
by the Lagrangian-Eulerian modeling numerical simulation via ANSYS
FLUENT, based on the setup described in Section 3.3.2. As for valida-
tions, the computed evaporation rate of the droplet in CFD is compared
with the results of both experiments and process model in Fig. 14, and
Fig. 14. Spray temperature profile along the ch am ber centerline.
13
Desalination xxx (xxxx) 115023
meets a good agreement. Since both the droplet size and travel distance
will af fect the instantaneous evaporation rate, the droplet sampled in
CFD follows the same Sauter mean diameter and drops along the verti-
cal central path as calibrated in experiments and defined in the process
model. The simulated overall evaporation rate, and the averaged outlet
temperature of vapor and wa ter, are also compared with experimental
measurements, which are summarized in Ta ble 4. The result shows the
numerical estimation of the overall thermal performance meets a good
agreement with experiments.
Fig. 15 illustrates the spray pattern of the median droplet diame-
ter (1.03 mm) with temperature changes along the droplet trajecto-
ries. In this case, all droplets are sufficiently cooled down to the
equilibrium temperature of 350 K or 77 °C before reaching the
chamber wa ll.
Fig. 16(a) shows the evaporation completeness of the sampled me-
dian diameter droplet along their trajectories, whereas Fig. 16(b)
shows the vapor generation rate distribution in the median plane of
the chamber. Both figures indicate that the flash evaporation is very
fa st, with most evaporation completed within the top half-space of
the chamber, indicating a great potential to reduce the height require-
ment in chamber design.
Fig. 17 shows the vapor temperature distribution in the median
plane. Apart from the strong non-uniform distribution along the ax ial
direction, some very hot vapor at a temperature close to that of inlet
spray is accumulated around the top of the chamber aw ay from the
vapor outlet while the vapor near the outlet has a much-lowered tem-
Table 4
Experime ntal validation of overall ch aracteristics.
CF D Experi ment
Evap or ation ra te 4. 30 kg /h 4. 19 kg /h
Ou tlet temp erature of the va por 357. 4 K 357. 9 K
Ou tlet temp erature of the wa ter 349. 9 K 350. 2 K
Fig. 15. Example of droplets temperature along their trajectories.
G. Guo et al.
Fig. 16. Evaporation process distributions of spray flash.
(a) Evaporation completeness along trajectories.(b ) Vapor generation rate in me dian plane.
Fig. 17. Vapor temperature distribution in the median plane.
perature. As expected, the vapor becomes cooler along the spray tra-
jectories.
Fig. 18(a) shows the velocity contour of the vapor in the median
plane of the chamber, and Fig. 18(b) shows the velocity path-line re-
leased from the median plane. The result indicates that, due to the
14
Desalination xxx (xxxx) 115023
downwa rd movement of the droplets, the generated vapor also comes
with an initial downwa rd velocity before it moves upwa rds by the
vacuum extraction. Near the vapor outlet, there is much-enhanced
vapor suction due to the small cross-section area of the vapor outlet.
This strong vapor suction region is luckily aw ay from the conical
spray coverage zone marked by droplet trajectories. Otherwise, a suf-
ficiently strong sucking flow could directly entrain some small
droplets or droplets of weak inertia into the vapor extraction tube.
The velocity scale in this figure is limited by 2 m/s to try presenting
the details of the velocity distribution in most weak-flowing areas.
The strong-flowing jetting and sucking area are represented as the red
zone in the contour, and vapor velocities in there could go much
higher than 2 m/s.
5. Conclusions
In this study, process characteristics of active vacuum aided spray
flash evaporation are investigated by an integrated experimental-
modeling-simulation approach. A Lagrangian-Eulerian modeling ap-
proach is developed, which is then numerically solved for a case study
using ANSYS FLUENT modified with UDF. Tw o experimental systems
are developed in this study: one is to measure the spray flash character-
istics such as evaporation rate, temperatures of extracted vapor and
liquid residue, and the droplet/vapor cooling along the vertical ax is of
the chamber; the other one is based on a combined technology of laser
marked shadowgraphy (LMS ), paralleled laser sweeping (PLS), and
particle image velocimetry (PIV) for the determination of droplet size
distribution of atomized spray. In order to improve the problematic ap-
proximation by the lumped heat capacity model (LHCM) used in CFD,
a dimensionless corrective approach for the evaporation expression in
the LHCM is proposed.
The thermal non-equilibrium phenomena in temperature are con-
spicuously obtained via both experiments and theoretical predictions,
which demonstrated the feature of highly efficient evaporation in
spray flash under active vacuuming. Such a non-equilibrium between
phases has the potential to be utilized in designing a desalination sys-
tem for achieving better thermal efficiency and yield capacity. Case
studies of the operation parametric effects on the spray flash character-
G. Guo et al. Desalination xxx (xxxx) 115023

Fig. 18. Velocity ma pping of vapor in me dian plane.

(a) Velocity contour.(b ) Velocity path-line released from me dian plane.

istics are demonstrated via both modeling predictions and experimental n number (–)
measurements. The operational pressure in vacuum chamber varies R universal gas constant (J/mol K)
from 25 kPa to 80 kPa, while the operating range in inlet spray temper- T temperature (K)
ature varies from 70 to 100 °C. The droplet size distribution of wa ter u velocity (m/s)
spray can be well fitted into a Rosin-Rammler distribution, with a size
range between 0.4 and 1.8 mm and a median size of 1.03 mm. The tem- Greeks
perature difference between the extracted vapor and the discharged
spray residue can be significantly impacted by the active vacuum ex- α evaporative coefficient (–)
traction and the superheating of input spray. For example, the temper- θ superheat ratio (–)
ature difference is linearly increased with the superheating level of in- μ dynamic viscosity (kg/m∙s)
put spray, with a slope of 0.47. Both simplified process model and full- β dimensionless parameter (–)
scale CFD simulation yield similar spray flash characteristics (with ex- ρ density (kg/m3)
perimental validation), such as spray cooling profile, evaporation rate Δ difference (–)
distribution, and the overall temperature difference between the ex-
tracted vapor and liquid residue, which partially verify the very weak Subscripts
vapor-droplet interactions on the droplet flash. However, the CFD sim-
ulation shows that there exist local regions of strong non-uniformity in A area
vapor temperature and velocity, which may be of importance to the cell the control volume
practical design of the evaporation system. d droplet, drag
e end
Nomenclature i order number
L lumped heat capacity
A projected area (m2) m modified phase, average, molar
C drag coefficient (–) p isobaric
d diameter (m) sat saturated condition
F force (N) u universal
g gravitational constant (m/s2) vd vapor-droplet
Ja Jakob number (–) ∞ fa r-field ambient
M molecular weight (kg/mol) a added mass
N group number (–) ch chamber
P pressure (Pa) E effective heat conduction
Re Reynolds' number (–) eff effective
t time (s) j size/trajectory number
V volume (m3) l liquid (water)
Bi Biot number (–) fg latent heat
c specific heat (kJ/kg K) s surfa ce
r radius (m) T turbulent
Fo Fourier number (–) v vapor
h heat (–) 0 initial condition
L dimensionless latent heat (–)
m mass (kg)

15
G. Guo et al.
Decl aration of competing interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
References
[1] H. Nass rull ah, S. F. Anis , R. Ha shai keh, N. Hi la l, Energy for desa li na tion: a
state-of-the-ar t revi ew, Desa li na tion. 491 (2020).
[2] F. E. Ahmed, R. Ha shai keh, N. Hi la l, Hy br id technolo gi es: the future of
energy efficient desa li na tion – a revi ew, Desa li na tion. 495 (2020) 114659.
[3] M. A. Shan non, P.W. Bohn, M. Elim elech, J.G. Geor gi ad is , B.J. Ma rĩ as , A. M.
Ma yes, Science an d technolo gy for wa ter puri fication in the comi ng decades,
Nature. 452 (2008) 301–310.
[4] O. Mi ya take, T. Tomi mu ra , Y. Ide, T. Fuji i, An experi mental study of sp ra y
flas h evap or ation, Desa li na tion. 36 (1981) 113–128.
[5] D.H. Troup, J.A. Ri char dson, Scal e nucleation on a heat tran sf er surf ace an d
its pr evention, Chem. Eng. Co mm un. 2 (1978) 167–180.
[6] O. Mi ya take, Y. Ko ito, K. Taga wa , Y. Ma ruta, Tran si ent char acteri stics an d
perf or ma nce of a novel desa li na tion sy stem ba sed on heat stor ag e an d sp ra y
flas hing, Desa li na tion. 137 (2001) 157–166.
[7] J. Ho u, H. Cheng, D. Wang, X. Gao, C. Gao, Experi mental investig ation of
lo w temp erature distil la tion coupled with sp ra y evap or ation, Desa li na tion.
258 (2010) 5–11.
[8] M. M. A. Ra j, K. K. Muruga vel, T. Ra ja seeniv as an , K. Sr ithar, A revi ew on
flas h evap or ation desa li na tion, Desa li n, Water Treat. 57 (2016)
13462–13471.
[9] I. Alatiqi, H. Ettouney, H. El-Dess ouky , K. Al-Ha jr i, Meas urements of
dyna mi c behavi or of a mu ltis tage flas h wa ter desa li na tion sy stem,
Desa li na tion. 160 (2004) 233–251.
[10] P. Guo, T. Li , Y. Wang, J. Li , Energy an d exergy an al ys is of a sp ra y-
evap or ation mu lti-effect distil la tion desa li na tion sy stem, Desa li na tion. 500
(2021) 114890.
[11] A. Lefebvre, V. Mcdonell , Atom ization an d Spra ys , Second Edition., 2017.
https: //doi. or g/10.1016/0009-2509(90)87140-N.
[12] A. H. Ar ag hi, M. Khia da ni, M. H. Sa da fi, K. Ho om an , A numeri cal mo del an d
experi mental veri fication for an al ys ing a new va cuum sp ra y flas h
desa li na tor util is ing lo w gr ad e energy , Desa li na tion. 413 (2017) 109–118.
[13] O. Mi ya take, T. Tomi mu ra , Y. Ide, M. Yu da , T. Fuji i, Effect of li quid
temp erature on sp ra y flas h evap or ation, Desa li na tion. 37 (1981) 351–366.
[14] Y. Ikegam i, H. Sa sa ki , T. Gouda, H. Uehara , Experi mental study on a sp ra y
flas h desa li na tion (influence of the direction of injection), Desa li na tion. 194
(2006) 81–89.
[15] S. Mutair , Y. Ikegam i, Experi mental investig ation on the char acteri stics of fl
as h evap or ation fr om superheated wa ter jets for desa li na tion, Desa li na tion.
251 (2010) 103–111.
[16] A. K. El-Fiqi, N.H. Ali, H. T. El-Dess ouky , H. S. Fa th, M. A. El-Hefni, Fl as h
evap or ation in a superheated wa ter li quid jet, Desa li na tion. 206 (2007)
311–321.
[17] W.A. Si ri gnan o, Fl uid Dyna mi cs an d Tandsp or t of Droplets an d Spra y,
Ca mb ri dge Univ ersi ty Press, 2010.
[18] C. Wang, R. Xu, Y. So ng, P. Jian g, Study on wa ter dr oplet flas h evap or ation
in va cuum sp ra y cool ing, Int. J. Heat Ma ss Tran sf . 112 (2017) 279–288.
[19] R. N. Dahm s, J.C. Oefelein, The si gnifican ce of dr op non-sp heri city in sp ra ys ,
Int. J. Multiph. Fl ow. (2016) 67–85.
[20] L. D. Ko ffma n, M. S. Plesset, L. Lees, Theory of evap or ation an d condensa tion,
Phys , Fl uids. 27 (1984) 876–880.
[21] A. H. Pers ad , C. A. Ward , Expr essi ons for the evap or ation an d condensa tion
coefficients in the Hertz-Knudsen rela tion, Chem. Rev. 116 (2016) 7727–7767.
[22] T. Ytrehus, S. Os tm o, Ki netic theory ap pr oa ch to interp hase pr ocesses, Int. J.
Multiph. Fl ow. 22 (1996) 133–155.
[23] H. Y. Erbil, Evap or ation of pure li quid sess il e an d sp heri cal susp ended dr ops:
a revi ew, Adv, Co ll oid Interf ace Sci. 170 (2012) 67–86.
[24] E.L. Cussler, Diffusion: Mass Transfer in Fluid System, Cambridge University Press,
2009.
[25] M. Bond, H. Struchtrup, Mean evap or ation an d condensa tion coefficients
ba sed on energy dependent condensa tion pr obab il ity, Phys . Rev. E - Stat.
Phys ics, Plas ma s, Fl uids, Rela t. Interd is cip. Top. 70 (2004) 21.
[26] C. A. Ward , G. Fa ng, Expr essi on for pr edicting li quid evap or ation flux:
statis tical ra te theory ap pr oa ch, Phys . Rev. E - Stat. Phys ics, Plas ma s,
Fl uids, Rela t. Interd is cip. Top. 59 (1999) 429–440.
[27] S. S. Penner, On the ki netics of evap or ation, J. Phys . Chem. A 56 (1952)
475–479.
[28] T.H. Hw an g, M. K. Mo al lemi , Heat tran sf er of evap or ating dr oplet flow in
lo w pr essure sy stems, Int, Co mm un. Heat Ma ss Tran sf . 15 (1988) 635–644.
[29] W.L. Cheng, H. Chen, L. Hu , W.W. Zhan g, Effect of dr oplet flas h evap or ation
on va cuum flas h evap or ation cool ing: mo deli ng, Int. J. Heat Ma ss Tran sf . 84
(2015) 149–157.
[30] Q. Chen, K. Thu, T.D. Bui, Y. Li , K. C. Ng, K. J. Chua , Develo pm ent of a mo del
for sp ra y evap or ation ba sed on dr oplet an al ys is , Desa li na tion. 399 (2016)
69–77.
[31] B. Ca i, Q. Wang, S. Yi n, H. Gu, H. Wang, H. Zhen, Energy an al ys is of sp ra y
flas h evap or ation fr om superheated upwa rd jets, Appl. Therm. Eng. 148
16
Desalination xxx (xxxx) 115023
(2019) 704–713.
[32] C. Wang, R. Xu, X. Chen, P. Jian g, B. Li u, Study on wa ter flas h evap or ation
under reduced pr essure, Int. J. Heat Ma ss Tran sf . 131 (2019) 31–40.
[33] G. Guo, C. Zhu, Z. Ji, Cf d si mu la tion of is ol ated sp ra y flas h evap or ation with
active va por extraction, in: 5th Therm. Fl uids Eng. Co nf., Begell house,
Co nnecticut, 2020: pp. 269–278.
[34] G. Guo, H. Deng, C. Zhu, Z. Ji, Non-volatile fraction effects in dispersed vacuum
spray flash evaporation, Press Int. Mech. Eng. Congr. Expo. 11 (2020) A010.
[35] S. Ko oij, R. Si js , M. M. Denn, E. Vill erma ux, D. Bonn, What determ ines the
dr op si ze in sp ra ys ?, Phys . Rev. X. 8 (2018) 31019.
[36] B. Zhan g, G. Guo, C. Zhu, Z. Ji, C. -H. Li n, Tran sp or t an d traj ectory of cough-
induced bimo da l aero so l in an ai r-conditioned sp ace, Indoor Buil t Envi ro n.
(2020) 1420326X2094116.
[37] B. Zhan g, G. Guo, C. Zhu, Z. Ji, Tran sp or t of aero so l by coughing in an ai r-
conditioned sp ace, Proc. Therm. Fl uids Eng, Summ er Co nf., Begell Ho use Inc,
2019, pp. 1341–1353.
[38] T.D. Fa nsler, S. E. Parr is h, Spra y meas urement technolo gy : a revi ew, Meas ,
Sci. Technol. 26 (2015) 12002.
[39] W.L. Cheng, W.W. Zhan g, H. Chen, L. Hu , Spra y cool ing an d flas h
evap or ation cool ing: the curr ent develo pm ent an d ap plication, Renew,
Sustai n. Energy Rev. 55 (2016) 614–628.
[40] B. Ca i, Y. Yi n, Y. Zheng, W. Wang, H. Gu, J. Ya o, H. Wang, Ma thematical
study of sp ra y flas h evap or ation in a sp ra y-as si sted seaw ater desa li na tion
cham ber, Desa li na tion. 465 (2019) 25–37.
[41] C. Zhu, Z. Ji, M. Zhou, G. Guo, B. Zhan g, US Patent No.10661194B2, n.d.
[42] A. Isma il , Co ntro l of mu lti-stag e flas h desa li na tion plan ts: a surv ey,
Desa li na tion. 116 (1998) 145–155.
[43] Q. Chen, M. K. Ja, Y. Li , K. J. Chua , Energy , exergy an d econom ic an al ys is of
a hybr id sp ra y-as si sted lo w-temp erature desa li na tion/therma l va por
comp ress ion sy stem, Energy . 166 (2019) 871–885.
[44] Q. Chen, S. J. Oh , Y. Li , M. K. Ja, Thermo dyna mi c optimi zation of a lo w-
temp erature desa li na tion sy stem dr iv en by sensible heat so urces, Energy . 192
(2020).
[45] H. E.S. Fa th, The non-equili br ium fa ctor an d the flas hing evap or ation ra te
insi de the flas h cham ber of a mu lti-stag e flas h desa li na tion plan t,
Desa li na tion. 114 (1997) 277–287.
[46] O. Mi ya take, T. Ha shim oto, N. Li or , The rela tionship between flow pa ttern
an d therma l non-equili br ium in the mu lti-stag e flas h evap or ation pr ocess,
Desa li na tion. 91 (1993) 51–64.
[47] L. X. Hu an g, K. Kuma r, A. S. Mujumd ar , A comp ar ativ e study of a sp ra y
dr yer with ro tary disc atom izer an d pr essure nozzle using comp utationa l
fluid dyna mi c si mu la tions, Chem. Eng. Process. Process Intensif. 45 (2006)
461–470.
[48] F. Xuening, C. Lei, D. Yu ma n, J. Mi n, F. Jinping, CF D mo deli ng an d an al ys is
of br ine sp ra y evap or ation sy stem integr ated with so la r coll ector,
Desa li na tion. 366 (2015) 139–145.
[49] M. H. Sa da fi, I. Jahn, K. Ho om an , Co ol ing perf or ma nce of so li d contai ning
wa ter for sp ra y as si sted dr y cool ing towers , Energy Co nvers. Ma na g. 91
(2015) 158–167.
[50] S. Mutair , Y. Ikegam i, On the evap or ation of superheated wa ter dr ops
form ed by flas hing of li quid jets, Int. J. Therm. Sci. 57 (2012) 37–44.
[51] N. Domb ro ws ki , W.R. Johns, The aero dyna mi c instab il ity an d disi ntegra tion
of vi scous li quid sheets, Chem. Eng. Sci. 18 (1963) 470.
[52] P.H. Ma rm ottant, E. Vill erma ux, On sp ra y form ation, J. Fl uid Mech. 498
(2004) 73–111.
[53] E. Vill erma ux, C. Cl an et, Li fe of a flap ping li quid sheet, J. Fl uid Mech. 462
(2002) 341–363.
[54] E. Vill erma ux, Fr ag mentation, Annu, Rev. Fl uid Mech. 39 (2007) 419–446.
[55] T.J. Thar ak an , A. Mukhopad hyay , A. Datta, M. A. Jog, Trends in
comp rehensiv e mo deli ng of sp ra y form ation, Int. J. Spra y Co mb ust. Dyn. 5
(2013) 123–180.
[56] Y. Sun, A. M. Alkhedhair , Z. Guan , K. Ho om an , Numeri cal an d experi mental
study on the sp ra y char acteri stics of full -cone pr essure sw ir l atom izers,
Energy . 160 (2018) 678–692.
[57] S. Dehaeck, H. V. Pary s, A. Hu bin, J.P.A. J.V. Beeck, La ser ma rk ed
shad owgr ap hy: a novel optical plan ar technique for the study of
mi crobubbles an d dr oplets, Exp, Fl uids. 47 (2009) 333–341.
[58] X. Wang, C. Zhu, Co ncentric evap or ating sp ra y jets in dilute ga s-so li ds pipe
flows, Powder Technol. 129 (2003) 59–71.
[59] S. Ghaemi , P. Ra himi , D.S. Nobes, Eval ua tion of StereoPIV meas urement of
dr oplet velo city in an efferv escent sp ra y, Int, J. Spra y Co mb ust. Dyn. 2
(2010) 103–123.
[60] H. Albr echt, N. Dama schke, M. Bory s, C. Tropea, La ser Doppler an d Phas e
Doppler Meas urement Techniques, 2013.
[61] A. R. Jones, Meas urement Science an d Technolo gy Spra y Meas urement
Technolo gy : A Revi ew Recent Ci tations, 2015.
[62] H. M. Lu, Q. Jian g, Si ze-dependent surf ace tension an d Tolm an ’s length of
dr oplets, La ngmu ir . 21 (2005) 779–781.
[63] R. Mei, R. J. Adri an , Fl ow pa st a sp here with an os cill ation in the fr ee-stream
velo city an d unsteady dr ag at finite Reynol ds numb er, J. Fl uid Mech. 237
(1992) 323–341.
[64] S. L. So o, Multiphase Fl uid Dyna mi cs, Science Press, 1990.
[65] T.L. Berg ma n, F. P. Incropera, D.P. DeWitt, A. S. La vi ne, Fundam entals of
Heat an d Ma ss Tran sf er, Wiley, 2011.
[66] S. Patankar , Numeri cal Heat Tran sf er an d Fl uid Fl ow, CR C Press, 2018.
[67] ANSY S Inc, ANSY S FL UENT 19.0 Theory Guide, ANSY S Inc., Ca nonsburg , PA,
G. Guo et al.
2019.
[68] S. A. Basha, K. R. Gopa l, In-cyli nder fluid flow, turb ulence an d sp ra y mo dels -
a revi ew, Renew, Sustai n. Energy Rev. 13 (2009) 1620–1627.
[69] J. Pors tendör fer, J. Gebhar t, G. Rö big, Effect of evap or ation on the si ze
distri bution of nebuli zed aero so ls , J. Aero so l Sci. 8 (1977) 371–380.
[70] P. Bisw as , C. L. Jones, R. C. Fl ag an , Distor tion of si ze distri butions by
condensa tion an d evap or ation in aero so l instruments, Aero so l Sci, Technol. 7
17
View publication stats
Desalination xxx (xxxx) 115023
(1987) 231–246.
[71] B. Zhan g, C. Zhu, Z. Ji, C. -H. Li n, Desi gn an d char acteri zation of a cough
si mu la tor, J. Breath Res. 11 (2017) 016014.
[72] A. G. Bail ey, W. Bala chan dr an , T.J. Will ia ms , The ro si n-ra mm ler si ze
distri bution for li quid dr oplet ensemb les, J. Aero so l Sci. 14 (1983) 39–46.