Professional Documents
Culture Documents
net/publication/349863740
CITATIONS READS
8 212
4 authors:
All content following this page was uploaded by Guangyu Guo on 15 March 2021.
Desalination
journal homepage: www.elsevier.com/locate/desal
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-
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. Desalination xxx (xxxx) 115023
an atomized spray [11], the spray flash can achieve a much higher flash evaporation via the significant differences in exposed evapora-
evaporation rate than the pool flash [4]. tion area as well as in the transport characteristics (such as residence
Specifically, spray flash in a vacuuming environment is a special time) of droplets of different sizes or injection velocities [11]. Unfortu-
mechanical process of droplet evaporation, where an atomized liquid nately, this polydispersion depends not only on the types of nozzles and
spray is injected into a vacuum chamber whose pressure is maintained physical mechanisms of atomization but also on the operation condi-
by a vacuum pump by extracting the generated vapor out of the cham- tions, which leads to the nonexistence of universal modeling to predict
ber [12]. This simple spray flash has many unique advantages, includ- the atomized spray characteristics at the nozzle tip or injection inlet
ing the low-temperature evaporation, high evaporation rate within a [11,38]. Consequently, the details in spray polydispersion have to be
relatively compact chamber, and no scaling problems [8]. Neverthe- experimentally determined case-by-case. The polydispersion in atom-
less, such a simple process involves very complicated phase transfer and ized spray also imposes greater challenges to the associated evapora-
transport mechanisms, with the integrated coupling of mass, momen- tion and transport modeling. For example, an overall averaged equiva-
tum, and energy among different phases of droplets and vapor under lent diameter for hydrodynamic transport (such as the hydraulic equiv-
highly non-equilibrium and non-uniform transport conditions. alent diameter of all droplets) can be quite different from an overall av-
According to previous experimental studies in recent decades, the eraged surfa ce-area equivalent diameter for evaporation. That means
evaporation process in a spray flash can be af fected by many spray no single averaged diameter can reasonably reflect both hydrodynamic
characteristics, such as the initial temperature of sprayed droplets [13], transport and evaporation characteristics in the modeling of an atom-
the direction and velocity of injection [14,15], and the spray flow rate ized spray flash. So fa r, most models in spray flash ignore such polydis-
[16]. For example, a higher level of superheating of the injected spray persion for simplicity by assuming all droplets have the uniform size,
normally leads to higher levels of evaporating rate, amount, and effi- mass flux, and velocity [28,39,40].
ciency. Even a spray flash with a low level of superheating, for in- However, spray polydispersion may have a significant impact on
stance, 2 °C, may also be effectively completed [15]. Likewise, a spray the flash evaporation under certain circumstances. For example, for a
upwa rdly injected against gravity may also evaporate better than a spray flash in an evaporation chamber of limited dimension, some fine
downwa rd one due to a longer residence time of droplets in the trans- droplets may be able to complete their evaporation within the cham-
port and evaporation process [15]. While the experimental studies can ber, while larger ones may not have enough time to complete the va-
result a few semi-empirical correlations of some parametric effects, a porization before colliding onto the chamber boundary or exiting out
comprehensive understanding and reliable quantitative predictions of of the chamber. Using a uniform size for the droplet evaporation and
the evaporation and transport characteristics of a spray flash have to transport (even averaged over a droplet size distribution) may misrep-
be obtained from the first-principle-based mechanistic modeling. resent the total surfa ce area of evaporation or phase transfer and the
In theory, droplet evaporation is basically governed by two coupled non-uniformity in phase transport. Hence a proper modeling considera-
mechanisms: the heat transfer to the interfa ce from the inner liquid side, tion is needed to account for this polydispersion effect, such as adopting
and the heat and mass transfer on the evaporating interfa ce from the a set of sub-grouped droplet sizes and mass fluxes to mimic the polydis-
gaseous ambient [17]. In the case of droplet flash evaporation of pure persed spray transport. Consequently, in this study, a combined experi-
substance in a vacuuming environment, the evaporation rate can be ment-modeling methodology is established to introduce the polydisper-
limited by the heat transfer from the liquid side due to the intense heat sion effect of atomized sprays and the vacuuming effect on the evapo-
consumption of phase change by vacuuming and no avail of heat trans- ration capacity in a spray flash, which is realized by adopting the La-
fer to the evaporating surfa ce from the ambient side [18]. Although the grangian approach for the discrete phase of droplets [12].
relative motion between droplets and extracted vapor flow may cause In a vacuum spray flash that is absent of external heating, the la-
asymmetric and enhanced convection [19], such asymmetry and en- tent heat is self-supplied by the evaporating droplets, resulting in a con-
hancement can be ignored when the pressure difference between the tinued decrease in droplet temperature. Moreover, with the vapor ex-
saturated vapor-liquid phase on the evaporating surfa ce of the droplet traction by vacuuming, the extracted vapor is expected to possess a
and the vacuuming environment is orders-of-magnitude higher than higher temperature than the remaining droplets or liquid residue. Such
that induced from the vapor flow over the droplets [11]. a temperature difference will demonstrate such a high thermal non-
So fa r, all evaporation models are based on the phase transfer theo- equilibrium process is very efficient in evaporation, which could benefit
ries under idealized conditions such as near phase equilibrium (where many industrial processes such as thermal distillation. Moreover, the
Chapman-Enskog theory can be applied, or Boltzmann probability dis- thermal energy in the vapor of higher temperatures could also be recov-
tributions become valid) and ideal gas approximation for the vapor ered to further improve energy efficiency [41]. However, in most indus-
phase [17,20–26]. In such approaches, the evaporation rate models are trial multi-stage flash desalination plants, the superheat level of the liq-
represented by the kinetic theory model [21,27] and diffusion- uid in each evaporator remains relatively low, in which the vapor ex-
controlled model [24], whereas the evaporation heat must be provided traction is passively condensation controlled by the next stage [42–46].
from the coupled heat transfer models governed by the heat transfer Therefore, in this study, the evaporation chamber is directly connected
from the vapor side (by conduction or convection) and ambient envi- with a vacuum source on purpose (named active vapor extraction) to
ronments (such as thermal radiations), heat conduction from droplet obtain a high thermal non-equilibrium condition during spray flash.
(with or without considerations of internal motions) or a combination Meanwhile, this study aims to investigate such interesting thermal non-
of all heat transfer modes. It must be noted that neither the vacuuming equilibrium between the extracted vapor and discharged spray residue,
flash is near phase equilibrium nor the saturated vapor on the evapora- characterized by the temperature difference between the two, during a
tion surfa ce follows the ideal gas law. Nevertheless, there has been no vacuum spray flash.
better modeling available for the estimation of droplets evaporation For vacuum spray flash in a bounded environment with active va-
rate in a vacuuming flash. As a result, both kinetic theory model and por extraction, due to the non-uniformed distribution of vapor sources
diffusion-controlled model have been extended into the evaporation and velocities caused by spray polydispersion, as well as the thermal
modeling of spray flash [28–34]. non-equilibrium of vapor exists in the full-field ambient all the time, a
A very important feature in flash evaporation modeling of an atom- full-scale computational fluid dynamics (CFD) study is desired to inves-
ized spray is its polydispersion or non-uniform distributions in droplet tigate such non-uniformity and thermal non-equilibrium of vapor. Sev-
size, droplet velocity, injection angle, as well as the number density of eral CFD studies on spray flash have been reported based on different
droplets or flow rate [35–37]. The polydispersion, especially in typi- applications [47–49]. The spray flash evaporation in all these CFD
cally widely-distributed droplet sizes, causes a strong non-uniformity in studies is based on the lumped heat capacity method of the droplet,
2
G. Guo et al. Desalination xxx (xxxx) 115023
which is due to the simplified point-treatment of droplets in the discrete distribution is then implemented into the process model and CFD model
phase modeling. However, direct use of the lumped heat capacity as input conditions of spray. A lab-scale experimental spray flash evap-
model without a proper modification in evaporation rate formulation oration system has been set up for the determination of the evaporation
can be erroneous, especially for vacuum spray flash in which the evapo- coefficient as well as model validation, in which the vacuumed evapo-
ration intensity is very strong [17,40,50]. Nevertheless, introducing the rator is connected to a vacuum pump to realize a timely vapor extrac-
effective heat conduction model into droplet trajectory models in a La- tion. The extracted vapor is condensed via external condensers and an-
grangian-Eulerian CFD simulation requires a sub-grid description of alyzed. With the evaporation coefficient empirically determined under
temperature distributions within each droplet, which would dramati- spray flash in a vacuum, the evaporation model is then implemented
cally increase the computational capacity requirement beyond practi- into the process model and CFD model. The parametric process model is
cal reality. Therefore, based on the comparative study between the based on a simplification that ignores the vapor-droplet interactions,
lumped heat capacity method and the effective heat conduction model whereas the full-field three-dimensional CFD model (via user-defined
for droplet flash, a dimensionless corrective expression is proposed in function and size/time-dependent correction) is developed to reflect the
this study to modify the evaporation rate equation so that a modified reality in vapor-droplet interactions as well as the non-equilibrium and
lumped heat capacity model can be used into the Lagrangian-Eulerian non-uniformity in the phase transfer and transport.
modeling and its associated CFD simulation.
In conclusion, to study the thermal non-equilibrium phenomena ex- 2. Experimental study
isted in a spray flash with active vapor extraction, an integrated exper-
iment-modeling-CFD methodology is aiming to be proposed in this pa- The main objectives of experiments include the measurements of to-
per, and several sub-tasks need to be achieved following the research tal evaporation rate, the temperature of feed-in atomized spray, the
logic: (1) determination of polydispersion characteristics of atomized temperature of extracted vapor, the temperature of liquid discharged
nozzle spray; (2) establishment of an experimental system for measure- from evaporation chamber, the temperature profile of spray flash
ments of some spray flash characteristics, which is also used for model along with the chamber height, the pressure of vacuum chamber, the
validations; (3) development of a Lagrangian-Eulerian process model flow rate of feed-in spray, as well as the atomized spray characteristics
to investigate multiple parametric effects on the vacuum spray flash; such as injection velocity and droplet size distribution.
(4) determination of the evaporation coefficient used in the evapora- Fig. 1 shows a schematic diagram of the vacuum spray flash sys-
tion model, which represents an empirical correction for the flash non- tem. The main components of the system consist of a feed circulation
equilibrium and various non-uniformities in an atomized spray (such as pump (Warren Rupp double diaphragm pumps, manufa cture (Mfr.)
droplet shape and size); (5) development of a corrective method for the number: WR10PP6XPP9), a heater of feed heating (Vulcan immersion
lumped heat capacity method used in discrete phase modeling of CFD, heater, Mf r. number: AUW250B), a nozzle (Bete Fog Nozzle, Mf r.
realized by comparing with a derived non-dimensional effective heat number:1/4WL1-1/2-60) for generating atomized spray, a vacuum-
conduction model within the droplet during the evaporation; (6) per- maintained evaporation chamber, a feed mixer of replenished supply
form a full-scale CFD simulation to study the phase coupling of non- and spray residue, a vapor condenser, a yield chamber, and a vacuum
uniformity and thermal non-equilibrium in vacuum spray flash and va- pump. As shown in Fig. 1, the system has two main flow paths: the
por transport inside the evaporation chamber. Specifically, the polydis- feed-spray path (marked as blue) and the vapor-condensate path
persed size distribution from spray atomization is experimentally deter- (marked as dotted green). The feed is pressurized and heated-up into a
mined via a laser-scanning aided technology. The polydispersed size flowrate-temperature regulated nozzle that is inserted into the evapo-
3
G. Guo et al. Desalination xxx (xxxx) 115023
ration chamber. An atomized spray is injected from the nozzle into the [57], paralleled laser sweeping (PLS) [58], and particle image ve-
evaporation chamber, where spray flash is formed due to the signifi- locimetry (PIV) [38,59].
cant pressure differences between the super-saturated droplets and the Fig. 2 shows a schematic diagram of the experimental system for
depressurized environment. The evaporation chamber is thermally in- characterizing atomized spray. An atomized spray is illuminated by a
sulated, with ports of vapor extraction and spray residue discharge. pseudo-laser-sheet, which is further enhanced by a reflector normal to
The discharged spray residue from the evaporation chamber is mixed the laser beam. This pseudo laser sheet is formed by a vertically sweep-
with replenished feed in the mixer (with a discharging port for the op- ing laser beam from a laser scanner with a high sweep frequency [58].
tional discharge of some or all feed residue) and then recirculated, The laser sheet is arranged crossing the medium vertical plane of the
through the pump and heater, back to the nozzle. The vapor, gener- spray. The images of droplets within the illumined volume of a laser
ated from spray flash, is extracted into a condenser where the vapor beam are recorded by a digital camera, as shown in more detail in Fig.
condensate is collected. The evaporation chamber, vapor-condenser, 3. The feed spray is preheated by a heater and regulated by a pressure
and a vacuum-regulator are connected in series to the vacuum pump. regulator. The pressure drop across the nozzle spray is measured by a
Coolant flow is used to ensure sufficient heat removal and hence va- differential pressure gauge, whereas the feed spray temperature and
por condensation in the condenser. An insignificant amount of non- flowrate are measured, respectively, by a thermocouple and a rotame-
condensable gases (NCG), degassed or co-vaporized from spray flash, ter.
cannot be condensed by the condenser and hence are extracted into Similar to the droplet-marking principle in LMS [57,60], a targeted
the vacuum pump and discharged into the ambient. droplet can be marked with glaring points by a laser beam with reflec-
In this study, the temperature measurements are based on thermo- tion. The image is pre-calibrated by a scale located on the laser sheet,
couples and thermocouple profile probe. The temperature profile probe and once the marked droplet is obtained, the size will be determined
(Omega ten sensing type k thermocouple profile probes) measures the based on pixel analysis. The sizes of these laser-marked droplets can
temperature distribution (along the probe direction) during the spray thus be determined through the recorded high-resolution images. It
flash process [4,13]. The pressure measurements are obtained using should be noted that a string of multiple exposures of the same droplet
vacuum pressure gages (IFM pressure sensor, model number: PG2794). in a recorded image can also be used to determine the velocity of the
The flow rates of inlet spray and feed replenishment are regulated by droplet. The difference between the size photographing and the veloc-
flow meters (IFM efector magnetic inductive flow meter, model num- ity photographing of droplets is how to coordinate the expose time of
ber: sm6004). The feed heating is maintained by an electric-powered the camera and the sweep frequency of the laser beam. For size deter-
heater. To reduce the surfa ce scaling and limit the amount of NCG, reg- mination, the exposure time of the camera must be short enough to
ular tap wa ter is used as the feed liquid, which is also helpful for the capture a sharp image, meanwhile it still needs to be long enough for
model validation since the feed is now a single-component condensable the laser beam to finish one sweep of the view field. For velocity deter-
liquid without the complication in concentration-related concerns. mination, the exposure time becomes non-critical, however the sweep
For a spray flash, the atomization characteristics from a nozzle frequency should be fa ster enough to sweep over multiple times of the
spray majorly depend on two curial parameters, Weber number and tracked droplet during the exposure.
Reynolds number, respectively representing the surfa ce tension and the In our case, the laser beam is generated by Laser Physics Reliant
viscous effects with respect to the inertia effect [51–56]. Specifically, it 1000 m argon laser, and the laser sweep is realized by Optical Flow
will strongly depend on the type of nozzle, the feed liquid, and various System laser sweeper (Model: SCN-CN2A). The photography area
operating conditions, including the nozzle pressure, ambient pressure, (marked as yellow zone in Fig. 3) by the camera (Canon EOS 6D Ma rk
flowrate, and feed temperature in our case [11]. Hence, the atomiza- II) is located at 139–259 mm along the vertical direction (defining noz-
tion characteristics such as polydispersion are highly case-sensitive. In zle exit is 0), and 0–179 mm along the horizontal direction (defining
this study, we have developed a spray measurement system, which is spray central ax is is 0). Fig. 4 illustrates an example of captured images
based on combined concepts of laser marked shadowgraphy (LMS ) of laser-marked droplets for helping further explanation. Here the red
4
G. Guo et al. Desalination xxx (xxxx) 115023
spots are the mark point on the droplet by laser (Helium-Neon in this an image is used for droplet size determination. In Fig. 4(b), two red
case). Fig. 4(a) shows a string of five images of a droplet, with the cam- spots are identified by circled marks. It is noted that many unfocused
era exposure time is five times longer than the sweep period of the laser. droplets represent the ones outside either the laser sheet or the focal
This original image is captured in the dark environment, but the back- zone of the camera. Assuming the spray is steady-state, we can take
ground color is inverted for a better presentation. During this exposure many images for the determination of droplet size distribution.
time, any droplets moving vertically along the sweeping plane will be
highlighted five times by the laser beam, and generate a string of five 3. Vacuum spray flash model
highlight points in the digital picture recorded. Thus, based on the cam-
era shutter speed, laser scanning period, and inter-point distance (or The modeling of spray flash aims to quantify various parametric ef-
numbers of pixels covered), the instantaneous velocity of the droplet fects and process characteristics on the spray evaporation, which is cru-
can be obtained. The average droplet velocity of the spray is deter- cial to optimize the design and operation of the spray flash system. The
mined by averaging multiple sets of data from different trajectories. key parameters include the atomized spray characteristics (such as
Fig. 4(b) shows a case with the camera exposure time is about the same droplet size/velocity distribution and mass flowrate), geometric system
as one period of the laser sweep. In that case, any droplets on the laser arrangements (such as chamber shape, size and orientation, spray noz-
sheet are highlighted only one-time by the sweeping laser beam. Such zle insertion, ports of vapor extraction, and liquid discharge), as well as
5
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)
(5)
6
G. Guo et al. Desalination xxx (xxxx) 115023
The droplet surfa ce temperature is determined from the overall heat sumed that the droplets are treated as point sources in the vapor flow so
transfer to the droplet surfa ce for the latent heat required by the flash- that the volumetric effect of droplets is ignored in the vapor flow.
induced evaporation. In the absence of external heating, especially un- The mass conservation equation is given by:
der convective cooling due to lower ambient temperature, the latent
heat for the droplet evaporation has to be derived from the thermal en- (7)
ergy of the droplet itself. Assuming the heat conduction within the
droplet is spherically-symmetric, the heat conduction equation within where the S m is the vapor generation rate per volume, which is the sum
the droplet can be expressed in the spherical coordinate: of individual vapor generation rates of all droplets within the control
volume Vcell. For a polydispersed spray, S m can be expressed by:
(6)
(7a)
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 where subscripts i and j denote the ith droplet of jth size group (or on jth
evaporation [50]: 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-
(6a)
tion rate of an individual droplet which is previously solved from the
where k is the molecular conductivity, and Pe is Peclet number of the Lagrangian model. N j depends on the local number density of droplets
droplet. The initial condition is given by spray temperature at nozzle in- of group j, n j, by:
jection:
(7b)
(6b)
Assuming there is no droplet breakup or coalescence, the mass con-
The boundary conditions are given by the symmetric condition at servation leads to a droplet number density conservation, which is ex-
the droplet center and heat transfer balance at the droplet surfa ce, re- pressed by:
spectively expressed as:
(7c)
(8)
(6d)
where the Fvd is the droplet-vapor momentum transfer per volume,
which is roughly approximated by the sum of individual drag forces of
where h m is the convective heat transfer coefficient, which may be ap- all droplets within the control volume Vcell. For a polydispersed spray,
proximated using Ranz and Ma rshall correlation with vapor-film prop- Fvd can be expressed by:
erties and relative droplet velocity [65]. In which the Nusselt number is
determined by:
(8a)
(6e)
where Fij is given by:
Thus, the Lagrangian model of a droplet flash and transport is es-
tablished. With six governing equations (Eqs. (1)–(6)) for six unknowns
(8b)
(m d, ud, d d, Tl, Psat and ), the problem is closed.
The hydrodynamic transport and evaporation of different-sized
The energy equation is approximated by [66]:
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- (9)
cific surfa ce and longer time duration, and the larger droplets will have
less evaporation completeness. As a result, using one average droplet The equation of state is given by:
size (such as Sauter mean diameter) to simulate polydispersed spray
(10)
evaporation may result in a severely biased estimation of the overall
evaporation rate. Therefore, in the Lagrangian modeling of a polydis- The turbulent viscosity μ T can be determined by a turbulence model,
persed spray, droplets are sub-grouped into N size-based groups (i.e., such as the k-ε model. The turbulent thermal conductivity KT can be re-
droplets in the same group will have similar sizes) according to the lated to μ T by gas kinetic theory.
droplet size distribution, which helps to account for this polydispersion Thus, for the Eulerian model, there are four independent Eqs. (7)–
effect on the spray transport and evaporation. (10) matching four independent variables (ρ, uv, p and Tv), which can be
solved by coupling with the Lagrangian model of droplets.
3.2. Eulerian mo del of vapor flow
3.3. ANSYS FLUENT-based numerical simu lation
One of the major objectives of this study is to estimate the extracted
vapor temperature and other vapor transport properties. The vapor The Lagrangian-Eulerian model above of spray flash can be numeri-
flow is modeled using the Eulerian approach, with the coupling of cally solved using the Discrete Phase Model of the commercial CFD
evaporating sprays from the Lagrangian trajectory model. It is as- solver ANSYS FLUENT. As each iteration of the computation, ANSYS
7
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)
8
G. Guo et al. Desalination xxx (xxxx) 115023
(14)
(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.
Fig. 7. Exem plified superheat within a droplet vs time at sm all β* and Bi
nu mbe r.
9
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)
10
G. Guo et al. Desalination xxx (xxxx) 115023
Table 2
Key geome tric and operational parame ters.
Para metric Va lue Unit
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
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
11
G. Guo et al. Desalination xxx (xxxx) 115023
12
G. Guo et al. Desalination xxx (xxxx) 115023
are quite close to each other, with an average value of about 0.47. The meets a good agreement. Since both the droplet size and travel distance
tiny variations on the slopes of each operation are appeared due to the will af fect the instantaneous evaporation rate, the droplet sampled in
non-linear relation between the saturated pressure and temperature of CFD follows the same Sauter mean diameter and drops along the verti-
the wa ter/vapor. We can also notice the deviation between experi- cal central path as calibrated in experiments and defined in the process
ments and modeling predictions increases with rising the superheat de- model. The simulated overall evaporation rate, and the averaged outlet
grees of the spray, in which the experimental measurements are lower temperature of vapor and wa ter, are also compared with experimental
than computational results. Such errors may be caused by enhanced lo- measurements, which are summarized in Ta ble 4. The result shows the
cal mixing of heat and mass between the two phases, which cannot be numerical estimation of the overall thermal performance meets a good
fully considered by this simplified process models. agreement with experiments.
It is noted that the stronger superheating of spray means a bigger Fig. 15 illustrates the spray pattern of the median droplet diame-
pressure difference between the saturated vapor from droplet evapora- ter (1.03 mm) with temperature changes along the droplet trajecto-
tion and the vacuuming environment. Due to such significant non- ries. In this case, all droplets are sufficiently cooled down to the
equilibrium in pressure, the evaporative coefficient α in Eq. (4) needs to equilibrium temperature of 350 K or 77 °C before reaching the
be calibrated since the evaporate rate equation is derivated from the as- chamber wa ll.
sumption of pseudo-equilibrium. Ba sed on the experimental results of Fig. 16(a) shows the evaporation completeness of the sampled me-
the case of 60 kPa, the α is calibrated as 0.01. Assuming the evapora- dian diameter droplet along their trajectories, whereas Fig. 16(b)
tion coefficient remains unchanged within the vacuuming range in this shows the vapor generation rate distribution in the median plane of
study, the modeling predictions on the droplet temperature distribu- the chamber. Both figures indicate that the flash evaporation is very
tions along with the chamber centerline (where a multi-point thermo- fa st, with most evaporation completed within the top half-space of
couple probe is located) can be compared to the measured temperature the chamber, indicating a great potential to reduce the height require-
profile, as shown in Fig. 14. The good agreement between the modeling ment in chamber design.
predictions and measurements also partially verifies the assumption of Fig. 17 shows the vapor temperature distribution in the median
an unchanged evaporation coefficient. We can also notice, the more in- plane. Apart from the strong non-uniform distribution along the ax ial
tense the evaporation is (the higher superheat level the spray is), the direction, some very hot vapor at a temperature close to that of inlet
larger deviation is obtained during the experimental measurements. We spray is accumulated around the top of the chamber aw ay from the
believe such error results from the larger fluctuation of temperature vapor outlet while the vapor near the outlet has a much-lowered tem-
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- Table 4
mal non-equilibrium between two phases. The averaged temperature Experime ntal validation of overall ch aracteristics.
profile along the chamber centerline of the spray from the specific CFD CF D Experi ment
case is also plotted and compared in the figure. In all cases, the droplets
Evap or ation ra te 4. 30 kg /h 4. 19 kg /h
approach the thermal equilibrium about half-wa y (0.25 m) before their
Ou tlet temp erature of the va por 357. 4 K 357. 9 K
exit from the chamber, indicating a sufficient length in the chamber de- Ou tlet temp erature of the wa ter 349. 9 K 350. 2 K
sign for the spray flash. However, to focus on the evaporating moment
more clearly, the plot range is cut-off at 0.3 m.
Fig. 14. Spray temperature profile along the ch am ber centerline. Fig. 15. Example of droplets temperature along their trajectories.
13
G. Guo et al. Desalination xxx (xxxx) 115023
5. Conclusions
14
G. Guo et al. Desalination xxx (xxxx) 115023
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. Desalination xxx (xxxx) 115023
16
G. Guo et al. Desalination xxx (xxxx) 115023
17