Numerical Simulation of Heat Transfer in A Horizontal Falling Film Evaporator of Mu

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 24

DES-13075; No of Pages 24

Desalination xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Desalination

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

Numerical simulation of heat transfer in a horizontal falling film evaporator of


multiple-effect distillation
Felix Wunder a, Sabine Enders b, Raphael Semiat c,⁎
a
Technische Universität Berlin, Germany
b
Karlsruher Institut für Technologie, Germany
c
Technion Haifa, Israel Institute of Technology, Israel

H I G H L I G H T S

• A 3D model of MED for water desalination is presented


• The equations that describe the process are presented and solved
• Simultaneous evaporation and condensation considering the inside pressure drop are modeled
• Conclusions are made regarding the effects of tube diameter-length-ratios

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

Article history: The presented work describes a mathematical model for the simulation of the heat transfer in a horizontal falling
Received 12 September 2016 film evaporator for sea water desalination in the context of a multi-effect distillation process. The simultaneous
Accepted 25 September 2016 evaporation and condensation of falling films on the inside and outside of a horizontal tube is mathematically de-
Available online xxxx
scribed and solved numerically. For the description of radial heat transfer, the pipe circumference is divided into
an impingement zone and a zone of laminar flow. Different model equations are implemented for both zones.
Moreover, the accumulation of condensate at the bottom of the pipe, as well as liquid motion in axial direction
is described. The inside pressure drop along the tube, as well as the temperature changes were calculated.
The model of outer heat transfer is validated for a limited range of Reynolds numbers. It is compared to experimental
data and numerical results from literature and it is shown how the application of the model may be extended to a
larger range of outside-Reynolds numbers. Regarding the phenomena inside the tube, the calculated results indicate
a significant relevance of the pressure drop on the heat transfer for increased pipe lengths. The reducing influence of
accumulating condensate on the overall heat transfer is identified as comparably low. Finally, various combinations
of pipe diameters and lengths are analyzed and the associated transmitted heat flows are graphically summarized.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction takes over the current market [4]. Although MSF is presently the most
popular thermal desalination process [5], it is expected that MED's mar-
Growing populations, increasing economic activities and a change in ket share will grow to the expenses of MSF [6,7], especially for the re-
climate have led to serious scarcity of drinking water in many countries covery of produced water from the oil industry.
[1]. Available resources of fresh water are unevenly distributed and the With the growing demand and shortage of fresh water resources,
transport of water from one place to another has proven to be uneco- the dependence on desalination is expected to increase continuously
nomical [2]. A possible solution to encounter the global shortage of followed by the consumption and environmental impact that eventual-
fresh water was found in the industrial desalination of seawater [3]. ly are expected to grow [8]. To further improve the economics of
The most popular thermal desalination types are multistage flash MED, the energy efficiency of the process needs to be enhanced. At
(MSF), multi-effect distillation (MED) while reverse osmosis (RO) bottom, the performance of MED is determined by the evaporation
process in its falling film evaporators. Accordingly, the clear under-
standing of heat transfer characteristics of falling film evaporation is
⁎ Corresponding author at: Technion IIT, Chemical Engineering Department, Technion
City 32000-02, Haifa, Israel.
crucial for the design and optimization of MED [9,10]. In order to save
E-mail addresses: felix.wunder@rwth-aachen.de (F. Wunder), sabine.enders@kit.edu material and to reduce energy consumption, it is the objective of
(S. Enders), cesemiat@technion.ac.il (R. Semiat). this study to find ways to ameliorate the thermodynamic performance

http://dx.doi.org/10.1016/j.desal.2016.09.020
0011-9164/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
2 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Nomenclature z axial
w wall
ϕ angular
Latin letters

Variable Description Unit


a1−6 parameters - of MED. A mathematical model that approximates the physical phenom-
A cross section m2 ena of MED horizontal falling film evaporation is developed, and sugges-
c1 −6 parameters -
tions to improve the design of condenser-evaporator tubes are made.
cp heat capacity J/g/K
C parameter - In the last decades a large number of analytical and (semi-) empiri-
d tube diameter m cal models of MED-plants, heat exchangers and evaporator tubes of
DH hydraulic diameter m multiple effects distillation have been published. The studies analyzed
f friction factor - plant-wide thermodynamic performance of MED processes as well as
E parameter -
microscopic heat transfer characteristic inside falling film evaporators.
F parameter -
g gravity constant kg/m/s El-Dessouky et al. [11] developed a steady-state model of an MED
h height of liquid M plant based on mass and heat balances to investigate the plant's thermal
h heat transfer coefficient W/m2/K performance and required specific heat transfer surface. The model com-
H vertical distance between two horizontal tubes, or parameter M
prises evaporator stages and auxiliary equipment like pre-heaters, con-
k conductivity W/m/K
L tube length m
densers and flash boxes. An empirical heat transfer correlation by Han
m; _
_ M mass flux kg/s/m2 and Flechter [12] was implemented for the evaporating liquid flow,
n number of - which is valid in a range of relatively high outside liquid Reynolds num-
p pressure, vapor pressure Pa bers for MED of 700 to 7000. For the heat transfer through the condensing
P parameter - two-phase flow, an empirical correlation by Shah [13] was used. El-
Ps vapor phase parameter -
q heat flux W/m2
Dessouky et al. [11] neglected the total pressure drop assuming the fric-
Q heat flow W tional pressure drop to be compensated by the negative momentum pres-
r radius m sure drop.
S length of interfacial cord m Al-Mutaz and Wazeer [14] developed a steady-state mathematical
t tube thickness m
model of an MED-plant with thermal vapor compression, to estimate
T temperature K
U local overall heat transfer coefficient W/m2/K the influences of number of evaporation stages, motive steam pressure,
U perimeter averaged overall heat transfer coefficient W/m2/K top brine temperature and temperature difference across effects. The
w fluid velocity m/s overall heat transfer coefficient was obtained from an experimental corre-
x angular distance M lation as a function of the brine temperature. The microscopic heat trans-
~
x vapor flow fraction -
fer inside the falling film evaporators was not explicitly modeled. Good
z axial distance -
agreement between calculated results and plant data was achieved.
De la Calle et al. [15] presented a dynamic model that simulates the
Greek letters start-up operation of an MED falling film evaporator. The model ema-
nates from a single tube model that is extended to a vertical tube col-
Variable Description Unit
umn by consecutively connecting the tubes through mass balances.
Γ mass flow per tube half per length kg/s/m The model of a tube bundle is obtained by assuming the same physics
δ film thickness m for all tube columns. However, the model simplifies the calculation of
ϵ void fraction - heat transfer by assuming constant falling film thickness.
ϕ angular coordinate -
Sideman et al. [16] studied the impact of various cross-sections of
Φfr two-phase multiplier -
ρ density kg/m3 evaporator tubes on the heat transfer performance. A two-dimensional
λ heat of phase change J/kg model of simultaneous evaporation and condensation was developed,
θ inclination angle - yielding the local distribution of heat transfer coefficients. Heat transfer
μ viscosity Pa s
through the bottom liquid flow was neglected, while the height of the
τi interfacial shear Pa
ξ dimensionless temperature variable
stratified flow was assumed to be constant over the tube length. An
empirical correlation was used to determine the averaged angle where
the condensate film meets with the bottom liquid pool. Since film thick-
Subscripts nesses are assumed to be small compared to the radius, Cartesian coordi-
nates were applied. In a later study by Sideman et al. [17] experimental
Variable Description data on the simultaneous evaporation-condensation process of MED
c ,con condensation was obtained. The results showed that the heat transfer coefficients re-
crit critical point main constant with increasing liquid feed. This contradicted the theoret-
d thermally developed
ically predicted results of decreasing overall heat transfer with increasing
ev evaporation
f ,fric frictional, film liquid load. Thus, Sideman et al. [17] implemented an empirical correction
g gas factor that was suggested by Chun and Seban [18] to take into account the
H hydraulic effects of rippling that improve the heat transfer performance.
i inside surface; impingement zone; interface Maron and Sideman [19] analyzed the effect tube incline on the con-
jet jet
l liquid, layer (stratified)
densation heat transfer coefficients in co- and counter-current flow. The
max max velocity interfacial shear on the stratified liquid and pressure drop were
mom momentum modeled using a friction factor given by the Blasius equation. Static
o , out outside surface and momentum pressure drop were not modeled. Though the pressure
s surface boundary
decrease along the tube axis is calculated, the decline of driving temper-
stat static
tot total ature forces along the tube length is not considered.
v evaporation Yang et al. [5] analyzed the heat transfer in a horizontal tube of MED
using semi-empirical correlations for heat transfer during condensation

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 3

by Schlünder [20] and pressure drop caused by interfacial shear by film analysis [5,16], or by using analytically obtained heat transfer coef-
Lockhart and Martinelli [21]. The heat transfer performance for different ficients, that were corrected by using an empirical factor [17]. However,
tube lengths are evaluated for tubes of the same inner diameter. The au- (semi-) empirical heat transfer correlations, like used by Yang et al. [5],
thors suggested increasing the tube length to compensate for negative El Dessouky et al. [11], Hou et al. [22] and Al-Mutaz and Wazeer [14] do
effects of interfacial shear stress and non-condensable gases. However, not allow investigating the distribution of heat transfer coefficients
equations to analyze the effects of tube incline and height of inside liq- around the tube perimeter. As Sideman et al. [16] showed, a clear under-
uid flow were not implemented. standing of distribution of heat transfer coefficients around the perime-
In a later study, Yang et al. [1] developed a mathematical model that ter is the basis for analyzing the impact of different cross section
simulates the simultaneous film condensation and evaporation inside geometries on the overall heat transfer coefficient.
and outside of a row of horizontal tubes, based on Nusselt's falling The literature survey leads to the conclusion that few studies focused
film theory. According to the authors, the calculated temperature distri- on the detailed analytical modeling of physical phenomena inside the
bution, film thicknesses and local heat transfer coefficients agree well condenser-evaporator tubes of MED. It is known, that high vapor veloc-
with experimental data. Though the effect of pressure drop induced de- ity inside condenser tubes has two effects. On the one hand, the height
crease of driving temperature difference is mentioned, equations for the of the stratified flow is kept low, since acceleration through interfacial
approximation of pressure drop are not presented in the paper. Tube in- shear is high [19]. Thereby the effective heat transfer area is kept high,
cline is not considered in the model and modeling of liquid bottom flow under the condition that heat transfer through the liquid bottom pool
is not discussed. Furthermore, the heat transfer coefficient is calculated is neglected. On the other hand, high vapor velocity translates into
based on pure Nusselt analysis and effects of ripples and impingement higher pressure losses, whereby the driving temperature difference be-
on the heat transfer are not considered. tween condensation and evaporation side is decreased [31]. Still, these
A steady-state distributed parameter model for the simulation of the two opposing effects have not yet been studied together in one model.
performance of a practical horizontal-tube falling film evaporator has Though, in many publications heat transfer through the bottom liquid
been developed by Hou et al. [22]. The model served to evaluate the pool was considered negligible [16,24,32,33].
heat transfer performance along the tube length and in a vertical tube In this work, the heat transfer through a condensing and an evapo-
row. For the condensation side a single phase-flow and a two-phase rating falling film is investigated in a two-dimensional analysis. This
flow region are assumed and different heat transfer and pressure drop work follows the approach of subdividing the perimeter into different
correlations are implemented accordingly. To estimate the two-phase heat transfer regions, to take into account the effect of a liquid impinge-
frictional pressure drop a correlation by Friedel [23] is implemented. ment zone on the heat transfer. The evaluation of heat transfer coeffi-
The two-phase heat transfer coefficient is calculated using a correlation cients through momentum, energy and continuity balances for a
for stratified flow by Jaster and Kosky [24] that is based on Nusselt's fall- laminar falling film is supplemented by a semi-empirical heat transfer
ing film solution [25], considers the changing void fraction along the correlation for impinging liquid as described by Chyu and Bergles [34].
tube axis and neglects heat transfer through the bottom liquid pool. Regarding the heat transfer phenomena on the inside, the effects of
However, Hou et al. wrongly cited the model by Jaster and Kosky [24] pressure drop and axial liquid flow at the bottom are evaluated simulta-
to be a correlation by Chato [26]. The Chato correlation [26] does not neously in one model. In previous studies only frictional pressure drop
take into account changing height of the liquid pool. For the single was considered [4,5,19] or pressure drop was completely neglected
phase region the pressure drop is calculated by the Fanning equation, since compensation of momentum and frictional pressure drop was as-
while the empirical Sieder-Tate correlation [27] is used to approximate sumed [11]. Therefore, frictional, static and momentum pressure drop
single phase heat transfer. For the evaporation side, an empirical corre- are considered and the contributions of each component to the overall
lation by Zhang [28] is used to calculate the outside heat transfer coeffi- pressure drop is discussed in this work. Furthermore, the effect of differ-
cients. According to Hou et al. [22] the model agrees well with practical ent tube diameters and lengths on axial pressure drop and on liquid bot-
plant data. The tubes are assumed to be horizontal. Though frictional tom flow is analyzed and the influence on the heat transfer performance
pressure drop is modeled, the effect of pressure drop on the condensa- is discussed. To estimate the decreasing saturation temperatures inside
tion temperature is not considered in the model. the tube, accurate correlations to predict the thermo-physical fluid
The correlations for condensation heat transfer by Jaster and Kosky properties are implemented.
[24] and Chato [26] consider the effects of liquid pool, but do not explic- This article commences with a brief introduction into the multiple-
itly calculate the height of the assumed liquid stratified flow. Chato's effects distillation process. Next, generally existing design specifications
[26] correlation assumes a constant value for the liquid height, while of MED are discussed and the most common specification is chosen for
Jaster and Kosky [24] considered the liquid bottom flow through the further analysis in this study. Subsequently, the theoretical fundaments
changing void fraction. However, the void fraction does not give infor- for the mathematical modeling of phenomena in a multi-effects distilla-
mation on the actual distribution of total liquid mass in the condensate tion falling film evaporators are presented. A model of an evaporator-
falling film and bottom liquid pool per cross section. According to Dob- condenser tube is implemented and the model equations are summa-
son and Chato [29], the correlation of Jaster and Kosky [24] has a mean rized. Results of parametric studies are presented, discussed and
deviation of 37%. concluded.
KouhiKamali et al. [4] investigated the effects of pressure losses in
MED on energy consumption and required heat transfer surface. Pres- 2. Multiple-effects distillation
sure drops on the condensation and evaporation side and in the
demister of MED heat exchangers were considered. Therefore, correla- The original motivation for the development of the multiple-effects
tions by Lockhart-Martinelli [21] and Friedel [23] were implemented. distillation process was the need to save energy by application of
The authors showed that both correlations produce nearly the same re- more equipment [6]. Nowadays, MED has proven to be the thermody-
sults. Case studies for actual MED plants were executed, while the plants namically most effective thermal desalination technique [35]. It
comprised tubes of the same diameter, but of different lengths. operates at highest temperatures between 64 and 70 °C. In comparison,
KouhiKamali et al. [4] concluded that optimizing the unit in sense of the top brine temperature of MSF is in the range of 90 to 110 °C. Beyond
pressure drop may increase the process efficiency. that, MED consumes less total energy than MSF to produce the same
Regarding previously developed models of MED, the mathematical amount of fresh water [36]. The thermodynamic advantage of MED pri-
approximation of outside physical phenomena was either achieved by marily lies in its highly efficient falling-film evaporators and the possi-
implementing empirical heat transfer correlations [18,28,30], by using bility to use low-enthalpy waste heat e.g. exhaust steam from power
heat transfer coefficients that were obtained from a Nusselt-type falling stations [1,35].

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
4 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

MED is non-sensitive to feed quality [4] and product qualities of less Since MED plants operate under vacuum pressure, air may leak into
than 20 ppm salinity are achieved, allowing direct use of the product in the system. Besides, nitrogen, oxygen and carbon dioxide are dissolve in
industrial processes and as boiler feed water. Compared to reverse os- the seawater. The gases partially escape the liquid phase when satura-
mosis, only simple pre-treatment measures are required. Rough sea tion concentration is reached in the evaporation process. Presence of
water filters sufficiently ensure safe operation and long lifetimes to non-condensable gases that accumulate at the tube wall leads to a sig-
more than twenty five years [35]. nificant decrease of heat transfer coefficients [40]. The evaporating tem-
Using low operating temperatures between usually 35 to 70 °C, alu- perature in each stage is elevated compared to pure water at the same
minum alloys can be used as tube material [37]. As a consequence, the pressure due to the salt concentration. Hence the driving force reduces
material costs are lower compared to MSF that requires materials of in each stage while the vapor is driven into minor supersaturation.
higher thermal resistance [35]. Still, desalination techniques that are The thermodynamic performance of the process is measured by the
based on polymer-made membranes have the lowest material cost. Gain-Output-Ratio, which is defined as the quantity of tons of product
However, experimental studies by Christmann et al. [38] conclude that water per ton of initially supplied steam [6].
the application of ultra-thin polymer tubes may be possible under cer-
tain process conditions of MED. M Distillate
GOR ¼ ð1Þ
While the current market share of MED is 12.5% of the thermal desa- M Primary steam
lination market, it has a share of 6% of the entire desalination market.
Single units of MED produce quantities of up to 36,000 m3/d of fresh The GOR of MSF is claimed to be between 8 and 12 while MED man-
water. Plants consisting of up to 14 units with overall production capac- ufacturers claim GORs of 10 to 16 [2]. By adding extra effects to the pro-
ities of up to 240,000 m3/d are nowadays in operation [39]. cess, the GOR may be easily increased [4].

2.2. Design variations


2.1. Process description
Currently many different design specifications of MED are in use. The
In most cases MED uses horizontal tube falling film evaporators in a design decision whether a plant operates in co-current, counter-current
serial arrangement [35]. The evaporators are called “stages” or “effects”. or cross-current mode is based upon the existing potential of scaling. In
A process with four evaporator stages is illustrated in Fig. 1. In the first co-current operation the feed is heated to the temperature of the hot
stage high pressure steam from an outside source enters the plant and end before evaporation starts. Aforesaid design version can avoid the
evaporates preheated sea water. The secondary generated steam is precipitation of calcium sulfate for example. In counter-current opera-
used to produce tertiary vapor in the second evaporator stage. This op- tion the sea water feed starts to evaporate at low temperatures, while
eration is repeated from effect to effect between the hot and cold ends of a cross-current design stipulates all stages to be fed with preheated
the plant. In nowadays plants low temperature differences between the feed water [6].
condensation and evaporation side allow up to 16 effect [6], whereby Furthermore, designs of MED may comprise thermal vapor com-
the temperature range is dictated by the inlet temperatures of sea pression (TVC) or mechanical vapor compression (MVC). The externally
water, as cooling fluid and hot steam [35]. supplied steam from a turbine may be high-pressure and hence hotter
The vapor exiting the last effect is condensed in a condenser unit, than the hot end temperature of about 70 °C of MED. The high-pressure
while the seawater feed is preheated. To achieve total condensation, ad- hot steam serves as the motive steam of TVC. It is mixed in a steam jet
ditional seawater for cooling is required (Fig. 1) [36]. Further preheating injector with low pressure steam from the last stage and taken as the
of seawater is usually completed by heat coupling with the discharging heating steam for the first stage. Thereby the required input energy of
brine [5]. Condensate accumulates in every stage, which eventually is the process is reduced [5]. TVC increases the energy efficiency of the
the desalinated product water. After the final evaporator stage the re- process, since the latent heat of vapor leaving the low temperature
maining water vapor and accumulated non-condensable gases are re- side of the plant is reused [6]. Moreover, the exergy destruction is re-
moved by a vacuum pump that maintains the pressure gradient over duced by decreasing the temperature difference between primary
the stages. The saturation pressures of the feed steam and the product steam and seawater feed. The mechanical vapor compression MED pro-
vapor that is condensed in the final stage determine the pressure gradi- cess is driven by electric power. Thus, application of the technique is at-
ent over the plant [6]. Pressure drops inside the tube and the demister tractive for remote areas with little population. The vapor produced in
lead to a decreased condensation temperature and thereby to a lower the last stage is mechanically compressed to raise its pressure and tem-
driving temperature gradient, which negatively affects the plants effi- perature before it is fed to the hot side to produce additional vapor. The
ciency [4]. process works without down condenser and cooling water [41].

Fig. 1. Cross-current MED plant with four stages [36].

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 5

2.3. Falling film evaporators for the evaporation process of MED could be reduce to 1.5 kWh of elec-
tricity equivalent. In total, an energy consumption of about 4 kWh/m3 of
A design of MED falling film evaporators consists of a bundle of hor- product is estimated, where energy for pumping, automatic control,
izontal tube, which are connected by headers at each end [2,42]. Like il- lighting etc. is already considered. Cooling water at the last stage is as-
lustrated in Fig. 2, seawater is introduced through spray nozzles on top sumed to be provided externally, while the additional energy demand
of the evaporator unit. Liquid falls on the tubes and runs around the pe- for the pumping of cooling water is about 1–2 kWh. In comparison,
rimeter as a slightly wavy film flow at low turbulence [6]. Simultaneous- the energy demand of MSF is about 6.9 to 8.4 kWh/m3 of product, de-
ly, saturated steam enters the tubes and condenses on the inner tube pending on the GOR [6]. The energy consumption of reverse osmosis
wall. The latent heat of condensation is used to evaporate a small lies between 3.5 and 4.2 kWh/m3 of product for a sea water feed with
amount of water on the outside [16]. 3.5% salinity at a recovery ratio of about 50% [46]. However, precipita-
For MED low temperature differences are limited to a minimum by tion softening as described by Segev et al. [47] has the potential to effi-
the increasing boiling point elevation that goes along with an increasing ciently remove scaling constituents from sea water. Thereby, the top
salinity of the seawater when water is evaporated. High temperature brine temperature of MED may be increased, resulting in an improve-
differences are undesirable since the film starts to boil, leading to dry ment of the thermodynamic efficiency. Other heat source of present-
spots that favor the risk of salt precipitation [6]. Consequently, the tem- day MED plants are amongst others geothermal energy [48] and solar
perature differences between the condensing vapor and evaporating power [3]. Using direct heating by fossil fuel would significantly in-
liquid are about 1.5 to 3 °C [35]. crease the cost of energy.
In order to reduce pressure drop, parallel arrangement of the con-
densate drainage and vapor flow is suggested [6]. The removal of con- 3. Theoretical fundaments
densate is facilitated by slight tilting of the tubes, whereby inclination
of 2–3° is prevalent [40]. With regard to the tube geometry, corrugated, The theoretical fundaments for the mathematical modeling of hori-
grooved and curved surfaces have been proposed in the past to zontal condenser-evaporator tubes are presented. Most publications
achieve significant increase of heat transfer coefficients of almost focus either on the phenomena of the outside evaporating film
200%. However, in most MED units smooth tubes are used [6]. (e.g. [30,34,45,49]) or on the physics of inside (condensing) two-
Apart from horizontal falling film evaporators, vertical tube evapora- phase flow (e.g. [37,50–52]).
tors are in use [6]. Nonetheless, the heat transfer coefficients of plain
horizontal tubes are about two times higher than those of vertical 3.1. Falling film mode
tubes [44]. Besides, horizontal tube falling film evaporators offer advan-
tages considering distribution problems, non-condensable gases and When the liquid seawater film flows from one tube to the tube
fouling [45]. Besides their application in desalination systems, falling below, the flow resembles in shape discrete droplets, jets or a continu-
film heat transfer units are widely used in the chemical, refrigeration ous sheet (Fig. 3). In case of low flow rates the liquid flows in droplet
and petroleum refining industry [34]. mode and changes to column and to sheet mode once the flow rate in-
creases. To estimate the shape of the impinging liquid flow on the tube,
Hu and Jacobi [45] conducted studies for a variety of flow rates, tube di-
2.4. Energy consumption of MED ameters and tube pitches. Based on their research, they proposed four
flow transition expressions:
To estimate the energy demand for MED, the primary steam is as- Droplet to Droplet Column Mode:
sumed to be extracted from a backpressure turbine at 70 °C. Neglecting
losses, an equivalent of about 17 kWh of electricity could be produced
from 1 ton of steam [2]. According to Semiat [6] the energy demand ReN ¼ 0:074 Ga0:302
L ð2Þ

Fig. 2. MED stage, consisting of spray nozzles, horizontal evaporator tubes and a demister [43].

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
6 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Droplet-Column to Column Mode: gradient in the stagnation zone (Fig. 5). Thus, the heat transfer coeffi-
cient of the stagnation zone is defined as:
ReN ¼ 0:096 Ga0:301
L ð3Þ
0   10:5
d w max
wjet w ρ
hs ¼ 1:03Pr kl @   A
1 jet l
Column to Column-Sheet Mode: 3 ð8Þ
d xin δinit μ l
δ
ReN ¼ 1:414 Ga0:233
L ð4Þ
The velocity of the jet stream is calculated from the equation of a free
Column-Sheet to Sheet: falling body

ReN ¼ 1:448 Ga0:236 ð5Þ pffiffiffiffiffiffiffiffiffi


L wjet ¼ 2gH ð9Þ

Accordingly, the falling film mode can be inferred from a given Reyn- where H is the height of the liquid film which is determined by the dis-
olds-Number. The modified Galileo-number is defined as [44]: tance between two vertical pipes. Knowing the jet velocity and mass
flow rate, the width of the jet is calculated from the conservation equa-
ρl σ 3
GaL ¼ ð6Þ tion:
μ 4l g

The Reynolds number of the draining mass flow on the tube surface 2Γ v
δin ¼ ð10Þ
is obtained from the following equation, Γin wjet ρl
v being half the liquid mass
flow that is introduced per tube length, and μl being the liquid viscosity
[16]:
The stagnation region is followed by a short jet impingement region,
4 Γ in
v where the heat transfer is considerably high because of impingement of
ReN ¼ ð7Þ
μl the liquid. The heat transfer coefficient for a laminar flow in the im-
pingement zone was estimated by Miyasaka and Inada [49]. Under lam-
To avoid dry spots that decrease heat transfer and may lead to pre- inar conditions the heat transfer coefficient in the impingement zone is
cipitation of salts, an even distribution of flow is wanted. Hence, MED calculated as:
should be operated with sufficient flow rates to ensure wet areas at
the bottom tubes of a row. Usually, for flow rates of Γin
v = 0.025 kg/m/s  
to Γin kl r o ϕ wmax ρl 0:8
v = 0.5 kg/m/s, the falling film mode in MED is most likely to be
1
hi ¼ 0:037 Pr 3 ð11Þ
droplet or column flow according to Eqs. (2) to (5) [6]. ro ϕ μ

3.2. Heat transfer regions around the perimeter where wmax being the velocity just outside the boundary layer. The an-
gles defining the stagnation and impingement zone are derived from
Chyu and Bergles [34] presented a model for saturated, non-boiling, Fig. 4 [34]:
falling-film evaporation on a horizontal tube. The model is based on four
defined heat transfer regions: the stagnation flow region, the impinge-
ment region, the thermal developing region and the fully developed re- δin
ϕs ¼ 0:6 ð12Þ
gion (Fig. 4). Incentive for the development of the model was that a pure ro
Nusselt-type conduction analysis does only provide valid results for the
heat transfer coefficients for thin films and low flow rates [34]. δin
ϕi ¼ 2:0 ð13Þ
With regards to the stagnation zone (region I in Fig. 4), Chyu and ro
Bergles [34] argued that a correlation for the heat transfer for a sheet
jet liquid striking a plain surface can be used since the film thickness The impingement region is followed by a thermal developing region.
is much smaller than the tube radius. Such a correlation was published Laminar, steady state flow is assumed. No nucleate boiling occurs and
in a study by Miyasaka and Inada [49]. The authors found that the stag- heat transfer only takes place by conduction, while evaporation only
nation zone is characterized by the velocity Wmax just outside the hydro- occurs at the liquid-vapor interface. Furthermore, shear and drag on
dynamic boundary layer and moreover depends on the velocity the liquid are neglected as well as surface tension effects. The wall

Fig. 3. Idealized horizontal tube falling-film modes a: droplet mode, b: column mode, c: sheet mode, d is the outer diameter of the tube, λ is the axial distance between droplet flows or two
liquid columns and s is the vertical distance between two horizontal pipes [45].

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 7

The average heat transfer rate from 0° to a variable angle ϕ is given


as
2  3
g ρl −ρg
qd ð0−ϕÞ ¼ F kl ðT w −T v Þ4 5 ð15Þ
3μ l Γ N

With kl being the liquid thermal conductivity, Tw the temperature at


the outside tube wall and Tv being the saturation temperature while

2 ∞   
F ¼1þ ∑ 1− exp −n2 πP ð16Þ
πP n¼1

and
!1
kl π 3μ l ρl 3
P¼ ro ϕ ð17Þ
ρl cp;l δ2v gΓ 2v

with δw is the liquid film thickness and Γv is the local mass flow per
length on one side of the tube. The angle ϕd that marks the angular po-
sition where a fully developed linear temperature profile is assumed, is
defined as
!1
ρl cp;l 3μ l Γ 4v 3
ϕd ¼ ð18Þ
kl π r o gρ5l

Fig. 4. Heat transfer regions in the outside evaporating film. ϕs ,ϕi ,ϕd are the angles that
describe the respective ends of the stagnation and impingement zone and the thermal 4. Condensation inside tube
developing region [34]

4.1. Axial two-phase flow


temperature is assumed to be uniform and the film thickness and veloc-
ity are considered to be constant in the thermal developing region [34]. The prediction of the governing flow pattern inside the tube is cru-
Latent heat transfer in the thermally developing region is negligible and cial for the accurate estimation of the heat transfer coefficient on the
most of the transferred heat is consumed for the superheating of the liq- condensation side. In general, the condensation process along the tube
uid to a fully developed linear temperature profile. The averaged heat axis may at first consist of a dry wall superheating zone and then have
transfer coefficient in the thermal development region is given by the a wet wall superheating zone, which is followed by a saturated con-
following correlation [34]: densing zone. Finally, a liquid sub-cooling zone may exist in case of
total condensation and further cooling to the outside evaporation tem-
perature [32].
ϕd qd ð0−ϕd Þ−ϕi qd ð0−ϕi Þ Flow patterns that are typical for condensation in horizontal tubes
hd ðϕi −ϕd Þ ¼ ð14Þ
ðϕd −ϕi ÞðT w −T s Þ are depicted in Fig. 6. In case of high flow rates, annular liquid flow oc-
curs with vapor and entrained liquid in the heart of the tube. With pro-
ceeding condensation, the vapor velocity decreases which causes a
decrease of the interfacial shear on the liquid film. Consequently, the liq-
uid film becomes thicker at the bottom of the tube. Newly formed con-
densate adds to the bottom film thickness, while increasing quantity of
liquid may lead to slug flow and finally to fully liquid flow [53].
The two fundamental factors controlling the flow regimes are shear
and gravity [26]. For low vapor velocities, gravitational forces dominate
and condensate primarily forms on the top portion of the tube, while
condensate accumulates in a liquid pool at the bottom of the tube
(Fig. 7). The bottom liquid is driven out axially by vapor flow and by
gravitational forces, in case of inclined tubes. Laminar film condensation
is the dominant heat transfer mechanism if the gravitational forces on
the fluid dominate [29]. In this case the heat transfer in the upper pro-
portion of the tube may be calculated by applying Nusselt's falling film
analysis [25,32]. For shear-dominated flow regimes, which correspond
to high vapor velocities, forced convective heat transfer is the
predominating heat transfer mode [29].
For low vapor flow rates, the interfacial shear on the condensate film
is negligible and Nusselt's falling film theory may be applied to model
the condensation in the upper portion of the tube, assuming laminar
Fig. 5. Distribution of velocity just outside the hydrodynamic boundary layer, x is the
one-directional flow [32]. Practically, condensation at the inner top of
angular coordinate, w the thickness of the liquid jet umax/uj is the velocity just outside the tube starts as drop condensation [6]. The phenomena of dropwise
the boundary layer divided by the velocity of the jet stream [34]. condensation leads to heat transfer coefficients that are normally ten

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
8 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Fig. 6. Flow patterns in horizontal flow during condensation [53].

times higher than in the case of film-wise condensation, while the tem- the same magnitude. They developed a model that considers the impact
perature difference is kept constant [54]. of capillary forces on the interface curvature but neglects the shear
Because of the low vapor mass flow rates that are introduced into forces based on the model by Taitel and Dukler [58]. They justified the
the horizontal tubes, it may be assumed that stratified flow is the neglect of shear forces by the dominating gravitational forces for in-
governing flow regime in the tubes of MED falling film evaporators. clined stratified flow. Lips and Meyer [50] state that models which do
Schausberg et al. [55] showed with a flow map based on research by not take into account the capillary forces, such as those of Fieg and
Breber et al. [56] that the stratified and transitional wavy flow pattern Roetzel [60] and Hussein et al. [61] provide satisfactory results in case
predominate in MED applications. The dashed box in Fig. 8 indicates of high void fractions. If the film thicknesses are small compared to
the usual operating conditions of Multiple-Effects Distillation. Likewise, the tube radius, the impact of surface curvature is insignificant.
Shen et al. [37] estimated based on a variety of flow maps (e.g. Baker et Consequently, the surface of the stratified flow can be assumed as
al. [57], Taitel and Dukler [58], Mandhane et al. [59]) that stratified- flat in an MED application and the bottom angle ϕb is determined by
wavy flow and fully-stratified flow are the only relevant flow patterns the height of the liquid pool h(z). The basis for the calculation of the
for MED. height of the liquid pool at every axial point is the equation of motion
for an inclined laminar stratified flow. It is given by Moalem-Maron
and Sideman [19] by
4.2. Inclined stratified flow
!
The tubes of MED falling film evaporators are commonly tilted about δ2 wl δ2 wl δpl
ηl þ ¼ −ρl g sinðθÞ ð19Þ
2 to 3°, which leads to an inclined stratified flow inside the evaporator δx2 δy2 δz
tubes [40]. As previous studies show, it is common to neglect the heat
transfer though the bottom stratified flow, since it is very small com-
pared to the heat transfer through the condensing film [16,24,32,33]. where wl is the liquid velocity in radial direction, pl is the pressure in the
Estimation of the angular position where the condensation heat liquid phase and θ is the inclination angle of the tube. For the numerical
transfer ends requires to calculate the angle ϕb, where the axially solution, the pressure gradients are assumed to be constant in every dis-
flowing bottom pool meets with the condensing film. This task de- crete axial increment with the length of Δz.
mands knowledge of the inside geometry of the two-phase flow.
According to Lips and Meyer [50], the liquid-vapor interface is not
flat for low mass flows, if the capillary and gravitational forces have δpl Δpl
≈ ð20Þ
δz Δz

Fig. 8. Condensation flow patterns Gv is the inside mass flow per area and t the
temperature [55]. Gv is translated into the Reynolds number of the gas phase by the
Fig. 7. Condensation in fully stratified flow. [32]. following equation: Reg ¼ Gμv di .
g

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 9

2 2
Acknowledging the fact, that δδyw2 l ≫ δδxw2 l for thin films, Eq. (19) can be The vapor Reynolds number in case of a two-phase flow is defined as
a function of the vapor velocity, vapor viscosity and the hydraulic diam-
further simplified to
eter of the vapor phase DH,g [58]:
δ2 wl Δpl
ηl ¼ −ρl g sinðθÞ ð21Þ
δy2 Δz DH;g wg ρg
Reg ¼ ð29Þ
μg
Assuming the classical no-slip boundary condition at the tube wall
and a velocity gradient of τi/ηl at the liquid-gas interface the solutio]n
of the differential equation is The hydraulic diameter of the gas phase is defined as the quotient
  of the surface area Ag, that is penetrated by the vapor flow, and the
1 Δpl   τ perimeter:
wl ¼ −ρl g sinðθÞ y2 −y2o − i ðy−yo Þ ð22Þ
2 ηl Δz ηl
4Ag
DH;g ¼ ð30Þ
where τi is the shear stress in the flow direction and yo is the vertical Sg þ Si
distance from the liquid surface to the bottom tube wall. Integrating
Eq. (22) over the liquid cross section under the assumption of con- If the thickness of the condensate film is neglected Sg and Sl may be
stant liquid density, an expression for the liquid mass flow M _ l ðzÞ is calculated from the inside geometry that is illustrated in Fig. 9 [51].
obtained.
Sg ¼ 2ϕb r i ð31Þ
Z Si =2 Z yo
_ l ðzÞ ¼ 2ρ
M wl dydx ð23Þ
l
0 0 Si ¼ 2 r i sinðπ−ϕb Þ ð32Þ

with S i as the length of the liquid cord, subtending the angle ϕb . Equally, the liquid and vapor cross sections Ag and Alare calculated
Eventually, integration leads to the following expression [16]: from geometry (see Fig. 9), neglecting the thickness of the condensate
film.
: r 4i ρl
M l ðzÞ ¼ C ½2sin3 ðϕb −πÞ cosðϕb −πÞ−15 sinðϕb −πÞ cosðϕb −πÞ
12 (24) 1
  Ag ¼ r 2i ðϕb þ sinð2ðπ−ϕb ÞÞ ð33Þ
þ 12cos2 ðϕb −πÞ þ 3 ðϕb −πÞ 2
τ  i
þ i 12 sinðϕb −πÞ−4sin3 ðϕb −πÞ−12ðϕb −πÞ cosðϕb −πÞ Al ¼ A−Ag ð34Þ
μ l ri C
ð24Þ
Because of the progressing condensation along the tube axis and
where C is a parameter, that is constant per axial segment. the changing liquid and vapor cross sections Ag and Althe liquid and
  vapor mean velocities change along the pipe and are calculated as
1 Δpl follows:
C¼ −ρl gsinðθÞ ð25Þ
μ l Δz
M_g
wg ¼ ð35Þ
For a given liquid mass flow rate M _ l ðzÞ, the angle ϕb is obtained by A g ρg
iteratively solving Eq. (24).
An explicit description of the interfacial shear fi between stratified M_l
liquid flow and vapor flow is required to provide a boundary condition wl ¼ ð36Þ
A l ρl
for Eq. (21). Taitel and Dukler [58] developed such a model for slightly
inclined tubes. The pressure drops are assumed to be equal in both
phases and the liquid-vapor surface of the bottom stratified flow is as- 4.3. Pressure drop model for a two-phase flow
sumed to be flat. The following equation is used to approximate the in-
terface shear: Through consideration of appropriate area and volume between the
 2 tubes, the interfacial shear affected pressure drops at the shell side of
f i ρl wg −wl MED effects are kept negligible [6]. Contrariwise, the axial pressure
τi ¼ ð26Þ
2 drop cannot be assumed as zero [4].
The pressure drops may be considered to be equal in both phases
Taitel and Dukler [58] suggested calculating the friction factor fi from [51]. Consequently, the liquid pressure drop Δpl equals the overall
Eq. (26) depending on the vapor phase Reynolds Number. Eq. (27) de- pressure drop Δpc on the condensation side of the tube. The pressure
scribes the friction factor for laminar flow, while Eq. (28) expresses drop per axial distance Δz is composed of the summated pressure
the friction factor for a transient or turbulent flow. drops due to momentum change, friction and difference in altitude
For Reg b 2000: [62].

16        
fi ¼ ð27Þ Δpc Δpc Δpc Δpc
Reg ¼ þ þ ð37Þ
Δz tot Δz fric Δz stat Δz mom

For Reg N 2000:


Correlations that describe the pressure drops are differentiated into
homogeneous and heterogeneous models. Homogeneous models as-
sume equal velocities in the vapor and liquid phase and provide satisfac-
f i ¼ 0:046 Reg−0:2 ð28Þ tory results in case of vapor flow fractions near 0 or 1. The vapor flow

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
10 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Fig. 9. Stratified flow in an inclined tube. D is the inner tube radius, h is the height of the liquid stratified flow, Si is the interfacial cord length, Sg is the proportion of the inner tube
circumference that is not covered with liquid, Sl is the circumference that is covered with liquid, θ is the angle of inclination and τwg,τi,τwl respectively are the gas to wall shear, the
interfacial shear and the liquid to wall shear [51].

fraction is defined as: Consequently, the Friedel correlation [23] is suggested to model the fric-
tional pressure drop. According to KouhiKamali et al. [4] the Friedel cor-
relation is widely used to approximate the pressure drops in MED and
M_g
~x ¼ ð38Þ has been proven to show good agreement with the experimental data.
_
Mtot The frictional pressure drop is calculated as the product of the pressure
drop of the liquid phase Δpl and a two-phase multiplier Φ2fr.

The more the flow velocities of both phases differ, the less exact do ðΔpÞFriedel
fric ¼ Δpz Φ2fr ð42Þ
homogeneous models approximate the physics of a two-phase flow.
Moreover, counter-current flow cannot be described with a homoge-
neous model. In a heterogeneous model (drift model) both phases are    
Δz 1
assumed to flow separately in the pipe with different velocities and Δpz ¼ 2 f l _ 2tot
m ð43Þ
di ρl
the effects of interfacial shear is considered. The heterogeneous two-
phase static pressure drop is described as follows [62]: The liquid and vapor friction factors depend on the Reynolds num-
ber. For laminar flow (Re b 2400) the factors are given by Eqs. (44)
   
Δp and (45).
¼ ϵρg þ ð1−ϵÞρl g sinðθÞ ð39Þ
Δz stat
16
fl ¼ ð44Þ
The void fraction ϵ is defined as the cross section of the tube that is ReFriedel
l
penetrated by vapor flow divided by the entire inner cross section.
16
Ag fg ¼ ð45Þ
ϵ¼ ð40Þ ReFriedel
g
Ai

For Reynolds numbers between 2400 and 100,000, fl and fg are calcu-
A change in momentum pressure results from acceleration or decel-
lated by the Blasius equation (Eqs. (46) and (47)) [23]. Friction factors
eration of the fluid or from a changing vapor flow fraction caused by
for higher Reynolds numbers than 100,000 are not considered, since
phase change. The momentum pressure drop is defined as
this operation mode can be excluded for MED.
[62]:
  −0:25
Δp : d   f l ¼ 0:079 ReFriedel ð46Þ
¼ M tot xw g þ ð1−  l
x Þw ð41Þ l
Δz mom dz
−0:25
f g ¼ 0:079 ReFriedel
g ð47Þ
For a condensing flow the outgoing kinetic energy is smaller than
the incoming kinetic energy. Thus the change in momentum pressure
in axial direction results in a pressure recovery. Neglecting the momen- In the context of the Friedel correlation the liquid Reynolds numbers
tum pressure drop leads to some conservative, common practice design for Eqs. (44) to (47) are defined as
[63].
Many models exist that predict the two phase frictional pressure _ tot di
m
ReFriedel ¼ ð48Þ
drop (Δpc/Δz)fric and a variety of those models is described by Thome
l
μl
in [31]. Moreno Quibén and Thome [64] proposed two-phase pressure
drop models that are based on the identified flow regimes. However, _ tot di
m
ReFriedel
g ¼ ð49Þ
the model for stratified flow lacks experimental validation. On that ac- μg
count, a conventional model for the frictional pressure drop is chosen
from literature. Whalley [65] made an extensive comparison of pressure
The two-phase multiplier Φ2fr of the Friedel correlation is specified as
drop models with a data base consisting of 25,000 data points and pro-
posed to choose a model upon m _ tot, the mass flux and the ratio of the liq-
uid to vapor viscosities μl/μg [31]. In case of MED the mass velocities are 3:24 F H
Φ2fr ¼ E þ ð50Þ
below 1000 kg/m2/s [66] and the viscosities ratio μl/μg is below 1000. Fr 0:045
H Wel0:035

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 11

The Froude number FrH, the parameters E, F, H and the liquid Weber with:
number Wel are defined by Eqs. (51) to (55)  
Tc
ξ¼ 1− ð58Þ
_ 2tot
m T crit
Fr H ¼ ð51Þ
gdi ρ2h
The parameters a1 to a6 have the following values

ρl f g a1 ¼ −7:86 a2 ¼ 1:84 a3 ¼ −11:79


xÞ2 þ ~x
E ¼ ð1−~ ð52Þ
ρg f l a4 ¼ 22:68 a5 ¼ −15:96 a6 ¼ 1:8:

F ¼ ~x
0:78
ð1−~

0:224
ð53Þ while pressure and temperature of water at the critical point are

!0:91     T crit ¼ 647:096 K and pcrit ¼ 22:064 MPa


ρl μ g 0:19 μ g 0:7
H¼ 1− ð54Þ
ρg μl μl
Moreover, equations of state that calculate the saturated vapor prop-
_2 d
m erties were proposed by Wagner and Pruss in their 1993 publication
Wel ¼ tot i ð55Þ [71]. The equations are obtained from fitting the residual part of the fun-
σρh
damental equation for the Helmholtz free energy to a consistent data set
while the homogeneous density is given by [62]: based on about 20,000 experimental data [70]. These equations are valid
between the triple and the critical point. The density of the saturated
!−1 vapor is calculated from the following equation [70]:
~x 1−~x
ρh ¼ þ ð56Þ  
ρg ρl ρg 2 4 8 18 37 71
ln ¼ c1 ξ6 þ c2 ξ6 þ c3 ξ6 þ c4 ξ 6 þ c5 ξ 6 þ c6 ξ 6
ρcrit
c1 ¼ −2:03 c2 ¼ −2:68 c3 ¼ −5:39 c4 ¼ −17:3
5. Thermophysical properties kg
c5 ¼ −44:76 c6 ¼ −63:92 ρcrit ¼ 322 3
m
5.1. Sea water ð59Þ

Sharqawy et al. [67] published equations for the properties of seawa- The equations for the saturated thermodynamic properties repre-
ter that are well applicable in the temperature and salinity ranges that sent the selected experimental data by Wagner and Pruss very well
are common for thermal and reverse osmosis desalination processes. and show uncertainties of not N0.01% to 0.5% [70].
In addition a MATLAB tool box for the calculation of the thermo-physical
properties of sea water was published by Sharqawy et al. [68]. For given 6. The mathematical model
salinities and temperatures, the provided functions calculate:
• latent heats of condensation λc and evaporation λv A mathematical model is presented that allows the calculation of
• liquid densities of the condensate ρl ,c and the evaporating film ρl,v heat transfer coefficients at the condensing and the evaporating side
• liquid conductivities on the inside kl , c and on the outside kl , v of the of a single tube of an MED falling film evaporator. Therefore, the simul-
tube taneous inside condensation of saturated vapor and outside evaporation
• the Weber number Wel and the surface tension σ, required for the of sea water are modeled. Due to the symmetry of the problem only one
Friedel correlation half of the tube is implemented in order to reduce computation time
• specific heat capacities of the condensate cp,c and the evaporating film (Fig. 10).
cp,v
• liquid viscosities on the inside and outside of the tube μl,c and μl,v 6.1. Basic assumptions

The formulation of the governing model equations is simplified by


Thereby, the provided code allows calculation of pure water proper- considering the following assumptions:
ties by simply setting the salinity to zero.
According to Sharqawy et al. [67] the validity of the approximation
for the surface tension σ is limited to a temperature range of 0–40 °C.
Notwithstanding, the calculated results for 75 °C show b0.03% deviation
from the experimental result by Vargaftik et al. [69] which may be asso-
ciated to round-off errors.

5.2. Pure water

Wagner and Pruss [70] adjusted the internationally accepted corre-


lations for steam and water properties, provided by the International
Association for the Properties of Water and Steam (IAPWS) to the inter-
national temperature scale of 1990 (ITS-90). The saturation tempera-
ture of pure water Tc at the condensation side of the tube is implicitly
given by the corresponding vapor pressure correlation by Wagner and
Pruss [70].
 
Pc T crit  1:5 3 3:5 4 7:5

ln ¼ a1 ξ þ a2 ξ þ a3 ξ þ a4 ξ þ a5 ξ þ a6 ξ ð57Þ
pcrit Tc Fig. 10. Three-dimensional model of simultaneous condensation and evaporation.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
12 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Falling film flow most relevant contribution to the overall local heat transfer coefficient.
Consequently, hi is used to approximate the overall heat transfer coeffi-
• The liquid impinges on the top of the tube in form of a sheet at satura-
cient between ϕs and ϕi. Accordingly, heat flow in the impingement
tion temperature zone is calculated as
• The film flow is steady and laminar in the thermally developed region
• Interfacial shear on the inside and outside falling films is neglected Z ϕi
• The uplifting impact of the vapor density is not considered in the mo- Q_ i ¼ LðT c −T v Þ hi ðro þ δv Þdϕ ð60Þ
0
mentum balances
• Inertial forces on the liquid films are negligible compared to gravita-
Where Tv is the evaporating temperature that is a function of the
tional forces
pressure and salts concentration (affected by the boiling point eleva-
• The film thickness is small compared to the radius
tion). The condensate vapor in the impingement zone on the inside
• Tube incline is small and the liquid films only move in the direction of
tube surface is determined by the transferred heat:
gravitational forces

Q_ i ¼ L Γ c;i λc ð61Þ
Heat transfer Thus, the condensed mass per tube length in the impingement zone
Γcon,i can be calculated as a function of the heat transfer coefficient in the
• Evaporation only takes place at the liquid-fluid-interface, where the
impingement zone, the latent heat of condensation λc and the effective
temperature is at saturation with the salt solution
temperature difference:
• Heat transfer is only by conduction across the film
• Heat transfer in axial direction is negligible Z ϕi
ðT c −T v Þ
• Heat transfer through the stratified flow is neglected Γ c;i ¼ hi ðr o þ δv Þdϕ ð62Þ
λc ϕs
• No nucleate boiling occurs in the film
• No scaling occurs on the tube surface
while the stagnation and impingement angles ϕs and ϕiare obtained
• The tube is perfectly wetted; dry patches do not exist
from Eqs. (12) and (13).

Fluid properties 6.3. Heat transfer and film flow in the fully developed region

• The effect of surface tension is negligible Heat transfer in the thermally developed region is modeled based on
• The fluid properties are constant along the perimeter the conservation equations of momentum, energy and mass. The model
• The liquid density is constant provided in this work considers finite wall thickness, non-linear tem-
perature gradients in the liquid films and avoids numerical error caused
by the use of Cartesian co-ordinates.
Inside two-phase flow

• The vapor feed is saturated 6.3.1. Mass conservation


• The presence of non-condensable gases is negligible The inside and outside mass balances per tube length are given as
• Stratified flow of the liquid condensate pool follows:
• The height of the stratified flow is small compared to the radius
Γ in out
v ¼ Γv þ Γv ð63Þ

Γ c ¼ Γ out
c ð64Þ
6.2. Two-dimensional model

Inasmuch as the assumed sheet falling mode concludes equal distri- The mass flows along the perimeter are a function of the tangential
bution of the feed along the tube, the heat transfer in axial direction is mean velocities wv;ϕ and wc;ϕ and film thickness δv and δc which change
neglected and the impact of inclination on the film flow is not consid- with the angular position.
ered, the problem can be reduced to a two dimensional problem. The
Γ v ¼ wv;ϕ δv ρl ð65Þ
model considers different heat transfer regions as described in Chyu
and Bergles [34]. An impingement and a thermally developed region
Γ c ¼ wc;ϕ δc ρl ð66Þ
are implemented. Considering heat transfer in the impingement zone
allows the calculation of the beginning condensation in these zones.
Thereby the incoming mass flow for the condensation side in the fully The total evaporating and condensing masses per tube length are
developed region is estimated. This is advantageous since the numerical obtained by integrating the evaporating and condensing masses per
_ v and m
area, m _ c over the circumference.
calculation routine requires an incoming mass flow unequal to zero. The
stagnation zone is not considered, because it is assumed to be very small Z ϕb
due to low liquid loads. Moreover, the thermal developing region is Γv ¼ _ v ðro þ δv Þdϕ
m ð67Þ
neglected because the liquid is assumed to impinge at saturation tem- ϕi

perature. In the thermally developed region evaporation and condensa- Z ϕb


tion are simultaneously simulated. Γc ¼ _ c ðr i −δc Þdϕ
m ð68Þ
ϕi
6.2.1. Heat transfer in the impingement region
The heat transfer coefficients of the impingement zone hi is calculat- Since evaporation and condensation only take place at the free sur-
ed by the correlations suggested by Chyu and Bergles [34] (Eq. (11)). faces, the energy balances around the inside and outside liquid vapor in-
However, Chyu and Bergles [34] did not model simultaneous inside con- terfaces respectively yield the equations for m _ ev :
_ con and m
densation and assumed uniform wall temperature. On the inside, con-
densation is assumed to begin as dropwise condensation on top of the δT
−kl _ c λc
¼m ð69Þ
tube. Thus the outside heat transfer coefficient is considered to be the δr

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 13

δT The heat fluxes at the tube wall are functions of the respective local
−kl _ v λv
¼m ð70Þ
δr temperature gradients and the conductivities of the condensing pure
steam kl ,c and the evaporating film of seawater kl,v.

6.3.2. Energy conservation δT δT


kl;c ðr Þ r δϕ L ¼ kl;v ðr o Þ r o δϕ L ð81Þ
The radial temperature gradients in Eqs. (69) and (70) are derived δr i i δr
from the equation of energy conservation for the evaporating and con-
densing film. Considering the assumptions made above, this equation
is given for the one dimensional conductive heat transfer of a steady Inserting Eqs. (87) and (89) into Eq. (82) leads to the following
state flow as: expression:

δT rδ2 T
0¼ þ 2 ð71Þ
δr δr T −T c T −T o
kl;c i  ¼ kl;v v  ð82Þ
ri r o þ δv
ln ln
r i −δc ro
The boundary conditions of the evaporating film are imposed by the
outside surface temperature of the tube To and the saturation tempera-
ture Tv, which corresponds to the pressure pv at the shell side. The film
thickness δv is a function of the angular position. The fact that the heat flux through the tube equals the entering and
exiting heat fluxes yields Eq. (83). This additional equation allows the
surface temperatures of the tube wall to be expressed as functions of
T ðr o Þ ¼ T o ð72Þ Ti and To, while the temperature gradient inside the tube wall is as-
sumed to be linear.

T v −T o ðT −T i Þ
T ðr o þ δv Þ ¼ T v ð73Þ o
kl   ¼ kw o ð83Þ
ro þ δv t
r o ln
ro
Respectively the boundary conditions of the inside evaporating film
are obtained. Since the presence of non-condensable gases is neglected,
the temperature at the vapor liquid interface Tc inside the tube is equal
to the saturation temperature that corresponds to the local vapor pres- 6.3.3. Momentum conservation
sure pc. The film thicknesses δc and δvand velocities wvϕ and wcϕ at each angu-
lar position are derived from the momentum equation. Simplification of
Tðri Þ ¼ Ti ð74Þ the Navier-Stokes equation with the assumptions made above leads to
the following differential equation, describing the motion of film flow
T ðr i −δc ðϕÞÞ ¼ T c ð75Þ along the perimeter of a tube
 
Integrating Eq. (71) twice subsequently yields the temperature gra- δ 1δ  ρ gsinðϕÞ
− r wϕ ¼ l ð84Þ
dients and temperature profiles for the outside (Eqs. (76) and (77)) and δr r δr μl
the inside (Eqs. (78) and (79)) films.
Assuming no slip at the inside and outside tube surface and no
δT T −T o shear at the vapor-liquid-interfaces leads to the following boundary
ðr Þ ¼ v  ð76Þ
δr r o þ δv conditions:
rln
ro
wv;ϕ ðr o Þ ¼ 0 ð85Þ
 
r
ln
r þ δv wc;ϕ ðr i Þ ¼ 0 ð86Þ
T ðr Þ ¼ T v −ðT o −T v Þ  o  ð77Þ
ro þ δv
ln
ro δwv;ϕ
ðr o þ δv ðϕÞÞ ¼ 0 ð87Þ
δr
δT T i −T c
ðr Þ ¼   ð78Þ
δr ri δwc;ϕ
rln ðri −δc ðϕÞ Þ ¼ 0 ð88Þ
r i −δc δr
 
r This yields for the r , ϕ -velocity profiles of the evaporating and con-
ln
r −δc densing film:
T ðr Þ ¼ T c þ ðT i −T c Þ  i  ð79Þ
ri
ln ( ! ! )
r i −δc g ρl sinðϕÞ 1 r3 þ 2 ðr o þ δv Þ3 r 3 þ 2 ðr o þ δv Þ3
wv;ϕ ¼ − ðr o þ δv Þ2 2ðr o þ δv Þ− o 2
−r o 2
þ r2
3μ l r r2o þ ðr o þ δv Þ r2o þ ðro þ δv Þ
The temperature profiles of the condensing and evaporating film are ð89Þ
linked by the fact, that the heat flux from the condensing film into the
tube wall equals the heat flux to the evaporating film:

( ! ! )
g ρl sinðϕÞ 1 r3 þ 2 ðri −δc Þ3 r3 þ 2 ðri −δc Þ3
wc;ϕ ¼ − ðr −δc Þ2 2ðr i −δc Þ− i −r i þ r2
3μ l r i ri þ ðri −δc Þ
2 2
r i þ ðri −δc Þ
2 2

Q_ in ¼ Q_ out ð80Þ ð90Þ

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
14 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

The mean velocities are obtained from Eqs. (91) and (92): shear stress is calculated using the correlation by Taitel and Dukler [58]
((27) to (29)), while the changing inside two-phase geometry and
Z r o þδðϕÞ
1 vapor and liquid velocities are estimated using Eqs. (30) to (33). To take
wv;ϕ ¼ wv;ϕ dr ð91Þ
δv ro account of changing thermo-physical properties, correlations for the den-
sity and temperature of saturated vapor, as described by Wagner and
Z ri
1 Pruss in [70], are implemented (Eqs. (57) to (59)).
wc;ϕ ¼ wc;ϕ dr ð92Þ
δc r i −δc
6.6. Implementation and program structure
Integration yields:
! Solving of the equation of motions for the falling film and stratified
 
1 g ρl sinðϕÞ r o þ δv ðr o þ δÞ2 r 3o þ 2 ðr o þ δv Þ3 flow requires discretization of the equations. To numerically solve the
wv;ϕ ¼ f− ½ ln − þ 2ðr o þ δv Þ
δv μl ro 3 r 2o þ ðro þ δv Þ2 described mathematical model, the tube geometry is sectioned into dis-
! crete elements. A MATLAB code is implemented that comprises two
1  r 3 þ 2 ðr þ δ Þ3 1 
− ðro þ δv Þ2 −r 2o o o v
þ ðr o þ δv Þ3 −r3o g ð93Þ nested “for-loops“, an outside loop for the axial discretization and an in-
6 r 2o þ ðro þ δv Þ2 9
side loop for the angular discretization.

  ! 6.7. Axial discretization


1 g ρl sinðϕÞ ri ðr i −δc Þ2 r3i þ 2 ðr i −δc Þ3
wc;ϕ ¼ f− ½ ln − þ 2ðr i −δc Þ
δc μl r i −δc 3 r2i þ ðr i −δc Þ2
In axial direction, the tube is divided into nz segments while each
!
1 2  r 3 þ 2 ðr −δ Þ3
i c 1 3  segment has the length of
− r −ðr i −δc Þ2 i
þ r −ðr i −δc Þ3 g ð94Þ
6 i r 2i þ ðri −δc Þ2 9 i
L
Δz ¼ ð99Þ
nz
Introducing Eqs. (65) and (66) into Eqs. (93) and (94) respectively
provides implicit correlations for the inside and outside film thickness The mass flow rate of the stratified flow is calculated for every axial
at a specific angular position. segment from the flow rate in the preceding axial segment and the
condensate mass flow at the bottom angle in the preceding segment
6.4. Heat transfer coefficients Γc(ϕb,z − Δz) = Γcon(z − Δz). So along the tube the condensate accumu-
lation is given by:
The heat transfer coefficient of the evaporating and condensing film
at the tube surface in the fully developed region are calculated as M _ l ðz−ΔzÞ þ 2Γ con ðz−ΔzÞΔz
_ l ðzÞ ¼ M ð100Þ
follows:
For a chosen incoming total mass flow the vapor mass flow is calcu-
ðr o þ δv Þm _ ev λv lated as
hv ¼ ð95Þ
r o ðT o −T v Þ
_ g ðzÞ ¼ M
M _ tot −M
_ l ðz−ΔzÞ ð101Þ
ðr −δc Þm _ con λc
hc ¼ i ð96Þ
r i ðT c −T i Þ In the first discrete axial section the bottom is assumed to be dry as
the vapor flow fraction ~x ¼ 1. Since the film thickness on the outside of
Introducing the conductivity of the tube wall kwthe overall heat the tube becomes mathematically infinite at 180° a finitely smaller
transfer coefficient Ud is obtained [54]: angle is chosen for ϕb in the first segment to ensure that a solution is ob-
tained. For subsequent axial segments, the bottom angle ϕb is obtained
1 from iteratively solving Eq. (24) for a given liquid mass flow. The liquid
Ud ¼   ð97Þ
r r ro r mass flow at the axial position z is calculated as follows:
þ ln þ
hc r i kw ri hv r o
Xz
_ l ðzÞ ¼
M 2Γ con ðz0 ÞΔz ð102Þ
The average overall heat transfer coefficient at a specific axial posi- z0 ¼0
tion of the tube is calculated by averaging and summing the heat trans-
fer contributions from each region [34]. In every segment the pressure drop is calculated and subtracted
from the inside pressure of the regarded segment to obtain the pressure
ϕi −ϕs ϕ −ϕi
U ðzÞ ¼ hi þ Ud b ð98Þ for the following axial segment.
π π
 
Δpc
pc ðzÞ ¼ pc ðz−ΔzÞ−Δz ð103Þ
6.5. Three-dimensional model Δz tot

The two dimensional model is extended to a three dimensional


Eq. (41) for the calculation of momentum pressure drop is
model of a horizontal pipe by introducing correlations that approximate
discretized as
the inside stratified flow and calculate the pressure change and
resulting change in thermo-physical fluid properties along the tube  
Δpc

axis. The motion of the liquid stratified flow is modeled based on Eqs. ≈M_ tot 1 ~xðzÞwg ðzÞ þ ð1−~xðzÞÞwg ðzÞ
(22) to (25) that were provided by Moalem-Maron and Sideman [19]. Δz mom Δz
ð104Þ
− ~xðz−ΔzÞwg ðz−ΔzÞ þ ð1−~xðz−ΔzÞÞwg ðz−ΔzÞ
While Moalem-Maron and Sideman [19] approximated the total pressure
drop in the liquid phase using a friction factor from the Blasius equation,
static and momentum pressure drop are considered in this work as well while the momentum pressure drop is neglected in the first segment.
(Eqs. (38) to (41). The Friedel correlation [23] (Eqs. (42) to (56)) is cho- Subsequently, the saturation temperature and density for the next
sen to approximate the two-phase frictional pressure drop. The interfacial section is recalculated by using Eqs. (57) to (59) that were provided

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 15

by Wagner and Pruss [70]. Further temperature dependent liquid and Furthermore, the effects of changing saturation temperature, effective
vapor properties, are recalculated in every segment using the MATLAB temperature difference, tube incline and diameter proportions are eval-
code provided by Sharqawy et al. [68]. uated in a three-dimensional analysis.

6.8. Angular discretization 7.1. Two-dimensional analysis

In every axial segment an angular calculation routine is executed. For the study of two-dimensional heat transfer phenomena, a tube
The angular partition between the stagnation angle and the bottom geometry is chosen as presented in Table 1. An analytical [16] and ex-
angle is subdivide into nϕ segments, yielding the angular increment perimental and analytical [17] analysis was conducted by Sideman et
al. based on the same geometry. However, the calculations in [16]
ϕb −ϕs neglected wall thickness and assumed linear temperature gradients.
Δϕ ¼ ð105Þ
nϕ Moreover, an impingement zone was not considered.
The results are presented for a tube of the first stage of an MED pro-
For every segment the mass flows are constant and the film thick- cess, where the temperature Tc of saturated hot steam is 70 °C and the
nesses are iteratively calculated accordingly. The mass flows on one salinity is 35,000 ppm. A driving temperature difference ΔT = Tc − Tv
tube half per axial distance are estimated for every angular segment, of 2 °C is assumed across the films, while the bottom angle is stipulated
whereby the output mass flows are the input mass flows for the to be ϕb = 179.97°.
succeeding segment.
7.1.1. Heat transfer in the stagnation and impingement region
_ con ðri −δc ÞΔϕ
Γ c ðϕÞ ¼ Γ c ðϕ−ΔϕÞ þ m ð106Þ
The calculated results show little impact of the impingement zone
_ ev ðr o þ δv ÞΔϕ according to Chyu and Bergles [34] on the predicted overall heat trans-
Γ v ðϕÞ ¼ Γ v ðϕ−ΔϕÞ þ m ð107Þ
fer performance. For ReN = 150, the stagnation zone reaches from 0° to
the stagnation angle ϕs = 0.14° and the impingement zone reaches
The total condensate flows and vapor product per axial segment are
from ϕs to the impingement angle ϕi = 0.47°. Though ϕs and ϕi increase
calculated as
with the Reynolds number, ϕi still has a value of below 3° for ReN = 900.
Xnϕ    Nonetheless, the average heat transfer coefficient of the impingement
Γc ¼ _ con r o −δv kϕ Δϕ
m ð108Þ
kϕ ¼1
zone hi is about 60% greater than hv(ϕi), the heat transfer coefficient
Xnϕ    in the first segment of the adjacent thermally developed region. Howev-
Γv ¼ _ con r o −δv kϕ Δϕ
m ð109Þ
kϕ ¼1 er, hi is 2.5 times smaller than the perimeter averaged heat transfer
coefficient Ud of the thermally developed region for ReN = 150. The im-
pact of the heat transfer coefficients of the impingement zone on the
6.9. Average overall heat transfer coefficient of a tube overall heat transfer coefficient is small. Neglecting hi the perimeter av-
eraged overall heat transfer coefficient U decreases by about 0.7% for
The average overall heat transfer coefficient for the horizontal tube is ReN = 150 and about 2% for ReN = 900.
calculated as the sum of the average overall heat transfer coefficients of Chyu and Bergles [34] did not model simultaneous inside condensa-
each axial segment, divided by the number of discrete segments. tion and their analysis for the thermally developed region was based on
XL the assumption of uniform wall temperature. Notwithstanding, the au-
U ðzÞ thors concluded that the most relevant contribution to the overall heat
U tube ¼ z¼0
ð110Þ transfer coefficient comes from the thermally developed region for low
nz
flow rates.
This chapter completed the aggregation of a new mathematical Numerically computed results through computational fluid dynam-
model for the simulation of heat transfer in a horizontal tube of an ics by Yang et al. [72] show that the zone of highest heat transfer exists
MED falling film evaporator. on top of the tube. This circumferential partition reaches from 0° to
about 20° for Reynolds Numbers of 600 to 1250 and is thus much larger
7. Results and discussion than the impingement zone described by Chyu and Bergles [34]. The
higher heat transfer in the first section of the tube perimeter is attribut-
The investigation of heat transfer performance of a horizontal evap- ed to a calculated recirculation flow.
orator-condenser tube yields the local heat transfer coefficients around
the circumference and along the tube axis. At first, a two dimensional 7.1.1.1. Heat transfer in the thermally developed region. The two-dimen-
analysis evaluates the effects of an impingement and a thermally devel- sional analysis of heat transfer in the thermally developed region dis-
oped region. Moreover, the distribution of local heat transfer coeffi- cusses the computed results for heat transfer coefficients, film
cients, film thicknesses and surface temperatures are studied around thicknesses, wall temperatures and mass transfer rates in dependence
the perimeter and compared to experimental and computed results on the axial coordinate. The calculated results are compared to experi-
from literature. Then, the calculated average overall heat transfer coeffi- mental data and previous analytically predicted data and the model as-
cients of a three-dimensional evaporator tube are compared to experi- sumptions are discussed.
mental results from other authors. Thereby, the validity of the model
is restricted to a certain parameter range of outside flow rates. 7.1.1.2. Heat transfer coefficients around the perimeter. For the thermally
developed region, Fig. 11 illustrates the inside, outside and local overall
heat transfer coefficients along the circumference. The local overall heat
Table 1
Plant geometry and process parameters. transfer coefficient of the thermally developed zone Ud is evaluated at
the inside tube wall. As Fig. 11 shows, the outside heat transfer coeffi-
Variable Value Unit
cient exhibits a maximum at ∅ = 83°, resulting from the changing
ri 0.0175 m film thickness around the perimeter. A locally high heat transfer coeffi-
ro 0.019 m cient corresponds to locally thin films. Because the evaporating film is
H 0.01 m
much thicker than the inside condensate film, the outside heat transfer

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
16 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

minimum of the outside wall temperature translates into a minimum


of the driving temperature force across the evaporating film. The in-
crease of To for ∅ N 138° is explained by the thermally isolating effect
of increasing film thicknesses in this circumferential partition.
The outside wall temperature changes by b0.4 °C between 0° and
180° which are about 0.6% of the value for the wall temperature at 0°.
However, 0.4 °C translate into 20% of the effective temperature
difference across the film of ΔT = 2 °C. Thus, considering the wall
temperature as uniform, as done by Chyu and Bergles [34] and
Kocamustafaogullaris and Chen [73] may lead to a noticeable over-pre-
diction of evaporated liquid, as a higher temperature gradient is
assumed across the evaporating film.
As expected, the temperature gradient across the outside film is by
far higher than the gradient in angular direction. The temperature gradi-
ent in radial direction has the dimension of 1 to 10 °C/mm, while the
temperature gradient in angular direction is not higher than 0.1 °C/mm
at Tc = 70 °C. This finding supports the neglect of heat transfer in angu-
lar direction. CFD results, as presented by Yang et al. [72], conclude that
a conduction and a convection sub-layer exist across the film. The au-
thors acknowledge that thereby models that neglect convection do
Fig. 11. Inside, outside and overall heat transfer coefficients. ReN = 150. still satisfactorily predict the heat transfer.

7.1.1.3. Distribution of film thickness around the perimeter. The variation


of outside and inside film thickness with ∅ is depicted in Fig. 13.
coefficient determines the magnitude of heat transfer. However, as While the outside film thickness mathematically approaches ∞ for
computational fluid dynamics (CFD) let assume, maximum heat trans- ϕ → 0 and ϕ → 180°, the film thickness on top of the tube is actually dy-
fer is expected to occur in the upper circumferential section [72]. namically changing since the liquid flows in droplets at ReN = 150 and
The obtained results for the distribution of heat transfer coefficients Tc = 70 °C.
around the perimeter are compared to the results of the calculations by According to Eqs. (2) and (3), transformation to droplet column is
Sideman et al. [16]. The results do qualitatively agree, regarding the expected to occur for ReN N 280, whereas the falling film mode changes
course of evaporation and condensation heat transfer coefficient over to column flow at ReN = 355. Equally the thickness on the downside of
ϕ, and have the same order of magnitude as the analysis of Sideman the tube periodically changes due to detaching droplets. The inside con-
et al. that was made on pure water. Heat transfer coefficients decrease densate film meets with the stratified layer at ϕb.
with the salinity of the liquid feed and augment with the saturation Hou et al. [74] measured the thickness of a falling film around a tube
temperature. Pure water yields about 5% higher heat transfer coeffi- with a diameter of do = 25.4 mm between the angles ϕ = 15° and ϕ =
cients than water at 35,000 ppm salinity. However, neglecting the 165°. Their experimental results and the results of a Nusselt falling film
wall thickness changes the estimated value for U d by not N0.01%. analysis are plotted in Fig. 14. As shown in Fig. 15, the model developed
Fig. 12 shows the outside and inside wall temperatures To and Ti in in this work leads to similar results for film thicknesses for the respec-
dependence of the angular coordinate. The inside wall temperature de- tive geometry and Reynolds number as the Nusselt analysis by Hou et
creases from the top of the tube to the bottom, resulting from an in- al. [74]. However, the temperature and salinity of the water in the ex-
creasing condensate film thickness. The driving temperature periments are not given in [74]. At 70 °C the calculated viscosity changes
difference between To and Tiis largest when∅ is about 83°. Thin films in- by approximately 10% between pure water and water with 35,000 ppm
crease the temperature gradients across the two films and tube wall,
resulting in higher temperature differences. Furthermore, a local

Fig. 13. Inside and outside film thicknesses δc and δv along the perimeter for ReN = 150
Fig. 12. Change of wall temperatures with the angular coordinate. ReN = 150. and Tc = 70 °C.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 17

ReN = 574 as used in the experiments by Hou et al. [74]. Yet, they
showed for Reynolds numbers below ReN = 200 that the impact of iner-
tia terms is negligible. The authors introduced a boundary condition on
top of the tube of non-zero velocity and deduced the inlet velocity pro-
file from the feeder height, assuming the liquid to flow as a jet. However,
they did not validate their predicted film thicknesses with experimental
data in [73].
The experimentally obtained data by Xu et al. [75] for the film thick-
ness agree well with the numerically estimated values of the new model
for low Reynolds numbers. At 35 °C an outside Reynolds number of
ReN = 150 corresponds to a liquid flow per tube length of about Γv =
0.03 kg/m/s. For this input parameters, the implemented code yields a
distribution of film thicknesses with a minimum of about 0.2 mm,
which agrees well with the experimental data reported there.
Concluding from the experimental results by Hou et al. [74], an in-
creasing vertical distance between two horizontal tubes leads to a de-
crease in average film thickness. This observation is explained by an
increase in liquid flow velocity. However, the model does not allow pre-
Fig. 14. Thickness distribution of an evaporating falling film around a circular tube, diction of this comportment. The liquid is assumed to stagnate at ϕ =0.
provided by Hou et al. [74]. δ the film thickness and s the distance the liquid falls before
striking on the tube surface. Measurements were carried out using a displacement
Hence, the quality of the calculated values decreases with enlarging
micrometer at ReN = 574. The outer diameter of the test tube was d0 =25.4 mm. inter-tube distance.
As the computations show, only about 1% of the liquid feed evapo-
rates. This suggests that computation time may be reduced by comput-
of salts. The viscosity is about 70% lower at 35 °C than at 70 °C, which ing momentum and energy balance separately, neglecting the impact of
translates into thicker films at low temperatures (Fig. 15). evaporated mass on the film thickness.
Unlike mathematically predicted, the minimum film thickness δv is In conclusion, the analytical solution of film thicknesses as provided
estimated to occur at a gradually higher angle than ϕ = 90° (Fig. 14). in this study agrees well with experimental data if outside Reynolds
Moreover, the calculated film thickness in the circumferential partition numbers remain low (~ 200). For higher Reynolds numbers, a model
from ϕ = 90° to ϕ = 165° is higher than the experimentally estimated needs to be developed that takes into account the effects of free falling
values. This may result from an under-prediction of the film velocity in height of liquid on the film velocity. Furthermore, inertial forces need
this angular section, due to non-consideration of inertial forces in the to be considered in the modeling equations for liquid flow rates.
model.
Kocamustafaogullari and Chen [73] considered inertial forces in their
falling film analysis. The authors solved the governing equations using 7.2. Three-dimensional model
the finite difference method and concluded that the effects become rel-
evant only for Reynolds numbers above ReN = 800, where the inertia 7.2.1. Overall heat transfer coefficients of an evaporator-condenser tube
terms have the same magnitude as viscous and gravity terms. Nonethe- The predicted overall heat transfer coefficients of a tube U tube in de-
less, comparison of the experimental results by Hou et al. [74] (Fig. 14) pendence of liquid feed rates were compared to the experimental re-
to the analytical results by Kocamustafaogullaris and Chen [73] (not sults of Shen et al. [76] and to the experimental and analytical results
shown) concludes that agreement between experimental data and of Sideman et al. [17]. For both cases the model uses as input parameters
computed film thicknesses may be improved by considering inertia the respective liquid feeds, outside and inside Reynolds numbers, tem-
terms in the momentum balance already for a Reynolds number peratures and tube geometries of the experiments. The overall heat
transfer coefficients were computed for Tc = 60 °C. For a water feed
rate per tube length of 0.03 kg/m/s, translating into ReN = 253, the com-
puted and experimentally estimated results agree well. Nonetheless,
the deviation between experimental and calculated values augments
with increasing liquid feed. While for Γv = 0.04 kg/m/s (ReN = 358)
the deviation is about 15%, the computed overall heat transfer coeffi-
cient is about 40% lower than the experimentally estimated data. In
the numerical analysis higher sprinkling rates lead to higher film thick-
nesses. This is unfavorable for heat transfer, yielding lower heat transfer
coefficients. Practically, a higher liquid flow may lead to growing film
velocities, which aggravates the convective heat transfer and increases
the waviness of the flow, causing eddies at the surface that enhance
the heat transfer.
Shen et al. [76], Xu et al. [77] found the overall heat transfer coeffi-
cient to increase with higher Reynolds numbers in their experiments.
In contrast, the experimental data provided by Sideman et al. [17]
show that the heat transfer coefficient is unaffected of changing Reyn-
olds number in the range of ReN = 120–360. This may be explained by
greater heights of free falling liquid. In the experiments by Shen et al.
[76] the distance between tubes was H = 33 mm, while the value for
φ H was not discussed in the publication by Xu et al. [77] and H was
only 10 mm in the experiments by Sideman et al. [17]. If tubes were
Fig. 15. Thickness δv of the evaporating film, depending on the angular coordinate ϕ and not fully wetted at lower Reynolds numbers, an increase of liquid feed
saturation temperature Tc for ReN = 574 and d0 = 25.4 mm. may have led to full wetting and thereby to an increase of measured

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
18 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

heat transfer coefficients. However, the factors that influence the exper- along the tube axis is illustrated in Fig. 17A. At the tube entrance, the
imental outcome may be affected by other parameters. For example, the mass flow rate M_ tot of the progressing vapor condensation, correspond-
heat transfer performance is negatively influenced by increasing con- ing to Reg = 1170, decreases, leading to a diminishing vapor velocity wg
centrations of non-condensable gases in the vapor feed. and Reynolds number in z-direction.
The experimental and numeric results of Sideman et al. [17] were The mean vapor and liquid velocities wg and wl are illustrated in Fig.
compared with the results obtained from the code developed in this 17B. While condensate accumulates in the liquid pool, the liquid veloc-
work. The predicted and measured values agree reasonably well for ity increases. Due to a growing interfacial surface, the liquid is accelerat-
ReN = 150. The under-prediction of heat transfer coefficients at low ed by higher shear forces. Simultaneously a growing height of the liquid
Reynolds numbers may be accounted to dynamic effects of droplet im- pool facilitates the establishment of velocity profiles in the cross section
pingement on the tube. Contrariwise, the model assumes a continuous with a higher mean velocity. As the difference between vapor and liquid
sheet striking on the top of the tube. For ReN = 300, the numeric results velocity decreases, the interfacial shear equally declines, resulting in di-
are about 30% smaller than the empirically estimated values. Differences minished liquid velocity. Hence, the flow pattern remains fully stratified
in the numeric values of Sideman et al. [17] may be accounted to different according to the flow map by Breber et al. [56], that was specified for
fluid properties in the calculations. Furthermore, unlike Sideman et al. MED application by Schausberg et al. [66] The flow map asserts that
[16,17] this study considers non-uniform bottom film thickness along transitional flow occurs for mass flows per area of Gv N 4 kg/m ² 2/s,
the tube length. The effects of bottom flow are discussed in detail below. which corresponds to a Reynolds number of Re g = 12,000. The forces
on a fluid element of the condensate falling film at ϕ =90° with the in-
7.2.2. Distribution of heat transfer coefficients in axial direction terfacial area ΔAfilm are contemplated. The calculated results for the
The same tube geometry as described in Table 1 is applied for the shear forces Fshear = τiΔAfilm in axial direction and gravitational forces
basic three dimensional analysis. The three dimensional geometry is Fg = gΔΑfilmδcρl have the same magnitude yet for much lower Reynolds
completed by the length L and the inclination angle θ. While the Reyn- numbers than Re g = 12,000. For a tube length of L = 1 m and an inlet
olds number is stipulated to be ReN = 150, the temperature difference Reynolds number of Reg = 1130, the shear forces on a fluid element of
ΔT between the initial condensation temperature Tc at the inlet of the the liquid film are approximately 10 to 20 times higher in the first
tube and the evaporation temperature Tv is two degree Celsius. At the fifth of the tube. On the other hand, modern MED stages can be as
tube inlet, the mass flow rate is M _ tot = 0.35 g/s, yielding an initial long as 7 m, depending on the tube diameter. Thus, the highest expected
mean vapor velocity of wg = 1.8 m/s. At the outlet of the tube, the Reg will be in the order of 8000. As a consequence, motion of the falling
vapor remaining flow ratio ~x is about 4%. film in axial direction is expected, leading to a smaller film thickness and
For an initial saturation temperature at the condensation side of Tc = liquid pool height in the first partition of the tube. In any case, modeling
70 °C, the perimeter averaged overall heat transfer coefficient is illus- the inside phenomena as stratified flow and one dimensional conden-
trated in Fig. 16. The initial decrease is explained by the starting conden- sate falling film over the entire tube length as suggested by Moalem-
sation, leading to increasing liquid flow at the bottom of the tube, Maron and Sideman [19], Sideman et al. [16] and Yang et al. [1] is
whereby the effective heat transfer area is reduced. Though condensate questionable.
continues to accumulate in the subsequent axial partition, the heat
transfer coefficient decreases at a lower rate, because the liquid pool is
7.2.4. Height of the stratified flow for different levels of tube incline
driven through shear by the vapor flow until almost all vapor is con-
The height of the stratified flow is plotted over the tube length in Fig.
densed. The perimeter averaged heat transfer coefficient U declines
18A. The impact of tube incline on the inside liquid flow is very little.
from the tube inlet to the tube outlet by about 5%, while the average
The height of the liquid flow at the tube outlet is by 20% lower at an in-
overall heat transfer coefficient of the tube is U tube = 3.425 kW/m2/K. clination of θ = 60 than at θ = 00. However, the effect of inclination on
the height of the stratified flow h and thereby on the bottom angle ϕb
7.2.3. Vapor and liquid flows inside the tube becomes only relevant in the last fifth of the tube. On the one hand, in-
For an initial saturation temperature at the condensation side of cline of the tube leads to a larger average bottom angle ϕbwhich trans-
Tc = 70 °C, the course of the Reynolds number of the gas phase Reg lates into a larger angular partition where heat transfer takes place.
Equally an incline of the tube creates an elliptical cross section which
leads to gradually higher heat transfer coefficients according to Fieg
and Roetzel [60], who studied the impact of tube incline on evaporation
heat transfer coefficients. On the other hand, the incline leaves parti-
tions of the tube dry and unused for heat transfer. While an incline of
6° reduces the outside heat transfer area by at least 0.4%, the increase
of the effective heat transfer area through larger bottom angles ϕb is es-
timated to be in the dimension of 0.01%. Consequently, a minimal in-
cline of the tubes is considered to be sufficient to enable removal of
condensate in case of a plant shut-down. During operation, the interfa-
cial shear is the prevailing force that affects the motion of the liquid bot-
tom pool.
The computed results by Moalem-Maron and Sideman [19]
show a lower height of the liquid pool of 0.02 to 0.08 mm and a
less steep increase of liquid pool height at the end of the tube.
Vapor was assumed to have totally condensed at the tube end,
while the effective temperature difference remained as dependent
parameter in the authors' calculations. However, comparability of
the results is limited since the authors did not indicate in [19] for
which incoming vapor mass flow the values were computed. High
vapor mass flows and low effective temperature differences trans-
late into low heights of liquid bottom pools, due to high shear and
Fig. 16. Perimeter averaged overall heat transfer coefficient U along the tube axis. low condensation rates.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 19

_ tot ¼ 0:35g=s; Tc(z=0) = 70 °C.


Fig. 17. A: Reynolds number of the gas phase Reg. B: Mean vapor and liquid velocities wg and wl along the axial tube coordinate z. ReN = 150, M

The bottom angle ϕb is dictated by the height of the stratified flow h this is partially explained by the comparably higher heat of evaporation
and plotted over the axial coordinate in Fig. 18B for an incline of θ = 2°. at low temperatures. On the other hand, the outside films are thicker at
ϕb marks the angular position from whereon heat transfer through the low temperatures, resulting from an increased viscosity. At 70 °C the
tube wall is neglected in this calculation. The bottom angle changes viscosity of seawater is by approximately 70% lower than at 35 °C,
from 180° at the tube entrance to about 165° at the tube outlet, which resulting in 25% higher minimum film thickness. At lower temperatures
corresponds to a decline of ca. 8% and results in a decline of the overall the increase of outside film thickness overweighs the impact of compa-
perimeter averaged heat transfer coefficient along the tube axis. How- rably thinner condensate films and a diminished height of the stratified
ever, as Fig. 16 shows, the perimeter averaged heat transfer coefficient flow inside the tube. As a conclusion, the lower temperature effects of
only declines by 5%. This is explained by the fact, that a decrease of MED yield less product than the higher temperature stages, assuming
the bottom angle to values near 180° has little impact on the overall equal heat transfer area per stage.
heat transfer coefficient, since the outside heat transfer coefficient is These results indicate, that a future solution of the scaling problem at
low too in the angular section close to 180°. In conclusion, the impact temperatures above 70 °C would allow improving the process efficiency
of the bottom liquid flow on the effective heat transfer area is low, as with small effort. The gain-output-ratio may be elevated with less mate-
long as the height of the liquid flow does not translate into a significant rial expenses by adding stages at the higher temperature end. It may be
reduction of the bottom angle ϕb. Thus, the design of condenser-evapo- interesting as well to increase the temperature driving force at lower
rator tubes needs to enable sufficient acceleration of liquid bottom flow temperatures.
by high enough vapor flow rates and some tube incline. Further, it is to mention that for all temperature levels, the heat
transfer coefficients are calculated with the same incoming mass
7.2.5. Effect of temperature level flow of M _ tot ¼ 0:35g=s (Reg = 1130). In praxis, the vapor quantity
Fig. 19 displays how the overall average heat transfer coefficients of produced per stage decreases with the saturation temperature,
a single tube U tube decreases with diminishing saturation temperatures. thus leading to diminished incoming vapor mass flow in direction
The heat transfer coefficient is about 30% lower at 35 °C than at 70 °C. of decreasing temperature. Controlling the temperature differences
Consequently, a lower amount of steam condenses inside the tubes in the different stages may improve the usage of the steam at the
and a smaller quantity of seawater is evaporated. On the one hand, lower temperature stages.

_ tot ¼ 0:35g=s;
Fig. 18. A: Height of the stratified flow along z for different angles of tube incline. B: Bottom angle ϕb, where the condensate film meets with the stratified flow. ReN = 150, M
Tc(z=0) = 70 °C.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
20 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

Fig. 19. Average overall heat transfer coefficient U tube for initial vapor saturation
temperatures Tc(z = 0) between 35 and 70 °C. For each temperature level, the tube is
_ tot ¼ 0:35g=s.
fed with a saturated vapor mass flow of ReN = 150, M
Fig. 21. Frictional, static, momentum and total pressure drop along the z-axis. L = 1 m,
ReN = 150, M _ tot ¼ 0:35g=s.
7.2.6. Effects of temperature difference
The angular averaged overall heat transfer coefficients U~ for different
ΔT are illustrated in Fig. 20. The tube-averaged overall heat transfer co- The analytical explanation agrees with the experimental results by,
efficient U tube decreases by about 4% from net driving force. ΔT = 1 °C, to Xu et al. [77] who found the inside heat transfer coefficients to decrease
ΔT = 2 °C. With increasing difference between the inside condensation with increasing driving temperature difference. The heat transfer coeffi-
temperature Tc and the outside evaporation temperature Tv, the heat cients of the condensation side were found to decrease by about 7% be-
flux increases and more vapor condenses. As a consequence, the inside tween ΔT = 3 °C and ΔT = 1 °C. However, the overall heat transfer
film thickness grows and at the same time, the amount of liquid flowing coefficients were not measured by Xu et al. in [77]. The impact of the
at the bottom increases, leading to a reduction of the effective heat condensation heat transfer coefficient may be negligible on the overall
transfer area. Hence, the average overall heat transfer coefficients in heat transfer coefficient for great difference between the dimensions
every axial segment are diminished by increasing ΔT. For ΔT = 2.5 °C of evaporation and condensation heat transfer coefficient.
and ΔT = 3 °C, total condensation has taken part at respectively z = In conclusion, less heat transfer area is required to evaporate a cer-
0.64 m and z = 0.81 and the vapor flow fraction ~x reaches zero before tain amount of feed seawater for larger Δ T in the effects of MED. On
the outlet. the other hand, the initially supplied energy from a primary steam can
be reused fewer times and less product fresh water is produced, reduc-
ing the gain-output ratio.

7.2.7. Pressure drop


The pressure gradients along the tube axis z are illustrated in Fig. 21.
The two-phase frictional pressure drop Δpfric is the prevailing contribu-
tion to the overall pressure increment Δptot. As a result of the rising liq-
uid velocity wl, the frictional pressure drop augments from z = 0 to z =
0.25 until the effect of diminishing vapor mean velocity causes a de-
crease of the pressure gradient. With increasing liquid flow along the
tube bottom, the static pressure increment Δpstat increases and causes
slight pressure recovery, while the contribution of momentum pressure
change Δpmom is negligible. For the analyzed geometry and M _ tot ¼ 0:35
g=s the overall pressure drop along the tube is 5.6 Pa, which leads to an
insignificant decrease of the inside saturation temperature Tc at the tube
end.

7.2.8. Effect of tube length


For larger tube lengths the pressure drops across the system and
their effect on the saturation temperature are expected to be slightly
more significant. Therefore, the heat transfer performance of tubes
with diameters and wall thicknesses as described above for lengths be-
Fig. 20. Perimeter averaged overall heat transfer coefficient depending on axial position tween one and eight meters is investigated. For the analysis, the vapor
for a condensation temperature of 70 °C inside the tube under different temperature mass flow that is fed to the tube is proportionally increased with the
_ tot ¼ 0:35g=s.
differences. L = 1 m, ReN = 150, M tube length (Table 2).

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 21

Table 2
Analyzed tube lengths and corresponding vapor feed mass flows and Reynolds Number.

Length L (m) 1 2 3 4 5 6 7 8

M_ tot (g/s) 0.35 0.7 1.05 1.4 1.75 2.1 2.45 2.8
Reg (-) 2260 1130 3390 4520 5650 6780 7910 9040

The highest mass flow of 2.8 g/s in Table 2 translates into a mass flux With lower condensation thinner falling film thicknesses appear,
to the tube of 3.6 kg/m2/s. For this mass flux at 70 °C, the flow regime is followed by a reduction of the height of the stratified flow. Hence, the
still stratified flow according to the flow pattern map by Schausberg et overall average heat transfer coefficient of a tube gradually increases
al. [66], though the flow regime may be transitional or wavy stratified with its length (Fig. 22C). However the increase is only marginal; U tube
in the first partition of the tube. is b 1% higher for a tube length of one meter than for an eight-meter-
The results of the analysis are illustrated in Fig. 22. As Fig. 22A shows long tube.
how the pressure drop Δptot over the tube length increases significantly The computed results contradict El-Dessouky's assumption in [11],
with L, whereby the frictional pressure drop Δpfric is the only relevant that frictional pressure drops are compensated by momentum pressure
contribution. The total pressure drop for a 1 m long tube is 5.6 Pa. In recovery. Moreover, non-consideration of the decrease of condensation
comparison, the pressure drop along an eight-meter-long tube is temperature caused by pressure losses as in the models of Hou et al. [22]
358 Pa, while the static pressure recovery is b 2% of the total pressure and Moalem Maron and Sideman [17], leads to an over prediction of
drop. Pressure drops are higher for longer tubes since the initial and av- heat transfer. The overall heat transfer performance is dictated mainly
erage Reynolds numbers Reg are significantly higher (Fig. 22B). More- by the evaporating film. An analysis of the accumulated gases along
over, the fluid covers a larger distance and more liquid condensate the tubes was not performed here. Non-condensable gases are the
accumulates. A higher liquid share per cross section aggravates the fric- main reason for the practically lower impact of inside heat transfer on
tional losses, as more energy is transferred from the vapor to drive the the overall heat transfer. Pressure losses inside the tube may increase
higher liquid mass to the outlet. the shear on the condensing layer and therefore further reduce the
With increasing pressure drop along the tube of increased length condensing temperature, reducing the initial driving temperature
(Fig. 22A) the vapor flux to the tube increases and the average overall difference for longer evaporator tubes and hence may reduce the evap-
heat transfer coefficient is slightly increased (22C), while the saturation oration capacity.
temperature Tc reduces (22D), whereby the driving force for the heat In practice, the pressure gradients across MED plants are in the range
transfer is reduced. While the temperature change between the inlet of 5–50 kPa, while pressure drops of b5 kPa per stage of MED are known
and outlet is negligible for one-meter-long tubes, the effective temper- [6]. Total pressure drop per stage is the sum of pressure losses inside the
ature difference decreases by about 13% along an eight-meter-long tube. evaporator tubes, on the shell side and in the demister. KouhiKamali et
As a consequence, a comparably lower flowrate of vapor condenses and al. [4] showed analytically that the pressure drops inside the tubes pre-
the vapor velocity wg remains relatively high. vail. Nonetheless, experimental estimation of pressure drops for

_ tot on the A: total pressure drop along the tube Δptot, B: the initial Reynolds number Reg (z = 0), C: the
Fig. 22. Effects of different tube length L and hence different inlet mass flows M
average overall heat transfer coefficient of the tube U tube and D: the saturation temperature at the tube outlet Tc(z=l).

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
22 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

different evaporator tube lengths is strongly needed to validate the it is interesting to calculate the heat flux q_ transferred for given tube
model predictions. geometries.
As Fig. 23 shows, the highest heat flux is attained with short tubes
and small diameters. This comportment results from higher shear forces
7.2.9. Effect of diameter-length ratio in the narrower tubes, affecting the liquid stratified flow. In the studied
Earlier computations of MED falling film evaporation were executed range, d0 = 25 mm is the most favorable diameter for tubes below 2 m
for one tube diameter only [1,4,16,17,19]. An analysis of different diam- length. Yet the differences are low. For longer tubes, above 3 m, a diam-
eter-length combinations could not be found in the literature, though eter of 45 mm yields the best performance, since pressure drops remain
KouhiKamali et al. [4] analyzed different practical plants with different at a moderate level and for a seven-meter-long tube, the heat flux for
tube lengths but same diameters. d0 = 45 mm is about 35% higher than for d0 = 25 mm. Accordingly,
Evidently, the fluid velocities in the tube impact the heat transfer by the decrease in effective temperature difference is the dominating
two effects. First, the shear stress controls the height of the stratified effect.
flow, dictating the effective heat transfer area. Second, the frictional For the first axial segment, where no bottom liquid flow exists, the
pressure losses significantly depend on the fluid velocities. Due to the averaged outside heat transfer coefficients at ReN = 150 almost have
low temperature differences in MED, pressure drops along the tube the same values for all discussed tube sizes. It is highest for tubes with
strongly affect the driving temperature difference through a reduction d0 = 25 mm, while it is 0.2–3% smaller for tubes of d0 = 35 mm and
of the condensation temperature in axial direction. To achieve a ther- d0 = 45 mm.
modynamically most efficient process, total condensation of the incom- The total tube area is commonly used as basis for economic calcula-
ing vapor inside the tube is desired. However, since it is difficult to tions of the capital costs of MED [78]. The diagrams presented in Fig. 24
adjust the exact flow that will be reduced to 0 at the outlet, little quan- may allow to determine for which tube geometry the highest heat
tity of the vapor remains uncondensed, yet joins the feed to the next transfer performance is expected for a given mean surface. Fig. 24B de-
stage. For the following analysis different diameter and length propor- picts the feasible heat transfer rates for given tube geometries. Together
tions are analyzed on their heat transfer performance for a vapor exit with the isosurface diagram in Fig. 24A the map in Fig. 24B shows the
at the outlet of ~x ¼ 2%. heat flux for different pairs of length and diameter while Fig. 24C allows
Increasing the outside diameter of short tubes between d0 = 25 mm estimating at which diameter-length-ratio a required heat flow and
and d0 = 45 mm increases the calculated overall heat transfer coeffi- thus mass transfer rate may be obtained at lowest material expenses.
cient by only 6–7%. The reason mainly lies in the change of the bottom For example, a required heat flow rate of 3430 W (Fig. 24C), may be
condensate angle, for low tube inclination. However, longer tubes attained with a seven-meter-long tube and a diameter of 37 mm
with lengths from one to seven meters and outer diameters of 25, 35 which corresponds to a mean surface of 0.8 m2. However, the same
and 45 mm present an increase of a few more percent of the overall heat flow rate may be realized with a 3.6 m long tube and a diameter
heat transfer coefficient with increased length. This is mainly due to of 45 mm, which translates into a 37.5% smaller mean heat transfer
the increase of vapor velocities inside longer tubes that affect the con- area of 0.5 m2 per tube. This calculation shows the considerably high
densate flow. prospects of optimization of diameter-length ratios to mitigate pressure
The increase in the vapor flow, increases also the internal pressure drops and to increase the effective heat transfer area in MED.
drops, leading to a decrease in inside saturation temperature and con- A comparison of attainable heat flows for bundles of tubes of differ-
densation ratio and thus to lower film thicknesses. The increase is ent diameters, is presented in Table 3. Equal distance between the tubes,
steepest for the inner diameter of d0 = 25 mm (di = 23 mm), since fric- a tube length of six meters and a mean area of 10 m2 is assumed for both
tional losses are higher for this tube. As Fig. 23 depicts, the driving tem- diameters.
perature difference reduces along the axis by 50% for d0 = 35 mm at 6 m However, the design of MED-evaporator tubes must take into ac-
tube length. For a 7 m tube with an outside diameter of 45 mm, ΔT de- count the effects of non-condensable gases in the vapor feed that may
clines by only little N10% along the tube axis, while the loss of effective significantly increase the required tube length to attain almost full con-
temperature difference surpasses 60% for d0 = 25 mm. densation at the tube end [40]. Likewise, additional tube length needs to
Two opposing effects occur at increasing pipe lengths and decreas- be considered for MED designs that operate with super-saturated vapor
ing diameters: on the one hand, the effective temperature decreases, feeds [6]. Moreover, optimization of the diameter length proportion
while on the other hand, overall heat transfer gradually improves and must take into account the entire design of an MED stage. Though com-
the effective heat transfer area of a single tube increases. Consequently, parably high values of di/L lead to a reduction of condensation pressure

Fig. 23. A: Inside saturation temperature Tc at the tube outlet for different tube length L and outer diameters d0, B: the overall heat transferred along different length tubes at different
diameters. Thickness is d0 − di = 1 mm. T(z = 0) = 55 °C.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
F. Wunder et al. / Desalination xxx (2016) xxx–xxx 23

to avoid pressure drops and interfacial shear at the shell side that also
affects the total material cost per effect. Furthermore, the optimum
tube geometry or material is not the same for every temperature level,
confronting the fact that the lowest manufacturing cost for tubes are
achieved if every effect uses tubes of the same design.
No attempt was made here to solve the generated vapor flow and
the pressure drop outside the tubes. The pressure drop outside the
tubes may affect the vapor temperature and hence reduce the driv-
ing force and heat flux. The impact of the liquid stratified bottom
flow on the overall heat transfer performance was found to be little,
as will probably be the effect of the shear flow outside the tube. The
analysis however, yielded information on the most economic combi-
nations of diameter-length for a given range of required heat flow
rates. In the future, experimental analysis of pressure drops for different
tube geometries is required to provide a basis for future improved plant
design.

8. Conclusion

A mathematical model of a horizontal tube of MED falling film evap-


oration has been developed. The model considers simultaneous falling
film condensation and evaporation. A perimeter partition of laminar
falling film flow is supplemented by an impingement zone, where
heat transfer is described using a semi-empirical correlation. Regarding
the inside phenomena of two-phase flow, non-agreement on the con-
sideration of pressure drops was found in the literature. However,
agreement on the neglect of heat transfer through the bottom liquid
flow was found. Consequently, a model was developed that considers
an inside liquid pool and its motion in axial direction, as well as pressure
drops along the tube axis. Frictional pressure drop was modeled using
the Friedel correlation. Moreover, equations that well approximate the
change of thermodynamic data and thermo-physical fluid properties
with decreasing pressure were implemented.
However, the consideration of an impingement zone has little im-
pact on the predicted overall heat transfer coefficient. Nonetheless, the
model predictability was proven in a certain range of outside liquid
Reynolds numbers. Through a comparison of the computed results to
models from literature and to experimental data, recommendations of
how to extend the model validity to a broader range of parameters
could be made. For higher Reynolds numbers, the consideration of iner-
tial forces in the momentum balance is recommended as much as the
consideration of convective terms in the heat balance and the use of a
correction term for the waviness of falling film flow.
For a liquid outside Reynolds number that produces results for the
overall heat transfer coefficient that agree well with experimental
data, the effects of bottom stratified flow and pressure drops were ana-
lyzed. The height of the liquid stratified flow was found to be little and
had thereby little impact on the overall heat transfer performance. The
analysis of pressure drops led to the conclusion, that momentum and
static pressure drop may be conveniently neglected. Contrariwise, the
frictional pressure drop attains the dimension of 100 Pa for tube lengths
of 3-5 m and higher and translates into significant decreases of driving
Fig. 24. A: Iso-surfaces Am depending on outer diameter d0 and tube length L. B: attainable
heat flux q_ for diameter-length ratios at outlet vapor qualities of ~x ¼ 2%. C: Attainable heat
temperature difference. Different diameter-length combinations were
flows Q_ for diameter-length ratios at outlet vapor qualities of ~x ¼ 2%. studied to investigate the opposing effects of low liquid pool height
and high pressure drops on the heat transfer performance for high
vapor velocities. The analysis yielded a summary of attainable heat
drops, the space between the horizontal tubes is decreased and may flow rates for given diameter-length combinations.
negatively affect the entire-stage performance. Another aspect of the However, empirical evidence is strongly needed. It is immanent to pro-
distance between tubes is to provide sufficient space for vapor product vide experimental data to weigh the impact of different diameter-tube ra-
tios on the thermodynamic performance of MED to improve the
Table 3 economics and sustainability of thermal seawater desalination in the future.
Attainable heat flows of tube bundles.

Diameter (mm) 25 45 Acknowledgement


Theoretical number of tubes (-) 21.2 11.8
Heat transfer/tube (MW) 0.7 5.2 Felix Wunder's stay at the Technion Haifa – Israel Institute of Tech-
Heat transfer/bundle (MW) 14.9 61.3
nology was supported by the Young Scientist Exchange Program by

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020
24 F. Wunder et al. / Desalination xxx (2016) xxx–xxx

the German Federal Ministry of Education and Research and the Israeli [36] H.M. Ettouney, H.T. El-Dessouky, R.S. Faibish, P.J. Gowin, Chem. Eng. Prog. 98 (12)
(2002) 32.
Ministry of Science and Technology (YSEP 90). He highly expresses his [37] S. Shen, R. Liu, Y. Yang, X. Liu, J. Chen, Desalin. Water Treat. 33 (1–3) (2012) 218.
gratitude for the financial support and likes to further thank his super- [38] J.B.P. Christmann, L.J. Krätz, H.-J. Bart, Appl. Therm. Eng. 38 (2012) 175.
visors Prof Sabine Enders and Prof Rafi Semiat for their time, support [39] H.M. Ettouney, M. Wilf, Seawater desalination: conventional and renewable energy
processes, in: A. Cipollina, G. Micale, L. Rizzuti (Eds.), Green Energy and Technology,
and consideration. Further thank is expressed to Avi Sagiv, Ala Garra, Springer, Heidelberg, New York, 2009.
Shiran Ring and Shiran Shultz who always helped with questions [40] G. Caruso, D. Vitale Di Maio, A. Naviglio, Desalination 309 (2013) 247.
concerning software tools or the Technion's infrastructure. [41] A.S. Nafey, H. Fath, A.A. Mabrouk, Desalination 230 (1–3) (2008) 1.
[42] M. Awad, E.S.R. Negeed, IJND 3 (3) (2009) 283.
[43] M. Khamis Mansour, M.A. Qassem, H. Fath, Desalin. Water Treat. (2014) 1.
References [44] J.R. Thome, Wolverine Heat Transfer Engineering Data Book - Chapter 14 Falling
Film Evaporation, 2009.
[1] L. Yang, C. Gu, Z. Xu, X. Zhang, S. Shen, Desalin. Water Treat. (2014) 1. [45] X. Hu, A.M. Jacobi, J. Heat Transf. 118 (3) (1996) 616.
[2] R. Semiat, Environ. Sci. Technol. 42 (22) (2008) 8193. [46] R. Semiat, J. Sapoznik, D. Hasson, Desalin. Water Treat. 15 (1–3) (2012) 228.
[3] P. Palenzuela, D. Alarcón, G. Zaragoza, J. Blanco, M. Ibarra, Desalin. Water Treat. 51 [47] R. Segev, D. Hasson, R. Semiat, Desalination 281 (2011) 75–79.
(4–6) (2013) 1229. [48] B. Rahimi, A. Christ, K. Regenauer-Lieb, H.T. Chua, Desalination 351 (2014) 202.
[4] R. KouhiKamali, A.S. Kojidi, M. Asgari, F. Alamolhoda, Desalin. Water Treat. 46 (1–3) [49] Y. Miyasaka, S. Inada, J. Chem, Eng. Japan/JCEJ 13 (1) (1980) 22.
(2012) 68. [50] S. Lips, J.P. Meyer, Int. J. Heat Fluid Flow 36 (2012) 83.
[5] L. Yang, S. Shen, H. Hu, X. Chen, Desalin. Water Treat. 24 (1–3) (2012) 101. [51] C. Tzotzi, N. Andritsos, Int. J. Multiphase Flow 54 (2013) 43.
[6] R. Semiat, in: UNESCO (Ed.), Encyclopedia of Life Support Systems, 2011. [52] J.M. Mandhane, G.A. Gregory, K. Aziz, Int. J. Multiphase Flow 1 (4) (1974) 537.
[7] I.S. Park, S.M. Park, J.S. Ha, Desalination 182 (1–3) (2005) 199. [53] J.W. Palen, G. Breber, J. Taborek, Heat Transfer Eng. 1 (2) (1979) 47.
[8] T. Mezher, H. Fath, Z. Abbas, A. Khaled, Desalination 266 (1–3) (2011) 263. [54] J.H. Lienhard, A Heat Transfer Textbook, Version 1.31, 3rd ed. Phlogiston Press, Cam-
[9] J. Uche, J. Artal, L. Serra, Desalination 152 (1–3) (2003) 195. bridge, Massachusetts, U.S.A., 2008
[10] S. Shen, S. Zhou, Y. Yang, L. Yang, X. Liu, Desalin. Water Treat. 33 (1–3) (2012) 300. [55] IDA World Congress on Desalination and Water Reuse, 2009.
[11] H.T. El-Dessouky, I. Alatiqi, S. Bingulac, H. Ettouney, Chem. Eng. Technol. 21 (5) [56] G. Breber, J.W. Palen, J. Taborek, J. Heat Transf. 102 (3) (1980) 471.
(1998) 437. [57] O. Baker, Fall Meeting of the Petroleum Branch of AIME, 1953 10–19.
[12] J. Han, L.S. Fletcher, Ind. Eng. Chem. Process. Des. Dev. 24 (3) (1985) 570. [58] Y. Taitel, A.E. Dukler, AICHE J. 22 (1) (1976) 47.
[13] M.M. Shah, ASHRAE Trans. 84 (4) (1978). [59] J.M. Mandhane, G.A. Gregory, K. Aziz, Int. J. Multiphase Flow 1 (4) (1974) 537.
[14] I.S. Al-Mutaz, I. Wazeer, Desalin. Water Treat. (2014) 1. [60] G.P. Fieg, W. Roetzel, Int. J. Heat Mass Transf. 37 (4) (1994) 619.
[15] A. De la Calle, L.J. Yebra, S. Dormido, in: M. Otter, D. Zimmer (Eds.),Proceedings of [61] H. Hussein, M.A. Mohamad, A.S. El-Asfouri, Renew. Energy 23 (3–4) (2001) 525.
the 9th International MODELICA Conference, September, 2012. [62] M. Kraume, Transportvorgänge in der Verfahrenstechnik: Grundlagen und
[16] S. Sideman, D. Moalem-Maron, R. Semiat, Desalination 17 (2) (1975) 167. apparative Umsetzungen, VDI, Springer, Berlin, 2004.
[17] S. Sideman, D. Moalem, R. Semiat, Desalination 21 (2) (1977) 221. [63] S. Bhramara, K.V. Sharma, T.K.K. Reddy, ARPN J. Eng. Appl. Sci. 4 (9) (2009) 64.
[18] K.R. Chun, R.A. Seban, J. Heat Transf. 93 (4) (1971) 391. [64] J. Moreno Quibén, J.R. Thome, Int. J. Heat Fluid Flow 28 (5) (2007) 1060.
[19] D. Moalem-Maron, S. Sideman, Int. J. Heat Mass Transf. 25 (9) (1982) 1439. [65] P.B. Whalley, in Heat Exchanger Design Handbook (Ed: G. F. Hewitt), Hemisphere,
[20] E.U. Schlünder, Heat Exchanger Design Handbook, Hemisphere Pub. Corp, Washing- Washington, DC.
ton, 1983. [66] P. Schausberg, J. Nowak, O. Medek, IDA World Congress on Desalination and Water
[21] R.W. Lockhart, R.C. Martinelli, Chem. Eng. Prog. 45 (1) (1949) 39. Reuse November, 2009.
[22] H. Hou, Q. Bi, X. Zhang, Int. Commun. Heat Mass Transfer 39 (1) (2012) 46. [67] M.H. Sharqawy, J.H. Lienhard, S.M. Zubair, Desalin. Water Treat. 16 (1–3) (2012)
[23] L. Friedel, European Two-Phase Flow Group Meeting, Ispra, Italy, Paper E2, 1979. 354.
[24] H. Jaster, P.G. Kosky, Int. J. Heat Mass Transf. 19 (1) (1976) 95. [68] M.H. Sharqawy, K.H. Mistry, Thermophysical properties of seawater, Center of Clean
[25] W. Nusselt, VDI-Z 60 (27) (1916) 541. Water and Clean Energy at MIT and KFUPM.
[26] J.C. Chato, American society of heating, Refrig. Air Cond. Eng. J. 4 (1962) 52. [69] N.B. Vargaftik, B.N. Volkov, L.D. Voljak, J. Phys. Chem. Ref. Data 12 (3) (1983) 817.
[27] E.N. Sieder, G.E. Tate, Ind. Eng. Chem. 28 (12) (1936) 1429. [70] W. Wagner, A. Pruss, J. Phys. Chem. Ref. Data 31 (2) (2002) 387.
[28] J.C. Zhang, (MSc Dissertation), Xi'an Jiaotong University, Xi'an, China 2010. [71] W. Wagner, A. Pruss, J. Phys. Chem. Ref. Data 22 (3) (1993) 783.
[29] M.K. Dobson, J.C. Chato, J. Heat Transf. 120 (1) (1998) 193. [72] L. Yang, Y. Liu, Y. Yang, S. Shen, Desalin. Water Treat. 64-71 (2016).
[30] Y. Fujita, M. Tsutsui, Heat transfer, in: G.F. Hewitt (Ed.), Proceedings of the Tenth In- [73] G. Kocamustafaogullari, I.Y. Chen, AICHE J. 34 (9) (1988) 1539.
ternational Heat Transfer Conference, 10th ed., IChemE Symposium Series, vol. 1, [74] H. Hou, Q. Bi, H. Ma, G. Wu, Desalination 285 (2012) 393, http://dx.doi.org/10.1016/
CRC Press, Rugby, Warwickshire, 1994. j.desal.2011.10.020.
[31] J.R. Thome, Wolverine Heat Transfer Engineering Data Book - Chapter 13 Two-Phase [75] L. Xu, S. Wang, Y. Wang, Y. Ling, Desalination 156 (1–3) (2003) 101.
Pressure Drops, 2006. [76] S. Shen, G. Liang, Y. Guo, R. Liu, X. Mu, Desalin. Water Treat. 51 (4–6) (2013) 830.
[32] J.R. Thome, Wolverine Heat Transfer Engineering Data Book - Chapter 8 Condensa- [77] L. Xu, M. Ge, S. Wang, Y. Wang, Desalination 166 (2004) 223.
tion Inside Tubes, 2006. [78] H.T. El-Dessouky, H.M. Ettouney, Fundamentals of Salt Water Desalination, 1st ed.
[33] S. Sideman, D. Moalem, R. Semiat, Desalination 17 (2) (1975) 167. Elsevier, Amsterdam, New York, 2002.
[34] M.-C. Chyu, A.E. Bergles, J. Heat Transf. 109 (1987) 983.
[35] A. Ophir, F. Lokiec, Desalination 182 (1–3) (2005) 187.

Please cite this article as: F. Wunder, et al., Numerical simulation of heat transfer in a horizontal falling film evaporator of multiple-effect
distillation, Desalination (2016), http://dx.doi.org/10.1016/j.desal.2016.09.020

You might also like