Vous êtes sur la page 1sur 12

Desalination 238 (2009) 128–139

A simplified simulation model of RO systems for


seawater desalination
Hyun-Je Oh, Tae-Mun Hwang, Sangho Lee*
Korea Institute of Construction Technology, Gyeonggi-Do, 411-712, Republic of Korea
Tel. +82 (31) 910-0320; Fax: +82 (31) 910-0291; email: s-lee@kict.re.kr

Received 15 November 2007; Accepted 31 January 2008

Abstract
Desalination of seawater has been considered as one of the most promising techniques for supplying fresh water
in the regions suffering water scarcity. Reverse osmosis (RO) is one of the major technologies for mid- and large-size
desalination plants because it offers a means of producing high quality of water from seawater with lower energy
consumption than other processes such as evaporation processes. In this study, RO systems for seawater desalination
were theoretically investigated to provide insight into the optimum process design. A simple model based on the
solution–diffusion theory and multiple fouling mechanisms was developed and used to analyze the performance of
RO systems. The effect of recovery ratio and permeate flux on the efficiency of the whole RO system was
investigated for a wide range of operating conditions. The model was also applied to optimize the design of RO
process for low energy requirement and high boron removal.

Keywords: Desalination; Seawater desalination; Reverse osmosis; Fouling mechanism; Process simulation;
Optimization

1. Introduction
as quality deterioration [1,2]. High pressure
Seawater desalination has been gaining reverse osmosis (RO) processes have been the
popularity as a feasible option for potable water technology of choice for seawater desalination in
production, as available water sources are the US and many other countries in the world
gradually depleting due to water scarcity as well [3,4]. The market share of RO desalination was
43% in 2004 and is forecasted to increase up to
*Corresponding author. 61% in 2015 [5]. This is because RO has many

Presented at the First International Workshop between the Center for the Seawater Desalination Plant and the European
Desalination Society, November 15–16, 2007, GIST, Gwangju, Korea.

0011-9164/09/$– See front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.desal.2008.01.043
H.-J. Oh et al. / Desalination 238 (2009) 128–139 129

advantages including low energy requirements, theory to predict RO performance and optimize
low operating temperature, small footprint, modu- energy requirement as well as permeate quality.
lar design, and low water production costs. The following assumptions were used for the
However, the performance of RO plants is model derivation:
quite sensitive to the quality of the feed water and C The solution–diffusion model is valid for the
plant operating conditions. This means that the transport of water and solute through the
availability of reliable RO models is of great membrane [Eqs. (1) and (2)].
importance for process design and operation [6, C An RO membrane module is made up of flat
7]. Unfortunately, it is difficult to obtain a rigor- channels with spacers.
ous mechanistic model of RO process, which C Diffusion coefficient is independent of solute
accounts for several important operating factors concentration.
such as permeate recovery, flux, feed tempera- C The brine concentration varies linearly along
ture, concentration polarization, and fouling [8]. an RO element [Eq. (9)].
Although the membrane makers have developed C The thin film theory is applicable for
several softwares to help possible customers to calculating concentration polarization effect
design an RO plant, they mainly focus on the [Eq. (8)].
performance analysis of some RO modules rather C Pressure drop in permeate side is neglected
than the optimization of RO process in terms of [Eq. (3)].
energy consumption and product water quality. C Osmotic pressure is proportional to the salt
Recently, a few works have focused on the concentration [Eq. (4)].
development of new RO models for the opti- C Mass transfer coefficient is constant for a
mization of membrane modules and desalination given fluid condition [Eq. (10)].
plants [8–11]. Nevertheless, these models are still C The transport constants for solutes except for
limited to consider the long-term performance of boron are same as that for NaCl. In other
RO systems because the impact of membrane words, binary solute system (salt and boron) is
fouling and compaction on RO performance are assumed for simplification.
often ignored. C The permeate flow rate is constant while feed
The main objective of this paper is to develop flow rate changes depending on the recovery.
a computer model for simulating and optimizing C Energy consumption by the high-pressure feed
the RO process for seawater desalination. The pump accounts for most energy use of RO
model can make predictions of any operating and process.
performance parameter of the RO plant regardless
of the type of membranes used. Using the model, Fig. 1 shows the flow in a RO membrane
the effects of various factors including recovery system for seawater desalination. The main com-
ratio, permeate flux, temperature, and fouling ponents of RO system are a pump unit, energy
mechanism were examined on the RO plant per- recovery device (ERD), and RO unit. A pump
formance. Moreover, the optimum operating unit supplies high feed pressure (Pf) and flow rate
conditions were explored to minimize energy (Qf) to RO unit. An energy recovery device
consumption and maximize boron rejection. transfers the energy from the concentrate stream
directly to feed flow to RO unit. An RO unit is a
physically packed group of RO vessels arranged
2. Model development in rows and an RO vessel comprises of 6–8
We have applied the solution–diffusion model membrane modules. Since there is a pressure drop
modified with the concentration polarization (Pd) along a vessel, the concentrate pressure (Pr)
130 H.-J. Oh et al. / Desalination 238 (2009) 128–139

Fig. 1. Schematic diagram of RO desalination process.

is lower than the feed pressure (Pf). The recovery 3. Results and discussion
ratio of RO system (Rec) is defined as the ratio of
3.1. Determination of model parameters
permeate flow rate (Qp) to feed flow rate (Qf).
Table 1 summarizes the model equations for Using the equations in Table 1, the model
RO process simulation. The composition of sea- parameters including Lv,0, Ls,salt, Ls,boron, γ1, and γ2
water listed in Table 2 was used in the modeling. were determined to model the feed pressure and
Two kinds of parameters are required for model rejection under a variety of conditions. The
calculations. The general parameters such as Qp, ROSA 6.1 software developed by Dow Chemical
Am, and gpump were obtained from previous works was used to provide data for calculating these
as listed in Table 3. On the other hand, the parameters. To begin, feed pressure, TDS rejec-
membrane-specific parameters such as Lv,0, Ls,salt, tion, boron rejection, and pressure drop were
Ls,boron, γ1, and γ2 were determined by comparing obtained for using ROSA 6.1 under various con-
model calculations with existing RO data. In this ditions. The FilmTec SW30HR-380 was chosen
paper, we applied ROSA 6.1 software [12] to for this calculation since the information on the
generate RO data under various operating con- geometry of feed channel is available. Next, the
ditions. Using these parameters, the equations in model equations in Table 1 were solved using a
Table 1 were simultaneous solved to predict feed guess for model parameters to compare model
pressure (Pf), specific energy consumption (E), calculations with the ROSA results. Then a non-
boron concentration in permeate (cp,boron), and linear regression process is used to find the values
TDS in permeate. The model calculation was for the model parameters with the goal to match
carried out for each element and the final result the data from the ROSA software.
was given as an averaged form. Fig. 2 compares the pressure and rejection
H.-J. Oh et al. / Desalination 238 (2009) 128–139 131

Table 1
Equations for the RO process simulation

Meaning Equation No. Ref.


Solvent transport J  Lv  Pf  Ploss  (1) [16–18]
Solvent transport parameter 1  A  (2) [8,19]
Lv   1  s 
1
1 ( T  293)
 Rc  Am 
2 Pf
293
Lv ,0 e
n
Ploss    i  Pd
Pressure loss (3) [16]
i

Osmotic pressure  i   cm ,i  c p ,i  RT (4) [8,16]

2 (5) [8]
 ud 
Pressure drop in an element Pd  1  h 
  
Solute transport J s ,i  Jc p ,i  Ls ,i  cm ,i  c p ,i  (6) [16,17,20]
1 ( T  273)
(7) [8]
Solute transport parameter Ls ,i  Ls ,i ,0 e 273

cm ,i  c p ,i (8) [17,20, 1]
J

Concentration polarization e ki

cb,i  c p ,i

Average concentration cm ,i  0.5 cm ,i x 0


 cm ,i x  lm  (9) —

0.17 0.77 (10) [8]


 ud      cb,i   Di 
0.4
Mass transfer coefficient ki  0.5510  h       
    Di      dh 
247.8
Viscosity (11) [22]
  2.414  105  10 T 140
Recovery ratio Rec  Q p Q f (12) —

Pf Q f   pump   PbQb  ERD


1
(13) —
Specific energy E
Qp
nm
(14) —
Qp J n Am
1 nm
Average flux J avg 
At
 n 1
nm

nm
J n

 Am
n 1
n 1
132 H.-J. Oh et al. / Desalination 238 (2009) 128–139

Fig. 2. Determination of model parameters by comparing the calculations with ROSA simulation results. Modeling
conditions: Javg = 12, 14, 18 L/m2-h; Rec = 20, 30, 40, 50, 60%; T = 5, 15, 25, 35EC.

predicted using the model with the ROSA soft- pressure drop in an element, γ1, and γ2, were
ware predictions. The flux and recovery ratio obtained to be 1.10×10!9 bar and 1.38. These
were ranged from 12 to 18 L/m2-h and from 20 to model parameters were used for the remainder of
60%, respectively. The temperature was also this paper.
varied from 5 to 35EC. For the wide range of
conditions that were studied and modeled, the
3.2. Effect of recovery and flux on RO
predictions of our model show very good agree-
performance
ment with ROSA software predictions. Based on
these calculations, Lv,0, Ls,salt, and Ls,boron were esti- The effectiveness of RO filtration depends on
mated to be 5.417×10!12 m2-s/kg, 5.13×10!9 m/s, a wide range of operation parameters including
and 6.54×10!7 m/s. In addition, the constants for recovery and flux. In Fig. 3(a), the variations in
H.-J. Oh et al. / Desalination 238 (2009) 128–139 133

Table 2 (a)
Composition of model seawater [3]

Component Concentration [mg/L]

Sodium, Na+ 10,800


Potassium, K+ 392
Calcium, Ca2+ 411
Magnesium, Mg2+ 1,290
Chloride, Cl- 19,439
Sulfate, SO42- 2,701
Bicarbonate, HCO3- 145
Carbonate, CO3 7.4
Boron, B 5
Silica, SiO2 2.9 (b)
Total dissolved solid 35,217

Table 3
Parameters for the RO process simulation [8,12]

Parameter Value
3
Permeate flow rate, Qp [m /day] 1090
Membrane area per one RO element, 35.3
Am [m2]
Brine channel length of RO element, 0.8665
hm [m]
Brine channel height of RO element, 0.84×10!3 Fig. 3. Dependence of specific energy, feed pressure, and
hm [m] permeate concentration on recovery. Modeling condition:
Brine channel width of RO element, 1.34 Javg = 14 L/m2-h; T = 25E ; gEDR = 85%. (a) Specific
wm [m] energy (") and feed pressure (Δ). (b) Boron (") and TDS
Number of elements in a pressure 6 (Δ) in permeate.
vessel, ne
Constant for solvent transport, α1 8.6464
Constant for solvent transport, α2 [bar!1] 0.0149 not only feed pressure but also feed flow rate [see
Constant for solute transport, β1 14.648 Eq. (13)]. As recovery increases, total energy
Pump efficiency, gpump 0.8 consumption increases to obtain high feed pres-
sure but less water can be pressurized to produce
the required amount of product water. Therefore,
specific energy and feed pressure were shown for the specific energy, which is total energy con-
recovery ratio (Rec) ranging from 15 to 70%. The sumption divided by feed flow rate, has a mini-
feed pressure increases as the recovery ratio mum value at an optimum recovery ratio. Under
increases, indicating that higher pressure is the given conditions, the minimum value for
required to obtain high recovery ratio. However, specific energy is 2.336 kWh/m3 at Rec = 51%. In
the specific energy decreases with increasing other words, this is 2.29 time of osmotic pressure
recovery ratio ranging from 15% to 50% and then (1 kWh/m3 = 36 bar).
increases with increasing recovery ratio over Fig. 3(b) shows the changes in boron con-
50%. This is because the specific energy depends centration and TDS in permeate as a function of
134 H.-J. Oh et al. / Desalination 238 (2009) 128–139

recovery ratio. As brine concentration is high at (a)


high recovery ratio, solute concentrations in
permeate increase with increasing recovery ratio
[see Eq. (6)]. Boron concentration in permeate,
which is crucial in seawater desalination for its
possible harmful effect, increases twice as the
recovery ratio increases from 15% to 70%. This
suggests that recovery ratio should be lowered
below a certain value to ensure high rejection of
solute.
Another critical operating parameter is the
flux, which was held constant in the previous
figures. Since the total permeate flow rate is kept (b)
constant at 1,090 m3/d, increasing flux (Javg) leads
to a decrease in total membrane area. For in-
stance, the total membrane area is 3,177 m2 at
Javg = 14.3 L/m2-h while it is 1,906 m2 at Javg =
23.8 L/m2-h. As shown in Fig. 4(a), specific
energy and feed pressure increases with increas-
ing flux due to the requirement of high feed
pressure [see Eq. (1)]. This implies that low flux
resulted in low specific energy consumption.
Although the energy requirement increases
with increasing flux, the solute rejection is Fig. 4. Dependence of specific energy, feed pressure, and
improved with flux. Both boron concentration permeate concentration on average flux. Modeling con-
and TDS in permeate decreases more than 50% as dition: Rec = 40%; T = 25EC ; gEDR = 85%. (a) Specific
a result of flux increase from 7 to 24 L/m2-h, as energy (G) and feed pressure ("). (b) Boron (G) and TDS
shown in Fig. 4(b). This is because solute flux (") in permeate.
(Js,i) is almost independent of solvent flux (J) for
high-rejection RO membranes. Rearranging varied from 5EC to 25EC for these calculations.
Eq. (6), solute concentration in permeate is esti- Specific energy increases with decreasing tem-
mated from the ratio of Js,i to J and thus it is perature because the solvent transport constant
inversely proportional to J. Accordingly, an opti- (Lv) is reduced at low temperature. On the other
mum range of average flux seems to exist for low hand, boron concentration in permeate decrease
specific energy and high rejection of solute. as temperature decrease, leading to a higher rejec-
tion of boron. Note that both the solvent transport
constant (Lv) in Eq. (2) and the solute transport
3.3. Effect of feed water temperature and energy constant (Ls,i) in Eq. (7) increase as temperature
recovery efficiency increases. Rearranging Eqs. (1) and (6), the solute
In addition to recovery and flux, feed water concentration in permeate is given by:
temperature plays an important role in RO cm ,i
filtration performance. Fig. 5 illustrates how c p ,i  (15)
temperature affects specific energy and boron
concentration in permeate. The temperature was
Lv
Ls ,i
 Pf  Ploss   1
H.-J. Oh et al. / Desalination 238 (2009) 128–139 135

(a) (a)

(b) Fig. 6. Effect of energy recovery efficiency on specific


energy at different recovery ratios. Modeling condition:
Javg = 14 L/m2-h; Rec = 40%; T = 25EC. (": gEDR = 90%;
Δ: gEDR = 80% ; ": gEDR = 60% ; G: gEDR = 0%).

recovery device (gERD) is important in energy con-


sumption of SWRO systems. As shown in Fig. 6,
specific energy without an energy recovery
device (gERD = 0%) is high compared with those
with energy recovery devices. The efficiency of
the energy recovery device is less important at a
high recovery ratio since brine flow rate (Qb) is
low in Eq. (13). For instance, specific energy at
gERD = 90% is less than half of that without
Fig. 5. Effect of feed water temperature on specific
energy recovery when Rec is below 28%. It
energy and boron concentration at different recovery
ratios. Modeling condition: Javg = 14 L/m2-h; Rec = 40%; becomes about 80% of that without energy
gEDR = 85%. (a) Specific energy. (b) Boron in permeate recovery at Rec = 70%.
(G: T = 25EC ; ": T = 15EC ; Δ: T = 5EC).

3.4. Optimization of operating conditions for


Therefore, the temperature dependence of solute
specific energy and boron rejection
rejection would be determined by a trade-off
between temperature dependence of Lv and that of Boron is naturally occurring and present in
Ls,i. The model calculation implies that the tem- seawater at an average concentration of 4.6 mg/L
perature dependence of Ls,i overwhelms that of Lv. [15]. However, if the concentration of boron is
Similar results were reported in an experimental too high, it is harmful to human health and crop
study by Hyung and Kim [3]. yields. Unfortunately, the RO process has not
Because of the relatively high energy require- been very effective in boron removal and the
ments, most SWRO systems are equipped with an rejection of boron by RO membranes has been
energy recovery device that recovers energy from found to be lower than 90% [3]. This implies that
the pressurized RO concentrate leaving the the operating conditions for RO should be care-
system [13,14]. Thus, the efficiency of energy fully controlled to reduce boron passage for a
136 H.-J. Oh et al. / Desalination 238 (2009) 128–139

single-pass SWRO plant. Therefore, SWRO sys-


tems should be optimized to maintain low boron (a)
in permeate as well as low energy consumption.
Fig. 7 shows the contours of specific energy
and boron concentration in permeate as a function
of flux and recovery ratio. Here, the two contours
were overlaid to determine optimum flux and
recovery ratio for a given performance guideline.
If the specific energy and boron concentration
have to be less than 2.5 kWh/m3 and 1 mg/L,
respectively, the flux and recovery ratio should
range from 12.5 to 19 L/m2-h and from 34 to
58%, respectively. The shaded area in Fig. 7(a)
indicates the ranges of operation conditions to
satisfy this performance guideline.
Decreasing temperature from 25EC to 15EC
significantly affects the operating conditions for
flux and recovery ratio. To meet the same guide-
line at 15EC, flux should be less than 14 L/m2-h
(b)
and the recovery should be less than 64%.
Therefore, the shaded area in Fig. 7(b) is quite
different from that in Fig. 7(a). Considering
seasonal variations of seawater temperature in
many countries except for the Middle East region,
it is evident from these results that temperature is
an important factor affecting the optimization of
SWRO systems.

3.5. Effect of fouling mechanism


Since most previous models for RO filtration
do not consider RO membrane fouling, a simple
approach to incorporate the effect of fouling on
specific energy and rejection was attempted in
our model. In Eq. (2), the fouling mechanisms in
the SWRO system are divided into (1) cake
Fig. 7. Contours of specific energy and boron concen-
formation (Rc) and (2) scale formation (As/Am). tration in permeate at different flux rates and recovery
Fouling due to deposition of particles and col- ratios. Modeling condition: gEDR = 85%. (a) T = 25EC.
loids is assumed to be attributed to cake forma- (b) T = 15EC.
tion. Only surface crystallization is considered as
scale formation mechanism, where flux decline
results from the blockage of the membrane Fig. 8 compares the specific energy for two
surface by lateral growth of the impermeable RO systems with different fouling mechanisms.
scale deposit on the membrane. The extent of fouling was given by the ratio of
H.-J. Oh et al. / Desalination 238 (2009) 128–139 137

(a) (a)

(b) (b)

Fig. 9. Effect of membrane fouling mechanism on boron


Fig. 8. Effect of membrane fouling mechanism on
concentration in permeate at different recovery ratios.
specific energy at different recovery ratios. Modeling
Modeling condition: Javg = 14 L/m2-h; T = 25E; gEDR =
condition: Javg = 14 L/m2-h; T = 25E; gEDR = 85%.
85%. (a) Cake formation. (b) Scale formation (G:Lv, f/ Lv
(a) Cake formation. (b) Scale formation (G: Lv, f/ Lv = 1;
= 1; ": Lv, f/ Lv = 0.75; Δ: Lv, f /Lv = 0.5).
": Lv, f/ Lv = 0.75; Δ: Lv, f/ Lv = 0.5).

solvent transport constant with and without foul- cases. As membrane fouling occurs due to cake
ing (Lpf /Lp). As fouling occurs, specific energy formation, the Lv,f decreases but the local flux at
increases due to the requirement of high feed the membrane surface does not change. On the
pressure. Regardless of fouling mechanisms, contrary, membrane fouling due to scale forma-
specific energy is same at same Lv,f /Lv. tion leads to a reduction in effective membrane
On the other hand, the boron rejection is quite area and the local flux at the membrane surface
different even at the same Lv.f /Lv, depending on increases. Since the solute rejection increases
fouling mechanisms. As shown in Fig. 9, the with increasing flux [as shown in Fig. 4(b)],
boron concentration in permeate is almost fouling due to scale formation reduces the boron
constant in the case of cake formation while it is concentration in permeate. These results suggest
quite different in the case of scale formation. This that the fouling mechanism is also important for
is because the local flux is different between two the prediction of SWRO systems.
138 H.-J. Oh et al. / Desalination 238 (2009) 128–139

4. Conclusions cm,i — Solute concentration at membrane


surface, mol/m3
In this work, RO systems for seawater
c̄ m,i — Average solute concentration at mem-
desalination were theoretically investigated using
brane surface, mol/m3
a simple model based on the solution–diffusion
cp,i — Solute concentration at permeate side,
theory and multiple fouling mechanisms. The
mol/m3
following conclusions can be drawn from this
dh — Hydraulic diameter, m
work:
Di — Solute diffusion coefficient, m2/s
C The predictions of our model were compared
E — Specific energy consumption, kWh/
and verified with the predictions of ROSA 6.1
m3
software. Although membrane manufacturers
hm — Brine channel height of an RO
supply software for their own membrane
element, m
products, our model presented in this work
J — Solvent flux [m/s]
can be used in any type of RO membrane with
Javg — Average flux for solvent over an RO
small adjustments for model parameters,
system, m/s
including water and salt transport constants.
Js,i — Solute flux, mol/m2-s
C Recovery ratio has to be optimized for low
ki — Mass transfer coefficient, m/s
energy consumption and high solute rejection.
Lv — Solvent transport parameter, m/s-Pa
Based on our calculation, the minimum spe-
Lv,0 — Intrinsic solvent transport parameter,
cific energy for a SWRO plant with a capacity
m/s-Pa
of 1,090 m3/d is 2.336 kWh/m3 at Rec = 51%,
Lv,f — Solvent transport parameter after
Javg = 14 L/m2-h, and T = 25EC.
membrane fouling, m/s-Pa
C Higher flux improves the solute rejection, but
Ls,i — Solute transport parameter, m/s
increases the specific energy consumption.
Ls,i,0 — Intrinsic solute transport parameter,
Increasing the feed water temperature reduces
m/s
the specific energy, but deteriorates the solute
lm — Brine channel length of an RO ele-
rejection. Energy recovery efficiency is an
ment, m
important factor affecting the specific energy
ne — Number of RO elements in a vessel
consumption, especially at low recovery ratio.
nm — Number of RO elements in an RO
C Using our model, the optimum recovery and
system
flux can be determine for a given condition of
Pd — Pressure drop along a RO system
specific energy and boron concentration in
Ploss — Pressure loss in an RO system, Pa
permeate. Temperature is also an important
Qb — Brine flow rate, m3/s
factor affecting the optimization of SWRO
Qf — Feed flow rate, m3/s]
system.
Qp — Permeate flow rate, m3/s
C Depending on the fouling mechanism, boron
R — Gas constant, J/mol-K
rejection may be different even at the same
REC — Recovery
fouling level.
t — Time, s
T — Temperature, K
u — Crossflow velocity in an RO brine
5. Symbols
channel, m/s
cb,i — Solute concentration at brine side, wm — Brine channel width of an RO ele-
mol/m3 ment, m
H.-J. Oh et al. / Desalination 238 (2009) 128–139 139

Greek [9] D.V. Gauwbergen and J. Baeyens, Modelling and


scale-up of reverse osmosis separation. Desalination,
ΔΠ — Osmotic pressure difference, Pa 139 (2001) 275.
ρ — Density, kg/m3 [10] M.G. Marcovecchio, S.F. Mussati, P.A. Aguirre,
ν — Kinematic viscosity, m2/s N.J., Scenna, Optimization of hybrid desalination
η — Viscosity, Pa-sec processes including multi stage flash and reverse
osmosis systems, Desalination, 182 (2005) 111–122.
[11] R.L. Stover, SWRO process simulator. Desalination,
Acknowledgements 221 (2008) 126–135.
[12] Dow, A guide to ROSA 6.0., Dow Chemical
This research was supported by a grant (code# Company, USA, 2004.
C106A152000106A085600212) from the Plant [13] P. Geisler, W. Krumm and T. Peters, Optimization of
Technology Advancement Program funded by the the energy demand of reverse osmosis with a
Ministry of Construction and Transportation of pressure-exchange system. Desalination, 125 (1999)
the Korean Government. 167–172.
[14] R.L. Stover, Energy recovery devices for seawater
reverse osmosis. EverythingAboutWater, 2006-11
(2006) 40–45.
References
[15] N. Nadav, Boron removal from seawater reverse
[1] M. Wilf and C. Bartels, Optimization of seawater RO osmosis permeate utilizing selective ion exchange
systems design. Desalination, 173 (2005) 1–12. resin. Desalination, 124 (1999) 131–135.
[2] S.A. Avlonitis, K. Kouroumbas and N. Vlachakis, [16] S. Lee and R.M. Lueptow, Membrane rejection of
Energy consumption and membrane replacement cost nitrogen compounds. Environ. Sci. Technol., 35
for seawater RO desalination plants. Desalination, (2001) 3008–3018.
157 (2003) 151–158. [17] M. Cheryan, Ultrafiltration and Microfiltration
[3] H. Hyung and J.H. Kim, A mechanistic study on Handbook, Technomic Publishing, Lancaster, PA,
boron rejection by sea water reverse osmosis mem- 1998.
branes. J. Membr. Sci., 286 (2006) 269–278. [18] S. Nicolas, B. Balannec, F. Beline, and B. Bariou,
[4] S. Atkinson, Japan’s largest sea-water desalination Ultrafiltration and reverse osmosis of small non-
plant uses Nitto Denko membranes. Membr. charged molecules: a comparison study of rejection
Technol., 2005 (2005) 10–11. in a stirred and an unstirred batch cell. J. Membr.
[5] G.W. Intelligence, Desalination markets 2005–2015 Sci., 164 (2000) 141–155.
A Global Assessment and Forecast, Oxford, UK, [19] S. Lee, J. Kim and C.H. Lee, Analysis of CaSO4
Media Analytics, 2004. scale formation mechanism in various nanofiltration
[6] A. Abbas and N. Al-Bastaki, Modeling of an RO modules. J. Membr. Sci., 163 (1999) 63–74.
water desalination unit using neural networks. Chem. [20] L.J. Zeman and A.L. Zydney, Microfiltration and
Eng. J., 114 (2005) 139–143. Ultrafiltration: Principles and Applications, Marcel
[7] M.G. Marcovecchio, P.A. Aguirre and N.J. Scenna, Dekker, New York, 1996.
Global optimal design of reverse osmosis networks [21] S. Kim and E.M.V. Hoek, Modeling concentration
for seawater desalination: modeling and algorithm. polarization in reverse osmosis processes. Desali-
Desalination, 184 (2005) 259–271. nation, 186 (2005) 111–128.
[8] S.A. Avlonitis, M. Pappas and K. Moutesidis, A [22] C.J. Seeton, Viscosity–temperature correlation for
unified model for the detailed investigation of liquids. Tribology Lett., 22 (2006) 67–78.
membrane modules and RO plants performance.
Desalination, 203 (2007) 218–228.