1992 (Boudinar)

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Desalination, 86 (1992) 273-290 273

Elsevier Science Publishers B.V.. Amsterdam

Numerical Simulation and Optimisation of


Spiral-Wound Modules

M. BEN BOUDINAR, W.T. HANBURY and S. AVLONITIS


Mechanical Engineering Department, Glasgow University, Glasgow G12 gQQ (Scotland)
(Received August 12, 1991; in revised form February 9, 1992)

SUMMARY

The performance of spiral-wound (SPW) modules can be improved by


optimising some key geometrical parameters for given operating conditions.
To this effect, a computer simulation program, which takes into account the
spiral geometry of the module as well as the main physical phenomena
occurring inside SPW modules, was developed. The differential equations
for the transport involved have been solved numerically using finite differ-
ences methods. The resulting computer program enables concentrations,
pressures and flow rates in the brine and permeate channels to be obtained
at any point in the module.
The investigation covered a wide range of feed conditions by using
experimental data provided from two different types of SPW modules. These
were the ROGA-4160HR and Filmtec FT30SW2.5” modules. The former
type dealt with data typical of brackish water desalination whereas the
second type provided data typically encountered in seawater desalination.
The predictions, based on the spiral model, agree very well for both
modules with the experimental data. Very good agreement was obtained for
the ROGA-4160HR module with deviations from experimental data of less
than 5% with respect to permeate flow and less than 7% with respect to
permeate concentration.
For the FT30 module, the results were somewhat less accurate with
maximum deviations of about 10% and 15% for the permeate flow and
concentration, respectively. Based on some preliminary experimental

001 l-9164/92/$05.00 0 1992 Elsevier Science Publishers B.V. All rights reserved.
274

evidence, it was suggested that a better prediction could be achieved if a


more accurate correlation for the mass transfer coefficient could be found.
As an illustrative example of the application of such model, the effects of
some geometrical variables on the ROGA module performance were
investigated with a view to its optimisation.

SYMBOLS

- total membrane area (m2)


AM
b - osmotic coefficient (bar)
C - concentration (ppm)
4 - diffusion coefficient (cm2/s)
E - friction parameter ( 1/cm2)
h - height (cm)
Jl - water flux (cm/s)
- solute flux (cm/s)
J2
K - brine spacer mixing efficiency (cf. Eqn. 1)
- water permeability coefficient (cm/s-bar)
kl
- salt permeability (cm/s)
2 - mass transfer coefficient (cm/s)
L - membrane length (cm)
M - brine spacer characteristic length (cf. Eqn. 1) (cm)
NLE - number of leaves
P - pressure (bar)
Pe - Peclet number (=vh/D )
PD - packing density (m2/m’)
Q - flow (cm3/s)
SC - Schmidt number ( =p/pDs)
T - temperature (“C)
VMOD - module volume (cm3)
v,v - velocity (cm/s)
W - membrane width (cm)
- brine spacer width (cm)
- membrane leaf glue line width (cm)

Greek letters

P - viscosity (g/cm-s)
i- - volume specific productivity (cm3s-‘/cm3.bar)
275

II - osmotic pressure (bar)


P - density (gcms3)
e - incremental angle

Subscripts
B - brine
BW - brine wall
M - membrane
P - permeate
PW - permeate wall

INTRODUCTION

Nowadays, reverse osmosis (RO) is a widely recognised separation


process which is being extensively used to provide potable water from
brackish and seawater. This success is largely due to the development of a
variety of membrane materials and module configurations. Among the
different module configurations used for RO, the form of a spiral wound
(SPW) is regarded as one of the most advantageous. Due to its large surface-
to-volume ratio, a SPW module has a high productivity per unit volume.
Besides this, SPW modules offer a good balance between fouling resistance
and ease of operation.
The potential market application of RO SPW modules has recently
motivated interest in the development of accurate design methods aiming at
improving the hydrodynamic and optimal arrangement of the module. As a
result of this research activity, analytical and numerical methods have been
presented. The former category includes, among others, the work of Rao
and Sirkar [l] who developed an explicit procedure to calculate the permeate
flux and concentration in the case of high rejecting membranes and of
Evangelista [2] and of Gupta [3] who derived models valid for looser
membranes.
Although analytical methods are very useful for fast estimation of the
process efficiency (i.e., productivity and product quality), they are, howev-
er, less suitable for module development and optimisation due to the many
approximations and assumptions required to make the problem mathemati-
cally more tractable. In this context, numerical techniques are more appro-
priate. Recently, a number of numerical modelling approaches such as those
of Evangelista [4], Rautenbach [5] and Chiolle [6] have been published. A
276

common particularity of these mathematical models termed as “slit” (SL)


model was to consider the SPW module as unwounded, i.e., the module was
assumed to consist of two flat spacer filled channels. According to their
authors, the “Slit Model”, which has been tested with available data, seems
to be valid for moderate feed concentrations. Indeed, they tested their results
with experimental data obtained from very dilute feed concentrations not
exceeding 2000 ppm. Although these data have, in practice, provided a
useful comparison for brackish water situations, they are, however, not
suitable for ascertaining the validity of correlations describing important
phenomena such as concentration polarisation.
In this work, an attempt is made to alleviate the possible limitations of the
SL mode1 by adopting a more realistic model which takes into account the
spiral geometry. The resulting computer program used a Finite Difference
Method (FDM) to solve the set of differential equations. The main objectives
of this work are:
(1) To compare the results of the spiral FDM program with experimental
data.
(2) To evaluate the influence of various module parameters on perfor-
mance and optimisation.

MATHEMATICAL MODEL

The mathematical treatment of the fluid dynamics in a SPW module is


very complicated because the velocities, pressures and concentrations
profiles in the brine and permeate channels are three dimensional. In
addition, the coupling between feed and permeate by the flow through the
membrane results in a further complication. Therefore, in order to simplify
the mathematical resolution, the following assumptions have been made:
(1) Negligible component of the brine and permeate flow velocities in the
radial and axial directions, respectively;
(2) The volumetric flow through the membrane is mainly made from the
solvent flux;
(3) Constant fluid density;
(4) For the calculation of the mass-transfer coefficient in channels
equipped with a spacer, the following relationship, which is in agreement
with experiments, Chiolle [6] has been assumed:
277

where K=OS and M=0.6 (cm).


(5) Validity of the Darcy’s Law concerning the pressure drop in porous
media.
(6) The pressure drop along the central permeate collector tube is
neglected (i.e., pressure considered as atmospheric).

An accurate modelling of the SPW modules must be based on the


momentum and material balances formulated in differential form for the feed
side and for the permeate side. At this stage, it should be noted that a further
assumption, which should not affect the results, has been introduced:
although the geometry is spiral, the differential equations governing the
phenomena have been taken in Cartesian coordinates. In other words, the
curvature of the permeate channel has been neglected. This assumption is
justified because the ratio of the channel height to mean module diameter is
small.
From the above assumptions, the following differential balances are
formulated. Starting with the permeate channel, we have:
*Darcy’s Law:
qkY)
= -p E, V&Y) (2)
6X

*Solvent material balance:


6 vp (GY) J, ,(%Y) + J&Y)
(3)
6x = h,

*Solute material balance:


6 C&,Y>
=

KJ:x,YI &&Y) +J&Y)&&9Y)) - q~9Y)v,,(~~Y) +J&,Y))l (4)


V&Y> h,

Similar expressions for the brine channel are obtained as follows:


*Darcy’s Law:
&&Y)
= -cc EBV&Y)
6Y
278

*Solvent material balance:


6 V&Y> Vi ikY> + J&Y)1
(6)
6y =- hB

*Solute material balances:


6 C&Y> 1
6Y = V&Y> h, (7)

[- (J&Y) C,,(X,Y) + J&Y) C,,kYN + C*kY)U&,Y) +J&,Y)>]

The initial and boundary conditions applicable in this case are as follows:
*PermeateJlow:

V&Y> = 0 for x=0 and 01y IL (8)

q(-GY)
for x=0 and OlylL (9)
6x =

Pp(-GY) = Pat, for x=W and OlylL (10)

q&Y) =
for x=0 and OlylL (11)
o

6X

*Brine flow:

vBtx,y> = vF for y=O and OlxlW (12)

&tx,Y) = c, for y=O and Olx<W (13)

PF(X,Y)= pF for y=O and 05x5 W (14)

In addition to the above equations, the analysis is based on the solution-


diffusion model which is widely used in RO literature. This model gives the
following relationships, valid at any point in the system for the solvent and
solute fluxes, respectively:

J, = k,[(P,-P,) - @,+‘>I (1%


279

J2 = 4 <CBW-
c,w) (16)

Assuming a relationship between osmotic pressure and concentration of the


form

ni = bCi (17)

where b is function of temperature and concentration, i.e., b=f(C,,Ti), and


obtained through a combination of a correlation giving the variation of the
osmotic pressure with both concentration and temperature (see Eqns. (26)
and (27)).

Eqn. (15) becomes:

J, = k, K& - pp>- b <&r Cpw)I (18)

Since the concentration of the permeate is determined by the relative


amounts of J, and J2 fluxes, Cp may be evaluated through the following
formula:

J2 UC,,- CPW)
c, = -z-=J2 (19)
Jl + J2 Jl Jl

For the concentration polarisation relationship, the stagnant film model


proposed in Sourirajan [7] is adopted:

J, = kin
[
c,w-CPW
‘B - ‘F’W
1 (20)

The set of equations, i.e., balances combined with the transport equa-
tions, are solved numerically using a finite difference method (FDM) which
requires, as shown in Fig. 1, a discretisation of the model in finite elements
along the brine flow as well as along the permeate flow.
The basic procedure used in the program consists of dividing the SPW
module in increments axially (Fig. la) and in treating each increment as an
individual element. Starting with the first increment and using the initial
conditions (i.e., Eqns. (12-14)), an initial guess for the value of the perme-
ate pressure at the closed end of the permeate channel is performed using an
analytical solution (cf. [9]). Then pressures, concentrations, fluxes and
280

i-th increments

A AY
Sk
1 2 3

Conditions

a) Axial discrctisalion along SPW module lcnglh.

Permeate channel

b) Radial discrclisalion along SI’W module widlh.


( Only a parlial view ol A-A cross section is showed).

Fig. 1. Physical model for the computer program.

velocities are calculated at any point of the mesh grid by an iterative


technique. The iteration is stopped when the conditions at the exit are
satisfied (permeate pressure is atmospheric). The values of the brine
conditions obtained will be used as initial conditions for the next increment;
and the procedure is repeated. Calculations are stopped when the reject end
of the module is reached.

RESULTS AND DISCUSSION

The accuracy of the proposed mathematical model has been assessed by


a comparison with the experiments of Taniguchi [7] (ROGA 4160HR) and
of Avlonitis [lo] (FT30). These data covered the desalination of brackish
281

and seawater, respectively. Table I shows the module geometrical character-


istics used in the calculations. Characteristic parameters for the membranes
and spacers are reported in Table II. These data have been collected and
worked out from different references specified at the bottom of each table.

TABLE I

Geometrical data of the modules

4” ROGA 4160-HRa 2.5” FT30b

3 1
W (cm) 143 110
L (cm) 88 85.4
hM (cm) 0.01 0.014
h, (cm> 0.07 0.077
h, (cm) 0.03 0.041
W,, (cm> - 133

aTaken from [7]. bTaken from [9].

TABLE II

Membranes and spacers characteristics

4” ROGA 4160-HRa 2.5” Fr30b

k, x ld 2.085 Wn. (25)


b x 16 1.444 Wn. (27)
EP 744,444 1,200,ooo
EB 183,673 25,008

‘Taken from [4]. bTaken from [9].

In addition, considering the wide range of the FT30 operating conditions


data, the variation of the membrane and fluid properties should be taken into
account. The FT30 membrane properties variation has been evaluated using
a small rectangular test cell described in Ben Boudinar [9]. The resulting
correlations for the water and the salt permeability were respectively as
follows:
282

Water permeability

-1.701 x 10_3P,
k, = k,, x lo-’ e (21)

where

k,, = 2.6719 + 1.801 x 1O-2 T+ 2.402 x 1O-3 T2 (22)

and salt permeability

h = 1.112 x 10-6~4.983xlO-~T (23)

Fluid properties such as viscosity, osmotic pressure and diffusivity were


correlated in terms of temperatures and/or concentrations using the data
presented in [ 111. The viscosity correlation was of the following form:

-2.008x 10-2T (24)


P = Poe [g/cm 51

where

p. = 1.4757 x 1O-2 + 2.4817 x lo-‘C + 9.3287 x lo-14C2 (25)

with

20 [“Cl 5 T I 45 [“Cl and 0 [ppm] ICI lo5 [ppm]

For the variation of the osmotic pressure, the data were given at 25°C as
below:

nT
= 0.23745 + 6.4784 x 1O-4C + 1.7753 x 10-9C2 (26)
2s
with: 0 [ppm] ICI lo5 [ppm].

Using Eqn. (26), the osmotic pressure at any temperature is given by:

‘T. ‘Tz
(27)
+ = T25
283

It should be noted that in Eqn. (27) the temperature is expressed in degrees


Kelvin.
The diffusion coefficient was correlated in terms of the temperature only:

D, = (0.72598 + 2.3087 x 10S2T+ 2.7657 x 10v4 T2) x 10m5 [cm2/s] (28)

with: 10 [“Cl 5 Tr45 [“Cl.


The comparison of the predicted results with the experimental data for the
ROGA module is listed in Table AI (Appendix A). Judging from this table,
it is seen that the agreement is very good, establishing thus confidence in the
procedure adopted.
A similar comparison has been carried out for the FT30 experiments.
These are reported in Tables AII, AI11 and AIV (Appendix A). As can be
seen, the predictions are, except for some rare cases, very satisfactory. After
a thorough investigation (see [9]), it was found that the most likely source
of error is due to the use of the established correlation for the mass transfer
coefficient (Eqn. (1)). This conclusion was reached after initiating an
experimental programme aimed primarily at assessing the influence of
various variables on the mass transfer coefficient. Based on our early
findings, these would suggest that, although the brine velocity plays a major
role in limiting the thickness of the mass transfer boundary layer at the
membrane surface (as considered in Eqn. (l)), other factors such as applied
pressure, flux and concentration should be taken into account. However, up
to now, we did not find sufficiently convincing mathematical correlation
describing specifically the concentration polarisation phenomena at the
severe feed conditions. Therefore, while still trying to get further under-
standing of this rather complicated phenomena, the use of Eqn. (1) appears
to give reasonably satisfactory results considering the wide range of experi-
mental data used.

APPLICATION OF THE MATHEMATICAL MODEL

The spiral model can be used to predict module performance (as seen in
the previous section) as well as for module optimisation. In the following,
as an illustrative example, we will be concerned with the geometrical
optimisation of the ROGA module at one particular set of operating condi-
tions. This restriction is due to the lack of correlations for the quantities
water and salt permeabilities with operating variables such as pressure,
284

temperature and feed concentrations. It is the authors’ opinion that any


attempt to investigate the effects of highly differing operating conditions on
optimal geometrical characteristics at constant values of k, and J+ should be
treated with caution.
In order to apply an optimisation procedure, it is first necessary to define
an objective function, then to identify the variables that affect it and finally
to find an equation relating the objective function to the variables. At this
stage it should be mentioned that, due to the scarcity of information on
module manufacturing and running energy costs, an economic optimisation
was not attempted. There is, however, no reason why, given the appropriate
data, an economic optimisation could not be made. Thus, as an illustrative
exercise, we have chosen to maximise the permeate flow per unit volume
and driving force.
In all cases, the volume specific productivity of the module (f), as defined
by Eqn. (29)), will be chosen as a target function.

In the following, unless otherwise stated, the geometrical data listed in


Table I were used for the generation of each plot.
Fig. 2 shows the variation of the module productivity with the number of
leaves when both the volume and the membrane area of the module are kept
constant (because of this, similar trends should be obtained when considering
the variation of the specific volume productivity). Also included in this plot
is the effect of varying the width of the leaf glue line. It is seen that in the
hypothetical case where the glue line width is negligible (i.e., W&=0), the
optimal leaf number, NUoP, is found in infinity. However, with a significant
glue line (as it is necessary in practice for protection against possible
leakage), the optimal leaf number exists in the range of M!X0P,=3-5,
depending on the glue line width. This behaviour is due to the fact that when
the number of leaves is less than NLEop,, the clear loss of performance is
caused by the permeate pressure drop resulting from the excessive width of
the permeate channel. On the other hand, when the number of leaves is
greater than NLE,,,, then the decrease in performance is attributed to the
reduced effective area of the membrane caused by the glue line.
Fig. 3 shows the influence of the length-to-width ratio on the volume
specific productivity with the number of leaves as a parameter. This plot
slows clearly that the beneficial effect of increasing the number of leaves
becomes less apparent for a number of leaves NLE> 4. Regarding the
285

50 -

45 -
- Wg = 0.0 [cm]
‘v) ------* Wg P 4.0 [cm]
.
----- Wg-E.O[cm]
OE 40-
s

a”
35 -
I’,= 35.6[bar] ; C,= 1940[ppm] ;

30;
I , 1 , , , , ,
Q,= 517[cc/s] ; A,= 7.55 [m*]
, , , , (
012345676 9 10 11 12

NLE : Number of Leaves

Fig. 2. Module productivity as a function of number of leaves with glue line width as
parameter (module volume is kept constant).

P,= 35.6 [bar] ; C,= 1940 [ppm]

- 1 LEAF
- *LEAVES
----- 3LEAVE.S
. . . . . I..., 4_VES
-.-.__._ .z$mES

- GLEAVES

1.5: . I . I * I * I . I
0 1 2 3 4 5
LPN

Fig. 3. Influence of membrane leaf geometry on specific module productivity.

optimal value of the length to width ratio, (L/W),, it is seen that this latter
tends to increase slightly with the increase in number of leaves. At the
optimal number of leaves NLJ&,=3-4, the corresponding (Lm,P, was
found to be of about 0.8.
All curves show, to some extent, a similar pattern highlighting a strong
influence of the (L./W) value on the volume specific productivity. This
286

behaviour can be explained by considering the pressure drop in both


channels:

l When(L/w) < WV, t, the sharp decrease in the value of f is due to the
P
considerable effect o the permeate pressure drop.
l When (L/w) > (L/w),,, the dependence of l on (L/W) is relatively less
obvious showing that the feed side pressure losses have less influence
than the permeate side pressure losses. Remaining in this region, it is
seen that as the number of leaves increases, the above-mentioned depen-
dence tends to decrease. This is due to the decrease in the brine velocity
(and consequently a decrease in the brine pressure drop) due to the fact
that the feed flow was maintained constant throughout.

Fig. 4 shows the volume specific productivity vs. the channel height of
the permeate channel for different membrane widths. The plot was per-
formed at a number of leaves, NLE=3, because it was shown previously
(Fig. 3) that this number was optimal.
All curves have a maximum which shifts towards thicker permeate
channels and lower volume specific productivity as the membrane width
increases. This behavior is due to the pressure drop in the permeate channel
which, in turn, influences the effective driving pressure. The permeate
pressure drop decreases with an increase in the permeate channel height but
increases with the membrane width. It is clearly shown that the high volume
specific productivity is obtained at the lower membrane width. For the
particular membrane width of the ROGA module (i.e., W= 143 cm), the
computed optimal permeate channel height is approximately 0.013 cm. This
value represents about half the actual permeate channel height (hP=
0.03 cm), presumably because other factors such as membrane support
should be considered.
The influence of the brine channel height on the volume specific produc-
tivity is shown in Fig. 5. Optimal thicknesses of the brine channel have been
evaluated for the conditions stated on the plot and for different module
lengths.
All the resulting curves reveal a sharp defined optimum which tends to
decrease with increase in module length. This is due to the pressure drop
which brings about a reduction of the module productivity. Consequently,
this would suggest that thicker brine channels must be used with an increase
in module length. According to Fig. 5, the optimum brine channel height for
the particular length of the ROGA module (i.e., L=88 cm) should be
approximately 0.01 cm. This value is much lower than the real one because
in actual situations modules are used up to a number of six in series in a
287

4 P,= 35.6[bar] ; C,= 1940[ppm] ;

vF= 16.2[cmk] ; 3 Leaves


1

W- 143 [cm]
W= 214.5 [cm]
WE 288 [cm]

0.02 0.04 0.06 0.08 0.10

h, [cm1

Fig. 4. Volume specific productivity as a function of height and width of the permeate channel.

P,= 35.6[bar] ; C,= 1940[ppm] ;


vF= 16.2 [cm/s] ; 3 Leaves

----- L= 132 [cm]


------* L= 176 [cm]

0.02 0.04 0.06 0.08 0.10

h, Ieml

Fig. 5. Volume specific productivity as a function of height and length of the brine channel.

single pressure vessel. Therefore, the optimum brine channel thickness


should be based on the total length of the modules contained in the pressure
vessel. In this context, it is clearly seen in Fig. 5 that the optimum brine
channel thickness increases as the length of the module increases. Further,
at this stage it should be noted that in practice, maximising 1:is not the only
aspect that has to be considered when designing a SPW module; other
factors, such as fouling and plugging tendencies which are likely to be
generated by very thin spacers are also important.
288

CONCLUSIONS

In view of the above analysis and observations, the following conclusions


were drawn:
(1) A new mathematical model, taking into account the main parameters
affecting the performance of a SPW module, was developed.
(2) The predictions of this mathematical model, when incorporated in an
FDM program, revealed that excellent agreement for the ROGA data were
obtained. For the FT30 data, the predictions, although very acceptable in
many cases, were less satisfactory when compared to those of the ROGA
module. It was particularly stressed that this relative weakness did not lie in
the formulation of the mathematical model but is more likely to be attribut-
able to the formula used in order to evaluate the mass transfer coefficient.
It is concluded that although Eqn. (1) seems to “work” quite well at the
moderate operating conditions (ROGA data), its applicability at the more
severe conditions (seawater, FT30) may still be doubtful. Therefore, in
order to address this problem, an improved mass transfer correlation is
required.
(3) An illustrative example of module optimisation was given, using one
set of operating conditions for the ROGA module. The dependence of the
volume specific productivity on geometrical characteristics was investigated.
Analyses of this type are suggested as being useful in the optimisation of the
module characteristics and operating conditions.

REFERENCES

1 K.K. Sirkar, P.T. Dang and G.H. Rao, Ind. Eng. Chem. Process Des. Dev., 21 (1982) 517.
2 F. Evangelism, Chem. Eng. J., 38 (1988) 33.
3 S.K. Gupta, Ind. Eng. Chem. Process Des. Dev., 24 (1985) 1240.
4 F. Evangelista and G. Jonsson, Chem. Eng. Comm., 72 (1988) 69.
5 R. Rautenbach and W. Dahm, Desalination, 65 (1987) 259.
6 A. Chiolle, G. Gianotti, M. Gramondo and G. Parrini, Desalination, 26 (1978) 3.
7 Y. Taniguchi, Desalination, 25 (1978) 71.
8 S. Sourirajan, Reverse Osmosis, Logos Press, London, 1970, pp. 180-189.
9 M. Ben Boudinar, Performance Prediction and Optimisation of Spiral Wound Modules, Ph.D.
Thesis, Glasgow University, 1991.
10 S. Avlonitis, Investigation and Prediction of Spiral Wound Reverse Osmosis Membrane
Performance, Ph.D. Thesis, Glasgow University, 1991.
11 U.S. Dept. of the Interior, Office of Saline Water Research and Development, Progress Report
No. 363, September, 1968.
APPENDIX A: TABULATED SAMPLE OF PERFORMANCE PREDICTIONS

TABLE AI TABLE AI1

Comparison between experimental* and predicted results for Comparison between experimental and predicted results for
the ROGA module at T=25(“C) the FT30 module at C,=25,000 @pm “artificial” seawater)

PF QF ‘F @Pm Q,, QpwM CP~ cpmM T PF QF QPEX QPNUM CPEX CPhVM


(bar) (cc/s) NaCl) (cc/s) (cc/s) @pm) @pm) (“C) 0-W (cc/s) (cc/d (cc/s) @pm) @pm)

35.6 517 1940 46.50 47.14 55.2 52.7 20 50 217.72 17.10 16.70 92 107
34.9 390 1895 47.50 46.45 57.2 54.2 20 55 219.83 19.33 19.15 89 97
34.5 265 1953 47.00 46.02 57.7 59.9 20 60 222.08 21.58 21.55 86 89
33.9 134 1899 46.10 45.09 71.7 71.5 20 70 226.32 25.82 26.21 79 78
33.9 104 1887 45.70 44.77 84.0 82.2 20 80 230.53 30.03 30.69 72 71
29.0 528 1982 35.70 37.39 63.4 64.6 25 50 220.75 20.25 18.86 98 123
27.9 375 1924 37.60 36.23 66.7 66.6 25 55 223.32 22.82 21.64 95 111
27.8 250 1973 37.70 36.27 67.6 72.1 25 60 225.80 25.30 24.37 89 102
27.2 122 1895 36.90 35.47 82.4 82.5 25 70 231.05 30.75 29.67 82 90
27.2 90 1879 37.00 35.28 92.3 92.1 25 80 235.43 34.95 34.75 72 82
34.8 399 pw** 49.20 49.30 - - 30 50 224.03 23.53 21.31 118 140
28.0 379 PW** 39.80 39.30 - - 30 55 226.73 26.23 24.48 108 127
30 60 230.10 29.60 27.58 100 117
*Data from [7]. **Pure water experiments. 35 50 227.17 26.67 24.01 129 162
35 55 230.73 30.23 27.61 121 147
TABLE AILI TABLE AN

Comparison of experimental and predicted results for the Comparison of experimental and predicted results for the
FT’30 module at C,=35,000 @pm “artificial” seawater) FT30 module at C,=40,000 @pm “artificial” seawater)

T pF QF QPEX QPNUM CPEX CPNUM PF QF QPEX QPNUMCPEX CPNVM


(“Cl (bar) (cc/s) (cc/s) (cc/s) @pm) @pm) (bar) (cc/s) (cc/s) @c/s) @pm) @pm)

20 50 212.48 11.99 11.97 211 197 20 50 210.02 09.52 09.77 275 265
20 55 214.98 14.48 14.28 178 170 20 55 212.10 11.60 12.01 228 223
20 60 217.08 16.58 16.54 158 151 20 70 218.10 17.60 18.48 158 160
20 70 220.95 20.45 20.93 137 128 20 80 221.97 21.47 22.58 135 139
20 80 223.73 23.23 25.13 129 113 25 50 211.10 10.60 10.82 332 308
25 50 213.88 13.38 13.35 248 226 25 55 213.68 13.12 13.35 277 259
25 55 216.68 16.19 15.97 207 196 25 60 216.08 15.58 15.84 232 225
25 60 218.82 18.32 18.53 179 175 25 70 220.23 19.73 20.66 182 185
25 70 223.32 22.82 23.49 141 147 25 80 223.70 23.20 25.26 150 160
25 80 227.27 26.77 28.24 129 131 30 50 212.00 11.50 11.98 372 358
30 50 215.48 14.98 14.90 279 261 30 55 215.38 14.88 14.84 330 300
30 55 219.02 18.52 17.87 238 226 30 60 217.50 17.00 17.64 276 261
30 60 221.05 20.55 20.75 220 202 30 70 222.70 22.20 23.08 220 214
30 70 226.22 25.72 26.36 178 170 30 80 227.90 27.40 28.30 189 186
30 80 230.85 30.35 31.73 158 151 35 50 213.98 13.48 13.23 382 417
35 50 216.98 16.48 16.57 322 303 35 55 216.78 16.28 16.44 378 349
35 55 219.93 19.43 19.91 277 262 35 60 219.83 19.33 19.60 356 304
35 60 223.42 22.92 23.18 243 234 35 70 225.38 24.88 25.71 273 249
35 80 231.05 30.55 31.56 222 217

You might also like