You are on page 1of 13

Computers and Chemical Engineering 30 (2005) 202–214

Pervaporation of hydrazine–water through hollow fiber module:


Modeling and simulation
Nazish Hoda 1 , Satyanarayana V. Suggala 2 , Prashant K. Bhattacharya ∗
Department of Chemical Engineering, Indian Institute of Technology Kanpur, Kanpur 208016, India
Received 18 March 2004; received in revised form 23 July 2005; accepted 29 August 2005
Available online 6 October 2005

Abstract
The mathematical model was formulated to simulate the performance of hollow fiber module for pervaporative separation of binary liquid
mixture of hydrazine/water. Separate equations were derived for feed flow in the tube side as well as on the shell side. The permeate flux through
the membrane was estimated using the detailed solution diffusion model. The solubility of the penetrant in the membrane was obtained from
Flory–Huggins equation. Margulas equation constant for the calculation of activity coefficient was estimated from least square error minimization
of VLE. Further, diffusion coefficient at infinite dilution (concentration) and plasticization coefficients were estimated by non-linear regression
of solution–diffusion model using the experimental flux data, taken from literature, of hydrazine–water through ethyl-cellulose membrane. The
performance equation for the entire module was derived by simplifying the mass, momentum and energy balance on both the feed and the permeate
sides. The obtained set of non-linear ordinary differential equations, coupled with the solution–diffusion model was solved numerically. The
property variation of feed and permeate sides, along the axial direction, was studied. Further, the influences of operating parameters on flux and
selectivity were also studied. It was found that separation was better for feed flow in the tube side compared to feed flow in the shell side. Also, the
shell side flow was restricted by operational constraint due to higher-pressure drop in the permeate side.
© 2005 Elsevier Ltd. All rights reserved.

Keywords: Pervaporation; Hollow fiber module; Hydrazine–water; Modeling; Simulation

1. Introduction produce anhydrous hydrazine is highly energy intensive due to


the formation of maximum boiling azeotrope at 71.5 wt.% of
The separation of hydrazine–water mixture is a problem of hydrazine. Hence, separation of hydrazine–water mixture is of
immense importance. Hydrazine and its derivatives are used for paramount interest. Pervaporation (PV), which overcomes some
a variety of commercial and non-commercial applications. One of the problems associated in conventional processes (Ravindra,
of the major non-commercial applications of the anhydrous Sridhar, & Khan, 1999; Satyanarayana, 2004) can be a potential
form of hydrazine is in the form of rocket fuel, in space alternative. The process may be an economical separation
shuttles and guided missile. Hydrazine solutions are used technique compared to conventional separation methods, such
commercially as blowing agent in agricultural applications, as distillation, especially in separations involving azeotropes
chemical and pharmaceutical industries. Methods (various (Duggal & Thompson, 1986), isomers (Wessling, Werner, &
reaction routes) for the production of hydrazine gives water Hwang, 1991), close boiling mixtures (Matsui & Paul, 2002)
as reaction product (Schmidt, 1984). Removal of water by and volatile organics (Meuleman, Willemsen, Mulder, & Strath-
conventional separation techniques (example distillation in the mann, 2001); apart from established industrial applications
presence of dehydrating agents or azeotropic distillation) to (Sander & Soukup, 1988) of dehydration of alcohols. Due to its
good separation efficiency and flux rates, PV results in savings
∗ Corresponding author. Tel.: +91 512 2597093; fax: +91 512 2590104. in energy costs, besides ensuring safety in operations. The US
E-mail addresses: hoda@cems.umn.edu (N. Hoda), svsatya7@yahoo.com Department of Energy identified Pervaporation membranes for
(S.V. Suggala), pkbhatta@iitk.ac.in (P.K. Bhattacharya).
1 Department of Chemical Engineering, University of Minnesota, MN 55455, organic–organic separations and reverse osmosis oxidation
USA. resistance membranes, as two of their highest ranking research
2 Department of Chemical Engineering, JNTU Ananthapur 515002, AP, India. priorities (Haggins, 1990).

0098-1354/$ – see front matter © 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compchemeng.2005.08.014
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 203

Nomenclature
in inlet
a activity j hydrazine
b plasticization coefficient l liquid phase mixture
CP specific heat (J/mol K) m membrane
dtube inside diameter of the fiber (m) p polymer
D diffusion coefficient (m2 /s) s mixture
D0 diffusion coefficient at infinite dilution (m2 /s) v vapor phase mixture
De effective
 diameter of the  z axial component
{h2 −(Rtube +t)2 }
shell = 4 tube2Rtube π (m)
Superscripts
F mass flow rate (kg/h) F feed side
f friction factor P permeate side
htube tube pitch [= 3(Rtube + t)] (m) n location
J mass flux (kg/m2 s) − average/molar
kL mass transfer coefficient in the boundary layer
(m/s)
L length of the module (m)
Lpot potting length (m) PV is a membrane separation technique, which has elements
M molecular weight (kg/mol) common to reverse osmosis and gas separation. In this process, a
Mc molecular weight between two cross-links liquid feed mixture is circulated in contact with a perm-selective
(kg/mol) non-porous membrane and permeate is evolved in the vapor state
N number of tubes from the opposite side of the barrier which is kept under low
P pressure (Pa) pressure. The commercialization of the PV process will depend
Psat saturation pressure (Pa) upon the appropriate design of module configuration. One of
p downstream pressure (Pa) the major disadvantages in the commercialization of pervapo-
R universal gas constant (J/mol K) rative separation is its lower stage cut. Therefore, it becomes
RHS right-hand side extremely important to have a module with a very high packing
Rshell radius of the shell (m) density, which hollow fiber modules may provide (Crowder &
Rtube radius of the tube (m) Cussler, 1998; Feng & Huang, 1997). Further, higher perme-
Re Reynolds number ation rate can be achieved and the modules may also be used
r radial coordinate for hollow fiber along with distillation columns for the development of hybrid
T temperature (K) PV-distillation process. Such a combination may be economi-
t membrane thickness (m) cally more favorable than the conventional separation scheme
u permeate velocity (m/s) (Bausa & Marquardt, 2000; Eliceche, Daviou, Hoch, & Uribe,
v feed velocity (m/s) 2002). However, hollow fiber module may also pose certain
V molar volume (m3 /mol) disadvantages; particularly, the building-up of the downstream
w weight fraction permeate pressure (Feng & Huang, 1995) which affects flux and
X axial coordinate in solution–diffusion model selectivity.
x mole fraction in liquid phase Hydrazine–water pervaporation work is very scarce and only
y mole fraction in vapor phase two such experimental work are available (Ravindra et al., 1999;
z axial coordinate for hollow fiber Satyanarayana & Bhattacharya, 2004); however, the work does
(0) module entrance not report any modeling of the process. In order to mathe-
matically model the process, the mass transport mechanism
Greek letters should be clearly understood. However, the mass transport of
∆ differential increment mixtures through a polymeric membrane is generally consid-
α selectivity ered to be complex as systems are often highly interactive. The
χ Interaction parameters equations that govern the transport phenomena are highly sys-
φ stage cut tem dependent and there is no universal model to represent
η viscosity (Pa s) all kinds of pervaporation processes. Hence, one of the objec-
ϕ volume fraction in membrane tives of this work is to develop a detailed model, simply based
λ latent heat of vaporization (J/kg mol) on solution–diffusion mechanism, for hydrazine–water system.
ρ density (kg/m3 ) The developed model is an extension of the earlier model (Cao
& Henson, 2002), applicable at high downstream vacuum and
Subscripts
for an average volume fraction of the component in the mem-
b bulk
brane. The present model considered the volume fraction vari-
i water
ation inside the membrane for flux estimation using a discrete
204 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

analogue method (Meuleman, Bosch, Mulder, & Strathmann, The mass transfer across the liquid boundary layer at the
1999). Therefore, developed model may also be applicable for feed side of the membrane can be described by Eq. (1). The
higher downstream pressures (low vacuum). However, the model work has been carried out with the assumption of negligible
developed by Cao and Henson (2002) was used for predicting temperature polarization. Since the product of flux and specific
the performance of a spiral wound module. The present model evaporation enthalpies of the permeating components are rela-
investigates similar performance for a hollow fiber module. tively small, the temperature difference between bulk flow and
The second objective of this work is to study the effect of feed the membrane surface can be neglected for process design cal-
flow on process performance for hydrazine–water separation. culations (Rautenback, Herion, & Meyer-Bluemenroth, 1991).
Model formulations for hollow fiber module were made con- Further, the heat transfer coefficient for the tube side feed flow
sidering mass, momentum and energy balance of both the feed is 287 W/m2 (Seider & Tate, 1936), while for the shell side
and the permeate sides. The simulations were carried out to esti- feed flow it is 9884 W/m2 (Zukauskas, 1972). Accordingly, the
mate the axial variation of properties like downstream pressure, temperature differences across the boundary layer for the two
feed flow rate, retentate temperature, permeate and retentate cases are obtained as 9.7e−3 and 2.7e−4 K, respectively. The
concentrations, etc. Further, a study was made to understand small temperature difference across the thermal boundary layer
the sensitivity of separation parameters to operating conditions. justifies the assumption and hence the effect of temperature
Finally, a comparison between performance for feed flow in the polarization may be considered to be negligible.
tube side and the shell side is discussed.
Ji = kL,i ρsl [wFib − wFiI ] (1)
2. Theories where wFib is the bulk liquid composition and wFiI is the con-
centration at the liquid–membrane interface. The mass transfer
2.1. Flux coefficient kL,i for the hollow fiber module can be estimated
using standard correlations (Cussler, 1998).
Five consecutive steps may describe the transport of solute The interaction parameter between the two liquid com-
from bulk feed side to bulk permeate side during PV through a ponents was calculated combining excess Gibbs free energy
non-porous membrane and are shown schematically in Fig. 1. (Prausnitz, Lichtenthaler, & Azevedo, 1986) with the
Flory–Huggins theory (Flory, 1953). The following relation for
1. Transport of a component from the bulk of the feed to the the binary liquid–liquid interaction parameter was derived and
membrane surface. is given by Eq. (2).
2. Sorption of the component into the membrane.     F  
3. Diffusion of the component in the membrane phase. 1 x F
ib
xjb Ax F xF
ib jb
4. Desorption of the component as vapor on the permeate side χij = F F xib F
ln +xjbF
ln + (2)
xib ϕjb ϕib
F ϕjb
F RT
of the membrane.
5. Transport of the component from the membrane surface to The Margules constant for hydrazine–water system was
the permeate bulk. determined by minimizing the least square error using the
vapor–liquid equilibrium data of hydrazine–water (Wilson,
This general transport mechanism can be applied for each Munger, & Clegg, 1952). The minimization was done in MAT-
component present in the liquid mixture. The permeation LAB using non-linear unconstrained optimization routine fmin-
through the membrane is estimated here, using the general- search. The value of Margules constant was found to be
ized solution diffusion model, i.e. from steps 2–4. The transport −5606.6 J/mol with a L2 Norm of 0.0013. Using this value,
resistance on the permeate side, i.e. step 5, is ignored as the liquid–liquid interaction coefficient was calculated from Eq. (2)
downstream pressure is very low. and the value was found to be −2.265. The high negative value
of interaction coefficient infers strong affinity between water and
hydrazine.
The sorption of the penetrant in the membrane can be pre-
dicted based on Flory–Huggins theory (Flory, 1953). This works
well even for semi-dilute and concentrated solution and has an
advantage over the Henry’s law used in some cases. Eq. (3) relate
the volume fraction and the activity of components in the mem-
brane for ternary system (membrane and two components).
Vi Vi
ln aim = ln φim + [1 − φim ] − φjm − φpm
Vj Vp
Vi
+ [χij φjm + χip φpm ][φjm + φpm ] − χjp φjm φpm
Vj
 
Vi ρp 2Mc  1/3 
+ 1− φpm − 0.5φpm (3)
Fig. 1. Mass transport steps during pervaporation process. Mc M
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 205

where χip is the interaction coefficient between solute and mem- Table 1
brane. The values were found to be 2.5 and 1.8 for water–ethyl Pure component physical property data
cellulose and hydrazine–ethyl cellulose, respectively (Ravindra, Property Water Hydrazine Ethyl cellulose
Sridhar, Khan, & Rao, 2000). Liquid density (kg/m3 ) 996.5 1002.6 1100
According to solution diffusion model, the flux of the com- Liquid viscosity (cP) 0.9 0.9 –
ponent, in absence of flow coupling and external forces, is Gas viscosity (cP) 0.0108 0.009 –
proportional to its chemical potential gradient. Further, if the Latent heat of 2.26 × 106 1.357 × 106 –
variation of pressure inside the membrane is neglected, the flux vaporization (J/kg)
Specific heat (J/kg K) 4200 3092 –
equation can be simplified as: Molecular weight 18 32 89448a
dϕim (g/g mol)
Ji = −ρsl Dim (4) Molecular weight – – 140a
dX between two cross
The diffusion coefficient of low molecular weight component links of polymer
in polymer is concentration dependent. Hence, the concentra- (g/g mol)
tion dependence of diffusion coefficient is taken into account. a Values are taken from ref. Ravindra et al. (1999).
Eq. (5) shows the functional dependence of diffusion coeffi-
cient on concentration. The diffusion coefficient is considered to very low pressure the fugacity coefficient as well as Poynting
vary exponentially with volume fraction of components (Brun, factor is equal to one. Hence, the activity may be given by:
Larchet, Bulvestre, & Auclair, 1985). The cross dependence is
chosen such that it accounts for the strong coupling effect in case p
aib
P
= aiIP = yib
P
(7)
of hydrazine–water mixture. The temperature dependence of dif- Pisat
fusivity is neglected. The diffusion coefficient typically shows
Further, permeate flux was calculated in an iterative manner
an Arrhenius type dependence on temperature. Accordingly, as
due to the difficulty of writing a single equation. The following
per calculation, for 10 ◦ C change in temperature, around 5% or
paragraph describes the procedure adapted to estimate flux.
less change in diffusion coefficient is observed. Since the simu-
The flux values are guessed. Starting from the downstream
lation result shows temperature variation of less than 10 ◦ C, this
side, the vapor phase activity was calculated using Eq. (7). The
seems to be an acceptable assumption. Besides this, availability
activity at the membrane interface on the downstream side is
of experimental data only at a single temperature prevented us
estimated by equating to the vapor phase activity. The volume
from using temperature dependent model.
fraction at location (n − 1) inside the membrane was calculated
Dim = Dim
0
exp[bii φim + bij φjm ] (5) from the values of volume fraction at location (n) using the
guessed flux value in Eq. (6). The procedure was continued till
Similar equation may also be written for component j. Where one reached the upstream side of the membrane. The activity
bii and bij are plasticization coefficients and Dim 0 is diffusion
inside the membrane was again calculated using Eq. (3). These
coefficient at infinite dilution. Meuleman et al. (1999) solved activities were compared with the corresponding liquid phase
for the flux using the discrete analogue method, which, how- activities at the membrane interface. The liquid phase activity
ever, took large computational time. The present work uses again at the membrane interface was estimated using Margules equa-
the discrete analogue method; however, by modifying the algo- tion. The concentration at the membrane–liquid interface was
rithm. Accordingly, the membrane was subdivided into n layers estimated by solving Eq. (1) for wFiI using the guessed values of
with thickness of
X and instead of activity; volume fraction flux. The procedure is repeated till a convergence is observed
was calculated at each subdivision, using Eq. (6). Meuleman with respect to assumed and calculated values of fluxes. The
et al. (1999) had calculated activity at each subdivision, using relative tolerance of 1e−6 was used. The membrane was sub-
an equation similar to Eq. (6), where volume fraction is replaced divided into 100 sublayers (n = 100). It was confirmed that the
by activity (and the second term on the RHS contains an addi- flux value is independent of n (Table 1).
tional volume fraction term in the denominator). However, in Model parameters (diffusion coefficient at infinite dilution
addition, they also had to solve the non-linear Flory–Huggins and plasticization coefficients) were estimated using exper-
equations at each subdivision to evaluate the volume fraction. imental flux data (Table 2) at different concentrations of
This increases the computational cost, although the flux value hydrazine–water through ethyl-cellulose membrane. The flux
obtained is quantitatively similar to that obtained using the pro- data is taken from work of Ravindra et al. (1999). They have
cedure mentioned here. used 45 ␮m thick aminated ethyl-cellulose membranes for sep-
(n−1) (n) Ji
φim = φim +
X (6)
0 ρ exp[b φ(n) + b φ(n) ]
Dim sl ii im ij jm Table 2
Experimental data of flux and selectivity obtained with ethyl-cellulose mem-
At equilibrium, chemical potential at the membrane interface brane (Ravindra et al., 1999)
is equal to that inside the membrane. This condition gives the
activity inside the membrane at the downstream side. On the wt.% hydrazine 0.20 0.30 0.40 0.64 0.70 0.85
J (kg/m2 s) ×105 4.721 4.162 1.111 0.833 0.833 1.305
downstream side, the activities were assumed to be equal to α 250.0 127.8 22.2 5.6 8.3 8.6
ratio of partial pressure to saturation vapor pressures, since at
206 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

Table 3 side conditions. The permeation through the membrane takes


Estimated Pareto-sets of model parameters place in the radial direction. The transport equation was derived
Pareto-set for diffusion model parameters based on the following assumptions: (i) steady incompressible
Parameters Set 1 Set 2 Set 3
flow, (ii) lubricating (Deen, 1998) feed side momentum equa-
tion, (iii) feed as well as permeate side variation in the radial
D10 (m2 /s) ×1013 3.24 3.05 3.36 direction is neglected, (iv) negligible axial diffusion compared
b11 72.64 111.67 110.53
b12 0.37 1.02 0.46
to convection, (v) constant VLE parameters in the axial direc-
D20 (m2 /s) ×1012 4.43 1.36 2.06 tion (Hickey & Gooding, 1998), (vi) negligible sensible heat
b21 0.55 1.29 1.49 (carried by permeates) and negligible reversible work and (viii)
b22 36.76 35.58 32.61 negligible heat generation due to viscous dissipation.

2.2.1. Tube side feed flow


arating hydrazine–water mixture. The ethyl cellulose used for 2.2.1.1. Feed side. The retentate side mass, momentum and
membrane preparation had a weight average molecular weight of energy balances for the case of feed flow in the tube side
89,448 and a poly dispersity index of 1.42. The PV experiments (Fig. 2a) on simplification yield the following equations in terms
were done at a permeate pressure of 0.05 mmHg. Given a set of N of composition, velocity, pressure and temperature. The detailed
data points, 2N equations are obtained by writing separate equa- derivations are given in Appendix A.1.
tions for components i and j. The parameters were determined by
solving the least square optimization problem using MATLAB dP F 8ηsl v
=− 2 (8)
non-linear unconstrained optimization routine fminsearch. The dz Rtube
estimated parameter values are listed in Table 3. Three Pareto-
dv 2J
sets are tabulated which were generated using different initial =− (9)
guess values. The L2norm for the fitting was 0.126. dz Rtube ρsl
F
dxib 2ρjl Vsl [Jxib
F − (J V ρ /M )]
i sl sl i
2.2. Model development: mass, momentum and energy = (10)
dz Mj vRtube ρsl
balances along with solution–diffusion model
dT 2Jλsl
=− (11)
To describe the hollow fiber module transport behaviour, a dz ρsl vRtube Cp,sl
mathematical model was formulated by writing mass, momen-
tum and energy balances. The transport equation was derived 2.2.1.2. Permeate side. The permeate side mass, species and
for the retentate (remaining feed) as well as for the permeate momentum balance provide coupled differential equation in
side of the module as the flux depends on feed and permeate terms of velocity, composition and pressure. Separating each

Fig. 2. (a) Hollow fiber module with tube side feed flow. (b) Hollow fiber module with shell side feed flow.
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 207

after rearrangement, the following equations are obtained equation in terms of velocity, composition and pressure. Sepa-
(details in Appendix A.1). ration of each after rearrangement, the following equations are
obtained.
P
dyib 2RT P NRtube [(Ji Msv /Mi )] − Jyib
P
= (12) P
dyib 2RT P (Ji Msv /Mi − Jyib
P)
dz pR2shell uMj = (20)
dz pRtube uMj

2RT P NRtube [(RT P Ji Msv /M) − (RT P Ji Mj /Mi )


2RT P (RT P Ji Msv /Mi ) − (RT P Ji Mj /Mi ) − RT P Jyib
P
du −RT P Jyib
P − Ju2 M yP ]
j ib −Ju Mj yib − 2pMsv u fyib Mj
2 P 3 P
= du
=
dz pR2shell yib
P M [M u2 − RT P ]
j sv dz pRtube yib
P M [M u2 − RT P ]
j sv
4Msv u3 fRshell (21)
− (13)
[Msv u2 − RT P ]De L
2RT P u[Msv Jyib
P − (M 2 J /M ) + M JyP
sv i i j ib
2RT P NR tube u[Msv Jyib
P − (Msv
2 J /M ) + M JyP
i i j ib dp +(Mj Ji Msv /Mi )] + 2pMsv u2 fyib
PM
j
= (22)
dp + (Mj Ji Msv /Mi )] dz Rtube yib
P M [M u2 − RT P ]
j sv
=
dz R2shell yib
P M [M u2 − RT P ]
j sv Further, the pressure drop in the potting length is given by,
4pMsv u2 fRshell 2fρsv [u(1)]2 Lpot
+ (14)
p =
[Msv u2 − RT P ]De L Rtube
(23)

The friction factor may be estimated by the following equation where Lpot is the potting length with an assumed value of 1 cm.
(Crowder & Cussler, 1998) The friction factor is computed as (Crowder & Cussler, 1998),
9.5 ρsv uDe 16 ρsv uDtube
f = Re > 6000, where Re = (15) f = , where Re = (24)
Re0.83 ηsv Re ηsv
where Re is the Reynolds number, u the permeate velocity, De where Re is the Reynolds number, u the permeate velocity, Dtube
the effective diameter of the shell, ρsv the permeate density and the inner diameter of the fibre, ρsv the permeate density and ηsv
ηsv is the permeate viscosity. is the permeate viscosity.
Boundary condition:
2.2.2. Shell side feed flow The pressure, velocity, composition and temperature for feed
2.2.2.1. Feed side. The retentate side mass, momentum and side are subject to the following initial condition:
energy balances for the case of feed flow in the shell side (Fig. 2b) P(0) = Pin ; v(0) = vin ; xib
F
(0) = xin ; T (0) = Tin (25)
on simplification yield the following equations in terms of com-
position, velocity, pressure and temperature: Since, the module operates without sweep gas flow, the perme-
  ate velocity is zero at the entrance. The permeate composition at
F
dxib 2N [Rtube + t] xib
FJ Ji the entrance is given by the cross flow flux fraction. The differ-
= Vsl − (16)
dz vR2shell Msl Mi ential equation for the permeate side composition; velocity and

 pressure are subject to the following boundary conditions:
dv −2 [Rtube + t] N Mj J Mi Mj Ji
= + − (17) Ji (0)
dz R2shell ρjl Msl ρil ρjl Mi yib
P
(0) = ; u(0) = 0; p(L) = pap . (26)
J(0)
where z = 0 and z = L denote the inlet point of feed and the out-
2N[Rtube + t]Mi Mj u[(1/ρil ) − (1/ρjl )] let point of permeate collection, respectively. pL is the applied
dP F −2fρsl v2 F J/M ) − (J /M )]
× [(xib sl i i permeate pressure.
= +
dz De R2shell [(Mi xib
F /ρ ) + (M xF /ρ )]
il j jb jl
Solution technique:

 The set of coupled ordinary differential equations for tube
4N[Rtube + t]ρsl v Mj Mi Mj side feed flow, i.e. Eqs. (8)–(14), and for shell side feed flow, i.e.
+ J+ − Ji (18)
R2shell ρjl ρil ρjl Eqs. (16)–(23), were solved numerically. The key observation is
that feed side is governed by initial value differential equation
while the permeate side is governed by boundary value differ-
dT 2N(Rtube + t)Jλsl
=− (19) ential equation. Hence, an iterative solution technique is used.
dz ρsl vR2shell Cp,sl The unknown permeate side pressure at z = 0 is guessed, trans-
forming the boundary value problem to an initial value problem.
2.2.2.2. Permeate side. The similar mass, species and momen- This initial value problem is then solved using Euler integra-
tum balance for the permeate side provides coupled differential tion. The permeate pressure value at z = L is obtained from the
208 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

Table 4
Chosen model parameters and operating conditions for simulation
Parameter Value

xin 0.64
Fin (kg/h) 180
Tin (K) 333
Pap (Pa) 100
Tap (K) 318
L (m) 1
Rshell (m) 0.4
Rtube (␮m) 100
t (␮m) 50
N 2.5 × 106

Euler integration of the corresponding permeate pressure equa-


tion. This value is compared with the actual permeate pressure Fig. 3. Feed side water concentration as a function of fiber length axial
value at z = L, pap . The procedure is repeated till a convergence position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m; Rshell = 0.4 m;
is observed with respect to actual and calculated values of per- Rtube = 100 ␮m; N = 2,500,000).
meate pressure. The relative tolerance of 1e−3 was used as the
convergence criterion. The spatial co-ordinate z was discretized larger in this case. This also compensates for the smaller perme-
into 1000 parts for the tube side case, while 8000 subdivisions ation area available, in this case, as compared to shell side case.
were used for the shell side case. The obtained solution was The available area in the case of shell side flow is higher since
confirmed to be independent of finer discretization. permeation takes place from the outer surface of the tube.

3. Results and discussion 3.1.2. Temperature variation: retentate (feed) side


The variation of feed side temperature as a function of posi-
3.1. Spatial variation tion is shown in Fig. 4. The feed side temperature decreases
linearly along the length for tube side feed flow, while the varia-
The obtained solution–diffusion model parameters, listed in tion is parabolic for the shell side case. The phase change during
Table 3, were used to predict the performance of the hollow fiber permeation causes the feed side temperature to decrease from
module, using the developed model equations for the separation 318 to 308 K, for the shell side feed flow; whereas, for the case
of hydrazine water mixture. The physical property data used in of feed flow, in tube side, it decreased from 318 to 302 K. The
present simulation are listed in Table 1. The hollow fiber module decrease of temperature is more for the case of tube side flow
parameters and the operating conditions used for the simulation because of the larger permeate flux, in this case. Further, the
for both tube and shell side flows are listed in Table 4. The hollow temperature drop suggests the use of inter stage heating with
fiber module parameters are taken from literature (Baker, 2000). multiple modules.
The Reynolds number for the tube side feed flow was 0.14, while
for shell side feed flow, it was 117. 3.1.3. Pressure variation: permeate side
Fig. 5 shows the ratios of local point permeate pressure to
3.1.1. Composition variation: retentate (feed) side applied pressure along the fiber length. The ratio decreases
The variation of retentate (feed) side water composition as
a function of position is shown in Fig. 3. The feed side com-
position of water shows a linear decrease along the length for
tube side feed flow, while the variation is parabolic for the case
of shell side. The parabolic nature of variation is attributed to
the large variation in the permeate side pressure; in this case, as
shown in Fig. 5. Hence, the permeation along most of the axial
length is small with a significant increase in permeation at the
end of the fibre length. However, the change in composition is
very small because of single pass and low permeation through
the ethyl-cellulose membrane. Further, the decrease in compo-
sition of water is more for the case of tube side, which is due
to larger permeation. Obviously, this is because of the smaller
permeating side pressure build-up in this case. The larger pres-
sure build-up in the shell side case is due to the smaller tube side Fig. 4. Feed side temperature as a function of fiber length axial
radius (narrow space) through which permeate flow takes place. position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m; Rshell = 0.4 m;
Hence, the driving force, i.e. difference of partial pressure, is Rtube = 100 ␮m; N = 2,500,000).
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 209

Fig. 6. Permeate velocity as a function of fiber length axial posi-


tion (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m; Rshell = 0.4 m;
Fig. 5. Ratio of downstream pressure to applied pressure as a function of Rtube = 100 ␮m; N = 2,500,000).
fiber length axial position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m;
Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000).

3.1.4. Pervaporate collection velocity variation: permeate


side
parabolically for both the cases of feed flows, i.e. the perme- The pervaporate collection velocity is defined as the cumula-
ate pressure is building-up along the fiber length from the entry tive volumetric permeate collected per unit cross-sectional area
point of downstream pressure (at z = L). The ratio is 31 for on the permeate side. The variation of pervaporate collection
the case of feed flow in the shell side and 8.2 for the case of velocity along the length for both sides feed flows is plotted in
tube side. The larger pressure build-up in the shell side case Fig. 6. The pervaporate collection velocity increases along the
is due to the smaller tube side radius through which perme- length for both the cases. The increase in velocity is 3.1 m/s for
ate flow takes place. Feng and Huang (1995) experimentally the case of shell side feed flow; whereas, it is 9.0 m/s for case of
observed the said ratio to be 45 for 1 m length of fiber, for feed flow in the tube side. The increase in velocity, in the two
the permeation of isopropanol through silicon rubber. However, cases, is due to available permeate side pressure gradient along
the pressure increase is small in case of spiral wound geom- the length (permeate side). Further, it may be observed that in
etry as reported by Cao and Henson (2002). They found this both the cases of feed flow the velocity gradient is steep at rear
ratio to be close to 1.1 in their simulation. The lower pressure end of the module. This gradient, in the case of shell side feed
increase in their case is due to larger cross-sectional area on flow, is steeper because of the larger pressure gradient in this
the permeate side compared to this case. Besides this, there case. The sharp increase in velocity at the rear end is caused by
are reasons related to geometric aspects of the module. The the steep decrease in pressure at that end which causes signifi-
permeate that comes out in case of spiral wound module is cant decrease in vapor density. Further, it may be mentioned here
point flux as opposed to the cumulative flux in case of hollow that the increase of fiber length and decrease of fiber diameter
fiber module. Hence, there may be larger pressure drops in hol- to reduce the cost of module leads to a significant increase in
low fiber module due to higher permeate flow rate. Secondly, collection velocity. This increase in collection velocities causes
permeate has to flow through a smaller length in case of spi- difficulty in operating the module and becomes one of the most
ral wound module. However, in case of hollow fiber module, important operational constraints.
the permeate travels throughout the module length before exit.
This may increase the pressure drop in the case of hollow fiber 3.1.5. Composition variation: permeate side
module. Fig. 7 shows the variation of the permeate side composi-
The increase of pressure build-up reduces the permeate flux tion with fiber length. The permeate composition of water goes
due to the decrease of driving force. This is one of the major through a minima in the cases of shell side feed flow. The minima
problems with the use of hollow fiber modules for the pervapora- in permeate composition of water is due to the counter balancing
tion process. However, only small increase of permeate pressure effects of increase in water permeation flux, i.e. water selectiv-
build-up for the pervaporation of hydrazine–water, in case of ity, at lower downstream pressure and the decrease in water flux
tube side flow, may not be a problem with the use of hollow due to decrease in feed composition of water. The minima in
fiber modules for its separation. The effect of permeate pressure the water flux, as shown in Fig. 8a, further validates this argu-
build-up may be reduced by increasing fiber diameter; but, at ment. The two minima are observed at the same location. The
the cost of high membrane area packing density. Hence, there is permeate composition shows a similar trend in the case of tube
a trade off between the module volume and separation attained. side flow; although, the minima gets shifted to the module end.
One may attempt to optimize the fiber dimensions for the hollow The decrease in permeate composition of water here is due to
fiber module. the dominating effect of decrease in water permeation flux due
210 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

to decrease in feed composition of water, which overcomes the


increase in water permeation at lower downstream pressure for
most of the module length. However, at the module end, the
decrease in downstream pressure becomes dominant and causes
the water flux to increase.

3.1.6. Flux variation


Fig. 8a shows the variation of hydrazine, water and total fluxes
with fiber length. The water and total flux goes through minima
along the fiber length for the case of shell side feed flow as
well as tube side feed flow. The minima were observed at an
axial position of 40 and 71 cm for shell and tube side flows,
respectively. However, the variations, in case of tube side flow,
are small as compared to the shell side case. The minimum,
observed in total and water flux, is due to the competing effect
Fig. 7. Permeate water concentration as a function of fiber length axial of decrease in downstream pressure and the decrease in feed
position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m; Rshell = 0.4 m; concentration of water. The hydrazine flux increases in both the
Rtube = 100 ␮m; N = 2,500,000). cases due to decrease in downstream pressure as well as increase
in hydrazine concentration in the feed. At z = L = 1 m, the total,
water, and hydrazine fluxes are 1.02 × 10−6 , 8.09 × 10−7 and
2.15 × 10−7 kg/m2 s, respectively, for the feed flow in the shell
side; whereas, for the tube side, these values are 1.05 × 10−6 ,
8.25 × 10−7 and 2.29 × 10−7 kg/m2 s, respectively. It is clearly
evident that the tube side fluxes are little bit higher than shell
side. This is attributed due to the fact that there is higher pressure
on the permeate side at z = 1 m (Fig. 5) in case of shell side
feed flow, as there is a significant pressure drop in the potting
length.
The ratio of local total flux to local total flux at z = 1 m (Jt /Jt
(1)) as a function of position is shown in Fig. 8b. Similarly, the
ratio of water fluxes as well as hydrazine fluxes are also shown
in Fig. 8b. The trends are similar to those observed for the flux
variation. However, the change in the ratios of fluxes is more for
shell side flow. This is because of a larger decrease in permeating
pressure for the case of shell side flow.

3.2. Comparison between the shell side and tube side feed
flows

The drop in the permeate pressure is more in the case of shell


side flow due to larger frictional losses caused because of the
low Reynolds number of permeate flow in tubes. The frictional
losses are comparatively less when the permeate flows in the
shell side for the case of tube side feed flow. The larger pres-
sure gradient of permeates in the tubes for the case of the shell
side feed flow causes abnormal increase in pervaporate collec-
tion velocity. This abnormal increase in velocity is due to the
Poiseuille’s nature of permeate flow through the tube, which
increases linearly with pressure gradient. This restricts the use
of shell side feed flow configuration at very low downstream
pressure (high vacuum), which increases collection velocity to
large value. So, despite the advantages of higher available per-
Fig. 8. (a) Hydrazine, water and total flux as a function of fiber meation area, the operational constraints become restricting for
length axial position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m;
Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000). (b) Relationships between frac-
feed flow at shell side. Also, the increased pressure build-up in
tional flux (flux w.r.t. corresponding flux at z = 1 m) and fiber length axial case of shell side feed flow causes reduction in stage cut. Hence
position (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; t = 50 ␮m; Rshell = 0.4 m; the range of operability is large for tube side feed flow, which
Rtube = 100 ␮m; N = 2,500,000). may be advantageous. However, the feed side pressure drop is
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 211

larger in the tube side case than that observed in shell side flow, 3.3.1. Effect of feed concentration
resulting in larger feed pumping cost and thus may be disadvan- Fig. 9a shows the effects of feed water concentration on stage
tageous. Also, the mechanical strength of the fiber may restrict cut for both the cases of feed flow (shell and tube side). The
tube side feed flow to lower velocity. stage cut decreases and then increases, exhibiting minima at
34.17 wt.% (or 48 mol%) water concentration for shell side feed
flow; whereas, for tube side feed flow, the minima is observed
3.3. Effect of operating variables at 27.27 wt.% (or 40 mol%). It has been explained earlier that
as the water concentration increases the solubility decreases and
The separation capability was investigated over a range of diffusivity increases. These two opposite trends result in minima
feed composition, feed flow rate, permeate pressure and mem- of stage cut at 34.17 wt.% water concentration. Experimentally,
brane thickness. The properties of interests are the selectivity, α Ravindra et al. (1999) obtained the minima at 36 wt.% concen-
and stage cut φ. tration of water. The difference may be attributed due to the fact
that limited experimental flux data are available in this range of
FP yib
P [1 − xF ]
ib water concentrations (Table 2). Further, stage cut is smaller in
φ= , α= (27)
FF xib
F [1 − yP ]
ib
the case of shell side flow as compared to tube side flow. This
is because of larger pressure drop in case of shell side feed flow
The ratio, φ gives a better understanding for interpreting separa- (as explained in Section 3.1.3) as against tube side flow.
tion. It depicts the fraction of feed that has been purified. Being Fig. 9b shows the effect of feed concentration on selectiv-
a ratio, it can be used for comparing the permeation in mod- ity. The selectivity of the membrane increases with increase in
ules with different feed flows. The module parameters, used for composition of water for either of the flows. This continuous
simulation, are given in Table 4. increase of selectivity with increase in water composition may
be explained because of decrease of swelling and increase of
diffusion coefficient. It is interesting to predict that at very low

Fig. 9. (a) Influence of water feed concentration on stage cut (ϕ) (Fin = 180 kg/h; Fig. 10. (a) Influence of feed flow rate on stage cut (ϕ) (xf = 50 mol%; p = 100 Pa;
p = 100 Pa; t = 50 ␮m; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000). (b) t = 50 ␮m; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000). (b) Influence
Influence of feed water concentration on selectivity (Fin = 180 kg/h; p = 100 Pa; of feed flow rate on selectivity (xf = 50 mol%; p = 100 Pa; t = 50 ␮m; L = 1 m;
t = 50 ␮m; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000). Rshell = 0.4 m; Rtube = 100 ␮m; N = 2,500,000).
212 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

composition of water (i.e. less than 20 wt.%), hydrazine perme- bulk concentration at higher flow rates. The increase of water
ation becomes more than that of water, and hence, results into concentration at the membrane surface decreases the solubility
selectivity value of less than 1, for both shell as well as tube side of feed solution in the membrane and this decreases the flux
feed flows. Therefore, analogous to distillation, here also there through the membrane.
is a formation of azeotrope during pervaporation of hydrazine- Fig. 10b shows the effect of feed flow rate on selectivity.
water mixtures. Such a phenomena has earlier also been demon- The selectivity of the membrane increases with increase in feed
strated by other authors (Bakish & Schneider, 1986; Nguyen & flow rate. This may be explained due to increase in mass trans-
Clement, 1991; Tusel & Bruschke, 1985; Mulder & Smolders, fer coefficient, which causes higher permeation flux of water.
1984) for systems like ethanol–water, etc. Still, pervaporation However, the selectivity was observed to be higher for the case
provides an advantage with shifting of azeotropic concentration of tube side flow as against shell side flow. This may be due to
of water. The shift is from 28.5 (by distillation) to 17.5 wt.% for larger permeate pressure drop in the case (Fig. 5) of shell side
tube side flow (refer inset of Fig. 9b). In case of shell side feed feed flow which may have resulted in lowering water flux.
flow, the azeotrope is observed at 28.42%. Further, shifting as
well as complete avoidance of azeotropic formation is possible 3.3.3. Effect of permeate pressure
with the selection of much improved selective membranes. The effect of downstream pressure on stage cut is shown
in Fig. 11a. Stage cut decreases with increase in downstream
3.3.2. Effect of feed flow rate pressure with obvious reasoning and approaches zero around
Fig. 10a shows the effect of feed flow rate on stage cut. The 4500 Pa. Further, stage cut is more for the tube side flow com-
stage cut continuously decreases with feed flow rate. This is pared to shell side flow, because of larger pressure drop in case of
primarily due to the reduction of residence time with increase in shell side flow (attributed to smaller cross-sectional area avail-
feed flow rate and hence lower time for permeation. Also, due to able for permeate flow). Likewise, the effect of downstream
increase in flow rate, there is increase of mass transfer coefficient pressure on selectivity is shown in Fig. 11b and the results depict
through the boundary layer. This will increase the concentration
of water at the membrane surface and may become equal to

Fig. 11. (a) Influence of downstream pressure on stage cut (ϕ) (xf = 50 mol%; Fig. 12. (a) Influence of membrane thickness on stage cut (ϕ) (xf = 50 mol%;
Fin = 180 kg/h; t = 50 ␮m; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m; Fin = 180 kg/h; p = 100 Pa; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m;
N = 2,500,000). (b) Influence of downstream pressure on selectivity N = 2,500,000). (b) Influence of membrane thickness on selectiv-
(xf = 50 mol%; Fin = 180 kg/h; t = 50 ␮m; L = 1 m; Rshell = 0.4 m; Rtube = 100 ␮m; ity (xf = 50 mol%; Fin = 180 kg/h; p = 100 Pa; L = 1 m; Rshell = 0.4 m;
N = 2,500,000). Rtube = 100 ␮m; N = 2,500,000).
N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214 213

a continuous decrease of selectivity. The increase in downstream Further, the θ component of velocity is zero assuming symmetry;
pressure results in decrease in flux. The decrease is more for the continuity equation under the assumption of incompressible
water flux compared to hydrazine flux, resulting in lower sepa- fluid gives the radial component of velocity:
ration factor at higher downstream pressure. Hence,
 
dvz r r3
3.3.4. Effect of membrane thickness vr = −2 − (A.2)
dz 2 4R2tube
The effect of membrane thickness on stage cut is shown in
Fig. 12a. An exponential decay trend is observed for tube side The radial component of velocity at r = R is equal to the perme-
as well as shell side flow. The decrease is obviously because of ation velocity and which is given by Eq. (9).
higher diffusional resistance with increase in thickness.
The selectivity values increase with increase in membrane
A.1.2. Mass balance at feed side
thickness for shell side as well as for tube side feed flows, as
The molar species balance on the feed side yields the follow-
shown in Fig. 12b. However, values of selectivity saturate after
ing equation on simplification:
30 ␮m thickness for the tube side feed flows but there is no such
trend observed in the case of shell side feed flow. The profile xib
F dv
z Mj 1 dxibF 2Ji
for shell side flow suggests that it is in the linear regime, which + vz =− (A.3)
Vsl dz ρjl Vsl dz
2 Rtube Mi
is displayed in case of tube side flow till a thickness value of
30 ␮m. Rearranging Eq. (A.3), Eq. (10) may be obtained, explicitly,
in terms of composition variation.
4. Conclusions
A.1.3. Momentum balance: permeate side
Detailed mathematical formulations were derived, combin- The permeate side momentum balance can be written as per
ing solution–diffusion model with mass, momentum and energy Bird et al. (1960). Further, vapor mass density can be estimated
balances. Simulated results were obtained from developed mod- by ideal gas law. The simplification yields the following equa-
els. The spatial property variations were observed in a hollow tions:
fiber module for hydrazine–water separation during pervapo- dp du dyP
ration. The tube side feed flow operation was found to provide [Msv u2 + RT P ] + 2Msv pu + pu2 [Mi − Mj ] ib
dz dz dz
better selectivity and flux compared to shell side feed flow. Also,
the shell side feed flow was observed to be restricted (operational 2fρsv u2 2Rshell
=− RT P (A.4)
constraint). Simulated results depicted formation of azeotrope De L
at 17.5 and 28.42 wt.% water concentration for tube side feed
flow and shell side feed flow, respectively; against azeotropic
water concentration of 28.5 wt.% as experienced by distillation.
A.1.4. Mass balance: permeate side
Results suggest use of better selective membrane or hybrid pro-
The overall mass balance on the permeate side was simplified
cess for the separation of hydrazine–water mixture.
as:
dp du dyP 2J
Appendix A Msv u + pMsv + pu[Mi − Mj ] ib = 2 NRtube RT P
dz dz dz Rshell
(A.5)
A.1. Tube side feed flow
The molar species balance yields the following equation on sim-
A.1.1. Momentum balance at feed side plification:
Under the lubricating assumption, the feed side velocity and
pressure equations were obtained by solving the axial com- dp P du dyP 2Ji
ponent of Navier Stokes equation. According to lubricating yib
P
u + yib p + pu ib = 2 NRtube RT P (A.6)
dz dz dz Rshell Mi
assumption (Deen, 1998) liquid flows in long narrow channels
or in thin films; they often have the characteristics of being uni- Eqs. (A.4)–(A.6) is simplified such that each equation contains a
directional and dominated by viscous stress. The other terms single derivative, providing Eqs. (12)–(14). Similar derivations
in the Navier Stokes equation are neglected based on order of can be done for the shell side feed flow.
magnitude analysis. The solution for the ‘z’ component Navier
Stokes equation, assuming pressure gradient to be independent References
of radial variation, gives the z component of velocity which is
averaged in the radial direction to obtain the following equation: Asada, T. (1988). Dehydration of organic solvents, some actual results of
pervaporation plants in Japan. In Third International Conference on per-
1 dP F 2 vaporation processes in the Chem. Industry.
vz = − R (A.1) Baker, R. W. (2000). Membrane technology and applications. New York:
8ηsl dz tube McGraw-Hill.
214 N. Hoda et al. / Computers and Chemical Engineering 30 (2005) 202–214

Bakish, R., & Schneider, W. (1986). Modelling of transport phenomena in Meuleman, E. E. B., Willemsen, J., Mulder, M. H. V., & Strathmann, H.
pervaporation membranes. In Proceedings of the first international con- (2001). EPDM as a selective membrane material in pervaporation. Journal
ference on pervaporation processes in the Chem. Ind. Atlanta, Vol. 23–26 of Membrane Science, 188, 235.
(pp. 133–141). Mulder, M. H. V., & Smolders, C. A. (1984). On the mechanism of separation
Bausa, J., & Marquardt, W. (2000). Shortcut design methods for hybrid mem- of ethanol/water mixtures by pervaporation: I. Calculation of concentra-
brane/distillation processes for the separation of nonideal multicomponent tion profiles. Journal of Membrane Science, 17, 289.
mixtures. Industrial and Engineering Chemistry Research, 39, 1658. Nguyen, Q. T., & Clement, R. (1991). Analysis of some cases of pseudo-
Bird, R. B., Stewart, W. E., & Lightfoot, E. N. (1960). Transport Phenomena. azeotropes in pervaporation. Journal of Membrane Science, 55(1–2),
New York: John Wiley & Sons. 1–19.
Brun, J. P., Larchet, C., Bulvestre, G., & Auclair, B. (1985). Modeling of Prausnitz, J. M., Lichtenthaler, R. N., & Azevedo, E. G. (1986). Molecular
the pervaporation of binary mixtures through moderately swelling, non thermodynamics of fluid-phase equilibria. Eaglewood Cliffs, NJ: Prentice-
reacting membranes. Journal of Membrane Science, 23, 257. Hall.
Cao, B., & Henson, M. A. (2002). Modeling of Spiral wound pervaporation Rautenbach, R., Herion, C., Franke, M., Abdul Fattah, A., Asfour, M., Bem-
modules with application to the styrene/ethyl benzene mixtures. Journal querer, C., et al. (1988). Investigation on mass-transport in asymmetric
of Membrane Science, 197, 117. pervaporation membranes. Journal of Membrane Science, 36, 445.
Crowder, R. O., & Cussler, E. L. (1998). Mass transfer resistance in hollow Rautenback, R., Herion, C., & Meyer-Bluemenroth, U. (1991). Engineer-
fiber pervaporation. Journal of Membrane Science, 173, 145. ing aspects of pervaporation: Calculation of transport resistances, module
Cussler, E. L. (1998). Diffusion: Mass transfer in fluid systems. New Delhi: optimization and plant design, Chapter 3 in pervaporation membrane
Cambridge University Press, First south Asian Edition. separation processes. In R. Y. M. Huang (Ed.), Membrane science and
Deen, W. H. (1998). Analysis of transport phenomena. New York: Oxford. technology series, Vol. 1. Elsevier.
Duggal, A., & Thompson, E. V. (1986). Dependence of diffusive perme- Ravindra, R., Sridhar, S., & Khan, A. A. (1999). Separation studies of
ation rates and selectivities on upstream and downstream pressures V1. hydrazine from aqueous solution by pervaporation. Journal of Polymer
Experimental results for the water/ethanol system. Journal of Membrane Science: Part B, 37, 1969.
Science, 27, 13. Ravindra, R., Sridhar, S., Khan, A. A., & Rao, A. K. (2000). Pervapora-
Eliceche, A. M., Daviou, M. C., Hoch, P. M., & Uribe, I. O. (2002). Opti- tion of water, hydrazine and monomethylhydrazine using ethyl cellulose
mization of azeotropic distillation columns combined with pervaporation membranes. Polymer, 41, 2795.
membranes. Computers and Chemical Engineering, 26, 563. Sander, U., & Soukup, P. (1988). Design and operation of a pervaporation
Feng, X., & Huang, R. Y. M. (1995). Permeate pressure build-up in shell side plant for ethanol dehydration. Journal of Membrane Science, 36, 463.
fed hollow fiber pervaporation membranes. Canadian Journal of Chemical Satyanarayana, S. V. (2004). Membrane preparation, characterization, mod-
Engineering, 73, 833. elling and optimization of pervaporation. Ph.D. thesis. Kanpur, India:
Feng, X., & Huang, R. Y. M. (1997). Liquid separation by membrane perva- Indian Institute of Technology.
poration, a review. Industrial and Engineering Chemistry Research, 36, Satyanarayana, S. V., & Bhattacharya, P. K. (2004). Pervaporation
1048. of hydrazine hydrate: Separation characteristics of membranes with
Flory, P. J. (1953). Principles of polymer chemistry. New York: Cornell Uni- hydrophilic to hydrophobic behaviour. Journal of Membrane Science, 238,
versity Press. 103–115.
Hickey, P. J., & Gooding, C. H. (1998). Modeling of spiral-wound membrane Schmidt, W. E. (1984). Hydrazine and its derivatives: Preparation, properties,
module for the pervaporation of volatile organic compounds from water. and applications. New York: John Wiley & Sons.
Journal of Membrane Science, 88, 47. Sieder, E. N., & Tate, C. E. (1936). Heat transfer and pressure drop of liquids
Haggins, J. (1990). Membrane Technology has achieved success, yet lags in tubes. Industrial & Engineering Chemistry, 28, 1429.
potential. Chemical and Engineering News, American Chemical Society, Tusel, G. F., & Bruschke, H. E. A. (1985). Use of pervaporation systems in
68(40), 22. the chemical industry. Desalination, 53, 327.
Matsui, S., & Paul, D. R. (2002). Pervaporation separation of aro- Wessling, M., Werner, U., & Hwang, S. T. (1991). Pervaporation of aromatic
matic/aliphatic Hydrocarbons by crosslinked poly(methyl acrylate- C8-isomers. Journal of Membrane Science, 57, 257.
co-acrylic acid) membranes. Journal of Membrane Science, 195, Wilson, R. Q., Munger, H. P., & Clegg, J. W. (1952). Vapour–liquid equi-
229. librium in the binary system hydrazine/water. Chemical Engineering
Meuleman, E. E. B., Bosch, B., Mulder, M. H. V., & Strathmann, H. Progress Symposium Series, 3, 115.
(1999). Modeling of liquid/liquid separation by pervaporation: Toluene Zukauskas, A. (1972). Heat transfer from tubes in cross flow. Advances in
from water. American Institute of Chemical Engineer’s Journal, 45, 2153. Heat Transfer, 8, 93.

You might also like