Professional Documents
Culture Documents
Modeling Membrane Gas Separation
Modeling Membrane Gas Separation
net/publication/329893574
CITATIONS READS
3 7,635
5 authors, including:
Some of the authors of this publication are also working on these related projects:
Development of the calculation block of the gas separation membrane module for modeling of technological processes in the ASPEN PLUS environment View project
All content following this page was uploaded by Kozlova Alina on 24 December 2018.
Abstract—The paper presents the results of the development of a unit for calculating mass transport in a mem-
brane gas separation apparatus in the Aspen Plus environment. The created user model has been verified by
comparing the results of the calculation of separation characteristics with available experimental data for two
cases. In the first case, the calculation results were compared with experimental data on the separation of the
ternary H2O/EtOH/N2 (13.3/5.5/81.2 mol %) vapor–gas mixture simulating the stream from the stripping
column of a laboratory membrane vapor separation unit with a plate-and-frame membrane module for eth-
anol recovery from fermentation broth. It has been shown that the calculated concentrations of the compo-
nents in the retentate and permeate streams are close to the experimental values obtained in the separation of
the ternary vapor–gas mixture. In the second case, the calculation results were compared with experimental
data obtained upon the separation of a binary N2/CH4 gas mixture (99/1 mol %) simulating a stream with a
low content of an easily penetrating impurity component using a disk-type membrane module. It has been
found that the calculated concentration of CH4 in the retentate is noticeably less than in the experiment,
especially at high stage cuts, a difference that is associated with the increasing influence of the longitudinal
diffusion of components at low speeds of the mixture in the channels of the membrane module, which dete-
riorates the separation characteristics of the apparatus with respect to the removal of impurities and requires
additional consideration of the deviation from the plug-flow mode in computational mathematical models
used for this case.
Keywords: membrane gas and vapor separation, mathematical modeling of membrane processes, Aspen Plus
DOI: 10.1134/S2517751619010049
1
2 KOZLOVA et al.
Feed JF
JR
stream cFi J(x) ci(x) J(x + Δ x) cRi Retentate
p
Δx
Membrane
0 x
M
L
J (x) p'
JP J '(x) c'(x) J '(x + Δ x)
Permeate i
cPi
1 Fig. 1. Schematic of mass transport in a membrane module for the countercurrent mode.
calculations. Aspen Plus contains a voluminous data- —change in a component flux upstream of the
base of physicochemical properties and thermody- membrane:
namic models, which allow prediction of properties
J i ( x + Δx ) = J i ( x ) − J i ( x ),
M
for both gaseous mixtures and liquids. With the use of (1)
the latest versions of Aspen Plus, it is possible to inte- —change in a component flux downstream the
grate design schemes created by the user in Excel [8], membrane:
thereby greatly expanding the possibilities of modeling
J i'( x + Δx ) = J i'( x ) − J i ( x ),
M
nonstandard blocks of the process scheme. Also, cus- (2)
tom models made in other software products can be
integrated into Aspen Plus. There are examples in the —component flux through the membrane:
literature of using the software platforms MATLAB J i ( x) = Qi B Δx Δpi ( x ),
M
(3)
[9] and FORTRAN [10] to create a custom model with
its subsequent export to Aspen Plus. —driving force for the transport of a component
Currently, Aspen Plus does not contain a ready across the membrane:
solution for modeling mass transport processes in Δpi ( x) = pci ( x ) − p'ci'( x ), (4)
membrane gas separation devices; however, the soft-
ware company (AspenTech©) provides, among other —mole fraction of the component upstream of the
things, the Aspen Custom Modeler (ACM) modeling membrane:
tool that focused on creating mathematical models of J i ( x)
ci ( x) = , (5)
its own, which can be used to design a custom mem-
brane block. The advantage of the model created in J i ( x)
i
ACM is the ease of exporting and binding it to Aspen
Plus, after which it can be used in the same way as —mole fraction of the component downstream of
other models already available in the Aspen Plus the membrane:
library.
J i'( x)
In this study, a gas separation membrane module ci'( x) = , (6)
was created in the ACM program and exported to
Aspen Plus for further work. The created block can
i
'
J i ( x)
MATHEMATICAL MODEL
OF MEMBRANE MODULE
JP = J '(0).
i
i (10)
The transfer of gases and vapors through nonpo- In this study, a plate-and-frame membrane mod-
rous polymeric materials is described in terms of the ule was chosen as the object of simulation. This type of
solution–diffusion mechanism [11]. To calculate mass module was designed by Stern [12] and was intended
transport in the membrane module (Fig. 1), the fol- for the recovery of helium from natural gas on an
lowing equations were used: industrial scale. The design of the module is similar in
configuration to that with flat sheet membranes, graphs and tables are automatically displayed in the cur-
which are often used in laboratory experiments in pri- rent unit of measurement, and the values input into forms
mary research. The module consists of flat sheet are automatically converted to basic units of measure-
membranes sandwiching a spacer made from drainage ment. After entering the information about the parame-
material [13]. With this design of the module and the ters and variables and spelling out mathematical equa-
feed stream composition, the crossflow rate becomes tions, streams are connected to the model; in the given
the main parameter that determines the mass trans- case, the block has three streams: feed stream as the inlet
port in the module. stream and the retentate and permeate as outlet streams.
In the study we considered three operating modes In ACM, streams are denoted through ports in which
1 of the membrane module: direct flow, countercurrent
parameters and variables that carry information, such as
flow, and crossflow. Mathematical models of mass flow temperature, pressure, stream component compo-
transport for all operation modes are based on the fol- sition, and energy characteristics (not considered in this
lowing assumptions: isothermal conditions, the paper), are built. After the complete information is
absence of a pressure gradient in the channel along the defined, the model can be tested. To do this, the compiler
membrane, ideal gas behavior of components, and built into ACM is run, which starts the calculation
constancy of membrane permeability. In the cases of according to the described model. At this stage, ACM
1 direct and countercurrent flows, the plug-flow mode
can identify errors resulting from building the model.
is realized in the feed and permeate channels; whereas Table 1 lists the parameters necessary for the calculation
the plug-flow mode for crossflow operation is realized of the custom model.
in the feed channel and the flow direction is perpen- To build a process flowchart using the created
dicular to the membrane in the permeate channel. block, the user must have the ACM file in which the
model was created. To export a model from ACM to
Aspen Plus, it is also necessary to have a C++ com-
CREATING A MODEL OF MEMBRANE GAS piler (Visual C ++ package from Visual Studio©) [14].
SEPARATION BLOCK IN ASPEN In order to add the custom block (Fig. 2) to the Aspen
CUSTOM MODELER Plus library, it is necessary to enable the option of
export to ACM and then run the import option for this
The principle of building a model in ACM is based model in Aspen Plus. Further, the custom model can
on writing equations describing the behavior of com- be used in the same way as any other elements already
ponents or their characteristics in the system under available in the Aspen Plus library.
consideration, for example, material balance equa-
tions, and any necessary additional conditions (for
example, boundary conditions). In the course of COMPARISON OF CALCULATION RESULTS
building a model, a certain sequence of actions is IN ASPEN PLUS WITH EXPERIMENTAL
observed. Work on model building begins with the RESULTS
definition of all parameters and variables involved in
the model. Aspen Custom Modeler has its own exten- The calculation block developed was tested on the
sive database of parameters and variables that can be basis of comparison of the calculated separation char-
selected in the user interface. Next, mathematical acteristics with available experimental data for the fol-
equations are written, which, as a rule, describe the lowing two cases:
material and heat balances. The order of the descrip- —comparison of the calculation results with the
tion of a variable or parameter can be studied in the experimental data obtained upon the separation of the
ACM assistant. To solve the equations, ACM uses ternary H2O/EtOH/N2 (13.3/5.5/81.2 mol %) vapor–
metric units (basic units of measurement); therefore, gas mixture [15] simulating the stream from the strip-
the user must ensure that the variables, parameters, ping column used in ethanol recovery from fermenta-
and coefficients in the model equations are written in tion broth by membrane vapor permeation in a labora-
basic units. However, the user can select the units for tory plate-and-frame membrane module, with the
the variables that are displayed in the graphical user module operating in the countercurrent mode (the 1
interface. This means that the values displayed in experimental part is detailed in [16]);
C++ compiler
Fig. 2. Scheme of operation of the ACM user block in Aspen Plus.
0.6
Experiment 100
Calculation 90
Concentration of CH4
0.5
in retentate, mol %
80
0.4
Stage cut, %
70
0.3 60
50
0.2 40
30 Experiment
0.1 20 Calculation
0 1
0.0E+00 5.0E–05 1.0E–04 1.5E–04 2.0E–04 0
Feed flow rate, mol/s 0.0E+00 5.0E–05 1.0E–04 1.5E–04 2.0E–04
Feed flow rate, mol/s
Fig. 3. Dependence of the CH4 concentration in the reten-
tate on the feed flow rate. Fig. 4. Dependence of the stage cut on the feed flow rate.
—comparison of the calculation results with exper- presence of the net spacer) and is not taken into
imental data obtained upon the separation of a binary account in mathematical models used for calculation.
N2/CH4 (99/1 mol %) [6] gas mixture simulating a stream
with a low concentration of easily penetrating impurity In the case of the removal of the easily permeable
component using a disk-type membrane module, with the impurity component from the binary gas mixture, it
1 module operating in the countercurrent mode.
was found that the calculated CH4 concentration in
the retentate is noticeably lower than the experimental
The comparison showed that in the case of separa- one, especially at low feed flow rates (high stage cuts,
tion of the ternary vapor–gas mixture, the calculated Figs. 3, 4). This is apparently due to the influence of
values for the concentrations of the components in the the longitudinal diffusion (mixing) of components in
retentate and permeate streams are basically close to the membrane module channels because of a long res-
the experimental values (Table 2). However, a notice- idence time of the feed upstream of the membrane
able deviation is observed for some of them, which can (tens of seconds), since the diffusion reduces the sep-
primarily be due to the influence of longitudinal mix- aration characteristics of the membrane apparatus, a
ing in the membrane module channels (due to the change that is especially noticeable upon the removal
SPELL: 1. countercurrent