Desalination 408 (2017) 127–132

Contents lists available at ScienceDirect

Desalination

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

Analysis and validation of the hydrodynamics of an electrodialysis cell

using computational fluid dynamics
R. Enciso, J.A. Delgadillo, O. Domínguez, I. Rodríguez-Torres ⁎
Instituto de Metalurgia, Facultad de Ingeniería, Universidad Autónoma de San Luis Potosí, 78210 San Luis Potosí, Mexico

H I G H L I G H T S

• Hydrodynamics and mass transfer in a electrodialysis reactor was studied via CFD.
• The theoretical results show a good correlation with those obtained experimentally.
• Theoretical and experimental mass transfer coefficients differ in max 16%.
• Major differences in the mass transfer coefficients are in the low velocity zones.
• CFD has shown the influence of the hydrodynamic regime on the process efficiency.

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

Article history: Due to the shortage of drinking water and the deterioration in its quality several processes have been developed
Received 1 September 2016 to remove the ions and impurities contained. A technology used for the ions removal is the electrodialysis, which
Received in revised form 11 January 2017 uses an electric field gradient and ion-exchange membranes to promote the movement of ions from treated
Accepted 12 January 2017
water to a suitable compartment.
Available online xxxx
Studies carried out using this technology have shown the influence of the hydrodynamic regime on the efficiency,
Keywords:
mass transfer rate on the process and its influence on the concentration polarization.
Electrodialysis Using the finite element method, this work describes the hydrodynamic behavior and mass transfer in a filter
Computational fluid dynamics press type electrodialysis reactor using computational fluid dynamics and experimental validation of the results
Hydrodynamic obtained with CFD.
Digital imagen analysis The theoretical results show a good correlation with those obtained experimentally for hydrodynamics, from
which the validation of the mass transfer and the distribution of potential on the electrodes and the solution is
possible. A second validation is made by comparing theoretical and experimental mass transfer coefficients
obtaining a maximum error of 16%.
© 2017 Elsevier B.V. All rights reserved.

1. Introduction combines ion-exchange and electrodialysis, it is important to take into

Water electrochemical treatment using ion-exchange material has a
wide use in electrodialysis [1,2] and electrodeionization [3,4]. Both
techniques utilize electrical current as the main driving force in matter
separation so mass transfer of charged particles is important. In fact
electrodeionization is a processes combining ion-exchange with elec-
trodialysis. The knowledge of hydrodynamics for both systems allows
the improvement of the performance of the processes as a whole [5,6].
The performance of electrodialysis process depends on several parame-
ters such as applied potential or current, or hydrodynamics in the dilut-
ed compartment. The hydrodynamics in the diluted compartment is
influenced by flow entrance effects. Given that electrodeionization
⁎ Corresponding author.
E-mail address: learsi@uaslp.mx (I. Rodríguez-Torres).
account the inlet and outlet geometry of the electrodializer [7].
The design of the necessary components of an electrodeionization
reactor is very similar to those used for a conventional electrodialysis,
resulting in the possibility to use the same strategy for the design of
the reactor given that EDI configuration was developed to improve the
electrical conductivity drawback at ED cell [5]. There are some pub-
lished works available that study the effects of the geometry of the
electrodializer over different process parameters. Pawlowsi et al. [8]
studied a reverse electrodialysis process using various concentrations
of NaCl solutions and studied the ohmic and non-ohmic resistances in
the stack varying the hydrodynamic conditions (1.5 cm s−1). Using a
chronopotentiometric study they determined the thickness of the
boundary layer finding that the process is strongly influenced by the
inlet effects. The presence of spacers reduces the thickness of this
boundary layer increasing the ohmic resistance on the membrane. In

http://dx.doi.org/10.1016/j.desal.2017.01.015
0011-9164/© 2017 Elsevier B.V. All rights reserved.
128 R. Enciso et al. / Desalination 408 (2017) 127–132
their findings they suggest that the length of the electrodializer is not
bigger than the zone in which the inlet is influencing. Zourmand et al.
[9] developed a bidimensional, stationary, isothermal model to predict
the mass transport of an ion in the electrodializer considering the con-
centration polarization phenomena and its influence in ion separation.
This model includes the Navier-Stokes and Nerst-Planck equations.
These equations where solved using the finite element method and
CFD on the dilute and concentrated compartment and the membranes,
being capable of predicting the local concentrations and ion fluxes, elec-
tric potential and velocity distribution in the electrodializer. This math-
ematical model is limited to current densities below the limiting
diffusion current and neglecting the effect of the spacers. Turek and
Mitko [10] developed a model to measure the RTD in a working reactor
to assess the effect of the applied current on the ion migration and the
electroosmotic flux of the water. The change of the RTD in the concen-
trated compartment was confirmed in the 0.38–0.84 cm s−1 range ob-
serving a substantial longitudinal dispersion, being the objective of
this methodology to obtain a method that allows to predict the proba-
bility of the formation of crystals on the membrane.
There are few studies that have reported the impact of hydrodynam-
ics on the efficiency of an electrodialysis reactor; however it is a funda-
mental design parameter. Furthermore, published studies show the
global influence of hydrodynamic on the rate of mass transfer efficiency
of the process and its relationship to concentration polarization. Often
optimizations or appropriate scaling are not possible since the mea-
sured response of the studied parameter (hydrodynamic regime) does
not allow to visualize or predict local variations, its effects and how to
modify them. Due to the complexity involved on the analysis of fluid
flow equations when the geometry of the system is complicated, com-
putational tools have been developed such as Computational Fluid Dy-
namics (CFD), that help describing the hydrodynamics in the
compartment globally and locally, as well as mass transfer (without re-
quiring a large number of experimental tests) [11].
The aim of this paper is to propose a simple methodology for the
experimental and theoretical description of an electrodialysis/
electrodeionization reactor using Digital Image Analysis and Computa-
tional Fluid to validate the results. The knowledge gain from this
paper will allow a better understanding of the mass transfer phenome-
na in an electrodeionization reactor.
2. Methodology
The reactor used consists of DSA and a stainless steel cathode, the
reactor compartments are delimited by ion exchange membranes and
the compartments formed between the membranes are of 50 × 80 mm
Fig. 1. Schematic illustration of the reactor geometry.
(Fig. 1). The fluid is fed by an inlet of 3 mm diameter located in the
lower compartment using a hose with a diameter of 1.6 mm and a dis-
charge identical to the inlet located at the top of the compartment. For
the hydrodynamics study, the continuity equation and the momentum
balance for a steady-state incompressible flow are solved. Calculating
the Reynolds number a value of Re = 17 for a mean residence time of
1 min was obtained, therefore, a laminar model was considered to simu-
late de system.
Simulations were carried on with a mesh conformed for 172,161 tet-
rahedral and hexahedral elements, which provides sufficient detail in
the areas with the highest gradients of speed and concentration. Simu-
lations were carried out with an Intel Core™ 2 Duo CPU T 5570 a
2.00 Ghz processor and 16GB of RAM. Details of the additional parame-
ters used for the simulation can be found on Table 1.
A 6 mm gap between spacers and an open channel configuration
were used due to their availability since this configuration was a first
test to assess the feasibility of the electrodializer for an
electrodeionization process (including further modelling).
To describe the effect of the hydrodynamic over mass transfer it is
necessary to know the flow distribution inside the cell. The validation
of the simulated hydrodynamic results was made using a flow distribu-
tion technique, replacing the ion-exchange membranes and electrodes
by transparent acrylic plates that allowed to obtain a video showing
the hydrodynamic behavior inside the reactor, then to carry on the Dig-
ital Image Analysis, a pulse of 1.0 mL of a blue dextran solution
(50 g L−1) was introduced in an injection port located 25 diameters
away from the input. Once the reactor was operating at steady state
the flow distribution was recorded digitally using a Canon Digital
Video GL2 3CCD camcorder.
The experimental results were compared with the simulated veloc-
ity profiles, determined by drawing a plane in the center of the channel
(z = 3 mm) and plotting the velocity profile of the cell with respect to x
axis, using data from simulation and local variations considering the
fluid velocity. The experimental trajectory of the fluid was obtained by
separating the video in to frames of 720 × 480 pixels and tracking the
path of the dye frame by frame that it was injected into the system.
The validation of the simulated results was obtained by comparing
experimental and simulated data. Detailed analysis allowed the detec-
tion of stagnant areas, channeling and flow distribution to optimize
the reactor operation.
The simulation of the primary current distribution was performed
using COMSOL Multiphysics 4.3a.
The imposed voltage on the simulated electrodialysis cell was 4.2 V
for the steel cathode and a voltage of 0 to the anode, and the current
density in the cathode was 2.5 A m−2.
The mass transport in the cell was modeled under a current density
of 2.5 A m−2 at laminar hydrodynamic conditions using results obtained
for the flow distribution and potential distribution of previous simula-
tions. For the Nernst-Planck equation the velocity term in the portion
of the convection contribution is obtained from the simulation of hydro-
dynamics, and the term of potential is obtained from the simulation of
the potential distribution.
For the bulk of the solution, the Nernst-Planck equation considering
diffusion, migration and convection was used, to solve the boundary
Table 1
Simulation parameters and geometry dimensions.
Reaction compartment volume 2.4 × 10−5 m3
Height and width of the compartment 0.08 × 0.05 m
Thickness of the compartment 6 × 10−3 m
Inlet and outlet diameter 1.6 × 10−3 m
Thickness of the ion-exchange membranes 1.5 × 10−4 m
Fluid viscosity 8.82 × 10−4 Pa s
Fluid density 994.4 kg m−3
Operation flow
2.5 × 10−2 L min−1
Residence time 1 min
 R. Enciso et al. / Desalination 408 (2017) 127–132 129

layer only the diffusion and migration terms were considered since the 3. Results and discussion of the experiment
velocity in the boundary layer is very close to zero; within the mem-
brane only the migration term was considered since it is the predomi- Using Digital Image Analysis (DIA) the fluid path was evaluated
nant mechanism inside the membrane. quantitatively and qualitatively and compared with the simulated re-
Data obtained from experiments using electrodialysis to 200 ppm sults as a first validation of the computational simulations. The fluid
KCl and varying the flow from 25 to 160 mL min−1 (from 0.1388 to flow shows regions were the fluid remains stagnant, rapid fluid flows
0.888 cm s−1) was conducted for an experimental validation, compar- (channeling) and recirculation zones (Fig. 2). Given that colorant diffu-
ing boundary layer thickness and mass transfer coefficients with those sion was neglected, the flow distribution using the videocamera was
provided in the simulation. The remaining characteristics of the solution easily registered.
which was used to conduct the experiments are the same as those used The data obtained from the fluid velocity at the center of the reactor
in the simulation. exhibit a laminar flow; which was confirmed calculating the Reynolds
Mass transfer coefficients were calculated using the Reynolds num- Number using
ber (Re) and replacing the Chilton - Colburn equation (Eq. (1)) which is
 
calculated jD, known as the Chilton - Colburn factor and finally the of 2BS
u ρ
mass-transfer coefficient (km) is calculated using Eq. (2). BþS
Re ¼
η  
2  0:04  0:006 3
 −1 = 0:0014 m s−1   994:4 kg m
l 0:04 þ 0:006
jD ¼ 1:85Re− =3
2 3
ð1Þ ¼ ¼ 16:4
de 8:82  104 Pa s−1

where, u is the flow velocity, B is the width of the cell, S is the height and
jD u ρ is the density.
km ¼
2 ð2Þ
υ =3 Assuming that the flow distribution is identical in all compartments,
D
due to duplicate geometries, same rate flows and pumps, the simulation
of only one of the compartments describes the other sections of the
where de is the hydraulic diameter, u is the linear velocity, υ is the kinet- reactor.
ic viscosity and D is the diffusion coefficient. km data obtained from Eq. The trajectories shown on Fig. 2 present a highly deviated hydrody-
(2) were compared with experimental simulated for different flow namic behavior, meaning that, there is a high percentage of the reactor
rates. being unused due to the stagnant zones, therefore affecting the

Fig. 2. Digital Image Analysis of the fluid flow frame by frame.

130 R. Enciso et al. / Desalination 408 (2017) 127–132

performance of the diluted compartment due to concentration polariza-

tion among being the principal cause for parasitic process.
In order to analyze the hydrodynamic results obtained by CFD, the
theoretical flow lines were calculated at 20 mm from the bottom of
the reactor and at the center of the compartment. These data, once plot-
ted, show the channeling, stagnant and recirculation volumes according
to their positions in the reactor. The Fig. 3 shows a clear accordance be-
tween the simulated and the experimental results, which were first
confirmed qualitatively and then quantitatively. The velocity profiles
shows that the turbulent description with k-e model is a flat velocity
profile, that is the expected profiles for flows in channels, whereas, the
laminar flow is more alike the experimental velocity measurements.
Plotted data show a very similar distribution between the two
results, small deviations can be attributable to the dye diffusion for
such low velocities, although the overall velocity profile is consistent
in the present zones and variations through the hole width of the
compartment.
Once the hydrodynamics validation was completed the electrical
field was simulated. When an electric field is imposed, ionic species mi-
grate towards electrodes. The profile of the velocity attained by an ion
depends on equipotential surfaces, which rest on cell geometry. The
combination of the chemical and electrical potential (diffusion-migra-
tion) is the base for the treatment of the ionic species. According to
Fig. 4 the cell geometry allows a well-defined an equipotential surfaces,
because the electrodes are of equal area occupying the entire opposite
sides of the cell.
The current distribution was calculated with the data of potential Fig. 4. Equipotential surfaces.
distribution. Fig. 5 shows an almost uniform distribution for the primary
current distribution. A slight increase is present in the zones that corre-
spond to the stagnant sections and a slight decrease at the inlet and out-
let of the reactor. These high current density zones have undesirable
consequences such as heating of the materials, degradation of the insu-
lating material, change on the electrical properties, voltage losses, and a
depleting in the lifetime of the electrochemical materials. The data form
the current distribution show a high uniformity in its density distribu-
tion throughout all the electrodialysis cell compartments, this will
have a direct impact in the current efficiency and the mass transfer
distribution.
Fig. 6 shows that mass transport concentration gradients follow the
patterns of the flow; the highest gradients are present on the zones with
the highest velocities and the smaller gradients are present in the zones
with the low velocities. On the diluted compartment a diminish on the
concentration of the potassium ion is visible since the fluid enters the
reactor, and tends to zero on the stagnant zones where the ion is

Fig. 3. Comparison between simulated (laminar and k-epsilon) and experimental velocity
profiles. Fig. 5. Current density distribution.
 R. Enciso et al. / Desalination 408 (2017) 127–132 131

Fig. 6. Concentration gradients in the diluted compartment.

depleted and with the given hydrodynamic conditions in that point, a
continuous supply of ions for convection is almost null. At this point,
to continue with the operation, the limiting current density will be
achieved rapidly, and water dissociation will be the responsible for the
ion transport
At the same time, in the same way, an increase of the concentration
at the proximity of the membrane is observed and ions are transported
through the membrane from the bulk, leave the solution in the diluted
compartment depleted of ions, which cross the membrane and finally
concentrate on the adjacent compartment.
Fig. 7 shows for the concentrated compartment, a concentration
near 5 mol m−3 at the zones where the flow velocity is higher, and re-
gions with low concentrations where the fluid is almost stagnant and
the ions almost depleted.
In order to validate the simulated results, local mass transfer coeffi-
cients where calculated for 4 different flow velocities (from 0.002 to
0.008 m s−1) using the Chilton-Colburn factor equations and presented
in Fig. 8 where an average deviation of 20%in the measurements is ob-
tained and attributable to the diffusion of the colorant used as a tracer
which leads to inaccurate measurements at low fluid velocities. In this
case, a good correlation was not achieved due to technical limitations
such as the accurate measurement of the fluid position due to diffusion
coefficient of the dye, nonetheless, the compared values have a minimal
discrepancy between experimental and theoretical results.
4. Conclusions
The present work validates the mathematical model and the simula-
tion as useful tools to analyze and comprehend the phenomena affect-
ing the performance of an electrodialysis reactor.
The methodology for the study of the experimental hydrodynamic
behavior using Digital Image Analysis offers an easy, cheap and relative-
ly simple way to obtain the necessary data of the fluid flow. The use of
dextran blue as tracer allows obtaining clear and defined images of
the fluid trajectories easily registered by the camera minimizing the ef-
fect of the diffusion of the dye.
The theoretical analysis of the Reynolds number and the study of the
velocity profiles using different turbulence model clearly shows that
this reactor operates under a laminar flow conditions. In case that the
reactor operates at the turbulent flow region the same methodology
can be applied using the k-e model to describe the fluid flow.

Fig. 7. Concentration gradients in the concentrated compartment.

132 R. Enciso et al. / Desalination 408 (2017) 127–132

1.2E-03 Acknowledgements
Simulated
1.0E-03 Experimental The authors gratefully acknowledge to Dra. Nubia V. Arteaga Larios
her support in the laboratory. R Enciso thanks to CONACyT the scholar-
ship granted number 360120.
8.0E-04
km (cm s-1)

Appendix A. Supplementary data

6.0E-04
Supplementary data to this article can be found online at http://dx.
4.0E-04
doi.org/10.1016/j.desal.2017.01.015.

2.0E-04
0.0E+00
2.0E-03 3.0E-03 4.0E-03 5.0E-03 6.0E-03 7.0E-03 8.0E-03
u (m s-1)
Fig. 8. Comparison between simulated and experimental mass transfer coefficients.
CFD results for the fluid distribution inside the reactor match the ex-
perimental results, allowing the capture of stagnant, recirculation and
high velocity zones. These results, provide de necessary validation to
continue with the subsequent simulations.
Current density distribution exhibits a uniform behavior on the cen-
tral portion with the characteristically rises on the borders.
For the mass transfer simulations, the results are consistent with the
hydrodynamic, in the sense that, the mass transfer is higher on the high
velocity zones than in dead zones. Mass transfer coefficients obtained
for the different velocities present a minimal deviation (being the ex-
ception the low velocity zones) when compared with the experimental
ones; which allows the validation of the mass transport on the electro-
dialysis reactor.
The experimental validation of the results allows the pursuit in fu-
ture studies to improve the performance of the reactor either by im-
proving the flow distribution through turbulence promoters, making
changes to the geometry or by varying the flow parameters.
References
[1] S. Koter, A. Warszawski, Electromembrane processes in environment protection,
Pol. J. Environ. Stud. 9 (2000) 45–56 (http://www.pjoes.com/pdf/9.1/45-56.pdf).
[2] A. Deghles, U. Kurt, Treatment of tannery wastewater by a hybrid
electrocoagulation/electrodialysis process, Chem. Eng. Process. 104 (2016) 43–50
(http://dx.doi.org/10.1016/j.cep.2016.02.009).
[3] L. Alvarado, C. Chen, Electrodeionization: principles, strategies and applications,
Electrochim. Acta 132 (2014) 583–597 (http://dx.doi.org/10.1016/j.electacta.2014.
03.165).
[4] X.H. Sun, H.X. Lu, J.Y. Wang, Brackish water desalination using electrodeionization
reversal, Chem. Eng. Process. 104 (2016) 262–270 (http://dx.doi.org/10.1016/j.cep.
2016.03.014).
[5] R. Enciso, I. Rodríguez-Torres, Improvement of the performance of the hydrodynamics
in the diluted compartment of an electrodeionization reactor, In: Symposium in Honor
of Professor Ralph E. White, Proceedings of the 226th Meeting of The Electrochemical
Society, 2014 October 5–9, The Electrochemical Society, Cancun (Mexico), 2014.
[6] A.N. Colli, J.M. Bisang, Evaluation of the hydrodynamic behavior of turbulence pro-
moters in parallel plate electrochemical reactors by means of the dispersion
model, Electrochim. Acta 56 (2011) 7312–7318 (http://dx.doi.org/10.1016/
electacta2011).
[7] S. Strahamann, Operating Principles of Electrodialysis and Related Processes: In Ion
Exchange Membrane Separation Processes, first ed. Elsevier, 2004.
[8] S. Pawlowski, P. Sistat, J.G. Crespo, S. Velizarov, Mass transfer modeling of desalina-
tion through an electrodialysis cell, J. Membr. Sci. 471 (2014) 72–83 (http://dx.doi.
org/10.1016/j.desal.2014.12.008).
[9] Z. Zourmand, F. Faridirad, N. Kasiri, T. Mohammadi, Mass transfer modeling of desa-
lination through an electrodialysis cell, Desalination 359 (2015) 41–51 (http://dx.
doi.org/10.1016/j.desal.2014.12.008).
[10] M. Turek, K. Mitko, Residence time distribution of the electrodialyzer under electric
field conditions, Desalination 342 (2014) 139–147 (http://dx.doi.org/10.1016/j.
desal.2013.11.042).
[11] H.K. Versteeg, W. Malalasekera, An Introduction to Computational Fluid Dynamics,
The Finite Volume Method, first ed.Longman, London, 1995.