A well-defined comparative study between stirred dead end and circular crossflow for microfiltration of china clay suspensions has been undertaken. The comparisons have been made with respect to convective mass transfer coefficients, permeation and rejection rates, and energy consumption. Similar operating and hydrodynamic conditions were implemented for the comparison. According to our experimental data the circular crossflow module was proven to perform better as compared with the stirred dead end system due to the higher mass transfer coefficients, higher permeation rates and lower energy consumption. The mass transfer coefficients observed are comparable to those previously found in vortex flow filtration and dead end flow filtration. The presence of Dean vortices in the circular crossflow module promotes flow instabilities in the curved channel flow path which reduce the concentration polarization effect during the filtration process. The concentration polarization effect however deteriorated due to solute build up (high solute concentration at the membrane surface) and decrease of the shear stress, i.e., the particle lift forces on the membrane surface. This resulted in deposition of particles on the membrane surface. In terms of energy consumption, for the same energy cost the limiting flux reached in circular crossflow was found to be higher than in the stirred dead end unit.

INTRODUCTION

Microfiltration (MF) is regarded as one of the oldest separation techniques among the pressure-driven membrane separation processes (Strathmann et al. 2011). Both MF and ultrafiltration (UF) membranes have been used extensively for the removal of particles, turbidity and microorganisms for water treatment (Gray et al. 2011; Shamsuddin et al. 2016). However, membrane fouling is a major impediment to membrane efficiency and it results in the reduction of membrane performance (Gao et al. 2011; Guo et al. 2012; Kochkodan et al. 2014). Despite the vast efforts to reduce the effect of membrane fouling by improving membrane properties, optimizing operating conditions and pre-treatment of feed water, fouling is unavoidable (Costa et al. 2006). Improved hydrodynamic conditions such as manipulating shear rates on membrane surfaces, improved design of the membrane modules, and induced flow instabilities are other useful methods in overcoming membrane fouling and concentration polarization (Jaffrin 2012).

Researchers have discussed various ways to induce these flow instabilities such as Taylor (Kroner & Nissinen 1988; Belfort et al. 1993; Park et al. 1994) and Dean vortices (Nunge & Adams 1971; Srinivasan & Tien 1971; Manno et al. 1998; Kaur & Agarwal 2002). Taylor vortices, which result from a rotating annular filter, were found to be one of the most successful techniques in reducing concentration polarization effects and membrane fouling. Both vortices have similarities in principle and use centrifugal forces to give rise to secondary flows which disrupt solute build-up on membrane surfaces, thus reducing concentration polarization and increasing permeation rates. However, Taylor vortices require substantially more energy than stationary Dean vortices. Hence, Taylor vortices have limited potential for upscaling as compared to Dean vortices. Experimental investigations undertaken by Belfort (1997), Belfort et al. (1993), Mallubhotla & Belfort (1997), Mallubhotla et al. (1995) and Mallubhotla et al. (1997) proved that Dean vortices effectively improve membrane filtration performances using curved channel modules. Kaur & Agarwal (2002) were the first researchers, to the best of our knowledge, who studied the effects of Dean vortices on filtration performance involving UF of protein suspensions in a circular thin flow channel module. They have experimentally calculated the mass transfer coefficients which were found to be higher than classical filtration models by a factor of 7–10.

The objective of this paper is to study the effects of Dean vortices on reducing concentration polarization and membrane fouling, and an increase of permeation fluxes in the case of MF of china clay suspensions through the study of hydrodynamics. While a significant amount of research has been done for reverse osmosis and UF which were widely used for desalination and removal of natural organic matter, MF for china clay particles has drawn less attention. Also, little research has been done into the influence of membrane configurations on the filtration performance. Thus, in this study, a comparison between circular crossflow and stirred dead end flow is attempted. Hydrodynamic conditions such as shear stress on the membrane wall which determines the mass transfer coefficients of particles needs to be investigated. As Becht et al. (2008) explained it is essential to define hydrodynamic conditions in great detail without leaving out the importance of similar operating conditions during comparison experiments. Hence, the aim of this paper is to investigate the hydrodynamic conditions of the two set-ups, i.e. circular crossflow and stirred dead end flow, using china clay suspension as the contaminant, and MF membranes for the filtration processes as detailed below.

EXPERIMENTAL PROCEDURE

Materials

MF experiments were performed with a mixed cellulose ester membrane (GSWP09000) consisting of cellulose acetate and cellulose nitrate which has an average pore size of 0.22 μm (Merck Millipore, Darmstadt, Germany). According to the manufacturer, the membranes are hydrophilic with thickness and porosity of 150 μm and 75%, respectively. A view of the clean membrane sample pictured using a scanning electron microscope (SEM) is shown in Figure 1(a). The particle size distribution of clay particles (Sigma-Aldrich, Dorset, UK) was evaluated using a Malvern-Sizer laser light scattering instrument (Malvern Instruments, Malvern, UK) and the result is shown in Figure 1(b).
Figure 1

(a) Membrane sample of 0.22 μm mixed cellulose ester, (b) particle size distribution of clay particles.

Figure 1

(a) Membrane sample of 0.22 μm mixed cellulose ester, (b) particle size distribution of clay particles.

Preparation of sample filtration

Prior to filtration experiments, the membrane was soaked in deionized water for 1 hour with the water changed every 20 minutes in order to remove any wetting agents. Measurement of pure water fluxes for each clean membrane was carried out. The ionic strength was adjusted to 0.01 M (0.585 g/l) by adding sodium chloride (NaCl), purchased from Sigma-Aldrich (Dorset, UK), into the china clay suspensions. In order to produce a homogeneous mixture, prior to the experiments the suspension was placed in an ultrasonic water bath for approximately 30 minutes at a temperature of 22 ± 2 °C. The pH of the china clay suspensions was adjusted to the selected pH values by adding various amounts of hydrochloric acid (HCl) or sodium hydroxide (NaOH), which were bought from Sigma-Aldrich (Dorset, UK), into the suspensions. The turbidity of the prepared suspensions was measured by a turbidity meter (model 20000; HF Scientific, Fort Myers, USA).

Membrane filtration apparatus

For the comparison of experiments two different configurations were used: a circular crossflow module manufactured by Amicon (Massachusetts, USA) and a stirred dead end apparatus (model XFUF07601) purchased from Merck Millipore (Darmstadt, Germany).

A schematic of the experimental setup is shown in Figure 2. The stirred dead end system (Figure 2(a) and 2(c)) has a fixed volume of 300 ml. The cell has an effective filtration surface area of 40 cm2 with a diameter of 76 mm. The feed reservoir was agitated by a flat blade paddle impeller (65 mm diameter and 9 mm height). Prior to filtration experiments the feed suspension was added into the feed reservoir. The membrane was placed at the bottom of the filtration cell while the pressure from the nitrogen cylinder was monitored by a pressure gauge and controlled by a pressure regulator (model 8286; Porter Instrument Co., Hatfield, USA). The speed of the flat blade paddle impeller was measured using a digital tachometer (Shenzhen Ever Good Electronic Co. Ltd, Shenzhen, China).
Figure 2

Schematic diagrams of filtration apparatus for (a) stirred dead end module, (b) circular crossflow module, and (c) side views of circular crossflow cell and stirred dead end flow, respectively.

Figure 2

Schematic diagrams of filtration apparatus for (a) stirred dead end module, (b) circular crossflow module, and (c) side views of circular crossflow cell and stirred dead end flow, respectively.

Figure 2(b) and 2(c) shows the circular crossflow module with an inside view of the filtration cell. The module has a feed volume of 600 ml and 40 cm2 effective filtration surface area. Both the feed and the retentate were recycled back into the feed reservoir at room temperature in order to maintain a constant suspension concentration throughout the filtration experiment. Figure 2(c) shows a flow pattern of the suspension in a circular channel over the membrane surface. There were three spirals with radii from 1 cm to 4.1 cm, with a channel spacing of approximately 1 cm. The spiral channel has the following dimensions: length (760 mm), width (9.5 mm) and height (0.38 mm) according to the manufacturer. The feed reservoir was pressurized by nitrogen which was adjusted to a predetermined pressure using manually operated valves. A pressure indicator was used to monitor pressure inside the feed vessel. Calibration of pressure gauges was conducted for both modules by validating the gauges with a precise gauge when changing pressure values. A new clean membrane was used and pre-treated for every new set of experiments. All experiments were carried out at room temperature (22 ± 2 °C). Permeate collection was taken at 1 minute intervals.

THEORY

In the circular flow module the difference in pressures between the internal and external walls of the circular channel flow gives rise to secondary flows known as Dean vortices. This phenomenon was shown to exist in such a module above a critical Reynolds (Re) number (Kaur & Agarwal 2002). Equation (1) can be used to calculate the Dean (De) number in the curved channel (Dean 1928): 
formula
1
where is the Re number above the critical Renumber (approximately 33–45), which was found experimentally by Brewster et al. (1959); is the equivalent hydraulic diameter calculated to be 0.0745 cm; and is the diameter of curvature of the channel path which has been calculated to be 4.51 cm (Shamsuddin et al. 2015).
As mentioned earlier, it is essential to have similar operating conditions for comparison but one must not leave out the importance of hydrodynamic conditions in order to satisfy the purpose of comparison (Shamsuddin et al. 2015). Therefore the calculation of shear stress in the circular flow system was made according to Becht et al. (2008), i.e., by solving the force balance across the membrane: 
formula
2
where is the transmembrane pressure, and is the length of the membrane channel (= 760 mm). A predetermined filtration pressure of 0.1 bar with a cross flow velocity of 1.156 m/s resulted in a flow profile pattern which corresponds to a Reynolds number of 867 and a shear stress of approximately 1.27 Pa.
However, a similar calculation for the case of stirred dead end filtration is not straightforward. According to Kosvintsev et al. (2005) the filtration cell has to be divided into two regions i.e. an inner region and an outer region. At the critical radius of the flat blade paddle impeller, the shear stress is highest but then decreases as it reaches the outer region. Therefore, in order to calculate the shear stress across the whole membrane, an average value of both the inner and outer regions should be calculated. Kosvintsev et al. (2005) developed the following correlation to find the critical radius: 
formula
3
where is the diameter of the flat blade paddle impeller, is the diameter of the filtration cell, h is the height of the flat blade paddle impeller, is the Reynolds number for the flat blade paddle impeller, and is the number of flat blade paddle impellers used.
The Reynolds number for both modules i.e. circular crossflow and stirred dead end can be calculated using Equation (4) and Equation (5), respectively: 
formula
4
 
formula
5
where is the dynamic viscosity of the fluid, is the density of the fluid, u is the fluid velocity, is the radius of the filtration cell, and is the angular velocity. Equation (5) was also used to calculate .
Shear stresses on the inner and outer regions are given by Equation (6) and Equation (7), respectively: 
formula
6
 
formula
7
where is the momentum boundary layer .

The stirred dead end module has a critical radius of 2.37 cm. In order for the system to achieve a similar shear stress as in the circular flow module, the flat blade paddle impeller requires a rotation speed of 145 rpm, which equals the shear stress of 1.27 Pa.

Cussler (2009) defined the mass transfer coefficient as the resistance to diffusion rate constant for solute movement in the boundary layer at the solid and liquid interface. The mass transfer coefficient was calculated according to the concentration polarization model proposed by Zydney & Colton (1986), Colton et al. (1975), and Blatt et al. (1970). The diffusion coefficient is defined as the ratio of molar flux and the driving force, and is determined by the Stokes-Einstein equation (Einstein 1905): 
formula
8
where is the Boltzmann constant (1.38 × 10−23 m2 kg s−2 K−1), T is the operating temperature in Kelvin, and is the average radius of china clay particles. Hence, the diffusion coefficient D according to Equation (8) was found to be .
The theory for calculation of the mass transfer coefficient for the circular crossflow module has been explained elsewhere (see e.g., Kaur & Agarwal 2002). The following Sherwood correlation was developed from our experimental results to describe the mass transfer coefficient for the circular crossflow module i.e. mass transfer of solutes from the membrane interface into the bulk phase, : 
formula
9
where is the Schmidt number which is the ratio of viscous diffusion rate and molecular diffusion rate .
The mass transfer coefficient for the stirred dead end system can be obtained from the typical mass transfer correlations. The mass transfer coefficient, , in the stirred dead end cell was obtained from the following Sherwood, , correlation (Mehta & Zydney 2006): 
formula
10

RESULTS AND DISCUSSION

Comparison of circular crossflow module and stirred dead end module

Filtration performance

Pure water fluxes of six clean cellulose ester membrane samples were measured for the investigation of hydraulic membrane resistance under a constant transmembrane pressure of 0.05 bar for both the circular crossflow system and stirred dead end system. Figure 3 illustrates total hydraulic resistances of membranes for both modules. The flux can be related to the total hydraulic resistance according to Darcy's law. Pure water flux measurement in the circular crossflow system (ranges between 550–650 l/hr.m2) is much higher than that in the stirred dead end system of approximately 120 l/hr.m2. Hydraulic resistance has an inversely proportional relationship with flux according to the following equation: 
formula
11
where J is the permeate flux , is the transmembrane pressure , is the dynamic viscosity , and is the total hydraulic resistance .
Figure 3

Total hydraulic resistances provided by membrane samples for circular crossflow module and stirred dead end module.

Figure 3

Total hydraulic resistances provided by membrane samples for circular crossflow module and stirred dead end module.

The differences in total hydraulic resistances for both modules are shown in Figure 3. Total hydraulic resistance in the circular flow module is much lower than in the stirred dead end. This is attributed to the effect of Dean vortices in the circular crossflow module. The flow pattern changes from typical laminar flow into an unstable laminar flow called Dean vortices when a fluid flows in the curved channel path at a Reynolds number above the critical Reynolds number. As a result of flow instabilities the resistance becomes lower according to Winzeler & Belfort (1993). The absence of such flow instabilities in the stirred dead end system results in much higher hydraulic resistance.

All filtration experiments were carried out with a suspension concentration of 0.4 g/l, ionic strength of 0.01 M, and the filtration pressures of 0.1 bar and 0.05 bar, respectively, for direct comparison between the circular flow module and the stirred dead end module The flow profile of the circular flow module is laminar which corresponds to the Reynolds number of 867 according to Equation (4). The crossflow velocity has been calculated as 1.156 m/s. The Reynolds number of the stirred dead end system was found to be 21,352 (turbulent flow) using Equation (5). Instead of keeping the Reynolds number uniform for both systems, the shear stresses on top of the membranes were made equal in order to maintain the entire operating conditions consistent for comparison. Equations (2)–(7) were used to calculate the shear stresses for both modules which were equal to 1.27 Pa.

The limiting flux for the circular crossflow module is six times greater than that for the stirred dead end module (Figure 4). After 25 minutes the flux varied within ±5%, hence, a steady state value was reached in the case of the circular crossflow module. A steady state flux was also reached in the case of the stirred dead end module, however, much faster (in less than 10 minutes) and of much lower value (Figure 4). This is attributed to the effect of Dean vortices present in the circular crossflow module, which depolarized solute build-up near the membrane interface: due to the higher wall shear stress particles were removed from the membrane surface. This resulted in intensive mixing between the boundary layer and the bulk phase (Bubolz et al. 2002). The formation of Dean vortices in the circular flow module slowed down formation of the steady state accumulation of solutes on the membrane surface in the early stage of the filtration process. Hence, the presence of Dean vortices results in improvement of mass transfer of solute from the membrane surface into the bulk solution.
Figure 4

Permeate fluxes for the circular crossflow module and the stirred dead end module.

Figure 4

Permeate fluxes for the circular crossflow module and the stirred dead end module.

Mass transfer coefficients

The mass transfer coefficients calculated for the circular crossflow system are in the range between 1.19 × 10−6 and 3.66 × 10−6 m/s calculated according to Equation (9). The mass transfer coefficients for the stirred dead end system are calculated according to Equation (10) and are found in the range between 1.12 × 10−7 and 4.31 × 10−8 m/s. According to Muller et al. (2003), for dead end filtration typical mass transfer coefficients were found to be about 5 × 10−8 m/s, whereas for crossflow filtration the range was 1 × 10−6 to 5 × 10−6 m/s, and for vortex flow filtration the range was 0.5 × 10−5 to 4 × 10−5 m/s. Hence, from our results the mass transfer coefficients for circular crossflow were comparable with those obtained for vortex flow filtration. We concluded that it was the presence of Dean vortices in the circular crossflow module that resulted in an improved mass transfer coefficient. The circular crossflow module showed better performance than the stirred dead end module as shown in Figures 3 and 4 under similar operating and hydrodynamic conditions. In the stirred dead end module the absence of such vortices led to a rapid build-up of solutes on the membrane surface thus the concentration polarisation effect took place. Although there was a stirrer to minimize the solute build-up it still could not reduce the concentration polarization effect as the filtration process progressed.

Observed rejection coefficients were found to be close to those in Table 1. The true or actual rejection percentage can be calculated using the following formula using the film model for concentration polarization (Blatt et al. 1970): 
formula
12
Table 1

Observed rejection coefficient for circular crossflow and stirred dead end modules

ConfigurationPermeate turbidity (NTU)Clay concentration in permeate (g/l)Observed rejection coefficient
Circular crossflow 0.35 ± 0.04 0.001302 ± 0.05 0.996744 ± 0.06 
Stirred dead end 0.91 ± 0.06 0.003317 ± 0.05 0.991708 ± 0.03 
ConfigurationPermeate turbidity (NTU)Clay concentration in permeate (g/l)Observed rejection coefficient
Circular crossflow 0.35 ± 0.04 0.001302 ± 0.05 0.996744 ± 0.06 
Stirred dead end 0.91 ± 0.06 0.003317 ± 0.05 0.991708 ± 0.03 

As the filtration process progressed, the observed rejection coefficients changed very slightly and lay within 3 to 6% variation as shown in Table 1. The true rejection percentages were calculated for both modules and were close to one. High solute concentration near the membrane surface led to the diffusion of the solute component in the opposite direction i.e. to the bulk. Therefore, concentration polarization did take place on the membrane surface but was more severe in the case of the stirred dead end module as seen in Figure 4. The reason why such a phenomenon took place in the circular crossflow module was because of the lower wall shear stress of 1.27 Pa and the high concentration of solute used (0.4 g/l). This resulted in a decreasing influence of the effect of Dean vortices as filtration progressed because, due to rapid solute build up and lower shear stress, the particle lift forces decreased thus resulting in a deposition of solutes on the membrane surface. Therefore, it is very important to search for the optimum operating and hydrodynamic conditions, i.e. higher wall shear stress (high operating pressure) and lower solute concentration might be desirable.

We investigated the influence of concentration of china clay particles on the flux decline in the circular crossflow system and the results are shown in Figure 5. The crossflow velocities and filtration pressure are kept the same for all the experiments. Generally, it can be seen from Figure 5 that the permeate flux decreases with the increasing solids concentration of the feed suspension. This observation is consistent with that of Hwang & Sz (2011).
Figure 5

Effect of suspension concentrations of 0.2, 0.4, and 0.6 g/l on permeate flux decline for the circular crossflow module.

Figure 5

Effect of suspension concentrations of 0.2, 0.4, and 0.6 g/l on permeate flux decline for the circular crossflow module.

SEM images of the top surface of the membranes were obtained at solids concentrations of 0.2 and 0.6 g/l in order to investigate the situation of membrane fouling after the MF process (Figure 6). At the higher concentration, there is a tendency for more china clay particles to accumulate on the surface of the membrane which leads to the formation of a cake layer. As seen from Figure 6, the higher the suspension concentration, the thicker the cake layer is since china clay particles were deposited on the membrane surface at a fixed suspension volume. The increasing thickness of the cake layer contributes to the resistance which confirms the observation mentioned above.
Figure 6

SEM images of membrane surfaces after the filtration process at different concentrations: (a) 0.2 g/l and (b) 0.6 g/l.

Figure 6

SEM images of membrane surfaces after the filtration process at different concentrations: (a) 0.2 g/l and (b) 0.6 g/l.

The typical impact of the filtration pressure on the flux decline is shown in Figure 7, which is consistent with Hwang & Sz (2011). However, the permeate fluxes also decrease more rapidly with increasing filtration pressure. This phenomenon is very significant with the MF of large particles like the china clay particles used in the experiments (Tarleton & Wakeman 1994). Figure 7 shows that the permeate fluxes are directly proportional to the filtration pressure.
Figure 7

Effect of filtration pressures of 0.01 bar, 0.05 bar and 0.1 bar on permeate flux decline. Suspension concentration is 0.4 g/l.

Figure 7

Effect of filtration pressures of 0.01 bar, 0.05 bar and 0.1 bar on permeate flux decline. Suspension concentration is 0.4 g/l.

In order to further verify the trends in terms of flux decline (Figure 7) caused by the change in the filtration pressure, SEM images were collected for measuring the thickness of the cellulose ester membrane samples after the filtration process. The thickness of the membrane is 150 μm as indicated by the manufacturer. As seen from Figure 8 different filtration pressures did not have a significant influence on the thickness of the membranes in the investigated range of transmembrane pressures. The thicknesses are the same (about 150 μm after the experiment) which indicates that the membrane characteristics remain constant for each membrane employed.
Figure 8

SEM images of the cross section of each membrane after the filtration process at different filtration pressures: (a) 0.01 bar, (b) 0.05 bar, and (c) 0.1 bar.

Figure 8

SEM images of the cross section of each membrane after the filtration process at different filtration pressures: (a) 0.01 bar, (b) 0.05 bar, and (c) 0.1 bar.

The decreased influence of Dean vortices was observed as filtration progressed due to rapid solute build up and lower shear stress which decreased the particle lift forces thus resulting in deposition of solutes on membrane surface. Therefore, it is very important to search for optimum operating and hydrodynamic conditions, i.e. higher wall shear stress (high operating pressure, greater than 0.1 bar) and a more dilute solute concentration might be desirable (less than 0.2 g/l), in order to benefit from the maximum effect of Dean vortices.

Energy consumption

Energy consumption is an important aspect in considering the feasibility of an industrial application. The energy consumption for both systems was calculated and compared under similar operating conditions. According to Manno et al. (1998), the energy dissipated per unit volume of permeate, E can be calculated using the following formula: 
formula
13
where Q is the feed flow rate, is the permeate flow rate, is the pressure difference between the inlet and outlet, and is the transmembrane pressure.
Figure 9 shows the limiting ratio (circular crossflow system to stirred dead end system) as a function of energy dissipated. For the same energy rate, the limiting flux reached in the circular crossflow system is always higher than in the stirred dead end system. Observations on energy consumption by any modules with Dean vortices effects were also made by Moulin et al. (1999) and Manno et al. (1998). They concluded that the presence of Dean vortices secondary flows, even at a fixed amount of energy dissipated, would result in more permeation fluxes compared to other conventional modules. With regard to the energy calculations both set-ups, i.e. circular crossflow and stirred dead end, operated at quite low pressures (0.1 bar and below) for the filtration experiments, hence, it was predicted that the amount of energy dissipated would be low as well. Also, the time required for the filtration experiment to complete at a filtration pressure of 0.1 bar was less than 1 hour. This system is scalable and the scalability of this analysis will be presented elsewhere.
Figure 9

Ratio of limiting flux in the circular crossflow system to stirred dead end system as a function of dissipated energy.

Figure 9

Ratio of limiting flux in the circular crossflow system to stirred dead end system as a function of dissipated energy.

CONCLUSIONS

A comparative study between a stirred dead end system and a circular crossflow system in MF of china clay suspensions was carried out. Comparison was made with respect to convective mass transfer coefficients, permeation and rejection rates, and energy consumption. Similar operating and hydrodynamic conditions were implemented. From our experimental data, the circular crossflow module was proven to perform better when compared to the stirred dead end system due to the higher mass transfer coefficients, higher permeation rates and lower energy consumption. The mass transfer coefficients are comparable to studies previously done in vortex flow filtration and dead end flow filtration. The presence of Dean vortices in the circular crossflow module promotes flow instabilities in the curved channel flow path which reduce the concentration polarization effect during the filtration process. For the same energy cost, the limiting flux reached in the circular crossflow system is always higher than in the stirred dead end system. Hence, it is proven that less energy was consumed in the circular crossflow module than in the stirred dead end module with higher permeation rates. From the study of hydrodynamics of both set-ups, the mass transfer coefficients of particles could be determined.

ACKNOWLEDGEMENTS

The authors would like to thank the Government of Brunei for the PhD studentship granted to Norazanita Shamsuddin at Loughborough University.

REFERENCES

REFERENCES
Belfort
G.
1997
Coiled membrane filtration system
.
US Patent 5626758
.
Belfort
G.
Mikulasek
P.
Pimbley
J. M.
Chung
K. Y.
1993
Diagnosis of membrane fouling using a rotating annular filter. 2. Dilute particle suspensions of known particle size
.
Journal of Membrane Science
77
(
1
),
23
39
.
Blatt
W. F.
Dravid
A.
Michaels
A. S.
Nelsen
L.
1970
Solute polarization and cake formation in membrane ultrafiltration: causes, consequences, and control techniques
. In:
Membrane Science and Technology
(
Flinn
J. E.
, ed.),
Plenum Press
,
New York
, pp.
47
97
.
Brewster
D. B.
Grosberg
P.
Nissan
A. H.
1959
The stability of viscous flow between horizontal concentric cylinders
. In:
Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences
251
,
76
91
.
Bubolz
M.
Wille
M.
Langer
G.
Werner
U.
2002
The use of dean vortices for crossflow microfiltration: basic principles and further investigation
.
Separation and Purification Technology
26
(
1
),
81
89
.
Colton
C. K.
Henderson
L. W.
Ford
C. A.
Lysaght
M. J.
1975
Kinetics of hemodiafiltration. I. In vitro transport characteristics of a hollow-fiber blood ultrafilter
.
The Journal of Laboratory and Clinical Medicine
85
(
3
),
355
371
.
Costa
A. R.
Norberta de Pinho
M.
Elimelech
M.
2006
Mechanisms of colloidal natural organic matter fouling in ultrafiltration
.
Journal of Membrane Science
281
,
716
725
.
Cussler
E. L.
2009
Diffusion: Mass Transfer in Fluid Systems
.
Cambridge University Press
,
New York, USA
.
Dean
W. R.
1928
Fluid motion in a curved channel. In: Proceedings of the Royal Society of London. London: Series A, Containing Papers of a Mathematical and Physical Character, pp. 402–420
.
Gao
W.
Liang
H.
Ma
J.
Han
M.
Chen
Z. L.
Han
Z. S.
Li
G. B.
2011
Membrane fouling control in ultrafiltration technology for drinking water production: a review
.
Desalination
272
(
1
),
1
8
.
Gray
S. R.
Dow
N.
Orbell
J. D.
Tran
T.
Bolto
B. A.
2011
The significance of interactions between organic compounds on low pressure membrane fouling
.
Water Science & Technology
64
(
3
),
632
639
.
Guo
W.
Ngo
H. H.
Li
J.
2012
A mini-review on membrane fouling
.
Bioresource Technology
122
,
27
34
.
Jaffrin
M. Y.
2012
Hydrodynamic techniques to enhance membrane filtration
.
Annual Review of Fluid Mechanics
44
,
77
96
.
Kochkodan
V.
Johnson
D. J.
Hilal
N.
2014
Polymeric membranes: surface modification for minimizing (bio)colloidal fouling
.
Advances in Colloid and Interface Science
206
,
116
140
.
Kosvintsev
S. R.
Gasparini
G.
Holdich
R. G.
Cumming
I. W.
Stillwell
M. T.
2005
Liquid-liquid membrane dispersion in a stirred cell with and without controlled shear
.
Industrial & Engineering Chemistry Research
44
(
24
),
9323
9330
.
Kroner
K. H.
Nissinen
V.
1988
Dynamic filtration of microbial suspensions using an axially rotating filter
.
Journal of Membrane Science
36
,
85
100
.
Mallubhotla
H.
Belfort
G.
1997
Flux enhancement during Dean vortices microfiltration. 8. Further diagnostics
.
Journal of Membrane Science
125
,
75
91
.
Mallubhotla
H.
Luque
S.
Belfort
G.
1997
Membrane filtration with self-cleaning spiral vortices
. In:
Euromembrane '97, Progress in Membrane Science and Technology
(
Kemperman
A. J. B.
Koops
G. H.
, eds).
University of Twente
,
The Netherlands
,
385
387
.
Manno
P.
Moulin
P.
Rouch
J. C.
Clifton
M.
Aptel
P.
1998
Mass transfer improvement in helically wound hollow fibre ultrafiltration modules yeast suspensions
.
Separation and Purification Technology
14
,
175
182
.
Moulin
P.
Manno
P.
Rouch
J. C.
Serra
C.
Clifton
M.J.
Aptel
P.
1999
Flux improvement by Dean vortices: ultrafiltration of colloidal suspensions and macromolecular solutions
.
Journal of Membrane Science
156
(
1
),
109
130
.
Muller
C.
Agarwal
G.
Melin
T.
Wintgens
T.
2003
Study of ultrafiltration of a single and binary protein solution in a thin spiral channel module
.
Journal of Membrane Science
227
,
51
69
.
Nunge
R. J.
Adams
L. R.
1971
Reverse osmosis in laminar flow through curved tubes
.
Desalination
13
,
17
36
.
Shamsuddin
N.
Das
D. B.
Starov
V. M.
2015
Filtration of natural organic matter using ultrafiltration membranes for drinking water purposes: circular cross-flow compared with stirred dead end flow
.
Chemical Engineering Journal
276
,
331
339
.
Shamsuddin
N.
Das
D. B.
Starov
V. M.
2016
Membrane-based point-of-use water treatment (PoUWT) system in emergency situations
.
Separation & Purification Reviews
45
(
1
),
50
67
.
Srinivasan
S.
Tien
C.
1971
Reverse osmosis in a curved tubular membrane duct
.
Desalination
9
,
127
139
.
Strathmann
H.
Giorno
L.
Drioli
E.
2011
Introduction to Membrane Science and Technology
.
Wiley-VCH Verlag & Company
,
Dresden, Germany
.
Tarleton
E. S.
Wakeman
R. J.
1994
Understanding flux decline in crossflow microfiltration: part 2. Effects of process parameters
.
Trans IChemE
72
(
A
),
431
440
.