Separation and Purification Technology xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

Separation and Purification Technology

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

Analysis on the nonuniformity of cake formation in rotating-disk

dynamic microfiltration
Kuo-Jen Hwang ⇑, Su-En Wu, Yu-Ling Hsueh
Department of Chemical and Materials Engineering, Tamkang University, Tamsui, New Taipei City 25137, Taiwan, ROC

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

Article history: This study analyzes the nonuniformity of cake formation and filtration flux in rotating-disk dynamic
Received 29 September 2016 microfiltration of fine particles. Because the installation of a rotating disk in the filter chamber generates
Received in revised form 21 November 2016 high shear stresses on the membrane surface, a rotating-disk filter can be used to achieve light cake mass
Accepted 30 December 2016
and high filtration flux. The effects of operating conditions—such as disk rotation speed and transmem-
Available online xxxx
brane pressure—on the filtration flux, membrane fouling, and cake properties are discussed both exper-
imentally and theoretically.
Keywords:
Because the shear stresses generated by a rotating disk are not uniformly distributed on the membrane
Microfiltration
Rotating-disk filter
surface, the cake formations as well as the filtration flux are functions of position and rotating-disk struc-
Local cake properties ture. In this study, the distributions of fluid velocity and shear stress on the membrane surface were sim-
Solid-liquid separation ulated using computational fluid dynamics. Combined with experimental data concerning local cake
properties, such as thickness, porosity, and specific filtration resistance, the mechanism of cake formation
based on a force balance model was analyzed. The result shows that the local filtration flux increases in
tandem with critical particle size. The agreement between the calculated results and the experimental
data demonstrates the reliability of the proposed method. The relationship between local filtration flux,
critical particle size, and operation conditions can be expressed as a nonlinear equation. This method esti-
mates the local cake mass and filtration rate, not only for optimizing operating conditions to achieve the
best filtration performance but also for designing high-efficiency disk structures or filter modules.
Ó 2017 Elsevier B.V. All rights reserved.

1. Introduction as seen in the vibratory shear-enhanced process (VSEP) system

by New Logic (California, USA), the SpinTek module with rotating
Membrane separation technology occupies an important posi- membranes (California, USA), the dynamic microfiltration
tion in chemical, mining, food, energy, materials, biochemistry, rotating-disk system by Pall Corporation (New York, USA), and
electronic, and environmental industrial processes. It is widely the filtration systems with rotating disks and rotating membranes
used in solid–liquid separation operations to achieve high selectiv- by Bokela, Novoflow, and Canzler (Germany) [14]. Jaffrin [15] also
ity and low pollution. However, membrane fouling is often reviewed varying systems of dynamic shear-enhanced filtration,
unavoidable during the filtration process. Many researchers have which consist of controlling the membrane shear rate to maintain
devoted their efforts to the improvement of separation efficiency filtration by rotating the disks or vibrating the membranes.
by mitigating cake growth or membrane fouling, thus reducing Because the high shear stress caused by a rotating disk can
operation costs. Several enhanced dynamic microfiltration meth- effectively reduce membrane fouling to enhance filtration flux,
ods for reducing fouling have also been proposed. rotating-disk microfilters have gradually been introduced in micro-
In dynamic filtration, the increase in membrane filtration per- filtration and are increasingly applied in solid–liquid separation.
formance, which is caused by the shear stress at the membrane, Rotating-disk microfilter operating performance, with regard to
is created by a disk that rotates near fixed membranes, membranes factors such as filtration rate and power consumption, depends sig-
that rotate around a shaft within an enclosure [1–10], or azimuthal nificantly on experimental operative conditions and the geometric
and longitudinal vibrations of membranes [11–13]. Several compa- parameters of rotating disks [15]. Jaffrin et al. [16] and Rios et al.
nies have investigated the recent technique of dynamic filtration, [17] have described hydrodynamic parameters as the most crucial
factors in rotating-disk filters and compared the effects of opera-
⇑ Corresponding author. tive conditions—such as transmembrane pressure, disk rotation
E-mail address: [email protected] (K.-J. Hwang). speed, fluid viscosity, and solute concentration—on the permeate

http://dx.doi.org/10.1016/j.seppur.2016.12.049
1383-5866/Ó 2017 Elsevier B.V. All rights reserved.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
2 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx

Nomenclature

A correction factor (–) Rm filtration resistance of membrane (1/m)

B correction factor (–) Rt total filtration resistance (1/m)
C1–C4 correction factors (–) So specific surface area of particles (m s
2)
dp diameter of particles (m) t filtration time (s)
dp,av average diameter of particles (m) vt centrifugal velocity (m/s)
dp,c critical diameter of particles (m) wc mean cake mass (kg/m2)
Fg net gravitational force (N) wr local cake mass (kg/m2)
Fi net interparticle force (N) aav average specific filtration resistance (m/kg)
Fl inertial lift force (N) ar local specific filtration resistance (m/kg)
Fp normal drag force (N) e average cake porosity (–)
Ft tangential drag force (N) e (r) local cake porosity (–)
k Kozeny constant (–) j velocity coefficient (–)
Lc cake thickness (m) l fluid viscosity (kg/s m)
Lc(r) local cake thickness (m) c shear rate on the membrane surface (1/s)
DP filtration pressure (Pa) cm mean shear rate on the membrane surface (1/s)
qs pseudo-steady state filtration flux (m3/m2 s) qs particle density (kg/m3)
qs(r) local filtration flux (m3/m2 s) q fluid density (kg/m3)
Q inlet feed flow rates (L/h) sm mean shear stress on the membrane surface (Pa)
r radius of rotating disk (m) m fluid kinematic viscosity (m2/s)
r0 radius of chamber (m) x rotating speeds (rpm)
Rc average filtration resistance of cake layer (1/m)
Rc(r) local filtration resistance of cake layer (1/m)

flux. The literature suggests that an increasing rate in the permeate Li et al. [28] employed PVDF and PES to recover the linseed oil
flux governed by the shear rate is a result of increasing rotating from an O/W emulsion and obtained a recovery of 76% supernatant
speeds. oil. Rios et al. [17] considered different mean pore size of ceramic
Equipping the rotating disk or using large blades [16,18,19] and polymeric membranes for concentrated biodiesel production,
helps to eliminate or mitigate membrane fouling and produce and concluded that ceramic membranes with a mean pore size of
much higher permeate fluxes, as compared with conventional 2 lm offered complete rejection at low TMP. Shear enhancements
cross-flow [20] and dead-end filtration [21]. Zhang et al. [22] used in rotating-disk ultrafiltration (30, 50, and100 kD) were used to
0.2-lm MF and 20-kD UF membranes for concentrating leaf pro- pretreat dairy wastewater [34], and the results showed that com-
tein in alfalfa juice. Three filtration modules were utilized and their plete pore blocking and cake formation are the main fouling mech-
performances compared; the results show that the shear-enhanced anisms, but that a high shear rate can reduce membrane pore
membrane module (RDM) improved separation, flux, and mem- blocking and irreversible fouling.
brane cleaning. In previous work, we designed a rotating-disk filter chamber
Understanding the effects of process conditions—such as hydro- for microalgae concentration [35]. In order to achieve high fil-
dynamic parameters and types of blades—on filtration perfor- tration fluxes and relatively low power consumptions, the effec-
mance is an essential step toward achieving optimal operations, tive design and optimized operation of rotating-disk dynamic
and using computational fluid dynamics (CFD) to simulate the filter have been discussed by the specific energy, which is
velocity and shear stress distributions in a dynamic filtration sys- defined as the power consumption per unit volume of the
tem is an effective, economical, and time-saving approach received filtrate [35]. The results were in agreement with previ-
[23,24]. A perforated disk for the dynamic filtration of microalgal ous experimental data [36], and by considering the filtration
harvesting was designed by Kim et al. [25]. Fluid velocity and shear flux and power consumption, we showed that Type A2 and C
stress were evaluated by CFD simulation, and the results showed disks had optimal designs. The present study evaluates local
that a perforated disk had more favorable performance in shear flux behavior and cake properties. To examine flux decline, a
stress and filtration than an unperforated disk did. The average dynamic rotating-disk filter was compared with the conven-
shear stress was log-scale-linearly related to flux and fouling resis- tional cross-flow membrane filtration process. Given that this
tance. The perforated disk (rotating at 600 rpm) showed a 2.3-fold study’s focus is to understand the effect of rotating speed on
higher plateau permeate flux of 323 L/m2/h compared with the filtration performance and flux behavior, mechanism of cake
unperforated disk (143 L/m2/h). formation, local cake properties, and particle size distribution
In contrast to the conventional separation process, the rotating- were investigated in the filtration experiment. The velocity
disk microfilter has been alternatively investigated in various high distribution and shear stress on the membrane surface were
quality products [26–28,17,29–31]. Some research has investi- simulated using CFD. The fluid flows in the rotating-disk filter
gated membrane fouling in a microfiltration system determined were simulated under various operative conditions, and experi-
by the membrane properties. Bouzerar et al. [26] used nylon and ments of cake thickness and filtration flux were reasonably esti-
PVDF membranes to separate a calcium carbonate suspension. mated by the simulated shear stress distribution. These results
Valentina et al. [32] investigated the microfiltration of skim milk were useful for designing a new disk structure and module con-
to separate casein micelles from whey proteins using a modified figuration. The methods proposed by this study enable estima-
dynamic filtration pilot (MSD). Moulai-Mostefa et al. [33] com- tions of the local cake mass and filtration flux and optimize
pared the rotating-disk module and a VSEP pilot with ultrafiltra- process efficiency and filtration performance through operating
tion for separating oil-in-water (O/W) emulsions. conditions.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx 3

2. Materials and experiments 100 rpm to 500 rpm, the TMP ranged from 20 kPa to 100 kPa, and
feed flow rate was set to 3  10
6 m3/s. The filtration conducted
2.1. Materials and membrane at least three times under each set of conditions to verify
reproducibility.
Polymethyl methacrylate (PMMA) spherical particles manufac-
tured by Soken Chemical & Engineering Co., Ltd (Japan) were used 2.3. Basic filtration equations and cake properties
in the rotating-disk dynamic microfiltration experiments. The den-
sity of the particulate sample was 1210 kg/m3 and the mean diam- According to a modification of Darcy’s law, the filtration rate
eter of the particles was 7.6 lm. The particle size distribution can be expressed as the driving force divided by the resistances
ranged from 1.5 lm to 25 lm. The particles were suspended in of the filter cake and the porous medium in series. Therefore, the
deionized water to prepare a 0.25 wt% suspension; the pH of the overall microfiltration resistance can be expressed by the basic fil-
suspension was kept at 7.0 and the temperature at 20 °C. A filter tration equation:
membrane made of mixed cellulose ester (Catalog #:
DP DP
A010A142C) was manufactured by the ADVANTEC Co. (Japan). qs ¼ ¼ ð1Þ
The mean pore size of the membrane was 0.1 lm and the mem-
lðRt Þ lðRm þ Rc Þ
brane filtration resistance was 4.82  1011 m
1. where qs is the filtration rate, DP is the filtration pressure, l is the
viscosity of the fluid, and Rc, Rm, and Rt are the filtration resistances
2.2. Rotating-disk dynamic filtration and analyses due to cake formation, the cleanliness of the membrane, and the
overall filtration resistance, respectively. During filtration, solid par-
A cylindrical filter chamber of a rotating-disk dynamic filter ticles are carried by the fluid and deposited on the membrane sur-
made of stainless steel #304 was designed and built in our labora- face to form a filter cake. The cake resistance can be calculated by
tory [36]. The diameter and height of the chamber were both determining the product of the cake mass and the average specific
38 mm. Two vanes, each with a height of 10 mm, a width of filtration resistance [37]:
10 mm, and a thickness of 1 mm, were placed beneath a rotating Rc ¼ aav wc ð2Þ
disk with a diameter of 30 mm. A low circular inlet and a high cir-
cular outlet, both with diameters of 9 mm, were connected to the where wc is the cake mass and aav is the average specific cake filtra-
chamber for feed inflow and concentrate outflow. A circular mem- tion resistance. The experimental data of qs, DP, l and Rm were sub-
brane was installed on the porous bottom plate to serve as a filter stituted into Eq. (1) to calculate the value of Rc. The average specific
medium. The filtration area was 1.12  10
3 m2. cake filtration resistance was then estimated using Eq. (2) once the
A schematic diagram of the dynamic microfiltration system is cake mass was measured. The value of the cake mass can be calcu-
shown in Fig. 1. A suspension was prepared in the suspension tank lated from the material balance of the cake:
and agitated using a magnetic mixer. The suspension was pumped wc ¼ qs ð1
eÞLc ð3Þ
into the rotating-disk microfilter using a peristaltic pump. The feed
flow rate was controlled and measured using a rotameter, and the where qs is the particle density, Lc is the cake thickness, and e is the
TMP was adjusted using a needle valve and indicated on the pres- average cake porosity. The porosity was calculated from the mate-
sure gauges. The rotating disk was connected to a motor with a rial balance once the cake mass and cake thickness were measured
shaft, and the disk rotating speed was kept constant during each in experiments. The average specific filtration resistance can be
experiment. The concentrate was recycled back to the suspension expressed as:
tank and the filtrate was collected in a receiver. The increase in fil-
kSo  ð1
eÞ
2
trate weight was measured using a load cell and the data were aav ¼ ð4Þ
recorded on a computer during filtration. When an experiment qs  e 3
was terminated, the cake that formed on the membrane surface in which k is the Kozeny constant and So is the specific surface area
was observed and its dry mass was measured. The particle sizes of the particles. The incompressible cake was formed by rigid
were measured using the LA 910 laser particle sizer (HORIBA, PMMA particles during filtration by the authors’ pervious study
Ltd., Japan). Disk rotation speed and TMP were major operating [42]. For spherical particles under compact packing, the value of k
parameters. The disk rotation speed used in this study ranged from is equal to 5.0. However, because k is a function of conditions such
as packing porosity, particle shape, particle surface roughness, and
flow condition, it was appropriate to measure it in experiments.
The specific surface area of the particles, So, is a function of particle
size and cake porosity, and the value of So is equal to 6/dp for a
spherical monosized particle. In general, for a spherical particulate
Concentrate sample with a size distribution, the value of So can be estimated
from the mean diameter of the particles, dp,av. Once the values of
kSo2 and cake porosity were obtained, the average specific filtration
resistance of the cake could be estimated theoretically from Eq. (4).

Rotating disk 2.4. Local cake properties and filtration performance

Feed Vane
z The value of the local cake mass can be calculated from the
x Membrane material balance of the cake:
Support
wr ¼ qs ½1
eðrÞ  Lc ðrÞ ð5Þ

Permeate where e(r) is the local cake porosity and Lc(r) is the local cake thick-
ness. The local porosity was calculated from the material balance
Fig. 1. Schematic diagram of the dynamic microfiltration system. once the local cake mass and local cake thickness were measured

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
4 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx

in experiments. The local specific filtration resistance can be also 2.5.2. Normal drag force
expressed as: The normal drag force can be calculated by a modification of
Stokes’ law, because the Reynolds number in the filtration direc-
kfS0 ðrÞg2  ½1
eðrÞ
ar ¼ ð6Þ tion is very small under most conditions:
qs  feðrÞg3
F p ¼ 3pldp qs C 3 ð10Þ
The value of So(r) can be estimated from the critical diameter of
particles at radius r position, dp,c. where C3 is a correction factor.

2.5. Force analysis on depositing particles 2.5.3. Net gravitational force

The net gravitational force exerted on a particle can be given by:
Fig. 2 shows a schematic diagram of the analyzed system, which p
ðqs
qÞgdp
3
Fg ¼ ð11Þ
is near the membrane surface. The major forces include the tan- 6
gential drag force due to the rotation effect (Ft), the tangential drag
where qs and q are the densities of the particle and fluid,
force (Fh), the centrifugal drag forces (Fc), the normal drag force due
respectively.
to filtrate flow (Fp), the inertial lift force (Fl), the net gravitational
force (Fg), and the net interparticle force (Fi). These forces were
2.5.4. Inertial lift force
analyzed as follows.
The inertial lift velocity of the particle in a linear tangential flow
can be calculated by [38]:
2.5.1. Tangential drag force
   3
In the rotating-disk dynamic microfiltration system used in this 61c2m dp
vl ¼  ð12Þ
study, the velocity distribution of a flow in the filter channel could 576m 2
be expressed by CFD. The total drag force (Ft) is a function of the
centrifugal (Fc) and tangential (Fh) drag forces, while Fc and Fh are where m is the fluid kinematic viscosity and cm is the mean shear
the drag forces in the r and h coordinate directions of a rotating- rate at the membrane surface. Therefore, the inertial lift force can
disk dynamic microfiltration system: be evaluated by this modification of Stokes’ law:

F t ¼ f ðF c ; F h Þ ð7Þ F l ¼ 3pldp v l C 4 ð13Þ

The centrifugal drag force can be estimated by the following where C4 is a correction factor.
equation

p  v 2  2.5.5. Net interparticle force
Fc ¼ qs d3p  t
C1 ð8Þ Interparticle forces should be taken into consideration, because
6 r the net interparticle force plays a major role in the deposition and
where dp is the particle diameter, vt is the centrifugal velocity, qs is packing of submicron particles. According to the DLVO theory, the
the density of the particle, and C1 is a correction factor. According to net interparticle force can be estimated by summing the van der
the analysis by Lu and Ju [43], the tangential drag force can be esti- Waals and electrostatic forces. The detailed calculation is described
mated by a modification of Stokes’ law, that is, in [39]. For a submicron particle (dp < 1 lm), the major forces
include the net interparticle force and the drag forces in both the
3
F h ¼ C2 pld2p c ð9Þ tangential and normal directions. In general, the van der Waals
4 force is an attractive force, whereas the electrostatic force is repul-
where c is the shear rate and C2 is a correction factor. The shear rate sive. However, the inertial lift force and the gravitational force
generated by a rotating disk can be expressed by simulated data. increase in importance as particle size increases; therefore, as
dp > 10 lm, the main forces affecting particle deposition include
inertial lift force and gravitational force. This implies that particle
size plays a major role in particle deposition. Hwang et al. [40] con-
cluded that drag forces are dominant in the selective deposition of
SiO2 particles. The tangential drag force is at least one order of
magnitude greater than that in the filtration direction. The devia-
tions resulting from ignoring the gravitational, lift, and interparti-
cle forces in rotating-disk microfiltration were small.

3. Numerical simulation methods

The fluid flows in the filter chamber and the shear stress on the
x membrane surface were simulated using the CFD software Fluent.
r Fl To prepare for the simulation, the geometry of the filter chamber
F was designed and 3-dimensional unstructured tetrahedral/hybrid
meshes were created using Gambit software. The mesh numbers
were approximately 437,000. Fluid flowed into the chamber from
Fp+Fg+Fi Fc the inlet in the x direction and flowed out from the outlet near
the top plate. The fluid flows were simulated using a segregated,
y steady-state, 3-D implicit numerical solver supplied by Fluent
Permeate and the governing equations of motion were coupled using the
SIMPLE algorithm. The conditions of the solid walls were set as
nonslip. The boundary conditions included one uniform velocity
Fig. 2. Forces on depositing particles in dynamic microfiltration. inlet and two pressure outlets: a concentrate outlet and a filtrate

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx 5

outlet. The convergence criteria were that the residuals of continu- obtained in this study was 0.26–3.50 Pa. The simulated results of
ity and velocity be below 1.0  10
4. the shear stress distribution at various disk rotation speeds and
the experimental figures of cake growth during 1-h filtration are
shown in Fig. 4. In the simulated results, the shear stress was
4. Results and discussion approximately zero at the center, increased toward the chamber
walls, and then decreased drastically near the wall regions because
The velocity distribution was simulated with Fluent. The typical of the no-slip condition on the wall. This trend can be reasonably
results on x-planes are shown in Fig. 3. Because only two vanes explained from the fluid velocity profiles (Fig. 3). The shear stress
were equipped on the rotating disk, the vanes, inlet, and outlet distributions were slightly asymmetric because of the feed inlet
conduits were observed only on the x plane. The disk rotation and concentrate outlet locations. The shear stress near the feed
speeds ranged from 100 to 500 rpm and the feed flow rate was inlet was higher than that at the concentrate outlet because the
3  10
6 m3/s. The fluid velocity in the filter chamber was 0– feed inlet was located at a lower level. Comparing the results under
0.785 m/s under operating conditions. The analysis of fluid velocity those three conditions, the local shear stresses increased with the
distribution, shown in Fig. 3a, revealed that the fluid velocity near disk rotation speed. As for experimental results, the cake growth is
the membrane surface was substantially increased by installing the indicated in white to match the color of PMMA particles. This sug-
rotating disk. The fluid velocities exhibited high magnitudes in the gests that the shear stress produced by the rotating disk plays a
regions near the rotating disks and vanes, especially in the region major role in determining cake growth and structure. Cake forma-
between the vane edges and the membrane surface (the maximum tion can be effectively mitigated by increasing the disk rotation
values were observed near the rim of the rotating disk). In the speed, which indicates that higher shear stress is correlated with
region below the rotating disk, the velocity magnitude increased thinner cake thickness. Consequently, less cake is formed under
with an increasing radial position in this rotation system. By con- higher disk rotating speeds than under lower rotating speeds. Fur-
trast, low-velocity profiles existed near the axial center beneath thermore, the cake formation was considerably nonuniform on the
the disk. The shear stress on the membrane surface was estimated entire membrane surface. When the cakes formed under disk-
from the simulated velocity profile, which was the key factor rotation conditions, the cake in the central region thickened
affecting cake formation and structure. However, the velocity vec- because of low shear stress, thinned gradually along the radial
tors near the rotating disk had almost the same velocity as the disk direction until it reached a minimum, and then thickened between
[Fig. 3b]; this can also be seen in the literature [24]. the disk edge and the chamber wall. This distribution evidently
A comparison of the mean shear stress at the membrane surface reflects the major role of shear stress on cake growth. In this study,
between the values was calculated for various disk rotation speeds, the local shear stress instead of the mean value was simulated;
using the empirical equation of [26] with distinct empirical coeffi- subsequently the formed cake was measured, and upon comparing
cients j and those obtained from simulations in the current study. the simulated and experimental results (Fig. 4), the relationship
Because j is a function of disk geometry and cannot be obtained between the formed cake and shear stress became obvious.
directly from theoretical derivations, it is better to determine by An analysis of the distributions of the cake weight at various
data fitting using the empirical equation. When the simulation in disk rotation speeds (Fig. 5) revealed that the cake mass decreased
this study was compared with the values of mean shear stress cal- from the center toward the chamber wall because the shear stres-
culated using the empirical equation with two j values, the simu- ses generated by the rotating disk were not uniformly distributed
lation results were found to be in agreement with the empirical on the membrane surface (Fig. 4). The local cake mass apparently
results of [26] when j was set as 1.2. The mean shear stress

Fig. 3. Simulated velocity distributions on the x-planes of a rotating-disk dynamic filter at three disk rotation speeds.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
6 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx

Fig. 4. Simulated shear stress distributions and cake formations at 1 h of operation at three disk rotation speeds.

12
1.1 8x10
0.25 wt% PMMA, Q=3x10 m /s-6 3 0.25 wt% PMMA, Q=3x10-6 m3/s
P = 100 kPa P = 100 kPa
= 100 rpm = 100 rpm
= 300 rpm 6x1012 = 300 rpm
0.8 = 500 rpm
= 500 rpm
(m/kg)
wr (kg/m2)

12
4x10
r

0.4
2x1012

0
0 0.0025 0.0075 0.0125
0.0025 0.0075 0.0125 r (m)
r (m)
Fig. 6. Local specific cake filtration resistances at different radial positions;
Fig. 5. Local cake mass at different radial positions; TMP = 100 kPa for x = 100– TMP = 100 kPa for x = 100–500 rpm.
500 rpm.

The particle size distributions in the original PMMA suspension

decreased in the central zone. The local cake thickness were mea- were compared with the cake at various disk rotating speeds
sured at the end of the experiments by using the vernier caliper (Fig. 7). The mean particle size may have decreased with increasing
with macroscopic observation. The results also showed that the disk rotation speed because of the higher shear stress acting on the
cake weight decreased as the rotating speed increased. membrane surface. Similarly, the average specific filtration resis-
The distributions of the specific filtration resistance at various tance increased with the increase in disk rotation speed because
disk rotation speeds are analyzed in Fig. 6. According to Eq. (6), of the decreases in both porosity and particle size of the cake.
once the values of the specific surface area of particles and porosity Hwang et al. [41] also found that size distribution becomes finer
are obtained, the local specific filtration resistance of the cake can as the cross-flow velocity increases because of the shear stress
be estimated. The local specific filtration resistance increased from on the membrane surface. This can be reasonably explained from
the center toward the chamber wall and the order of magnitude the size-dependent particle deposition. The critical particle diame-
was as high as 12. The opposite trend was observed with the dis- ters in the filter cakes formed under various operating conditions
tributions of the cake weight because the particle size and porosity will be discussed later in this paper.
decreased with increasingly radial positions. Bouzerar et al. [26] Fig. 8 shows a comparison between the calculated results and
noted that the total filtration resistance becomes equal to the experimental data of the average specific filtration resistance and
membrane resistance at a very high rotating speed (above cake mass at the pseudosteady state under various operating con-
1800 rpm) for r < 0.03 m. ditions. The symbols depict the experimental data and the curves

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx 7

-4
16 1.2x10
0.25 wt% PMMA 0.25 wt% PMMA, Q=3x10-6 m3/s
Q=3x10-6 m3/s, P = 20 kPa Experimental data
Original Calculated value
-4
12 = 100 rpm 1.0x10
= 300 rpm

q s(r) (m3/m2s)
= 500 rpm
f(dp) (%)

8.0x10-5
8

6.0x10-5
4

-5
4.0x10
-6 -6 -6 -6 -6
4.0x10 4.5x10 5.0x10 5.5x10 6.0x10
0 dp,c (m)
1 10 100
dp ( m) Fig. 9. Local filtration flux effects on critical particle size under various operating
conditions.
Fig. 7. Particle size distributions in original PMMA suspension and the cake at
various disk rotation speeds at TMP = 20 kPa.
critical particle size and the calculated value of local filtration flux
can be expressed as:
0.6 1x1013
0.25 wt% PMMA qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
dp;c
Aq2s dp;c r2 x4 þ Bl2 c2
2
Q=3x10-6 m3/s, P = 20 kPa qs ðrÞ ¼  ð14Þ
f cl
0
exp. cal.
av v.s
wc v.s
where qs (r) is the local filtration flux, fc0 , A, and B are coefficients
0.4
that can be estimated from the given critical particle size and oper-
(m/kg)
wc (kg/m2)

1x1012 ative conditions.

Considering all experimental local cake properties, the local fil-
av

tration flux can be estimated as a function of the radial positions

0.2 under various operating conditions. The experimental value of
local filtration flux can be expressed by:

DP
1x1011 qs ðrÞ ¼ ð15Þ
lðRm þ Rc ðrÞÞ
0 5x1010
0 200 400 600 where Rc (r) is the local cake filtration resistance, which can be esti-
(rpm) mated given ar and wr. The result shows that the local filtration flux
increases with increasing critical particle size, and the agreement
Fig. 8. Comparison of calculated results and experimental data for average specific
cake filtration resistances and cake mass at various disk rotation speeds at between the calculated results and the experimental data demon-
TMP = 20 kPa. strates the reliability of the proposed method. Once the cake prop-
erties are obtained from the operative conditions, the filtration flux
represent the results calculated according to the proposed theories.
Because a higher disk rotation speed results in a larger shear stress, -4
4x10
the particles on the membrane surface become less stable and less
0.25 wt% PMMA
cake mass forms. Moreover, the trend in the average specific filtra- Q=3x10-6 m3/s, P = 20 kPa
tion resistance of the cake can be explained by the variations in the = 500 rpm
particle size distribution and packing porosity of the cakes. Com- 3x10
-4 = 300 rpm
pared to the data shown in Fig. 8, the calculated results and the = 100 rpm
experimental data agree substantively.
q s (r) (m3/m2s)

The particles with diameters smaller than a critical particle size

will deposit on the septum, and the probability of particle deposi- 2x10-4
tion significantly reduces to a negligible value when dp exceeds this
value. The effect of local filtration flux on the critical particle size at
rotating-disk structures under various operating conditions is
shown in Fig. 9. The critical particle size, dp,c, for particle deposition 1x10
-4

under given conditions was obtained by the experimental method

proposed by Lu and Ju [43]. The mechanism of cake formation was
developed using a force balance model.
0
The forces exerted on the PMMA particles deposited under the
0 0.2 0.4 0.6 0.8
operative conditions considered in this study were described in
r/r0
Section 2.5. The parameters Fg and Fl are small compared with
the drag forces associated with fluid flow, and the Fi is also consid- Fig. 10. Local filtration flux at different r/r0 with various disk rotation speeds at
erably smaller than the drag forces. Thus, the relationship between TMP = 20 kPa.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
8 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx

can be estimated accurately. The results provide methods of esti- [7] V.S. Espina, M.Y. Jaffrin, L.H. Ding, Comparison of rotating ceramic membranes
and polymeric membranes in fractionation of milk proteins by microfiltration,
mating the local filtration flux, optimizing operating conditions,
Desalination 245 (2009) 714–722.
and designing high-efficiency disk structures or filter modules. [8] F. Liebermann, Dynamic cross flow filtration with Novoflow’s single shaft disk
The local filtration flux for different radial positions at various filters, Desalination 250 (2010) 1087–1090.
disk rotation speeds is shown in Fig. 10. The local filtration flux [9] Z. Tu, L. Ding, Microfiltration of mineral suspensions using a MSD module with
rotating ceramic and polymeric membranes, Sep. Purif. Technol. 73 (2010)
increased with the rotating speed because of a higher shear stress 363–370.
and centrifugal drag force acting on the particles. However, filtra- [10] P. Meyer, A. Mayer, U. Kulozik, High concentration of skim milk proteins by
tion flux values are functions of position because the local cake ultrafiltration: characterisation of a dynamic membrane system with a
rotating membrane in comparison with a spiral wound membrane, Int. Dairy
properties are not uniformly distributed on the membrane surface J. 51 (2015) 75–83.
[18,26]. The calculated flux can account for the mean experimental [11] O. Al-Akoum, M.Y. Jaffrin, L. Ding, P. Paullier, C. Vanhoutte, An hydrodynamic
filtration flux increasing with the rotating speed (namely, the investigation of microfiltration and ultrafiltration in a vibrating membrane
module, J. Membr. Sci. 197 (2002) 37–52.
decrease in cake filtration resistances, as previously described). [12] O. Al-Akoum, M. Mercier-Bonin, L. Ding, C. Fonade, P. Aptel, M.Y. Jaffrin,
This result in a reasonably accurate prediction of filtration flux Comparison of three different systems used for flux enhancement: application
and demonstrates the appropriate use of the earlier theoretical to crossflow filtration of yeast suspensions, Desalination 147 (2002) 31–36.
[13] N. Moulai-Mostefa, M. Frappart, O. Akoum, L. Ding, M.Y. Jaffrin, Separation of
derivation. The proposed method for analyzing the local and mean
water from metal working emulsions by ultrafiltration using vibratory
filtration characteristics in rotating-disk dynamic microfiltration is membranes, J. Hazard. Mater. 177 (2010) 978–982.
reliable. [14] L. Ding, M.Y. Jaffrin, J. Luo, Dynamic filtration with rotating disks, and rotating
or vibrating membranes, in: Progress in Filtration and Separation, Elsevier Ltd.,
2015, pp. 27–59 (Chapter 2).
[15] M.Y. Jaffrin, Dynamic shear-enhanced membrane filtration: a review of
5. Conclusions rotating disks, rotating membranes and vibrating systems, J. Membr. Sci. 324
(1–2) (2008) 7–25.
The local cake properties and filtration flux in rotating-disk [16] M.Y. Jaffrin, L. Ding, O. Akoum, A. Brou, A hydrodynamic comparison between
rotating disk and vibratory dynamic filtration systems, J. Membr. Sci. 242 (1–
dynamic microfiltration were analyzed through experiments
2) (2004) 155–167.
involving a suspension of fine PMMA and CFD simulations of fluid [17] S.D. Rios, E. Clavero, J. Salvado, X. Farriol, C. Torras, Dynamic microfiltration in
flows in the filter and the shear stress on the membrane surface. microalgae harvesting for biodiesel production, Ind. Eng. Chem. Res. 50 (2011)
The simulation results of mean shear stress obtained in this study 2455–2460.
[18] R. Bouzerar, M.Y. Jaffrin, L. Ding, P. Paullier, Influence of geometry and angular
were 0.26–3.50 Pa, which agrees with the results of Bouzerar et al. velocity on performance of a rotating disk filter, AIChE J. 46 (2) (2000) 257–
[26] when j was set as 1.2. 265.
Combined with experimental data on local cake properties, [19] A. Brou, L. Ding, P. Boulnois, M.Y. Jaffrin, Dynamic microfiltration of yeast
suspensions using rotating disks equipped with vanes, J. Membr. Sci. 197
such as thickness, porosity, and specific filtration resistance, the (2002) 269–282.
mechanism of cake formation was analyzed with a force balance [20] M.Y. Jaffrin, Dynamic filtration with rotating disks, and rotating and vibrating
model. The results demonstrate that the local filtration flux membranes: an update, Curr. Opin. Chem. Eng. 1 (2012) 1–7.
[21] Z. Zhu, J. Luo, L. Ding, O. Bals, M.Y. Jaffrin, E. Vorobiev, Chicory juice
increases with increasing critical particle size. The relationship clarification by membrane filtration using rotating disk module, J. Food Eng.
between local filtration flux, critical particle size and the condi- 115 (2) (2013) 264–271.
tions of operation can be expressed as a nonlinear equation: [22] W. Zhang, N. Grimi, M.Y. Jaffrin, L. Ding, Leaf protein concentration of alfalfa
juice by membrane technology, J. Membr. Sci. 489 (2015) 183–193.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi [23] C. Torras, J. Pallarès, R. Garcia-Valls, M.Y. Jaffrin, CFD simulation of a rotating
dp;c
Aq2s dp;c r2 x4 þ Bl2 c2
2
qs ðrÞ ¼  disk flat membrane module, Desalination 200 (2006) 453–455.
f cl
0
[24] C. Torras, J. Pallarès, R. Garcia-Valls, M.Y. Jaffrin, Numerical simulation of the
flow in a rotating disk filtration module, Desalination 235 (1–3) (2009) 122–
The cake formation and filtration flux were nonuniform on the 138.
entire membrane surface because the distributions of cake proper- [25] K. Kim, J.Y. Jung, J.H. Kwon, J.W. Yang, Dynamic microfiltration with a
perforated disk for effective harvesting of microalgae, J. Membr. Sci. 475
ties, such as mean particle diameter, porosity, and average specific (2015) 252–258.
filtration resistance, are functions of position. These results provide [26] R. Bouzerar, L. Ding, M.Y. Jaffrin, Local permeate flux-shear-pressure
a way to not only achieve local filtration performance but also relationships in a rotating disk microfiltration module: implications for
global performance, J. Membr. Sci. 170 (1) (2000) 127–141.
design high-efficiency disk structures or filter modules. [27] R. Bouzerar, M.Y. Jaffrin, A. Lefevre, P. Paullier, Concentration of ferric
hydroxide suspensions in saline medium by dynamic cross-flow filtration, J.
Membr. Sci. 165 (2000) 111–123.
Acknowledgments [28] L. Li, L. Ding, Z. Tu, Y. Wan, D. Clausse, J.L. Lanoisellé, Recovery of linseed oil
dispersed within an oil-in-water emulsion using hydrophilic membrane by
The authors wish to express their sincere gratitude to the Min- rotating disk filtration system, J. Membr. Sci. 342 (2009) 70–79.
[29] Z. Zhu, J. Luo, L. Ding, O. Bals, M.Y. Jaffrin, E. Vorobiev, Chicory juice
istry of Science and Technology of the Republic of China for its clarification by membrane filtration using rotating disk module, J. Food Eng.
financial support. 115 (2013) 264–271.
[30] J. Luo, L. Ding, Y. Wan, P. Paullier, M.Y. Jaffrin, Application of NF-RDM
(nanofiltration rotating disk membrane) module under extreme hydraulic
References conditions for the treatment of dairy wastewater, Chem. Eng. J. 163 (2010)
307–316.
[31] J. Luo, Z. Zhu, L. Ding, O. Bals, Y. Wan, M.Y. Jaffrin, E. Vorobiev, Flux behavior in
[1] C.K. Choi, J.Y. Park, W.C. Park, J.J. Kim, A study on dynamic separation of silica
clarification of chicory juice by high-shear membrane filtration: evidence for
slurry using a rotating membrane filter: 2. Modelling of cake formation, J.
threshold flux, J. Membr. Sci. 435 (2013) 120–129.
Membr. Sci. 157 (1999) 177–187.
[32] V.S. Espina, M.Y. Jaffrin, M. Frapparta, L.H. Ding, Separation of casein micelles
[2] J. Engler, M.R. Wiesner, Particle fouling of a rotating membrane disk, Water
from whey proteins by high shear microfiltration of skim milk using rotating
Res. 34 (2) (2000) 557–565.
ceramic membranes and organic membranes in a rotating disk module, J.
[3] M.Y. Jaffrin, G. He, L.H. Ding, P. Paullier, Effect of membrane overlapping on
Membr. Sci. 325 (2008) 872–879.
performance of multishaft rotating ceramic disk membranes, Desalination 200
[33] N. Moulai-Mostefa, O. Akoum, M. Nedjihoui, L. Ding, M.Y. Jaffrin, Comparison
(2006) 269–271.
between rotating disk and vibratory membranes in the ultrafiltration of oil-in-
[4] L.H. Ding, M.Y. Jaffrin, M. Mellal, G. Heb, Investigation of performances of a
water emulsions, Desalination 206 (2007) 494–498.
multishaft disk (MSD) system with overlapping ceramic membranes in
[34] W. Zhang, L. Ding, Investigation of membrane fouling mechanisms using
microfiltration of mineral suspensions, J. Membr. Sci. 276 (2006) 232–240.
blocking models in the case of shear-enhanced ultrafiltration, Sep. Purif.
[5] S. Bhattacharjee, S. Ghosh, S. Datta, C. Bhattacharjee, Studies on ultrafiltration
Technol. 141 (2015) 160–169.
of casein whey using a rotating disk module: effects of pH and membrane disk
[35] K.J. Hwang, S.E. Wu, Disk structure on the performance of a rotating-disk
rotation, Desalination 195 (2006) 95–108.
dynamic filter: a case study on microalgae microfiltration, Chem. Eng. Res. Des.
[6] C. Bhattacharjee, P.K. Bhattacharya, Ultrafiltration of black liquor using
94 (2015) 44–51.
rotating disk membrane module, Sep. Purif. Technol. 49 (2006) 281–290.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049
 K.-J. Hwang et al. / Separation and Purification Technology xxx (2017) xxx–xxx 9

[36] K.J. Hwang, S.J. Lin, Filtration flux-shear stress-cake mass relationships in [40] K.J. Hwang, S.Y. Wang, E. Iritani, N. Katagiri, Fine particle removal from
microalgae rotating-disk dynamic microfiltration, Chem. Eng. J. 244 (2014) seawater by using cross-flow and rotating-disk dynamic filtration, J. Taiwan
429–437. Inst. Chem. Eng. 62 (2016) 45–53.
[37] F.M. Tiller, J.R. Cramp, F. Ville, A revised approach to the theory of cake [41] K.J. Hwang, Y.L. Hsu, K.L. Tung, Effect of particle size on the performance of
filtration, in: P. Sanasundaran (Ed.), Fine Particles Processing, vol. 2, Amer. Inst. cross-flow microfiltration, Adv. Powder Technol. 17 (2) (2006) 189–206.
Min. Met. Petr. Eng., New York, 1980, pp. 1549–1558. [42] W.M. Lu, K.L. Tung, S.M. Hung, J.S. Shiau, K.J. Hwang, Constant pressure
[38] P.R. Vasseur, G. Cox, The lateral migration of a spherical particle in two- filtration of mono-dispersed deformable particle slurry, Sep. Sci. Technol. 36
dimensional shear flows, J. Fluid Mech. 78 (1976) 385. (11) (2001) 2355–2383.
[39] K.J. Hwang, H.C. Liu, W.M. Lu, Local properties of cake in cross-flow [43] W.M. Lu, S.C. Ju, Selective particle deposition in cross-flow filtration, Sep. Sci.
microfiltration of submicron particles, J. Membr. Sci. 138 (1998) 181–192. Technol. 24 (7,8) (1989) 517.

Please cite this article in press as: K.-J. Hwang et al., Analysis on the nonuniformity of cake formation in rotating-disk dynamic microfiltration, Separ. Purif.
Technol. (2017), http://dx.doi.org/10.1016/j.seppur.2016.12.049