w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0

Available online at www.sciencedirect.com

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

Water hammer reduces fouling during natural water

ultrafiltration

F. Broens a, D. Menne b, I. Pothof c, B. Blankert d, H.D.W. Roesink d, H. Futselaar d,

R.G.H. Lammertink e,f, M. Wessling b,e,*
a
Convergence B.V., Munsterstraat 18, 7418 EV Deventer, The Netherlands
b
Chemical Process Engineering-AVT.CVT, RWTH Aachen University, Turmstr. 46, 52056 Aachen, Germany
c
Deltares Rotterdamseweg 185, P.O. Box 177, 2600 MH Delft, The Netherlands
d
Pentair X-Flow B.V., Marssteden 50, 7547 TC Enschede, P.O. Box 739, 7500 AS Enschede, The Netherlands
e
Membrane Technology Group, University of Twente, 7500 AE Enschede, The Netherlands
f
Soft Matter, Fluidics and Interfaces, University of Twente, 7500 AE Enschede, The Netherlands

article info abstract

Article history: Todays ultrafiltration processes use permeate flow reversal to remove fouling deposits on
Received 21 June 2011 the feed side of ultrafiltration membranes. We report an as effective method: the opening
Received in revised form and rapid closing of a valve on the permeate side of an ultrafiltration module. The sudden
21 October 2011 valve closure generates pressure fluctuations due to fluid inertia and is commonly known
Accepted 5 December 2011 as water hammer. Surface water was filtrated in hollow fiber ultrafiltration membranes
Available online 14 December 2011 with a small (5%) crossflow. Filtration experiments above sustainable flux levels
(>125 l (m2h)
1) show that a periodic closure of a valve on the permeate side improves
Keywords: filtration performance as a consequence of reduced fouling. It was shown that this effect
Ultrafiltration depends on flux and actuation frequency of the valve. The time period that the valve was
Water hammer closed proved to have no effect on filtration performance. The pressure fluctuations
Fouling generated by the sudden stop in fluid motion due to the valve closure are responsible for
the effect of fouling reduction. High frequency recording of the dynamic pressure evolution
shows water hammer related pressure fluctuations to occur in the order of 0.1 bar. The
pressure fluctuations were higher at higher fluxes (higher velocities) which is in agreement
with the theory. They were also more effective at higher fluxes with respect to fouling
mitigation.
2011 Elsevier Ltd. All rights reserved.

1. Introduction the same time, membrane filtration processes are often

1.1. General introduction
Ultrafiltration has proven to be a viable treatment process for
purification of surface water and wastewater due to its high
and constant retention of bacteria, viruses and natural
organic matter NOM (Shannon et al., 2008; Cho et al., 2000). At
simple in operation and have a low footprint in comparison to
conventional processes. Even though the advantages have led
to strong increase in installed membrane area, ultrafiltration
for water filtration still copes with a major challenge: fouling
of the filter causes higher permeation resistance, thus
requiring more energy for filtrating the same volume. In order
to operate in a sustainable way, periodic cleaning is required

* Corresponding author. Chemical Process Engineering-AVT.CVT, RWTH Aachen University, Turmstr. 46, 52056 Aachen, Germany.
E-mail address: matthias.wessling@avt.rwth-aachen.de (M. Wessling).
0043-1354/$ e see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2011.12.011
1114 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0
for better efficiency. Hence, efficient control of fouling is
critical in membrane operation and much research has
focussed the last years on ways to describe, predict and
control fouling (Chang et al., 2002; Bacchin and Aimar, 2005;
Aimar and Bacchin, 2010).
An important concept has been introduced during the last
decade: the critical and sustainable flux (Bacchin et al., 2006).
In general a membrane filtration process is operated below the
critical or sustainable flux (Field et al., 1995). The determina-
tion of the critical flux or the sustainable flux can be
approached with different methods (Le Clech et al., 2003; van
de Ven et al., 2008; van der Marel et al., 2009). However, such
maximum flux levels in ultrafiltration are clearly below
100 l (m2h)
1). Today there is no research at controlling and
diminishing the fouling rate above the critical flux.
Crucial to the operation of such membrane filtration
processes is the concept of flow reversal, often known as
blackpulsing (Ma et al., 2001) and backflushing (Levesley and
Hoare, 1999). Here, a part of the permeate is returned to the
feed phase through the membrane to detach the deposited
matter on and in the membrane. Together with a subsequent
forward flush, the deposit will be washed out the membrane
feed channel.
The concept of dynamically altering the filtration condi-
tions has also been extended into for instance oscillatory
flow (Blanpain-Avet et al., 1999) or vibrating membranes
(Girones i Nogue et al., 2006). In this article, we report a novel
method for controlling and reducing the effects of fouling
tested for the filtration of natural water. We quickly stop the
permeate flow by closing the valve on the permeate side
briefly: this causes an effect called water hammer to occur.
Extensive filtration experiments, together with ultra-fast
pressure recordings, demonstrate that the concept of water
hammer is a new and strong tool to reduce fouling at high
flux levels.
1.2. Water hammer
Water hammer is an effect that commonly occurs when
a valve is closed suddenly in a pipeline system and a resultant
pressure wave propagates through the pipe. Such water
hammers are of significant interest in large hydraulic piping
systems such as drinking water pipeline networks. The
evolving pressure spikes are caused by the fluids inertia:
momentum is conserved at all times and the motion of the
fluid is converted into pressure forces (Wylie and Streeter,
1978). The fundamentals of the water hammer phenomenon
are comprehensively described by Wylie and Streeter (1978)
and Kartvelishvili (1994).
For the maximum pressure fluctuation the Joukowsky
pressure surge formula can be used. This equation is
a simplified form of the conservation of momentum for
incompressible flow.
Dp rf $af $Dv
Dv is the change in average velocity, rf is the fluid density,
and af is the speed of sound in the medium, which depends
both on the fluid and the pipe wall material.
Eq. (1) is only valid if the valve closes within the charac-
teristic time 2L=af time, where L is the pipe length before
the valve: The valve has to be shut completely before the
pressure wave, which has been generated from partially
closing the valve, is returned (Wylie and Streeter, 1978).
Hence, the pipe length (before the valve) and the time the
valve requires for closing are critical. The most important
variable is the average flow velocity and correlates linearly
with the maximum pressure surges.
The water hammer can be realized by placing a valve on
the permeate side which opens and closes frequently during
the permeation phase of a filtration cycle. As such, the
filtration cycle proceed as normally done and described in
detail by van de Ven et al. (2008) while the valve functions
frequently during permeate production. We hypothesize that
a water hammer on the permeate piping causes an alteration
of hydrodynamic conditions in the membrane which in turn
will affect the deposition phenomena on the membrane
surface.
2. Experimental
2.1. Filtration process
The experiments were carried out with an automated setup as
schematically shown in Fig. 1, was used. The setup was fully
controlled by a Labview-based program. The flow meters used
were Krohne Optimass (Coriolis based) 2050C mass flowme-
ters (0.05 kg/h accuracy). The valves were solenoid Buschjost
8208066 series valves, with a closing time of 30 ms. All tubing
consisted of flexible Festo polyurethane 4 mm tubing. The
pumps were Ismatec piston pumps with a REGLO-2 head and
a maximum capacity of about 4.5 l/h. Digital pressure sensors
(Endress Hauser BMC 131-ATT) were used, with a maximum
of 2.18 bar and a 0.02 bar accuracy. All tanks and storage
vessels were made of polyethylene.
The used setup automatically performed filtrations,
backwashes and chemical cleanings at chosen lengths and
intervals. Crossflow was controlled by a metering valve and
measured with a balance. The metering valve was set to
a constant orifice, resulting in a constant resistance value
giving higher flow at higher pressure. The valves highlighted
in gray (Fig. 1) could be operated to generate the desired
water hammer effect. A length of 1.8 m 4 mm Festo poly-
urethane tubing was used between the valve and the module,
to ensure the valve closes faster then the characteristic time
of 2L=af .
The chemicals for the chemical cleaning step consisted of
a caustic (first) and an acidic cleaning step. The caustic solu-
tion consisted of 400 ppm (volume) NaOCl in water, after
which NaOH was added up to reach a pH of approximately
11.5. The acidic solution consisted of a HNO3 solution at pH
1.5. Membranes were cleaned in 4 cycles of caustic/acidic
cleanings with respectively 25 and 15 min soak times in the
(1) module.
The setup automatically logged every 4 s the flux
(l (m2h)
1)), temperature ( C) and transmembrane pressure
(TMP) as the difference between feed and permeate pressure:
TMP pfeed
ppermeate (2)
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0 1115

Fig. 1 e Process diagram of the experimental filtration setup.

The feed pressure was measured at the entrance of the 2.3. Filtration settings
module, the permeate pressure was measured at the entrance
of the permeate connection (on top of the module). The filtration system carries out constant flow permeation. In
practice, one tries to fix the permeate flux at a level that
guarantees a sustainable operation: the membrane resistance
2.2. Membrane characteristics
and the filtration resistance stay constant over many filtration
cycles.
In practice for (natural) water filtration, ultrafiltration is
Here we report filtration experiments which are carried out
mainly the selected membrane type for filtration: It has typi-
intentionally at significantly higher flux levels. This causes the
cally a 1:1,000,000 blocking ratio for viruses, bacteria and large
membrane and filtration resistance to drift over all the filtra-
organic molecules. Ultrafiltration is mainly performed using
tion cycles toward higher values. Normally, this has to be
hollow-fiber/tubular membranes and is operated in dead-end
avoided, however, we are able to observe the water hammer
mode. The performed experiments in this paper were carried
effect better. In fact, if the water hammer reduces the drift in
out with a small bleed crossflow (5% of the feed) to allow
membrane and filtration resistance, one would be able to
possible dislodged fouling to be carried away with the bleed
permeate at higher fluxes hence intensifying the process
flow. This crossflow was intentionally low to resemble a dead-
efficiency. For these experiments the chosen net flux was
end filtration, which is the most used operation mode in
between 125 and 300 l (m2h)
1) and was the same for the
natural water filtrations.
experiments with and without valve control.
For all experiments, ultrafiltration membranes (Norit) were
used. The fibers had an average inner diameter of 1.5 mm and
a nominal pore size of 0.3 mm. The modules were constructed
using a transparent acrylic housing; this allowed trapped air Table 1 e Fixed parameters in the experimental setup.
bubbles in the module to be spotted. Air bubbles were always
Parameter Fixed value Unit
removed (bubbles dampen pressure spikes) by tilting the
module, until all bubbles were flushed out through the Filtration steps 24 e
Backwash steps 24 e
permeate connection.
Chemical cleanings per cycle 1 e
The acrylic housing had a 13 mm outer diameter, a 2 mm
Filtration time 11 min
wall thickness and was 0.50 m in length. On both ends G1=400 Crossflow flux 5 %
male tapered threading was tapped, allowing various nozzles Backwash time 40 s
to fit directly on the module. On the top (modules were operated Backwash flux 125 l (m2h)
1)
vertically), directly next to the potting, a female G1=800 thread- Caustic Chem.clean time 25 min
ing was tapped as permeate port. The permeate port was placed Acidic chem.clean time 15 min
Chem.clean flux 150 l (m2h)
1)
on top to assure that virtually all air automatically flushed out.
Membrane area 0.0158 m2
7 fibers were potted per module, leaving after glueing
Valve close time 30.0 ms
0.48 m effectively, giving a 0.0158 m2 surface area (Table 1).
1116 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0
2.4. Valve settings
A limited number of parameters was chosen to be varied
(shown in Table 2); these are the parameters that have the
highest influence on the water hammer effect according to Eq.
(1). The parameters are liquid velocity (given through the
magnitude of the flux) and valve operation (open time/close
time and cycle times). The actuation valve could open and
close at variable intervals up to a minimum of 0.4 s. An
important constant here is the time the valve needs to close:
30 ms. This was a property of the valve and could not be
adjusted. The time at which the valve is closed decreases the
permeate production time and should hence be as small as
possible.
The total time of one valve actuation cycle being closed and
opened is described as the frequency of valve operation:
1
fcycle
tclosed topen
The frequency can be increased for instance by reducing the
closed valve time and/or the open valve time.
2.5. High frequency equipment
A high frequency data acquisition setup was used to measure
pressure spikes, at a rate of 3000 Hz - 15 kHz maximum. The
equipment was supplied by Deltares and consisted of two
high frequency pressure sensors attached to a signal ampli-
fier. The signal was processed by Deltmeasure, a program
written for this purpose and provided by Deltares. Both
sensors were calibrated and had a 0.005 bar accuracy. The
permeate pressure sensor was placed in the middle of the
module, the feed pressure sensor directly on top of the
module.
3. Results and discussion
3.1. Variation of valve cycle (open/close) frequency
Typical data obtained during a filtration run are shown in
Fig. 2(a). The left figure shows all the filtration data logged
during a filtration sequence with more than 25 permeation
phases at a set flux value with constant filtration time. The
right figure shows the data extracted from each permeation
phase: the left y-axis gives the measured initial trans-
membrane pressure, the right y-axis shows the calculated
initial resistance using Darcys law. Different valve open/close
frequencies were tested (according to Eq. (3)), by varying the
time the valve was opened (between 10 and 30 s) and keeping
Table 2 e Varied parameters.
Parameter Variable range Unit
Filtration flux 125e300 l (m2h)
1)
Open valve time 5e30 s
Closed valve time 0.75e5 s
the time it was closed constant (2.6 s). The experiments with
valve actuation were carried out longer than the base case
such that over 5 h of time no lower filtrated volume than
without valve control was obtained.
Fig. 2(a) clearly shows that the actuation of the permeate
valve causing the water hammer has a beneficial effect on the
filtration performance. At all valve settings the initial filtration
resistance as well as the final filtration resistance (can be
extracted from Fig. 2(a)) are lower than the reference case. The
lower the frequency of valve actuation (longer intermediate
time at 20 s for instance), the more the resistance approaches
the reference case. But at an intermediate time of 10 s, the
initial resistance reduces by about 30%. This would immedi-
ately result in 30% energy savings. However, increasing the
frequency by further decreasing the intermediate time would
not make much sense because the intermediate time is at 10 s
only 4 times larger than the non-productive close valve time of
(3) 2.5 s.
Further experiments on the time the valve could be kept
closed at an intermediate time of 10 s showed that shortening
even below close valve times of 1 s would be equally good.
3.2. Variation of flux
Several fluxes were tested at the valve open time of 10 s and
a close time of 0.75 s (a frequency of 0.093 Hz). Every test was
ran from the same batch of canal water on the same day.
Fig. 3(a) and (b) show the 5% crossflow filtration at
125 l (m2h)
1) over 5 h filtration time (black line). This was
compared to a 5% crossflow reference of the same batch,
performed directly after each other with a thorough chemical
cleaning in between. The experiment consisted of 24 cycles of
11 min filtration and a 40 s backwash at constant flux and
temperature.
Fig. 3(a) shows very little difference between the refer-
ence and the filtration cycles with water hammer. Also, the
drift of the initial filtration resistance is low in comparison
to all data reported in this paper at higher fluxes (see also
the figures below). These measurements and others done at
lower fluxes show no difference between reference and
water hammer modified filtration cycles suggesting that the
water hammer only is effective under conditions of serious
fouling, at fluxes above a sustainable flux. Fig. 3(c) and (e) are
carried out at a flux of 150 l (m2h)
1) and 185 l (m2h)
1) flux
respectively. It can be seen that there is a significant
reduction between filtration resistance of the standard and
valve controlled experiments at fluxes higher than
125 l (m2h)
1).
As mentioned above, we found that the time in which the
valve was closed had no significant impact on the effective-
ness of influencing the fouling phenomena occurring at the
feed side of the hollow fiber. The actual process of closing the
valve effects the filtration resistance, but not the time that it
was closed. These observations are in agreement with the
theory on water hammer dynamics which predicts very rapid
pressure spikes to be generated by the sudden change in fluid
motion damping out over time (ms range).
Variation of flux showed to have a significant impact on the
initial filtration resistance over time: but for all fluxes higher
than 125 l (m2h)
1) the water hammer integration into the
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0
Fig. 2 e Filtration cycles for different valve open/close frequencies, at a constant flux of 125 l (m2h)L1) and constant closed
valve times of 2.6 s. Frequency varied through different intermediate times 10, 15, and 20 s in comparsion to the reference
crossflow experiment. (a) Transmembrane pressures as logged during perme-ation phases. (b) Same as 2(a), but shown as
initial transmembrane pressure and initial resistance(dual axis).
filtration process showed a significant reduction in filtration
resistance. An average resistance reduction of 32% for the
150 l (m2h)
1) (Fig. 3(c)) experiment and 48% for the
185 l (m2h)
1) Fig. 3(e) experiment was found.
The process of closing/opening a valve on a 0.1 Hz
frequency or higher at a flux of 150 l/m2/h or higher creates an
effect that reduces the fouling. This could be either due to the
generated pressure spikes to be more effective and/or
a change of dominant fouling phenomenon that is affected by
the pressure spikes.
Concentration polarization and fouling phenomena
contribute to the overall resistance at high fluxes. Since the
water hammer effect works better at higher fluxes we
hypothesize that the flow valve control causes a distortion of
the hydrodynamic conditions at the membrane fluid inter-
face. The nature of fouling, i.e. pore blocking, pore constric-
tion, adsorption, and cake formation, may be influenced as
well, however at this point it is impossible to unravel the
effect of water hammer on fouling mechanisms, its preven-
tion or disruption of fouling deposition.
3.3. High frequency measurements
In order to understand the dynamics of the water hammer
phenomenon better, detailed pressure spike measurements
were carried out to observe and identify these pressure spikes.
With the high frequency setup (described in Section 2.5),
measurements of the dynamics in pressure changes were
recorded for both feed and permeate at a rate of 3000
measurements per second. A typical result of such
a measurement is shown in Fig. 4(a). The noise in the signal of
both pressure sensors is caused by the pump rotations of the
piston pump. When the valve is closed, a distinct peak appears
in the permeate as well as in the feed channel. The peak is
sharper and richer in features on the permeate side. On the
shoulder of the large peak smaller fluctuations occur which we
attribute to the phenomenon of water hammer (Fig. 4(c)).
1117
The pressure changes are best visible when observed as the
deviation of the actual transmembrane pressure at that time.
Normalized TMP can be calculated as:
TMPav pf
pp (4)
 
TMPnorm pf t
pp t
TMPt0 (5)
Measurements of pressure changes at different close valve
times and fluxes were performed as well. The following
experiments were carried out: a) high frequency pressure
recordings at different valve closed times and b) high
frequency pressure recording at different fluxes.
The linear slope in Fig. 4(b) is due to the valve being shut,
while the pump keeps pumping. At longer valve shut times,
the normalized pressure TMP continues to decrease linearly
due to the increasing permeating volume in the constant
permeate volume. The smaller fluctuations in the beginning
are the actual water hammer pressure changes. They quickly
dampen out after approximately 3 reflections. The different
closed valve times show identically small spikes in the first
700 ms. The time the valve is closed has no effect, except for
the total pressure increase as a consequence of the closed
valve. On the basis of the previous observations that show no
influence on the valve closure time, we find it likely that the
small pressure changes induce hydrodynamic conditions in
the feed phase which suppress the extent of fouling.
The observed water hammer spikes are about 0.05e0.1 bar.
The calculated values and measured values are shown in
Table 3: They are in reasonable agreement with the theoretical
expected values calculated by the Joukowsky Eq. (1). The
deviation between prediction and actually measured pressure
fluctuations may be found in the difference of the assumption
of the underlying theory and our actual situation. The model
for instance assumes an impermeable wall, an inelastic and
rigid pipe. Small, yet invisible air bubbles may also reduce the
water hammer related pressure fluctuations. All these may
result in a reduction of maximum pressure in the order of 50%.
1118 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0

Fig. 3 e Filtration cycles at different fluxes ((a) (c) and (e)). At a close valve time of 0.75 s and an open valve time of 10 s.
(a) Transmembrane pressures at 125 l (m2 h)L1 flux. (b) Same as 3(a), initial transmembrane pressure and inital resistance.
(c) Transmembrane pressures at 150 l (m2 h)L1 flux. (d) Same as 3(c), initial transmembrane pressure and inital resistance).
(e) Transmembrane pressures at 185 l (m2 h)L1 flux. (f) Same as 3(e), initial transmembrane pressure and inital resistance.

After reaching a minimum in normalized TMP at 750 ms (at a) the steep change in pressure is an increase in trans-
the instant the valve is opened), the TMP increases very membrane pressure, resulting shortly in a higher flux and
sharply. This sharp increase in TMP is expected to have no (or hence shortly high fouling deposit on the surface. As
negative) effect on the filtration resistance because: a consequence of the shortly increased flux, the overall
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0 1119

Fig. 4 e Different high frequency measurements, corresponding to the filtration experiments from Fig. 3 with a fixed valve
close time of 0.75 s and a 10 s close interval. (a) Feed and permeate pressure recording of a 125 l (m2 h)L1 flux filtration.
(b) Pressure difference recording for different valve closed times at 150 l (m2 h)L1. (c) Pressure recording for fluxes of 125 to
300 l (m2 h)L1. (d) Pressure recording (Figure 4(c)) plotted as normalized TMP.

permeation over time is the same in comparison to
a filtration without valve control.
b) For longer valve closed times this slope is also much
sharper (see Fig. 4(b)). In Fig. 2(a) it was shown that varying
valve closed times had no effect. Hence, it can be concluded
that the small pressure fluctuations before the maximum
are responsible for reducing filtration resistance.
and (d) long-term performance. For instance the pressure
spikes are damped over the length due to an imperfect elastic
behavior of the tube wall. Therefore we hypothesize that
a valve closer to the module will result in a more effective
water hammer.

4. Conclusions
Yet unanswered but still important questions are the
influences of (a) distance between valve and membrane, (b)
Water hammer related pressure fluctuations could be induced
module size, (c) material fatigue for membranes and valves,
and recorded during the filtration with an ultrafiltration
process operated on surface water. The magnitude of the
pressure fluctuations can be influenced through the permeate
flux, e.g. velocity in the permeate piping system. The water
Table 3 e Calculated expected maximum pressures in
a lab scale membrane module. hammer affects the filtration process positively at high fluxes
when the sustainable flux has most likely been surpassed. The
Flux (l/m2/h) v (m/s) pmax,middle (bar) Deviation (%)
water hammers effect becomes more beneficial at systems
125 0.0089 0.028 53 that show severe fouling. It suggests that fouling at high flux
150 0.011 0.035 49 level can be systematically reduced and controlled. This in
185 0.013 0.042 52
turn would imply that todays ultrafiltration processes may be
250 0.018 0.056 47
300 0.022 0.070 43
operated at much higher fluxes making it superfluous to
operate below the sustainable flux.
1120 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 1 1 1 3 e1 1 2 0
Acknowledgment
Matthias Wessling acknowledges the support through the
Alexander-von-Humbold Professorship.
references
Aimar, P., Bacchin, P., 2010. Slow colloidal aggregation and
membrane fouling. Journal of Membrane Science 360 (1e2),
70e76.
Bacchin, P., Aimar, P., 2005. Critical fouling conditions induced by
colloidal surface interaction: from causes to consequences.
Desalination 175 (1), 21e27.
Bacchin, P., Aimar, P., Field, R.W., 2006. Critical and sustainable
fluxes: theory, experiments and applications. Journal of
Membrane Science 281 (1e2), 42e69.
Blanpain-Avet, P., Doubrovine, N., Lafforgue, C., Lalande, M., 1999.
The effect of oscillatory flow on crossflow microfiltration of
beer in a tubular mineral membrane system e membrane
fouling resistance decrease and energetic considerations.
Journal of Membrane Science 152 (2), 151e174.
Chang, I., Le Clech, P., Jefferson, B., Judd, S., 2002. Membrane fouling
in membrane bioreactors for wastewater treatment. Journal of
Environmental Engineering-ASCE 128 (11), 1018e1029.
Cho, J., Amy, G., Pellegrino, J., 2000. Membrane filtration of natural
organic matter: comparison of flux decline, nom rejection, and
foulants during filtration with three UF membranes.
Desalination 127 (3), 283e298.
Field, R., Wu, D., Howell, J., Gupta, B., 1995. Critical flux concept
for microfiltration fouling. Journal of Membrane Science 100
(3), 259e272.
Girones i Nogue, M., Akbarsyah, I.J., Bolhuis-Versteeg, L.A.M.,
Lammertink, R.G.H., Wessling, M., 2006. Vibrating
polymeric microsieves: antifouling strategies for
microfiltration. Journal of Membrane Science 285,
323e333.
Kartvelishvili, L.N., 1994. Water hammer: basics and current
state of the theory. Hydrotechnical Construction 28 (9),
559e568.
Le Clech, P., Jefferson, B., Chang, I., Judd, S., 2003. Critical flux
determination by the flux-step method in a submerged
membrane bioreactor. Journal of Membrane Science 227 (1e2),
81e93.
Levesley, J., Hoare, M., 1999. The effect of high frequency
backflushing on the microfiltration of yeast homogenate
suspensions for the recovery of soluble proteins. Journal of
Membrane Science 158 (1e2), 29e39.
Ma, H., Hakim, L., Bowman, C., Davis, R., 2001. Factors
affecting membrane fouling reduction by surface
modification and backpulsing. Journal of Membrane
Science 189 (2), 255e270.
Shannon, M.A., Bohn, P.W., Elimelech, M., Georgiadis, J.G.,
Marinas, B.J., Mayes, A.M., 2008. Science and technology for
water purification in the coming decades. Nature 452 (7185),
301e310.
van de Ven, W.J.C., vant Sant, K., Punt, I.G.M., Zwijnenburg, A.,
Kemperman, A.J.B., van der Meer, W.G.J., Wessling, M., 2008.
Hollow fiber dead-end ultrafiltration: influence of ionic
environment on filtration of alginates. Journal of Membrane
Science 308 (1e2), 218e229.
van der Marel, P., Zwijnenburg, A., Kemperman, A., Wessling, M.,
Temmink, H., van der Meer, W., 2009. An improved flux-step
method to determine the critical flux and the critical flux for
irreversibility in a membrane bioreactor. Journal of Membrane
Science 332 (1e2), 24e29.
Wylie, E., Streeter, V., 1978. Fluid Transients. McGraw-Hill Inc.