WO2019194747A1 - Membrane filtration system and method - Google Patents

Abstract

A membrane filtration system comprising: at least one membrane module to perform filtration of a feed fluid in which the at least one membrane module is submerged during use of the membrane filtration system; at least one spacer provided next to and spaced apart from the at least one membrane module in parallel with the at least one membrane module; and a vibration mechanism provided to vibrate only the at least one spacer to create a scouring effect on the at least one membrane module during use of the membrane filtration system, wherein vibration of the at least one spacer is in only directions parallel to the at least one membrane module.

Description

MEMBRANE FILTRATION SYSTEM AND METHOD

FIELD OF THE INVENTION

This invention relates to a membrane filtration system and method, and in particular to a membrane filtration system and method using submerged membrane module with fouling control.

BACKGROUND OF THE INVENTION

Membrane separation technology has been widely applied in many fields. However, the occurrence of membrane fouling due to the unfavourable interactions of the filtered substances in the solution with membranes reduces the membrane performance and increases the operating cost. Although membrane fouling is inevitable, it can be alleviated by effective physical, chemical, and biological fouling control strategies depending on the type of fouling that is dominant (Drews, 2010; Fane and Fell, 1987). In the submerged flat sheet membrane filtration (such as microfiltration and ultrafiltration) processes, periodic physical cleaning methods are frequently adopted to maintain sustainable membrane performance.

The conventional physical cleaning approach, such as gas sparging, is widely applied in the submerged flat sheet membrane filtration process, which promotes turbulence and instabilities inside membrane modules (Cui et al., 2003; Wibisono et al., 2014). However, high energy consumption (up to 70% of total energy) incurred by the gas blower and diffuser (Judd, 2011; Verrecht et al., 2010) and relatively low shear rates are the major limitations of gas sparging.

Alternatively, air/liquid flow-facilitated particle scouring in submerged flat sheet membrane filtration processes has been considered as an energy-efficient physical fouling control method and has been successfully applied in membrane bioreactors (Aslam et al., 2017). For example, the liquid flow- facilitated particle fluidization only consumes less than 10% of the energy consumption of gas sparging (Kim et al., 2011). The particle movement enhances shear-induced diffusion and provides mechanical abrasion forces along the membrane due to the physical nature of the particles (Wu et al., 2017). A stronger membrane scouring force can be achieved at a higher particle momentum, but this requires greater energy consumption (Wang et al., 2016). In addition, this method is also limited by the difficulty in achieving homogeneous shear over the whole membrane surface because of size segregation (Wang et al., 2017) and by the possibility of membrane integrity loss (Wu et al., 2015).

Another shear-induced membrane fouling control strategy, involving rotating and vibrating the flat sheet membranes has received attention (Al Akoum et al., 2002; Bilad et al., 2012; Serra and Wiesner, 2000). Vibration of membrane module/element can be achieved by a motor, or a magnetic device, or the use of resonance (Jaffrin, 2008). Such motion increases the back-transport of potential foulants and reduces the concentration polarization and cake formation. However, because rotating/vibrating membrane involves the movement of membrane cassette and permeate water inside of membrane module, it requires a relatively high vibration energy as part of the vibration energy is consumed to move the membrane cassette and permeate water (i.e., without relating to fouling control) and complicated membrane module/reactor design (especially mechanical connection).

SUMMARY OF INVENTION

Disclosed in this application is a membrane filtration system comprising vibrating spacers as a turbulence promoter to alleviate membrane fouling in a submerged flat sheet membrane filtration system. In the commercial flat sheet membrane module, a space is generally provided between two flat sheet membrane elements to allow bubbles to scour the membrane surfaces. Thus, the vibrating spacer is located in such a space and closely situated to the membrane surface. This guarantees no additional modification is required for the commercial flat sheet membrane module configuration. The spacers are connected to a vibration mechanism, which allows them to undergo vibrations parallel to the membrane surface. The movement of the spacers can induce disruption in the boundary layer and can promote shear stress on the membrane surface. The motion trajectory of each spacer guarantees it to cover the membrane area for effective fouling control. One of the major advantages of spacer vibration is that the spacers can be tailor-made with light materials, ensuring minimizing energy consumption of the vibration motor.

According to a first exemplary aspect, there is provided a membrane filtration system comprising: at least one membrane module to perform filtration of a feed fluid in which the at least one membrane module is submerged during use of the membrane filtration system; at least one spacer provided next to and spaced apart from the at least one membrane module in parallel with the at least one membrane module; and a vibration mechanism provided to vibrate only the at least one spacer to create a scouring effect on the at least one membrane module during use of the membrane filtration system, wherein vibration of the at least one spacer is in only directions parallel to the at least one membrane module.

The at least one spacer may be spaced apart from the at least one membrane module by a distance ranging from 0.1 mm to 10 mm.

The at least one spacer may have a non-continuously-flat surface facing the at least one membrane module.

The non-continuously-flat surface may comprise one of: a surface having grooves thereon, and a surface having a plurality of protrusions thereon. The grooves may have a width of 0.5 mm, a depth of 1 mm, and adjacent grooves may be 3 mm apart. The plurality of protrusions may have a wave-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis, the z axis being parallel to the at least one membrane module.

The wave-like pattern may comprise a series of parallel and alternating grooves and ridges oriented perpendicularly to a direction of vibration of the at least one spacer.

The plurality of protrusions may have a hill-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis and a plurality of adjacent sinusoidal surfaces thereon along a second (y) axis, the y axis being orthogonal to the z axis, the y axis and the z axis being parallel to the at least one membrane module.

Each of the plurality of protrusions may have a protrusion height ranging from 0.5 mm to 2.35 mm and peak-to-peak distance between adjacent protrusions ranges from 3mm to 7mm.

The plurality of protrusions may have the form of an accordion fold.

The at least one spacer may have a material thickness of 0.3 mm.

The at least one spacer may have an overall cross-sectional height ranging from 1.3 mm to 5 mm.

The vibration mechanism may provide vibration to the at least one spacer at a frequency ranging from 1 Hz to 5 Hz and an amplitude ranging from 2 mm to 20 mm.

The at least one spacer may be longer and wider than each membrane in the at least one membrane module by at least an amplitude of vibration of the at least one spacer respectively.

The at least one spacer may have a surface roughness ranging from 2.9 nm to 26.3 nm.

The at least one spacer may have a plurality of through holes.

The at least one membrane module may comprise at least one flat sheet membrane. The at least one membrane module may comprise a plurality of hollow fibre membranes.

The at least one spacer may comprise a first spacer provided on a first side of the at least one membrane module and a second spacer provided on a second side of the at least one membrane module.

The at least one spacer may comprise multiple spacers and the at least one membrane module may comprise multiple membrane modules, wherein the spacers and membrane modules are provided in an alternating arrangement.

The number of spacers may be greater than the number of membrane modules by one.

The at least one spacer may comprise at least a rigid frame.

The at least one spacer may comprise a rigid sheet.

A method of membrane filtration comprising the steps of:

(a) submerging at least one membrane module configured to perform filtration of a feed fluid in the feed fluid;

(b) vibrating only at least one spacer in only directions parallel to the at least one membrane module during filtration, the at least one spacer provided next to and spaced apart from the at least one membrane module in parallel with the at least one membrane module.

The vibration may be at a frequency ranging from 1 Hz to 5 Hz and an amplitude ranging from 2 mm to 2 cm.

The vibration may be intermittent at cycles comprising up to 3 minutes of continuous vibration followed by up to 1 minute without vibration.

The vibration may comprise at least one of: linear movement of the at least one spacer and non-linear movement of the at least one spacer.

BRIEF DESCRIPTION OF FIGURES

In order that the invention may be fully understood and readily put into practical effect there shall now be described by way of non-limitative example only exemplary embodiments of the present invention, the description being with reference to the accompanying illustrative drawings.

FIG. 1 is a schematic front view illustration of an exemplary embodiment of a membrane filtration system of the present invention in an experimental configuration.

FIG. 2 is a schematic side view illustration of the system of FIG. 1.

FIG. 3(a) is a schematic diagram of an experimental set-up of the membrane filtration system in a first experimental study.

FIG. 3(b) is a schematic perspective view of a membrane module of the membrane filtration system.

FIG. 4 shows perspective and side views of an exemplary embodiment of a 2-D spacer of the

membrane filtration system in the first experimental study.

.FIG. 5 shows perspective and side views of an exemplary embodiment of a 3-D spacer of the

membrane filtration system in the first experimental study.

FIG. 6 shows graphs of size distribution of alginate (measured by Zetasizer, Malvern, UK), bentonite (measured by Mastersizer, Malvern, UK), and a mixture of bentonite and alginate (measured by Mastersizer, Malvern, UK) used in the first experimental study.

FIG. 7 shows a simulation model of (a) a full filtration setup and (b) a small section of the filtration setup showing part of a spacer and membrane surface in the first experimental study.

FIG. 8 is a schematic illustration of a side view of spacers sandwiching a membrane module depicting shear stresses and turbulence during vibrational movement of the spacers.

FIG. 9(a) is a chart of average membrane fouling rates under different test conditions in the first experimental study.

FIG. 9(b) is a chart of TMP profiles over 1 hr filtration time for repeated control experiments in the first experimental study.

FIG. 10 shows the effect of spacer type on membrane performance in the first experimental study (a) at 1 mm distance and (b) at 0.1 mm distance between spacer and membrane module (frequency of 2 Hz and amplitude of 1.2 cm).

FIG. 11 is a schematic illustration of a side view of different types of spacers in arrangement with the membrane module tested in the first experimental study.

FIG. 12 is a graph showing the effect of gas sparging on membrane performance.

FIG. 13 is graphs of x-axis component of velocity (a and d), y-axis component of velocity (b and e), and shear rate (c and f) profiles near the membrane surface with different vibrating spacers at different distances between spacers and membrane based on CFD simulation (frequency of 2 Hz and amplitude of 1.2 cm).

FIG. 14 shows images of x-axis component of velocity, y-axis component of velocity, and shear rate profiles near the membrane surface with different vibrating spacers at different distances between spacers and membrane based on CFD simulation (frequency of 2 Hz and amplitude of l.2cm)

FIG. 15 shows the effect of (a) vibration frequency (distance of 0.1 mm and amplitude of 1.2 cm) and (b) vibration amplitude (distance of 0.1 mm and frequency of 2 Hz) of 3-D spacer on membrane fouling reduction.

FIG. 16 is graphs of x-axis component of velocity (a and d), y-axis component of velocity (b and e), and shear rate (c and f) profiles near the membrane surface at different vibration frequencies (distance of 0.1 mm and amplitude of 1.2 cm) (a, b, and c) and different amplitudes

(distance of 0.1 mm and frequency of 2 Hz) (d, e, and f) based on CFD simulation of 3-D spacer vibration.

FIG. 17 shows images of x-axis component of velocity, y-axis component of velocity, and shear rate profiles near the membrane surface at different vibration frequencies (distance of 0.1 mm and amplitude of 1.2 cm) and different amplitudes (distance of 0.1 mm and frequency of 2 Hz) based on CFD simulation of 3-D spacer vibration.

FIG. 18 shows the effect of intermittent vibration of 3-D spacer on membrane fouling reduction at a vibration frequency of 2 Hz and amplitude of 1.2 cm.

FIG. 19 shows (a) calculated power consumption per membrane area for vibration spacer at different frequencies and amplitudes and (b) the relationship between membrane fouling rate and spacer vibration based power consumption per membrane area.

FIG. 20 is an AFM image of a clean membrane surface with roughness (R_a) at 45.4±2. l nm

FIG. 21 is a graph of extended operation of membrane filtration with the 3-D spacer (Flux at 20 L/m²h, distance of 0.1 mm, frequency of 2 Hz, and amplitude of 1.2 cm),

FIG. 22 is a bar chart of the TOC and turbidity of the feed and permeate under different conditions.

FIG. 23 shows five different spacer configurations studied in a second experimental study.

FIG. 24 is a schematic diagram of an experimental set-up of the membrane filtration system in the second experimental study.

FIG. 25 shows photographs of 3D printed spacers tested in the second experimental study.

FIG. 26 is a graph of particle size distribution of the model foulants used in the second experimental study.

FIG. 27 is simulation images showing x-axis component of velocity (a-c) and shear rate (d-f) near the membrane surface with 3D spacer vibration. (a,d) Spacer 1; (b,e) Spacer 2; (c,f) Spacer 3 (at a permeate flux of 20 L/m²h).

FIG. 28 is graphs of x-axis (a), y-axis (b) and z-axis (c) component of velocity and surface average shear rate (d) near the membrane surface for vibrating Spacer 1, Spacer 2 and Spacer 3 at a permeate flux of 20 L/m²h.

FIG. 29 is a graph of TMP profiles of membrane filtration with vibrating Spacer 1 and Spacer 3 at a permeate flux of 20 L/m²h.

FIG. 30 is images showing the potential flow jet during vibration of Spacer 3S and Spacer 3L.

FIG. 31 is simulation images showing x-axis component of velocity (a-c) and shear rate (d-f) near the membrane surface with 3D spacer vibration. (a,d) Spacer 3; (b,e) Spacer 3S; (c,f) Spacer 3L (at a permeate flux of 20 L/m²h).

FIG. 32 is graphs of x-axis and y-axis component of velocity and turbulence kinetic energy (a and b) and shear rate (c) near the membrane surface with vibrating Spacer 3, Spacer 3S, and Spacer 3L (at a permeate flux of 20 L/m²h).

FIG. 33 is a graph of TMP profiles of membrane filtration with vibrating Spacer 3, Spacer 3S, and Spacer 3L at a permeate flux of 20 L/m²h.

FIG. 34 is a graph of TMP profiles of membrane filtration with vibrating Spacer 3, Spacer 3S, and Spacer 3L at a permeate flux of 40 L/m²h.

FIG. 35 is a setup of a spacer vibration in a hollow fibre membrane filtration system tested in a third experimental study.

FIG. 36 is a setup of a hollow fibre membrane vibration in a hollow fibre membrane filtration system tested in the third experimental study.

FIG. 37 is a graph of effect of vibrating different spacers on critical flux in the third experimental study.

FIG. 38 is a plot of membrane performance for vibrating spacers and vibrating membranes at different vibrating frequencies.

FIG. 39 is a plot of relationship between membrane fouling rates and energy consumptions.

FIG. 40 is plots of the effect of vibration frequency (Hz) on dTMP/dt (kPa/hr) for (a) membrane vibration - horizontal arranged fiber (0% looseness) and (b) spacer vibration - horizontal arranged fiber (0% looseness).

FIG. 41 is graphs of comparison between membrane orientation for the vibrating membrane system (a) tight membrane and (b) 3% looseness membrane.

FIG. 42 is graphs of dTMP/dt versus spacer vibration frequency for vertical and horizontal arranged membrane modules with fiber looseness of (a) 0%, (b) 0.5% and (c) 3%.

FIG. 43 is a plot of effect of hollow fiber looseness (i.e., tight, 0.5% looseness and 3% looseness) on membrane movement at different frequencies. Note the fiber movement velocities in spacer vibration system are all around 0 mm/s.

FIG. 44 is graphs of dTMP/dt versus fiber looseness for (a) membrane vibration system, and (b) spacer vibration system at different frequencies.

FIG. 45 is graphs showing a comparison of the fouling control performance of spacer vibration with membrane vibration at hollow fiber looseness of (a) 0%, (b) 0.5% and (c) 3%. Horizontal arranged hollow fiber membrane module achieved lower fouling rate for both systems.

FIG. 46 is graphs of dTMP/dt versus energy consumption (/') for membrane and spacer vibration systems with (a) tight, (b) 0.5% looseness and (c) 3% looseness fibers.

FIG. 47 shows a simulation model of (a) a full filtration setup and (b) a small section of the filtration setup showing part of a spacer and membrane surface in the second experimental study.

FIG. 48 is graphs showing a comparison of (a) TKE_T and (b) mean surface-averaged shear rate along the membrane surface induced by the vibration of Spacer 1, Spacer 2 and Spacer 3.

FIG. 49 is graphs showing effect of perforations on the evolution of the (a) x-directional velocity, (b) y-directional velocity, (c) z-directional velocity, (d) surface-averaged shear rate and (e) local shear rate at the membrane surface at a permeate flux of 20 L/m²h.

FIG. 50 is bar charts showing effect of perforations on the (a) TKE_T, (b) mean surface-averaged shear rate and (c) mean local shear rate along the membrane surface at permeate fluxes of 20, 40 and 60 L/m²h.

FIG. 51 is a bar chart showing effect of perforations on the membrane fouling rate at a permeate flux of 20 L/m²h.

FIG. 52 is bar charts showing fouling rates for the three spacers at a permeate flux of (a) 40 L/m²h and (b) 60 L/m²h.

FIG. 53 is a flow chart of an exemplary embodiment of a method of membrane fdtration.

DETAILED DESCRIPTION

Exemplary embodiments of the membrane filtration system 100 and method of membrane filtration 200 will be described with reference to FIGS. 1 to 53 in which the same reference numerals are used across the figures to refer to the same or similar parts.

As shown in FIGS. 1, 2, 35 and 53, in general, the membrane filtration system 100 comprises at least one membrane module to perform membrane filtration of a feed fluid (200) in which the at least one membrane module 90 is submerged (201) during use of the membrane filtration system 100 for membrane filtration (200). The system 100 also comprises at least one spacer 10, 20 provided next to and spaced apart from the at least one membrane module 90 in parallel with the at least one membrane module 90. A vibration mechanism 30 is provided to vibrate only the at least one spacer 10, 20 (202) to create a scouring effect on the at least one membrane module 90 during use of the membrane filtration system 100. Vibration of the at least one spacer 10, 20 is in only directions parallel to the at least one membrane module 90, i.e., all movements of the at least one spacer 10, 20 are within a plane of the at least one spacer 10, 20 that is parallel to a plane of the at least one membrane module 90. The vibration may comprise linear or non-linear movements of the at least one spacer 10, 20, or a combination of both linear and non-linear movements. The at least one membrane module 90 may include at least one flat sheet membrane (FIGS. 1, 2 and 3) or a plurality of hollow fibre membranes (FIGS. 35 and 36) and is not vibrated by the vibration mechanism 30. The at least one spacer 10, 20 preferably has an uneven or non-continuously-flat surface facing the at least one membrane module 90.

In exemplary embodiments of the system 100 as shown in FIGS. 1, 2, 3 and 35, the at least one spacer may comprise a first spacer 10 comprising a rigid sheet provided on a first side of one membrane module 90 in parallel with the membrane module 90 and a second spacer 20 comprising a rigid sheet provided on a second side of the membrane module 90 in parallel with the membrane module 90. In alternative embodiments (not shown), only one spacer may be provided. In other alternative embodiments (not shown), multiple spacers may be provided together with multiple membrane modules, wherein the spacers and membrane modules are provided in an alternating arrangement, that is, in a“spacer - membrane module - spacer - membrane module - spacer and so on” arrangement. Where multiple spacers and multiple membrane modules are provided, the number of spacers may be one more than the number of membrane modules.

First Experimental Study

The aim of this study is to investigate the feasibility of spacer vibration in controlling submerged flat sheet membrane fouling and examine the different vibration factors influencing fouling mitigation efficiency, including spacer configuration, the distance between spacers and membranes, and vibration conditions (frequency, amplitude, and intermittent vibration).

A mixture of particulate (bentonite) and macromolecular (sodium alginate) was used as model foulants (Kola et al., 2012). Three-dimensional computational fluid dynamics (3-D CFD) simulations were also performed in order to illustrate the profiles of flow velocity and shear rate near the membrane surface. A preliminary estimation of the energy required to generate oscillating spacer motion was calculated and compared with conventional gas sparging method. This study highlights the feasibility of using spacer vibration for membrane fouling control in submerged membrane filtration systems, such as (1) anaerobic and aerobic membrane bioreactors in treating different types of water and wastewater; (2) membrane separation processes that where cake layer fouling is predominant.

In this study, an exemplary system 100 was set up as shown in FIG. 3(a) in which the membrane module 90 (FIG. 3(b) comprised two parallel pieces of flat sheet membranes 91 (effective membrane area of 5 cm ^c 7 cm = 35 cm² per piece of membrane 91; so total membrane area is 70 cm²) were mounted into a membrane module 90 of 8 cm ^c 12 cm, in a sandwich-like module configuration. Each flat sheet membrane 91 was made of polyvinylidene fluoride (PVDF) and had a nominal pore size of 0.08 pm. The membrane module 90 held on a module support was submerged in a tank 80 (effective volume of 2 L, 13.8 cm (L) ^c 8.2 cm (W) ^c 17.7 cm (H)) containing feed fluid 300.

Different sets of two spacers 10, 20 comprising rigid sheets were provided, one on each side of the membrane module 90. The spacers 10, 20 were all each 10 cmx lO cm in size to ensure full coverage of at least the surface area of the membranes 91 in the membrane module 90. The two spacers 10, 20 were connected to the vibration mechanism 30. The vibration mechanism 30 comprised a motor 31 driving a crank 32 connected to a first end of a slider 33 to effect reciprocating linear motion of a spacer holder 34 provided at a second end of the slider 33. The two spacers 10, 20 were attached to the holder. The membrane module 90 and two spacers 10, 20 were provided vertically. The two spacers 10, 20 were each provided at a fixed distance away from the membrane module 90. Two fixed distances of 0.1 mm and 1 mm were tested. Distance between each spacer 10, 20 and the membrane module 90 was regulated by screws (not shown) linking the spacer holder 34 with each spacer 10, 20. Vibration frequency ranged from 1 Hz to 2.5 Hz and amplitude (i.e., maximum displacement of both spacers 10, 20) ranged from 0.8 cm to 2 cm.

A first set of the spacers 10, 20 (not shown, hereinafter referred to as l-D spacers in this First Experimental Study) that were tested comprised acrylic plates of thickness or overall cross-sectional height of 3 mm and weight of 27 g each having a flat smooth surface facing the membrane module 90.

A second set of the spacers 10, 20 as shown in FIG. 4 (referred to as 2-D spacers in this First Experimental Study) had non-continuously-flat surfaces facing the membrane module 90, the non- continuously-flat surfaces each having grooves formed on a flat surface with an exemplary surface roughness R_a of 3.9 ± 1.0 nm. The 2-D spacers 10, 20 were made of acrylic material and had a weight of 27 g. As shown in FIG. 4, the 2-D spacers 10, 20 had an overall cross-sectional height of 3 mm, with grooves having a width of 0.5 mm, a depth of 1 mm, and adjacent grooves being 3 mm apart. The grooves were formed by laser engraving.

A third set of the spacers 10, 20 as shown in FIG. 5 (referred to as 3-D spacers in this First Experimental Study) had non-continuously-flat surfaces facing the membrane module 90, the non- continuously-flat surfaces each comprising a surface having a plurality of protrusions thereon and a surface roughness R_a of 25.6 ± 0.7 nm. The plurality of protrusions had a hill-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis and a plurality of adjacent sinusoidal surfaces thereon along a second (y) axis, the y axis being orthogonal to the z axis and both the z and y axes being parallel to the membrane module. As shown in FIG. 5, the 3-D spacers 10, 20 had a hill or protrusion height of 0.5 mm (protrusion height being the height of each protrusion from a central plane of the spacer), an overall cross-sectional height of 1.3 mm, a peak-to-peak distance between adjacent protrusions of 3 mm, and material thickness of 0.3 mm.

The sinusoidal surfaces of the 3-D spacers 10, 20 followed a general sine function given in Eq. 1 below: F (x) as A$m(Bx~~C) -f D

[Eq. 1]

2

For example, the parameters A = 0.45, B p , D = 0 were selected for the configuration

3

of the 3-D spacers 10, 20.

The 3-D spacers 10, 20 were formed by 3-D printing with a EOSINT P395 Selective Laser Sintering (SLS) machine (EOS e -Manufacturing Solutions, Germany) using polyamide-l2 based powder PA2200 material (EOS GmbH Electro Optical Systems, Germany). SLS is a powder-based 3-D printing technique whereby the 2-D cross sections of the spacer model at each layer height are scanned by a carbon dioxide (C0₂) laser to sinter the powder particles together. The processing parameters for PA2200 were based on the EOS part properties profile. Residual support materials were removed in an abrasive blasting machine. It is noted that for mass production of 3-D spacers, other materials and manufacturing technique (such as moulding) can be used.

Filtration Procedure

A mixture of bentonite (1 g/L, Sigma- Aldrich, US) and alginate (1 g/L, Sigma- Aldrich, US) was used as a model solution 300 to simulate particulate and macromolecular foulants (the size distribution is presented in FIG. 6). The membrane module 90 was operated at a constant permeate flux of 20 U/m²h and the permeate flow was achieved by a peristaltic pump 220. The suction pressure was measured at 6-s interval by a pressure transducer 210, which was connected to a computer with a data-logging system (Uabview, NI, US).

The filtration experiment was performed for 1 h, which was typically the time needed to reach a quasi-steady state as emphasized in previous studies (Ui et al., 2013; Pourbozorg et al., 2017). For comparison, the conventional gas sparging for fouling control, without spacers, was performed. In this study, an air flow rate at 0.5-4 U/min (equivalent to specific membrane aeration demand ranging from 8.6 to 68.6 Nm³/m²h) was employed to achieve comparable membrane fouling control effectiveness as vibrating spacers. It is noted that such specific membrane aeration demand was much higher than that in real submerged flat sheet membrane systems (such as membrane bioreactors, MBRs) due to generally more compactness for the real submerged flat sheet membrane system.

Before each experiment, the clean membrane module 90 was filtered with distilled water for 10 min and the transmembrane pressure (TMP) data were recorded (TMP_m, ~5 kPa at 20 U/m²h). The foulant mixture was then added into the filtration tank and the filtration was performed for 1 h (the final TMP was defined as TMP,). The TMP induced by membrane fouling (TMP_f) was calculated as TMP_f = TMP, - TMP_m and was ploted against filtration time. The average membrane fouling rate (kPa/h) was calculated as fouling rate = (TMP, - TMP_m)/filtration time.

At the end of filtration, the membrane module 90 was carefully taken out from the tank 80 and the cake layer on the membranes was removed by flushing the membranes with tap water for 2 min. The physically-cleaned membrane was filtered with distilled water for 10 min and the TMP data were recorded (TMR_{ώeG cleaning})· The TMP increase due to cake layer fouling (TMP_C) was calculated as TMP_C = TMP, -

After the experiments had been conducted, the membrane was cleaned using 0.5% of hypochlorite for 5 min under ultrasonication conditions (28 Hz, 60 W) and a membrane clean water test was carried out to ensure no membrane integrity loss before reuse. For each test condition, repeated filtration experiments (n>3) were performed and the averaged data obtained.

Three-dimensional (3-D) computational fluid dynamics (CFD) simulation

The 3-D CFD simulations of a section of the filtration system 100 comprising the 3-D spacers 10, 20 described above with reference to FIG. 5 were carried out using Comsol Multi -Physics 5.1. The Multi-Physics package used in the simulation relates to the fluid-structure interaction with laminar flow fluid and the governing equations (the Navier-Stokes and continuity equations) are described in the Appendix. Since the filtration setup is symmetrical in the x- and y-axes (FIG. 7(a)), a small section (FIG. 7(b)) comprising part of one spacer and membrane surface of the full system 100 was selected and simulated. The model solution was assumed to be incompressible at 25°C, which is similar to the lab room temperature during the filtration experiments. The fluid velocity perpendicular to membrane surface was simulated based on the experimental flux at 20 L/m²h. An open boundary condition was set for the top of the tank to account for the water surface movements generated by spacer vibration.

In the simulation, the 3-D spacer was assumed as solid polyamide material from Comsol materials database, with sinusoidal surface similar to that used in the experiment, i.e., a layer thickness of 1.3 mm. The movement of the spacer was described with a movement of 0 in both x and y-axis and amplitude ^c £ίh(2p x frequency x time) in the z-axis. The induced liquid flow velocity and shear rate at x-axis that are associated with back-transport of foulants were then simulated from 0 to 2 s with an interval of 0.1 s. Different predefined mesh sizes with tetrahedral elements gave similar results, thus the coarse mesh was selected for the tank and the membrane, while a fine mesh was used for the 3-D spacer to shorten the computation time while ensuring that small details on the 3-D spacer were properly meshed.

Turbulence kinetic energy (TiCE) calculation Turbulence kinetic energy (TKE), the mean kinetic energy associated with eddies in turbulent flow, has proven to be associated with membrane fouling alleviation by gas sparging (Yeo et ah, 2007) or turbulence promoter (Pourbozorg et ak, 2016, 2017). TKE can be quantified based on the velocity fluctuations in /-direction (ii ') described by Eq. 2 below:

The velocity fluctuations in /-direction (u{) can be calculated based on Eq. 3 below: u[— Ui— iq [Eq.3]

Where u, is the velocity in /-direction and the mean velocity in /-direction.

In this study, as the movement of the spacers 10, 20 was parallel to the membrane surface (z-axis) at a very close distance (1 mm and 0.1 mm) and back-transport of foulants was along the x-axis and y-axis, TKE_X values were calculated using Eq. 4-5 and TKE_y values were calculated using Eq. 6-7 below:

u_x' = u_x - v7_x [Eq.5]

TKE_y = i ( z) [Eq. 6]

Where u_x and u_y are the velocity in x-direction and -di recti on respectively, obtained by CFD simulation;

and vJ are the mean velocity in x-direction and v-dircction: u 'and u_y' is the velocity fluctuations in x-direction and y-di rection.

Turbidity, TOC examination, and AFM measurement After each filtration experiment, the turbidity and total organic carbon (TOC) of the permeate sample was examined to check whether membrane integrity was maintained. The turbidity was measured by a turbidity meter (Hach, US). The total organic carbon (TOC) of the sample (after being filtered through a 0.45 pm hydrophilic filter) was monitored using a TOC analyzer (Shimadzu, Japan). The roughness of the spacer surface was measured by an atomic-force microscopy (AFM, Park Systems, South Korea).

Effect of spacer configuration on membrane fouling reduction

In this study, as shown in FIG. 8, the spacers (l-D, 2-D, and 3-D spacers 10, 20) were located at each side of the membrane module 90, which is expected to induce turbulence (curved arrows) and shear stress (grey arrows) along the membrane surfaces 91, in directions opposite to movement (white arrows) of the spacers 10, 20. Similarly to membrane vibration systems, the boundary fluid layers around the membrane are thought to be mobilized in the membrane filtration system 100 with vibrating spacers (Li et ah, 2013). A series of experiments were conducted to determine the impact of different types of spacers and spacer locations on membrane performance and to compare the efficiencies of membrane fouling alleviation by spacer vibration and gas sparging. The averaged membrane fouling rates under different test conditions are summarized and depicted in FIG. 9.

After filtration, the cake layers were physically removed and the cake layer fouling was identified. The results indicated that the cake layer fouling was predominant (over 95% of the total fouling) under each test condition (data not shown).

FIG. 10 shows the effect of spacer type on membrane performance at (a) 1 mm distance and (b) 0.1 mm distance between each of the spacers 10, 20 and the membrane module 90, at a spacer vibration frequency of 2 Hz and amplitude of 1.2 cm. FIG. 11 shows the different types of spacers 10, 20 in arrangement with the membrane module 90 during the experiments. The effect of gas sparging on membrane performance is shown in FIG. 12. Under different spacer vibration conditions, two different stages of TMP profiles were noticed, as can be seen in FIGS. 10(a) and 10(b), in agreement with those under gas sparging conditions (FIG. 11). An initial TMP jump was thought as the rapid accumulation of cake layers that could not be readily removed by spacer vibration, possibly due to the stronger interaction of the clean membrane surface with the foulants. Furthermore, a slow linear rise of TMP (i.e., slow cake layer deposition) was observed as possibly the induced shear force by spacer vibration was comparable to the interaction force between foulants. Compared to the control condition (dead-end; TMP_f reached ~37 kPa), vibrating spacers could slow down the initial TMP increase phase and significantly enhance the membrane performance (TMP_f reached -7-17 kPa), regardless of spacer types and locations. When the spacers 10, 20 were located at 1 mm of distance from the membrane module 90, the average membrane fouling rate for vibrating l-D, 2-D, and 3-D spacers 10, 20 was 15.3±2.1, l7.3±3.4, and l3.7±1.8 kPa/h, respectively (FIG. 10(a) and FIG. 9), indicating almost comparable influence of spacer type on the membrane fouling alleviation (/;>0.05. T-test). When the distance between the spacers 10, 20 and membrane module 90 was reduced to 0.1 mm, the average membrane fouling rate was l3.3±4.5, 8.9±3.2, and 6.9±1.3 kPa/h for l-D, 2-D, and 3-D spacer, respectively (FIG. 10(b) and FIG. 9). It was observed that the 3-D spacers 10, 20 displayed slightly higher efficiency in mitigating membrane fouling than the 2-D spacers 10, 20, both of which had significantly better fouling control performance than the l-D spacers 10, 20 (/;>0.05. T-test). In addition, almost 13%, 48%, and 49% of membrane performance was improved at a distance of 0.1 mm compared to those at a distance of 1 mm for the l-D, 2-D, and 3-D spacers 10, 20, respectively.

Further CFD model simulation was performed to elucidate the liquid flow velocity at x-axis and y- axis and shear rate profiles (FIG. 13 and 14). The turbulence kinetic energy values (TKE_X and TKE_y) were calculated based on the liquid velocity profile (FIG. 13). From FIG. 13, it can be observed that the 3-D spacer movement could induce more liquid velocity fluctuations at both x-axis and v-axis directions, therefore higher TKE_X (3.0x 10 ⁶ and 5.3 x lO ⁶ m²/s² at 1 and 0.1 mm of distance respectively) and TKE_y (4.8^c 10⁷ and 3.5^c 10 ⁶ m²/s² at 1 and 0.1 mm of distance respectively) was achieved to enhance back-transport of foulants compared to the movement of l-D (TKE_X of 2.4 ^c 10 ⁸ and 5.2x 10⁷ m²/s² and TKE_y of l .Ox 10 ⁷ and 3.8x 10⁷ m²/s² at 1 and 0.1 mm of distance respectively) or 2-D spacer (TKE_X of 2.3 KG⁴' and 5.0 KG⁸ m²/s² and TKE_y of 3.4x l0 ³⁵ and 9. K M ⁷ m²/s² at 1 and 0.1 mm of distance respectively) at the same vibration conditions. This suggests that the surface morphology of the 3-D spacer (FIG. 5) could enhance the turbulence along the membrane surface, as a result, alleviating membrane fouling.

In addition, the 3-D spacer movement generated less shear rate compared to the l-D and 2-D spacers at both distances of 0.1 mm and 1 mm, as shown in FIGS. 13(c) and (f). It is well known that shear rate for a fluid between two parallel plates is defined as the ratio of moving plate velocity and distance between two plates. In this study, the three spacers had the same moving velocity under the same frequency and magnitude condition. As the same distance of the hill-like part of the spacer to the membrane surface was maintained, the distance from the each point of the spacer to the membrane surface for the three spacers were different (FIG. 11), i.e., the overall distance of the 3-D spacer to the membrane surface was greater than those of the l-D and 2-D spacers. This may lead to less shear rate induced by the 3-D spacer. While, at a distance of 0.1 mm, the shear rate induced by 2-D spacer was slightly lower the l-D spacer, but at a distance of 1 mm, the shear rate induced by 2-D spacer was higher than l-D spacer. It is noted that grooves in the 2-D spacer were deep (1 mm) and narrow (0.5 mm) (FIG. 4), which may limit the mass transfer rate of liquid inside of these grooves. These liquid flows possibly led to an influence on the secondary flow shear rate induced by the 2-D spacer movement, depending on the distance between the membrane and spacer.

In a previous study on the use of air sparging for hollow fibre fouling reduction (Yeo et al., 2007), it was found the fouling alleviation can be correlated to both the mean shear and TKE around the fibres induced by the bubbling, and both effects had almost comparable contribution in fouling reduction. In this study, although the contributions of TKE and shear rate to membrane fouling alleviation were not elucidated, it was also thought that the combined effects of TKE and shear rate could determine membrane fouling reduction by spacer vibration. Possibly, this could explain the phenomena that there was almost no difference in membrane performance for all the spacers at a distance of 1 mm (FIG. 10(a)), while the 3-D spacer could significantly enhance the membrane performance compared to l-D and 2-D spacer at a distance of 0.1 mm (FIG. 10(b)).

Furthermore, the CFD simulation results (FIG. 13 and FIG. 14) reveal that reducing the distance between spacers and membrane surface from 1 mm to 0.1 mm, both turbulence kinetic energy values at x-axis and y-axis and shear rate increased, almost regardless of spacer configuration. This corresponded well with the experimental results (FIGS. 10(a) and (b)).

In addition, the membrane performance by gas sparging at an air flow rate of 0.5-4 L/min was examined (FIG. 12). It was observed that the membrane fouling reduction with vibrating 3-D spacers at a condition of distance of 0.1 mm, frequency of 2 Hz, and amplitude of 1.2 cm was comparable to that with air sparging at flow rate of 1 L/min. This observation is similar to hollow fibre membrane vibration systems which also had comparable membrane performance as gas sparging (Kola et al., 2014; Li et al., 2013). It is worth noting that the local shear rate along the membrane surface in the presence of gas sparging is associated with bubble path, size, and shape (Cui et al., 2003). It is therefore not justified to simply compare membrane performance without consideration of energy consumption. Further comparison of energy consumptions of spacer vibration and gas sparging for fouling control will be discussed below.

Effect of spacer vibration condition on membrane fouling reduction Vibration frequency and amplitude

As the 3-D spacer showed the most effective membrane fouling alleviation (FIG. 10), the influences of 3-D spacer vibration frequency (1-2.5 Hz) and amplitude (0.8-2 cm) on membrane fouling efficiency were examined at continuous vibration and at a distance of 0.1 mm. FIG. 15 plots the TMP profiles at different vibration frequencies. By increasing the frequency from 1 Hz to 1.5 Hz, and to 2 Hz, the membrane fouling rate was significantly decreased from l l.l±0.9 to 9.7±1.8, and to 6.9±1.3 kPa/h (FIG. 9). Further increasing vibration frequency from 2 to 2.5 Hz had little influence on the membrane fouling rate (6.9±1.3 vs. 7.5±0.8 kPa/h. /;>().05) (FIG. 9). FIG. 15b shows the impact of vibration amplitude on the membrane fouling development. Similarly, increasing the amplitude from 0.8 to 1.2, to 2.0 cm, the membrane fouling rate dropped from 9.3±1.2 to 6.9±1.3 kPa/h, and to 6.l±0.8 kPa/h (FIG. 9). Similarly, insignificant improvement of membrane performance was noticed with increasing amplitude from 1.2 to 2 cm. Furthermore, the CFD simulation results (FIG. 16 and FIG. 17) revealed that increasing vibration frequency and amplitude enhanced the magnitude of fluid motion away from the membrane surface at x-axis and v-axis directions (i.e., turbulence kinetic energy) as well as increased shear rate. Previous studies with fluidized granular activated carbon (GAC) as scouring agents for membrane fouling control have demonstrated that membrane fouling reduction due to the GAC scouring effect was strongly correlated with the fluidized GAC momentum, which was positively related to the fluidized GAC mass and velocity (Wang et ah, 2016). In addition, in a recent study with hollow fibre membrane fouling control by turbulence generators (perforated plates) (Pourbozorg et ah, 2016, 2017), membrane fouling alleviation was improved by increasing turbulence kinetic energy, which was positively correlated with turbulence amplitude and frequency.

It should be noted that with increasing the frequency from 2 to 2.5 Hz or the amplitude from 1.2 to 2 cm, the TKE at both x-axis and v-axis directions as well as the shear rate induced by the vibrating 3-D spacers were enhanced (FIG .16). The improved secondary flow turbulence and shear rate are expected to remove more foulants away from the filtration zone. However, the filtration experiments revealed that when vibration frequency or amplitude was beyond the‘threshold’ level (2 Hz and 1.2 cm respectively), the membrane fouling alleviation efficiency reached a‘plateau’ (FIG. 9). It is due to the shear force at frequency above 2 Hz or at amplitude above 1.2 cm would exceed the permeate drag force, leading to no further benefits for fouling control. Such‘plateau’ phenomenon is also observed in membrane vibration systems (Kola et ah, 2012) and gas bubbling systems (Cui et ah, 2003).

Intermittent vibration

In the spacer vibration membrane filtration system 100, the energy consumed by the motor 31 for vibrating spacers 10, 20 is a major contribution to the overall energy requirement. Previous studies with membrane vibration (Bilad et ah, 2012; Li et ah, 2014) have found that vibration relaxation at an optimal time interval offers adequate membrane fouling control while ensuring reduced energy consumption. It was therefore necessary to investigate intermittent spacer vibration with continuous permeation in this study in order to achieve high membrane performance and energy efficiency. FIG. 18 depicts the membrane fouling development at different intermittent modes for 3-D spacer vibrations (at a frequency of 2 Hz, amplitude of 1.2 cm, distance of 0.1 mm). When the intermittent mode was performed at 1 min of vibration and 9 min of no vibration, within the first 4-cycles, nearly the same TMP_f level was observed during the no vibration period as the control, although during the 1 min of spacer vibration the TMP_f was reduced modestly by ~2-6 kPa. This suggests that the cake layer formed within the 9-min of no vibration period could not be fully removed by the turbulence and shear stress due to spacer vibration (l-min). Further extending the operation period for this intermittent mode (1 min on/9 min off), the TMP_f tended to be slightly lower than that without any spacer vibration. Such finding is possibly related to the fact that the 3-D spacer was close to the membrane surface (0.1 mm) and could prohibit the further accumulation of the cake layers on the membrane surface when the cake layer thickness reached a certain level.

In further tests, the spacer vibration time was fixed at 1 min and the non-vibration time varied. It was observed that, by increasing the ratio of vibration time to idle time from 1/9, to 1/3, and to 1/1, the membrane fouling mitigation efficiency significantly improved (FIG. 18). However, the membrane performance was only partially recovered during the spacer vibration period. This illustrates that at a fixed vibration time period (i.e., fixed turbulence and shear effectiveness), fouling control efficiency of spacer vibration was associated with the thickness of the formed cake layers during the idle period. By extending the vibration period from 1 min to 3 min (i.e., increasing the turbulence and shear effectiveness) with 1 min of idle time, the TMP_f after 3-min spacer vibration was almost similar to that with continuous spacer vibration. In addition, the membrane performances at the intermittent modes of 1 min on/l min off and 5 min on/5 min off were relatively comparable. These findings emphasize that cake layer removal efficiency was dependent on the cake layer thickness and vibration intensity, which is consistent with the findings in the study with fluidized GAC as scouring agents (Wu et ak, 2016).

Comparison of power consumption of spacer vibration and gas sparging

The energy consumption associated with membrane fouling control is the major contributor of the overall energy consumption in a membrane separation process. Therefore, reduction of energy associated with fouling control is an important target to achieve sustainable membrane operation. In this study, as the spacer vibration system was attempted to replace the coarse bubble aeration for membrane fouling control in the submerged flat sheet system (i.e., the energy consumption for other components were the same), the power consumptions for spacer vibration and coarse bubble aeration were compared.

In a spacer vibration system, the power consumption is defined by Eq. 8 below (Genkin et ah, 2006):

Power (W) = ma2FA [Eq.8] where m is the moving mass (kg), a the average acceleration (m/s²), F the frequency (Hz) and A the amplitude (m). The average acceleration (m/s²) is calculated as 2AF². In this study, the 3-D spacer had a weight of 5.3 g with an area of 0.01 m² (note that one spacer was located between two membrane sheets, thus the effective membrane area was 0.02 m² per 0.01 m² spacer area). The power consumption per membrane area (W/m²) in terms of spacer vibration was estimated and plotted based on amplitude and frequency in FIG. 19. It was observed that increasing vibration amplitude and frequency lead to higher power consumption for membrane fouling mitigation. It was also noted that the membrane fouling alleviation by spacer vibration was not linearly correlated with its required power consumption. Therefore, an optimal vibration condition needed to be identified towards achieving lower fouling rate with less power consumption based on the relationship between membrane fouling and its required power consumption.

The gas sparging energy is generally calculated based on the power consumption of blowers for the submerged membrane configuration. In this study, the specific membrane aeration demand ranged at 8.6-68.6 Nm³/m²h (i.e, air flow rate at 0.5-4 L/min), much higher than that in real submerged membrane bioreactors (MBRs) due to generally more compactness for the real submerged flat sheet membrane system. To make a fair comparison, a specific membrane aeration of 0.3 Nm³/m²h in conventional MBRs was chosen (Judd and Judd, 2006). A power consumption of 0.019 kWh/Nm³ air for coarse bubble aeration was taken for calculating the gas sparging energy for membrane fouling control (Maere et ah, 2011). Thus, the power consumption of gas sparging was estimated to be 5.7 W/m², which was almost three-order of magnitude higher than that of the spacer vibration system in terms of membrane fouling control.

Implications and perspective

This study experimentally proved the feasibility of a physical membrane fouling control strategy by vibrating spacers in a submerged flat sheet membrane system. Two major advantages of spacer vibration compared to other physical membrane cleaning approaches have been highlighted. For example, (1) the geometry of spacers can be tailor-made. This will allow the spacers to be easily integrated into the existing commercial flat sheet membrane module products. In addition, apart from membrane fouling control, the spacers could also be designed to enhance process performance. For example, the spacers can be made wholly or in part of adsorbent materials (regeneration needs to be considered), which can improve micropollutant removal in membrane -based water treatment. For example, the spacer may be made of an adsorbent material which has a rigid nature. Alternatively, the spacer may comprise a rigid sheet which is coated with a layer of adsorbent material. In another example, the spacer may comprise a rigid frame surrounding and supporting a layer of adsorbent material within the rigid frame. For biotreatment, the spacers can be made of biocarrier-like materials, which act as biocarriers to facilitate biofilm growth in biofilm membrane reactors. (2) The spacer can be made of low density materials (e.g., 530 g/m² in this study) and spacer vibration can be performed at a low frequency (2 Hz) and small amplitude (1.2 cm) for effective membrane fouling control, therefore requiring extremely lower energy consumption compared to air sparging. Under the same vibration conditions (frequency and amplitude), spacer vibration also uses less energy compared to membrane vibration due to much heavier weight of membrane modules, especially with permeate inside it. In addition, the light-weight spacer nature guarantees that less mechanical support is required to connect the motor with the spacers compared to that with the vibrating membranes.

It is worth noting that when the spacers are very close to the membrane surface, membrane integrity may be adversely affected if a rough spacer surface is present. In this study, the tested spacers had very smooth surfaces (R_a at 3.9±1.0 nm for the l-D and 2-D spacers, at 25.6±0.7 nm for the 3-D spacer) and their roughness was even lower than that of the clean membrane surface (FIG. 20, R_a at 45.4±2. l nm). In addition, to assess the extent of this potentially negative effect during a long-term filtration, a membrane filtration experiment was performed for 4.5 days (FIG. 21). The permeate turbidity and TOC data (FIG. 22) examined after each experimental condition illustrated that the vibrating spacers did not lead to membrane integrity loss during the tests. It is foreseen that spacer vibration as an energy-efficient fouling control approach can be applied in membrane bioreactors for wastewater treatment and other membrane separation processes where cake layer fouling is predominant.

Conclusions

The first experimental study and analysis illustrated the membrane fouling control efficiency of spacer vibration under different conditions. The conclusions can be drawn as below:

(1) At a distance of 0.1 mm between the spacers and membrane, the 3-D spacer more efficiently alleviated membrane fouling than l-D and 2-D spacers.

(2) Increasing vibration frequency and amplitude of the 3-D spacer, the membrane fouling alleviation was improved due to enhanced turbulence and shear rate as illustrated by CFD simulation. However, when vibration frequency or amplitude was beyond the ‘threshold’ level (2 Hz and 1.2 cm respectively), the membrane fouling alleviation efficiency reached a‘plateau’.

(3) Intermittent vibration of the 3-D spacer at an interval of 3 min on/l min off was able to achieve comparable membrane fouling control as continuous vibration, offering a solution to reduce energy consumption. (4) Spacer vibration (<0.008 W/m²) consumes significantly less power in terms of fouling control than gas sparging (5.7 W/m²) under the comparable fouling control effectiveness.

In this study, vibrating spacers were investigated to control fouling in a submerged flat sheet membrane filtration system. The impact of spacer types (l-D, 2-D, and 3-D), parallel distance of spacers to the membrane module, vibration frequency and amplitude, and intermittent vibration interval on membrane fouling potential were examined. The results showed that vibration of 3-D spacers (sinusoidal geometry) could achieve up to 48% and 25% fouling reduction compared to l-D (flat plate) and 2-D (flat plate with grooves) spacers respectively when the spacers were located at a distance of 0.1 mm to the membrane module. Increasing vibration frequency of the 3-D spacer from 1 to 2.5 Hz and increasing amplitude from 0.8 to 2 cm improved membrane fouling control efficiency by 33% and 34% respectively. The improved membrane performance was attributed to combined effects of the enhanced turbulence kinetic energy associated with back-transport of foulants and shear rate on the membrane surface as illustrated by the three-dimensional computational fluid dynamics (3- D CFD) simulations. An experimental and power consumption comparison of spacer vibration and gas sparging for fouling control was also performed. This study highlights the feasibility of vibration spacers in enhancing submerged flat sheet membrane performance.

Second Experimental Study

The aim of this study was to optimize 3D spacer configuration by computational fluid dynamics (CFD) simulation and experimental testing. In this study, membrane filtration performances with vibration of different spacers at different permeate fluxes were compared using a mixture of bentonite and sodium alginate as model foulants. Several types of 3D spacers, namely, hill-like pattern, wave-like pattern, wave-like pattern with perforated small holes and wave-like pattern with perforated large holes were designed and prepared by 3D printing techniques. CFD simulations were performed to compare the flow profiles (in terms of kinetic turbulence energy) and shear rates near the membrane surface with vibrating these 3D spacers. Meanwhile, the effect of 3D spacer configuration on the membrane performance was illustrated experimentally in a lab-scale submerged flat sheet membrane filtration system performed at a constant flux, with a mixture of bentonite and sodium alginate as feed. The transmembrane pressure (TMP) development patterns from the experiments were combined with the CFD simulation results to give insights on the important design parameters of 3D spacers for maximizing the effectiveness of spacer vibration-based membrane fouling control.

CFD Simulations

Since the filtration setup (FIG. 47(a)) was symmetrical in the x and y axes, a small section (FIG. 47(b)) of the setup was selected for simulation. Three-dimensional (3D) computational fluid dynamics (CFD) simulations of the section of the filtration setup as shown in FIG. 47(b) were carried out using Comsol Multi -Physics 5.1. The simulations aimed to provide a basis to identify key design parameters in terms of liquid velocity and shear rate near the membrane surface. The Navier-Stokes and continuity equations were used as governing equations (Appendix), describing the laminar flow fluid-structure interaction. The liquid in the tank was set to be incompressible liquid water at 25°C, which is similar to the lab room temperature during performing the filtration experiments. The permeate flux across the membrane was set to 20, 40 or 60 L/m²h. An open boundary condition was set for the top of the tank as the moving spacer led to movements of the surface water.

In this study, five types of pairs of 3D spacers as shown in FIG. 23 were designed and tested, hereinafter referred to as Spacer 1, Spacer 2, Spacer 3, Spacer 3S and Spacer 3L. All five types of the 3D spacers had a material thickness of 0.3 mm as shown in FIG. 23, in which the vertical double- ended arrows in the left column indicate direction of vibration or movement of the spacers.

Spacer 1 (FIG. 23(a)) was identical with the 3-D spacer (shown in FIG. 5) that was tested in the first experimental study as described above.

Spacer 2 (FIG. 23(b)) comprised a non-continuously-flat surface having a plurality of protrusions thereon, the plurality of protrusions having a wave-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis, the z axis being parallel to the membrane module. Spacer 2 can thus also be said to comprise a series of parallel and alternating grooves and ridges formed by the wave-like pattern on the spacer. Spacer 2 had a protrusion height of 2.35 mm, an overall cross-sectional height of 5 mm, and a peak-to-peak distance between adjacent protrusions of 7 mm. As can be seen in FIG. 23(b), the Spacer 2 was oriented with its grooves and ridges parallel to a direction of spacer vibration.

Spacer 3 (FIG. 23(c)) was identical to Spacer 2 but oriented orthogonally to the orientation of spacer 2 in the study so that Spacer 3 was oriented with its grooves and ridges perpendicular to the direction of spacer vibration.

Spacer 3S (FIG. 23(d)) was similar to Spacer 3 but included through holes of 1 mm diameter in the spacer.

Spacer 3L (FIG. 23(e)) was similar to Spacer 3 but included through holes of 2 mm diameter in the spacer. The through holes in spacer 3S and 3L were at positions that were 50° to the horizontal axes, because preliminary simulations indicate that this maximized the fluid flow towards the membrane through the upward-facing perforations and back-transport away from the membrane through the downward facing perforations during spacer vibration.

Solid polyamide from Comsol materials database was selected as spacer material and the liquid velocity and shear rate induced by these spacers were simulated at an amplitude of 1.2 cm and a frequency of 2 Hz. The movement of the spacer was simulated with a prescribed movement of 0 in both x and y-axis and amplitude ^c siiiln ^c frequency ^c time ) in the z-axis. A time -dependent study was then carried out from 0 to 1 s with an interval of 0.02 s. A point was generated along z-axis at the 0.05 mm away from the centre of the membrane to obtain plots of shear rate and x-axis component of velocity at a 0.2 mm distance from the surface of the membrane over the period of ls. Different predefined mesh sizes with tetrahedral elements gave similar results, thus the coarse mesh was selected for the tank and the membrane, while a fine mesh was used for the 3D spacer to shorten the computation time while ensuring that small details on the 3D spacers are properly meshed.

The various 3D-spacers were simulated and the local x-directional velocities, y-directional velocities, z-directional velocities, shear rates and surface-averaged shear rate were compared. The local x- directional velocities, y-directional velocities, z-directional velocities and shear rates were taken near the perforations or protrusions to investigate the localized effect of the unique designs, while the surface-averaged shear rates reflect the representative shear rates across the entire membrane.

Turbulence kinetic energy (TKE) calculation

Other than the x-directional velocity, y-directional velocity, z-directional velocities and shear rate, previous studies have shown the turbulence kinetic energy (TKE) is associated with membrane fouling alleviation by gas sparging (Yeo et ah, 2007) or turbulence promoter (Pourbozorg et ah, 2016, 2017). TKE is the mean kinetic energy associated with eddies in turbulent flow and can be calculated based on the fluctuating mean velocity described by Eq. 9 below:

[Eq. 9] where w,’ is derived using Eq. 10:

Where Ui is the velocity in the i-direction and Ui is the root mean square (RMS) value of the amplitude of fluctuations from the mean velocity obtained from the peaks and troughs of the velocity fluctuations over time in the i-direction.. In this study, the movement of the spacers along the z-axis was partially translated into fluid motion along the x and y axis, which resulted in eddies aiding membrane fouling alleviation through increasing shear and improving back-transport of foulants. Hence, the average TKE_X, TKE_y and TKE_Z were calculated based on the CFD simulated velocities along the x, y and z directions (i.e.. m_c, m_n and m_z) using the following equations:

[Eq. 11]

U_x ~ U ®i_x [Eq. 12] [Eq. 13] [Eq. 14]

ll_z = u_z— u_z [Eq. 15]

TKE_z = i (¾) [Eq. 16] ft, , ft and f- are the mean velocities in the x-, y- and z-directions. respectively; and u , u\ and u'_y are the velocity fluctuations in the x-, y- and z-directions, respectively. The total TKE, TKE_T can then be derived from the summation of TKE_X, TKE_V and TKE_Z as follows.

TKEc = TKE_X + TKE_V + TKE_: [Eq. 17]

Filtration system setup

FIG. 24 shows the schematic diagram of a lab-scale submerged flat sheet membrane filtration system 100. A flat sheet membrane module 90 consisting of two pieces of polyvinylidene fluoride (PVDF) membranes (a nominal pore size of 0.08 pm, effective filtration area of 5 cm ^c 7 cm = 35 cm²) was submerged into a tank 80 having an effective volume of 2 L (dimensions being 13.8 cm (L) ^c 8.2 cm (W) x 17.7 cm (El)). Two pieces of spacers 10, 20 (10 cmx lO cm) were provided on each side of the membrane module 90 in a sandwich configuration. A vibration mechanism 30 was provided to vibrate the spacers 10, 20 by mounting the spacers 10, 20 onto a spacer holder 34 connecting to a motor 31, which drove the spacers 10, 20 in linear reciprocating motion via a crank-slider mechanism. The spacer holder 34 can be adjusted to ensure that the distance between each spacer 10, 20 and membrane surface was ~0.l mm and the spacers 10, 20 fully cover at least each membrane in the membrane module 90. In this study, vibration frequency was fixed at 2 Hz and the amplitude of the spacer motion was fixed at 1.2 cm, which was the same as the vibration condition for CFD simulation. A constant permeate flux was achieved by regulating the speed of the peristaltic pump 220 (Masterflex, US). The suction pressure was monitored by a pressure transducer 210 connecting to a Data Acquisition module 240 (NI-DAQ; National Instruments). The measured values were logged in a computer 240 equipped with Labview data-logging software (National Instrument, US).

31) printed spacers

In this study, based on the spacer design (FIG. 23), four types of the spacers, namely Spacer 1, Spacer 3, Spacer 3S, and Spacer 3U with the same thickness of 0.3 mm were prepared (FIG. 25) using 3D printing technique. Solidworks design software (SOUIDWORKS Corp., US) was used to model these spacers based on the spacer design details (FIG. 23) and EOSINT P395 Selective Uaser Sintering (SUS) machine (EOS e -Manufacturing Solutions, Germany) was used to print the spacers. The spacer material was polyamide-l2 based powder PA2200 (EOS GmbH Electro Optical Systems, Germany). In the SLS machine, a carbon dioxide (C0₂) laser was used to sinter powdered materials via layer by layer of cross sections of the 3D spacer. After printing, the residual support materials were removed using an abrasive blasting machine. It is noted that for mass production of 3-D spacers, other materials and manufacturing technique (such as molding) can be used.

Filtration experiments

The feed solution 300 containing a mixture of 1 g of bentonite (Sigma-Aldrich, CAS No. 1302-78-9) and 1 g of sodium alginate (Sigma-Aldrich, CAS No. 9005-38-3) was used, which simulates particulate and macromolecular foulants respectively. The particulate size distribution of the feed was measured using the Malvern Mastersizer, shown in FIG. 26.

The filtration experiments were performed at permeate fluxes of 20, 40 or 60 L/m²h (LMH) for 1 h (the time taken to reach a quasi-steady state, considered as short-term tests (Li et ah, 2013; Pourbozorg et ah, 2017)) or a long-term period of time. For each test, the transmembrane pressure (TMP) of the clean membrane was recorded by filtering distilled water for 10 min. After that, the model foulant solution was filtered and the development of total TMP (TMP_t) with time was recorded. The TMP resulting from membrane fouling (TMP_f) was calculated by subtracting TMP_m from TMP,. For each condition, multi repeats of filtration experiments were performed and the averaged data were presented in the figures. To examine whether membrane integrity was compromised by vibrating spacers, at the end of filtration experiment, the turbidity of the permeate was measured using a turbidity meter (Hach, US).

Effect of spacer configuration and orientation

In this study, vibrating spacers were used to creating turbulence and shear stress along membrane surface to mitigate membrane fouling in a flat sheet membrane filtration system. Unlike typical rotating disk along the membrane where the movement of the disk occurs in one direction (Uee et ah, 1995), vibrating spacers resulted in reciprocating motion, which led to the ineffectiveness of using flat spacers to increase shear rate. This can be explained by the positive z-axis motion on the upward stroke of the flat spacer being nullified by the negative z-axis motion on the downward stroke of the flat spacer. In order to increase turbulence, shear rate, and foulant back transport near the membrane surface, the design of the spacer should focus on maximizing the conversion of the z-axis motion of the spacer to x- and y-axis fluid motion. Thus, the spacer with hill-like surface (Spacer 1) and the spacer with wave-like surface (Spacer 2 and Spacer 3) were designed (FIG. 23). The CFD simulation in terms of velocity and shear rate near the membrane surface during spacer vibration was performed in order to optimize the spacer design parameters. The simulation results are presented in FIGS. 27 and 28.

Although Spacer 2 and Spacer 3 had the same configuration (FIG. 23), they induced dissimilar velocity and shear rate profiles (FIGS. 27 and 28) when they moved in different patterns, i.e., groove and ridges being oriented parallel to movement or vibration direction for Spacer 2, and groove and ridges being oriented perpendicular to movement or vibration direction for Spacer 3. FIG. 27 presents the contour plots of the x-directional velocity (i.e., represents the trajectory away from the membrane) and shear rate near the membrane surface at 0.5s, which is the end of one cycle of the sinusoidal movement. All three spacers gave distinctly different x-directional velocity and shear rate profiles, which indicates that both 3D-spacer configuration and orientation may play important roles in dictating the extent of membrane fouling mitigation. It was observed that vibrating Spacer 3 (FIGS. 27(c) and (f)) could induce a significant improved x-axis velocity and shear rate along the membrane surface compared to vibrating Spacer 2 (FIGS. 27(b) and (e)). Especially, with vibrating Spacer 3, (1) the motion of the fluid near the peaks of the grooves and ridges appeared to be intensified as the valley longitudinal waves trapped and moved the pockets of fluid within the trough of the waves (FIG. 27(c)) and (2) evenly distributed shear force along the membrane could be obtained (FIG. 27(f)). Notably, Spacer 1, which can be considered to be a combination of both Spacer 2 and Spacer 3 configurations, while having shallower peaks and troughs, allowed the fluid to freely move along the z-axis with troughs aligned transverse to the z-axis, reducing the velocity and shear force in those regions (FIGS. 27(a) and (d)). FIG. 28 displays the quantitative evolution of x-directional velocity, y-directional velocity, z- directional velocity and shear rate, which reflects periodicity due to the oscillatory motion of the spacers. Compared to the induced turbulence in the y-direction (FIG. 28(b)), more turbulence occurred in the x-direction (FIG. 28(a)). In addition, vibrating Spacer 3 could induce 10.3-times higher turbulence along the x-direction and 4.4-times higher turbulence along the y-axis than vibrating Spacer 1. On the other hand, the averaged steady-state shear rate along membrane surface induced by vibrating Spacer 3 was 75% higher than that of vibrating Spacer 1. Furthermore, the z- directional velocity (FIG. 28(c) was most dominant in terms of magnitude, followed by the y- directional (FIG. 28(b)) and x-directional (FIG. 28(c)) velocity for all spacers. Also, the surface- averaged shear rate (FIG. 28(d)) was the greatest for Spacer 3, followed by Spacer 1 then Spacer 2. These indicate that both spacer configuration and orientation play crucial roles in determining liquid velocity and shear rate along the membrane surface during spacer vibration.

Furthermore, the total turbulence kinetic energy (TKE_T) values were calculated based on the CFD simulation results and presented in FIG. 48(a). The TKE_T produced by vibrating Spacers 1 and 2 were similar, while vibrating Spacer 3 produced a TKE_T more than twice that of Spacers 1 and 2. Notably, the mean surface-averaged shear rate in FIG. 48(b) produced by vibrating Spacer 3 was also twofold and threefold that respectively of Spacers 1 and 2, which consistently indicates the superiority of Spacer 3.

To further confirm the CFD simulation results, membrane filtration experiments using Spacer 1 and Spacer 3 were performed at a permeate flux of 20 L/m²h. As can be observed in FIG. 29, vibrating Spacer 3 could achieve better membrane performance than vibrating Spacer 1 during a 4.5-day filtration period of time. Overall, the CFD simulation and filtration experiments revealed that the configuration of Spacer 3 could benefit to enhance secondary flow on the membrane surface, more significantly alleviating membrane fouling.

Effect of perforations on the spacers

In a previous study (Pourbozorg et al., 2017) using perforated plates as turbulence promoter to mitigate membrane fouling, it was found that the kinetic turbulence energy (i.e., fouling reduction effectiveness) strongly depended on the plate solidity. To further increase shear force across the membrane and improve back transport of foulants, angled through holes were placed in Spacer 3 in order to effectively allow the feed to be pushed in and out of the spacers (Spacer 3S and 3L in Figure 2). These designs were intended to create jets of fluid flow across one period of the wave pattern for both the upward stroke and downward stroke of the vibrating spacer (FIG. 30). These jets of fluid flow would increase the membrane shear rate at the points where they strike the membrane and facilitate to transport foulants accumulated near the membrane out through the matching holes to the bulk feed.

CFD simulation in terms of x-axis and y-axis velocity as well as shear rate was performed at a permeate flux of 20 L/m²h, with the spacers vibrated in a direction parallel to the membrane at a frequency of 2 Hz, and amplitude of 1.2 cm and the simulation results are presented in FIGS. 31 and 32. The images shown in FIG. 31 indicated that the addition of the holes did not significantly improve both shear rate near the membrane surface and fluid movement away from the membrane. Further analysis of turbulence kinetic energy revealed that the presence of greater-sized holes in the spacer (Spacer 3L) could slightly improve the turbulence energy levels along the x-axis (TKE_x=8.53 l x l0 ⁵m²/s²) and y-axis (TKE_y=l .835 x l0 ⁶m²/s²) than the spacer without holes (TKE_x=8.27x l0 ⁵m²/s² and TKE_y= 1.535 x l0 ⁶m²/s²). With adding smaller-sized holes in the spacer (Spacer 3S), the turbulence energy along the x-axis (TKE_c=3.184^c l0 ⁵m²/s²) was decreased and that along y-axis (TKE_y=4.0l2x l0 ⁶m²/s²) was improved compared to the spacer without holes (Spacer 3, TKE_x=8.27x l0 ⁵m²/s² and TKE_y=l .535 x l0 ⁶m²/s²), achieving slightly lower overall all turbulence energy than that with Spacer 3.

The membrane filtration experiments using Spacer 3, Spacer 3S, and Spacer 3L were conducted at a permeate flux of 20 L/m²h. It was observed in FIG. 33 that after a dramatically increase of TMP, almost constant TMP development was noticed when these three types of spacers were vibrated at the same vibrating conditions (frequency of 2 Hz and amplitude of 1.2 cm). As a result, there was almost no obvious difference of the TMP profiles. Possibly, at a relatively lower permeate flux of 20 L/m²h, the shear force induced by vibrating these three spacers could conquer the driving force causing foulant disposition. Therefore, such mere difference of turbulence energy and shear rate along the membrane surface could not lead to significant improvement of membrane performance.

To further compare the fouling control effectiveness of Spacer 3, Spacer 3S, and Spacer 3 L, a higher permeate flux at 40 L/m²h was employed for membrane filtration tests. The filtration results in FIG. 34 show vibrating Spacer 3S (i.e., with smaller holes) led to improved membrane fouling control efficiency compared to vibrating Spacer 3 and Spacer 3L (i.e., with greater-sized holes).

FIG. 49 further quantifies the evolution of the x-directional velocity, y-directional velocity, z- directional velocity, surface -averaged shear rate and local shear rate near the perforations at the membrane surface. In particular, compared to Spacer 3, Spacers 3S and 3L conferred a slight decrease of the surface-averaged shear rate across the membrane surface (FIG. 49(d)), but a slight increase in local shear rate near the perforations because of the jets of fluid flow created (Fig. lOe). This indicates that the perforations caused the local shear rate near the perforations to increase, but diminished the overall surface-averaged shear rate. This is due to the fluid flow through the perforations increasing z- directional velocity (FIG. 49(c)) but reducing x-directional velocity (FIG. 49(a)), which means less of the vibrating motion is being converted to shear force as fluid is allowed to escape from the perforations. This indicates the possibility of increasing perforation density for improved fouling mitigation. Further analysis of the turbulence kinetic energy (TKE) in FIG. 50 reveals that, vis-a-vis the spacer without perforations (i.e., Spacer 3), the presence of smaller (Spacer 3S) and larger (Spacer 3L) perforations increased TKE_T respectively by 40.3% and 6l.l%.Furthermore, similar to the observations made in Fig. 10, the mean surface-averaged shear rates of the perforated spacers were lesser than that of the non-perforated spacer (FIG. 50(b)), though the spacer with the larger perforations (i.e., Spacer 3L) exhibited a larger shear near the perforations (FIG. 50(c)). These observations are similar to previous research using oscillating turbulence generator in a liquid filled tank, whereby the shear resulting from jets and wakes created by the perforations cancelled out each other (Pourbozorg et ah, 2016, 2017; Cheng et ah, 2001; Silva et. al, 1994). In other words, the perforated spacer became more similar to the turbulence promoter described in Pourbozorg et al. (Pourbozorg et al., 2016, 2017). However, these conflicting results obtained from the simulation makes it difficult to determine if the spacers with perforations were superior to that without perforation at a permeate flux of 20 L/m²h.

The filtration experiments were carried out using Spacers 3, 3S and 3L at a permeate flux of 20 L/m²h. FIG. 51 presents the fouling rates (i.e., dTMP_f/dt) of the spacers with and without perforations. The fouling rates were similar for all three spacers, which shows that the perforations did not improve membrane fouling mitigation at the permeate flux of 20 L/m²h. This may in part be due to the already high effectiveness of Spacer 3 in mitigating membrane fouling (FIG. 48), and the enhancement in turbulence and local shear being negated by the decreased surface-averaged shear (FIG. 50). In view of the previous studies that such perforations are superior (Pourbozorg et al., 2017), higher permeate fluxes that cause more drastic fouling were further investigated.

The CFD simulation was performed at higher permeate fluxes of 40 and 60 L/m²h to assess the effectiveness of the perforations on Spacer 3. FIG. 50 reveals that more differences among the three spacers at higher permeate fluxes. With increasing permeate flux, TKE_T decreased for both Spacer 3 and 3S, and surface-averaged shear rate and local shear rate decreased for Spacer 3 but increased for Spacer 3S. The change in permeate flux negligibly affected these parameters for Spacer 3L. Therefore, Spacer 3S exhibited potential to out-perform the other two spacers at even higher permeate fluxes, because it is the only spacer that gave a positive correlation between shear rates (i.e., both overall and local) and permeate flux (FIG. 50(b) and (c)).

The membrane filtration experiments were also carried out for the three spacers at the higher permeate fluxes of 40 and 60 L/m²h. It was observed, at the higher fluxes, Spacer 3S gave the lowest fouling rates (FIG. 52 (a) and (b)). This was further affirmed by statistical analysis at a confidence level of 90%, which indicates that Spacer 3S exhibited a better performance than Spacer 3 at the two higher fluxes. This shows that the perforations did indeed help in fouling mitigation. However, along with the higher TKE_T and mean local shear rate comes the lower mean surface averaged shear rate, which results from reciprocating movements and possibility of introducing more foulant through the perforations directly to the membrane. Hence, the number of perforations per unit area and perforation size has to be further optimized when designing such vibrating spacers. It should be noted that a discrepancy appears to exist between simulations and experiments in that, while the simulations indicate that surface-averaged mean local shear rates (FIG. 50(c)) and overall TKE magnitudes (FIG. 50(a)) were higher for Spacer 3L relative to Spacer 3S, the experiments indicate instead that the fouling rates for Spacer 3S were lower than Spacer 3L (FIG. 52). However, agreement between the simulations and experiments exists with respect to the increasing efficacy of Spacer 3S to mitigate membrane fouling at higher permeate fluxes. The discrepancy can be attributed to the lack of consideration of the colloidal foulants in the feed in the simulations, which means that the role of the perforations in the movement of the particulates that inevitably affects the fouling extent was not accounted for.

Furthermore, the possibility of introducing more foulant through the perforations directly to the membrane exists, which explains the comparatively higher fouling rate observed for Spacer 3L that had larger perforations. Nonetheless, it is clear that perforations can be included and optimized to further improve the effectiveness of fouling mitigation by 3D-spacers.

Conclusions

Both simulation and experimental results revealed that the wave-like Spacer 3 (groove and ridge direction perpendicular to spacer movement direction) could alleviate more membrane fouling than the other designed spacers (e.g., hill-like Spacer 1, wave-like Spacer 2 with groove and ridge direction parallel to spacer movement direction), regardless of tested permeate fluxes. Furthermore, two types of wave-like spacers (Spacers 3S and 3L) were designed based on CFD simulation, in which perforations were placed on the spacer in order to enhance shear rate and provide a path for the back transport of foulants into the bulk feed. The membrane filtration tests showed that the presence of either large holes (2 mm diameter) or small holes (1 mm diameter) did not significantly reduce membrane fouling at a lower permeate flux (20 L/m²h), but the presence of holes significantly improved membrane performance at a higher permeate flux (40 L/m²h or 60 L/m²h).

In this study, both simulation and experimental results clearly revealed that the spacer morphological design could have a significant impact on the fouling mitigation effectiveness of the vibrating spacer. The following conclusions can be drawn:

(1) The simulation results show that the spacer groove and ridge orientation affected the shear rate and back transport near the membrane surface.

(2) Vibrating the wave-like spacer (groove and ridge orientation perpendicular to movement or vibration direction) induced the higher TKE and shear rate at the membrane surface than vibrating hill -like spacer. The filtration results were in an agreement with the simulation findings.

(3) The perforations on the spacer (with smaller-sized holes) provided slight improvements in shear rate and back transport. Perforations on the 3D-spacers were not effective at lower permeate fluxes. This is because the non-perfbrated spacer performed well as-is, so the addition of perforations did not improve the performance further at the lower fouling extent associated with the low permeate flux. Thirdly, at the higher permeate fluxes, different sizes of perforations exhibited different performances, which indicates the importance of optimizing the perforations in terms of perforations per unit area and perforation size. Specifically, the smaller perforations performed the best and were found to be increasingly effective at higher permeate fluxes, while the larger perforations performed slightly better compared to that without perforations at some tested fluxes (e.g., flux of 40 L/m² h).

Third Experimental Study

In a third experimental study, the effect of a spacer vibration system which comprised using a motor 31 to vibrate spacers 10, 20 provided in a sandwich configuration on each side of a membrane module 90 of hollow fibre membranes (FIG. 35) using Spacer 1 (FIG. 23(a) and Spacer 2 (FIG. 23(b)) was compared with a fibre vibration system which comprised using a motor 31 to vibrate the membrane module 90 comprising hollow fibre membranes without any spacers provided (FIG. 36).

The simulated wastewater was a mixture of inorganic particulate (bentonite) and organic macromolecular (sodium alginate). The hollow fiber membranes were commercially available PVDF membranes from Tritech (OD = 2 mm, China) with nominal pore size of 0.1 pm that targeted for wastewater treatment. To ensure that the spacers 10, 20 did not contact the membrane module 90, especially for the fibers having higher looseness values, the spacers 10, 20 were each located 5 mm or 10 mm away from the membrane module 90.

It can be seen from FIG. 37 that the presence of vibrating spacer 2 led to better membrane performance. Compared to the hollow fibre membranes without any looseness, the membranes with 5% of looseness could achieve better membrane performance with vibrating spacers. The spacers were positioned approximately 1 cm from the hollow fibre.

FIG. 38 shows that under the same vibrating frequency condition, vibrating membranes could mitigate more membrane fouling compared to vibrating spacers. Similarly, in order to achieve comparable membrane fouling control effectiveness, vibrating membranes could be performed at a relatively lower frequency compared to vibrating spacers. However, as shown in FIG. 39, vibrating membranes appear to consume more energy compared to vibrating spacers at the similar membrane fouling rates. This is mainly due to greater mass/weight of membrane modules than the spacers.

Increasing the frequency had a greater effect on the dTMP/dt of the spacer vibration system than the membrane vibration, as evident in the steeper slope of dTMP/dt versus vibration frequency seen in FIG. 40(b) compared to FIG. 40(a).

FIGS. 41 and 42 showed that the hollow fibers with horizontal position (i.e., orthogonally arranged with the vibrating direction) were more beneficial in membrane fouling mitigation for both vibration systems (i.e., spacer vibration and membrane vibration).

FIG. 43 shows that the effect of spacer vibration frequency and fiber looseness on the fiber movement velocity is negligible, which signifies that the intensive turbulence and shear stress on fiber surface held a key for the fouling mitigation efficiency.

The effect of hollow fiber looseness on dTMP/dt is insignificant at higher frequencies. However, a significant decrease trend was found at lower frequencies, particularly for the spacer vibration system at 1 Hz, as shown in FIG. 44.

FIGS. 45 and 46 show a comparative analysis of the vibration systems of the fouling mitigation. There appeared to be a sharp decrease trend from the correlation between dTMP/dt versus power requirement ( ) for the spacer vibration system, whereas the relationship between the dTMP/dt and P tended to plateau for the membrane vibration system. At the same P value, the dTMP/dt is similar for both vibration systems.

Collectively, the spacer vibration system was more cost-effective than the membrane vibration with respect to its lower energy consumption, and was able to achieve competitive fouling mitigation efficiency at higher frequencies (i.e., above 2 Hz). The generated turbulence and unsteady-state shear were more important than the fiber movement in mitigating the membrane fouling, which underscores the feasibility of spacer vibration enabling the fouling mitigation at lower energy consumption. This beneficial effect would be more advantageous with closer distance of the spacers to the membrane module.

Whilst there has been described in the foregoing description exemplary embodiments of the present invention, it will be understood by those skilled in the technology concerned that many variations in details of design, construction and/or operation may be made without departing from the present invention. For example, spacers having other configurations besides those described and shown in the figures may be provided, such as those having a triangular wave-form cross-section, rectangular wave-form cross section, dove-tail wave-form cross-section and so on.

Appendix

Governing equations and boundary conditions

The eontmuhy and Navier-Stokes equations for a 3D steady-state incompressible flow used in the laminar flow stationary study are as follows:

where ¾., li_gand n_s are the velocity components m v-. y- and x-direcikms, p is the pressure, m an /3 are respectively the dynamic viscosity and density of the fluid. REFERENCES

Al Akoum, O., Jaffrin, M. Y., Ding, L. H., Paullier, P., Vanhoute, C., (2002). An hydrodynamic

investigation of microfiltration and ultrafiltration in a vibrating membrane module, J.

Membrane Sci., 197 (1-2), pp. 37-52.

Aslam, M., Charfi, A., Lesage, G., Heran, M., Kim, J., (2017). Membrane bioreactors for wastewater treatment: A review of mechanical cleaning by scouring agents to control membrane fouling, Chem. Eng. J., 307, pp. 897-913.

Bilad, M. R., Mezohegyi, G., Declerck, P., Vankelecom, I. F. J., (2012). Novel magnetically induced membrane vibration (MMV) for fouling control in membrane bioreactors, Water Res, 46 (1), pp. 63-72.

Cheng, N. S., Law, A. W. K., (2001). Measurements of turbulence generated by oscillating grid, J.

Hydraul. Eng., 127, pp. 201-208.

Cui, Z. F., Chang, S., Fane, A. G., (2003). The use of gas bubbling to enhance membrane processes, J.

Membrane Sci., 221 (1-2), pp. 1-35.

Drews, A., (2010). Membrane fouling in membrane bioreactors-Characterisation, contradictions, cause and cures, J. Membrane Sci., 363 (1-2), pp. 1-28.

Fane, A. G., Fell, C. J. D., (1987). A Review of Fouling and Fouling Control in Ultrafiltration,

Desalination, 62, pp. 117-136.

Genkin, G., Waite, T. D., Fane, A. G., Chang, S., (2006). The effect of vibration and coagulant

addition on the filtration performance of submerged hollow fibre membranes, J. Membrane Sci., 281 (1-2), pp. 726-734.

Hopfmger,E.J., Toly, J.A., (1976). Spatially Decaying Turbulence and Its Relation to Mixing across Density Interfaces, J. Fluid Mech, 78, pp. l55-&.

Jaffrin, M. Y., (2008). Dynamic shear-enhanced membrane filtration: A review of rotating disks, rotating membranes and vibrating systems, J. Membrane Sci., 324 (1-2), pp. 7-25.

Judd, S., 2011. The MBR book (second edition): Principles and applications of membrane bioreactors for water and wastewater treatment Elsevier.

Judd, S., Judd, C. (Eds.). (2006). The MBR book : principles and applications of membrane

bioreactors in water and wastewater treatment. Amsterdam ; Boston ; London: Elsevier.

Kim, J., Kim, K., Ye, H., Lee, E., Shin, C., McCarty, P. L., Bae, J., (2011). Anaerobic Fluidized Bed Membrane Bioreactor for Wastewater Treatment, Environ. Sci. Technol, 45 (2), pp. 576- 581. 28

Kola, A., Ye, Y., Le-Clech, P., Chen, V., (2014). Transverse vibration as novel membrane fouling mitigation strategy in anaerobic membrane bioreactor applications, J. Membrane Sci., 455, pp. 320-329.

Kola, A., Ye, Y., Ho, A., Le-Clech, P., Chen, V. K., (2012). Application of low frequency transverse vibration on fouling limitation in submerged hollow fibre membranes, J. Membrane Sci. ,

409, pp. 54-65.

Krzeminskia, P., Leveretteb, L., Malamisc, S., Katsoud E., (2017). Membrane bioreactors - A review on recent developments in energy reduction, fouling control, novel configurations, LCA and market prospects, J. Membrane Sci., 527, pp. 207-227.

Lee, S.S., Burt, A., Russotti, G., Buckland, B., (1995). Microfiltration of Recombinant Yeast-Cells Using a Rotating-Disk Dynamic Filtration System, Biotechnology and Bioengineering, 48, pp. 386-400.

Li, T., Law, A. W. K., Fane, A. G., (2014). Submerged hollow fibre membrane filtration with

transverse and longitudinal vibrations, J. Membrane Sci., 455, pp. 83-91.

Li, T., Law, A. W. K., Cetin, M., Fane, A. G., (2013). Fouling control of submerged hollow fibre

membranes by vibrations, J. Membrane Sci., 427, pp. 230-239.

Maere, T., Verrecht, B., Moerenhout, S., Judd, S., Nopens, I., (2011). BSM-MBR: A benchmark

simulation model to compare control and operational strategies for membrane bioreactors, Water Res., 45 (6), pp. 2181-2190.

Pourbozorg, M., Li, T., Law, A. W. K., (2016). Effect of turbulence on fouling control of submerged hollow fibre membrane filtration, Water Res., 99, pp. 101-111.

Pourbozorg, M., Li, T., Law, A. W. K., (2017). Fouling of submerged hollow fiber membrane

filtration in turbulence: Statistical dependence and cost-benefit analysis, J. Membrane Sci., 521, pp. 43-52.

Serra, C. A., Wiesner, M. R., (2000). A comparison of rotating and stationary membrane disk filters using computational fluid dynamics, Journal of Membrane Science, 165 (1), pp. 19-29.

Silva, I.P.D.D., Fernando, H.J.S., (1994). Oscillating grids as a source of nearly isotropic turbulence, Phys. Fluids, 6, pp. 2455-2464.

Tan, W. S., Suwamo, S. R., An, J., Chua, C. K., Fane, A. G., Chong, T. H., (2017). Comparison of solid, liquid and powder forms of 3D printing techniques in membrane spacer fabrication, J. Membrane Sci., 537, pp. 283-296.

Verrecht, B., Maere, T., Nopens, I., Brepols, C., Judd, S., (2010). The cost of a large-scale hollow

fibre MBR, Water Res., 44 (18), pp. 5274-5283.

Wang, J. W., Wu, B., Liu, Y., Fane, A. G., Chew, J. W., (2017). Effect of fluidized granular activated carbon (GAC) on critical flux in the microfiltration of particulate foulants, J. Membrane Sci. , 523, pp. 409-417.

Wang, J. W., Zamani, F., Cahyadi, A., Toh, J. Y., Yang, S. L., Wu, B., Liu, Y., Fane, A. G., Chew, J.

W., (2016). Correlating the hydrodynamics of fluidized granular activated carbon (GAC) with membrane -fouling mitigation, J. Membrane Sci., 510, pp. 38-49.

Wang, Z.W., Ma, J.X., Tang, C.Y.Y., Kimura, K., Wang, Q.Y., Han, X.M., (2014). Membrane

cleaning in membrane bioreactors: A review, J. Membrane Sci., 468, pp. 276-307. Wibisono, Y., Comelissen, E. R., Kemperman, A. J. B., van der Meer, W. G. J., Nijmeijer, K., (2014).

Two-phase flow in membrane processes: A technology with a future, J. Membrane Sci., 453, pp. 566-602.

Wu, B., Wong, P. C. Y., Fane, A. G., (2015). The potential roles of granular activated carbon in

anaerobic fluidized membrane bioreactors: effect on membrane fouling and membrane integrity, Desalin. Water Treat., 53 (6), pp. 1450-1459.

Wu, B., Wang, Y. Y., Lim, W., Chew, J. W., Fane, A. G., Fiu, Y., (2016). Enhanced performance of submerged hollow fibre microfiltration by fluidized granular activated carbon, J. Membrane Sci. , 499, pp. 47-55.

Wu, B., Zamani, F., Fim, W. K., Fiao, D. X., Wang, Y. Y., Fiu, Y., Chew, J. W., Fane, A. G., (2017).

Effect of mechanical scouring by granular activated carbon (GAC) on membrane fouling mitigation, Desalination, 403, pp. 80-87.

Yeo, A. P. S., Faw, A. W. K., Fane, A. G., (2007). The relationship between performance of

submerged hollow fibers and bubble -induced phenomena examined by particle image velocimetry, Journal of Membrane Science, 304 (1-2), pp. 125-137.

Claims

1. A membrane filtration system comprising:

at least one membrane module to perform filtration of a feed fluid in which the at least one membrane module is submerged during use of the membrane filtration system;

at least one spacer provided next to and spaced apart from the at least one membrane module in parallel with the at least one membrane module; and

a vibration mechanism provided to vibrate only the at least one spacer to create a scouring effect on the at least one membrane module during use of the membrane filtration system, wherein vibration of the at least one spacer is in only directions parallel to the at least one membrane module.

2. The membrane filtration system of claim 1, wherein the at least one spacer is spaced apart from the at least one membrane module by a distance ranging from 0.1 mm to 10 mm.

3. The membrane filtration system of claim 1 or claim 2, wherein the at least one spacer has a non- continuously-flat surface facing the at least one membrane module.

4. The membrane filtration system of claim 3, wherein the non-continuously-flat surface comprises one of: a surface having grooves thereon, and a surface having a plurality of protrusions thereon.

5. The membrane filtration system of claim 4, wherein the grooves have a width of 0.5 mm, a depth of 1 mm, and adjacent grooves are 3 mm apart.

6. The membrane filtration system of claim 4, wherein the plurality of protrusions have a wave-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis, the z axis being parallel to the at least one membrane module.

7. The membrane filtration system of claim 6, wherein the wave-like pattern comprises a series of parallel and alternating grooves and ridges oriented perpendicularly to a direction of vibration of the at least one spacer.

8. The membrane filtration system of claim 4, wherein the plurality of protrusions have a hill-like pattern comprising a plurality of adjacent sinusoidal surfaces thereon along a first (z) axis and a plurality of adjacent sinusoidal surfaces thereon along a second (y) axis, the y axis being orthogonal to the z axis, the y axis and the z axis being parallel to the at least one membrane module.

9. The membrane filtration system of any one of claims 6 to 8, wherein each of the plurality of protrusions has a protrusion height ranging from 0.5 mm to 2.35 mm and peak-to-peak distance between adjacent protrusions ranges from 3mm to 7mm.

10. The membrane filtration system of any one of claims 6 to 9, wherein the plurality of protrusions have the form of an accordion fold.

11. The membrane filtration system of any one of claims 6 to 10, wherein the at least one spacer has a material thickness of 0.3 mm.

12. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer has an overall cross-sectional height ranging from 1.3 mm to 5 mm.

13. The membrane filtration system of any one of the preceding claims, wherein the vibration mechanism provides vibration to the at least one spacer at a frequency ranging from 1 Hz to 5 Hz and an amplitude ranging from 2 mm to 20 mm.

14. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer is longer and wider than each membrane in the at least one membrane module by at least an amplitude of vibration of the at least one spacer respectively.

15. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer has a surface roughness ranging from 2.9 nm to 26.3 nm.

16. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer has a plurality of through holes.

17. The membrane filtration system of any one of the preceding claims, wherein the at least one membrane module comprises at least one flat sheet membrane.

18. The membrane filtration system of any one of the preceding claims, wherein the at least one membrane module comprises a plurality of hollow fibre membranes.

19. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer comprises a first spacer provided on a first side of the at least one membrane module and a second spacer provided on a second side of the at least one membrane module.

20. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer comprises multiple spacers and the at least one membrane module comprises multiple membrane modules, wherein the spacers and membrane modules are provided in an alternating arrangement.

21. The membrane filtration system of claim 20, wherein the number of spacers is greater than the number of membrane modules by one.

22. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer comprises at least a rigid frame.

23. The membrane filtration system of any one of the preceding claims, wherein the at least one spacer comprises a rigid sheet.

24. A method of membrane filtration comprising the steps of:

(a) submerging at least one membrane module configured to perform filtration of a feed fluid in the feed fluid;

(b) vibrating only at least one spacer in only directions parallel to the at least one membrane module during filtration, the at least one spacer provided next to and spaced apart from the at least one membrane module in parallel with the at least one membrane module.

25. The method of claim 24, wherein the vibration is at a frequency ranging from 1 Hz to 5 Hz and an amplitude ranging from 2 mm to 2 cm.

26. The method of claim 24 or 25, wherein the vibration is intermittent at cycles comprising up to 3 minutes of continuous vibration followed by up to 1 minute without vibration.

27. The method of any one of claims 24 to 26, wherein the vibration comprises at least one of: linear movement of the at least one spacer and non-linear movement of the at least one spacer.