Impurities and colloidal substances are two of many fouling conditions that reduce the membrane filtration performance used in wastewater treatment. This study investigates the potential of fluidic-oscillation-generated microbubbles (MBs) to defoul the filtration membrane. Cartridge filters for microfiltration (MF) of 1 μm pore size were fouled using surface seawater collected from the Hull coastal area. The seawater was circulated at 5.8 L/min to actuate colloidal substance deposition on the membrane surface. The recorded feed channel pressure drop (ΔP) across the membrane filters showed rapid fouling occurred in the first 8 hrs of the circulation. Fluctuations of ΔP during the next 8 hrs were observed showing the colloids filling the pores of the membrane, and remaining steady for 2 hrs showing the membrane was completely fouled. The filtration membrane was cleaned and defouled using fluidic-oscillator-generated MBs. The fouled membranes were sparged with 1 L/min of air scouring for ∼1 to ∼2 hrs to remove the deposited colloids and impurities on the surface of the membrane. The membrane, analysed by Scanning Electron Microscopy (SEM), UV254 and Electrical Conductivity (EC) meter, showed the extent of MBs-mediated removal of the deposited colloidal particle from the membrane surfaces. This study found that the highest defouling rate occurs with MBs generated by fluidic oscillator (closed vent), followed by MBs generated by fluidic oscillator (opened vent) and MBs generated without fluidic oscillator at 9.53, 6.22, and 3.41 mbar/min, respectively.
Membrane filtration approaches such as microfiltration (MF), ultrafiltration (UF), nanofiltration (NF) and reverse osmosis (RO) are very important for wastewater treatment and waste recovery (Bhattacharya et al. 2013; Galanakis et al. 2013; Giacobbo et al. 2015). They provide many advantages such as high selectivity, capacity and feasibility. However, they are easily fouled by biofoulings, and organic and colloidal substances which restrict the permeation rate and reduce the process efficiency. In general, fouling is usually caused by the deposition of small colloidal particles on the membrane surface and inner walls of membrane pores which results in the formation of a cake layer (Chesters et al. 2013a, 2013b; Zhao & Yu 2014). Conventional defouling methods such as chemical cleaning and pre-treatment usually are destructive and cause waste problems. Recently, innovative studies were conducted to explore the potential of MBs to clean the filtration membrane. Mechanisms such as creating MB pulsating-like action (Wilson et al. 2013), the behaviour of MBs by adsorption (Akuzawa et al. 2010), and swarm velocity (Lee & Lee 2002) clearly described the role of MBs in cleaning applications. Agarwal et al. (2012) and Wibisono (2014) listed four steps of cleaning using MB: (1) generation of smaller MBs; (2) MBs burst to generate high-pressure spot and shear force; (3) continuous biofilm matrix disruption; and (4) biofilm detachment. Based on the cleaning mechanisms mentioned, it is important to generate smaller microbubble high-pressure spots, however, it is unlikely that small MBs will be generated by only depending on the various sizes of pores, shear and materials of the diffuser system. Zimmerman et al. (2009) generate a smaller size of MBs from the diffuser pore using fluidic oscillation by oscillating the feed air stream by pinching off the bubbles, known as a hemispherical cap. Thus, a fluidic oscillation microbubble is generated to assist and compare with the conventional bubble cleaning method to restore membrane performance. MF membranes used as pre-treatment for desalination usually have shorter filter lifetimes due to fouling (Baker 2004). This research has been mainly to exploit the advantage of using a cheaper way of producing smaller microbubbles in cleaning MF membranes (Zimmerman et al. 2011). Using MB cleaning, the performance of the filtration membrane has been developed positively by prolonging the membrane life and alleviating energy consumption (Fazel et al. 2013; Wibisono et al. 2015). Environmentally, this research will significantly bring the food and chemical industries towards green waste management by reducing waste production and replacement of chemical cleaning agents (Mercier-Bonin et al. 2004; Chesters et al. 2013a, 2013b).
Membrane fouling and defouling
Fouling is usually caused by the deposition of small colloidal particles on the inner walls of membrane pores. The blockages are a build-up of particles in the form of a cake layer on the membrane surface and membrane pore opening. The effect of permeation flux reduction due to fouling is twofold. First, pore blocking and cake formation lead to an increase in flow resistance. After that, the presence of colloidal particles deposited on the membrane surface hampers liquid mixing. Thus, a relatively high concentration of solutes persists near the membrane surface which causes a reduction of the solvent flux crossing the membrane (Henry et al. 2012).
P: Gas pressure
PL: Liquid pressure
σ: Surface tension of the liquid
db: Bubble diameter
Fluidic oscillator microbubble defouling
The mechanisms of using MBs to defoul filtration membranes exemplified in Figures 1 and 2 show great potential in utilising MBs for controlling and preventing membrane fouling. Lu et al. (2008) concluded that higher gas flowrate and smaller bubble sparging limit fouling better on hollow fibre MF membranes. MBs applied to MF membranes successfully enhance membrane performance by reducing transmembrane pressure (TMP) more effectively, enhance the critical flux, and induce lighter cake formation (Yu et al. 2003; Lu et al. 2008; Hwang & Wu 2009). Lee et al. (2014) found that MBs are able to adhere to particulates and colloidal matter, thus causing them to float, disrupt the gel layer, and provide pyrolytic decomposition of protein.
Zimmerman & Tesar (2010) patented a method of producing smaller microbubbles using a fluidic oscillator. This device acts as an amplifier by oscillating the gas passing through. Zimmerman (2014) listed several applications of using smaller bubbles generated by a fluidic oscillator to strip components of liquid such as gas transfer in bioreactors, anaerobic digesters, and particle separation. Various applications of fluidic-oscillator-generated microbubbles have been studied: better oil emulsion separation (Hanotu et al. 2013), higher separation efficiency via microbubble distillation (Al-yaqoobi et al. 2016), better algal growth (Kamaroddin et al. 2016), and efficient yeast recovery (Hanotu et al. 2014).
In this paper, the study of defouling will be conducted using MBs generated by fluidic oscillation. A fluidic oscillator connected to the diffuser as shown in Figure 3 is able to produce a smaller bubble size (Hanotu et al. 2013). Instead of relying on the structure of a porous material for the nozzles to generate smaller bubbles, fluidic oscillations divert the jet overcoming the Coanda effect to enable the pinching-off of the hemispherical cap of bubble formation, resulting in nearly mono-dispersed, uniformly released microbubbles (Zimmerman et al. 2008). This device has no moving parts and is able to produce smaller microbubbles at higher energy efficiency (Tesař 2014a).
MATERIALS AND METHODS
Experimental design and setup
In this study, two main phases of experiment were conducted:
Membrane fouling by circulating the seawater – mainly increase in pressure drop.
Microbubble sparging for membrane defouling – optimised membrane performance.
Microfiltration membrane defouling
The filtration system was developed to remove colloidal substances from surface seawater and circulated at 5.4–5.8 L/min. The filtration housing is fitted to a 10 inch tubular MF membrane as shown in Table 1. The pressure drop of the filtration system is recorded every 1 minute. A flowmeter and pressure transducers (P1 and P2) are connected and recorded using an Arduino Data Logger. For the experimental start-up, the tubular unit is circulated with tap water for 24 hr to allow the soaking process and pressure to balance before being fed with seawater. Five millilitres of the effluent sample is collected at 1 minute intervals from the beginning and with every imposed pressure-drop level. One tubular unit runs for 3 days. Once the pressure drop is constantly above 1.4 bar, the system is sparged with microbubbles from the air-scouring unit. The air-scouring unit consists of a control box (pressure regulator, valve, pressure gauge) connected to the fluidic oscillator and diffuser. The air flow rate injected with the feed at approximately 1 L/min using the alumina diffuser produces 100–1,000 μm size microbubbles to defoul the filtration membrane. The fluidic oscillator is operated with a feedback loop length of 50 mm. MBs sparge the membrane and membrane samples are analysed by Scanning Electron Microscopy (SEM).
|Type||Sediment cartridge (guard) filter|
|Micron rating||1 micron|
|Cartridge dimension||ID: 30 mm; OD: 65 mm; L: 255 mm|
|Membrane system setup||Cross-flow|
|Temperature||Room ∼ 22.7–25.1(°C)|
|Pressure initiation||2 bar|
|Type||Sediment cartridge (guard) filter|
|Micron rating||1 micron|
|Cartridge dimension||ID: 30 mm; OD: 65 mm; L: 255 mm|
|Membrane system setup||Cross-flow|
|Temperature||Room ∼ 22.7–25.1(°C)|
|Pressure initiation||2 bar|
Main membrane characteristics, pore sizes, and composition
Table 1 presents information on the main membrane characteristics. A microfiltration membrane (MF) of 1 micron pore size was operated using crossflow configuration at room temperature during summer time. Seawater fed to the membrane was filtered specifically for the impurities and colloids contained.
Seawater and membrane sources
Seawater was collected from the East Riding of Yorkshire, England, at Spurn and stored at room temperature (21 ± 4 °C) prior to all tests. UV254 and pH of the seawater were at 0.034 cm−1 and 8.0 respectively.
For microbubble generation, the fluidic oscillator is connected to the air sources as shown in Figure 4. There are many advantages to using the fluidic oscillator in terms of cost-effectiveness, robustness, reliability, immobile parts and no requirement of electricity (Zimmerman et al. 2011). The bubbles generated by fluidic oscillation have a low energy consumption that distinguishes the method from other methods such as ultrasonic and rotary disk which require a significant supply energy (Abdulrazzaq et al. 2015). Zimmerman et al. (2008) explained this device acting as a fluidic amplifier with the potential to pinch off the hemispherical cap bubble. This early break-off of bubble formation at the diffuser aperture offers the smallest possible bubble size. Figure 2 and Equation (1) illustrate the relationship of the smaller microbubble having a higher surface-area-to-volume ratio which leads to higher momentum transfer rates, especially for scrubbing the surface of the membrane (Agarwal et al. 2012; Abdulrazzaq et al. 2015). A fluidic oscillator mainly consists of three parts: one inlet for air supply, two mid-ports for the feedback loop, and two exit ports as the oscillation channel outlet. The arrangement as shown in Figure 4 oscillates the gas flow between two paths under constant pressure of gas (Zimmerman & Tesar 2010). A remarkable feature of this system is that the frequency of the oscillation is adjustable by manipulating the air flow rate and the length of the feedback loop (Tesař 2014b). Figure 4 shows the effect of microbubble size generated (a) with and (b) without the fluidic oscillator.
MB operating conditions
MBs were generated using the scouring unit, which was connected through the alumina diffuser at the bottom of the filtration housing. The air was injected through the diffuser at flow rate and pressure of 1 L/min and 2.2 bar respectively. The bubble size generated was in the range of 100–1,000 micron. The following MB conditions were generated:
1 L/min of flow with slightly open vent valve
1 L/min of flow with fully closed vent valve
Non-fluidic-oscillator-generated microbubbles/steady flow sparging
Data collection and measurement
Arduino pressure transducer and flowmeter
Two pressure transducers were installed at the inlet, P1, and outlet, P2, to measure the pressure drop while the flowmeter was connected after the circulation pump. For analogue reading both of these instruments were connected to an Arduino Uno Data Logger. The data were collected at 1 minute intervals using a PLX-DAQ Excel sheet.
Continuous monitoring of the pH value, Total Dissolved Solids (TDS), Electrical Conductivity (EC) and temperature of the feed were inferred using the Continuous Monitor Hydroponics trimeter. The sensors were placed in the feed tank. The nutrition controller collected the pH value, TDS and temperature of the system. The sample was collected and analysed using a UV/Vis spectrophotometer.
UV absorbance and SEM
The UV absorbance of the water was measured at 254 nm using the UV/Vis spectrophotometer (Jenway 6705). At the beginning of the experiment, during microbubble sparging, and if there were fluctuations of P2, UV absorbance was tested at intervals of 1 minute for 20 minutes. With steady pressure drop, the absorbance was measured in intervals of 30 mins to 1 hr. The surface of the membranes after the experiment was dried at 50 °C for one night and coated using gold. The gold-coated membrane surface was examined for colloidal deposition and removal under the Scanning Electron Microscope.
RESULTS AND DISCUSSION
Effect of microbubbles on fluid properties
Figure 5 shows the EC value collected from the trimeter nutrition controller. Both of the values decreasing over time indicate that the membrane was fouled. The rapid decrement of the absorbance and EC show the dissolved solids were deposited on the surface of the membrane during the first 500 minutes, which is roughly after 8 hrs of circulation. The values remain constant for ∼2 hrs showing the membrane filtration efficiency has dropped because its ability to filter more particles is now limited. This finding is in agreement with the study conducted by Gwenaelle et al. (2017) which stated that fouling could be initiated after just 15 minutes of filtration. When the microbubbles were introduced to the system after the 700th minute, the absorbance value varied from 0.019 cm−1 to 0.0225 cm−1. It could be assumed that some of the deposited particles on the membrane surface scrubbed by the microbubbles were recirculated through the filtration system. Antithetically, there were no changes in EC value after MBs due to constant salt concentration as MF does not separate ions. The EC value however was observed to have remarkable changes over the 1,000 min filtration period due to some colloid breakdown and ion charge on the surface of the membrane, which has been explored by Thomas & Cremers (1970). Both values, however, continue to decrease over time and after MB treatment.
Overall fluidic-oscillator-generated microbubble cleaning
Pressure is the main probe for examining the properties and results of this study. Increasing transmembrane pressure drop (TMP) means that the filter is continuing to filter the impurities from the seawater and particles are deposited on the surface of the membrane. In this study, the microbubbles were introduced at the 600th minute to remove the deposited particles from the surface of the membrane.
Fluctuations in Figure 6 show the following experimental configuration in order:
Slightly open vent valve: better TMP reduction
Fully closed vent valve: best TMP reduction
Non-fluidic-oscillator-generated MBs: slowest TMP reduction
Fluidic oscillator defouling with vent valve
Figure 6 shows the pressure drop recorded for each 1-minute interval for the whole filtration cycle of 72 hrs (3 days). The fouling baseline was calculated and windowed over ten points of recorded TMP. The difference between the inlet and outlet pressure gives the TMP across the membranes. At the 450th minute, the pressure transducer recorded small fluctuations of pressure which heralded that impurities were starting to be deposited and filling the membrane surfaces and pores. Rapid fouling was observed from 500 min to 572 min and started to record a constant pressure drop until minute 702. The TMP fluctuates at minute 732 showing the pressure fluctuations due to pressure release from the diffuser once microbubbles were introduced at 705 min. These preliminary data show the positive relationship between the bubbles and cleaning due to shear forces, drag forces, and strong velocity fluctuations induced by the bubble flow (Nagaoka et al. 2006).
After the sparging processes were stopped, the TMP started to increase. At 2,000 minutes, a similar trend of fouling was observed for which the TMP remained constant showing the fouling had reached saturation. Fluidic-oscillator-generated MBs once again were introduced with zero vent flow to the system. This resulted in a higher defouling rate, where the bubbles possess sufficient or higher shear and drag force to detach the deposited particles. Higher TMP is also recorded after 1,000 and 2,500 minutes showing the MB defouling was not able to restore the performance of the membrane to its initial conditions. The slowest defouling rate was recorded as non-fluidic-oscillator-generated MBs were sparged to the system as shown in Figure 6. It required approximately 125 minutes to reduce the pressure drop before it was fouled. The data show that MBs generated by fluidic oscillator without vent valve flow have the highest efficiency of defouling followed by MBs generated by fluidic oscillator with vent valve and MBs generated without fluidic oscillator. The defouling assumption was in line with the dissolved particles as shown in Figure 5 where the fluid quality improved over time. This meant that most of the particles were filling the pores and MB sparging created additional forces for better defouling.
Fluidic oscillator and defouling rate
The highest TMP for each defouling was recorded at the time elapsed. Three defouling methods were applied, where the highest recorded defouling rates achieved were by using fluidic oscillator (condition: I) at 9.53 mbar/min followed by fluidic oscillator (condition: II) at 6.22 mbar/min and the lowest defouling rate without fluidic oscillator (condition: III) at 3.41 mbar/min. Attributable to greater flows with more shear and drag force, the MBs generated under condition I by using the fluidic oscillator have the highest defouling rate of 9.53 mbar/min. MBs generated under condition II by fluidic oscillation show half of the defouling rate followed by the condition-III-generated MBs, in agreement with Lee et al. (2014). This finding has also been studied by Wibisono et al. (2015), who stated that higher velocity and more bubble flow (as in condition II) reflect positive improvement in membrane process. Under condition III, Zimmerman et al. (2008) stated that the bubbles are tenfold larger in size. Wu et al. (2012) explained the limitation of larger size bubble on fouling control for the deposition of small particles.
Generally, the most efficient MF defouling is achieved by scouring the MF under fluidic-oscillator-generated MBs. This finding shows that the fluidic oscillator generates smaller MBs. This results in a higher efficiency of the cleaning effect in scrubbing the colloids and impurities deposited on the surface of the MF (Lee et al. 2014; Wibisono 2014). Zimmerman (2014) explained the MBs generated by fluidic oscillations would inhibit repulsion between bubbles and particles for better particle separations, which is also in agreement with the study conducted by Akuzawa et al. (2010) and Agarwal et al. (2011). This, however, leads to a different finding in using the fluidic oscillator for a cleaning effect. The highest defouling was observed while using oscillator without a flow of air in the vent valve (condition: II). The basic inference from this is that the greater flow of air to the diffuser results in more bubbles generated compared with the oscillator with the open vent valve. Manipulation of oscillator frequency by changing feedback loop length and bleeding flowrate is crucial to ensure smaller bubble generation (Zimmerman et al. 2008; Brittle et al. 2015). Figure 4 illustrates the characteristics and function of the fluidic oscillator which need to be further investigated.
Colloid deposition and its removal – SEM
Figure 7(a) shows SEM images of the colloids deposited on the surface of the MF membrane. Because MF has large pores, the filtration process will basically remove large-size molecules such as colloids. However, some salt particles might also be present due to the process of drying prior to SEM analysis. Figure 7(b) shows the defouled membrane filter after MB scouring. It can be clearly seen that MBs generated by fluid oscillation scrubbed all of the impurities from the surface of the membrane. Nevertheless, not all of the impurities were removed. The result obtained is similar to the ones conducted by Gwenaelle et al. (2017) where not 100% of impurities will be removed by MBs. It has also been suggested that combining MBs with other chemicals such as coagulant may help in improving the rate of impurities eliminated by MBs. The image however only shows the removal of the impurities by the final sparging process (after 72 hrs).
The MBs are able to increase the effectiveness of membrane cleaning and defouling. The fluidic-oscillator-generated MBs resulted in a higher defouling efficiency of the filtration membrane. The TMPs recorded were able to distinguish the relationship of fluidic oscillator and defouling rate. This, however, does not reflect the relationship of the fluidic oscillator and MB properties for defouling mechanisms without figurative data. Thus the following future works are very important to test all the hypotheses inferred from this study.
General assumptions and a preliminary relationship between fluidic oscillator and defouling were made as the membranes were defouled more rapidly with fluidic oscillations. This, however, requires further data measurement and analysis to distinguish a figurative relationship below.
Installation of the bleeding valve to the fluidic oscillator outlet (system inlet).
Bleeding valve flow rate to determine relationship between the flow rate of air and defouling.
The oscillation frequency of the fluidic oscillator for better fouling and defouling.
Microbubble size distribution at different oscillation frequency.
C. Harun would like to thank the Ministry of Higher Education (MOHE) Malaysia for financial support. We wish to thank Dr Pratik D. Desai and Dr Michael Hines for discussions and technical support. Many thanks to Mr Elliott Gunard and Mr Andrew G. Patrick for helping me to design and fabricate the defouling rig. This research was supported by the Engineering and Physical Sciences Research Council (EPSRC) by a grant number EP/I019790/1.