Evaluation of fouling deposition, fouling reversibility and energy consumption of submerged hollow fiber membrane systems with periodic backwash

Journal of Membrane Science 452 (2014) 319–331

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Evaluation of fouling deposition, fouling reversibility and energy consumption of submerged hollow fiber membrane systems with periodic backwash Ebrahim Akhondi a,b,c, Filicia Wicaksana a,d, Anthony Gordon Fane a,b,c,n a Singapore Membrane Technology Centre, Nanyang Environment and Water Research Institute, Nanyang Technological University, 1 Ceantech Loop, Singapore 637141, Singapore b Nanyang Environment and Water Research Institute, Nanyang Technological University, 1 Cleantech Loop, Singapore 637141, Singapore c Division of Environmental and Water Resources, School of Civil and Environmental Engineering, Nanyang Technological University, 50 Nanyang Avenue, Singapore 639798, Singapore d Department of Chemical and Materials Engineering, the University of Auckland, Auckland 1142, New Zealand

art ic l e i nf o

a b s t r a c t

Article history: Received 4 June 2013 Received in revised form 11 October 2013 Accepted 12 October 2013 Available online 22 October 2013

The filtration behavior and energy consumption of submerged hollow fiber membranes were investigated in a dead-end mode under constant flux operation with periodic backwash. Various operating parameters such as filtration flux, feed concentration, backwash duration, backwash strength, and the aid of air scouring during backwash were investigated to optimize the filtration performance and energy consumption. Baker's yeast was used as the model foulant. The transmembrane pressure (TMP) profiles obtained with this model foulant indicated the existence of two regions: a low fouling region followed by a high fouling region. In the low fouling region, periodic backwash could effectively remove the fouling layer. The percentage of reversible fouling by backwash was significantly lower in the second region which led to a rapid increase in TMP. The Taguchi experimental design method was used to determine the critical parameters, the effect of individual parameters and the optimum conditions. Since the transition into the second region of filtration is not favorable, the fouling rate in the first 10 cycles, the net permeate volume and specific energy consumption at the end of the first region were chosen as the design method responses. The results showed that filtration flux and feed concentration have significant effects on the membrane fouling rate. In order to maximize the net permeate volume, filtration flux, backwash duration and backwash strength were found to be the most important parameters, which must be carefully selected. To minimize the specific energy consumption, filtration flux, backwash duration and aeration rate during backwash were the most important factors. & 2013 Elsevier B.V. All rights reserved.

Keywords: Submerged hollow fiber membrane Fouling Backwash Energy consumption Taguchi's method

1. Introduction Low-pressure membrane systems, including microfiltration (MF) and ultrafiltration (UF) membranes have become a common technology for drinking water treatment. The application of submerged membrane systems have become increasingly popular in surface water treatment, wastewater recovery for indirect potable reuse and pre-treatment of reverse osmosis (RO) processes because of the lower energy consumption and the higher permeate quality. Submerged hollow fiber membranes with greater packing density and the feasibility of backwashing have been widely used [1–4].

n Corresponding author at: Singapore Membrane Technology Centre, Nanyang Environment and Water Research Institute, Nanyang Technological University, 1 Ceantech Loop, Singapore 637141, Singapore. Tel.: þ 65 6790 5272. E-mail address: [email protected] (A.G. Fane).

0376-7388/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2013.10.031

Membrane fouling has been the most severe problem and barrier to the increased use of MF and UF membrane technologies [5,6]. The degree of fouling in submerged membrane systems is a complex function of feed characteristics, membrane properties and operating conditions [7–9]. Inadequate hydrodynamic management and unsuitable membrane properties are factors that encourage or exacerbate membrane fouling [10,11]. Fouling control strategies are required to avoid decline in permeability due to severe membrane fouling in submerged hollow fiber systems [12–14]. It is possible to minimize fouling by both choosing a suitable membrane material with less adsorption of substances in the feed water and optimizing the operating conditions in the system [15–18]. Two types of fouling phenomena are distinguished for microfiltration and ultrafiltration. The first type is known as filtration-induced macrosolute or particle deposition, which occurs as external fouling or cake formation on the top surface of the membrane. This fouling is caused by

320

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

particles that are larger than the membrane pores, and is often reversible and nonadhesive [12].The second, which can occur with or without filtration, is macrosolute (e.g., humic substances, proteins, etc.) or particle adsorption onto the membrane or into the membrane pores [19,20]. This attachment depends on the specific intermolecular interactions between the particles and the membrane. This type of fouling is usually irreversible and adhesive [21–23]. Application of backwashing (reversal of permeate flow through the pores) with air scouring during backwash is a common physical approach to remove fouling in submerged systems [24]. Backwash is believed to loosen and detach the fouling cake from the membrane surface that can easily be removed by crossflow or air bubbles. However in some cases the cake layer might serve as a secondary layer to protect the membrane from internal fouling by macromolecular components. Thus, frequent backwash could provide additional opportunities for macromolecules to enter the membrane pores [25]. Ye et al. [26] observed a transition from a mixed cake layer of particulates and macromolecules at the beginning of the experiment to a fouling structure dominated by the macromolecules after some filtration/cleaning cycles. The results of a study with sea water showed that cyclical cleaning can reorganize the foulant structures and change subsequent fouling patterns [27]. The first few cycles showed more significant irreversible fouling than the following cycles while the percentage of irreversible fouling became constant after some cycles. The reason is that there is more chance of pore blocking for the new membrane than the used membrane. Thus pore blocking can be the most dominant fouling mechanism in the first few cycles [27,28]. Filtration duration, backwash duration, and backwash flow rate are important parameters in the fouling of submerged hollow fiber membranes. Ye et al. [27] investigated the effect of filtration duration (from 1200 to 5400 s per cycle) on membrane fouling of real seawater filtration while other operation parameters were kept constant. It was found that the final TMP after 16 h filtration and the percentage of reversible fouling which can be removed by backwash did not increase when filtration duration increased from 1200 to 3600 s. Further increase in filtration duration (from 3600 to 5400 s) promoted membrane fouling due to formation of a more compact cake layer which was more irreversible. Chua et al. [29] investigated the effect of backwash duration, on the membrane fouling of a pilot pressurized hollow fiber membrane module. Prolonged backwash duration was found to be more effective than increasing air scouring duration in controlling membrane plugging. Studies by Ye et al. [27] showed that the final TMP and fouling rate decreased by more than 50% with the increase of backwash duration from 10 to 30 s while no obvious improvement was observed for further increase of backwash duration. Also, the increase of backwash duration up to 30 s resulted in a slight increase in the percentage of fouling removed by backwash. However it dropped slightly after further increase in backwash duration. This indicates that excess backwash volume might also cause membrane blockage or change the structure of the fouling cake due to residual impurities in the backwash fluid. The effect of backwash flowrate on the membrane fouling limitation for real seawater filtration was investigated with other conditions kept constant by Ye et al. [27]. They observed the lowest final TMP after 16 h of filtration in the case where the backwash flux was 1.5 times the filtration flux. Further increase in backwash flux to twice the filtration flux was found to increase the final TMP. This implies that backwash changes the fouling rate during the filtration cycle. Similarly, the percentage of fouling removed by backwash increased slightly to a maximum for a backwash flux of 1.5 times the filtration flux and dropped slightly when the backwash flux was twice that of the filtration flux. These

results indicate the existence of an optimum backwash flux in fouling limitation. Similar to excessive backwash duration, it seems that excessive backwash flux also causes convection of impurities from the permeate side to membrane pores or the residual fouling layer which results in less reversible fouling and higher fouling rate. The existence of an optimum backwash flux in fouling limitation was also reported by Chua et al. [29]. They found that an increase of backwash flow rate up to twice that of the permeate flow rate resulted in process improvement, but no further benefit was observed for further increase in the backwash flow rate. Overall, an increased backwash flux was found to be slightly more effective than increased backwash duration when the same amount of backwash volume was being used [27,30]. It has been reported in studies that air scouring during backwash can assist fouling removal and improve backwash efficiency [31,32]. While the backwash is expected to detach the cake layer from the fiber, air scouring removes the loosened deposits away from the membrane surface [31,32]. The effect of aeration during backwash on the membrane fouling during seawater filtration was investigated by Ye at al. [27]. Their results showed that backwash with a moderate air flow rate had a lower final TMP and slowed down the fouling rate more effectively than a high air flow rate. In other words, high air flow rate actually limited the benefit of air scouring and did not improve reversibility. This unexpected high fouling rate observed in the high air flow rate was attributed to the bubble shape difference (slugs vs. small bubbles) at different air flow rates. The Taguchi method is a fractional factorial type of method for designing experiments. This method involves using orthogonal arrays to organize the parameters affecting the process and the levels at which they should be varied. Unlike the factorial design methods that test all possible combinations, the Taguchi method tests several pairs of combinations and collect the necessary data to determine the effect of each parameter and the most influential parameters. The aim of this study was to investigate the filtration behavior and energy consumption of submerged hollow fiber membranes over an extended period of time in dead-end mode under constant flux operation with periodic backwash. Previous studies, point to the importance of several factors that may interact. Therefore in this work the Taguchi experimental design method was used to determine the critical parameters, the contribution of individual parameters and their interactions and the optimum conditions [33,34]. The operating parameters studied were filtration flux, feed concentration, backwash duration, backwash strength, and the aid of air scouring during backwash in order to optimize the filtration performance and energy consumption.

2. Materials and methods 2.1. Equipment setup The experiments were performed using a bench scale submerged hollow fiber membrane system. A schematic diagram of the submerged hollow fiber unit is shown in Fig. 1. The membrane bundle comprised seven fibers of 15 cm length. The membrane bundle was immersed in a Plexiglas tank with 10 cm width, 30 cm length and 40 cm height. The bottom end of the hollow fibers bundle was sealed and permeate was extracted from the upper part of the bundle. The amount of permeate withdrawn was controlled by a peristaltic pump (Cole-Parmer, model 07523-80) which was used for suction of permeate and also backwashing of hollow fibers. At the same time, the same amount of permeate was fed into the tank to maintain a constant feed concentration by the peristaltic pump B (Cole-Parmer, model 07523-80). Permeate was

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

321

Fig. 1. The diagram of dead-end submerged hollow fiber membrane filtration system.

Table 1 Selected parameters and levels. Parameter

Level 1

Level 2

Level 3

Level 4

1. Flux (LMH) 2. Concentration (g/l) 3. Aeration rate(m/s) 4. BW time (min) 5. BW strength (times of filtration flux)

15 0.2 0.00055 0.5 0.5

25 0.5 0.0011 1 1

50 1 0.00165 2 1.5

75 2 0.00275 3 2

used to backwash the membrane. The rotation direction of the permeate pump was reversed and the solenoid valve in the air line was opened during the backwash. The flow rate of the permeate stream was calibrated with an electronic precision balance (Setra EL-410S) and was monitored during the experiments by a digital flowmeter (McMillan Co.). The pressure of the permeate stream were monitored with a pressure transducer (Cole-Parmer, model 68073-02). Air was introduced into the membrane tank during backwash by a blower (HIBLOW HP20) via a calibrated rotameter (DWYER 1–5 l/min) and through a series of nozzles positioned in the tank under the membrane module. The reported aeration rate throughout the manuscript is the superficial velocity of air (m/s). The operation of pumps, pressure transducer, air pump, solenoid valve, and digital balance was controlled by the CX-Supervisor program. The membranes used in the experiments were ZeeWeed HF membranes manufactured by GE Water & Process Technologies. The membrane was a Polyvinylidene Fluoride (PVDF) hollow fiber ultrafiltration membrane with 1.91 mm OD, 0.9 mm ID, and nominal pore size of 0.035 μm (UF). The hollow fiber modules used in all experiments were potted in the lab. A new membrane module was used for each experiment and the membrane modules were not reused. The number of fibers, packing density and the module length were kept the same in all experiments. The membrane area of the membrane modules was 0.00626 m2. Washed yeast was used as feed suspension in all experiments. The suspension was made by adding 10 g Lesaffre's instant dry yeast (Sacchromyces cerevisiae) to 1000 mL of MilliQ water. Washing was achieved by centrifuging the yeast suspension at 2000 rpm for 15 min, discarding the supernatant, and resuspending the pellet. This procedure was repeated two times. The washing was aimed to remove the residual proteins arising out of any broken yeast and soluble macromolecules (such as extra cellular polymeric substances). Baker's yeast was chosen as the model foulant because it contains biomass and colloids that represent particulate and polysaccharides fouling. A concentration range of 200–2000 mg/l (Table 1) was used to represent the MLSS concentration in wastewater treatment plants. Since membrane fouling at lower feed

concentration (such as the water treatment application) was less pronounced, higher concentration range was selected to accelerate membrane fouling. 2.2. Experimental design Taguchi's method was used to achieve meaningful results by the minimum number of experiments. The orthogonal array described by the symbol L16 is sufficient to examine the effects of five parameters at four levels; the 16 experiments being equivalent to 54 ¼ 625 combinations. Operating parameters such as flux, feed concentration, backwash duration, backwash strength, and aeration rate were select as the major influential parameters involved in this study. Table 1 lists parameters and levels which were examined in this study. The L16 array, which has 16 rows and the assignment of the parameters to the columns, are listed in Table 2. The filtration time was fixed at 15 min for all experiments. The experiments were not performed randomly. The respective ranges for each parameter were chosen based on previous studies and applications. However, the combination of levels in each experiment and the number of experiments were dictated by Taguchi's method. Some of the experiments were performed three times. The reported values are the average values of the three replicates. 2.3. Data analysis In order to analyze the results, Design-Expert 8 software (StatEase, Inc.) was used. The effect of operation conditions in membrane fouling limitation and energy consumption were evaluated by using four measurable response variables, which are discussed as follow: 1. Fouling rate is the average of TMP increase rate over 10 cycles (from 2nd cycle to 11th cycle). Fouling rate ¼

11 ∑ii ¼ ¼ 2 f ouling rate ðiÞ 10

ð1Þ

322

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

Table 2 Taguchi L16 orthogonal array. Experiment

Flux (LMH)

Concentration (g/l)

Aeration (m/s)

BW time (min)

BW strength (times of Filtration flux)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

15 15 15 15 25 25 25 25 50 50 50 50 75 75 75 75

0.2 0.5 1 2 0.2 0.5 1 2 0.2 0.5 1 2 0.2 0.5 1 2

0.00055 0.0011 0.00165 0.00275 0.0011 0.00055 0.00275 0.00165 0.00165 0.00275 0.00055 0.0011 0.00275 0.00165 0.0011 0.00055

0.5 1 2 3 2 3 0.5 1 3 2 1 0.5 1 0.5 3 2

0.5 1 1.5 2 2 1.5 1 0.5 1 0.5 2 1.5 1.5 2 0.5 1

TMP i f inal  TMP i initial Δt

ð2Þ

where TMP i initial and TMP i f inal are the initial and the final TMP values of filtration cycle i, and Δt is the filtration duration of each cycle. 2. Net permeate volume ¼ filtrate volume backwash volume Net permeate volume is the collected permeate volume at the beginning of the high fouling rate region. This region is when dTMP/dt changes rapidly, as discussed below (Section 3.1). 3. Number of cycles, this is the number of filtration cycles before transition into the high fouling rate region. 4. Total specific energy consumption¼(Filtration energy consumptionþ Backwash energy consumptionþAeration energy consumption)/Net permeate volume¼EP þ EB (see below).

The pumping specific energy consumption during both filtration and backwash ðWh m  3 Þ can be computed as [35] EP ¼

ðQ P ΔP P ðΔt Fil =3600Þηs Þ þ ðQ BW ΔP BW ðΔt BW =3600Þηs Þ Q P Δt Fil  Q BW Δt BW

ð3Þ

where Q P and Q BW are the mean permeate or backwash flow-rate respectively ðm3 s  1 Þ, ΔP P the pressure head loss through the suction pump (Pa), ΔP BW the pressure head loss during backwash, Δt Fil;BW , the time of filtration or backwash step (s) and ηs , the suction pump yield (asummed equal to 0.7). Specific energy consumption ðwh m  3 Þ due to air bubbling has been evaluated as [35,36]: 2 3 !1  1=λ Q A 4 104 yþ P  15 EB ¼ P 2:73  105 ηðλ  1Þ Q P 1000PT λ

ð4Þ

where P is the blower inlet pressure (Pa), T is the air temperature (K), λ is the ratio of specific heat capacity at constant pressure to specific heat capacity at constant volume (E 1.4), η, the blower 1 efficiency (generallyE0.7), Q A the aeration flow-rate ðm3 h Þ, Q p 1 the permeate flow rate ðm3 h Þ and y the membrane aerator depth (m). Reversibility (%) represents the percentage of fouling that can be removed by the backwash in the corresponding cycle. Reversibility ð%Þ ¼

TMP fi inal  TMP initial iþ1 ΔTMP i

ð5Þ

60 50

TMP (kPa)

Fouling rate ðiÞ ¼

40

(a)

30

(b)

20 10 0 -10

0

5

10

15

20

25

30

35

40

45

50

Time (hr)

Fig. 2. TMP vs. time profile of (a) continuous dead-end filtration and (b) dead-end filtration with periodical backwash and air scouring during backwash. At filtration flux of 15 l/m2 h, concentration of 0.5 g/l of washed yeast; backwash mode for (b): 15 min filtration with 1 min backwash at the same flux and 0.0011 m/s aeration rate.

where TMP i f inal is the final TMP value of filtration cycle i, TMP i þ 1 initial is the initial TMP value of filtration cycle i þ1, and ΔTMP i indicates the TMP increase during the filtration cycle i.

3. Results and discussion Initially some general observations are provided at low flux (Section 3.1) and high flux (Section 3.2). The results of the statistical design are discussed in Sections 3.3, 3.4 and 3.5. 3.1. Dead-end filtration at low filtration flux Fig. 2 shows the TMP vs. time profile at a flux of 15 l/m2 h both without and with intermittent backwash. It shows that periodical backwash has extended the filtration process. Under the continuous mode of filtration, the TMP was relatively constant over 15 h filtration and then the TMP started to rise rapidly. For the filtration with periodical backwash, the duration of the constant TMP region was prolonged to about 30 h. Backwash seems to be gradually less effective after 35 h filtration and backwash. Three regions are evident in the TMP vs. time profiles. These enlarged regions are shown in Fig. 3. In the first region of filtration, fouling rate was measured to be very low and the fouling layer seems to be fully removed during backwash; for the conditions of this experiment the first region was from 0 h to approximately 30 h. In the second region (Fig. 3(b)), fouling was still low but some irreversible fouling was detectable by means of TMP monitoring at the beginning of consecutive cycles. In the last region that occurred after 40 h filtration and backwash, a high fouling rate was observed and reversibility was observed to be lower than

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

323

Fig. 3. (a)–(c) TMP profiles at different stages of filtration with backwash. (d) Initial and final TMPs in filtration cycles. At filtration flux of 15 l/m2 h, concentration of 0.5 g/l of washed yeast; backwash mode: 15 min filtration with 1 min backwash at the same flux where the aeration rate is 0.0011 m/s.

3.2. Dead-end filtration at high filtration flux

Fig. 4. The fouling rate and the reversibility in each cycle. At filtration flux of 15 l/m2 h, concentration of 0.5 g/l of washed yeast, 15 min filtration with 1 min backwash at the same flux where the aeration rate is 0.0011 m/s.

the previous regions. This suggests that even though backwash removed part of the fouling cake, the foulant cake structure may have changed to a more compact structure. The increased fouling rate while the deposition rate is constant during constant flux dead-end filtration may be caused by increased local flux due to loss of effective filtration area by irreversible fouling. Fig. 3 (d) depicts initial and final TMPs during the experiment with periodical backwash. Final TMPs overlap initial TMPs in the first region of filtration which implies that hydraulic resistance due to foulant deposition was negligible. However in the later regions the initial and final TMPs diverge signifying greater fouling rates for the same amount of fouling deposition. The fouling rate and the reversibility in each cycle are shown in Fig. 4. The fouling rate ðdTMP=dtÞ was measured to be 0.4 kPa/h in the first cycle that increased to 1, 1.5, 4.1, 20 kPa/h in the 80th; 120th; 140th; and 160th cycle, respectively. The reversibility was observed to be widely scattered in the first few hours of filtration. Noise in pressure readings and the low TMP at the same time caused these variations in reversibility which are presumably not meaningful. Further increase of filtration duration decreased the variability of the reversibility of the foulant to a value of about 80% that indicates significant membrane fouling.

Fig. 5 shows the TMP vs. time profile at a flux 75 l/m2 h both without and with intermittent backwash. It is obvious that periodical backwash limits the membrane fouling. Under the continuous mode of filtration, the TMP increased around 40 kPa in less than 2 h and the fouling rate (as dTMP/dt) was constant over this period. For filtration with periodical backwash, a significant portion of the membrane permeability was restored by backwash at the end of each cycle over a period of 415 h. Eventually TMP rise was 40 kPa after 18 h of filtration with periodical backwash. Two regions were distinguished in the TMP vs. time profile but a low fouling rate region, similar to the first region of filtration at flux 15 l/m2 h, was not observed. The reversibility dropped suddenly at the beginning of the second region due to a rapid change in the cake layer resistance. The fouling rate and the reversibility in each cycle are shown in Fig. 6 for filtration and backwash at a flux of 75 l/m2 h. The fouling was less reversible in the first few cycles than the following cycles while the percentages of reversible fouling became constant after the 6th cycle. The reason for lower reversibility is that there are more chances of pore blocking for the new membrane than the used membrane. Thus in the first few cycles, pore blocking contributes significantly to membrane fouling which cannot be removed by backwash. Similarly, Ye et al. [27] found less reversible fouling in the first few cycles than the following cycles of filtration of seawater. Also, Howe et al. observed more significant fouling in the first filtration run than individual filtration runs later filtering with natural surface water [28]. The reversibility drops very sharply at the beginning of the second region which was after about 65 cycles of operation. This rapid decrease in backwash efficiency resulted in very high fouling rates in the subsequent cycles.

3.3. Experiment responses values The four-level L16 orthogonal table used for the optimization and the corresponding fouling rates and long term reversibility

324

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

80 70

dead-end with periodical backwash

TMP (kPa)

60

continuous dead-end filtration

50 40 30 20 10 0 0

2

4

6

8

10

12

14

16

18

20

Time(hr) Fig. 5. TMP vs. time profiles of continuous dead-end filtration (red data points) and dead-end filtration with periodical backwash and air scouring during backwash (blue data points). At filtration flux of 75 l/m2 h, concentration of 2 g/l of washed yeast; backwash mode: 15 min filtration with 2 min backwash at the same flux where the aeration rate was 0.00055 m/s. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

80

120 100

60

80

50 40

60

30

40

20

20

10 0

Reversibility (%)

dTMP/dt (kPa)

70

0

10

20

30

40

50

60

0 70

Filtration cycle Fig. 6. The fouling rate and the reversibility in each cycle. At filtration flux of 75 l/m2 h, concentration of 2 g/l of washed yeast, 15 min filtration with 2 min backwash at the same flux where the aeration rate is 0.00055 m/s.

Table 3 Experiment responses values. Experiment

Fouling rate (kPa/h)

Net permeate volume (L/m2)

Total specific energy consumption (Wh m  3)

Number of cycles

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

0.02 0.22 0.72 1.98 0.42 0.72 1.48 4.8 1.44 3 6.6 15 2.64 4.25 10.7 18.6

223.5 447 431.03 319.28 175.61 143.68 162.84 95.79 431.03 454.98 399.11 47.89 558.75 681.67 989.78 1021.71

2.96 9.11 25.7 75.4 12.8 10.08 5.96 6.9 23.3 20.1 13.3 13.05 74.3 30.4 58.9 66.2

61 128 137 142 39 33 27 16 43 39 37 5 34 39 59 63

obtained from sixteen candidate runs are shown in Table 3. Design-Expert 8 software was used to analyze the responses. 3.4. Evaluation and validation of models 3.4.1. Fouling rate model evaluation Table 4 shows the analysis of variance for the fouling rate responses. The Model F-value of 89.84 implies the model is significant. There is only a 0.17% chance that a “Model F-Value” this large could occur due to noise. Values of “Prob4F” less than

0.0500 indicate model parameters are significant. In this case filtration flux, feed concentration, and aeration rate were found to be significant parameters. A signal to noise ratio greater than 4 is desirable and in this set of experiments the ratio of 32.6 indicates an adequate signal and this model can be used to navigate the design space. The sum of squares for a parameter is the amount of information that can be attributed to the parameter as it changes. The percent contribution is obtained by summing all the parameter sums of squares (SS) and then taking each individual SS and

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

dividing by the total SS and multiplying by 100. When all parameters have the same degrees of freedom, the % contributions can be used to determine which are larger contributors than others. Table 4 Analysis of variance for the fouling rate responses. Source

Sum of squares

Mean square

F value P value Prob4F

Model A-Flux B-concentration C-Aeration D-backwash duration E-backwash strength

22.128 11.32 9.67 0.958 0.107

1.83 3.77 3.22 0.32 0.036

89.84 184.47 157.52 15.61 1.758

0.0017 0.0007 0.0008 0.0246 0.3273

0.061

0.020

0.57

0.6727

325

As shown in Fig. 7, the most important parameters determining fouling rate were filtration flux and feed concentration which had the biggest roles in fouling rate (51.2% and 43.75% respectively). The next effective parameter was aeration rate (4.3%). The effect of the backwash-related factors was evaluated to be much lower. Fig. 8 depicts a graph of the actual response (fouling rate) values vs. the predicted response values. This plot helps to detect a value, or group of values, that are not easily predicted by the model. However the data points are close to and split evenly by the 451 line so all the responses values were predicted by the model.

3.4.2. Evaluation of the net permeate volume model Table 5 shows the analysis of variance for the net permeate volume responses. It should be recalled that this represents the net permeate volume before transition to a high fouling rate

Fig. 9. Effect of parameters on the net permeate volume. Fig. 7. Effect of parameters on the fouling rate.

Fig. 8. The actual fouling rate values vs. the predicted fouling rate values.

Table 5 Analysis of variance for the net permeate volume responses.

Fig. 10. The actual permeate volume values vs. the predicted permeate volume values by the model.

Table 6 Analysis of variance for the total specific energy consumption responses.

Source

Sum of squares

Mean square

F value

P value Prob4F

Source

Sum of squares

Mean square

F value

P value Prob4 F

Model A-Flux B-concentration C-Aeration rate D-backwash duration E- backwash strength

1270442.51 967103.77 53038.69 10738.73 137481.18

104975.32 322367.92 17679.56 3579.58 45827.06

29.33 90.06 4.94 0.20 12.8

0.0089 0.0019 0.1112 0.8888 0.0324

2.321 1.368 0.123 0.252 0.700

0.257977 0.456167 0.042219 0.084211 0.233554

4.486015 7.93237 0.501349 1.464357 4.061317

0.0408 0.0165 0.7075 0.3155 0.0681

102080.15

34026.72

9.51

0.0484

Model A-Flux B-concentration C-Aeration rate D-backwash duration E-backwash strength

0.224

0.072795

1.724225 0.3328

326

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

(required cleaning etc). The Model F-value of 29.33 implies the model is significant. There is only a 0.89% chance that a “Model F-Value” this large could occur due to noise. Values of “Prob4F” less than 0.0500 indicate model terms are significant. In this case filtration flux, backwash duration and backwash strength were significant model terms. The signal to noise ratio must be greater than 4 for an adequate signal and a ratio of 17.465 indicates an adequate signal here. Based on this analysis of variance (ANOVA) results, this model can be used for further investigations.

Fig. 11. Effect of parameters on the total specific energy consumption.

The contribution of each parameter to the net permeate volume is shown in shown in Fig. 9. The parameter with the most effect on the net permeate volume was filtration flux with 76.1%. Backwash duration and backwash strength had the next biggest roles after filtration flux with 10.8% and 8% respectively. The feed concentration was found to be the next most important parameter (4.1%). The aeration rate during backwash was found to have the lowest effect on the permeate volume. Within the range of fluxes tested an increasing flux leads to greater net permeate volume, however we anticipate that beyond a certain flux the effect of fouling will decrease net permeate volume. Further discussion is given in Section (3.5.1). Fig. 10 depicts the actual net permeate volume values vs. the predicted net permeate volume values by the model. The data points are scattered evenly by the 451line so all the responses values were well predicted by the model. 3.4.3. Evaluation of total specific energy consumption model Table 6 shows the analysis of variance for the total specific energy consumption responses. The Model F-value of 4.48 implies the model is significant. Filtration flux and backwash duration are the most significant parameters. A signal to noise ratio of 6.505 indicates an adequate signal. Figs. 11 and 12 depict the contribution of each parameter to the total specific energy consumption and the actual total specific energy consumption values vs. the predicted total specific energy consumption values by the model. The parameters with the most effect on the total specific energy consumption were filtration flux with 51.3% and backwash duration with 26.3%. Aeration rate was found to be the next important parameter with 9.5% effect. Finally, backwash strength and feed concentration had lower influence on the total energy consumption. The effect of flux on specific energy is discussed further in Section 3.6. 3.5. The effects of operation conditions on fouling and long term performance of membrane

Fig. 12. The actual total specific energy consumption values vs. the predicted total specific energy consumption values by the model.

3.5.1. Effect of filtration flux Fig. 13(a) depicts the effect of imposed filtration flux on the membrane fouling rate. The deposition rate of the particles onto the membrane surface is faster at higher flux and higher feed concentration in dead-end filtration and there is some evidence that the fouling rate was higher at higher imposed flux for the same deposition rate. For instance, for the same deposition rate when filtering at (a) flux 25 l/m2 h and concentration of 1 g/l and (b) flux 50 l/m2 h and concentration of 0.5 g/l, the fouling rates

Fig. 13. The effect of filtration flux on (a) the fouling rate and (b) the net permeate volume at the end of the low fouling region of filtration. The other parameters were set at the average.

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

ðdTMP=dtÞ were measured to be 1.43 and 3.0 kPa/h respectively. These results show that the fouling potential of particles is dependent on the applied flux during filtration. The change in fouling rate may be due to the different cake layer structure [37]. The lower filtration flux can cause the cake layer to grow in thickness rather than density while high flux filtration can form a dense cake layer with higher specific cake resistance. Similarly Rodriguez et al. [38] reported the fouling potential to be directly proportional to the applied flux during filtration where this proportionality is related to the compression of the cake deposit occurring at high flux values. The higher the flux, the less porous the cake and therefore the higher the fouling rate. Fig. 13(b) shows the effect of imposed filtration flux on the net permeate volume at the end of the low fouling region of filtration. Since the transition into the second region of filtration with a very high fouling rate and low reversibility is not favorable, the net permeate volume at the end of the low fouling region was chosen as the design method response. The net permeate volume represents prediction for long-term fouling reversibility. The high net permeate volume at the lowest flux of 15 l/m2 h was due to the lose cake structure. The results suggest that this loose cake could be easily washed away by air scouring during backwash. The lowest net permeate volume observed was at a filtration flux of 25 l/m2 h and it proportionally increased for fluxes above 25 l/m2 h. Since the backwash strength was set at 0.5–2 times filtration flux, the higher net permeate volume at high fluxes was because of the higher backwash fluxes. It has also been found

327

in other studies that increase of backwash flow rate leads to the process improvement and can prolong the process [27,29]. At a filtration flux of 25 l/m2 h the cake structure is denser than at 15 l/m2 h while the backwash strength is not large enough to remove the cake layer. The imposed flux also has an impact on the specific energy consumption and this is discussed in Section 3.6. 3.5.2. Effect of feed concentration The effect of feed concentration on the fouling rate is shown in Fig. 14(a) while other operation conditions are kept constant. The results show that increase in feed concentration resulted in higher fouling rate. The increased fouling rate is not only because of a greater deposition rate due to higher feed concentration but may be because of different particle size distributions at different feed concentrations. At higher feed concentrations, particulates have more chance of coagulation and formation of a wider particle size distribution. Filtration of a mixture of small and large particles results in a cake with lower porosity due to the filling of the voidage between large particles with small particles [39,40]. Fig. 14(b) shows the effect of feed concentration on the net permeate volume. The non-linear rate of increase in net permeate volume with increase in feed concentration may be attributed to different fouling mechanisms at different feed concentrations. It can be seen that the net permeate volume increased when feed concentration increased from 0.2 to 1 g/l. The net permeate volume then decrease at higher concentrations. The reasons for

Fig. 14. The effect of feed concentration on (a) the fouling rate and (b) the net permeate volume at the end of the low fouling region of filtration. The other parameters were set at the average.

Fig. 15. The effect of aeration rate during backwash on (a) the fouling rate and (b) the net permeate volume at the end of the low fouling region of filtration. The other parameters were set at the average.

328

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

Fig. 16. The effect of backwash duration on (a) the fouling rate and (b) the net permeate volume at the end of the low fouling region of filtration. The other parameters were set at the average.

Fig. 17. The effect of backwash duration on the number of cycles in the low fouling region of filtration.

the maximum in this plot are not obvious but may be a result of more effective backwash for more homogeneous cakes at modest solids load competing with increasingly thicker cakes requiring removal at higher solids load. The other factors at higher solid load may be the increased viscosity of suspension. This could decrease the effectiveness of air scouring, as observed by Wicaksana et al. [41,42] for bubbling in a submerged membrane system with increased feed viscosity.

3.5.3. Effect of air scouring during backwash The effect of aeration during backwash on the membrane fouling during filtration and the net permeate volume is evaluated in this section. One of the benefits of the air scouring is that it moves the loose cake away from the membrane surface and the effective filtration zone. Based on the quantitative measure of the influence of individual factors (Fig. 7), the effect of aeration rate during backwash on the membrane fouling was modestly significant. Fig. 15 shows the effect of aeration rate during backwash on the fouling rate and the net permeate volume at the end of the low fouling region of filtration. When the air flow rate increased from 0.0011 m3/m2 s to 0.0016 m3/m2 s the fouling rate dropped noticeably to a lower level. However, further increase of the air flow rate did not limit the fouling rate. Since the fouling formation in the first few cycles is mainly

because of pore blocking of the membrane surface, higher air flow rates during backwash helps to remove fine particles and avoid higher local fluxes in the next cycles. Similarly Ye et al. [27] observed that moderate air flow rate slowed down the fouling rate during filtration, while the fouling rate tended to increase again with further increase in the air flow rate. In another study similar trends were attributed to the break-up of chemical flocs [43]. The influence of air flow rate during backwash on the net permeate volume was not significant. A slightly decrease in the net permeate volume appeared when the superficial air flow velocity increased from 0.00055 to 0.00277 m3/m2 s. This means that fouling reversibility was reduced at higher air flow rates and the air scouring during the backwash influenced the structure of the cake. Effect of backwash duration: Fig. 16(a) shows that the fouling rate did not vary considerably with increase in the backwash duration from 30 s to 3 min. In other words, even 30 s of backwash was as effective as 3 min backwash. The ANOVA results also indicated that influence of backwash on the fouling rate was very small. However the backwash did influence the number of cycles before the transition to high fouling rate. Fig. 16(b) shows that the net permeate volume increased to a maximum with the increase of backwash time up to 2 min. The net permeate volume then dropped slightly as the backwash time further increased to 3 min. It seems reasonable that an increase in backwash duration could be useful for fouling removal but the reason for decline in the net permeate volume for backwash durations greater than 2 min is possibly the permeate wastage. However further investigation showed that high backwash did not increase the number of cycles. Fig. 17 shows the effect of backwash duration on the number of cycles that filtration can be continued before transition to the severe fouling region. Similar to the net permeate volume, the number of cycles increased to its maximum for a backwash duration of 2 min while further increment reduced the number of cycles from 64 to 63 cycles. Ye et al. [27] similarly found that excessive backwash duration might foul the membrane or the remaining fouling cake due to residual impurities in the backwash flux. Effect of backwash strength/flux: The effect of backwash strength on the membrane fouling limitation for washed yeast suspension is shown in Fig. 18(a). Similar to the backwash duration, backwash strength (in terms of times of filtration flux) was found to have a very low influence on the fouling rate compared to the other operating parameters. It appears that variation of backwash strength in this range does not influence the fouling rate during the filtration cycle. In contrast with this Ye et al. [27] found a

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

329

Fig. 18. The effect of backwash strength on (a) the fouling rate and (b) the net permeate volume at the end of the low fouling region of filtration. The other parameters were set at the average.

Fig. 19. The effect of filtration flux on the specific energy consumption. The other parameters were set at the average.

Fig. 21. The effect of backwash duration on the specific energy consumption. The other parameters were set at the average.

impurities out of the membrane pores and cake structure. When the backwash flux is high, blocked pores can be open and the cake layer can be more effectively removed but at the same time the lumen-side of the membrane might also be fouled by residuals or impurities in the permeate convicted by the high velocity backwash solution. The existence of an optimum backwash flux in fouling limitation has been reported in other studies [27,29]. 3.6. Energy consumption evaluations

Fig. 20. The effect of aeration rate on the specific energy consumption. The other parameters were set at the average.

minimum fouling rate at a backwash strength of 1.5 times the filtration flux. It seems that the fouling rate depends on the membrane and feed properties. The different results might be due to differences in feeds and membranes properties. Fig. 18(b) shows the effect of backwash strength on the net permeate volume. The net permeate volume increased to a maximum as backwash strength increased from 0.5 to 1 times of filtration flux, then it decreased again to a minimum when the backwash flux was 1.5 times of filtration flux and finally the net permeate volume increased when backwash strength was twice that of filtration. The results confirmed the existence of an optimum backwash flux in fouling limitation. Low backwash flux cannot displace the

The ANOVA results showed that filtration flux, backwash duration, and aeration rate are the most influential parameters in the energy consumption with relative impact of 51.3%, 26.3%, and 9.4% respectively. Fig. 19 shows the effect of flux on energy consumption, based on fixed average values of the other parameters. It is obvious in Fig. 19 that the pump energy demand is lower than the blower energy demand for low filtration fluxes while it is much higher than that of the blower for high filtration fluxes. The permeate pump energy demand increases as the filtration flux increases from 15 to 75 l/m2 h. When the operating flux is higher the pressure drop across the membrane and fouling layer is greater, and this greater pressure difference has a higher energy demand. In contrast, the higher flux would increase the production rate while the aeration rate during backwash is constant over the time so the energy demand will be less in terms of kWh m  3 of permeate. When the system is operated at the maximum flux of 75 l/m2 h the total energy consumption is 5 times more than that at a filtration at flux of 25 l/m2 h. Similarly, Parameshwaran et al.[44] and Tangsubkul et al. [45] observed that operation at lower flux will required significantly less energy for the feed pump.

330

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

Fig. 20 depicts the effect of aeration rate on the energy consumption. Even though a higher aeration rate could limit the fouling rate it caused the net permeate volume to be reduced due to a more irreversible fouling cake. According to these findings, the increments in aeration rate did not decrease the pump energy demand while the blower energy demand was proportional to the aeration rate. The total energy demand increased as the aeration rate increased. The effect of backwash duration on the specific energy consumption is shown in Fig. 21. Both the pump specific energy consumption and blower specific energy consumption increased with increase in backwash duration. The reason is that more permeate water was consumed for longer backwash durations and that decreased the system productivity. Low productivity while the pump and blower are working results in higher energy consumption per unit volume of permeate. 3.7. Optimization of operating conditions Optimum operating conditions for maximizing the permeate volume in the low fouling region of filtration for yeast concentrations of 0.2, 0.5, and 2 g/l were: filtration flux, 75 l/m2 h; aeration rate, 0.00055 m/s; backwash duration, 2 min; backwash strength, 1 times of filtration flux; for concentration of 1 g/l the optimum backwash strength was predicted to be 0.5 times of filtration flux while the other parameters were the same as other concentrations. For minimizing the total energy consumption, optimum operating conditions were: filtration flux, 25 l/m2 h; aeration rate, 0.00055 m/s; backwash duration, 0.5 min; and backwash strength, 0.5 times of filtration flux. There has to be a tradeoff between the specific energy consumption and net permeate volume. The specific energy consumption would be high when the operating conditions were optimized for maximum permeate volume. On the other hand, the net permeate volume would be low when the system was optimized for minimum total specific energy consumption.

4. Conclusions The Taguchi experimental design method was used to determine the critical parameters, the effect of individual parameters and the optimum conditions for submerged membrane filtration of a yeast suspension. The experiments conducted showed that backwash provided effective removal of the fouling layer and prolonged the filtration process for submerged membrane systems especially at high imposed fluxes. However, excessive backwash duration and strength resulted in permeate wastage, severe pore blocking, and high specific energy consumption. Some of the deposits still were attached to the membrane after backwash. These deposits contribute additional resistance to the filtration in subsequent cycles eventually requiring removal by chemical cleaning. The analyses showed that filtration flux and feed concentration have significant effects on membrane fouling rate. The net permeate volume at the end of the low fouling region which represents fouling reversibility was most influenced by filtration flux, backwash duration and backwash strength which must be carefully selected. To minimize the specific energy consumption, filtration flux, backwash duration and aeration rate during backwash were the most important factors. There has to be a tradeoff between the specific energy consumption and net permeate volume for optimal operation of this type of membrane system. Application of the Taguchi experimental design method is a useful approach for this analysis.

Acknowledgments This research was performed at the Singapore Membrane Technology Centre which is supported by the Singapore Economic Development Board and Nanyang Technological University. References [1] A.G. Fane, A. Yeo, A. Law, K. Parameshwaran, F. Wicaksana, V. Chen, Low pressure membrane processes—doing more with less energy, Desalination 185 (2005) 159–165. [2] S. Nakatsuka, I. Nakate, T. Miyano, Drinking water treatment by using ultrafiltration hollow fiber membranes, Desalination 106 (1996) 55–61. [3] J. Günther, P. Schmitz, C. Albasi, C. Lafforgue, A numerical approach to study the impact of packing density on fluid flow distribution in hollow fiber module, J. Membr. Sci. 348 (2010) 277–286. [4] G. Di Profio, X. Ji, E. Curcio, E. Drioli, Submerged hollow fiber ultrafiltration as seawater pretreatment in the logic of integrated membrane desalination systems, Desalination 269 (2011) 128–135. [5] C. Gü ell, R.H. Davis, Membrane fouling during microfiltration of protein mixtures, J. Membr. Sci. 119 (1996) 269–284. [6] S. Chellam, J.G. Jacangelo, T.P. Bonacquisti, Modeling and experimental verification of pilot-scale hollow fiber, direct flow microfiltration with periodic backwashing, Environ. Sci. Technol. 32 (1998) 75–81. [7] M. Raffin, E. Germain, S.J. Judd, Influence of backwashing, flux and temperature on microfiltration for wastewater reuse, Sep. Purif. Technol. 96 (2012) 147–153. [8] P. Aimar, Slow colloidal aggregation and membrane fouling, J. Membr. Sci. 360 (2010) 70–76. [9] C.Y. Tang, T.H. Chong, A.G. Fane, Colloidal interactions and fouling of NF and RO membranes: a review, Adv. Colloid. Interface Sci. 164 (2011) 126–143. [10] S. Vigneswaran, D.Y. Kwon, H.H. Ngo, J.Y. Hu, Improvement of microfiltration performance in water treatment: is critical flux a viable solution?, Water Sci. Technol., 41, 2000, 2000; 309–315. [11] Z.F. Cui, S. Chang, A.G. Fane, The use of gas bubbling to enhance membrane processes, J. Membr. Sci. 221 (2003) 1–35. [12] Y. Li, W. Zhang, X. Zhang, C. Chen, J. Wang, Characterization of fouling in immersed polyvinylidene fluoride hollow fibre membrane ultrafiltration by particles and natural organic matter, Desalination Water Treat. 18 (2010) 309–314. [13] T. Ohn, M.S. Jami, E. Iritani, Y. Mukai, N. Katagiri, Filtration behaviors in constant rate microfiltration with cyclic backwashing of coagulated sewage secondary effluent, Sep. Sci. Technol. 38 (2003) 951–966. [14] J.C.T. Lin, D.J. Lee, C. Huang, Membrane fouling mitigation: membrane cleaning, Sep. Sci. Technol. 45 (2010) 858–872. [15] H. Ma, C.N. Bowman, R.H. Davis, Membrane fouling reduction by backpulsing and surface modification, J. Membr. Sci. 173 (2000) 191–200. [16] H. Ma, L.F. Hakim, C.N. Bowman, R.H. Davis, Factors affecting membrane fouling reduction by surface modification and backpulsing, J. Membr. Sci. 189 (2001) 255–270. [17] S. Nakatsuka, I. Nakate, T. Miyano, Drinking water treatment by using ultrafiltration hollow fiber membranes, Desalination 106 (1996) 55–61. [18] D. Rana, T. Matsuura, Surface modifications for antifouling membranes, Chem. Rev. 110 (2010) 2448–2471. [19] P. Xiao, F. Xiao, D.-s. Wang, T. Qin, S.-p. He, Investigation of organic foulants behavior on hollow-fiber UF membranes in a drinking water treatment plant, Sep. Purif. Technol. 95 (2012) 109–117. [20] H. Huang, N. Lee, T. Young, A. Gary, J.C. Lozier, J.G. Jacangelo, Natural organic matter fouling of low-pressure, hollow-fiber membranes: effects of NOM source and hydrodynamic conditions, Water Res. 41 (2007) 3823–3832. [21] K. Kimura, H. Yamamura, Y. Watanabe, Irreversible fouling in MF/UF membranes caused by Natural Organic Matters (NOMs) isolated from different origins, Sep. Sci. Technol. 41 (2006) 1331–1344. [22] D. Jermann, W. Pronk, S. Meylan, M. Boller, Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production, Water Res. 41 (2007) 1713–1722. [23] E. Aoustin, A.I. Schäfer, A.G. Fane, T.D. Waite, Ultrafiltration of natural organic matter, Sep. Purif. Technol. 22–23 (2001) 63–78. [24] N. Hilal, O.O. Ogunbiyi, N.J. Miles, R. Nigmatullin, Methods employed for control of fouling in MF and UF membranes: a comprehensive review, Sep. Sci. Technol. 40 (2005) 1957–2005. [25] J. Wu, P. Le-Clech, R.M. Stuetz, A.G. Fane, V. Chen, Novel filtration mode for fouling limitation in membrane bioreactors, Water Res. 42 (2008) 3677–3684. [26] Y. Ye, V. Chen, P. Le-Clech, Evolution of fouling deposition and removal on hollow fibre membrane during filtration with periodical backwash, Desalination 283 (2011) 198–205. [27] Y. Ye, L.N. Sim, B. Herulah, V. Chen, A.G. Fane, Effects of operating conditions on submerged hollow fibre membrane systems used as pre-treatment for seawater reverse osmosis, J. Membr. Sci. 365 (2010) 78–88. [28] K.J. Howe, A. Marwah, K.-P. Chiu, S.S. Adham, Effect of membrane configuration on bench-scale MF and UF fouling experiments, Water Res. 41 (2007) 3842–3849. [29] K.T. Chua, M.N.A. Hawlader, A. Malek, Pretreatment of seawater: results of pilot trials in Singapore, Desalination 159 (2003) 225–243.

E. Akhondi et al. / Journal of Membrane Science 452 (2014) 319–331

[30] T. Zsirai, P. Buzatu, P. Aerts, S. Judd, Efficacy of relaxation, backflushing, chemical cleaning and clogging removal for an immersed hollow fibre membrane bioreactor, Water Res. 46 (2012) 4499–4507. [31] C. Serra, L. Durand-Bourlier, M.J. Clifton, P. Moulin, J.-C. Rouch, P. Aptel, Use of air sparging to improve backwash efficiency in hollow-fiber modules, J. Membr. Sci. 161 (1999) 95–113. [32] Y. Bessiere, C. Guigui, P.J. Remize, C. Cabassud, Coupling air-assisted backwash and rinsing steps: a new way to improve ultrafiltration process operation for inside-out hollow fibre modules, Desalination 240 (2009) 71–77. [33] R.K. Roy, Design of Experiments Using the Taguchi Approach: 16 Steps to Product and Process Improvement, Wiley, New York, 2001. [34] R.L. Mason, R.F. Gunst, J.L. Hess, in: Robert L. Mason, Richard F. Gunst, James L. Hess (Eds.), Statistical Design and Analysis of Experiments: With Applications to Engineering and Science, 2nd ed.,J. Wiley, New York, 2003. [35] A. Massé, H.N. Thi, P. Legentilhomme, P. Jaouen, Dead-end and tangential ultrafiltration of natural salted water: influence of operating parameters on specific energy consumption, J. Membr. Sci. 380 (2011) 192–198. [36] B. Verrecht, S. Judd, G. Guglielmi, C. Brepols, J.W. Mulder, An aeration energy model for an immersed membrane bioreactor, Water Res. 42 (2008) 4761–4770. [37] J. Zhang, H.C. Chua, J. Zhou, A.G. Fane, Factors affecting the membrane performance in submerged membrane bioreactors, J. Membr. Sci. 284 (2006) 54–66.

331

[38] S.G.S. Rodríguez, M.D. Kennedy, G.L. Amy, J.C. Schippers, Flux dependency of particulate/colloidal fouling in seawater reverse osmosis systems, Desalination Water Treat. 42 (2012) 155–162. [39] S.S. Madaeni, Effect of large particles on microfiltration of small particles, J. Porous Mater. 8 (2001) 143–148. [40] L.N. Sim, Y. Ye, V. Chen, A.G. Fane, Crossflow sampler modified fouling index ultrafiltration (CFS-MFI UF)—an alternative fouling index, J. Membr. Sci. 360 (2010) 174–184. [41] F. Wicaksana, A.G. Fan, V. Chen, The relationship between critical flux and fibre movement induced by bubbling in a submerged hollow fibre system, Water Sci. Technol. 51 (2005) 115–122. [42] F. Wicaksana, A.G. Fane, V. Chen, Fibre movement induced by bubbling using submerged hollow fibre membranes, J. Membr. Sci. 271 (2006) 186–195. [43] B.-B. Choi, Y.-J. Choi, J.-S. Choi, S. Lee, H.-J. Oh, Energy management in submerged microfiltration systems by optimum control of aeration, Desalination 247 (2009) 233–238. [44] K. Parameshwaran, A.G. Fane, B.D. Cho, K.J. Kim, Analysis of microfiltration performance with constant flux processing of secondary effluent, Water Res. 35 (2001) 4349–4358. [45] N. Tangsubkul, K. Parameshwaran, S. Lundie, A.G. Fane, T.D. Waite, Environmental life cycle assessment of the microfiltration process, J. Membr. Sci. 284 (2006) 214–226.