- Email: [email protected]
Accelerated testing for fouling of microfiltration membranes using model foulants
Desalination 343 (2014) 113–119
Contents lists available at ScienceDirect
Desalination journal homepage: www.elsevier.com/locate/desal
Accelerated testing for fouling of microfiltration membranes using model foulants Yoon-Jin Kim a, Taekgeun Yun a, Sangho Lee a,⁎, Dohee Kim b, Jongdeok Kim b a b
School of Civil and Environmental Engineering, Kookmin University, Jeongneung-Dong, Seongbuk-Gu, Seoul, 136-702, Republic of Korea LG Electronics, Advanced Research Institute, 16 Woomyeon-dong, Seocho-gu, Seoul 137-724, Republic of Korea
H I G H L I G H T S • Fouling of hollow fiber membranes was evaluated by means of accelerated testing experiments. • A simple model based on pseudo-cake filtration model was applied to estimate the normalized fouling rates, θ/J2. • In the accelerated testing of membranes, the fouling rate was less sensitive to foulant concentration than to flux.
a r t i c l e
i n f o
Article history: Received 5 July 2013 Received in revised form 31 October 2013 Accepted 18 January 2014 Available online 18 February 2014 Keywords: Hollow fiber Fouling Microfiltration Mathematical models Accelerated testing Fouling rate
a b s t r a c t Hollow fiber microfiltration (MF) membranes have been widely employed for water and wastewater treatment. Nevertheless, membrane fouling is still one of the most serious issues in operating hollow fiber membrane systems. In this research, fouling of hollow fiber MF membranes was evaluated by means of accelerated testing experiments. A single fiber filtration unit was used to perform the fouling experiments in an accelerated way. Permeate flux and foulant concentrations were used as fouling control parameters to adjust the rate of fouling. Four model foulants used were kaolin, silica, natural organic matters (NOM), and alginate. A simple theoretical model was applied to investigate the fouling characteristics of the membrane by the model foulants. The analysis showed that there was a nonlinear correlation between the fouling rate and the fouling control parameters. The “normalized” fouling rate, θ/J2, was found to be useful to quantify the fouling rate and to implement a long-term simulation of TMP changes. © 2014 Elsevier B.V. All rights reserved.
1. Introduction Microfiltration or ultrafiltration (MF/UF) membrane has been gaining popularity as a feasible option for advanced water and wastewater treatment [1,2]. The use of MF/UF has been studied by researchers since the mid-1990s and cost reduction in these technologies in the mid-2000s led to the installation of MF/UF plants [3]. MF and UF systems typically utilize hollow fiber modules. A major advantage of hollow fiber membrane modules over other configurations of membranes is the high membrane surface area to footprint ratio achieved by low aspect ratio (diameter-to-length ratio) of fibers. Moreover, they provide cost-effective methods of removing particles and pathogenic microorganisms from treated water [4]. However, membrane fouling is still one of the most serious shortcomings in hollow fiber MF/UF systems [3,5]. The problem lies with the fact that membrane fouling is difficult to predict and control [6]. Fouling behavior is influenced by various factors, including membrane surface properties, the nature of the particle or dissolved foulants, and ⁎ Corresponding author. Tel.: +82 2 910 4529; fax: +82 2 910 4939. E-mail address: [email protected] (S. Lee). 0011-9164/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.desal.2014.01.016
feed water properties [7,8]. Accordingly, it is highly desirable to have an accelerated fouling test method that is short in duration, utilizes a minimum amount of test solution, only requires a small membrane area, and is representative of the large-scale process [9]. Accelerated life testing, which is defined as the process of testing a product by subjecting it to conditions in excess of its normal service parameters, has been widely applied to many industries to predict longterm performance of a product [10,11]. Nevertheless, relatively few works have been done in the field of MF/UF membranes. Previous works on accelerated testing of MF/UF membranes have focused on the chemical degradation and aging [12–14]. Although fouling status of membrane was identified as an important aging factor [14], little information is available on the accelerated testing conditions of membrane fouling. This study focuses on developing accelerated testing protocols as a tool to predict long-term performance and lifetime of hollow fiber MF membranes. Using model foulants such as kaolin, silica, natural organic matters (NOM), and alginate, fouling propensity of the membrane was quantitatively analyzed. A theoretical model was applied to analyze the fouling performance of hollow fiber MF membranes. The flux and
114
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
foulant concentrations were used as important parameters to accelerate fouling. Using the filtration model and the parameters obtained from the experiments, the fouling rates could be estimated.
algorithm in Fig. 1, a simulation model was developed using Matlab (Fig. 2).
2. Theory
3. Experimental section
2.1. Analysis of data from accelerated fouling test
3.1. Experimental setup
We applied a simple filtration model to estimate the fouling rates of dead-end microfiltration. Although there are a lot of different fouling mechanisms depending on the characteristics of foulants, pseudo-cake filtration model was adopted in this study for simple analysis of data. Based on this model, the permeate flux (J) on the TMP can be described by Darcy's law [15].
A schematic diagram of a laboratory-scale, submerged hollow fiber membrane system for accelerated fouling test is shown in Fig. 3. The system consisted of 12 filtration tanks, allowing the simultaneous testing of MF fibers at the same time. Each tank had a working volume of 1 L and MF fiber was immersed vertically in the reactor. A magnetic stirrer was positioned just below the membrane and the stirring speed was controlled by a magnetic stirrer plate. The MF fibers were made of polyvinylidene fluoride (PVDF) with the nominal pore size of 0.2 μm. They had an internal diameter of 0.7 mm and an external diameter of 1.3 mm. The length of the fiber was 18 cm. Since the fiber was relatively short, the pressure drop along the fiber was neglected. Permeate from the membrane module was pulled by a peristaltic pump (EW-07551-00, Cole-Parmer, USA). A permeate volume was frequently measured by collecting permeate volume using a mass cylinder. The transmembrane pressure was continuously measured using a pressure transducer (ISE40A-01-R, SMC, Japan) and a data logger (usb-6008, NI. U.S.A.), which were connected to a computer. The temperature of solution was kept constant at 20 °C. Total recycle mode, where both the retentate from the MF loop and permeate were recycled into the tank, was adopted to keep the reactor volume constant during the operation time.
ΔP ¼ ηðRm þ Rc ÞJ
ð1Þ
where ΔP is the transmembrane pressure (TMP); η is the viscosity of water; Rm is the membrane resistance; Rc is the cake resistance; and J is the permeate flux. The cake resistance (Rc) is given by Rc ¼
αmc Am
ð2Þ
where α is the specific cake resistance; mc is the mass of cake deposited on the membrane; and Am is the membrane area. Here, mc is proportional to the flux of the foulants: mc ¼ JAm c f t
ð3Þ
where cf is the effective foulant concentration. Note that cf is different from cb, which is the bulk concentration of foulant. This implies that all foulants cannot approach the membrane surface due to the back transport. By combining Eqs. (1), (2), and (3), ΔP is given by 2
ΔP ¼ ηRm J þ ηαc f J t ¼ ηJRm þ θt
ð4Þ
where θ is the rate of membrane fouling in dead-end filtration tests. Under constant flux conditions, θ can be calculated from the slope of the plot between t and ΔP. To accelerate the fouling rate (θ), either J or cf may be increased. However, θ may not be linearly proportional to J (or cf). Accordingly, it is important to understand the correlations between θ and J (or cf), which should be experimentally determined.
3.2. Model foulants and test conditions Model foulants used in this study were kaolin (Sigma Aldrich), silica (Sigma Aldrich, Ludox colloidal silica AM-30), NOM (IHSS, Swannee river natural organic matter), and alginate (Sigma Aldrich, alginic acid sodium salt from brown algae). The concentrations of the foulants were 2, 5, 10, 20 mg/L, respectively. The flux step method was applied to investigate the effect of applied flux on fouling rate by adjusting flux between 50 L/m2 h and 200 L/m2 h. Prior to each filtration test, all membranes were stabilized using deionized water during 500 min.
2.2. Using the simulation based on the mathematical model Once θ is known, it can be used for long-term simulation of hollow fiber MF systems. Of course, there should be a substantial difference between the results from this simple simulation and the actual data from a pilot or full scale plants. The information from the simulation should be used just for initial projection of the performance of MF membranes. Considering the situation that periodic backwash is applied, Eq. (4) is modified as 2
ΔP ðt þ Δt Þ ¼ ΔP ðt Þ þ αηJ c f Δt
ð5Þ
ΔP ðt Þjafter backwash ¼ ΔP ðt Þjbefore backwash −JηRc ðt Þβ
ð6Þ
where β is the constant to describe the effect of backwash. To consider the effect of viscosity, the following correlation was used [16]: η ¼ 2:414 10
−5
247:8 10ðT−140Þ
ð7Þ
The flow chart for the simulation is shown in Fig. 1. Here, β may be determined from a set of experiments. Based on the
4. Results and discussion 4.1. Effect of flux and foulant concentrations on transmembrane pressure To examine the basic properties of the MF membranes, a set of filtration tests were out using colloidal silica under the following operating conditions: flux, 50, 100, 150, and 200 L/m2 h; concentrations of the silica, 2, 5, 10, and 20 mg/L. The results are shown in Fig. 4. As the flux increases, the fouling rate, which is the slope of the TMP curve, increases. Nevertheless, the fouling rates at same flux were different in some cases. For example, the fouling rate at 2 mg/L of silica was 9.05 × 10− 5 bar/min when the flux increases from 100 L/m2 h to 150 L/m2 h. On the other hand, it was 3.02 × 10−4 bar/min when the flux decreases from 200 L/m2 h to 150 L/m2 h. It is evident that there is a hysteresis in TMP changes due to the irreversibility of membrane fouling. If the fouling layer is removed by relaxing the flux, the first and second fouling rates should be identical. Accordingly, the difference between these fouling rates, which may be defined as the degree of hysteresis, may be used as a quantitative measure to describe the characteristics of fouling.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
115
Fig. 1. Flow chart for simulation of continuous hollow fiber MF systems.
4.2. Membrane fouling rate by single model foulant Under similar operation conditions as the previous one, filtration experiments were also carried out using the other model foulants, including kaolin, NOM, and alginate. As a result of these tests, the slopes of the TMP curve (θ) were calculated. When the slopes were calculated, initial fouling at the beginning of each flux step was not considered to increase the reproducibility of the results. As indicated in Eq. (4), θ is proportional to J2. Accordingly, the intrinsic characteristics of foulants, ηαcf, should be determined by dividing θ by J2. The “normalized” fouling rate, θ/J2, is shown as a function of flux in Fig. 5. Although the concentrations were the same, the fouling propensities were different for different foulants. This is because those foulants
have different fouling potentials even at the same concentration. The order of fouling propensity is inversely reflected in the size of particles and hydrophobicity of organic matters [16,17]: θ/J2 (NOM) N θ/J2 (alginate) N θ/J2 (silica) N θ/J2 (kaolin). The kaolin, which is denoted as the black color, did not result significant membrane fouling. However, it is known that the kaolin may cause fouling in combination with other foulants. In the cases of the silica (red color) and NOM (green color), the fouling rates were different even at the same flux (the first and second runs at 50 LMH). This suggests that irreversible fouling occurs in the filtration of silica and NOM. The alginate (yellow color) initially resulted in rapid fouling, and then led to similar fouling rate regardless of applied flux. This is attributed to the initial adsorptive fouling by the interaction between alginate and membrane surface.
116
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
Fig. 2. A Matlab-based simulation program for MF systems.
Similar trends are observed at different concentrations, as illustrated in Figs. 6, 7, and 8. Except for alginate, θ/J2 increases with increasing flux. Nevertheless, θ/J2 is not strongly dependent on the foulant concentrations. According to Eq. (3), θ/J2 is proportional to cf, which may not be identical to cb. Based on the results, it is likely that cf does not increase even with increasing cb. As mentioned above, the degree of hysteresis in θ/J2 may be used to quantify the irreversible nature of fouling. For instance, NOM showed high degree of hysteresis at 2 mg/L but alginate showed low degree of hysteresis at 2 mg/L (Fig. 5). At 5 mg/L, both NOM and alginate exhibit relatively low degree of hysteresis (Figs. 6 and 7). This suggests that the
degrees of hysteresis may provide additional information on the characteristics of fouling at a given condition. Fig. 9 shows the dependence of θ/J2 on foulant concentration for the model foulants. The results confirm that θ/J2 does not significantly change with the concentration. Except for kaolin, the θ/J2 values range from 5 × 10− 9 to 10 × 10−9 bar/min (L/m2 h)2. It should be noted that fouling rate is not linearly proportional to the foulant concentration. There are several reasons: first, the foulants may be added in an excess amount and thus the saturation of foulant deposition occurs during the tests; second, adsorption of foulants may be limited by the membrane surface rather than by the foulant concentration. These results suggest that the fouling rate is more sensitive to flux than foulant concentration. If the fouling rate should be increased during the accelerated testing, it is better to adjust the flux than the foulant.
4.3. Simulation of continuous MF system using θ/J2
Fig. 3. Schematic diagram of experimental setup for MF tests: (a) multi-channel peristaltic pump, (b) pressure transducer, (c) hollow fiber, (d) magnet stirrer, (e) data logger and (f) desktop.
Although the focus of this research was experimental, we applied the simple model described in Fig. 1 to aid in interpreting the continuous operation of MF systems and to predict membrane performance over a longterm filtration. Fig. 10 shows the examples of simulation results using experimental θ/J2. Since the backwash interval for most pilot- or full-scale membrane systems for surface water treatment ranges from 15 min to 60 min, the simulation was performed using a backwash interval of 30 min. In our model, the effect of the backwash conditions is reflected by β in Eq. (6). Results showed that the TMP profiles can be simulated using the parameter (θ/J2) obtained from the accelerated testing. The increase in TMP within the backwash interval is attributed to θ/J2 but the overall increase in TMP is affected by β. In this simulation, β is assumed to be 0.9 but different β values would result in different results.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
117
Fig. 4. Changes in transmembrane pressure (TMP) as a function of filtration time: (A) 2 mg/L silica, (B) 5 mg/L silica, (C) 10 mg/L silica, and (D) 20 mg/L silica.
(1) A simple model based on pseudo-cake filtration model was applied to estimate the fouling parameters. The model could be successfully applied to interpret experimental data obtained using a laboratory-scale MF system.
(2) Although foulant concentration and flux were used to control the rate of fouling, the fouling rate was less sensitive to foulant concentration than to flux. This is because the effective foulant concentration approaching the membrane surface is different from the bulk concentration. (3) The normalized fouling rates, θ/J2, were difference for different foulants even at the same concentration. The alginate initially resulted in rapid fouling, and then led to similar fouling rate regardless of applied flux, suggesting the initial adsorptive fouling. (4) Using the θ/J2, which was estimated from the accelerated test, long-term operations of MF systems could be simulated. Nevertheless, additional approaches, including the consideration
Fig. 5. Effect of flux on membrane fouling rates under various foulant concentrations (2 mg/L).
Fig. 6. Effect of flux on membrane fouling rates under various foulant concentrations (5 mg/L).
Accordingly, a new technique to estimate β should also be developed for the long-term modeling of MF system with backwashing. 5. Conclusions In this work, an accelerated testing method was investigated to analyze fouling propensity of hollow fiber MF membranes. The following conclusions can be drawn:
118
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
Fig. 7. Effect of flux on membrane fouling rates under various foulant concentrations (10 mg/L).
of backwash effect, should be done to quantitatively predict the long-term performance of MF systems based on the results of the accelerated testing. Nomenclature ΔP trans-membrane pressure (bar) η fluid viscosity (Pa s) Rm membrane resistance (m−1) Rc cake resistance (m−1) J permeate flux (Lm−2 h−1) α specific cake resistance mc mass of cake deposited on the membrane (mg)
Fig. 8. Effect of flux on membrane fouling rates under various foulant concentrations (20 mg/L).
Am cf cb θ t β
membrane area (m2) effective foulant concentration (mg/L) bulk concentration of foulant (mg/L) rate of membrane fouling filtration time (h) constant to describe the effect of backwash
Acknowledgements This research was supported by LG Electronics, Advanced Research Institute.
Fig. 9. Effect of foulant concentration on membrane fouling rates for various model foulants: (A) kaolin, (B) silica, (C) NOM, and (D) alginate.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
119
References [1] Zeman, A.L. Zydney marcel dekker, Microfiltration and Ultrafiltration - principles and applications 1996. 648. [2] A.G. Fane, S. Chang, E. Chardon, Submerged hollow fibre membrane module—design options and operational considerations, Desalination 146 (2002) 231–236. [3] W. Gao, H. Liang, J. Ma, M. Han, Z.-l. Chen, Z.-s. Han, G.-b. Li, Membrane fouling control in ultrafiltration technology for drinking water production: a review, Desalination 272 (2011) 1–8. [4] T. Lebeau, C. Lelievre, H. Buisson, D. Cleret, L.W. Van der Venter, P. Cote, Immersed membrane filtration for the production of drinking water: combination with PAC for NOM and SOCs removal, Desalination 117 (1998) 219–231. [5] W. Neubrand, S. Vogler, M. Ernst, M. Jekel, Lab and pilot scale investigations on membrane fouling during the ultrafiltration of surface water, Desalination 250 (2010) 968–972. [6] 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. [7] B. Sarkar, A combined complete pore blocking and cake filtration model during ultrafiltration of polysaccharide in a batch cell, J. Food Eng. 116 (2013) 333–343. [8] Y.S. Polyakov, A.L. Zydney, Ultrafiltration membrane performance: effects of pore blockage/constriction, J. Membr. Sci. 434 (2013) 106–120. [9] D. Mosqueda-Jimenez, R. Narbaitz, T. Matsuura, Membrane fouling test: apparatus evaluation, J. Environ. Eng. 130 (2004) 90–99. [10] S.H. Mohammadian, D. Aït-Kadi, F. Routhier, Quantitative accelerated degradation testing: practical approaches, Reliab. Eng. Syst. Saf. 95 (2010) 149–159. [11] S.J. Bae, S.-J. Kim, J.I. Park, C.W. Park, J.-H. Lee, I. Song, N. Lee, K.-B. Kim, J.-Y. Park, Lifetime prediction of a polymer electrolyte membrane fuel cell via an accelerated startup–shutdown cycle test, Int. J. Hydrog. Energy 37 (2012) 9775–9781. [12] C. Regula, E. Carretier, Y. Wyart, M. Sergent, G. Gésan-Guiziou, D. Ferry, A. Vincent, D. Boudot, P. Moulin, Ageing of ultrafiltration membranes in contact with sodium hypochlorite and commercial oxidant: experimental designs as a new ageing protocol, Sep. Purif. Technol. 103 (2013) 119–138. [13] I. Levitsky, A. Duek, E. Arkhangelsky, D. Pinchev, T. Kadoshian, H. Shetrit, R. Naim, V. Gitis, Understanding the oxidative cleaning of UF membranes, J. Membr. Sci. 377 (2011) 206–213. [14] E. Zondervan, A. Zwijnenburg, B. Roffel, Statistical analysis of data from accelerated ageing tests of PES UF membranes, J. Membr. Sci. 300 (2007) 111–116. [15] H. Darcy, Les fontaines publiques de La Ville Dijon, 1856. [16] S. Lee, P.K. Park, J.H. Kim, K.M. Yeon, C.H. Lee, Analysis of filtration characteristics in submerged microfiltration for drinking water treatment, Water Res. 42 (2008) 3109–3121. [17] K.-J. Hwang, C.-Y. Liao, Effects of membrane morphology and operating conditions on microfiltration particle fouling, J. Taiwan Inst. Chem. Eng. 43 (2012) 46–52. Fig. 10. Model simulation using different θ/J2 values. Conditions: backwash interval, 30 min; β, 0.9; J, 50 L/m2 h. (A) θ/J2 = 5 × 108. (B) θ/J2 = 10 × 108.
Contents lists available at ScienceDirect
Desalination journal homepage: www.elsevier.com/locate/desal
Accelerated testing for fouling of microfiltration membranes using model foulants Yoon-Jin Kim a, Taekgeun Yun a, Sangho Lee a,⁎, Dohee Kim b, Jongdeok Kim b a b
School of Civil and Environmental Engineering, Kookmin University, Jeongneung-Dong, Seongbuk-Gu, Seoul, 136-702, Republic of Korea LG Electronics, Advanced Research Institute, 16 Woomyeon-dong, Seocho-gu, Seoul 137-724, Republic of Korea
H I G H L I G H T S • Fouling of hollow fiber membranes was evaluated by means of accelerated testing experiments. • A simple model based on pseudo-cake filtration model was applied to estimate the normalized fouling rates, θ/J2. • In the accelerated testing of membranes, the fouling rate was less sensitive to foulant concentration than to flux.
a r t i c l e
i n f o
Article history: Received 5 July 2013 Received in revised form 31 October 2013 Accepted 18 January 2014 Available online 18 February 2014 Keywords: Hollow fiber Fouling Microfiltration Mathematical models Accelerated testing Fouling rate
a b s t r a c t Hollow fiber microfiltration (MF) membranes have been widely employed for water and wastewater treatment. Nevertheless, membrane fouling is still one of the most serious issues in operating hollow fiber membrane systems. In this research, fouling of hollow fiber MF membranes was evaluated by means of accelerated testing experiments. A single fiber filtration unit was used to perform the fouling experiments in an accelerated way. Permeate flux and foulant concentrations were used as fouling control parameters to adjust the rate of fouling. Four model foulants used were kaolin, silica, natural organic matters (NOM), and alginate. A simple theoretical model was applied to investigate the fouling characteristics of the membrane by the model foulants. The analysis showed that there was a nonlinear correlation between the fouling rate and the fouling control parameters. The “normalized” fouling rate, θ/J2, was found to be useful to quantify the fouling rate and to implement a long-term simulation of TMP changes. © 2014 Elsevier B.V. All rights reserved.
1. Introduction Microfiltration or ultrafiltration (MF/UF) membrane has been gaining popularity as a feasible option for advanced water and wastewater treatment [1,2]. The use of MF/UF has been studied by researchers since the mid-1990s and cost reduction in these technologies in the mid-2000s led to the installation of MF/UF plants [3]. MF and UF systems typically utilize hollow fiber modules. A major advantage of hollow fiber membrane modules over other configurations of membranes is the high membrane surface area to footprint ratio achieved by low aspect ratio (diameter-to-length ratio) of fibers. Moreover, they provide cost-effective methods of removing particles and pathogenic microorganisms from treated water [4]. However, membrane fouling is still one of the most serious shortcomings in hollow fiber MF/UF systems [3,5]. The problem lies with the fact that membrane fouling is difficult to predict and control [6]. Fouling behavior is influenced by various factors, including membrane surface properties, the nature of the particle or dissolved foulants, and ⁎ Corresponding author. Tel.: +82 2 910 4529; fax: +82 2 910 4939. E-mail address: [email protected] (S. Lee). 0011-9164/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.desal.2014.01.016
feed water properties [7,8]. Accordingly, it is highly desirable to have an accelerated fouling test method that is short in duration, utilizes a minimum amount of test solution, only requires a small membrane area, and is representative of the large-scale process [9]. Accelerated life testing, which is defined as the process of testing a product by subjecting it to conditions in excess of its normal service parameters, has been widely applied to many industries to predict longterm performance of a product [10,11]. Nevertheless, relatively few works have been done in the field of MF/UF membranes. Previous works on accelerated testing of MF/UF membranes have focused on the chemical degradation and aging [12–14]. Although fouling status of membrane was identified as an important aging factor [14], little information is available on the accelerated testing conditions of membrane fouling. This study focuses on developing accelerated testing protocols as a tool to predict long-term performance and lifetime of hollow fiber MF membranes. Using model foulants such as kaolin, silica, natural organic matters (NOM), and alginate, fouling propensity of the membrane was quantitatively analyzed. A theoretical model was applied to analyze the fouling performance of hollow fiber MF membranes. The flux and
114
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
foulant concentrations were used as important parameters to accelerate fouling. Using the filtration model and the parameters obtained from the experiments, the fouling rates could be estimated.
algorithm in Fig. 1, a simulation model was developed using Matlab (Fig. 2).
2. Theory
3. Experimental section
2.1. Analysis of data from accelerated fouling test
3.1. Experimental setup
We applied a simple filtration model to estimate the fouling rates of dead-end microfiltration. Although there are a lot of different fouling mechanisms depending on the characteristics of foulants, pseudo-cake filtration model was adopted in this study for simple analysis of data. Based on this model, the permeate flux (J) on the TMP can be described by Darcy's law [15].
A schematic diagram of a laboratory-scale, submerged hollow fiber membrane system for accelerated fouling test is shown in Fig. 3. The system consisted of 12 filtration tanks, allowing the simultaneous testing of MF fibers at the same time. Each tank had a working volume of 1 L and MF fiber was immersed vertically in the reactor. A magnetic stirrer was positioned just below the membrane and the stirring speed was controlled by a magnetic stirrer plate. The MF fibers were made of polyvinylidene fluoride (PVDF) with the nominal pore size of 0.2 μm. They had an internal diameter of 0.7 mm and an external diameter of 1.3 mm. The length of the fiber was 18 cm. Since the fiber was relatively short, the pressure drop along the fiber was neglected. Permeate from the membrane module was pulled by a peristaltic pump (EW-07551-00, Cole-Parmer, USA). A permeate volume was frequently measured by collecting permeate volume using a mass cylinder. The transmembrane pressure was continuously measured using a pressure transducer (ISE40A-01-R, SMC, Japan) and a data logger (usb-6008, NI. U.S.A.), which were connected to a computer. The temperature of solution was kept constant at 20 °C. Total recycle mode, where both the retentate from the MF loop and permeate were recycled into the tank, was adopted to keep the reactor volume constant during the operation time.
ΔP ¼ ηðRm þ Rc ÞJ
ð1Þ
where ΔP is the transmembrane pressure (TMP); η is the viscosity of water; Rm is the membrane resistance; Rc is the cake resistance; and J is the permeate flux. The cake resistance (Rc) is given by Rc ¼
αmc Am
ð2Þ
where α is the specific cake resistance; mc is the mass of cake deposited on the membrane; and Am is the membrane area. Here, mc is proportional to the flux of the foulants: mc ¼ JAm c f t
ð3Þ
where cf is the effective foulant concentration. Note that cf is different from cb, which is the bulk concentration of foulant. This implies that all foulants cannot approach the membrane surface due to the back transport. By combining Eqs. (1), (2), and (3), ΔP is given by 2
ΔP ¼ ηRm J þ ηαc f J t ¼ ηJRm þ θt
ð4Þ
where θ is the rate of membrane fouling in dead-end filtration tests. Under constant flux conditions, θ can be calculated from the slope of the plot between t and ΔP. To accelerate the fouling rate (θ), either J or cf may be increased. However, θ may not be linearly proportional to J (or cf). Accordingly, it is important to understand the correlations between θ and J (or cf), which should be experimentally determined.
3.2. Model foulants and test conditions Model foulants used in this study were kaolin (Sigma Aldrich), silica (Sigma Aldrich, Ludox colloidal silica AM-30), NOM (IHSS, Swannee river natural organic matter), and alginate (Sigma Aldrich, alginic acid sodium salt from brown algae). The concentrations of the foulants were 2, 5, 10, 20 mg/L, respectively. The flux step method was applied to investigate the effect of applied flux on fouling rate by adjusting flux between 50 L/m2 h and 200 L/m2 h. Prior to each filtration test, all membranes were stabilized using deionized water during 500 min.
2.2. Using the simulation based on the mathematical model Once θ is known, it can be used for long-term simulation of hollow fiber MF systems. Of course, there should be a substantial difference between the results from this simple simulation and the actual data from a pilot or full scale plants. The information from the simulation should be used just for initial projection of the performance of MF membranes. Considering the situation that periodic backwash is applied, Eq. (4) is modified as 2
ΔP ðt þ Δt Þ ¼ ΔP ðt Þ þ αηJ c f Δt
ð5Þ
ΔP ðt Þjafter backwash ¼ ΔP ðt Þjbefore backwash −JηRc ðt Þβ
ð6Þ
where β is the constant to describe the effect of backwash. To consider the effect of viscosity, the following correlation was used [16]: η ¼ 2:414 10
−5
247:8 10ðT−140Þ
ð7Þ
The flow chart for the simulation is shown in Fig. 1. Here, β may be determined from a set of experiments. Based on the
4. Results and discussion 4.1. Effect of flux and foulant concentrations on transmembrane pressure To examine the basic properties of the MF membranes, a set of filtration tests were out using colloidal silica under the following operating conditions: flux, 50, 100, 150, and 200 L/m2 h; concentrations of the silica, 2, 5, 10, and 20 mg/L. The results are shown in Fig. 4. As the flux increases, the fouling rate, which is the slope of the TMP curve, increases. Nevertheless, the fouling rates at same flux were different in some cases. For example, the fouling rate at 2 mg/L of silica was 9.05 × 10− 5 bar/min when the flux increases from 100 L/m2 h to 150 L/m2 h. On the other hand, it was 3.02 × 10−4 bar/min when the flux decreases from 200 L/m2 h to 150 L/m2 h. It is evident that there is a hysteresis in TMP changes due to the irreversibility of membrane fouling. If the fouling layer is removed by relaxing the flux, the first and second fouling rates should be identical. Accordingly, the difference between these fouling rates, which may be defined as the degree of hysteresis, may be used as a quantitative measure to describe the characteristics of fouling.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
115
Fig. 1. Flow chart for simulation of continuous hollow fiber MF systems.
4.2. Membrane fouling rate by single model foulant Under similar operation conditions as the previous one, filtration experiments were also carried out using the other model foulants, including kaolin, NOM, and alginate. As a result of these tests, the slopes of the TMP curve (θ) were calculated. When the slopes were calculated, initial fouling at the beginning of each flux step was not considered to increase the reproducibility of the results. As indicated in Eq. (4), θ is proportional to J2. Accordingly, the intrinsic characteristics of foulants, ηαcf, should be determined by dividing θ by J2. The “normalized” fouling rate, θ/J2, is shown as a function of flux in Fig. 5. Although the concentrations were the same, the fouling propensities were different for different foulants. This is because those foulants
have different fouling potentials even at the same concentration. The order of fouling propensity is inversely reflected in the size of particles and hydrophobicity of organic matters [16,17]: θ/J2 (NOM) N θ/J2 (alginate) N θ/J2 (silica) N θ/J2 (kaolin). The kaolin, which is denoted as the black color, did not result significant membrane fouling. However, it is known that the kaolin may cause fouling in combination with other foulants. In the cases of the silica (red color) and NOM (green color), the fouling rates were different even at the same flux (the first and second runs at 50 LMH). This suggests that irreversible fouling occurs in the filtration of silica and NOM. The alginate (yellow color) initially resulted in rapid fouling, and then led to similar fouling rate regardless of applied flux. This is attributed to the initial adsorptive fouling by the interaction between alginate and membrane surface.
116
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
Fig. 2. A Matlab-based simulation program for MF systems.
Similar trends are observed at different concentrations, as illustrated in Figs. 6, 7, and 8. Except for alginate, θ/J2 increases with increasing flux. Nevertheless, θ/J2 is not strongly dependent on the foulant concentrations. According to Eq. (3), θ/J2 is proportional to cf, which may not be identical to cb. Based on the results, it is likely that cf does not increase even with increasing cb. As mentioned above, the degree of hysteresis in θ/J2 may be used to quantify the irreversible nature of fouling. For instance, NOM showed high degree of hysteresis at 2 mg/L but alginate showed low degree of hysteresis at 2 mg/L (Fig. 5). At 5 mg/L, both NOM and alginate exhibit relatively low degree of hysteresis (Figs. 6 and 7). This suggests that the
degrees of hysteresis may provide additional information on the characteristics of fouling at a given condition. Fig. 9 shows the dependence of θ/J2 on foulant concentration for the model foulants. The results confirm that θ/J2 does not significantly change with the concentration. Except for kaolin, the θ/J2 values range from 5 × 10− 9 to 10 × 10−9 bar/min (L/m2 h)2. It should be noted that fouling rate is not linearly proportional to the foulant concentration. There are several reasons: first, the foulants may be added in an excess amount and thus the saturation of foulant deposition occurs during the tests; second, adsorption of foulants may be limited by the membrane surface rather than by the foulant concentration. These results suggest that the fouling rate is more sensitive to flux than foulant concentration. If the fouling rate should be increased during the accelerated testing, it is better to adjust the flux than the foulant.
4.3. Simulation of continuous MF system using θ/J2
Fig. 3. Schematic diagram of experimental setup for MF tests: (a) multi-channel peristaltic pump, (b) pressure transducer, (c) hollow fiber, (d) magnet stirrer, (e) data logger and (f) desktop.
Although the focus of this research was experimental, we applied the simple model described in Fig. 1 to aid in interpreting the continuous operation of MF systems and to predict membrane performance over a longterm filtration. Fig. 10 shows the examples of simulation results using experimental θ/J2. Since the backwash interval for most pilot- or full-scale membrane systems for surface water treatment ranges from 15 min to 60 min, the simulation was performed using a backwash interval of 30 min. In our model, the effect of the backwash conditions is reflected by β in Eq. (6). Results showed that the TMP profiles can be simulated using the parameter (θ/J2) obtained from the accelerated testing. The increase in TMP within the backwash interval is attributed to θ/J2 but the overall increase in TMP is affected by β. In this simulation, β is assumed to be 0.9 but different β values would result in different results.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
117
Fig. 4. Changes in transmembrane pressure (TMP) as a function of filtration time: (A) 2 mg/L silica, (B) 5 mg/L silica, (C) 10 mg/L silica, and (D) 20 mg/L silica.
(1) A simple model based on pseudo-cake filtration model was applied to estimate the fouling parameters. The model could be successfully applied to interpret experimental data obtained using a laboratory-scale MF system.
(2) Although foulant concentration and flux were used to control the rate of fouling, the fouling rate was less sensitive to foulant concentration than to flux. This is because the effective foulant concentration approaching the membrane surface is different from the bulk concentration. (3) The normalized fouling rates, θ/J2, were difference for different foulants even at the same concentration. The alginate initially resulted in rapid fouling, and then led to similar fouling rate regardless of applied flux, suggesting the initial adsorptive fouling. (4) Using the θ/J2, which was estimated from the accelerated test, long-term operations of MF systems could be simulated. Nevertheless, additional approaches, including the consideration
Fig. 5. Effect of flux on membrane fouling rates under various foulant concentrations (2 mg/L).
Fig. 6. Effect of flux on membrane fouling rates under various foulant concentrations (5 mg/L).
Accordingly, a new technique to estimate β should also be developed for the long-term modeling of MF system with backwashing. 5. Conclusions In this work, an accelerated testing method was investigated to analyze fouling propensity of hollow fiber MF membranes. The following conclusions can be drawn:
118
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
Fig. 7. Effect of flux on membrane fouling rates under various foulant concentrations (10 mg/L).
of backwash effect, should be done to quantitatively predict the long-term performance of MF systems based on the results of the accelerated testing. Nomenclature ΔP trans-membrane pressure (bar) η fluid viscosity (Pa s) Rm membrane resistance (m−1) Rc cake resistance (m−1) J permeate flux (Lm−2 h−1) α specific cake resistance mc mass of cake deposited on the membrane (mg)
Fig. 8. Effect of flux on membrane fouling rates under various foulant concentrations (20 mg/L).
Am cf cb θ t β
membrane area (m2) effective foulant concentration (mg/L) bulk concentration of foulant (mg/L) rate of membrane fouling filtration time (h) constant to describe the effect of backwash
Acknowledgements This research was supported by LG Electronics, Advanced Research Institute.
Fig. 9. Effect of foulant concentration on membrane fouling rates for various model foulants: (A) kaolin, (B) silica, (C) NOM, and (D) alginate.
Y.-J. Kim et al. / Desalination 343 (2014) 113–119
119
References [1] Zeman, A.L. Zydney marcel dekker, Microfiltration and Ultrafiltration - principles and applications 1996. 648. [2] A.G. Fane, S. Chang, E. Chardon, Submerged hollow fibre membrane module—design options and operational considerations, Desalination 146 (2002) 231–236. [3] W. Gao, H. Liang, J. Ma, M. Han, Z.-l. Chen, Z.-s. Han, G.-b. Li, Membrane fouling control in ultrafiltration technology for drinking water production: a review, Desalination 272 (2011) 1–8. [4] T. Lebeau, C. Lelievre, H. Buisson, D. Cleret, L.W. Van der Venter, P. Cote, Immersed membrane filtration for the production of drinking water: combination with PAC for NOM and SOCs removal, Desalination 117 (1998) 219–231. [5] W. Neubrand, S. Vogler, M. Ernst, M. Jekel, Lab and pilot scale investigations on membrane fouling during the ultrafiltration of surface water, Desalination 250 (2010) 968–972. [6] 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. [7] B. Sarkar, A combined complete pore blocking and cake filtration model during ultrafiltration of polysaccharide in a batch cell, J. Food Eng. 116 (2013) 333–343. [8] Y.S. Polyakov, A.L. Zydney, Ultrafiltration membrane performance: effects of pore blockage/constriction, J. Membr. Sci. 434 (2013) 106–120. [9] D. Mosqueda-Jimenez, R. Narbaitz, T. Matsuura, Membrane fouling test: apparatus evaluation, J. Environ. Eng. 130 (2004) 90–99. [10] S.H. Mohammadian, D. Aït-Kadi, F. Routhier, Quantitative accelerated degradation testing: practical approaches, Reliab. Eng. Syst. Saf. 95 (2010) 149–159. [11] S.J. Bae, S.-J. Kim, J.I. Park, C.W. Park, J.-H. Lee, I. Song, N. Lee, K.-B. Kim, J.-Y. Park, Lifetime prediction of a polymer electrolyte membrane fuel cell via an accelerated startup–shutdown cycle test, Int. J. Hydrog. Energy 37 (2012) 9775–9781. [12] C. Regula, E. Carretier, Y. Wyart, M. Sergent, G. Gésan-Guiziou, D. Ferry, A. Vincent, D. Boudot, P. Moulin, Ageing of ultrafiltration membranes in contact with sodium hypochlorite and commercial oxidant: experimental designs as a new ageing protocol, Sep. Purif. Technol. 103 (2013) 119–138. [13] I. Levitsky, A. Duek, E. Arkhangelsky, D. Pinchev, T. Kadoshian, H. Shetrit, R. Naim, V. Gitis, Understanding the oxidative cleaning of UF membranes, J. Membr. Sci. 377 (2011) 206–213. [14] E. Zondervan, A. Zwijnenburg, B. Roffel, Statistical analysis of data from accelerated ageing tests of PES UF membranes, J. Membr. Sci. 300 (2007) 111–116. [15] H. Darcy, Les fontaines publiques de La Ville Dijon, 1856. [16] S. Lee, P.K. Park, J.H. Kim, K.M. Yeon, C.H. Lee, Analysis of filtration characteristics in submerged microfiltration for drinking water treatment, Water Res. 42 (2008) 3109–3121. [17] K.-J. Hwang, C.-Y. Liao, Effects of membrane morphology and operating conditions on microfiltration particle fouling, J. Taiwan Inst. Chem. Eng. 43 (2012) 46–52. Fig. 10. Model simulation using different θ/J2 values. Conditions: backwash interval, 30 min; β, 0.9; J, 50 L/m2 h. (A) θ/J2 = 5 × 108. (B) θ/J2 = 10 × 108.