Membrane fouling insight during reverse osmosis purification of pretreated olive mill wastewater

Separation and Purification Technology 168 (2016) 177–187

Contents lists available at ScienceDirect

Separation and Purification Technology journal homepage: www.elsevier.com/locate/seppur

Membrane fouling insight during reverse osmosis purification of pretreated olive mill wastewater J.M. Ochando-Pulido ⇑, M.D. Víctor-Ortega, A. Martínez-Ferez Chemical Engineering Department, University of Granada, 18071 Granada, Spain

a r t i c l e

i n f o

Article history: Received 9 January 2016 Received in revised form 10 May 2016 Accepted 26 May 2016 Available online 27 May 2016 Keywords: Olive mill wastewater Modeling Fouling Reverse osmosis Wastewater reclamation

a b s t r a c t The chronological sequence and the relative importance of the individual fouling mechanisms were investigated on a polymeric RO membrane used for final purification of secondary-treated olive mill wastewater (OMW2ST). An insight of membrane fouling under different operating conditions is reported. The goodness of the fitting (R2 ? 0.99) points for the formation of a cake or gel layer over the membrane as the dominant fouling mechanism in the transient and steady-state operation stages. A non-uniform cake was built-up, as shown in the HR-SEM analysis on the fouled membrane, principally formed by the organic matter (188.7 ± 37.9 mg L
1) and residual iron colloids carried by the OMW2ST feedstream. However, about 39.0% of the initial flux drop could be attributed to concentration polarization on the boundary region of the membrane. Moreover, it is relevant that the partial blocking model describes correctly the incipient fouling formation on the membrane, disregarding the operating conditions. This behavior was attributed to surface defects, more permeable zones of the membrane that may act as pores during the initial and intermediate operation stages, and in particular to the small organic particles (up to 63.3% between 3 and 30 kDa) remaining in the OMW2ST feedstream. Ó 2016 Elsevier B.V. All rights reserved.

1. Introduction Fouling issues represent the main handicap for the definite implementation of membrane processes. As a proof of this, over the last 5 years more than 3000 papers have been published in international scientific journals trying to shed light on this deleterious issue [1]. Despite this incredible effort made by a plethora of researchers, the reduction of membrane fouling remains still today as one of the main challenges of the broad applied membrane technology [2]. The reason is that fouling on membranes does not only lead to sensible capital losses, due to the need of a premature module substitution, but also gives rise to unexpected increases of investment costs during the development and design of membrane plants, as engineers tend to design membrane processes with an excessive oversized capacity to ensure the long-run performance [3]. Also, the operating costs are triggered during the operation of the membrane plant to maintain the target productivity, because the operating conditions have to be intensified to cope with the dynamic fouling build-up. Furthermore, this also obliges to stops for in-situ cleaning procedures, which hinder the cost-efficiency

⇑ Corresponding author. E-mail address: [email protected] (J.M. Ochando-Pulido). http://dx.doi.org/10.1016/j.seppur.2016.05.024 1383-5866/Ó 2016 Elsevier B.V. All rights reserved.

of the process both due to the frequent need for shut-downs in the continuous operation and the expenditure in cleaning reagents. In this regard, one of the most problematic applications of membrane technology is the case of wastewater treatment processes, in which membranes have gained significant use in the last few years. In particular, reverse osmosis (RO) membranes are able to ensure the compliance of the most stringent regulations, and have already been implemented in many wastewater treatment plants for the management of the industrial effluents of a wide variety of sectors, comprising stainless steel [4], energy cogeneration [5], nuclearpower [6], textile [7,8] and agro-food industries [9,10], among others. In these cases fouling carries a negative technical and economic impact on the process since the permeate stream, that is purified water with a quality grade compatible with irrigation use, has a limited economic value. If of irreversible nature, fouling reduces the membrane service lifetime drastically. In any case, fouling alters the selectivity of the membrane and depletes its productivity, making the integration of the membrane operation in wastewater treatment plants economically unfeasible [11–13]. Fouling is a complex phenomenon which involves different mechanisms comprising pore blocking, plugging or constriction, cake, gel and biofilm formation as well as cake-enhanced concentration polarization [11–13]. As the main factors that determine the extent of fouling build-up on membranes we can point the

178

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

membrane type, structure, active layer roughness and mean porosity, the hydrodynamic operating conditions inside the module and the characteristics of the effluent in contact with the membrane, namely the composition and concentration. Olive mill wastewater (OMW) is a highly polluted effluent byproduced in olive oil industries. Although typically circumscribed to the Mediterranean Basin, these industries are now expanding to other countries such as France, the USA, Australia and China, and thus the problem related to these effluents is not an issue of a particular region anymore; on the contrary, it is becoming a task of global concern. The difficulty in treating OMW resides in its high content in recalcitrant organic compounds, most of which are resistant to conventional and biological processes [14]. Thereby, many treatments have been investigated in the last decades [15–22], but their complexity or lack of cost-efficiency have hindered their implementation at real industrial scale, thus the problem with regard to OMW is still far from being resolved. In the last years, as a result of the advances in new membrane materials and module configurations capable of offering enhanced technical and economical performances, the use of membrane technology has been pointed as a feasible solution for the purification of these problematic effluents, in conjunction with other processes, usually in the form of integrated pretreatments upstream the membrane operation [13,23–25]. In this regard, OMW contain high concentrations of a wide range of solutes in the shape of suspended solids and colloidal particles which are all very likely to cause membrane fouling, such as organic pollutants, as well as inorganic matter that may also lead to deleterious scaling problems [26]. The critical and threshold flux theories, which sustain that there is boundary that divides a nil or low membrane fouling operation framework from a high and exponential fouling build-up region, have been highlighted to be very useful as a tool for the design and control of membrane plants [11–13]. In particular, this has been underlined by Stoller and Ochando for the treatment of OMW with microfiltration (UF), ultrafiltration (UF), nanofiltration (NF) and RO membranes [13,23,26,27]. However, despite the usefulness of the critical and threshold flux theories, they do not describe the type of fouling occurring on the membranes. Fouling mechanisms are also important to fully understand what is happening between the membrane and the effluent, in order to take the adequate decisions with respect to the design of the membrane plant, adoption of properly-tailored pretreatment process and set-up of optimized operating conditions. In the present paper, the performance of a RO membrane used for final purification of secondary-treated OMW, thoroughly described in previous works by the Authors [9,10,22], is examined. Inhibition or minimization of the membrane fouling build-up common in all membrane facilities is the key to make the process feasible. With this purpose, the dynamic behavior of the RO membrane under different operating conditions was studied by fitting to the fouling laws, comprising the different types of pore plugging, blocking and cake formation. The chronological sequence and the relative importance of the individual fouling mechanisms were investigated. The insight of the membrane fouling mechanisms is imperative to achieve the proper steady-state control of the process to ensure stable performance of the RO operation. To the Authors’ knowledge, no previous work on the analysis of the dynamic fouling mechanisms occurring during membrane filtration of this kind of wastewater can be found in the scientific literature up to the date. 2. Experimental 2.1. Analytical methods The analytical procedures were performed by using reagents with analytical grade purity (over 99%) and carried out in triplicate.

Chemical oxygen demand (COD), total suspended solids (TSS), total phenolic compounds (TPh), total iron, electroconductivity (EC) and pH measurements were performed following standard methods [28]. Ionic concentrations were analyzed with a Dionex DX-120 ion chromatograph, as described in previous work [9,10]. Microphotographs of the active layer of the membrane were performed with a high resolution scanning electron (HR-SEM) microscope (Carl Zeiss SMT model). Small coupons (2 cm  2 cm) of the virgin RO membrane and of the fouled membrane after the RO experiments were cut at the center and extreme zones to carry out the HR-SEM analysis, together with complementary elemental microanalysis of the virgin and fouled membrane layers. 2.2. Particle size distribution and molecular weight cut off Particle size distribution (PSD) was examined in the OMW2ST feedstream prior to the RO membrane operation. The PSD was carried out with a liquid sampler (model LS-200, Beckman Coulter) by taking 1 mL OMW2ST samples for their analysis. Complementary to this, molecular weight cut off (MWCO) of the organic matter remaining in the secondary-treated OMW stream (OMW2ST) was also examined, following a procedure already tuned in previous work by the Authors [29]. The method performed consisted in filtering 100 mL samples of OMW2ST through 0.45 lm nitrate cellulose filters (Sartorius). After this, 15 mL of the filtered samples were poured into Falcon tubes provided with membranes presenting different MWCO (ranging from 3 to 100 kDa). The Falcon tubes were subsequently centrifuged at 4000 rpm during 3 min, after which the COD of the centrifugedfiltered samples was analyzed. 2.3. Effluent characteristics OMW samples were collected from various olive oil mills located in the provinces of Jaén and Granada (Spain). The samples were gathered during the olive oil production period, after which they were promptly characterized in the laboratory. After this, OMW samples were conducted to a secondary treatment optimized by the Authors in former work, where the full details can be found [9,22], which was finally transferred to an industrial scale. The effluent exiting the secondary treatment (OMW2ST) presents the characteristics briefly reported in Table S1 in the Supplementary Material (SM). The ulterior objective was the final purification of the OMW2ST effluent by means of RO, with the goal of complying with the quality standards to reuse it whether for irrigation purposes or in the olive oil production process, as water for cleaning procedures or in the centrifuges [9]. 2.4. Membrane plant The membrane plant (Prozesstechnik GmbH, Basel, Switzerland) used for final purification of OMW2ST is shown in Fig. 1. The bench-scale plant was equipped with a non-stirred doublewalled tank (5 L maximum volume), where the effluent was fed (2 L), while a diaphragm pump (Hydra-Cell model D-03) could be set to drive the OMW2ST stream to a plate-and-frame membrane module, with dimensions 3.9 cm height  33.5 cm length  14.2 cm width, provided with a flat-sheet RO membrane. The tested RO membrane was a commercial thin-film composite (TFC) one of polyamide/polysulfone, provided by GE Water and Process Technologies (SC model). The characteristics of this membrane are reported in Table S2 in the SM. After being received from the manufacturer, the membrane coupons were equilibrated with MilliQÒ water until a constant and stable permeate flux was observed; then the hydraulic permeability of the membrane was

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

Fig. 1. Bench-scale membrane plant.

179

in contact with the membrane, then those solutes can likely enter the membrane pores and cause fouling issues due to pore constriction. Otherwise, if the particles in the effluent present a bigger size than the pores of the membrane, these molecules can whether cause pore blocking or lead to the formation of a cake over the membrane layer. In case of molecules of a size in the near range of the membrane pores diameter, plugging or sealing of the pores can occur [31,32]. To sum up, the blocking laws can be grouped in the following fouling mechanisms: (i) complete pore blocking or pore sealing, (ii) intermediate pore blocking, (iii) standard pore blocking or pore filling (iii), and (iv) cake or gel layer formation (see Fig. 2). Complete pore blocking assumes that each molecule only fouls the membrane by sealing one pore each, such that no solute particle deposits over another (A caption). Otherwise, intermediate pore blocking also takes into account the possibility that solute molecules may settle over others, not just only blocking pores (B caption). On the other hand, standard blocking considers that solutes smaller than the diameter of the pores (Dp) can enter them and thus fill the pores and lead to their constriction, reducing the volume inside the pore proportionally to the number of particles deposited into it (C caption). Finally, cake or gel layer formation mechanism, which occurs normally for particles of bigger size (dp) than the membrane Dp, supposes that the particles become deposited over the membrane surface without blocking any pore, whether because the membrane is dense and presents strictly no pores, or because the pores are already covered by other particles and thus are not susceptible to being blocked anymore (D caption). The equations that correspond to each of these fouling mechanisms are hereafter reported [31]:

determined by measuring the pure water flux over its admissible pressure range, both tasks performed following the procedures described elsewhere [9,10]. The full characteristics of the membrane plant can be found in previous work by the Authors [9,10]. To sum up the main features, the system operating pressure was monitored by a digital pressure gauge (Endress + Hauser, Ceraphant T PTC31) and could be finely tuned (PTM set point ± 0.01 bar) by means of a spring loaded pressure-regulating valve located on the concentrate outlet (SSR4512MM-SP, Swagelok); the flowrate (0.1 L h
1 precision) could be set by means of a feed flow rate valve; whereas the operating temperature was automatically regulated (Tset point ± 0.1 °C) through a proportional-integral-derivative (PID) electronic temperature controller (Yokogawa model UT100), which acted on a magnetic valve in the refrigeration loop so as to recirculate cooling water from a chiller (PolyScience model 7306) into the doublewalled tank jacket. During the purification of OMW2ST, semicontinuous operation was adopted, which implied recycling continuously the concentrate stream back to the feedwater tank, whereas steadily collecting the permeate stream, which was proportionally replaced by fresh OMW2ST. Samples of the permeate and concentrate streams were periodically taken during the RO experiments for their analysis. The effect of the operating parameters - net pressure (15–35 bar), temperature (15–30 °C) and feed flowrate (3–6 L min
1) - on the membrane fouling mechanisms were investigated. Fluctuations in the feedstock composition made the evaluation of the membrane performance more difficult, therefore experiments were replicated twice. The membrane permeability was completely recovered (by checking the pure water permeability of the OMWST-fouled membrane after the cleaning assay K0w vs. that of the virgin RO membrane Kw) for the successive RO experiments by carrying out an in-situ cleaning protocol with 0.1–0.2% w/v NaOH, sodium dodecyl sulfate (SDS) and citric acid solutions (Panreac S.A.), following the procedure already proposed by the Authors [30].

where Jp0 makes reference to the initial permeate flux and K is a constant with suffix i that indicates the type of fouling mechanism: complete blocking (Kb), intermediate blocking (Ki), standard blocking (Ks) and cake formation (Kc). Subsequently, it was observed that by differentiation of the filtered volume data to obtain the permeate flux, linear expressions could be obtained for the fouling mechanisms. As a result, a linear dependence was observed when the experimental permeate flux data were represented in the form t/V vs. t for standard blocking. Analogously, a linear trend was confirmed in the case of cake or gel layer formation on the membrane when the experimental t/V filtration data were represented vs. the filtrated volume V. In a same way, it was observed that complete blocking was the predominant form of membrane fouling if the permeate flux Jp vs. the filtrated volume V is linear (i0 ), whereas if the inverse of the permeate flux 1/Jp vs. filtration time t presents a linear behavior the controlling mechanism is likely to be intermediate blocking (ii0 ) [31]:

2.5. Fouling mechanisms

J p ¼ J p0
K b  V

ði Þ

An insight into the fouling mechanisms, as a result of the interaction between the membrane and the particles carried by the effluent in contact with it, is necessary to adopt properly-tailored pretreatment process and optimized operating conditions. As a general rule, if the selected membrane presents pores with a mean diameter larger than the size of the particles in the effluent

1 1 ¼ þ Ki  t J p J p0

ðii Þ

K b  V ¼ J p0  ð1
e
K b t Þ

ðiÞ

K i  V ¼ lnð1 þ K i  J p0  tÞ

ðiiÞ

Ks  t t 1 ¼
2 V J p0

ðiiiÞ

Kc  V ¼

2t 2
V J p0

ðivÞ

0

0

As stated by Wang and Tarabara [33,34], the main advantage of using the linear expressions of the blocking laws is the fact that they permit quick and easy identification of the predominant

180

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

Fig. 2. Scheme of membrane fouling mechanisms: (A) pore sealing or complete pore blocking, (B) intermediate pore blocking, (C) standard blocking or pore filling/constriction, and (D) cake or gel layer formation [32].

fouling mechanism, such that no additional data processing is needed, but only linear least squares fitting, with an easily determinable correlation coefficient (R2).

2.6. Experimental RO flux data fitting procedure The experimental data obtained during the OMW2ST purification RO experiments under the different operating conditions assayed were fitted to the fouling models following the procedure described by Wang and Tarabara [33]. The goal was to identify and address the relative impact and the chronological sequence of each of the possible fouling mechanisms. Least squares regression was used to determine the best fitting of the dynamic RO runs flux data to the linear expressions of the fouling mechanisms models, on the basis of the coefficient of determination (R2). With this purpose, the method consisted firstly in the identification of a ‘core’ portion of the corresponding permeate flux-time data profile which fitted accurately (R2 ? 0.99) any of the patterns of the fouling mechanisms. Subsequently, the flux data core was extended to the adjacent flux data points, evaluating if each additional flux data value effectively belonged to the 95% confidence interval based on the least squares regression of the expanded ‘core’ to the linear equations of the fouling models [33–35]. Hence, if the following data point was confirmed to be within the confidence interval, the point was definitely included in the core, then further proceeding with the evaluation of the next adjacent flux data value. The procedure was stopped when a data point notbelonging to the core at the 95% confidence interval was found [33–35].

Within this framework, a clear identification and differentiation of the possible membrane fouling mechanisms is imperative to achieve the proper steady-state control of the process and ensure a stable performance of the membrane unit. Hence, the operation of the selected TFC RO membrane during the final purification of OMW2ST was examined in order to assess the fouling mechanism or mechanisms occurring during the dynamic semicontinuous operation. Beforehand, the intrinsic resistance of the RO membrane (Rm) was estimated on the basis of pure water permeate flux (Jw) measurements within the admissible operating pressure range of the membrane (Fig. 3). It can be observed that the Rm sensibly decreased with an increment in the operating temperature, that is, 7.6% by increasing the temperature from 15 to 22 °C (2.78  1014 m
1 for the former vs. 2.57  1014 m
1 for the latter temperature), and to 30.7% upon an increment from 22 to 30 °C (1.78  1014 m
1). This was attributed to the increase of the water diffusivity through the membrane thickness due to the lower viscosity of the solvent at higher temperature conditions, in sum to swelling changes in the physical properties of the membrane, leading to higher permeability and thus a reduction of the mass transfer resistance opposed by the membrane [36]. Moreover, the p of the feedstream, calculated using Van’t Hoff ’s equation, was found to be in the range 1.45–1.52 bar for temperature values of 15–30 °C, respectively. The measured pure water permeability coefficient of the RO membrane (Kw) was equal to 1.41, whereas lower values of the permeability coefficient (K) were confirmed for the OMW2ST stream (0.86 L h
1 m
2 bar
1). The difference observed between Kw and K coefficients, 39.0%, is attributed to concentration polarization taking place on the boundary region of the RO membrane, as the concentration of particles on

3. Results and discussion In the present work, the performance of a RO membrane used for final purification of secondary-treated OMW to comply with the water quality standards to reuse it whether for irrigation purposes or internally in the olive oil production process, thoroughly described in previous works by the Authors [9,10,22], was examined. For this purpose, inhibition or at least minimization of the membrane fouling build-up which irretrievably occurs in all membrane facilities is key to ensure the feasibility of the process. In order to take the adequate decisions with respect to the membrane plant design, the adoption of properly-tailored pretreatment process, optimized operating conditions and cleaning protocols, the insight of the mechanisms by which fouling occurs between the membrane and the foulants in the effluent are necessary. Fouling problems have conditioned the reliability of membrane processes. This feeling still persists at industrial scale and it is thereby necessary to understand and validate membrane fouling phenomena theories and their description in order to succeed in filling the knowledge gap.

Fig. 3. Permeability measurement on the selected RO membrane: pure water (Kw) 15 °C ( ), 22 °C ( ) and 30 °C ( ); pretreated OMW2ST (K) 22 °C ( ).

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

the membrane surface reaches its maximum value after a short initial filtration [37]. This is in good agreement with the high roughness of the RO membrane, as shown in the HR-SEM microphotographs of the active layer (Fig. 4). Some authors have pointed for a relationship between the membrane permeability and the membrane layer morphology, stating that membranes that present major surface roughness can offer higher permeability values as a result of their higher effective contact surface [38]. Subsequently, the dynamic permeate flux-time profiles of the RO membrane were fitted to the different types of fouling mechanisms comprising (i) complete pore blocking or pore sealing, (ii) intermediate pore blocking, (iii) standard pore blocking or pore filling (iii), and (iv) cake or gel layer formation (described in Section 2.5). The results of the fitting of the experimental permeate flux profiles data of the final OMW2ST RO purification upon the different operating conditions examined - operating pressure (15–35 bar), temperature (15–30 °C) and feed flowrate (3–6 L/min) - to the fouling mechanisms are hereafter reported in Fig. 5. The adjustment of the dynamic OMW2ST RO permeate flux profiles to the complete and standard blocking models was confirmed not to lead to satisfactory results (R2 < 0.95 for both cases). Given that RO membranes do not have definable pores and generally the dominant separation mechanism is based on solutiondiffusion, this was somewhat expectable. On the other hand, good fitting results were obtained for both intermediate blocking and cake formation fouling build-up mechanisms. The adjustment of the experimental RO dynamic flux data to both intermediate and cake models reported in Fig. 5 are given in the form 1/Jp vs. t for the intermediate blocking mechanism (left figures) whereas t/V vs. V for the cake or gel layer formation (right figures). Linearity of both fittings pin-points for both intermediate blocking and cake formation fouling mechanisms effectively occurring on the RO membrane during OMW2ST purification. Furthermore, the values of the fouling mechanism constants (Ki and Kc), calculated from the correlation of the experimental dynamic RO flux data to both fouling models, are hereafter reported in Table 1. As a general pattern, it was observed that the cake or gel layer formation is the controlling fouling build-up mechanism during the final stage of all the membrane experiments disregarding the operating conditions set. The goodness of the fitting (R2 ? 0.99) indicates that the formation of a cake or gel layer over the membrane is the dominant fouling mechanism in the intermediate and steady-state stages of the OMW2ST RO purification trials (Fig. 5).

Fig. 4. HR-SEM image (10x) of the active layer of the virgin RO membrane.

181

This was lately confirmed in the HR-SEM analysis performed on the fouled RO membrane after OMW2ST purification (Fig. 6). A nonuniform cake was formed on the skin layer of the selected membrane after the RO purification of OMW2ST [22,29]. The nonuniformity of the formed cake is due in part to the polymeric net spacer used to promote the turbulence over the membrane surface to reduce polarization effects and fouling build-up, but also to the roughness of the membrane enhancing the deposition of particles on the skin layer. This fouling layer represents the main additional resistance to that intrinsic of the membrane. The formed cake layer was principally composed by the significant organic matter load carried by the OMW2ST feedstream (188.7 ± 37.9 mg L
1, Table S1 in SM). Moreover, the microanalyses performed on the fouled RO membrane layer (Fig. 6) showed peaks of iron, which confirm the presence of inorganic fouling in the form of residual colloidal iron on the fouled surface of the RO membrane derived from the secondary treatment applied to the raw OMW2 stream [22,29]. Also, it is important to underline the role of certain ionic species such as calcium ions, which promote the aggregation of the organic matter by intra and intermolecular bridge formation mechanisms [38]. However, a non-linearity of the dynamic permeate flux data curves to the cake formation Eq. (iv) was noted during the early stages of the OMW2ST RO purification operation in all the experiments. This raised the question of the implication of another different fouling mechanism during the initial periods of the OMW2ST RO purification, responsible for the incipient permeate flux decay once the concentration polarization stablished, prior to the formation of a cake or gel layer over the membrane. Some authors have underlined similar results when studying the flux behavior of NF and RO membranes, without a clear identification of the responsible fouling mechanisms within the initial stages, whereas others have pointed to the blocking laws as possible responsible of the initial fouling typology in saline-rejection membranes (NF and RO) [39–41]. In the present work, the initial stages of the dynamic permeate flux profiles were found to fit accurately (R2  0.99) the intermediate pore blocking mechanism (Eq. (ii0 )) for all the RO experimental runs. In the intermediate or partial pore blocking mechanism, molecules of a size (dp) similar to that of the pores of the membrane (Dp) can deposit over the pores or over the membrane surface, such that the pores can be partially blocked at their entrance. Thus, the intermediate blocking model takes into account the possibility that the solutes may deposit over others formerly settled, assuming that each position of the membrane has the same probability of being covered, and that the probability that one particular particle settles on one place is equal to the ratio between the free spaces of the membrane layer and the already fouled ones [41–43]. Furthermore, it was observed in the fitting of the dynamic permeate flux data of the OMW2ST RO purification that there are intermediate stages of transition from one mechanism to the other. In this intermediate transition stages, the permeate flux data are concurrently consistent with both the partial or intermediate blocking and cake or gel layer formation models, being the effects of both mechanisms proportional. The impacts and chronological sequence of the fouling mechanisms on the selected RO membrane as a function of the operating conditions are reported in Fig. 7. Given the fact that RO membranes are considered to be homogeneous surfaces, it is relevant that the intermediate blocking model describes correctly the incipient fouling formation on the selected RO membrane based on the fitting of the experimental permeate flux data, disregarding the operating conditions. This was also pin-pointed by other researchers in former works [33,34,41,42].

182

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

Fig. 5. Fitting of the permeate flux data profiles to the intermediate (left figures) and cake-formation (right figures) fouling mechanisms. A caption: 35 bar, 2.55 m s
1, 22 °C; B caption: 25 bar, 2.55 m s
1, 22 °C; C caption: 15 bar, 2.55 m s
1, 22 °C; D caption: 25 bar, 5.09 m s
1, 15 °C; E caption: 25 bar, 5.09 m s
1, 22 °C; F caption: 25 bar, 5.09 m s
1, 30 °C.

This behavior may have a physical explanation in the fact of the existence of permeation variability on the active layer of the RO membrane, as a result of surface defects, also called pore defects. These more permeable zones of the membrane may therefore act as pores during the initial and intermediate operation stages. Due to the higher permeability of these regions of the membrane surface, they are more prone to colloidal matter deposition, thus promoting the possibility of intermediate blocking phenomena [42–45]. The concentration and settlement of organic and inorganic colloidal matter on the active layer of polyamide membranes is typically favored by their rough surface and high tendency to

permeation as a result of their high effective contact surface [38,46,47]. The reduction of the hydrodynamic shear force in the boundary regions of the valleys of this ridge-and-valley structured membrane and the higher local flux thus enhances fouling in these areas of the membrane just from the beginning of the operation. Particle size distribution (PSD) performed on the OMW2ST effluent reveal a high concentration of sub-micron (dp < 2 lm) particles (75.1%, that is, a number of particles np equal to 17519), whereas 24.9% (np equal to 5830) in the supra-micron range (2 lm < dp < 50 lm). Moreover, the analysis of the molecular weight cut off (MWCO) of the organic solutes remaining in

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

183

Fig. 5 (continued)

Table 1 Calculated values for intermediate or partial blocking (Ki) and cake or gel layer fouling constants (Kc) as a function of the RO operating conditions. Operating conditions 22 °C, 22 °C, 22 °C, 15 °C, 22 °C, 30 °C,

35 bar, 25 bar, 15 bar, 25 bar, 25 bar, 25 bar,

3 L/min 3 L/min 3 L/min 6 L/min 6 L/min 6 L/min

Ki, m
1
5

9.0  10 5.0  10
5 6.7  10
5 1.8  10
5 2.8  10
5 4.2  10
5

Kc, s m
2 2.5  10
5 1.4  10
5 2.3  10
5 4.5  10
6 5.5  10
6 3.4  10
6

Ki: intermediate fouling constant; Kc: cake or gel layer fouling constant.

OMW2ST effluent, reported in Fig. 8, highlight there is a very significant presence of low molecular weight colloidal organic particles (MWCO between 3 and 30 kDa, up to 63.3%). In particular, the concentration in the OMW2ST effluent stream of organic particles presenting a size below 3 kDa was found to be around 32.5%. OMW2 is related to humic compounds because it is dark colored,

contains phenolic compounds and shares some of the properties of humic substances. In particular, low molecular weight phenolic compounds have molecular weights mainly in the range 200–500 Da [48,49]. This fact is very relevant for the membrane operation, as these dissolved organic pollutants and small colloids are very prone to cause severe membrane fouling by getting trapped in the valleys of the ridge-and-valley structured active layer and in the membrane surface voids and defects, and also in the nodules of the spacers and in the membrane support [45,46]. Small particles may lead to blocking and plugging of the pores of the membrane, given the fact that pore blocking and plugging may be statistically significant when the size of the particles is in the near range of the membrane mean pore diameter [46–48]. Winfield indicated that, for RO and NF membranes, solid particles greater in size than 5 lm play a minor role in determining the rate of membrane fouling than sub-micron particles [46]. Subsequently, after the membrane is colmated no more surface defects would be available for partial blocking, and therefore only the cake formation controls the fouling formation mechanism on

Fig. 6. HR-SEM microphotograph (A caption: 10x) and microanalysis (B caption) of the active layer of the fouled RO membrane.

184

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

Fig. 7. Impacts and chronological sequence of fouling mechanisms (j = intermediate blocking, = cake layer) on the selected RO membrane as a function of the op. conditions: pressure (15–35 bar), crossflow velocity (2.55–5.09 m s
1) and temperature (15–30 °C).

Fig. 8. Organic pollutants (measured as COD) molecular weight cut off (MWCO).

the membrane [33,34]. This is consistent with the higher rejection of organic and inorganic matter registered during the initial stages of the OMW2ST purification operation with the selected RO membrane (98.8–100% for the first 750–1000 mL volume vs. 95.9–96.5% in the steady-state RO membrane performance). Other authors have reported similar results in the short run operation of NF and RO membranes, just from the very first moment that colloidal particles were introduced in purely electrolytic feedstreams [33,34,50–52]. A focus on Table 1, where the values of the intermediate fouling (Ki) and cake or gel layer fouling constants (Kc) are reported, as

well as on Figs. 5 and 7, permits a further analysis of the impacts and chronological sequence of the fouling mechanisms on the selected RO membrane as a function of the operating conditions. Upon the maximum applied operating pressure (35 bar), the highest values of both Ki and Kc were observed. At this pressure, the value of Ki was found to be 25.6–44.5% higher than at 15–25 bar, whereas 8.0–44.0% higher for Kc, respectively (Table 1). Also, it was noted that the formation of the gel layer or cake began earlier at the maximum operating pressure (Fig. 7), due to the major transport of solutes to the membrane surface leading to a rapid colmatation of the surface defects of the membrane, quickly leading to the build-up of the cake. On the other hand, at the medium operating pressure tested (25 bar), it seems to exist an equilibrium between the double electric layers (membrane-colloid) repulsion and the hydrodynamic force (permeation drag) resulting from the convective transport towards the membrane. Thus, lower fouling by both intermediate blocking (Ki) and cake formation (Kc) mechanisms was registered at this operating pressure (Table 1), and the formation of the cake layer was observed to develop at a later period (Fig. 7). Otherwise, minor values of both Ki and Kc constants were registered when the crossflow velocity was doubled (2.55–5.09 m s
1), that is, 44.0% lower for Ki whereas 60.7% lower for Kc (Table 1). The enhanced turbulence over the membrane, and therefore major shear force, hinders the partial blocking of the membrane surface defects, and the transition from this fouling mechanism to the cake formation is narrowed (Fig. 7). Hence, the simple deposition of solutes over the membrane layer seems to be controlling, but in

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

a minor degree as a result of the sweeping effect of the higher turbulence. In fact, in the experimental RO runs upon the highest crossflow, the value of the gel layer or cake formation constant Kc initially experimented a higher value that afterwards became reduced until a stable value was attained in the steady-state. This points to the fact of a more dense and localized initial settlement of foulants on the membrane surface, due to the membrane roughness, that loosens during the operation time as a result of the effective turbulence (NReynolds in the range 1.3  104–2.6  104) promoted in the membrane module. Regarding the effect of the operating temperature, it was noted that the impact of the intermediate or partial blocking mechanism was reduced between 35.7 and 57.1% (Ki) and occurred for a shorter period (Fig. 7) at the lowest temperature if compared to the higher ones examined (22 and 30 °C, respectively). This was attributed to the higher hydrodynamic diameters of the organic pollutants upon decreasing temperature, which enhances their aggregation into sub-micron and supra-micron aggregates as a result of the minor solubility, as reported by Jin et al. [36,50]. Analogously, the fact of a lower mean diameter of the fouling particles upon major temperature (30 °C) favors their diffusivity across the membrane, and also the probability of intermediate or partial blocking of the surface or pore defects of the membrane. This is reflected in both the higher Ki values observed (Table 1), as well as the longer filtration period consistent with this fouling mechanism (Fig. 7). As pointed by Stoller [13], the higher the number of particles in the feedstream presenting a modal diameter (dp) in a range close to that of the membrane layer defects (Dp), that is, in the near range between 10 < dp/Dp < 10, the higher the tendency of quick and hardly-reversible membrane fouling. Furthermore, during the transient and steady-state filtration period, in which the formation of a cake or gel layer is the controlling fouling mechanism, a lower value of Kc was found upon higher operating temperature (Table 1). At major temperature, the lower mean size and major solubility of the organic foulants lead to a more loosened and porous cake, opposing lower filtration resistance. On another hand, minor formation of a cake or gel layer was noted at the lowest temperature examined (15 °C). As a result of their major size at this temperature value, fouling particles are more easily swept by the turbulence promoted in the RO membrane module [36,50]. These results are supported by the normalized fouling build-up rate values D(Rf/Rm)/Dt (min
1) reported in Table 2, calculated by means of the resistances-in-series model thoroughly described elsewhere [53,54].

Table 2 Results of the experimental RO runs for OMW2ST purification upon the different operating conditions studied for the selected membrane. Operating conditions

Jp0, L h
1 m
2

Jpss, L h
1 m
2

Rf, m
1

D(Rf/Rm)/Dt, h
1

22 °C, 35 bar, 3 L/min 22 °C, 25 bar, 3 L/min 22 °C, 15 bar, 3 L/min 15 °C, 25 bar, 6 L/min 22 °C, 25 bar, 6 L/min 30 °C, 25 bar, 6 L/min

32.1

n.o.

1.29  1014

64.2  10
2

21.8

15.6

1.82  1014

12.0  10
2

15.7

12.4

9.00  1013

14.1  10
2

18.7

16.5

5.25  1013

3.3  10
2

24.6

21.1

5.39  1013

7.6  10
2

35.6

31.8

3.76  1013

7.0  10
2

J0: initial permeate flux; Jss: steady-state permeate flux; Rf: fouling resistance; n.o.: not observed.

185

As it can be seen, similar fouling build-up rates were attained at operating pressures ranging from 15 to 25 bar (12.0  10
2– 14.1  10
2 h
1), but increased in the order of 5 times (64.2  10
2 h
1) upon incrementing the PTM up to 35 bar. Upon an increase of the flowrate, the fouling build-up rate became reduced by more than 45.5%. Otherwise, similar fouling build-up was noticed upon high temperatures, exhibiting fouling build-up rate values ranging between 7.0  10
2 h
1 and 7.6  10
2 h
1 for 22–30 °C, whereas strikingly minor fouling resistance was found at the lowest temperature (3.3  10
2 h
1, that is, 52.5–56.2% fouling build-up rate reduction). Although it is commonly assumed that for NF and RO membranes cake filtration is the controlling mechanism of colloidal fouling, the results of this research work on RO membrane purification of OMW2ST indicate that a certain extent of the fouling attained on the selected membrane cannot be attributed to cake formation. Other authors have also warned of this fact [33,34,39– 41,50–56]. As stated by Wang and Tarabara [33], given the fact that colloidal deposition in the early stages of filtration is the major cause of irreversible membrane fouling, the importance of more detailed understanding of the colloid-membrane interactions susceptible of being interpreted as partial blocking of surface or pore defects is further necessary. 4. Conclusions The chronological sequence and the relative importance of the individual fouling mechanisms were investigated on a RO membrane used for final purification of secondary-treated OMW (OMW2ST). To the Authors’ knowledge, no previous work on the analysis of the dynamic fouling mechanisms occurring during membrane filtration of this kind of wastewater can be found in the scientific literature up to the date. The goodness of the fitting (R2 ? 0.99) indicates that the formation of a cake layer over the membrane is the dominant fouling mechanism in the transient and steady-state operation stages, confirmed by HR-SEM analysis on the fouled membrane: a nonuniform cake was formed, principally by the organic matter and residual iron colloids carried by the feedstream. However, about 39.0% of the initial flux drop could be attributed to concentration polarization on the boundary region of the membrane. Moreover, it is relevant that the partial blocking model describes correctly the incipient fouling formation on the membrane, disregarding the operating conditions. This behavior was attributed to surface defects, more permeable zones of the membrane that may act as pores during the initial and intermediate operation stages. The concentration in the OMW2ST effluent stream of organic particles presenting a size below 3 kDa was found to be around 32.5%. In particular, low molecular weight phenolic compounds have molecular weights mainly in the range 200–500 Da. Upon the maximum applied operating pressure (35 bar) the value of Ki was found to be 25.6–44.5% higher than at 15–25 bar, whereas 8.0–44.0% higher for Kc, respectively, and the formation of the cake began earlier. On the other hand, at medium operating pressure (25 bar) it seemed to exist an equilibrium between the membrane-colloid repulsion and the permeation drag, thus lower fouling by both intermediate blocking (Ki) and cake formation (Kc) mechanisms took place, and the formation of the cake layer started to develop at a later period. Moreover, minor values of both Ki and Kc constants were registered when the crossflow velocity was doubled (2.55–5.09 m s
1), that is, 44.0% lower the former whereas 60.7% lower for the latter. The enhanced turbulence over the membrane, therefore major shear force, hinders the partial blocking of the membrane surface defects. Regarding the effect of the operating temperature, the impact of the partial blocking mechanism was reduced between

186

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187

35.7 and 57.1% (Ki) and occurred for a shorter period at the lowest temperature if compared to the higher ones examined, attributed to the higher hydrodynamic diameters of the organic pollutants upon decreasing temperature. On another hand, minor build-up of the cake layer was also noted at this temperature, as a result of the major size of the fouling particles, more easily swept by the turbulence promoted in the RO membrane module. The insight of the membrane fouling mechanisms, to attain inhibition or minimization of the membrane fouling common in all membrane facilities, is imperative to achieve the proper steady-state control of the process to ensure stable performance of the operation and make the process feasible. Acknowledgments The Spanish Ministry of Science and Innovation is also gratefully acknowledged for having funded the projects CTQ200766178 and CTQ2010-21411, as well as the University of Granada. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.seppur.2016.05. 024. References [1] Data Taken from Scopus, Aug 2015, . [2] R.W. Baker, Membrane Technology and Applications, John Wiley & Sons Ltd., England, 2004. [3] United States, Environmental Protection Agency, Office of Water: Membrane Filtration Guidance Manual, EPA 815-R-06-009, 21–23, 2005. [4] J. Lee, T. Kwon, I. Moon, Performance of polyamide reverse osmosis membranes for steel wastewater reuse, Desalination 189 (2006) 309–322. [5] A.N. Samodurov, S.E. Lysenko, S.L. Gromov, A.A. Panteleev, E.B. Fedoseeva, The use of reverse osmosis technology for water treatment in power engineering, Therm. Eng. 53 (6) (2006) 439–443. [6] V.N. Epimakhov, M.S. Oleinik, L.N. Moskvin, Reverse osmosis filtration based water treatment and special water purification for nuclear power systems, Atom. Energy 96 (4) (2004) 234–240. [7] T. Srisukphun, C. Chiemchaisri, T. Urase, K. Yamamoto, Foulant interaction and RO productivity in textile wastewater reclamation plan, Desalination 250 (2010) 845–849. [8] M. Liu, Z. Lü, Z. Chen, S. Yu, C. Gao, Comparison of reverse osmosis and nanofiltration membranes in the treatment of biologically treated textile effluent for water reuse, Desalination 281 (2011) 372–378. [9] J.M. Ochando-Pulido, G. Hodaifa, S. Rodriguez-Vives, A. Martinez-Ferez, Impacts of operating conditions on reverse osmosis performance of pretreated olive mill wastewater, Water Res. 46 (15) (2012) 4621–4632. [10] J.M. Ochando-Pulido, S. Rodriguez-Vives, A. Martinez-Ferez, The effect of permeate recirculation on the depuration of pretreated olive mill wastewater through reverse osmosis membranes, Desalination 286 (2012) 145–154. [11] R.W. Field, G.K. Pearce, Critical, sustainable and threshold fluxes for membrane filtration with water industry applications, Adv. Colloid Interface Sci. 164 (2011) 38–44. [12] P. Le-Clech, V. Chen, T.A.G. Fane, Fouling in membrane bioreactors used in wastewater treatment, J. Membr. Sci. 284 (1–2) (2006) 17–53. [13] M. Stoller, Effective fouling inhibition by critical flux based optimization methods on a NF membrane module for olive mill wastewater treatment, Chem. Eng. J. 168 (2011) 1140–1148. [14] P. Paraskeva, E. Diamadopoulos, Technologies for olive mill wastewater (OMW) treatment: a review, J. Chem. Technol. Biotechnol. 81 (2006) 475– 1485. [15] M. Annesini, F. Gironi, Olive oil mill effluent: ageing effects on evaporation behavior, Water Res. 25 (1991) 1157–1960. [16] E.S. Aktas, S. Imre, L. Esroy, Characterization and lime treatment of olive mill wastewater, Water Res. 35 (2001) 2336–2340. [17] K. Al-Malah, M.O.J. Azzam, N.I. Abu-Lail, Olive mills effluent (OME) wastewater post-treatment using activated clay, Separ. Purif. Technol. 20 (2000) 225–234. [18] E.K. Papadimitriou, I. Chatjipavlidis, C. Balis, Application of composting to olive mill wastewater treatment, Environ. Technol. 18 (1) (1997) 10–107. [19] R. Sarika, N. Kalogerakis, D. Mantzavinos, Treatment of olive mill effluents. Part II. Complete removal of solids by direct flocculation with poly-electrolytes, Environ. Int. 31 (2005) 297–304.

[20] J.B. De Heredia, J. Garcia, Process integration: continuous anaerobic digestion ozonation treatment of olive mill wastewater, Ind. Eng. Chem. Res. 44 (2005) 8750–8755. [21] Ü. Tezcan Ün, U. Altay, A.S. Koparal, U.B. Ogutveren, Complete treatment of olive mill wastewaters by electrooxidation, Chem. Eng. J. 139 (2008) 445–452. [22] G. Hodaifa, J.M. Ochando-Pulido, S. Rodriguez-Vives, A. Martinez-Ferez, Optimization of continuous reactor at pilot scale for olive-oil mill wastewater treatment by Fenton-like process, Chem. Eng. J. 220 (2013) 117–124. [23] J.M. Ochando-Pulido, A. Martinez-Ferez, A focus on pressure-driven membrane technology in olive mill wastewater reclamation: state of the art, Water Sci. Technol. 66 (12) (2012) 2505–2516. [24] C.A. Paraskeva, V.G. Papadakis, E. Tsarouchi, D.G. Kanellopoulou, P.G. Koutsoukos, Membrane processing for olive mill wastewater fractionation, Desalination 213 (2007) 218–229. [25] E. Turano, S. Curcio, M.G. De Paola, V. Calabrò, G. Iorio, An integrated centrifugation–ultrafiltration system in the treatment of olive mill wastewater, J. Membr. Sci. 206 (2002) 519–531. [26] M. Stoller, B. De Caprariis, A. Cicci, N. Verdone, M. Bravi, A. Chianese, About proper membrane process design affected by fouling by means of the analysis of measured threshold flux data, Sep. Purif. Technol. 114 (2013) 83–89. [27] J.M. Ochando-Pulido, M. Stoller, Kinetics and boundary flux optimization of integrated photocatalysis and ultrafiltration process for two-phase vegetation and olive washing wastewaters treatment, Chem. Eng. J. 279 (2015) 387–395. [28] A.E. Greenberg, L.S. Clesceri, A.D. Eaton, Standard Methods for the Examination of Water and Wastewater, 22nd ed., APHA/AWWA/WEF, Washington, DC. Cabs., 2006. [29] J.M. Ochando-Pulido, G. Hodaifa, M.D. Victor-Ortega, S. Rodríguez-Vives, A. Martinez-Ferez, Reuse of olive mill effluents from two-phase extraction process by integrated advanced oxidation and reverse osmosis treatment, J. Hazard. Mater. 263 (1) (2013) 158–167. [30] J.M. Ochando-Pulido, M.D. Victor-Ortega, A. Martinez-Ferez, On the cleaning procedure of a hydrophilic reverse osmosis membrane fouled by secondarytreated olive mill wastewater, Chem. Eng. J. 260 (2015) 142–151. [31] J. Hermia, Constant pressure blocking filtration laws - application to powerlaw non-newtonian fluids, Trans. Inst. Chem. Eng. 60 (1982) 183–187. [32] W.R. Bowen, J.I. Calvo, A. Hernández, Steps of membrane blocking in flux decline during protein microfiltration, J. Membr. Sci. 101 (1995) 153–165. [33] F. Wang, V.V. Tarabara, Pore blocking mechanisms during early stages of membrane fouling by colloids, J. Colloid. Interf. Sci. 328 (2008) 464–469. [34] F. Wang, V.V. Tarabara, Coupled effects of colloidal deposition and salt concentration polarization on reverse osmosis membrane performance, J. Membr. Sci. 293 (2007) 111–123. [35] D.C. Montgomery, A.E. Peck, G.G. Vining, Introduction to Linear Regression Analysis, fourth ed., Wiley-Interscience, N.J., 2006. [36] X. Jin, A. Jawor, S. Kim, E. Hoek, Effects of feed water temperature on separation performance and organic fouling of brackish water RO membranes, Desalination 239 (2009) 346–359. [37] H. Choi, K. Zhang, D.D. Dionysiou, D.B. Oerther, G.A. Sorial, Effect of permeate flux and tangential flow on membrane fouling for wastewater treatment, Sep. Purif. Technol. 4 (1) (2005) 68–78. [38] M. Elimelech, X. Zhu, A. Childress, S. Hong, Role of membrane surface morphology in colloidal fouling of cellulose acetate and composite polyamide reverse osmosis membranes, J. Membr. Sci. 127 (1997) 101–109. [39] S.F.E. Boerlage, M.D. Kennedy, M.P. Aniye, E.M. Abogrean, G. Galjaard, J.C. Schippers, Monitoring particulate fouling in membrane systems, Desalination 118 (1998) 131–142. [40] E.M. Tracey, R.H. Davis, Protein fouling of track-etched polycarbonate microfiltration membranes, J. Colloid Interface Sci. 167 (1994) 104–116. [41] S.S. Madaeni, T. Mohammadi, M.K. Moghadam, Chemical cleaning of reverse osmosis membranes, Desalination 134 (1–3) (2001) 77–82. [42] M.C. Vincent-Vela, B. Cuartas-Uribe, S. Álvarez-Blanco, J. Lora-García, Analysis of fouling resistances under dynamic membrane filtration, Chem. Eng. Process. 50 (2011) 404–408. [43] B. Blankert, B.H.L. Betlem, B. Roffel, Dynamic optimization of a dead-end filtration trajectory: blocking filtration laws, J. Membr. Sci. 285 (2006) 90–95. [44] P. Aimar, Recent progress in understanding particle fouling of filtration membranes, in: Fifth International Membrane Science & Technology Conference, Sydney, Australia, 2003. [45] B. Mi, C.L. Eaton, J.H. Kim, C.K. Colvin, J.C. Lozier, B.J. Marinas, Removal of biological and non-biological viral surrogates by spiral-wound reverse osmosis membrane elements with intact and compromised integrity, Water Res. 38 (2004) 3821. [46] B. Winfield, A study of the factors affecting the rate of fouling of reverse osmosis membranes treating secondary sewage effluents, Water Res. 13 (1979) 565–569. [47] E. Gwon, M. Yu, H. Oh, Y. Ylee, Fouling characteristics of NF and RO operated for removal of dissolved matter from groundwater, Water Res. 37 (2003) 2989–2997. [48] M. Stoller, A three year long experience of effective fouling inhibition by threshold flux based optimization methods on a NF membrane module for olive mill wastewater treatment, Chem. Eng. Trans. 32 (2013) 37–42. [49] M. Stoller, A. Chianese, Influence of the adopted pretreatment process on the critical flux value of batch membrane processes, Ind. Eng. Chem. Res. 46 (2007) 2249–2253.

J.M. Ochando-Pulido et al. / Separation and Purification Technology 168 (2016) 177–187 [50] E.M.V. Hoek, M. Elimelech, Cake-enhanced concentration polarization: a new fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 37 (2003) 5581–5588. [51] D.C. Sioutopoulos, S.G. Yiantsios, A.J. Karabelas, Relation between fouling characteristics of RO and UF membranes in experiments with colloidal organic and inorganic species, Desalination 350 (2010) 62–82. [52] E.O. Akdemir, A. Ozer, Investigation of two ultrafiltration membranes for treatment of olive oil mill wastewater, Desalination 249 (2009) 660–666. [53] M. Cheryan, Ultrafiltration and Microfiltration, Technomic Publishing Company Inc., Lancaster, PA, 1998.

187

[54] V. Chen, A.G. Fane, S. Maedani, I.G. Wenten, Particle deposition during membrane filtration of colloids: transition between concentration polarization and cake formation, J. Membr. Sci. 125 (1997) 109–122. [55] M.K. Purkait, P.K. Bhattacharya, S. Dea, Membrane filtration of leather plant effluent: Flux decline mechanism, J. Membr. Sci. 258 (2005) 85–96. [56] T. Mohammadi, M. Kazemimoghadam, M. Saadabadi, Modeling of membrane fouling and flux decline in reverse osmosis during separation of oil in water emulsions, Desalination 157 (2003) 369–375.