- Email: [email protected]
UF process for drinking water production
DESALINATION ELSEVIER
Desalination
147 (2002) 95-100 www.elsevier.com/locate/desal
Impact of coagulation conditions on the in-line coagulationKJF process for drinking water production C. Guigui”*, J.C. Roucha, L. Durand-Bourlierb, V. Bonnelyeb, P. Aptel” “Laboratoire de GEnie Chimique-CNRS UMR 5503-Universite’ Paul Sabatier; 118 Route de Narbonne, 31062 Toulouse-Cedex, France Tel. +33 (05) 61559773; Fax 33 (OS) 61559760; email: [email protected] “ONDEO, 38 rue du Prksident Wilson, 78230 Le Pecq, France Received
15 February 2002; accepted 4 April 2002
Abstract An in-line coagulation (without settling)/UF process has been studied to improve membrane performance and water quality for surface water treatment. Using coagulation before UF increases permeate quality; the extent of dissolved organic matter removal is controlled by the coagulation step. Efficient coagulation conditions for a coagulation/settling process can be applied for the in-line coagulation/UF process and membrane fouling is reduced. Floe cake resistance is lower than resistance due to the unsettled floe and the uncoagulated organics. For an insideout hollow-fiber system, the impact of coagulation dose depends on filtration mode: for instance, in cross-flow with a feed-and-bleed configuration, a reduction of coagulant dose induces an increase of the mass transfer resistance even in quasi-stable hydrodynamic operating conditions. Keywords: Water treatment;
Fouling;
Hybrid process; Coagulation;
1. Introduction
for larger installations using good-quality surface water. Today, the objectives are to extend membrane technology to poorer-quality sources for the removal of color, taste, dissolved organic matter (DOC) and disinfection by-products (DBP). This requires ultrafiltration technology to be integrated into a multiple treatment line. The
tive
Ultrafiltration (UF) membranes have rapidly become an efficient alternative to conventional treatment for drinking water production. Primarily used for turbidity and microorganism removal, membrane processes are now economically attrac*Corresponding
Ultrafiltration
addition
author.
Presented at the International July 7-12, 2002.
Congress on Membranes
of coagulant
and Membrane
Processes
0011-9 164/02/$- See front matter 0 2002 Elsevier Science B.V. All rights reserved PII: SO0 I I-9 164(02)00582-9
before
UF or MF units with
(ICOM),
Toulouse,
France,
96
C. Guigui et al. /Desalination
or without settling (in-line coagulation) may increase natural organic matter (NOM) removal for a better reduction of DBP. In the conventional process (coagulation/settling/ conventional filtration), the coagulation step is used to reduce the concentration of NOM and suspended solids (MES) after settling or sand filtration and to minimize the formation of DBP. This primary physical-chemical step controls floe formation and thus the efficiency of the following unit operations. The floe structure may differ depending on the coagulation conditions (coagulant, pH, dose). A lot of researches have been devoted to coagulation mechanisms. However, it remains difficult to accurately identify the preponderant physical-chemical phenomena because of the complexity of NOM and floe characterization. Nevertheless, the literature reports two different coagulation mechanisms: sweep coagulation and coagulation by charge neutralization. [l-3]. The first is reported to be dominant at low pH (5-6) and low concentration; the second is preponderant at higher pH and concentration. The floe structure (porosity, density, cohesion force) may differ as a function of these physical-chemical conditions: the floe formed at the lowest pH and dose are reported to be denser and less porous than those formed at higher pH and dose (sweep coagulation) [41. In addition, the impact of coagulation on the conventional process and especially on the liquidsolid separation depends on the hydrodynamic conditions during the mixing and the transport of floe. The shear rate is one of the important operating parameters since it controls collision frequency and hence floe size, its density and its rupture. From an industrial point of view, the literature reports that for a coagulation-flocculation/settling process, the optimal coagulation conditions should be a pH higher than 6, a coagulant dose between 30 and 100 ppm (in FeCl,) and a shear rate around 100-200 s-l at the coagulant injection time. But, the shear rate during floe formation is not the most important operating parameter. For a coagulation-
147 (2002) 95-100
flocculation/conventional filtration process, the coagulant concentration should be lower, the pH under 6 and in this case, the control of the shear rate up to 300 s-r is very important. Moreover, according to Masschelein [5], ferric floe is more resistant to stress forces than aluminum floe and in the combination of coagulation-flocculation/ conventional filtration aluminum floe would be apt to break up. According to these points, the same questions still remain for a coagulation/settling/W (or MF) process and for an in-line coagulation/W process: what are the optimal coagulation conditions for the combined process in terms of permeate quality and membrane fouling, is the impact of coagulation conditions modified according to the filtration mode (including geometry and nature of membrane) and how does the combined process perform? The overall goal of this research was to address such questions for an in-line coagulation/UF membrane system using inside/out hollow fibers (HF). This was a multiphase project with the initial work focusing on control of membrane fouling using hydrodynamic forces (as Dean vortices) at given coagulation conditions. A labscale study [6] has demonstrated that in crossflow filtration, the feed-and-bleed configuration reduces backwash frequency and critical conditions have been found where filtration operates in a quasi non-deposit regime. Experiments on a semi-industrial scale [7] reached the same conclusion. In addition to that, the second objective was to evaluate the impact of coagulation conditions on fouling and water quality for the two filtration modes (cross-flow and dead-end) in order to better understand the mechanisms that could affect membrane performances: the results of this study are reported in the present paper.
2. Experimental The Canal du Midi (Toulouse, France) served as the source of surface water. The average characteristics were: DOC 3.9 mg.l-I, UV,,, 7.2 m-l and turbidity 21 NTU. FeCl,, FeSO, and PACL were
C. Guigui et al. /Desalination
used as coagulant. Typical jar-tests (JT) (rapid mix, slow mix regime and settling) were performed to evaluate the effect of coagulant dose and pH. The UF membranes used were cellulosic hollow fibers (Aquasource, France) with an internal diameter of 0.93 mm. In addition, flat sheet membranes were cast from the same collodion as used to spin hollow fibers. Filtration experiments (cross-flow and dead-end) using hollow-fiber modules were performed with a lab-scale pilot. Batch experiments were performed in dead-end with an unstirred amicon cell (Amicon 8200). 3. Results and discussion 3. I. Coagulation/settling
process
At the same metal cation concentration (mol/l), the conventional jar-test demonstrated that the use of ferric chloride allows higher NOM removal compared to other coagulants. But, for this coagulant, an optimal coagulant dose was not found and NOM removal increases with the increase of coagulant dose up to a plateau. In these conditions, the coagulation efficiency is about 60% corresponding to a Fe concentration of around 3-4x 1W mol/l. In more detail, the analysis demonstrated that the UV removal is higher than DOC removal in the same conditions of coagulation. The variation of pH has a greater impact on coagulation efficiency so as the concentration of coagulant decreases. A pH of around 5-6 leads to the highest DOC removal whereas a pH around 6-7 induces the highest UV removal. This higher efficiency at pH of around 6 should be due to the higher positive charge of the hydrolyzed metal species, which may react with the negatively charged NOM. The comparison of these results with the literature remains difficult since the choice of coagulant and the coagulant conditions depend on the nature and interaction of the NOM. However, it is interesting to note that many authors indicate that coagulation is more efficient to remove UV than DOC [8-lo] and that UV is correlated with
97
147 (2002) 95-100
the NOM of the highest molecular weights. Besides, Vilge-Ritter [ 1l] and Gray [ 121 have concluded that ferric chloride preferentially binds to polyhydroxyaromatic compounds, which have high molecular weight. Results about influence of the coagulation conditions would give an important indication on the NOM of Canal du Midi water. 3.2. CoagulationAJF
process
Preliminary dead-end experiments show that under unstirred and stirred conditions, the rate of fouling and the permeate quality are similar whatever the nature of the coagulant (at the same coagulation conditions). For this reason, the filtration experiments were performed with ferric chloride. 3.2.1. Water quality (Fig. I) The level of NOM removal depends on the pH. The reduction of NOM is slightly higher for a pH around 7 and the impact of the pH on the coagulation efficiency was greater for the highest dose of coagulant. Besides, as for the coagulationflocculation/settling experiments, it may be interesting to distinguish the efficiency of the coagulation/UP process on UV and DOC removal.
6.5 PH
Fig. 1. NOM removal by in-line coagulationAJF. Dead-end filtration (TMP=l bar). Impact of FeCl, dose (in ppm-Fe) and pH. Raw water: UV = 0.083 cm-‘, DOC = 5.4 ppm, turbidity = 25.4 NTU).
98
C. Guigui et al. /Desalination
Whatever the coagulation conditions, UV removal was greater than DOC removal. The tendency was that an increase of the pH (up to a neutral pH) induces an increase of the removal even if the impact of the pH variations is greater on DOC removal than on UV removal. About DOC removal, the impact of pH is all the more important at high concentration. Then, at pH 7.5 and 20 ppm in Fe, NOM removal was about 50%. It is important to keep in mind that, as retention of the floe by the UF membrane is total, there is an accumulation of floe on the membrane in the unstirred dead-end experiments. This increases the residence time and the floe concentration in the filtration cell which may cause an increase of adsorption of uncoagulated NOM on floe. This phenomenon would be enhanced when the floe is abundant and the largest, or in other words, at a high coagulant concentration and neutral pH. In conclusion of this first part, it may be noted that good coagulation conditions (nature of coagulant, pH and dose) generally used in a coagulation/ settling treatment, should lead to good performance in terms of water quality for an in-line coagulation/UF process with dead-end filtration. However, if there are some fluctuations around this optimal point, the impact of the pH or the coagulant concentration might be different in a coagulation/settling treatment than in a coagulation/UF process. In particular, the impact of the pH is greater at low coagulant dose in the coagulation/settling process whereas it is greater at high concentrations in a coagulation/UF process. Moreover, to evaluate the membrane contribution on NOM removal in the in-line coagulation/ UF process, the supematant of the coagulatedsettled water was ultrafiltrated: more than 80% of the total NOM removal was accomplished during the coagulation step (at the optimum coagulant dose). One last point concerns the influence of the transmembrane pressure (TMP) on permeate quality and fouling. In these experiments, the coagulation step was performed at atmospheric pressure and the pressure was increased during
147 (2002) 95-100
the filtration step. The results show that there is no impact of the TMP on the permeate quality. 3.2.2. Membrane fouling UF batch-cell experiments showed that coagulation improves permeation flux. Both pH and concentration had an impact on the rate of fouling. As expected in dead end filtration, the cake resistance was greater at high coagulant doses. The influence of the pH was the same whatever the coagulant dose and the rate of fouling was the lowest at pH around 7.5. These experimental results have been analyzed with Hermia laws [ 131. It appears that there is a modification of fouling mechanism at neutral pH. This suggests that there is a modification of floe structure and/or cake porosity (at similar size). So, in unstirred dead-end filtration, controlling the coagulation pH remains important. To understand fouling phenomena, different feed fractions were prepared (coagulation at neutral pH): (1) raw water without coagulation, (2) supematant of a JT experiment, (3) floe of a JT redispersed in distilled water. After the deposit of a floe cake on the membrane, ultrafiltration , ,, ,..._” ,,.._ ,, ..“..
0
10
.._ _...”
20
30
.I- .
40
50
Time (min)
Fig. 2. Variations of the ratio (flux/initial flux) vs. time for UF of (1) raw water, (2) supematant, (3) floe redispersed in distilled water, (4) supematant on the floe cake, (5) distilled water on the floes cake ([Fe] = 20ppm).
C. Guigui
et al. /Desalination
was also performed with distilled water (5) and with the supematant fraction (4). Fig. 2 shows that the fouling rate is the greatest with the raw water (1) and that the supernatant (2) is a higher foulant feed than the floe suspension (3). Moreover, the resistance of a pre-deposit floe cake is stable vs. time when distilled water is filtrated (5) while it increases when the supematant of JT is filtrated (4). This last point might be in relation with data about permeate quality: the floe cake on the membrane could retain the uncoagulated NOM and/or the smallest floe. Moreover, in these experiments, there is no major difference in the NOM content between the supematant of JT and the permeate of the JT supematant filtrated on floe cake; so we speculate that the reduction of fouling between the direct filtration of the supernatant and the filtration of the supematant on a floe cake, is certainly due to the retention of the smallest floe (preventing pore blocking) by the floe cake. About the influence of TMP on membrane fouling, results show that when the TMP increases, there is an increase of the specific cake resistance (Fig. 3). But, if the pressure decreases, the specific resistance returns to its initial value: the compaction is reversible.
0
10
20
30
40
50
60
70
AP (kPa)
Fig. 3. Variations of the specific resistance as a function of the compressive pressure in the cake. [Fe] = 20 ppm, pH = 7.5.
147 (2002)
95-100
99
From the Carman-Kozeny equation [ 141, the specific resistance depends on floe size and cake porosity. The impact of the pressure may be on the one hand, on the floe deformation and on the other hand, on the cake porosity. It remains difficult to distinguish which effect is predominant but the global effect is an increase of the cake resistance. The floe from coagulation at a high coagulant dose and a neutral pH has a lower density and so are highly compressible. This structure may explain the rearrangement of the cake when the pressure increases. This result is similar to those of Lee and ~011. [15]: they have shown that aluminum floe was compressible and the floe formed by sweep coagulation is more compressible than floe formed by charge-neutralization. However, although the floe cake is compressible, the compression is not irreversible: when the pressure is reduced, the cake is relaxed and regains a more porous structure. But, this apparent reversibility does not necessarily mean that the cake returned to its initial structure: the piling of the floe is certainly different, but in terms of water permeability, the cake resistance is the same. From a process point of view, this is an important point, in so far as, the filtration process is generally performed at constant feed flow rate: the deposit of floe causes an increase in TMP during the filtration period and a reversible effect may be an important point in the backwash procedure. However, operation in a constant TMP mode of filtration would certainly increase the duration of the filtration. With HF membranes, the impact of the coagulant dose depends on the filtration mode: in dead-end mode, a lower dose of coagulant induces a reduction of the transmembrane pressure during filtration but the backwash efficiency decreases. In cross-flow mode with a feed-and-bleed configuration, a reduction of the amount of coagulant added induces an increase of the fouling resistance even in quasi-stable hydrodynamic operating conditions (Fig. 4). However, this rate of fouling remains very low compared to that caused by raw water filtration. In these conditions, the impact of coagulation conditions is less than in dead-end
C. Guigui et al. /Desalination
0
0.1
0.2
0.3
0.4
0.5
0.6
vol(m13/rn*)
resistance vs. filtrate volume per unit of surface area. (coagulant dose in ppm-Fe, UF in cross-flow mode: Re = 1210 and constant flux J = 12Ol/h.m*). Fig. 4. Fouling
filtration. This may be explained by the importance of the shear rate during the process (characterization of floe in the recirculating loop may be interesting to explain this difference). 4. Conclusion In-line coagulationAJF process can be used to treat poor quality surface water: permeate quality is increased and membrane fouling is reduced. Coagulation conditions determined by jar-test experiments are efficient to good performance in line coagulation/UF process. The permeate quality remains controlled by coagulation efficiency and the impact of physical-chemical coagulation conditions depends on the filtration process. It can be noted the fraction of interstitial fluid which contents the uncoagulated water (and/or the smallest floe) may induce a higher fouling than floe cake. References [l ]
[2]
G.A. Edwards and A. Amirtharajah, Removing color caused by humic acids, J. AWWA, (1985) 50-57. J.K. Edzwald and J.E. Tobiason, Enhanced coagulation: USArequirements and a broader view, Proc. International
147 (2002) 95-100 IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 73-80. [31 CR. O’Melia, WC. Becker and K.K. Au, Removal of humic substances by coagulation, Proc. International IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 55-64. [41 W.R. Knocke, Effects of coagulation on sludge thickening and dewatering, J. AWWA, 86(6) (1987) 89. W.J. Masschelein, Lacoagulation, in: Processus unitaires [51 de traitement de l’eau potable, Cedeboc, New York, Ch. 5, 1992. [61 C. Guigui, V. Bonnelye, L. Durand-Bourlier, J.C. Rouch and I? Aptel, Combination of coagulation and UF for drinking water production: impact of process configuration and module design, Water Sci. Tech.: Water Supply, 1(5/6) (2001) 107-118. [71 C. Guigui, V. Bonnelye, L. Durand-Bourlier, N. Abidine, J.C. Rouch and P Aptel, A novel approach for the UF of surface water: combination of coagulation and Dean vortices in a feed and bleed process configuration, AWWA Membrane. Tech. Conf., San Antonio, Texas, March 4-7 (2001), CD-Rom. R. Bian, Y. Watanabe, N. Tambo and G Ozawa, Removal PI of humic substances by UF and NF membranes systems, Proc. International IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 131-139. S.W. Krasner and G. Amy, Jar-test evaluations of PI enhanced coagulation. Utilities should be evaluating the amenability of their supplies to enhanced coagulation, J. AWWA, 85(10) (1995) 93-107. 101 E. Lefebre and B. Leg&e, Coagulation par Fe(II1) de substances humiques extraites d’eaux de surface: effet du pH et de la concentration en substances humiques, Wat. Res., 24(5) (1990) 591-606. 111 A. Vilge-Ritter, Etude des mecanismes d’elimination de la matiere organique des eaux de surface par coagulationfloculation a l’aide de sels d’aluminium ou de fer, PhD Thesis, Universitt d’Aix-Marseille III, France, 1997. [I21 K.A. Gray and A. H. Simpson, Use of pyrolysis gas chromatography-mass spectrometry to study the nature and the behavior of natural organic matter in water treatment. Water Disinfection and natural organic matter characterization and control. Proc. 5th Gotenburg Symposium, September 28-30, 1992, Nice, France. J. Hermia, Constant pressure blocking filtration laws. Application to power-law non-newtonian fluids, Tram Inst. Chem. Eng., 60 (1982) 183-187. P.C. Carman, Fundamental principles of industrial filtration, Trans. Inst. Chem. Eng., 16 (1938) 168-183. J.D. Lee, S.H. Lee, M.H. Jo, P-K. Park, C.H. Lee and J.W. Kwak, Effect of coagulation conditions on membrane filtration characteristics in coagulation-MF process for water treatment, Env. Sci. Technol., 34(17) (2000) 3780.
Desalination
147 (2002) 95-100 www.elsevier.com/locate/desal
Impact of coagulation conditions on the in-line coagulationKJF process for drinking water production C. Guigui”*, J.C. Roucha, L. Durand-Bourlierb, V. Bonnelyeb, P. Aptel” “Laboratoire de GEnie Chimique-CNRS UMR 5503-Universite’ Paul Sabatier; 118 Route de Narbonne, 31062 Toulouse-Cedex, France Tel. +33 (05) 61559773; Fax 33 (OS) 61559760; email: [email protected] “ONDEO, 38 rue du Prksident Wilson, 78230 Le Pecq, France Received
15 February 2002; accepted 4 April 2002
Abstract An in-line coagulation (without settling)/UF process has been studied to improve membrane performance and water quality for surface water treatment. Using coagulation before UF increases permeate quality; the extent of dissolved organic matter removal is controlled by the coagulation step. Efficient coagulation conditions for a coagulation/settling process can be applied for the in-line coagulation/UF process and membrane fouling is reduced. Floe cake resistance is lower than resistance due to the unsettled floe and the uncoagulated organics. For an insideout hollow-fiber system, the impact of coagulation dose depends on filtration mode: for instance, in cross-flow with a feed-and-bleed configuration, a reduction of coagulant dose induces an increase of the mass transfer resistance even in quasi-stable hydrodynamic operating conditions. Keywords: Water treatment;
Fouling;
Hybrid process; Coagulation;
1. Introduction
for larger installations using good-quality surface water. Today, the objectives are to extend membrane technology to poorer-quality sources for the removal of color, taste, dissolved organic matter (DOC) and disinfection by-products (DBP). This requires ultrafiltration technology to be integrated into a multiple treatment line. The
tive
Ultrafiltration (UF) membranes have rapidly become an efficient alternative to conventional treatment for drinking water production. Primarily used for turbidity and microorganism removal, membrane processes are now economically attrac*Corresponding
Ultrafiltration
addition
author.
Presented at the International July 7-12, 2002.
Congress on Membranes
of coagulant
and Membrane
Processes
0011-9 164/02/$- See front matter 0 2002 Elsevier Science B.V. All rights reserved PII: SO0 I I-9 164(02)00582-9
before
UF or MF units with
(ICOM),
Toulouse,
France,
96
C. Guigui et al. /Desalination
or without settling (in-line coagulation) may increase natural organic matter (NOM) removal for a better reduction of DBP. In the conventional process (coagulation/settling/ conventional filtration), the coagulation step is used to reduce the concentration of NOM and suspended solids (MES) after settling or sand filtration and to minimize the formation of DBP. This primary physical-chemical step controls floe formation and thus the efficiency of the following unit operations. The floe structure may differ depending on the coagulation conditions (coagulant, pH, dose). A lot of researches have been devoted to coagulation mechanisms. However, it remains difficult to accurately identify the preponderant physical-chemical phenomena because of the complexity of NOM and floe characterization. Nevertheless, the literature reports two different coagulation mechanisms: sweep coagulation and coagulation by charge neutralization. [l-3]. The first is reported to be dominant at low pH (5-6) and low concentration; the second is preponderant at higher pH and concentration. The floe structure (porosity, density, cohesion force) may differ as a function of these physical-chemical conditions: the floe formed at the lowest pH and dose are reported to be denser and less porous than those formed at higher pH and dose (sweep coagulation) [41. In addition, the impact of coagulation on the conventional process and especially on the liquidsolid separation depends on the hydrodynamic conditions during the mixing and the transport of floe. The shear rate is one of the important operating parameters since it controls collision frequency and hence floe size, its density and its rupture. From an industrial point of view, the literature reports that for a coagulation-flocculation/settling process, the optimal coagulation conditions should be a pH higher than 6, a coagulant dose between 30 and 100 ppm (in FeCl,) and a shear rate around 100-200 s-l at the coagulant injection time. But, the shear rate during floe formation is not the most important operating parameter. For a coagulation-
147 (2002) 95-100
flocculation/conventional filtration process, the coagulant concentration should be lower, the pH under 6 and in this case, the control of the shear rate up to 300 s-r is very important. Moreover, according to Masschelein [5], ferric floe is more resistant to stress forces than aluminum floe and in the combination of coagulation-flocculation/ conventional filtration aluminum floe would be apt to break up. According to these points, the same questions still remain for a coagulation/settling/W (or MF) process and for an in-line coagulation/W process: what are the optimal coagulation conditions for the combined process in terms of permeate quality and membrane fouling, is the impact of coagulation conditions modified according to the filtration mode (including geometry and nature of membrane) and how does the combined process perform? The overall goal of this research was to address such questions for an in-line coagulation/UF membrane system using inside/out hollow fibers (HF). This was a multiphase project with the initial work focusing on control of membrane fouling using hydrodynamic forces (as Dean vortices) at given coagulation conditions. A labscale study [6] has demonstrated that in crossflow filtration, the feed-and-bleed configuration reduces backwash frequency and critical conditions have been found where filtration operates in a quasi non-deposit regime. Experiments on a semi-industrial scale [7] reached the same conclusion. In addition to that, the second objective was to evaluate the impact of coagulation conditions on fouling and water quality for the two filtration modes (cross-flow and dead-end) in order to better understand the mechanisms that could affect membrane performances: the results of this study are reported in the present paper.
2. Experimental The Canal du Midi (Toulouse, France) served as the source of surface water. The average characteristics were: DOC 3.9 mg.l-I, UV,,, 7.2 m-l and turbidity 21 NTU. FeCl,, FeSO, and PACL were
C. Guigui et al. /Desalination
used as coagulant. Typical jar-tests (JT) (rapid mix, slow mix regime and settling) were performed to evaluate the effect of coagulant dose and pH. The UF membranes used were cellulosic hollow fibers (Aquasource, France) with an internal diameter of 0.93 mm. In addition, flat sheet membranes were cast from the same collodion as used to spin hollow fibers. Filtration experiments (cross-flow and dead-end) using hollow-fiber modules were performed with a lab-scale pilot. Batch experiments were performed in dead-end with an unstirred amicon cell (Amicon 8200). 3. Results and discussion 3. I. Coagulation/settling
process
At the same metal cation concentration (mol/l), the conventional jar-test demonstrated that the use of ferric chloride allows higher NOM removal compared to other coagulants. But, for this coagulant, an optimal coagulant dose was not found and NOM removal increases with the increase of coagulant dose up to a plateau. In these conditions, the coagulation efficiency is about 60% corresponding to a Fe concentration of around 3-4x 1W mol/l. In more detail, the analysis demonstrated that the UV removal is higher than DOC removal in the same conditions of coagulation. The variation of pH has a greater impact on coagulation efficiency so as the concentration of coagulant decreases. A pH of around 5-6 leads to the highest DOC removal whereas a pH around 6-7 induces the highest UV removal. This higher efficiency at pH of around 6 should be due to the higher positive charge of the hydrolyzed metal species, which may react with the negatively charged NOM. The comparison of these results with the literature remains difficult since the choice of coagulant and the coagulant conditions depend on the nature and interaction of the NOM. However, it is interesting to note that many authors indicate that coagulation is more efficient to remove UV than DOC [8-lo] and that UV is correlated with
97
147 (2002) 95-100
the NOM of the highest molecular weights. Besides, Vilge-Ritter [ 1l] and Gray [ 121 have concluded that ferric chloride preferentially binds to polyhydroxyaromatic compounds, which have high molecular weight. Results about influence of the coagulation conditions would give an important indication on the NOM of Canal du Midi water. 3.2. CoagulationAJF
process
Preliminary dead-end experiments show that under unstirred and stirred conditions, the rate of fouling and the permeate quality are similar whatever the nature of the coagulant (at the same coagulation conditions). For this reason, the filtration experiments were performed with ferric chloride. 3.2.1. Water quality (Fig. I) The level of NOM removal depends on the pH. The reduction of NOM is slightly higher for a pH around 7 and the impact of the pH on the coagulation efficiency was greater for the highest dose of coagulant. Besides, as for the coagulationflocculation/settling experiments, it may be interesting to distinguish the efficiency of the coagulation/UP process on UV and DOC removal.
6.5 PH
Fig. 1. NOM removal by in-line coagulationAJF. Dead-end filtration (TMP=l bar). Impact of FeCl, dose (in ppm-Fe) and pH. Raw water: UV = 0.083 cm-‘, DOC = 5.4 ppm, turbidity = 25.4 NTU).
98
C. Guigui et al. /Desalination
Whatever the coagulation conditions, UV removal was greater than DOC removal. The tendency was that an increase of the pH (up to a neutral pH) induces an increase of the removal even if the impact of the pH variations is greater on DOC removal than on UV removal. About DOC removal, the impact of pH is all the more important at high concentration. Then, at pH 7.5 and 20 ppm in Fe, NOM removal was about 50%. It is important to keep in mind that, as retention of the floe by the UF membrane is total, there is an accumulation of floe on the membrane in the unstirred dead-end experiments. This increases the residence time and the floe concentration in the filtration cell which may cause an increase of adsorption of uncoagulated NOM on floe. This phenomenon would be enhanced when the floe is abundant and the largest, or in other words, at a high coagulant concentration and neutral pH. In conclusion of this first part, it may be noted that good coagulation conditions (nature of coagulant, pH and dose) generally used in a coagulation/ settling treatment, should lead to good performance in terms of water quality for an in-line coagulation/UF process with dead-end filtration. However, if there are some fluctuations around this optimal point, the impact of the pH or the coagulant concentration might be different in a coagulation/settling treatment than in a coagulation/UF process. In particular, the impact of the pH is greater at low coagulant dose in the coagulation/settling process whereas it is greater at high concentrations in a coagulation/UF process. Moreover, to evaluate the membrane contribution on NOM removal in the in-line coagulation/ UF process, the supematant of the coagulatedsettled water was ultrafiltrated: more than 80% of the total NOM removal was accomplished during the coagulation step (at the optimum coagulant dose). One last point concerns the influence of the transmembrane pressure (TMP) on permeate quality and fouling. In these experiments, the coagulation step was performed at atmospheric pressure and the pressure was increased during
147 (2002) 95-100
the filtration step. The results show that there is no impact of the TMP on the permeate quality. 3.2.2. Membrane fouling UF batch-cell experiments showed that coagulation improves permeation flux. Both pH and concentration had an impact on the rate of fouling. As expected in dead end filtration, the cake resistance was greater at high coagulant doses. The influence of the pH was the same whatever the coagulant dose and the rate of fouling was the lowest at pH around 7.5. These experimental results have been analyzed with Hermia laws [ 131. It appears that there is a modification of fouling mechanism at neutral pH. This suggests that there is a modification of floe structure and/or cake porosity (at similar size). So, in unstirred dead-end filtration, controlling the coagulation pH remains important. To understand fouling phenomena, different feed fractions were prepared (coagulation at neutral pH): (1) raw water without coagulation, (2) supematant of a JT experiment, (3) floe of a JT redispersed in distilled water. After the deposit of a floe cake on the membrane, ultrafiltration , ,, ,..._” ,,.._ ,, ..“..
0
10
.._ _...”
20
30
.I- .
40
50
Time (min)
Fig. 2. Variations of the ratio (flux/initial flux) vs. time for UF of (1) raw water, (2) supematant, (3) floe redispersed in distilled water, (4) supematant on the floe cake, (5) distilled water on the floes cake ([Fe] = 20ppm).
C. Guigui
et al. /Desalination
was also performed with distilled water (5) and with the supematant fraction (4). Fig. 2 shows that the fouling rate is the greatest with the raw water (1) and that the supernatant (2) is a higher foulant feed than the floe suspension (3). Moreover, the resistance of a pre-deposit floe cake is stable vs. time when distilled water is filtrated (5) while it increases when the supematant of JT is filtrated (4). This last point might be in relation with data about permeate quality: the floe cake on the membrane could retain the uncoagulated NOM and/or the smallest floe. Moreover, in these experiments, there is no major difference in the NOM content between the supematant of JT and the permeate of the JT supematant filtrated on floe cake; so we speculate that the reduction of fouling between the direct filtration of the supernatant and the filtration of the supematant on a floe cake, is certainly due to the retention of the smallest floe (preventing pore blocking) by the floe cake. About the influence of TMP on membrane fouling, results show that when the TMP increases, there is an increase of the specific cake resistance (Fig. 3). But, if the pressure decreases, the specific resistance returns to its initial value: the compaction is reversible.
0
10
20
30
40
50
60
70
AP (kPa)
Fig. 3. Variations of the specific resistance as a function of the compressive pressure in the cake. [Fe] = 20 ppm, pH = 7.5.
147 (2002)
95-100
99
From the Carman-Kozeny equation [ 141, the specific resistance depends on floe size and cake porosity. The impact of the pressure may be on the one hand, on the floe deformation and on the other hand, on the cake porosity. It remains difficult to distinguish which effect is predominant but the global effect is an increase of the cake resistance. The floe from coagulation at a high coagulant dose and a neutral pH has a lower density and so are highly compressible. This structure may explain the rearrangement of the cake when the pressure increases. This result is similar to those of Lee and ~011. [15]: they have shown that aluminum floe was compressible and the floe formed by sweep coagulation is more compressible than floe formed by charge-neutralization. However, although the floe cake is compressible, the compression is not irreversible: when the pressure is reduced, the cake is relaxed and regains a more porous structure. But, this apparent reversibility does not necessarily mean that the cake returned to its initial structure: the piling of the floe is certainly different, but in terms of water permeability, the cake resistance is the same. From a process point of view, this is an important point, in so far as, the filtration process is generally performed at constant feed flow rate: the deposit of floe causes an increase in TMP during the filtration period and a reversible effect may be an important point in the backwash procedure. However, operation in a constant TMP mode of filtration would certainly increase the duration of the filtration. With HF membranes, the impact of the coagulant dose depends on the filtration mode: in dead-end mode, a lower dose of coagulant induces a reduction of the transmembrane pressure during filtration but the backwash efficiency decreases. In cross-flow mode with a feed-and-bleed configuration, a reduction of the amount of coagulant added induces an increase of the fouling resistance even in quasi-stable hydrodynamic operating conditions (Fig. 4). However, this rate of fouling remains very low compared to that caused by raw water filtration. In these conditions, the impact of coagulation conditions is less than in dead-end
C. Guigui et al. /Desalination
0
0.1
0.2
0.3
0.4
0.5
0.6
vol(m13/rn*)
resistance vs. filtrate volume per unit of surface area. (coagulant dose in ppm-Fe, UF in cross-flow mode: Re = 1210 and constant flux J = 12Ol/h.m*). Fig. 4. Fouling
filtration. This may be explained by the importance of the shear rate during the process (characterization of floe in the recirculating loop may be interesting to explain this difference). 4. Conclusion In-line coagulationAJF process can be used to treat poor quality surface water: permeate quality is increased and membrane fouling is reduced. Coagulation conditions determined by jar-test experiments are efficient to good performance in line coagulation/UF process. The permeate quality remains controlled by coagulation efficiency and the impact of physical-chemical coagulation conditions depends on the filtration process. It can be noted the fraction of interstitial fluid which contents the uncoagulated water (and/or the smallest floe) may induce a higher fouling than floe cake. References [l ]
[2]
G.A. Edwards and A. Amirtharajah, Removing color caused by humic acids, J. AWWA, (1985) 50-57. J.K. Edzwald and J.E. Tobiason, Enhanced coagulation: USArequirements and a broader view, Proc. International
147 (2002) 95-100 IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 73-80. [31 CR. O’Melia, WC. Becker and K.K. Au, Removal of humic substances by coagulation, Proc. International IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 55-64. [41 W.R. Knocke, Effects of coagulation on sludge thickening and dewatering, J. AWWA, 86(6) (1987) 89. W.J. Masschelein, Lacoagulation, in: Processus unitaires [51 de traitement de l’eau potable, Cedeboc, New York, Ch. 5, 1992. [61 C. Guigui, V. Bonnelye, L. Durand-Bourlier, J.C. Rouch and I? Aptel, Combination of coagulation and UF for drinking water production: impact of process configuration and module design, Water Sci. Tech.: Water Supply, 1(5/6) (2001) 107-118. [71 C. Guigui, V. Bonnelye, L. Durand-Bourlier, N. Abidine, J.C. Rouch and P Aptel, A novel approach for the UF of surface water: combination of coagulation and Dean vortices in a feed and bleed process configuration, AWWA Membrane. Tech. Conf., San Antonio, Texas, March 4-7 (2001), CD-Rom. R. Bian, Y. Watanabe, N. Tambo and G Ozawa, Removal PI of humic substances by UF and NF membranes systems, Proc. International IAWQ-IWSA Joint Specialist, Trondheim, Norway, 2000, pp. 131-139. S.W. Krasner and G. Amy, Jar-test evaluations of PI enhanced coagulation. Utilities should be evaluating the amenability of their supplies to enhanced coagulation, J. AWWA, 85(10) (1995) 93-107. 101 E. Lefebre and B. Leg&e, Coagulation par Fe(II1) de substances humiques extraites d’eaux de surface: effet du pH et de la concentration en substances humiques, Wat. Res., 24(5) (1990) 591-606. 111 A. Vilge-Ritter, Etude des mecanismes d’elimination de la matiere organique des eaux de surface par coagulationfloculation a l’aide de sels d’aluminium ou de fer, PhD Thesis, Universitt d’Aix-Marseille III, France, 1997. [I21 K.A. Gray and A. H. Simpson, Use of pyrolysis gas chromatography-mass spectrometry to study the nature and the behavior of natural organic matter in water treatment. Water Disinfection and natural organic matter characterization and control. Proc. 5th Gotenburg Symposium, September 28-30, 1992, Nice, France. J. Hermia, Constant pressure blocking filtration laws. Application to power-law non-newtonian fluids, Tram Inst. Chem. Eng., 60 (1982) 183-187. P.C. Carman, Fundamental principles of industrial filtration, Trans. Inst. Chem. Eng., 16 (1938) 168-183. J.D. Lee, S.H. Lee, M.H. Jo, P-K. Park, C.H. Lee and J.W. Kwak, Effect of coagulation conditions on membrane filtration characteristics in coagulation-MF process for water treatment, Env. Sci. Technol., 34(17) (2000) 3780.