Ognier Et Al., 2002, Ds

DESALINATION ELSEVIER

Desalination

146 (2002) 141-147 www.elsevier.com/locate/desal

Characterisation and modelling of fouling in membrane bioreactors S. Ognier*, C. Wisniewski, A. Grasmick Laborutoire Genie des Procedes de Montpellier 2, CC 024, Place Eugene Bataillon 34095 Montpellier Cedex 05 France Tel. +33 (4) 67 14 48 54; Fax +33 (4) 67 14 48 54; emails: [email protected] [email protected] [email protected] Received 7 February 2002; accepted 6 March 2002

Abstract

A membrane bioreactor used for denitrification of a synthetic substrate was studied in term of membrane fouling. For standard pH and temperature conditions, subcritical conditions were defined to ensure the process stability. The stepwise method was used to determine the critical flux for the deposition of colloidal particles. Under standard physicochemical conditions, only a low and constant fouling resistance was observed if the permeate flux was maintained below the critical flux. The influence of physicochemical variations was then investigated by varying pH and temperature in the biological reactor. It was observed that, when the pH value was higher than a critical one, the membrane was rapidly fouled. This maximum admissible pH value decreased when the temperature increased. On analysing the reversible nature of fouling and the variations of ionic concentrations with the pH, the role of carbonate calcium precipitation was pointed out. By using classical filtration models, it was shown that the fouling mechanism could be the deposition of CaCO, particles formed in the bulk suspension by bulk crystallisation. Keywords: Membrane

bioreactor;

Membrane

fouling; Subcritical

1. Introduction Critical flux is an interesting notion to define optimal hydrodynamic conditions; subcritical conditions can be defined to avoid macroscopic *Corresponding author. Presented ut the International July 7-l 2, 2002.

Congress on Membranes

regime; Precipitation

deposits from building up on the membrane surface [ 11. However, the membrane permeability can decrease during the operation due to the interactions between soluble compounds and membrane material, which do not depend on hydrodynamic conditions. Therefore, the stability and Membrane

Processes

001 l-91 64/02/$- See front matter 0 2002 Elsevier Science B.V. All rights reserved PII: SO0 II-9

164(02)00508-8

(ICOM),

Toulouse, France,

142

S. Ognier et al. /Desalination

of the system depends not only on hydrodynamic conditions but also on biological and physicochemical suspension properties, and it is of utmost importance to define subcritical hydrodynamic conditions as well as physico-chemical and biological conditions to obtain a stable filtration regime. In these conditions, long filtration periods without having to use chemical cleaning procedures can be obtained. Fan et al. [2] showed that a membrane bioreactor (MBR) system for the treatment of raw municipal wastewater can be run continually over 70 d with a stable transmembrane pressure. However, the biological and physico-chemical properties of the suspension are not always stable due to influent composition or temperature changes. As these variations are weak, they are not always taken into account when defining the operating conditions. Therefore, subcritical conditions are generally defined experimentally in fixed biological and physicochemical conditions. Based on these considerations, the objective of this work was (i) to ensure that stable filtration conditions could be obtained in an MBR under constant hydrodynamic, biological and physico-chemical conditions and (ii) to study the influence of weak variations of two physicochemical parameters (pH and temperature) on process stability. The fouling phenomena were analysed by using classical filtration models. 2. Experimental 2.1. Membrane bioreactor Experiments were conducted on a pilot MBR, which consisted of a 20-l bioreactor tank and a ceramic ultrafiltration membrane module (Membralox@) having a 0.24-m* surface area and a mean pore size of 0.05 pm and a resistance of 5x10” m-l. The recirculated pump integrated to the system ensured the perfect mixing of the reactor and made the retentate circulate with a I .6 m/s tangential velocity in the membrane module, corresponding to a wall shear stress of

146 (2002) 141-147

13 Pa. A cooling system kept the whole system at a constant temperature of 25?1 “C. A constant permeation flux was maintained by using a suction pump (Watson-Marlow 505 RS). 2.2. Denitrifcation

process

The system worked in denitrification of a synthetic substrate. The biodenitrification was realised by an activated sludge, mixed culture, taken from the aeration tank from the municipal wastewater treatment plant in Montpellier (France) and acclimated to the synthetic substrate used in the experiments. The synthetic substrate was prepared by diluting potassium nitrate and ethanol in tap water so that the concentrations were 200 mg,,,/l and 1000 mg,n/l, respectively. In these conditions, the ratio COD/N was equal to 5. (NH,),HPO, was also added so that COD/P = 150. The reaction of denitrification can be written as follows: SCH,CH,OH + 12 NO;+ + 9 H,O + 2 CO,

10 HCO,- + 6 N,

Hydroxide and hydrogenocarbonate ions are metabolically produced by the reaction of denitrification. In theory, the increase of alkalinity is equal to 3.6 mg CaCO3/mg N-NO,--N denitrified. The bioreactor pH value could increase to 8.5 or more when no acidic solution was added to the substrate (consequently, chlorydic acid was added to the substrate to maintain the pH value between 7.5 and 8). By keeping the biological parameters (hydraulic retention time, sludge retention time, organic loading) constant, the biomass concentration was stable, equal to a constant value of 1.5 g&l. Table 1 presents the main biological characteristics of the system. 2.3. Fouling characterisation TMP evolution was monitored in the MBR by recording data of pressure transducers P,, P,

S. Ognier et al. /Desalination 146 (2002) 141-147 Table I Biological Volumetric

l

conditions loading rate, gcoD/lld

3

Hydraulic retention time, h

8

Sludge retention time, d

3.5

Specific denitrification

rate, gNlgv&d

Yield, g&gco, Biomass concentration,

0.4 0.2

g&l

1.5

and P,. P, and P, are the pressures of the retentate measured at the input and the output of the membrane module and P, is the pressure on the

permeate side. During the filtration operation, P, and P, are constant and P, decreases due to membrane fouling. TMP was calculated by the relation:

To characterise the nature of the fouling, several cleaning methods were tested: intermittent filtration, forward flush with water, back flush with water and slight acid cleaning. Except for the intermittent filtration, the membrane resistance to a water permeation was determined after each cleaning method. TMP was measured when filtering pure water at lo,20 and 30 l.m-*.h-‘.The tangential velocity was the one used during the filtration operation. Details of the different methods in chronological order of their applications are as follows: l Intermittent filtration. The suction pump was switched off for approximately 10 min, then the filtration operation was reinitiated. During the intermittent filtration, the recirculation pump continued to make the retentate circulate. l Forward flush. The filtration ws stopped when the reactor was filled up with pure water. Then, the water was recirculated for 10 min without filtering. l Back flush. Pure water was filtered in the opposite direction of the normal filtration operation with a TMP of 0.5 bar.

l

143

Acidic cleaning. Last, the membrane was chemically cleaned with a slightly diluted solution of nitric acid (HNO,) at room temperature. Complete chemical cleaning. To restore the initial permeability of the membrane, a complete chemical cleaning was done. An alkaline solution (diluted hydroxide sodium solution) and then an acid solution (diluted nitric acid solution) were filtered at 60°C.

3. Results and discussion 3. I. Definition of operating conditions ensuring process stability The operating conditions of the MBR were defined under “standard” conditions of pH and temperature, that is to say, a pH value between 7.5 and 8 and a temperature equal to 25kl”C. The objective of this preliminary study was to define hydro-dynamic conditions where no deposition of colloidal particles on the membrane occurred. Therefore, the increase of membrane resistance is controlled and a stable regime should be obtained. In theory, such conditions are possible when the permeate flux is inferior to the critical flux value. To determinate the critical flux value in the MBR, the stepwise method was used. The permeate flux was stepwise increased with a step length of 30 min. Below the critical flux value, the TMP stabilised rapidly after each flux increase, and the stabilised value of TMP increased linearly with the flux imposed. Above the critical flux value, this linear relationship did not apply any more due to a deposition phenomenon. Fig. 1 shows the TMP measured for each flux value. The subcritical regime corresponds to the first part of the curve where the resistance stays constant (2x10’* m-l) for permeate flux values below 38 l.m-*.h-‘. This resistance differed from the clean membrane one due to an instantaneous fouling phenomenon taking place at the very beginning of the filtration. Above 38 l.m-2.h-‘, the fouling resistance increased

S. Ognier et al. /Desalination 146 (2002) 141-147

Fig. I. Flux of permeate vs. TMI?

dramatically, indicating that the filtration regime was supracritical. Considering these results, it could be concluded that a 10 l.m-*.h-’ value for the permeate flux, less than the critical value, would ensure the process stability. However, the results obtained with the stepwise method did not mean that there was no membrane fouling when the flux of permeate was below the critical value. Actually, a slow fouling phenomenon could not necessarily be observed with the stepwise method due to the relatively short duration of the steps and the precision of the measurements. To ensure that the conditions were stable under standard conditions of pH and temperature, longterm experiments were conducted. During the whole filtration run (5 weeks), the fouling rate was very low, with an increasing rate of the resistance below 10” m-‘.d-’ (less than 0.003 bar/d). But this resistance proved to be irreversible as neither a forward flush, nor a back flush were efficient. So, a permeate flux value of 10 l.m-2.h-1 allowed a stable fouling regime.

removal efficiencies stayed almost constant for all the physicochemical conditions tested. To study the influence of pH on process stability, the pH value in the bioreactor was allowed to vary between 7.5 and 9. This was done by adjusting the substrate pH value to different values (the added volume of chlorydric acid changed). It was observed that, when the value of the pH increased too much, a continuous increase of the TMP could be observed. This phenomenon is illustrated in Fig. 2 where the TMP evolution is shown with time and with pH variations. In this experiment, the temperature was equal to 25kl”C. Continuous increases of the TMP can be observed; these are numbered on the graph from 1 to 4. These TMP always increases corresponded to pH values greater than 8.5. However, if the suspension pH was decreased by acidifying the substrate, the system could become stable again and the fouling was partly removed. Additional experiments were conducted to study the effect of temperature that was varied from 22 to 32°C by adjusting the cooling system. As mentioned previously, TMP increases were observed when the pH values were shown to depend on temperature (Fig. 3): for high temperatures, the fouling occurred for pH values inferior to those for low temperatures. To chatacterise the phenomenon leading to these rapid TMP increase, the nature of the fouling resistance was investigated by testing several

-9

3.2. Analysis of the influence of physicochemical conditions on process stability In the range of pH values tested, it is reported in literature that the biological activity does not undergo significant changes [3]. Similar results were obtained in the present srudy since VSS concentration, soluble COD concentration and

8

Fig. 2. Influence of pH increases on TARPevolution.

I,

S. Ognier et al. /Desalination

I

15

17

19

21

23

25

Fig. 3. Influence

of temperature

29

27

Temperature

31

33

35

(“C)

on critical pH value.

cleaning methods. Two experiments were conducted: (1) Case A was initiated during an intensive fouling phase, (2) Case B was done at the end of the filtration run presented in Fig. 2, when the process was stabilised again (t = 575 h). The results are presented in Table 2. As shown by these results, filtration resting is totally ineffective in removing the fouling resistance in both cases. This result signifies that the fouling mechanism is not the formation of a reversible deposit on the membrane surface. However, in the first case, half of the fouling resistance can be eliminated by the forward flush. As the cleaning methods of filtration resting and forward flush differ only in the use of water (the hydrodynamic conditions are identical), the cleaning efficiency of the forward flush points out once again the crucial role of physicochemical conditions. The forward flush effectiveness could be due to the use of water with a neutral pH value. The importance of the pH value in fouling removal had been already noticed during the experiments. This result was confirmed by the resistance values Table 2 Membrane

Case A Case B

resistance

145

146 (2002) 141-147

obtained after acid cleaning: in both cases, acid cleaning proved to be very efficient in removing the fouling resistance that remained after the forward flush. The irreversible fouling can be of organic (biofilm, metabolites, etc.) or inorganic (precipitated salts) nature. Alkaline cleaners are generally considered as the most effective against biofilms and organic foulants whereas acidic cleaning is required to ensure the removal of inorganic precipitants [4]. Therefore, the effectiveness of the acidic cleaning indicates that the fouling could be mainly due to precipitation phenomena. The continuous fouling increase observed is not in disagreement with this hypothesis. Actually, if precipitation can be instantaneous, the continuous feed of hard tap water and the biological reaction can induce continuously a salt precipitation as long as the suspension pH is beyond the critical value for precipitation. To determine the nature of the precipitants, the influence of pH on the suspension composition was analysed. When the pH was increased, only hydroxide ions (OH-) and carbonate ions (CO:-) concentrations were increased. Therefore, the precipitation was supposed to depend on the concentrations of hydroxide and/or carbonate ions. As the substrate was prepared with hard tap water (Ca2+= 120 mg/l and Mg*+ = 8 mg/l), the reaction quotient was compared with the solubility product for the main hydroxide and carbonate precipitates involving calcium and magnesium ions. Table 3 presents the solubility product at 25°C the reaction quotient calculated at pH 9 with the calcium and magnesium concentrations in the tap water used for the substrate. These calculations indicate that two inorganic crystals, CaCO, and Mg(OH),, can precipitate

values obtained after the different cleaning methods tested (m-l)

Before intermittent filtration

After intermittent filtration

After forward

20x10” 9x1012

20x10” 9x1012

10x10’* 6.5. lOI

flush

After back flush

After acidic cleaning

8.6x10’* 6.5~10’~

1.9x10’* 1.8x1o’2

146

S. Ognier et al. /Desalination

146 (2002) 141-147

Table 3 Solubility products at 25°C and reaction quotient calculated at pH 9

Case A Case B

Before intermittent filtration

After intermittent filtration

After forward flush

After back flush

After acidic cleaning

20x1 012 9x1012

2ox10’2 9x1012

1ox1o’2 6.5.10’*

86x10’* 6.5~10’~

1.9x10t2 1.8~10’~

when the pH value exceedes 9. In the case of CaCO, precipitation, the theoretical “critical” pH value is given by the relation where KS is the solubility product of CaCO, and K3 the constant . . dissociation of HCO,-: pH,,.,, = -log [[ Ca ‘+]x [HCO;]x+‘j s

Contrary to other common precipitates, CaCO, is characterised by the decrease of its solubility when the temperature is increased. So, in the case of CaCO,, the value of pHc,,, decreases when the temperature is increased. This decrease of pHo,, related with the temperature could not be observed with the other salts present in the suspension. It can be therefore concluded that the decrease of the pH value corresponding to the strong increase of TMP when the temperature is increased, is in favour of CaCO, precipitation. There are two ways to explain the resistance increase due to the CaCO, formation: (i) the crystal particles are formed in the bulk phase (bulk crystallisation) and deposit on the membrane or (ii) the crystal grows on the surface of the membrane material (heterogeneous crystallisation). Even if the hydrodynamic cleaning methods of intermittent filtration and back flush proved to be inefficient to remove the fouling resistance, the hypothesis

of a deposition due to bulk crystallisation cannot be eliminated. Actually, heterogeneous crystallisation could take place on the mineral deposit present on the membrane and therefore induce chemical binding between particles. It was reported elsewhere that a cake of CaCO, particles can be irreversible due to the cohesive properties of the particles [5]. To further investigate the fouling mechanism, the intensive fouling phases numbered 1,2,3 and 4 were analysed by using usual filtration models: standard blocking law, complete blocking law, intermediate blocking law and cake filtration law [6]. At first, these models have been developed for frontal filtration mode: back-transport forces are not taken into consideration. The results obtained for the first fouling phase are presented in Table 4. Except for the complete blocking law (the hypothesis of this mode1 are generally reported as too restrictive), it is shown that the models fit the experimental data quite well. This result is coherent with the building up of an irreversible deposit: no particles back-transport is possible due to the severe particles adhesion, so the models apply satisfactorily. However, one can note that the agreement is much better when the cake filtration and the intermediate blocking law are

Table 4 Analysis of 1st fouling phase by using usual filtration laws

Linear relation Correlation coefficient

Standard blocking law

Complete blocking law

Intermediate blocking law

Cake filtration law

Rm”*vs. V 0.963

R-’ vs. V 0.918

InR vs. v 0.990

R vs. V 0.983

S. Ognier et al. /Desalination

considered. Considering that better fittings are obtained with models describing an external fouling mechanism, the fouling would be located at the pore entrance or on the membrane surface rather than in the whole membrane matrix. This result allows one to think that the fouling could be caused by the deposition of CaCO, particles formed in the bulk suspension (bulk crystallisation) on the membrane surface. Actually, a pore constriction mechanism should have been obtained with heterogeneous crystallisation. However, it is of course difficult to base conclusions on the only use of the models and further research would be necessary to confirm this hypothesis.

146 (2002) 141-147

l

The fouling mechanism could be the deposition of CaCO, particles formed in the bulk suspension by bulk crystallisation. Further research would be necessary to confirm this hypothesis.

References [I] J.A. Howell, Subcritical flux operation of micro[2]

[3]

4. Conclusions An MBR for denitrification was studied in terms of process stability. Unusual membrane fouling in an MBR system was investigated. The following conclusions could be drawn: l In an MBR for denitrification, the great alkalinity of the suspension can cause the precipitation of calcium carbonate for pH values between 8 and 9. l The role of precipitation can be pointed out as a cause of system instability, even if the system works in subcritical conditions.

147

[4]

[5]

[6]

filtration, J. Membr. Sci., 107 (1995) 165-171. C. Wrsniewski, F. Persin, T. Cherif, R. Sandeaux, A. Grasmick and C. Gavach, Denitrification of drinking water by the association of an electrodialysis and a membrane bioreactor: feasability and application, Desalination, 139 (2001) 199-205. X.-J. Fan, V. Urbain, Y. Qian and J. Manem, Ultrafiltration of activated sludge with ceramic membranes in a cross-flow membrane bioreactor process, Water Sci. Technol., 41(10-l 1) (2000) 243-250. R. Liikanen, J. Yli-Kuivila, R. Laukkanen, Efficiency of various chemical cleanings for nanofiltration membrane fouled by conventionally-treated surface water, J. Membrane Sci., 195 (2002) 265-276. A. Ould-Dris, M.Y. Jaffrin, D. Si-Hassen and Y. Neggaz, Analysis of cake build-up and removal in cross-flow microfiltration of CaCO, suspensions under varying conditions, J. Membr. Sci., 175 (2000) 267-283. J.A. Suarez and J.M. Veza, Dead-end microfiltration as advanced treatment for wastewater, Desalination, 127 (2000) 47-58.