Journal of Membrane Science 375 (2011) 268–275
Contents lists available at ScienceDirect
Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci
Ultrafiltration as an advanced tertiary treatment of anaerobically digested swine manure liquid fraction: A practical and theoretical study Raquel López-Fernández a,∗∗ , Carolina Aristizábal a , Rubén Irusta a,b,∗ a
Environmental Division, CARTIF Centro Tecnológico, Parque Tecnológico de Boecillo, 205, 47151 Valladolid, Spain Department of Chemical Engineering and Environmental Technology, Higher Technical College of Industrial Engineering, University of Valladolid, Paseo del Cauce 59, 47011 Valladolid, Spain b
a r t i c l e
i n f o
Article history: Received 20 September 2010 Received in revised form 22 March 2011 Accepted 27 March 2011 Available online 6 April 2011 Keywords: Anaerobic digestion EGSB Specific cake resistance Swine manure Ultrafiltration
a b s t r a c t The removal of the organic matter of the swine manure liquid fraction has been carried out by an integral process based on two stages: anaerobic digestion in an expanded granular sludge bed (EGSB) reactor followed by ultrafiltration (UF) as a tertiary treatment. The lab-scale EGSB reactor was operating for 39 weeks with a hydraulic retention time (HRT) of 3.8 days; leading to a 70% tCOD removal and a biogas production of 0.15 Nm3 CH4 kg−1 tCOD removed. The UF process was studied in two different geometry and configuration membrane modules: external tubular (E-T) and submerged hollow fiber (S-HF). Both lab-scale systems have been compared in terms of the filtration selectivity and productivity and the S-HF has been the most selective and productive configuration in the filtration of the EGSB effluent. The whole process (EGSB + S-HF) provides a permeate flow without solids and yields a tCOD removal around 90%. Finally this paper proposes a satisfactory model for the UF of the EGSB effluent in the S-HF module. Both the membrane intrinsic resistance (RM ) and the specific cake resistance (˛) have been calculated. © 2011 Elsevier B.V. All rights reserved.
1. Introduction Traditionally swine manure has been used as a direct fertiliser on agricultural land. However, due to the present trend of raising large herds of livestock in smaller areas, the soil available is not often enough to spill all the manure produced. Nutrients such as nitrogen or phosphorous and the organic matter are the most problematic components of the manure. While the formers are culprits for contamination of soil and groundwater and for eutrophication of surface waters, the latter is believed to be responsible for acute water pollution incidents and for odour problems [1,2]. Anaerobic digestion is an established bioconversion technology for the organic fraction with simultaneous production of biogas. Most of the current full scale plants operate with anaerobic codigestion of pig manure with other organic wastes [3]. However, the degradation has proven to be difficult mainly due to the inhibition of the methanogenic bacteria by high amounts of ammonia in the digester [4,5]. Thus, the treated effluents use to contain high amounts of suspended solids or persistent organic substrates,
∗ Corresponding author at: Department of Chemical Engineering and Environmental Technology, Higher Technical College of Industrial Engineering, University of Valladolid, Paseo del Cauce 59, 47011 Valladolid, Spain. Tel.: +34 983 546504; fax: +34 983 546521. ∗∗ Co-corresponding author. E-mail addresses:
[email protected] (R. López-Fernández),
[email protected] (R. Irusta). 0376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2011.03.051
which prevent them from being directly used for irrigation or spread on soils. In order to improve the efficiency of the process and increase the biogas production and nutrient removal, more sophisticated treatment systems are being implemented [6]. The use of anaerobic digestion in combination with membrane technology is an alternative to achieve retention of microorganisms and allow operation with high biomass concentrations. This design was known under the acronym ADUF (Anaerobic Digestion with Ultrafiltration) and was originally intended at the early 90s as a treatment solution for highly COD laden wastewater like brewery and starch effluent [7]. The ADUF system has been applied for the treatment of pig manure at pilot scale in the BIOREK® concept [8]. The process consists of a digester (mesophilic or thermophilic) coupled to an external tubular UF membrane, leading to an anaerobic membrane bioreactor (AnMBR). The integral treatment is completed by an ammonia stripping step and a final reverse osmosis (RO) to achieve 90% COD removal and up to 99.9% ammonia removal efficiencies. It has also been studied the performance of a two stage anaerobic system based on an acidogenic reactor with a submerged UF membrane followed by a methanogenic reactor, with a lower part design as an upflow anaerobic sludge blanket (UASB) reactor. COD removal efficiency in this case was 80% and methane gas production was 0.32 Nm3 CH4 kg−1 tCOD removed [9]. The expanded granular sludge bed (EGSB) reactors were developed to overcome problems that can occur in the UASB such as preferential flows, hydraulic short cuts and dead zones. The use of higher upflow liquid velocities (4–10 m h−1 ) and higher
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
height/diameter ratios permit the partial expansion (fluidization) of the granular sludge bed, improving the biomass–wastewater contact. Soluble pollutants are efficiently treated in EGSB reactors but suspended solids are not substantially removed from the wastewater stream due to the high upflow velocities applied [10]. EGSB reactors have recently been applied to treat brewery, starch or palm oil mill wastewater [11]. EGSB reactors have shown to be suitable for the treatment of biodegradable toxic of inhibitory compounds such as metals [12]. Since they operate with high recirculation ratios, the inlet is dilute to levels no longer dangerous for bacterial activity [13]. Basing on the capacity of these reactors to overcome inhibition, this work proposes the use of an EGSB reactor for the anaerobic digestion of swine manure where, as it has been already commented, the ammonia interference is one of the major problems [4,5]. As far as we are concerned, EGSB reactors have never been used in this application, so one of the aims of the present research will be to explore the feasibility of this novel treatment. The treatment of the liquid fraction of the swine manure (hereinafter called LFSM for simplicity) has been completed with a UF membrane in order to prevent the non-degraded suspended solids washout (Fig. 1). This membrane-coupling system could be an alternative for the effective treatment compared to the already cited, AnMBR in the BIOREK® process, or the two stage anaerobic system based on UASB reactors and UF membranes. Compared to these anaerobic digesters, higher loading rates are expected to be achieved in the EGSB reactor. There is a precedent of this membrane-coupled EGSB reactor proposed herein, but for treating domestic wastewater under moderate to low temperature at lab scale [14]. This paper is specially focused on the UF process applied as tertiary treatment of the EGSB effluent, as it is often the critical step because of the membrane clogging and the viscosity effect on mass transfer. Nevertheless, the final objective of this study is the integration of both treatments (EGSB and UF) in an AnMBR reactor which will be further achieved by the UF concentrate recirculation to the feed of the anaerobic reactor (drawn as a broken line in Fig. 1). 2. Materials and methods 2.1. Feed swine manure Swine manure was collected from a farm located in Avila (Spain). The waste was filtered by a 0.5 mm screen to remove large particles
269
Table 1 Liquid fraction swine manure (LFSM) characterisation. Parameter
Unit
Average value ± SD
pH Total alkalinity tCOD sCOD BOD5 TS VS TSS VSS TKN NH4 -N Total P
– g L−1 g L−1 g L−1 g L−1 g L−1 g L−1 g L−1 g L−1 g L−1 g L−1 g L−1
8.01 6.888 9.433 3.194 3.950 14.494 9.501 8.157 5.228 1.710 1.413 0.203
± ± ± ± ± ± ± ± ± ± ± ±
0.12 0.207 0.472 0.160 0.592 0.725 0.475 0.408 0.261 0.068 0.071 0.010
and characterised according to the standard methods (Table 1) [15]. Then, it was stored at 4 ◦ C prior to use. Special attention should be paid to the NH4 -N concentration of 1.4 g L−1 , unusually low due to the presence of seaweed in the feeding of the pigs from this current farm, that increases the nitrogen adsorption (reduction at origin). This value is below the estimated limit for inhibition of unadapted methanogenic cultures (1.5–2.5 g L−1 ) that could increase up to 4 g L−1 after adaptation [5]. 2.2. EGSB reactor The EGSB reactor was constructed of Plexiglas with a working volume of 3.2 L (52 mm i.d. and with a quiescent region of 220 mm i.d. at the top). The total volume of the reactor, including the sedimentation part, was 12.2 L. The reactor was installed in a chamber room maintained at a constant temperature of 30 ◦ C. A recirculation pump was used to expand the bed and to control the up flow velocity. Biogas production was measured with an inverted tube (Fig. 2). Methane in the biogas and the volatile fatty acids (VFA) were both analysed by gas chromatography with a flame ionisation detector (FID). The reactor was inoculated with 3 L of granular sludge taken from an UASB reactor treating wastewater from a paper mill factory to achieve a final sludge concentration of 14.2 g TS L−1 and 9.9 g VS L−1 . The collapsed bed height was 99 cm and the expanded height 150 cm. After inoculating the reactor, the granular sludge was acclimated in a batch process with the LFSM for two weeks. Macro and micronutrients were added as follows: (NH4 )2 HPO4 (0.25 g L−1 ); KH2 PO4 (0.02 g L−1 ); MgSO4 ·7H2 O (0.0015 g L−1 ); CaCl2 (0.001 g L−1 ); H3 BO3 (0.005 g L−1 ); EDTA
Fig. 1. Treatment of the liquid fraction of the swine manure by using EGSB and UF technologies.
270
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
Fig. 4. Submerged hollow fiber (S-HF) module.
Fig. 2. Lab scale EGSB reactor.
(0.1 g L−1 ). At the same time, NaHCO3 was added for buffering capacity. 2.3. Ultrafiltration modules 2.3.1. External-tubular (E-T) membrane module The E-T module consisted of an external monotubular membrane provided by Koch Membrane Systems (CTG-HFM 251 model). It was made in polyethersulfone (PES), with a mean molecular weight cut-off (MWCO) of 100 kDa and a filtration area of 0.017 m2 . The unit had a feed tank (T1) of 20 L, a permeate tank (T2) of 10 L, a recirculation pump (P1), a heat exchanger (HE), four needle valves (V1–V4) to control the flow, a temperature indicator (TI), two flowmeters (FI1 and FI2) and three manometers (PI1–PI3) (Fig. 3). 2.3.2. Submerged hollow fiber (S-HF) membrane module The S-HF module consisted of a hollow fiber membrane provided by Zenon – GE Water Process & Technologies (ZW-10 model).
It was made in polyvinylidene fluoride (PVDF), with a nominal pore size of 0.04 m and a filtration area of 0.93 m2 . It was submerged in a 32 L polyethylene tank (T1). Permeate line was composed by a suction pump (P1), a flowmeter (FI1) and a manometer (PI1). An aeration system and a chemical cleaning device were also installed, with no temperature control (Fig. 4). 2.3.3. Experimental procedure The characterisation of the E-T and the S-HF membranes was carried out by the permeability tests with water and diluted LFSM at various temperatures. The latter dilution was calculated by assuming a 90% COD removal of the previous anaerobic treatment. Finally, these tests were conducted with the effluent from the EGSB reactor at 30 ◦ C in order to select the most suitable module for the tertiary treatment. All the filtration experiments were carried out with total permeate recirculation (total reflux) so the volume in the feed tanks was kept constant. For the E-T module, the experimental was carried out by modifying TMP and measuring Jp , keeping the feed flow rate constant (QF = 1200 L h−1 ). The filtration in the S-HF module was achieved by a slight negative vacuum and the JP vs. TMP curves were obtained by modifying Jp and measuring TMP. A constant aeration of 2 Nm3 h−1 was maintained in the feed tank during all the experimentation. To ensure that the performance of the membranes modules was almost the same in all the study, cleaning was performed after every experiment. The E-T membrane was chemically cleaned by a basic and a basic-hypochlorite (200 mg L−1 ) cycle. If necessary, a final acid/basic cycle was carried out. The S-HF membrane was cleaned by submerging the module in a 500 mg L−1 NaClO solution for 30 min. If necessary, it was finally treated with a 1000 mg L−1 citric acid solution. 3. Results and discussion 3.1. EGSB reactor
Fig. 3. External tubular (E-T) module.
The EGSB reactor has been operating for 39 weeks with a hydraulic retention time (HRT) of 3.8 days at 30 ◦ C. During the startup, the volumetric loading rate (VLR) was gradually increased from 0.44 to 2.76 g tCOD L−1 d−1 (Table 2). This variation was imposed to the process by varying the feed concentration while keeping a constant feed flow of 0.842 L d−1 . During the last step, the raw LFSM was used directly without diluting. The up flow velocity was increased from 4 m h−1 to 8 m h−1 .
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
271
Fig. 5. COD removal evolution in the EGSB reactor.
Table 2 EGSB operation parameters.
tively. Regarding the S-HF module, the AW were 292.2; 294.8; 322.8; 355.5 and 368.0 L h−1 m−2 bar−1 at 15, 20, 25, 30 and 35 ◦ C.
Stage
Week
tCOD (g L−1 )
sCOD (g L−1 )
VLR (g L−1 d−1 )
1 2 3 4 5 6
1–5 6–11 12–14 15–17 18–32 33–39
1.7 3.7 5.4 6.1 6.7 10.5
0.5 1.2 1.3 1.3 1.8 2.5
0.44 0.97 1.42 1.60 1.76 2.76
The pH and the VFA/alkalinity ratio (R) were kept stable during all the different operation stages: pH ≈ 7.5 and R < 0.3. Alkalinity in the reactor increased gradually from 2.5 to 7.5 g CaCO3 L−1 so the amount of added NaHCO3 was reduced from the initial 1 g g−1 COD to 0.50 by the end of the 4th stage. Suspended solid removals were about 75% and tCOD about 70% (Fig. 5). This moderate efficiency was due to the low biodegradable fraction of the swine and the subsequent solid accumulation, as reported in Refs. [2,4–6]. Biogas composition was 66% of methane and 34% of carbon dioxide. Finally, values about 0.15 Nm3 of CH4 kg−1 tCOD removed were obtained. 3.2. Ultrafiltration modules comparison 3.2.1. Membranes permeability with water (AW ) Water permeability test results for E-T and S-HF membranes are shown in Fig. 6. For the E-T module, AW values of 887.8; 986.0 and 1081.8 L h−1 m−2 bar−1 were obtained at 25, 30 and 35 ◦ C respec-
3.2.2. Membranes permeability and selectivity with diluted LFSM The UF of the diluted LFSM (tCOD = 0.90–0.99 g L−1 and sCOD = 0.34–0.39 g L−1 ) was carried out. For the E-T module the linearity of Darcy’s law was only applicable for pressures below a critical TMP value (TMPC,SM ) [16], and higher pressures did not undergo significant increases in the JP (Fig. 7a). It was obtained a TMPC,SM ≈ 1.4 bar for the three temperatures analysed. ASM values of 57.1; 60.7 and 62.6 L h−1 m−2 bar−1 were obtained at 25, 30 and 35 ◦ C respectively. When the filtration was carried out in the SHF module, the linear relationship was maintained in the entire interval above 30 L h−1 m−2 for the three temperatures as can be seen in Fig. 7b. The ASM values obtained were 212.0; 248.8 and 282.4 L h−1 m−2 bar−1 at 18, 25 and 29 ◦ C respectively. The influence of the temperature on the permeability values AW and ASM for each module has been analysed (Fig. 8). As expected, a linear relationship (A over the T) was found in all the cases and increasing the temperature also increased the membrane permeability in both configurations. Comparing the results obtained in Fig. 8, it is remarkable that in the E-T module, the permeability decreased drastically (93%) when diluted LFSM was filtered instead of water. However, it only decreased around 25% for the S-HF module. This behaviour could be explained attending to fouling trend of the each membrane, that is a complex matter, and depends on many factors such as mem-
Fig. 6. Water permeability tests: E-T (a) and S-HF (b).
272
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
Fig. 7. Permeability tests with diluted LFSM: E-T (a) and S-HF (b).
brane configuration or operational conditions [17]. Among these factors, one of the most important is the membrane material, that is different in our study being the E-T module made in PES and the S-HF in PVDF. Those two membranes are different in pore size, morphology and hydrophobicity, so their fouling propensity was different. It has been recently published the fouling comparison of two UF membranes made in PES and PVDF and it was observed that the EPS had higher affinity with the PES membrane [18]. This result is in concordance with our experimental observations, where the PVDF membrane is the more fouling resistant. The selectivity of the filtration of the diluted LFSM is directly associated with the membrane rejection and is also function of the operational conditions. In all the cases, the ultrafiltration provides a permeate flow without solids and the rejection values obtained for both modules are shown in Fig. 9. In the S-HF module, the increase in JP slightly favoured the selectivity of the filtration. On the other hand, increasing the temperature, the rejections decreased. However, the selectivity behaviour in the E-T system was not clearly defined and the results obtained were quite random. Finally, rejection values in the S-HF were higher than in the E-T: while the former showed a COD removal about 70%, the latter gave values in the interval 55–70%. 3.2.3. Membranes permeability and selectivity with the EGSB reactor effluent During the EGSB operation, the effluent was stored in a cool chamber (4 ◦ C). Before being used in the UF experiments, it
Fig. 8. Temperature dependence on membrane permeability (AW and ASM ).
was warmed at 27–30 ◦ C and characterised (tCOD = 1.978 g L−1 , sCOD = 0.832 g L−1 and SST = 1.247 g L−1 ). The filtration was carried out in both modules at total reflux conditions and at 30 ◦ C. In the E-T module, it was obtained a TMPC,EGSB of 1.3 bar and the AEGSB of 16.27 L h−1 m−2 bar−1 . The filtration provides a permeate flow without solids and yields to a tCOD removal around 60–65% (Fig. 10a). When the filtration was carried out in the S-HF module, it was obtained a JC,EGSB of 20 L m−2 h−1 , the AEGSB was 241.6 L h−1 m−2 bar−1 and tCOD removal about 70–75% (Fig. 10b). In concordance with the previous experimentation, the S-HF membrane was the more productive and selective unit when the digested effluent from the EGSB reactor was ultrafiltrated. If the combined treatment is considered (anaerobic digestion in the EGSB reactor followed by UF in the S-HF module) a tCOD removal of 90% was obtained, a high value compared to others habitually achieved by anaerobic conventional treatments for these types of effluents [19]. The highest loading rate used in the EGSB reactor was 2.76 g COD L−1 d−1 (see Table 2) and 3.8 days of HRT. These operational conditions and tCOD removal efficiencies are similar to the already cited BIOREK® process, that operates with a VLR of 3 g COD L−1 d−1 , a HRT of 6 days and yields to a 90% tCOD removal [8]. In the case of the two stage anaerobic system coupled to a UF membrane, these values were 3.75 g COD L−1 d−1 , 1–2 days, and 80% tCOD removal respectively [9].
Fig. 9. E-T and S-HF filtration selectivity.
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
273
Fig. 10. UF of the EGSB effluent: E-T (a) and S-HF (b).
3.3. Theoretical analysis of the filtration of the EGSB reactor effluent in the S-HF module In order to carry out this analysis, the continuous filtration of the EGSB effluent in the S-HF module was studied at different permeate fluxes (JP1 = 3.91; JP2 = 9.35 and JP3 = 14.71 L h−1 m−2 ) at 30 ◦ C and without gas sparging, which is equivalent to conventional dead end filtration. As expected, the TMP increased faster when the JP applied was higher (Fig. 11), specially for the two higher JP values. This rapid increase of the TMP is explained because of the dead end filtration conditions with no aeration provided in the membrane tank. On the other hand, rejection values were in the interval 70–80% and increased slightly with the time. The JP applied did not show a clear effect on the selectivity of the filtration as can be seen in Fig. 11. In concordance to the filtration theory, the transmembrane pressure TMP has a linear relationship with the time (Eq. (1)): TMP − TMP0 = · ˛ · us · ω(t)
(1)
where TMP0 is the initial TMP, is the fluid viscosity, ˛ is the specific cake resistance, us is the superficial velocity of filtration and ω(t) is the mass of solids per unit area in the cake. The initial TMP is proportional to , the membrane intrinsic resistance of the hollow fiber membrane (RM ), and us (i.e. the permeate flux, Jp ) (Eq. (2)): TMP0 = · RM · us
(2)
In order to find RM , the results obtained in the water permeability test at 30 ◦ C shown in Fig. 6b (temperature at which the effluent
Fig. 11. Continuous dead end filtration of the EGSB effluent in the S-HF module.
of the EGSB was filtered), were analysed. The slope on the TMP0 vs. us (JP ) representation enables to know this parameter according to Eq. (2). A linear regression was fitted to the data (correlation coefficient R = 0.997) and all the experimental TMP0 values were inside the 95% confidence interval (represented as grey lines in Fig. 12). Using the straight line slope value (1.010 × 108 ± 6.94 × 107 Pa s m−1 ) and the water viscosity at 30 ◦ C, = 0.7977 × 10−3 kg m−1 s−1 , a final value of RM = 1.266 × 1012 m−1 was calculated (Eq. (2)). When the effluent from the EGSB reactor was filtrated in the S-HF module at 30 ◦ C (see Fig. 11), the membrane resistance (RM ) is only a component of the total resistance which includes also the fouling and the concentration polarisation phenomenon. For a compressible cake, Eq. (1) is applicable during all the experimentation. Then, combining Eqs. (1) and (2): TMP = · RM · us + · ˛ · us · ω(t)
(3)
The cumulative specific mass of solids in the cake depositing on membrane surface ω(t), can be found through the unsteady state mass balance of the system comprised the membrane tank including the permeate recirculation (Eq. (4)): ω(t) =
VC0 (1 − e−(QF /V )t ) aF
(4)
where V is the tank volume (0.032 m3 ), C0 is the initial tCOD (tCOD0 = 2.024 kg m−3 ), aF is the membrane filtration area (0.93 m2 ), QF is the volumetric flow filtered (three values have been tested corresponding with the three experiments shown in Fig. 11). Therefore, the TMP vs. ω(t) plot gives the membrane resistance (RM ) and the effluent viscosity () from the y-intercept (see Eq. (3)). The slope is related to the specific resistance (˛), which represents the effect of gradually accumulating layers of solid in the absence of stirring. In Figs. 13–15 it can be seen the evolution of TMP over the time for the three flows tested.
Fig. 12. Correlation of TMP0 with the flux using the water permeability tests at 30 ◦ C in the S-HF module.
274
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
Fig. 13. Correlation of TMP with ω(t) for QF = 3.64 L h−1 .
Fig. 16. Compressible cake model behaviour for QF = 8.69 L h−1 .
Fig. 14. Correlation of TMP with ω(t) for QF = 8.69 L h−1 .
Fig. 17. Compressible cake model behaviour for QF = 13.68 L h−1 .
simplicity is the following two-parametric potential model: ˛ = ˛0 (TMP)n
(5)
where n is the compressibility coefficient (n = 0 when the cake is incompressible). Combining Eqs. (1) and (5) and integrating: (TMP − TMP0 )1−n = · us · ω(t) ˛0 · (1 − n)
(6)
By taking logarithm, Eq. (6) can be transformed into a linear relationship as follows:
Fig. 15. . Correlation of TMP with ω(t) for QF = 13.68 L h−1 .
Ln[TMP − TMP0 ] =
As it can be seen in Fig. 13, a linear regression was fitted to the data using a QF = 3.64 L h−1 (correlation coefficient R = 0.9924) and all the experimental TMP values were inside the 95% confidence interval. The viscosity of the EGSB effluent at 30 ◦ C could be calculated from the y-intercept (3610.29 ± 58.69 Pa), assuming the RM determined in the experiment with water at this temperature (see Fig. 12). The value of ˛ = 1.607 × 1014 m kg−1 was obtained from the slope (4.582 × 105 ± 3.26 × 104 m s−2 ). Nevertheless, when the filtration was carried out at higher flows (QF = 8.69 and 13.68 L h−1 ) there is not a linear relationship between TMP and ω(t) (Figs. 14 and 15). This could be explained because the cake began to be compressible and, as a consequence, the specific cake resistance (˛) was a function of the transmembrane pressure applied. Several models explain this effect but the most used for its
1 1 Ln[˛0 · (1 − n)] + Ln[ · us · ω(t)] 1−n 1−n (7)
The constants ˛0 and n are two adjustable parameters according to the experimental conditions. Therefore, the Ln(TMP − TMP0 ) vs. Ln[·us ·ω(t)] plot gives the n parameter from the slope and the specific resistance (˛0 ) from the y-intercept (Figs. 16 and 17). The parameters obtained from the aforementioned compressible cake model are summarised in Table 3: The degree of compressibility is given by the value of n in the Eq. (7). When n < 1, the cake has a low/moderate compressibility [20]. So, in this case, according to the n values the compressibility was moderate, and it increased with the flow rate as expected. Regard-
Table 3 ˛0 and n values. Flow rate (L h−1 )
Slope
Confidence interval (95%)
n
y-Intercept
Confidence interval (95%)
˛0 (m kg−1 )
8.69 13.68
1.6218 1.7695
(1.523–1.721) (1.693–1.846)
0.383 0.435
48.672 52.053
(46.268–51.076) (50.226–53.880)
1.752 × 1013 1.055 × 1013
R. López-Fernández et al. / Journal of Membrane Science 375 (2011) 268–275
ing the specific cake resistance (˛0 ), the trend was the opposite and it decreased when the flow rate increased. 4. Conclusions (1) A lab-scale EGSB reactor has been innovatively applied for the treatment of the LFSM. Yields of 60–70% tCOD removal and a biogas production of 0.15 Nm3 CH4 kg−1 tCOD removed were obtained, operating with a HRT of 3.8 days and a VLR of 2.76 g tCOD L−1 d−1 . (2) Two different membranes in geometry and configuration have been studied as a tertiary treatment. The UF of the EGSB effluent in the S-HF membrane have resulted more productive and selective than the E-T. The filtration in the S-HF module can be carried out with a critical JC,EGSB of 20 L m−2 h−1 , a permeability AEGSB of 241.6 L h−1 m−2 bar−1 and tCOD rejections of 70%. (3) The anaerobic digestion of the LFSM in a EGSB reactor, followed by UF in a S-HF membrane as a tertiary treatment offers a good technological solution to treat this kind of wastewater. These serial processes (EGSB + S-HF) provide a permeate flow without solids and yields to a whole tCOD removal about 90%. (4) The filtration theory and Darcy’s law can be used to explain the membrane fouling when the EGSB effluent was filtered in the SHF module at low permeate flow rates (less than 5 L h−1 ). Values of RM = 1.266 × 1012 m−1 and ˛ = 1.607 × 1014 m kg−1 have been determined. (5) When higher flow rates than 5 L h−1 were used, it was necessary to consider the compressible cake model in order to correlate the time variation of the experimental TMP. The cake was slightly compressible in both cases (n < 0.45). On the other hand, the ˛ values estimated were lower than when the cake was not compressible, and they decreased when the flow rate increased. Acknowledgements This research has been financially supported by the Spanish Ministry of Industry, Tourism and Trade through the FIT-3101002006-17 project. R. López-Fernández acknowledges the assistance of the Spanish Ministry of Education and Science for the award of a Torres Quevedo contract (PTQ04-3-0430). C. Aristizábal also acknowledges the Iberoamerican program of Science and Technology for the Development (CYTED) for the award of a grant.
275
References [1] A. Luo, J. Zhu, P.M. Ndegwa, Removal of carbon, nitrogen, and phosphorus in pig manure by continuous and intermittent aeration at low redox potentials, Biosystems Engineering 82 (2) (2002) 209–215. [2] D.P Chynoweth, A.C. Wilkie, J.M. Owens, Anaerobic processing of piggery wastes: a review, in: Proceedings of the ASAE Annual International Meeting, Orlando, FL, USA, July 11–16, 1998, pp. 1–39. [3] W. Verstraete, P. Vandevivere, New and broader applications of anaerobic digestion, Critical Reviews in Environmental Science and Technology 28 (2) (1999) 151–173. [4] K.H. Hansen, I. Angelidaki, B.K. Ahring, Improving thermophilic anaerobic digestion of swine manure, Water Research 33 (8) (1999) 1805–1810. [5] K.H. Hansen, I. Angelidaki, B.K. Ahring, Anaerobic digestion of swine manure: inhibition by ammonia, Water Research 32 (1) (1998) 5–12. [6] D. Karakashev, J.E. Schmidt, I. Angelidaki, Innovative process scheme for removal of organic matter, phosphorous and nitrogen from pig manure, Water Research 42 (2008) 4083–4090. [7] W.R. Ross, J.P. Barnard, J. le Roux, H.A. Viliers, Application of ultrafiltration membranes for solid-liquid separation in anaerobic digestion systems: the ADUF process, Water SA 16 (2) (1990) 85–91. [8] J. du Preez, B. Norddahl, K. Christensen, The BIOREK® concept: a hybrid membrane bioreactor concept for very strong wastewater, Desalination 183 (2005) 407–415. [9] S.-M. Lee, J.-Y. Jung, Y.C. Chung, Novel method for enhancing permeate flux of submerged membrane system in two-phase anaerobic reactor, Water Research 35 (2) (2001) 471–477. [10] L. Seghezzo, G. Zeeman, J.B. van Lier, H.V.M. Hamelers, G. Lettiga, A review: the anaerobic treatment of sewage in UASB and EGSB reactors, Bioresource Technology 65 (1998) 175–190. [11] Y. Zhang, L. Yan, L. Chi, X. Long, Z. Mei, Z. Zhang, Startup and operation of anaerobic EGSB reactor treating palm oil mill effluent, Journal of Environmental Science 20 (2008) 658–663. [12] I. Colussi, A. Cortesi, L. Della Vedova, V. Gallo, F.K. Cano Robles, Start-up procedures and analysis of heavy metals inhibition on methanogenic activity in EGSB reactor, Bioresource Technology 100 (2009) 6290–6294. [13] G.R. Zoutberg, P. de Been, The biobed EGSB (expanded granular sludge bed) system covers shortcomings of the upflow anaerobic sludge blanket reactor in chemical industry, Water Science Technology 35 (10) (1997) 183–188. [14] L.-B. Chu, F.-L. Yang, X.-W. Zhang, Anaerobic treatment of domestic wastewater in a membrane-coupled expanded granular sludge bed (EGSB) reactor under moderate to low temperature, Process Biochemistry 40 (2005) 1063–1070. [15] APHA, AWWA, WPCF, Standard Methods for the Examination of Waster and Wastewater, 12th ed., Díaz de Santos, Madrid, 1992. [16] G.F. Crozes, J.G. Jacangelo, C. Anselme, J.M. Laîné, Impact of ultrafiltration operating conditions on membrane irreversible fouling, Journal of Membrane Science 124 (1) (1997) 63–76. [17] A. Drews, Membrane fouling in membrane bioreactors—characterisation, contradictions, cause and cures, Journal of Membrane Science 363 (2010) 1–28. [18] G. Zhang, S. Ji, X. Gao, Z. Liu, Adsorptive fouling of extracellular polymeric substances with polymeric ultrafiltration membranes, Journal of Membrane Science 309 (1–2) (2008) 28–35. [19] R.D. Costa, C.R.G. Tavares, E.S. Cossich, Stabilization of swine wastes by anaerobic digestion, Environmental Technology 28 (10) (2007) 1145–1151. [20] P.B. Sørensen, P. Moldrup, J.A.A. Hansen, Filtration and expression of compressible cakes, Chemical Engineering Science 51 (6) (1996) 967–979.