- Email: [email protected]
The effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed reactor
Available online at www.sciencedirect.com
Chemical Engineering and Processing 46 (2007) 1349–1356
The effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed reactor C. Arnaiz a,∗ , P. Buffiere b , J. Lebrato c , R. Moletta b a
Departamento de Ingenier´ıa Qu´ımica y Ambiental, Escuela Universitaria Polit´ecnica, Universidad de Sevilla, Virgen de Africa 7, 41011 Sevilla, Spain b Laboratoire de Biotechnologie de l’Environnement, INRA, Avenue des Etangs, 11100 Narbonne, France c Grupo Tratamiento de Aguas Residuales, Escuela Universitaria Polit´ ecnica, Universidad de Sevilla, Virgen de Africa 7, 41011 Sevilla, Spain Received 11 July 2006; received in revised form 25 October 2006; accepted 28 October 2006 Available online 5 December 2006
Abstract This paper describes the application of the inverse fluidization technology to the anaerobic digestion of wine distillery wastewater. In this reactor, a granular floating solid is expanded by a current of gas. The carrier particles (ExtendospheresTM ) were chosen for their large specific surface area and their low energy requirements for fluidization. The experimental procedure was defined to examine the effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed (ITB) reactor. Moreover, in order to evaluate treatment efficiency, the active biomass concentration was estimated using the phospholipids analysis. The ITB bed reactor appeared to be a good option for anaerobic treatment of high strength wastewater, particularly for the treatment of wine distillery wastewater. The system attained high organic loading rate (OLR) with good chemical oxygen demand (COD) removal rates and it exhibited a good stability to the variations in OLR and HRT. It was found that the main advantages of this system are: low energy requirement because of the low fluidization velocities required; there is no need of a settling device, because solids accumulate at the bottom of the reactor, so they can be easily drawn out and particles with high-biomass content can be easily recovered. Lipid-phosphate concentration has been revealed as a good method for biomass estimation in biofilms since it only includes living biomass. The comparison of the measured concentration of volatile attached solids with the estimated active biomass concentration by means of phospholipids analysis indicated that extremely high active biomass concentrations can be maintained in the system. © 2006 Elsevier B.V. All rights reserved. Keywords: Anaerobic process; Granular floating carrier; Industrial wastewater; Inverse fluidization technology; Phospholipid analysis; Shock loads
1. Introduction Fluidized bed reactors have been developed to provide biological treatment of high strength organic wastewater. This technology has been applied to two main areas of wastewater treatment: nutrient removal in anoxic reactors [1] and organic carbon removal in aerobic and anaerobic reactors [2–5]. Wastewater which contain a high concentration of organic wastes are discharged from many industries including slaughterhouses, breweries, dairies, food processing and wine distilleries. Anaerobic digestion of these types of wastewater is an economical and attractive method because of the low energy consumption, the low production of sludge, the low nutrient
∗
Corresponding author. Tel.: +34 95 4552812; fax: +34 95 4282777. E-mail address: [email protected] (C. Arnaiz).
0255-2701/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.cep.2006.10.017
requirement and the potential energy recovery from the methane produced [6]. In the field of anaerobic digestion, however, fluidized bed reactors have not been extensively used at full-scale. This lack of industrial success may result from a combination of several negative points of traditional fluidization or up-flow fluidization: a high level of maintenance because of their complexity, the need for liquid recycling, or hydrodynamic problems [7]. Several systems have been investigated to adapt the fluidization process for the anaerobic treatment of wastewater such as the inverse fluidized beds or down-flow fluidized beds [8–10] and, more recently, the inverse turbulent bed (ITB) [11,12]. In the ITB configuration, the originality arises from the use of a carrier with a specific density lower than the liquid and a fluidization only ensured by an up-flow current of gas. Previous studies on ITB show several advantages compared to classical up-flow and down-flow fluidization. First, bed height control results automatically from the location of the injection
1350
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
device. Second, a gas injection is simpler than a liquid recycling, reducing clogging problems. Third, the low energy requirement, because of the low fluidization velocities required [11,12]. In order to be an efficiently engineered process, an anaerobic ITB should be tolerant of fluctuations in process parameters such as influent chemical oxygen demand (COD) or influent organic loading rate (OLR). In many industrial processes, wastewater are subject to daily or seasonal variations of flow and load. Any treatment plant must be capable of meeting the average load and be able to accommodate shock loads with a minimum of degradation to the effluent and a minimum of process disruption. The stability and response of a reactor can be quantitatively evaluated from both the overall performance of the system and the concentration of intermediates or end products. COD removal and VFA concentration, if kept in acceptable levels after transient changes in organic load, suggest that the system is suitable for treating variation of the input loading. However, some form of load equalization would be desirable for a maximum substrate removal efficiency on an industrial scale. The loading capacity and fluctuations tolerance of a system, and the subsequent wastewater depollution yield, are dictated by the amount of active biomass available to grow on the biodegradable fraction of the influent. Therefore, in order to evaluate the treatment efficiency of a system, biomass concentration is the most basic, necessary information. Although some researchers measured volatile solids as an index of attached biomass, the index also accounts for inert organic matter like solid metabolites and unviable biomass so that is not directly related to degradation activity. In other words, active biomass concentration is more essential for evaluating treatment efficiency. In spite of its importance, the active biomass concentration has been reported in very few papers which deal with attached biomass, because it is very difficult to measure experimentally (time-consumption, low repeatability and overestimation) [12–16]. Phospholipids, a cell wall component, offer many advantages over other assays for selective cellular biomass estimation in biofilms, and their determination by colorimetric methods is relatively simple, reproducible and sensitive [14]. Lipids have been widely used in environmental samples [17–19] and recently applied to aerobic and anaerobic wastewater biofilm [12,20]. The effects of sudden changes in process parameters on the performance of the anaerobic ITB have not previously been reported. This study was undertaken in order to assess the stability of an anaerobic ITB operated at 35 ◦ C and treating an industrial wine distillery wastewater when subjected to transient changes in OLR. In this work, phospholipid content was assumed to represent the living cells inside the biofilm.
Fig. 1. Laboratory-scale anaerobic inverse turbulent bed bioreactor.
liquid temperature at 35 ± 2 ◦ C. Influent was pumped continuously into the reactor by a peristaltic pump. The input OLR was increased/decreased by reducing/rising the HRT while maintaining constant the COD concentration in the influent. Effluent was discharged through a port on the low part of the column, connected to an outlet tube that kept the liquid level in the reactor. Recycling of the biogas was ensured by a peristaltic pump at a constant flow rate of 0.5 l min−1 (gas superficial velocity of 6 m h−1 , 1.65 mm s−1 ). The gas injection point is located 0.2 m above the bottom of the reactor and the gas is delivered through a perforated rubber tube. The biogas production is measured by a gas flow meter (Sho-RateTM ). 2.2. Influent composition The substrate used in this work is an industrial wine distillery wastewater, supplemented with trace elements and nitrogen, with a total organic carbon concentration (TOC) of 8–12 kg m−3 , equivalent to a COD of 20–30 kg m−3 . The average characteristics of the influent are given in Table 1. 2.3. Physical properties of the carrier material
2. Experimental 2.1. Reactor design The experimental set-up used in this study is shown in Fig. 1. The reactor consisted in a PVC tubular section of 0.08 m internal diameter and 1 m height with a conic bottom, with a total volume of 5 l. The system was equipped with a water jacket keeping the
Table 2 shows physical properties of the support material used in this work. Commercially available ExtendosphereTM (provided by PQ Hollowsphere Ltd.) is a light mineral granular material mainly composed of silica. The shape of the particle is perfectly spherical, and the surface of the material present small crevices. Average particle diameter is 175 m. It required a low gas velocity for being expanded (1.5 mm s−1 ), calculated
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356 Table 1 Influent composition Compound
Concentration (g l−1 )
Solution 1
K2 HPO4 CaCl2 NH4 Cl MgSO4 Yeast extract Solution 1 Wine distillery Wastewater
0.022 g 0.002 g 0.1 g 0.013 g 0.032 g 0.5 ml Up to 1 l
FeCl3 (g l−1 ) H3 BO3 (g l−1 ) CuSO4 ·5H2 O (g l−1 ) NaI (g l−1 ) MnCl2 ·4H2 O (g l−1 ) Na2 MoO4 ·2H2 O (g l−1 ) ZnSO4 ·7H2 O (g l−1 ) CoCl2 ·6H2 O (g l−1 )
0.5 0.05 0.1 0.01 0.04 0.02 0.04 0.05
by the correlation of pressure-drop experimental data at different fluidization velocities [10].
1351
orthophosphoric acid at 10%. The carbon dioxide contained in the samples was previously eliminated by bubbling oxygen gas for 2 min. Gas was analyzed by gas chromatography with a Shimadzu GC-8A apparatus with argon carrier (3 bar) using a catharometer detector (90 mA). CO2 and NO2 were separated in a Hayesep Q precolumn (80–100 mesh, 2 m × 1/8 in.); O2 , H2 , N2 and CH4 were separated in a second column with a molecular sieve ˚ (80–100 mesh, 2 m × 1/8 in.). Oven temperature was 35 ◦ C; 5A temperature of both injector and detector was 100 ◦ C. The chromatograph was coupled to a Shimadzu CR5A integrator. pH was measured with a Mettler Toledo 1100 Calimatic pH meter.
2.4. Experimental procedure
2.6. Biomass determination
The reactor was filled with the solid carrier material up to 40% of its active volume (working volume). The reactor was inoculated with a mixture of anaerobic sludge and anaerobically treated distillery wastewater and the OLR gradually increased until after a period of 90 days an active bacterial population had been established. A detailed account of the reactor start-up procedure has been described previously [11]. In this work, the reactor was operated at OLR from 9.5 to 30.6 kgCOD m−3 day−1 (0.0007–0.0024 kgCOD m−2 day−1 ). The reactor was monitored for temperature, flow rate, pH and biogas production. Biogas composition, volatile attached solids to the carrier (VAS), attached lipid phosphate to the carrier (ALP), volatile fatty acids (VFA) and TOC were routinely analyzed.
Biofilm development within the reactor was measured by determining VAS on washed samples of bioparticles according to Livingston and Chase [21]. A sample of biomass-laden particles (0.5–1.5 ml) was withdrawn from the reactor and gently washed with 100 ml of distilled water to remove any unattached biomass. This sample was then transferred to a weighed crucible and placed in an oven at 105 ◦ C for 24 h to remove all unbound moisture. The crucible was reweighed and placed in a furnace at 560 ◦ C for 1 h to burn off all the biomass present, followed by a further reweighing. Because ExtendosphereTM itself showed some ignition loss, the obtained values were corrected accordingly. The biomass loading was calculated as the difference between the two weightings divided by the total volume of support material present in the sample. The procedure used in this study in order to determine ALP was a modification of that found in [19]. Samples of biomassladen particles were withdrawn from the reactor and gently washed with distilled water to remove any unattached biomass. The procedure consisted of: (a) 3 ml of effluent for SLP and 0.5–1 ml of bioparticles for ALP were added into 70 ml screwcap test tubes. Then, 20 ml of chloroform, 20 ml of methanol and 20 ml of deionized water were added to the samples. The extraction mixture was gently shaken for 10 min and allowed to stand up to complete phase separation. (b) To facilitate recovery of the chloroform, the aqueous (upper) phase was aspirated from the test tubes with the aid of a vacuum pump and subsamples of 5 ml of the chloroform layer were transferred into 10 ml screw-cap test tubes. At this point, lipids can be stored at −20 ◦ C. (c) The chloroform was removed under a stream of nitrogen, and phosphate was liberated from lipids by adding 2.7 ml of a potassium persulfate solution (5 g added to 100 ml of 0.36N H2 SO4 ) and the sealed test tubes were heated in an oven at 105 ◦ C for 1 h. (d) Phosphate release by persulfate digestion was determined by adding 0.6 ml of an ammonium molybdate solu-
2.5. Analytical methods VFA and TOC of the discharged effluent and biogas composition were determined daily through off-line analysis. Liquid samples were centrifuged at 10,000 rpm for 10 min before analysis to remove suspended solids. VFA analysis was done using a gas chromatograph with a flame ionization detector Chrompack CP 9000, nitrogen being the carrier gas (335 kPa). The column was a semi capillar Econocap FFAP (15 m length and 0.53 mm diameter). Injector and detector temperatures were 250 and 275 ◦ C, respectively. The temperature of the oven was programmed to rise from 80 to 120 ◦ C during the analysis with an elevation of 10 ◦ C min−1 . The chromatograph was coupled with an integrator Shimadzu CR3A. TOC was titrated by UV oxidation with a Dohrman DC 80 apparatus. Carbon compounds were oxidized in potassium persulfate at low temperature and the formed carbon dioxide was detected by infrared absorption. Samples were diluted twice with
Table 2 ExtendosphereTM physical properties (SSA, specific surface area; Umf , minimum fluidization velocity; dp , particle diameter) Apparent density (kg m−3 )
Density (kg m−3 )
SSA (m2 m−3 )
Umf (mm s−1 )
dp (m)
Shape
400
690
20,000
1.50
175
Spheres
1352
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
tion [2.5% of (NH4 )6 Mo7 O24 ·4H2 O in 5.72N H2 SO4 allowed to stand for 10 min] and 2.7 ml of a malachite green solution (0.111% polyvinyl alcohol dissolved in water at 80 ◦ C is allowed to cool, and 0.011% malachite green is then added and allowed to stand for 30 min). (e) The absorbance at 610 nm was then read using a spectrophotometer (Beckman DU® 640). The concentrations of phosphate were calculated by using the regression line from a standard curve obtained by digesting 10, 20, 40, 60, 80, 100 and 150 l of a 1 mM glycerol–phosphate solution. In the phospholipid analysis, solvents for lipid extraction were of high quality, lipid standard [dl-(-phosphatidylethanolamine, dipalmitoyl (C16:0)], calcium glycerol phosphate and malachite green were of reagent quality (Sigma) and polyvinyl alcohol was 98% hydrolyzed (average molecular weight, 13,000–23,000; Aldrich Chemical Co., Inc.). Glassware was washed with phosphate-free detergent, rinsed five times with tap water and two to three times with deionized water, and airdried. Glassware was rinsed with chloroform just before used. Potassium persulfate solution must be replaced monthly. Bioparticles were observed with an optical microscope OLYMPUS BX 60 and a 2 mm Leitz-Wetzlar graduated slide with 0.01 mm intervals. 3. Results 3.1. Influence of transient changes in influent flowrate on process stability The performance of the laboratory-scale system under conditions of continuous loading with transient changes is shown in Fig. 2. Fig. 2a is a plot of the alkalinity (expressed in kgCaCO3 m3 ) and VFA (expressed in kg m−3 ) at the outlet of
the reactor. Fig. 2b corresponds to the input OLR (expressed in kgCOD m−3 day−1 ) and COD removal efficiency (expressed in %). Fig. 2c represents the evolution of biogas and methane production (expressed in m3 m3REACTOR day−1 ). 3.1.1. Period 1 The reactor was initially operated at an OLR of 15.8 kgCOD m−3 day−1 and a HRT of 22.8 h. The OLR was increased stepwise and, after 31 days of operation, the OLR achieved by the reactor was 23.1 kgCOD m−3 day−1 with a HRT of 15.6 h. COD removal ranged during this period from 78 to 87%. The VFA concentration remained between 1.3 and 2.4 kg m−3 . Gas production increased significantly from 6.1 to 12.5 m3 m3REACTOR day−1 , while the percentage of CH4 remained almost constant with an average value of 60.4 ± 1.1%. 3.1.2. Period 2 During this period, the input OLR was fixed at 11.0 kgCOD m−3 day−1 (HRT of 29.2 h) and it was increased stepwise to 28.2 kgCOD m−3 day−1 (HRT of 11.8 h) in 28 days (from day 37 to day 65). Fig. 2b shows that the system was not affected, even when OLR was increased more quickly than in Period 1, and COD removal was always between 83 and 92%. The VFA content and the gas production were also not affected. The VFA concentration remained always below 1.26 kg m−3 and gas production increased continuously from 6.2 to 15.2 m3 m3REACTOR day−1 , with a CH4 content of 64.3 ± 0.7%. 3.1.3. Period 3 From day 75 up to the end of this study, the mode of operation was abruptly changed in order to verify the stability of the system facing strong disturbances. The input OLR was dropped down from 27.2 kgCOD m−3 day−1 (HRT of 11.7 h) to an average of 16.4 kgCOD m−3 day−1 (HRT of 19.3 h) since day 75 to day 90. After this period, the input OLR was roughly increased to an average of 28.1 kgCOD m−3 day−1 (HRT of 11.8 h) since day 91 to day 117. Fig. 2c shows that the system was partly affected. Nevertheless, COD removal was always between 70 and 84%. It is noticeable that the VFA content was also affected, but it was always below 2.6 kg m−3 . The average gas production from day −1 91 to day 117 was 15.1 ± 0.8 m3 m−3 REACTOR day . The CH4 content in the gaseous phase was of 61.1 ± 1.7%. 3.2. Influence of transient changes in influent flowrate on biomass growth and carrier colonization During the three operational periods, biomass accumulation was monitored as VAS (mg per ml of support) and ALP (nmol Pi per ml of support). In this study, phospholipid analysis was assumed to represent the active cells inside the biofilm. Fig. 3 shows the accumulation of biomass on carrier particles during the time of operation.
Fig. 2. Performance of the anaerobic inverse fluidized bed reactor treating industrial wine distillery wastewater.
3.2.1. Period 1 The initial VAS into the reactor (12.8 mg ml−1 ) dropped down to 8.8 mg ml−1 the day 26, but the system recovered at
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
1353
4. Discussion
Fig. 3. Biomass accumulation on carrier particles during the operation periods.
the end of this period up to a value of 13.1 mg ml−1 . The effect of continuous increasing in the input loading rate on ALP was similar to the described for the VAS, with a decrease from 40.7 nmol Pi ml−1 to a minimum of 14.3 nmol Pi ml−1 by day 21. Nevertheless, this parameter did not recover itself completely at the end of this period, and the measurement of ALP gave a value of 29.3 nmol Pi ml−1 by day 35. 3.2.2. Period 2 During this period, the VAS concentration increased up to a value of 20.5 mg ml−1 , while ALP dropped down from a maximum of 49.1 nmol Pi ml−1 the day 54 up to 20.0 nmol Pi ml−1 at the end of the period. 3.2.3. Period 3 The abrupt change in the mode of operation of the reactor during this period did not affect negatively either VAS or ALP concentration. VAS and ALP increased up to a value of 23.4 mg ml−1 and 30.7 nmol Pi ml−1 , respectively. The main objective in measuring lipid-phosphate concentration was to find the living cell biomass. Fig. 4 shows VAS against ALP during the operation period. It can be seen that when x = 0, y = 0. This means that when there is no living biomass, expressed as lipid concentration, there are still attached solids. This amount (y = 0) gave a value of 10.7 mgVAS ml−1 CARRIER .
Fig. 4. VAS vs. ALP during the operation periods.
The response to transient changes in OLR of an anaerobic ITB operated at 35 ◦ C in the treatment of industrial wine distillery wastewater appears to be stable and reliable at the operational conditions of this work. There were not noticeable changes about instability in process parameters. VFA concentration was always below 2.6 kg m−3 . CH4 content in the gaseous phase was always between 56.3 and 66.3% while COD removal never fell below 70%. The specific methane production gave a mean value of 0.33 m3 of CH4 per kg of removed COD. This value is lightly inferior to the stoichiometric theoretical of 0.35 m3 of CH4 per kg of removed COD, typical of an anaerobic system. Nevertheless, this data is similar to that obtained by other authors who argue that synthesis of new microorganisms is the reason for this deviation from the theoretical ratio [22]. In spite of three periods of transient changes in the input OLR, it was possible to obtain COD removal between 70 and 92% at OLR from 9.5 to 30.6 kgCOD m−3 day−1 . A comparison with previously studied reactors in our laboratory, treating the same kind of effluent in a continuous loading mode of operation, can be done. Buffiere et al. [23] reported carbon removal rates between 75 and 92% at OLR from 2 to 18 kgCOD m−3 day−1 working with a classical up-flow fluidized bed using 385 m pozzolana particles. In an inverse fluidized bed (with a down-flow liquid fluidization using perlite as biomass carrier), the OLR was increased from 3 to 15 kgCOD m−3 day−1 , but the reactor was destabilized and the carbon removal was only 55% at the end of the experiment and the input load had to be decreased [10]. In a similar reactor (an inverse turbulent bed with an up-flow gas stream using ExtendosphereTM as biomass carrier), the OLR was increased from 2 to 23.1 kgCOD m−3 day−1 with a carbon removal efficiency of about 77% [11]. COD removal attained in this work can be also compared with those obtained from some up-flow anaerobic fluidized bed reactors treating similar substrates in a continuous loading mode of operation. COD removal were observed to decrease from 96.6 to 81.5% at HRT of 60.0 and 11.0 h, respectively, and OLR of 5.7 and 32 kgCOD m−3 day−1 working with an up-flow thermophilic anaerobic fluidized bed reactor using open pore sintered glass beds as biomass carrier [22]. A subsequent applied organic overload produced a fast decrease in pH, which resulted in poor effluent quality (low efficiency of substrate removal, 33.5%). Other wine distillery waste studies have been reported using different anaerobic processes such as anaerobic filter [24] and the UASB reactor [25], with 80 and 83% of COD removal at organic loads of 12.0 and 13.2 kgCOD m−3 day−1 and HRT of 1.4 and 2.4 days, respectively. In this study, total biomass amount is measured in terms of volatile solids concentration and lipid-phosphate concentration. The main disadvantage of the first method is that its estimation includes not only active microorganisms, but also inert mass, exopolymers and absorbed organic matter on flocs and biofilms. Phospholipids, present on bacterial membrane up to 90–98%, do not form part of cell reserves and are easily degradable during bacteria lysis [17]. Therefore, their estimation only includes
1354
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
living biomass. When the relation between these two measures is calculated (VAS/ALP), an estimation of living biomass with regards to total biomass is obtained. Initial biomass concentration of pre-colonized bioparticles used in this study was 12.8 mgVAS ml−1 CARRIER and 40.7 nmolALP ml−1 . At the end of the study, final CARRIER −1 biomass concentration was 23.4 mgVAS mlCARRIER and 30.7 nmolALP ml−1 CARRIER . It means that a biomass concentration almost duplicated in terms of VAS but 25% reduced in terms of ALP multiplied by 1.5 the organic matter treated. It is important to point out a better overall carbon removal rate in the reactor at the end of this study in spite of a biomass measured as lipid-phosphate concentration much lower than at the beginning of the study. It could be explained if a more active biomass is admitted. That increase in the biomass activity cannot be explained by a greater ratio of cells in the biofilm matrix, since ALP decreased significantly, but it suggests a change in biofilm composition. On the other hand, there does not always have to be a relation between the concentration of biomass and the performance parameters of a reactor. A priori, it seems clear that in a system in which the organic matter is degraded only and exclusively by the action of microorganisms, the consumption rate of substratum and the growth or production rate of microbial cell mass must be deeply related. Nevertheless, several factors make these rates not proportional: it is necessary to take into account that the microorganisms cannot be considered chemical reactors since they have the capability of adaptation, and population displacements can occur. This strong selection is not uncommon in reactors fed with industrial wastewater, in which very specific substrata could select very specific groups of bacteria, progressively more active [12,22,26]. Some authors indicate that the friction effects due to high turbulence and shear of fluidized bed reactors may favour growth of thin, dense and active biofilms [22,26]. However, microscopic observations of bioparticles at the end of this work confirmed the presence of biomass on the carrier as local outgrowth colonies, not as uniform covering of the particles (Fig. 5). The type of attached growth or colonial morphology (homogeneous layer, cluster) is a characteristic of the bacterial strains and may significantly influence on spatial distribution of cells [14,27]. Based
Fig. 5. Biofilm coverage of the carrier material at the end of the study.
on these data, the active attached biomass concentration into a fluidized bed reactor could be a consequence of both turbulence and high shear conditions, and progressive biological selection. It is interesting to point out the abrupt decrease of ALP at day 56 (Fig. 3). Since there was not any noticeable change in the monitored parameter, it was assumed that a change in the biofilm composition took place. An important point of the relationship between total biomass amount measured as volatile solids and total biomass amount measured as lipid-phosphate concentration is that is not of y = ax type, as it could be expected, but of y = ax + b type (Fig. 4): when living biomass is completely stabilized, there are still volatile solids into the reactors (within the biofilm). Biomass consists of a biodegradable fraction and an inert fraction, which cannot be biologically degraded. This inert fraction is composed by deadend products [28,29] and rest of extracellular matrix. Based on data of this study, biomass inert fraction would be represented by the ordinate at the y-axis of straight lines shown in Fig. 4: 10.7 mgVAS ml−1 CARRIER . An important fact for effluent quality is the volatile suspended solids concentration. Although detailed measurements were not performed routinely during this study, effluent volatile suspended solids concentration reached a value of 1.46 ± 0.19 mg ml−1 at higher OLR (between day 91 and the end of the study). This value is only lightly superior to that achieved by other authors working with traditional up-flow or down-flow configuration at similar operation parameters [22,26]. The relationship between total biomass amount measured as volatile suspended solids and total biomass amount measured as suspended lipid-phosphate concentration gave y = 0.05x + 0.58. The ordinate at the y-axis of this straight line is almost 20 times inferior to the ordinate for the attached biomass. It can be explained if we assumed that biomass inert fraction would be represented by the ordinate at the y-axis of straight lines so higher ordinates were those obtained for attached biomass, since exopolymer matrix is more abundant in biofilms, up to 80% of volatile solids [30]. One of the major advantages of the anaerobic fluidized bed system over a suspended microbial system is the low volatile suspended solids concentration in the effluent, even at very high loading (hydraulic and organic) conditions. However, in classical up-flow or down-flow fluidized beds, overgrowth of biomass and an important biogas production may lead to operational problems such as particle carry-over, or to fluidization problems such as gas spouting or bed contraction [7]. Concerning the ITB reactor, in which the biogas itself is used as fluidizing agent, these hydrodynamic limitations seem to be overcome. Furthermore, biogas production in the reactor helps to fluidization, which enables to reduce the energy consumption due to biogas recycling at high organic loading rates. Indeed, the bottom of the reactor stands for an internal settler and no external settling device is needed to separate detached biomass from the liquid. Additionally, this settling zone can be used during the start-up period to recover unattached inoculum in order to recycle it into the reactor. From this point of view, inverse fluidization is a very promising technology for the treatment of high strength wastewater. Besides its high carbon removal potential, it over-
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
comes the limitations of traditional fluidized bed reactors. No liquid recycling is required, which prevent from pipe clog up and other maintenance problems. On an industrial scale, recycling of biogas is much easier than recycling of liquid. It is also a robust system, which can tolerate fluctuations in process parameters, such as OLR, with a minimum effluent degradation and its subsequent recovery. 5. Conclusions The ITB reactor appeared to be a good option for anaerobic treatment of high strength wastewater, particularly for the treatment of wine distillery wastewater. The systems attained high OLR with good COD removal rates and it exhibited a good stability to the variations in OLR. At the operational conditions of this work, it was possible to obtain COD removal between 70 and 92% at OLR from 9.5 to 30.6 kgCOD m−3 day−1 . The biofilm at the end of the study was very active and specialized, which can explain the stability of the reactor facing disturbances. Lipid-phosphate concentration has been revealed as a good method for attached biomass estimation since it only includes living biomass. This point is of particular importance in biofilms, in which exopolymer matrix constitutes an important fraction of volatile solids. Acknowledgments This work was financed in part by a grant from the Ministry of Education and Cultures of Spain and by a grant from the Water Management Professional Training Project of the European Community’s Leonardo da Vinci Programme, both to the first author. Appendix A. Nomenclature
ALP COD dp HRT ITB OLR PVC SSA TOC Umf VAS VFA
attached lipid phosphate (nmol ml−1 CARRIER ) chemical oxygen demand (kg m−3 ) particle diameter (m) hydraulic retention time (day) inverse turbulent bed reactor organic loading rate (kgCOD m−3 day−1 ) polyvinyl chloride specific surface area (m2 m−3 ) total organic carbon (kg m−3 ) minimum fluidization velocity (m h−1 ) volatile attached solids (mg ml−1 CARRIER ) volatile fatty acids (kg m−3 )
References [1] M. Green, M. Shnitzer, S. Tarre, B. Bogdan, G. Shelef, C.J. Sorden, Fluidized bed reactor operation for groundwater denitrification, Water Sci. Technol. 29 (1994) 509–515. [2] L. Nikolov, D. Karamanev, The inverse fluidization: a new approach to biofilm reactor design to aerobic wastewater treatment, Stud. Environ. 42 (1991) 177–182.
1355
[3] B. Rusten, L. Hem, H. Ødegaard, Nitrification of municipal wastewater in moving-bed biofilm reactors, Water Environ. Res. 67 (1995) 75–86. [4] J.J. Heijnen, A. Mulder, W. Enger, F. Hoeks, Review on the application of anaerobic fluidized bed reactors in waste-water treatment, Chem. Eng. J. 41 (1989) B37–B50. [5] G.K. Anderson, I. Ozturk, C.B. Saw, Pilot-scale experiences on anaerobic fluidized-bed treatment of brewery wastes, Water Sci. Technol. 25 (1990) 157–166. [6] J.M. Stewart, S.K. Bhattacharya, R.L. Madura, S.H. Mason, J.C. Schonberg, Anaerobic treatability of selected organic toxicants in petrochemical wastes, Water Res. 29 (1995) 2730–2738. [7] P. Buffiere, C. Fonade, R. Moletta, Mixing and phase hold-up variations due to gas production in anaerobic fluidized bed digesters: influence on reactor performance, Biotechnol. Bioeng. 60 (1998) 36–43. [8] Y. Chan Choi, D. Seog Kim, S. Joo Park, S. Koo Song, Wastewater treatment in a pilot scale inverse fluidised bed biofilm reactor, Biotechnol. Lett. 9 (1995) 35–40. [9] D. Garcia-Calderon, P. Buffiere, R. Moletta, S. Elmaleh, Influence of biomass accumulation on bed expansion characteristics of a down flow anaerobic fluidized bed reactor, Biotechnol. Bioeng. 57 (1998) 136–144. [10] D. Garcia-Calderon, P. Buffiere, R. Moletta, S. Elmaleh, Anaerobic digestion of wine distillery wastewater in downflow fluidized bed, Water Res. 32 (1998) 3593–3600. [11] P. Buffiere, J.P. Bergeon, R. Moletta, The inverse turbulent bed: a novel bioreactor for anaerobic digestion, Water Res. 34 (2000) 673–677. [12] C. Arnaiz, P. Buffiere, S. Elmaleh, J. Lebrato, R. Moletta, Anaerobic digestion of dairy wastewater by inverse fluidization: the inverse fluidized bed and the inverse turbulent bed reactors, Environ. Technol. 24 (2003) 1431–1443. [13] T. Kuba, H. Furumai, T. Kusuda, A kinetic study on methanogenesis by attached biomass in a fluidized bed, Water Res. 24 (1990) 1365–1372. [14] V. Lazarova, J. Manem, Biofilm characterization and activity analysis in water and wastewater treatment, Water Res. 29 (1995) 2227–2245. [15] M.C. Arnaiz, C. Ruiz, E. Gomez, I. Garcia, E. Escot, E. Aguilar, J.M. Medialdea, J.C. Gutierrez, J. Lebrato, Evaluation of the efficacy of materials as anaerobic wastewaters treatment supports by phospholipid analysis, in: Proc. of the International Specialty Conference on Microbial Ecology of Biofilms: Concepts, Tools and Applications, Lake Bluff, Illinois, USA, 1998, pp. 294–300. [16] M.C. Arnaiz, Depuraci´on biol´ogica de aguas residuales industrials, Desarrollo de tecnolog´ıa con lechos fluidizados, Thesis Doc., University of Sevilla, Spain, 2002. [17] D.C. White, R.J. Bobbie, J.S. Herron, J.D. King, S. Morrison, Biochemical measurements of microbial mass and activity from environmental samples, in: J.W. Costerton (Ed.), Native Aquatic Bacteria: Enumeration, Activity and Ecology, ASTM Spec. Tech. Publ., University of Calgary, 1979. [18] D.L. Balkwill, F.R. Leach, J.T. Wilson, J.F. Mcnabb, D.C. White, Equivalence of microbial biomass measures based on membrane lipid and cell wall components, adenosine triphosphate, and direct counts in subsurface aquifer sediments, Microbiol. Ecol. 16 (1988) 73–84. [19] R.H. Findlay, M.G. King, J. Watling, Efficacy of phospholipid analysis in determining microbial biomass in sediments, Appl. Environ. Microbiol. 55 (1989) 2888–2893. [20] T.C. Zhang, P.L. Bishop, Density, porosity and pore structure of biofilms, Water Res. 28 (1994) 2267–2277. [21] A.G. Livingston, H.A. Chase, Development of a phenol degrading fluidized bed bioreactor for constant biomass holdup, Chem. Eng. J. 45 (1991) B35–B47. [22] M. Perez, L.I. Romero, D. Sales, Organic matter degradation kinetics in an anaerobic thermophilic fluidised bed bioreactor, Anaerobe 7 (2001) 25–35. [23] P. Buffiere, C. Fonade, R. Moletta, Continuous operation of a fluidized bed bioreactor for anaerobic digestion: residence time influence on degradation kinetics, Biotechnol. Lett. 17 (1995) 833–888. [24] M. Perez, L.I. Romero, D. Sales, Degradaci´on anaerobia termof´ılica de vinazas de vino: efecto del pH de la alimentaci´on, Tecnolog´ıa del Agua 158 (1996) 41–45. [25] A.M. Craveiro, B.M. Rocha, W. Schmidell, Water Treatment Conference Aquatech’86, Amsterdam, 1996, pp. 307–319.
1356
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
[26] P. Buffiere, R. Moletta, Relations between carbon removal rates, biofilm size and density of a novel anaerobic reactor: the inverse turbulent bed, Water Sci. Technol. 41 (2000) 253–260. [27] M.A. Siebel, W.G. Characklis, Observations of binary population biofilms, Biotechnol. Bioeng. 37 (1991) 778–789. [28] W.W. Eckenfelder, Principles of Water Quality Management, CBI Publishing Co., Boston, 1980.
[29] G.R. Marais, Theory, Design and Operation of Nutrient Removal Activated Sludge Processes, Water Research Commission, Pretoria, 1984. [30] Y.M. Nelson, L.W. Lion, M.L. Shuler, W.C. Ghiorse, Modeling oligotrophic biofilm formation and lead adsorption to biofilm components, Environ. Sci. Technol. 30 (1996) 2027–2035.
Chemical Engineering and Processing 46 (2007) 1349–1356
The effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed reactor C. Arnaiz a,∗ , P. Buffiere b , J. Lebrato c , R. Moletta b a
Departamento de Ingenier´ıa Qu´ımica y Ambiental, Escuela Universitaria Polit´ecnica, Universidad de Sevilla, Virgen de Africa 7, 41011 Sevilla, Spain b Laboratoire de Biotechnologie de l’Environnement, INRA, Avenue des Etangs, 11100 Narbonne, France c Grupo Tratamiento de Aguas Residuales, Escuela Universitaria Polit´ ecnica, Universidad de Sevilla, Virgen de Africa 7, 41011 Sevilla, Spain Received 11 July 2006; received in revised form 25 October 2006; accepted 28 October 2006 Available online 5 December 2006
Abstract This paper describes the application of the inverse fluidization technology to the anaerobic digestion of wine distillery wastewater. In this reactor, a granular floating solid is expanded by a current of gas. The carrier particles (ExtendospheresTM ) were chosen for their large specific surface area and their low energy requirements for fluidization. The experimental procedure was defined to examine the effect of transient changes in organic load on the performance of an anaerobic inverse turbulent bed (ITB) reactor. Moreover, in order to evaluate treatment efficiency, the active biomass concentration was estimated using the phospholipids analysis. The ITB bed reactor appeared to be a good option for anaerobic treatment of high strength wastewater, particularly for the treatment of wine distillery wastewater. The system attained high organic loading rate (OLR) with good chemical oxygen demand (COD) removal rates and it exhibited a good stability to the variations in OLR and HRT. It was found that the main advantages of this system are: low energy requirement because of the low fluidization velocities required; there is no need of a settling device, because solids accumulate at the bottom of the reactor, so they can be easily drawn out and particles with high-biomass content can be easily recovered. Lipid-phosphate concentration has been revealed as a good method for biomass estimation in biofilms since it only includes living biomass. The comparison of the measured concentration of volatile attached solids with the estimated active biomass concentration by means of phospholipids analysis indicated that extremely high active biomass concentrations can be maintained in the system. © 2006 Elsevier B.V. All rights reserved. Keywords: Anaerobic process; Granular floating carrier; Industrial wastewater; Inverse fluidization technology; Phospholipid analysis; Shock loads
1. Introduction Fluidized bed reactors have been developed to provide biological treatment of high strength organic wastewater. This technology has been applied to two main areas of wastewater treatment: nutrient removal in anoxic reactors [1] and organic carbon removal in aerobic and anaerobic reactors [2–5]. Wastewater which contain a high concentration of organic wastes are discharged from many industries including slaughterhouses, breweries, dairies, food processing and wine distilleries. Anaerobic digestion of these types of wastewater is an economical and attractive method because of the low energy consumption, the low production of sludge, the low nutrient
∗
Corresponding author. Tel.: +34 95 4552812; fax: +34 95 4282777. E-mail address: [email protected] (C. Arnaiz).
0255-2701/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.cep.2006.10.017
requirement and the potential energy recovery from the methane produced [6]. In the field of anaerobic digestion, however, fluidized bed reactors have not been extensively used at full-scale. This lack of industrial success may result from a combination of several negative points of traditional fluidization or up-flow fluidization: a high level of maintenance because of their complexity, the need for liquid recycling, or hydrodynamic problems [7]. Several systems have been investigated to adapt the fluidization process for the anaerobic treatment of wastewater such as the inverse fluidized beds or down-flow fluidized beds [8–10] and, more recently, the inverse turbulent bed (ITB) [11,12]. In the ITB configuration, the originality arises from the use of a carrier with a specific density lower than the liquid and a fluidization only ensured by an up-flow current of gas. Previous studies on ITB show several advantages compared to classical up-flow and down-flow fluidization. First, bed height control results automatically from the location of the injection
1350
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
device. Second, a gas injection is simpler than a liquid recycling, reducing clogging problems. Third, the low energy requirement, because of the low fluidization velocities required [11,12]. In order to be an efficiently engineered process, an anaerobic ITB should be tolerant of fluctuations in process parameters such as influent chemical oxygen demand (COD) or influent organic loading rate (OLR). In many industrial processes, wastewater are subject to daily or seasonal variations of flow and load. Any treatment plant must be capable of meeting the average load and be able to accommodate shock loads with a minimum of degradation to the effluent and a minimum of process disruption. The stability and response of a reactor can be quantitatively evaluated from both the overall performance of the system and the concentration of intermediates or end products. COD removal and VFA concentration, if kept in acceptable levels after transient changes in organic load, suggest that the system is suitable for treating variation of the input loading. However, some form of load equalization would be desirable for a maximum substrate removal efficiency on an industrial scale. The loading capacity and fluctuations tolerance of a system, and the subsequent wastewater depollution yield, are dictated by the amount of active biomass available to grow on the biodegradable fraction of the influent. Therefore, in order to evaluate the treatment efficiency of a system, biomass concentration is the most basic, necessary information. Although some researchers measured volatile solids as an index of attached biomass, the index also accounts for inert organic matter like solid metabolites and unviable biomass so that is not directly related to degradation activity. In other words, active biomass concentration is more essential for evaluating treatment efficiency. In spite of its importance, the active biomass concentration has been reported in very few papers which deal with attached biomass, because it is very difficult to measure experimentally (time-consumption, low repeatability and overestimation) [12–16]. Phospholipids, a cell wall component, offer many advantages over other assays for selective cellular biomass estimation in biofilms, and their determination by colorimetric methods is relatively simple, reproducible and sensitive [14]. Lipids have been widely used in environmental samples [17–19] and recently applied to aerobic and anaerobic wastewater biofilm [12,20]. The effects of sudden changes in process parameters on the performance of the anaerobic ITB have not previously been reported. This study was undertaken in order to assess the stability of an anaerobic ITB operated at 35 ◦ C and treating an industrial wine distillery wastewater when subjected to transient changes in OLR. In this work, phospholipid content was assumed to represent the living cells inside the biofilm.
Fig. 1. Laboratory-scale anaerobic inverse turbulent bed bioreactor.
liquid temperature at 35 ± 2 ◦ C. Influent was pumped continuously into the reactor by a peristaltic pump. The input OLR was increased/decreased by reducing/rising the HRT while maintaining constant the COD concentration in the influent. Effluent was discharged through a port on the low part of the column, connected to an outlet tube that kept the liquid level in the reactor. Recycling of the biogas was ensured by a peristaltic pump at a constant flow rate of 0.5 l min−1 (gas superficial velocity of 6 m h−1 , 1.65 mm s−1 ). The gas injection point is located 0.2 m above the bottom of the reactor and the gas is delivered through a perforated rubber tube. The biogas production is measured by a gas flow meter (Sho-RateTM ). 2.2. Influent composition The substrate used in this work is an industrial wine distillery wastewater, supplemented with trace elements and nitrogen, with a total organic carbon concentration (TOC) of 8–12 kg m−3 , equivalent to a COD of 20–30 kg m−3 . The average characteristics of the influent are given in Table 1. 2.3. Physical properties of the carrier material
2. Experimental 2.1. Reactor design The experimental set-up used in this study is shown in Fig. 1. The reactor consisted in a PVC tubular section of 0.08 m internal diameter and 1 m height with a conic bottom, with a total volume of 5 l. The system was equipped with a water jacket keeping the
Table 2 shows physical properties of the support material used in this work. Commercially available ExtendosphereTM (provided by PQ Hollowsphere Ltd.) is a light mineral granular material mainly composed of silica. The shape of the particle is perfectly spherical, and the surface of the material present small crevices. Average particle diameter is 175 m. It required a low gas velocity for being expanded (1.5 mm s−1 ), calculated
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356 Table 1 Influent composition Compound
Concentration (g l−1 )
Solution 1
K2 HPO4 CaCl2 NH4 Cl MgSO4 Yeast extract Solution 1 Wine distillery Wastewater
0.022 g 0.002 g 0.1 g 0.013 g 0.032 g 0.5 ml Up to 1 l
FeCl3 (g l−1 ) H3 BO3 (g l−1 ) CuSO4 ·5H2 O (g l−1 ) NaI (g l−1 ) MnCl2 ·4H2 O (g l−1 ) Na2 MoO4 ·2H2 O (g l−1 ) ZnSO4 ·7H2 O (g l−1 ) CoCl2 ·6H2 O (g l−1 )
0.5 0.05 0.1 0.01 0.04 0.02 0.04 0.05
by the correlation of pressure-drop experimental data at different fluidization velocities [10].
1351
orthophosphoric acid at 10%. The carbon dioxide contained in the samples was previously eliminated by bubbling oxygen gas for 2 min. Gas was analyzed by gas chromatography with a Shimadzu GC-8A apparatus with argon carrier (3 bar) using a catharometer detector (90 mA). CO2 and NO2 were separated in a Hayesep Q precolumn (80–100 mesh, 2 m × 1/8 in.); O2 , H2 , N2 and CH4 were separated in a second column with a molecular sieve ˚ (80–100 mesh, 2 m × 1/8 in.). Oven temperature was 35 ◦ C; 5A temperature of both injector and detector was 100 ◦ C. The chromatograph was coupled to a Shimadzu CR5A integrator. pH was measured with a Mettler Toledo 1100 Calimatic pH meter.
2.4. Experimental procedure
2.6. Biomass determination
The reactor was filled with the solid carrier material up to 40% of its active volume (working volume). The reactor was inoculated with a mixture of anaerobic sludge and anaerobically treated distillery wastewater and the OLR gradually increased until after a period of 90 days an active bacterial population had been established. A detailed account of the reactor start-up procedure has been described previously [11]. In this work, the reactor was operated at OLR from 9.5 to 30.6 kgCOD m−3 day−1 (0.0007–0.0024 kgCOD m−2 day−1 ). The reactor was monitored for temperature, flow rate, pH and biogas production. Biogas composition, volatile attached solids to the carrier (VAS), attached lipid phosphate to the carrier (ALP), volatile fatty acids (VFA) and TOC were routinely analyzed.
Biofilm development within the reactor was measured by determining VAS on washed samples of bioparticles according to Livingston and Chase [21]. A sample of biomass-laden particles (0.5–1.5 ml) was withdrawn from the reactor and gently washed with 100 ml of distilled water to remove any unattached biomass. This sample was then transferred to a weighed crucible and placed in an oven at 105 ◦ C for 24 h to remove all unbound moisture. The crucible was reweighed and placed in a furnace at 560 ◦ C for 1 h to burn off all the biomass present, followed by a further reweighing. Because ExtendosphereTM itself showed some ignition loss, the obtained values were corrected accordingly. The biomass loading was calculated as the difference between the two weightings divided by the total volume of support material present in the sample. The procedure used in this study in order to determine ALP was a modification of that found in [19]. Samples of biomassladen particles were withdrawn from the reactor and gently washed with distilled water to remove any unattached biomass. The procedure consisted of: (a) 3 ml of effluent for SLP and 0.5–1 ml of bioparticles for ALP were added into 70 ml screwcap test tubes. Then, 20 ml of chloroform, 20 ml of methanol and 20 ml of deionized water were added to the samples. The extraction mixture was gently shaken for 10 min and allowed to stand up to complete phase separation. (b) To facilitate recovery of the chloroform, the aqueous (upper) phase was aspirated from the test tubes with the aid of a vacuum pump and subsamples of 5 ml of the chloroform layer were transferred into 10 ml screw-cap test tubes. At this point, lipids can be stored at −20 ◦ C. (c) The chloroform was removed under a stream of nitrogen, and phosphate was liberated from lipids by adding 2.7 ml of a potassium persulfate solution (5 g added to 100 ml of 0.36N H2 SO4 ) and the sealed test tubes were heated in an oven at 105 ◦ C for 1 h. (d) Phosphate release by persulfate digestion was determined by adding 0.6 ml of an ammonium molybdate solu-
2.5. Analytical methods VFA and TOC of the discharged effluent and biogas composition were determined daily through off-line analysis. Liquid samples were centrifuged at 10,000 rpm for 10 min before analysis to remove suspended solids. VFA analysis was done using a gas chromatograph with a flame ionization detector Chrompack CP 9000, nitrogen being the carrier gas (335 kPa). The column was a semi capillar Econocap FFAP (15 m length and 0.53 mm diameter). Injector and detector temperatures were 250 and 275 ◦ C, respectively. The temperature of the oven was programmed to rise from 80 to 120 ◦ C during the analysis with an elevation of 10 ◦ C min−1 . The chromatograph was coupled with an integrator Shimadzu CR3A. TOC was titrated by UV oxidation with a Dohrman DC 80 apparatus. Carbon compounds were oxidized in potassium persulfate at low temperature and the formed carbon dioxide was detected by infrared absorption. Samples were diluted twice with
Table 2 ExtendosphereTM physical properties (SSA, specific surface area; Umf , minimum fluidization velocity; dp , particle diameter) Apparent density (kg m−3 )
Density (kg m−3 )
SSA (m2 m−3 )
Umf (mm s−1 )
dp (m)
Shape
400
690
20,000
1.50
175
Spheres
1352
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
tion [2.5% of (NH4 )6 Mo7 O24 ·4H2 O in 5.72N H2 SO4 allowed to stand for 10 min] and 2.7 ml of a malachite green solution (0.111% polyvinyl alcohol dissolved in water at 80 ◦ C is allowed to cool, and 0.011% malachite green is then added and allowed to stand for 30 min). (e) The absorbance at 610 nm was then read using a spectrophotometer (Beckman DU® 640). The concentrations of phosphate were calculated by using the regression line from a standard curve obtained by digesting 10, 20, 40, 60, 80, 100 and 150 l of a 1 mM glycerol–phosphate solution. In the phospholipid analysis, solvents for lipid extraction were of high quality, lipid standard [dl-(-phosphatidylethanolamine, dipalmitoyl (C16:0)], calcium glycerol phosphate and malachite green were of reagent quality (Sigma) and polyvinyl alcohol was 98% hydrolyzed (average molecular weight, 13,000–23,000; Aldrich Chemical Co., Inc.). Glassware was washed with phosphate-free detergent, rinsed five times with tap water and two to three times with deionized water, and airdried. Glassware was rinsed with chloroform just before used. Potassium persulfate solution must be replaced monthly. Bioparticles were observed with an optical microscope OLYMPUS BX 60 and a 2 mm Leitz-Wetzlar graduated slide with 0.01 mm intervals. 3. Results 3.1. Influence of transient changes in influent flowrate on process stability The performance of the laboratory-scale system under conditions of continuous loading with transient changes is shown in Fig. 2. Fig. 2a is a plot of the alkalinity (expressed in kgCaCO3 m3 ) and VFA (expressed in kg m−3 ) at the outlet of
the reactor. Fig. 2b corresponds to the input OLR (expressed in kgCOD m−3 day−1 ) and COD removal efficiency (expressed in %). Fig. 2c represents the evolution of biogas and methane production (expressed in m3 m3REACTOR day−1 ). 3.1.1. Period 1 The reactor was initially operated at an OLR of 15.8 kgCOD m−3 day−1 and a HRT of 22.8 h. The OLR was increased stepwise and, after 31 days of operation, the OLR achieved by the reactor was 23.1 kgCOD m−3 day−1 with a HRT of 15.6 h. COD removal ranged during this period from 78 to 87%. The VFA concentration remained between 1.3 and 2.4 kg m−3 . Gas production increased significantly from 6.1 to 12.5 m3 m3REACTOR day−1 , while the percentage of CH4 remained almost constant with an average value of 60.4 ± 1.1%. 3.1.2. Period 2 During this period, the input OLR was fixed at 11.0 kgCOD m−3 day−1 (HRT of 29.2 h) and it was increased stepwise to 28.2 kgCOD m−3 day−1 (HRT of 11.8 h) in 28 days (from day 37 to day 65). Fig. 2b shows that the system was not affected, even when OLR was increased more quickly than in Period 1, and COD removal was always between 83 and 92%. The VFA content and the gas production were also not affected. The VFA concentration remained always below 1.26 kg m−3 and gas production increased continuously from 6.2 to 15.2 m3 m3REACTOR day−1 , with a CH4 content of 64.3 ± 0.7%. 3.1.3. Period 3 From day 75 up to the end of this study, the mode of operation was abruptly changed in order to verify the stability of the system facing strong disturbances. The input OLR was dropped down from 27.2 kgCOD m−3 day−1 (HRT of 11.7 h) to an average of 16.4 kgCOD m−3 day−1 (HRT of 19.3 h) since day 75 to day 90. After this period, the input OLR was roughly increased to an average of 28.1 kgCOD m−3 day−1 (HRT of 11.8 h) since day 91 to day 117. Fig. 2c shows that the system was partly affected. Nevertheless, COD removal was always between 70 and 84%. It is noticeable that the VFA content was also affected, but it was always below 2.6 kg m−3 . The average gas production from day −1 91 to day 117 was 15.1 ± 0.8 m3 m−3 REACTOR day . The CH4 content in the gaseous phase was of 61.1 ± 1.7%. 3.2. Influence of transient changes in influent flowrate on biomass growth and carrier colonization During the three operational periods, biomass accumulation was monitored as VAS (mg per ml of support) and ALP (nmol Pi per ml of support). In this study, phospholipid analysis was assumed to represent the active cells inside the biofilm. Fig. 3 shows the accumulation of biomass on carrier particles during the time of operation.
Fig. 2. Performance of the anaerobic inverse fluidized bed reactor treating industrial wine distillery wastewater.
3.2.1. Period 1 The initial VAS into the reactor (12.8 mg ml−1 ) dropped down to 8.8 mg ml−1 the day 26, but the system recovered at
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
1353
4. Discussion
Fig. 3. Biomass accumulation on carrier particles during the operation periods.
the end of this period up to a value of 13.1 mg ml−1 . The effect of continuous increasing in the input loading rate on ALP was similar to the described for the VAS, with a decrease from 40.7 nmol Pi ml−1 to a minimum of 14.3 nmol Pi ml−1 by day 21. Nevertheless, this parameter did not recover itself completely at the end of this period, and the measurement of ALP gave a value of 29.3 nmol Pi ml−1 by day 35. 3.2.2. Period 2 During this period, the VAS concentration increased up to a value of 20.5 mg ml−1 , while ALP dropped down from a maximum of 49.1 nmol Pi ml−1 the day 54 up to 20.0 nmol Pi ml−1 at the end of the period. 3.2.3. Period 3 The abrupt change in the mode of operation of the reactor during this period did not affect negatively either VAS or ALP concentration. VAS and ALP increased up to a value of 23.4 mg ml−1 and 30.7 nmol Pi ml−1 , respectively. The main objective in measuring lipid-phosphate concentration was to find the living cell biomass. Fig. 4 shows VAS against ALP during the operation period. It can be seen that when x = 0, y = 0. This means that when there is no living biomass, expressed as lipid concentration, there are still attached solids. This amount (y = 0) gave a value of 10.7 mgVAS ml−1 CARRIER .
Fig. 4. VAS vs. ALP during the operation periods.
The response to transient changes in OLR of an anaerobic ITB operated at 35 ◦ C in the treatment of industrial wine distillery wastewater appears to be stable and reliable at the operational conditions of this work. There were not noticeable changes about instability in process parameters. VFA concentration was always below 2.6 kg m−3 . CH4 content in the gaseous phase was always between 56.3 and 66.3% while COD removal never fell below 70%. The specific methane production gave a mean value of 0.33 m3 of CH4 per kg of removed COD. This value is lightly inferior to the stoichiometric theoretical of 0.35 m3 of CH4 per kg of removed COD, typical of an anaerobic system. Nevertheless, this data is similar to that obtained by other authors who argue that synthesis of new microorganisms is the reason for this deviation from the theoretical ratio [22]. In spite of three periods of transient changes in the input OLR, it was possible to obtain COD removal between 70 and 92% at OLR from 9.5 to 30.6 kgCOD m−3 day−1 . A comparison with previously studied reactors in our laboratory, treating the same kind of effluent in a continuous loading mode of operation, can be done. Buffiere et al. [23] reported carbon removal rates between 75 and 92% at OLR from 2 to 18 kgCOD m−3 day−1 working with a classical up-flow fluidized bed using 385 m pozzolana particles. In an inverse fluidized bed (with a down-flow liquid fluidization using perlite as biomass carrier), the OLR was increased from 3 to 15 kgCOD m−3 day−1 , but the reactor was destabilized and the carbon removal was only 55% at the end of the experiment and the input load had to be decreased [10]. In a similar reactor (an inverse turbulent bed with an up-flow gas stream using ExtendosphereTM as biomass carrier), the OLR was increased from 2 to 23.1 kgCOD m−3 day−1 with a carbon removal efficiency of about 77% [11]. COD removal attained in this work can be also compared with those obtained from some up-flow anaerobic fluidized bed reactors treating similar substrates in a continuous loading mode of operation. COD removal were observed to decrease from 96.6 to 81.5% at HRT of 60.0 and 11.0 h, respectively, and OLR of 5.7 and 32 kgCOD m−3 day−1 working with an up-flow thermophilic anaerobic fluidized bed reactor using open pore sintered glass beds as biomass carrier [22]. A subsequent applied organic overload produced a fast decrease in pH, which resulted in poor effluent quality (low efficiency of substrate removal, 33.5%). Other wine distillery waste studies have been reported using different anaerobic processes such as anaerobic filter [24] and the UASB reactor [25], with 80 and 83% of COD removal at organic loads of 12.0 and 13.2 kgCOD m−3 day−1 and HRT of 1.4 and 2.4 days, respectively. In this study, total biomass amount is measured in terms of volatile solids concentration and lipid-phosphate concentration. The main disadvantage of the first method is that its estimation includes not only active microorganisms, but also inert mass, exopolymers and absorbed organic matter on flocs and biofilms. Phospholipids, present on bacterial membrane up to 90–98%, do not form part of cell reserves and are easily degradable during bacteria lysis [17]. Therefore, their estimation only includes
1354
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
living biomass. When the relation between these two measures is calculated (VAS/ALP), an estimation of living biomass with regards to total biomass is obtained. Initial biomass concentration of pre-colonized bioparticles used in this study was 12.8 mgVAS ml−1 CARRIER and 40.7 nmolALP ml−1 . At the end of the study, final CARRIER −1 biomass concentration was 23.4 mgVAS mlCARRIER and 30.7 nmolALP ml−1 CARRIER . It means that a biomass concentration almost duplicated in terms of VAS but 25% reduced in terms of ALP multiplied by 1.5 the organic matter treated. It is important to point out a better overall carbon removal rate in the reactor at the end of this study in spite of a biomass measured as lipid-phosphate concentration much lower than at the beginning of the study. It could be explained if a more active biomass is admitted. That increase in the biomass activity cannot be explained by a greater ratio of cells in the biofilm matrix, since ALP decreased significantly, but it suggests a change in biofilm composition. On the other hand, there does not always have to be a relation between the concentration of biomass and the performance parameters of a reactor. A priori, it seems clear that in a system in which the organic matter is degraded only and exclusively by the action of microorganisms, the consumption rate of substratum and the growth or production rate of microbial cell mass must be deeply related. Nevertheless, several factors make these rates not proportional: it is necessary to take into account that the microorganisms cannot be considered chemical reactors since they have the capability of adaptation, and population displacements can occur. This strong selection is not uncommon in reactors fed with industrial wastewater, in which very specific substrata could select very specific groups of bacteria, progressively more active [12,22,26]. Some authors indicate that the friction effects due to high turbulence and shear of fluidized bed reactors may favour growth of thin, dense and active biofilms [22,26]. However, microscopic observations of bioparticles at the end of this work confirmed the presence of biomass on the carrier as local outgrowth colonies, not as uniform covering of the particles (Fig. 5). The type of attached growth or colonial morphology (homogeneous layer, cluster) is a characteristic of the bacterial strains and may significantly influence on spatial distribution of cells [14,27]. Based
Fig. 5. Biofilm coverage of the carrier material at the end of the study.
on these data, the active attached biomass concentration into a fluidized bed reactor could be a consequence of both turbulence and high shear conditions, and progressive biological selection. It is interesting to point out the abrupt decrease of ALP at day 56 (Fig. 3). Since there was not any noticeable change in the monitored parameter, it was assumed that a change in the biofilm composition took place. An important point of the relationship between total biomass amount measured as volatile solids and total biomass amount measured as lipid-phosphate concentration is that is not of y = ax type, as it could be expected, but of y = ax + b type (Fig. 4): when living biomass is completely stabilized, there are still volatile solids into the reactors (within the biofilm). Biomass consists of a biodegradable fraction and an inert fraction, which cannot be biologically degraded. This inert fraction is composed by deadend products [28,29] and rest of extracellular matrix. Based on data of this study, biomass inert fraction would be represented by the ordinate at the y-axis of straight lines shown in Fig. 4: 10.7 mgVAS ml−1 CARRIER . An important fact for effluent quality is the volatile suspended solids concentration. Although detailed measurements were not performed routinely during this study, effluent volatile suspended solids concentration reached a value of 1.46 ± 0.19 mg ml−1 at higher OLR (between day 91 and the end of the study). This value is only lightly superior to that achieved by other authors working with traditional up-flow or down-flow configuration at similar operation parameters [22,26]. The relationship between total biomass amount measured as volatile suspended solids and total biomass amount measured as suspended lipid-phosphate concentration gave y = 0.05x + 0.58. The ordinate at the y-axis of this straight line is almost 20 times inferior to the ordinate for the attached biomass. It can be explained if we assumed that biomass inert fraction would be represented by the ordinate at the y-axis of straight lines so higher ordinates were those obtained for attached biomass, since exopolymer matrix is more abundant in biofilms, up to 80% of volatile solids [30]. One of the major advantages of the anaerobic fluidized bed system over a suspended microbial system is the low volatile suspended solids concentration in the effluent, even at very high loading (hydraulic and organic) conditions. However, in classical up-flow or down-flow fluidized beds, overgrowth of biomass and an important biogas production may lead to operational problems such as particle carry-over, or to fluidization problems such as gas spouting or bed contraction [7]. Concerning the ITB reactor, in which the biogas itself is used as fluidizing agent, these hydrodynamic limitations seem to be overcome. Furthermore, biogas production in the reactor helps to fluidization, which enables to reduce the energy consumption due to biogas recycling at high organic loading rates. Indeed, the bottom of the reactor stands for an internal settler and no external settling device is needed to separate detached biomass from the liquid. Additionally, this settling zone can be used during the start-up period to recover unattached inoculum in order to recycle it into the reactor. From this point of view, inverse fluidization is a very promising technology for the treatment of high strength wastewater. Besides its high carbon removal potential, it over-
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
comes the limitations of traditional fluidized bed reactors. No liquid recycling is required, which prevent from pipe clog up and other maintenance problems. On an industrial scale, recycling of biogas is much easier than recycling of liquid. It is also a robust system, which can tolerate fluctuations in process parameters, such as OLR, with a minimum effluent degradation and its subsequent recovery. 5. Conclusions The ITB reactor appeared to be a good option for anaerobic treatment of high strength wastewater, particularly for the treatment of wine distillery wastewater. The systems attained high OLR with good COD removal rates and it exhibited a good stability to the variations in OLR. At the operational conditions of this work, it was possible to obtain COD removal between 70 and 92% at OLR from 9.5 to 30.6 kgCOD m−3 day−1 . The biofilm at the end of the study was very active and specialized, which can explain the stability of the reactor facing disturbances. Lipid-phosphate concentration has been revealed as a good method for attached biomass estimation since it only includes living biomass. This point is of particular importance in biofilms, in which exopolymer matrix constitutes an important fraction of volatile solids. Acknowledgments This work was financed in part by a grant from the Ministry of Education and Cultures of Spain and by a grant from the Water Management Professional Training Project of the European Community’s Leonardo da Vinci Programme, both to the first author. Appendix A. Nomenclature
ALP COD dp HRT ITB OLR PVC SSA TOC Umf VAS VFA
attached lipid phosphate (nmol ml−1 CARRIER ) chemical oxygen demand (kg m−3 ) particle diameter (m) hydraulic retention time (day) inverse turbulent bed reactor organic loading rate (kgCOD m−3 day−1 ) polyvinyl chloride specific surface area (m2 m−3 ) total organic carbon (kg m−3 ) minimum fluidization velocity (m h−1 ) volatile attached solids (mg ml−1 CARRIER ) volatile fatty acids (kg m−3 )
References [1] M. Green, M. Shnitzer, S. Tarre, B. Bogdan, G. Shelef, C.J. Sorden, Fluidized bed reactor operation for groundwater denitrification, Water Sci. Technol. 29 (1994) 509–515. [2] L. Nikolov, D. Karamanev, The inverse fluidization: a new approach to biofilm reactor design to aerobic wastewater treatment, Stud. Environ. 42 (1991) 177–182.
1355
[3] B. Rusten, L. Hem, H. Ødegaard, Nitrification of municipal wastewater in moving-bed biofilm reactors, Water Environ. Res. 67 (1995) 75–86. [4] J.J. Heijnen, A. Mulder, W. Enger, F. Hoeks, Review on the application of anaerobic fluidized bed reactors in waste-water treatment, Chem. Eng. J. 41 (1989) B37–B50. [5] G.K. Anderson, I. Ozturk, C.B. Saw, Pilot-scale experiences on anaerobic fluidized-bed treatment of brewery wastes, Water Sci. Technol. 25 (1990) 157–166. [6] J.M. Stewart, S.K. Bhattacharya, R.L. Madura, S.H. Mason, J.C. Schonberg, Anaerobic treatability of selected organic toxicants in petrochemical wastes, Water Res. 29 (1995) 2730–2738. [7] P. Buffiere, C. Fonade, R. Moletta, Mixing and phase hold-up variations due to gas production in anaerobic fluidized bed digesters: influence on reactor performance, Biotechnol. Bioeng. 60 (1998) 36–43. [8] Y. Chan Choi, D. Seog Kim, S. Joo Park, S. Koo Song, Wastewater treatment in a pilot scale inverse fluidised bed biofilm reactor, Biotechnol. Lett. 9 (1995) 35–40. [9] D. Garcia-Calderon, P. Buffiere, R. Moletta, S. Elmaleh, Influence of biomass accumulation on bed expansion characteristics of a down flow anaerobic fluidized bed reactor, Biotechnol. Bioeng. 57 (1998) 136–144. [10] D. Garcia-Calderon, P. Buffiere, R. Moletta, S. Elmaleh, Anaerobic digestion of wine distillery wastewater in downflow fluidized bed, Water Res. 32 (1998) 3593–3600. [11] P. Buffiere, J.P. Bergeon, R. Moletta, The inverse turbulent bed: a novel bioreactor for anaerobic digestion, Water Res. 34 (2000) 673–677. [12] C. Arnaiz, P. Buffiere, S. Elmaleh, J. Lebrato, R. Moletta, Anaerobic digestion of dairy wastewater by inverse fluidization: the inverse fluidized bed and the inverse turbulent bed reactors, Environ. Technol. 24 (2003) 1431–1443. [13] T. Kuba, H. Furumai, T. Kusuda, A kinetic study on methanogenesis by attached biomass in a fluidized bed, Water Res. 24 (1990) 1365–1372. [14] V. Lazarova, J. Manem, Biofilm characterization and activity analysis in water and wastewater treatment, Water Res. 29 (1995) 2227–2245. [15] M.C. Arnaiz, C. Ruiz, E. Gomez, I. Garcia, E. Escot, E. Aguilar, J.M. Medialdea, J.C. Gutierrez, J. Lebrato, Evaluation of the efficacy of materials as anaerobic wastewaters treatment supports by phospholipid analysis, in: Proc. of the International Specialty Conference on Microbial Ecology of Biofilms: Concepts, Tools and Applications, Lake Bluff, Illinois, USA, 1998, pp. 294–300. [16] M.C. Arnaiz, Depuraci´on biol´ogica de aguas residuales industrials, Desarrollo de tecnolog´ıa con lechos fluidizados, Thesis Doc., University of Sevilla, Spain, 2002. [17] D.C. White, R.J. Bobbie, J.S. Herron, J.D. King, S. Morrison, Biochemical measurements of microbial mass and activity from environmental samples, in: J.W. Costerton (Ed.), Native Aquatic Bacteria: Enumeration, Activity and Ecology, ASTM Spec. Tech. Publ., University of Calgary, 1979. [18] D.L. Balkwill, F.R. Leach, J.T. Wilson, J.F. Mcnabb, D.C. White, Equivalence of microbial biomass measures based on membrane lipid and cell wall components, adenosine triphosphate, and direct counts in subsurface aquifer sediments, Microbiol. Ecol. 16 (1988) 73–84. [19] R.H. Findlay, M.G. King, J. Watling, Efficacy of phospholipid analysis in determining microbial biomass in sediments, Appl. Environ. Microbiol. 55 (1989) 2888–2893. [20] T.C. Zhang, P.L. Bishop, Density, porosity and pore structure of biofilms, Water Res. 28 (1994) 2267–2277. [21] A.G. Livingston, H.A. Chase, Development of a phenol degrading fluidized bed bioreactor for constant biomass holdup, Chem. Eng. J. 45 (1991) B35–B47. [22] M. Perez, L.I. Romero, D. Sales, Organic matter degradation kinetics in an anaerobic thermophilic fluidised bed bioreactor, Anaerobe 7 (2001) 25–35. [23] P. Buffiere, C. Fonade, R. Moletta, Continuous operation of a fluidized bed bioreactor for anaerobic digestion: residence time influence on degradation kinetics, Biotechnol. Lett. 17 (1995) 833–888. [24] M. Perez, L.I. Romero, D. Sales, Degradaci´on anaerobia termof´ılica de vinazas de vino: efecto del pH de la alimentaci´on, Tecnolog´ıa del Agua 158 (1996) 41–45. [25] A.M. Craveiro, B.M. Rocha, W. Schmidell, Water Treatment Conference Aquatech’86, Amsterdam, 1996, pp. 307–319.
1356
C. Arnaiz et al. / Chemical Engineering and Processing 46 (2007) 1349–1356
[26] P. Buffiere, R. Moletta, Relations between carbon removal rates, biofilm size and density of a novel anaerobic reactor: the inverse turbulent bed, Water Sci. Technol. 41 (2000) 253–260. [27] M.A. Siebel, W.G. Characklis, Observations of binary population biofilms, Biotechnol. Bioeng. 37 (1991) 778–789. [28] W.W. Eckenfelder, Principles of Water Quality Management, CBI Publishing Co., Boston, 1980.
[29] G.R. Marais, Theory, Design and Operation of Nutrient Removal Activated Sludge Processes, Water Research Commission, Pretoria, 1984. [30] Y.M. Nelson, L.W. Lion, M.L. Shuler, W.C. Ghiorse, Modeling oligotrophic biofilm formation and lead adsorption to biofilm components, Environ. Sci. Technol. 30 (1996) 2027–2035.