A New VFA Sensor Technique for Anaerobic Reactor Systems Peter F. Pind,* Irini Angelidaki,* Birgitte K. Ahring BioCentrum-DTU, Technical University of Denmark, Building 227, DK-2800 Lyngby, Denmark; telephone: +45 4525 6183 or +45 45 25 61 86; fax: +45 4588 3276; e-mail:
[email protected] Received 24 April 2002; accepted 5 September 2002 DOI: 10.1002/bit.10537
Abstract: A key parameter for understanding and controling the anaerobic biogas process is the concentration of volatile fatty acids (VFA). However, this information has so far been limited to off-line measurements using labor-intensive methods. We have developed a new technique that has made it possible to monitor VFA online in one of the most difficult media: animal slurry or manure. A novel in situ filtration technique has made it possible to perform microfiltration inside a reactor system. This filter enables sampling from closed reactor systems without large-scale pumping and filters. Furthermore, due to its small size it can be placed in lab-scale reactors without disturbing the process. Using this filtration technique together with commercially available membrane filters we have constructed a VFA sensor system that can perform automatic analysis of animal slurry at a frequency as high as every 15 minutes. Reproducibility and recovery factors of the entire system have been determined. The VFA sensor has been tested for a period of more than 60 days with more than 1000 samples on both a full-scale biogas plant and lab-scale reactors. The measuring range covers specific measurements of acetate, propionate, iso-/n-butyrate and iso-/n-valerate ranging from 0.1 to 50 mM (6–3000 mg). The measuring range could readily be expanded to more components and both lower and higher concentrations if desired. In addition to the new VFA sensor system, test results from development and testing of the in situ filtration technique are being presented is this article. © 2003 Wiley Periodicals, Inc. Biotechnol Bioeng 82: 54–61, 2003.
Keywords: anaerobic manure treatment; in situ filtration; on-line measurement; VFA
INTRODUCTION Volatile fatty acids (VFA) are some of the most important intermediates in the anaerobic biogas process. Conversion of the VFA through the acetogenic and acetoclastic step into methane and carbon dioxide is the most important conversion in the biogas process. It is well recognized that monitoring the specific concentration of VFA can give vital information on process status (Ahring et al., 1995; Alonso,
Correspondence to: Birgitte K. Ahring *Present address: Environment & Resources DTU, Technical University of Denmark, Building 115, DK-2800 Lyngby, Denmark. Contract grant sponsor: Danish Energy Council Contract grant number: EFP no. 1383/98-0014
© 2003 Wiley Periodicals, Inc.
1992; Cobb and Hill, 1991; Hickey and Switzenbaum, 1991; Hill, 1982; Hill and Bolte, 1989; Hill and Holmberg, ¨ ztu¨rk, 1991; Pind et al., 1988; Mo¨sche and Jo¨rdening, 1999; O 1999; Rozzi et al., 1997; Sørensen et al., 1991). Several authors have shown that especially the isoforms of butyrate and valerate are fast and reliable indicators of changes in the process balance (Ahring et al., 1995; Cobb and Hill, 1991; Hill and Bolte, 1989; Hill and Holmberg, 1988), while propionate and butyrate levels are known to increase if the hydrogen level increases to inhibiting levels (Ahring et al., ¨ ztu¨rk, 1991). The 1995; Mo¨sche and Jo¨rdening, 1999; O actual effect of VFA levels on the different microorganisms involved in the biogas process is quite complex and dependent on the actual species involved, e.g., methanogens (Mladenovska and Ahring, 2000; Mo¨sche and Jo¨rdening, 1998; Vavilin and Lokshina, 1996). Inhibition by increased VFA concentration is also known to be dependent on pH (Meyer and Heinzle, 1998; Mo¨sche and Jo¨rdening, 1999). Sudden increases in VFA concentration will cause pH to decrease if the alkalinity is low. This very complex interaction of inhibition, substrate affinity, and pH dependency leads to the conclusion that no general assumptions on inhibitory levels of VFA are possible. Instead, a more complex evaluation of the VFA concentration has to be conducted (Ahring et al., 1995; Bjo¨rnsson et al., 1997; Pind et al., 1999; Pullammanappallil et al., 1998; Renard et al., 1988; 1991). Such a complex evaluation necessitates access to much and frequent information on VFA concentrations. VFA is easily measured using GC or HPLC, provided that all particulate matter has been removed from the sample. When dealing with anaerobic waste treatment, the presence of particulate matter is often high. Only a few reports exist on development of an on-line VFA monitoring systems. Slater et al. (1990) placed a filter on the recirculation loop of a fluidized-bed reactor and transferred the final permeate to a GC with a modified injection port allowing on-line analysis every 12 minutes. However, no data from these measurements and no validation of the suitability of this procedure were ever published. Ryhiner et al. (1993) used a 0.45 m filter on a similar recirculation loop acidifying the permeate with 1% formic acid. The sample was transferred to a specially designed
flow-through vial and then injected into the GC. Acetic, propionic, butyric and valeric acids, including their isoforms, were analyzed and followed for a period of 5 hours. Again, no validation of the procedure was published and this method has to our knowledge not been included in other studies. Zumbusch et al. (1994) used a membrane with a normal molecular weight cut-off (NMWC) of 20,000 g (equal to 20 kDa) for sampling in a UASB reactor. The permeate was pumped through a gas separator and then analyzed in a HPLC using an injection valve. The accuracy of the HPLC was much lower than normally achieved by GC; however, the HPLC was not specifically optimized for this purpose. During the 40 hours the system was operated it showed problems with fouling. To our knowledge, no on-line VFA measurement technique has previously been developed for use on full-scale anaerobic systems treating solid wastes. The published attempts all suffer from problems with membrane fouling and difficulties with the transfer of a representative sample to a GC or HPLC. Fouling of the membrane reduces the permeate flow. A high permeate flow can be obtained by using membranes with a relatively large surface area (0.1–1 m2), but large surface areas result in large hold-up volumes and large flows in the membrane cartridge, thus increasing the demands for flow rates. Instead, small hollow fiber membrane cartridges can be used with a relatively high surface/ hold-up volume ratio. At present, commercial membrane filters are available that can perform ultrafiltration on small sample volumes, provided that the maximum particle size does not exceed 0.1–1 mm. However, solid waste often contains particulate matter of much larger sizes. In the present study we present a newly developed methodology for on-line VFA measurements in manure reactors along with a complete test and validation of the VFA sensor. The development of the sensor required the use of sample purification by filtration, which introduces the possibility of concentration changes in the sample. Therefore, verification of sample reproducibility for each part of the sampling technique has been performed in both laboratory and full-scale applications.
Figure 1.
Schematic layout used for the new VFA sensor.
phosphoric acid and removes precipitates and gasses. The flux/hold-up volume ratio was optimized resulting in a complete breakthrough-time of less than 10 min. Finally, a GC unit is used to perform the actual VFA analysis. Because of the presence of many organic compounds in the ultrafiltered medium extra time is used for burn-out of the GC-column (5 min). This increases the analysis time from optimally 3–5 min to 8–10 min, excluding time required to cool the column. Detailed Sensor System A schematic layout of the sample preparation system is shown in Figure 2. The rotating prefilter (1) has a pore size of 60 m and an effective area of 25 cm2. The prefilter is placed directly in the biogas reactor or in the recirculation loop in a full-scale reactor. Patent application has been filed for both the microfiltration technique and the whole on-line system (Danish patent application numbers PA 2000 01013 and 2000 01014). The hold-up volume between the cleaning valves (6) is 55 mL and is initially flushed for 30 s with prefiltered medium prior to running a recirculation pump (5). The recirculation pump increases the flow and pressure to 1 m/s and 0.8–0.9 atm in an ultra-membrane (4): UFP-
MATERIALS AND METHODS Sensor Principle A schematic layout of the sensor system is shown in Figure 1. The first part is a rotating filter for performing the prefiltration that was placed directly in the biogas reactor. The prefiltration technique was able to supply a sufficient amount of sample to a recirculation loop on a membrane cartridge. Furthermore, the prefiltration technique resulted in a small retention time, less than 1 min. Second, an ultrafiltration unit was used which was able to remove most of the organic particles larger than approximately 100,000 Da. Third, a sample preparation unit mixes the sample with
Figure 2. Schematic illustration of the sample preparation system used for on-line VFA analysis in biogas systems. 1: rotating prefilter placed in situ. 2: sampling port for control extraction by syringe (prefiltered samples). 3: peristaltic pump. 4: ultra-membrane. 5: recirculation pump. 6: three-way valve for bypassing cleaning fluid for regeneration of the ultra-membrane. 7: sampling port for control extraction by syringe (ultrafiltered samples). 8: nonreturn valve. 9: back-flushing pump. 10 and 11: linked peristaltic pumps pumping equal amounts of ultra-filtered sample and 1% (w/v) phosphoric acid (15), respectively. 12: waste pump. 13: overflow pump. 14: pump back-flushing with neutralizing liquid (16). 17: Minifilter. 18: Flow-cell (vial).
PIND ET AL.: A NEW VFA SENSOR TECHNIQUE FOR ANAEROBIC REACTORS
55
Complete VFA sensor tests were conducted on a similar reactor, fed every 6 h, with an average organic loading of 4.45 g VS/L.d (15 d HRT). The temperature was kept at 55°C. Manure composition for both reactor experiments as well as the full-scale reactor is shown in Table I. Gas was collected from the headspace of the effluent storage. The gas flow was monitored using a modified liquid displacement technique similar to the one previously described (Angelidaki et al., 1992), using paraffin oil as liquid in a U-tube with a working volume of 10 mL.
100-E-4A membrane (A/G Technology Corporation) 100,000 NMWC, area 420 cm2. The hold-up volume (approximately 20 mL) of the ultra-membrane cartridge is flushed with filtered permeate for 6 min prior to sampling. The permeate is returned through a nonreturn valve (8) while the flushing is performed, thus minimizing the sampling amount removed from the reactor system. Equal amounts of sample and 1% (w/v) phosphoric acid (15) are then mixed by a peristaltic pump (10) and (11). This sample mixture is first removed by (12), flushing all previous sample hold-up volumes and afterwards passing it through a mini-filter (17), capturing possible precipitate formed in the sample mixture. An open sample vial or container (18) is flushed with the sample mixture, allowing degassing of carbonate. Overflow from the vial (18) is removed by an overflow pump (13). The sample is then transferred by a autoinjector syringe to a GC. The sample is analyzed by using the auto-injector system on a Shimadzu GC-14A equipped with a fused silica column (Nukol Poly(ethylene glycol), 0.53 ID, 15 m, 0.5 m film). Injection: 150°C, detector 250°C, column temperature 90°C (1.5 min) increased to 110°C (during 2 min) increased to 195°C (during 2.12 min) and maintained for 5 min. After injection, the flow-cell (18) and mini-filter (17) are back-flushed with neutralizing liquid (16) to a waste container by (12) and (14). Finally, the flow-cell and tubes are emptied by (12) and the system is now ready for the next sample. The complete set-up (pumps, valves, motors, and GC) is controlled by a specially designed computer program allowing precise and reproducible sample preparation. Validation and testing of the individual parts of the sensor system was performed in both laboratory and full-scale application as described below.
Analytical Methods Gas composition was measured as previously described (Angelidaki et al., 1990). Samples for VFA analysis was taken either from the reactor content (5 mL) , the prefiltration loop [2 mL, (D); Fig. 3] and/or the ultrafiltration loop [1 mL, (7); Fig. 2] by a syringe. Reactor samples were extracted through a sample tube placed on top of the reactor [(F) in Fig. 3] using a 50-mL syringe to flush the tube prior to sampling. Reactor samples were diluted with tap water, and 1 mL was centrifuged after acidification with 30 L 17% (w/v) phosphoric acid in an Eppendorf tube. 0.5 mL prefiltrate or ultrafiltrate sample were mixed with 0.5 mL 1% (w/v) phosphoric acid in an Eppendorf tube and were centrifuged afterwards. The supernatant was analyzed by a GC equipped with a flame-ionization detector. Manually collected samples were analysed on a HP 5890 GC equipped with a HP FFAP column (Bonded and modified crosslinked poly ethylene glycol), 0.53 ID, 30 m, 1.0 m film). Injection: 175°C, detector 200°C, column temperature 100°C (0.5 min) increased to 125°C (during 3 min) increased to 200°C (during 1.67 min) and maintained for 7 min. 1 L sample was injected. All samples were compared to standards in the range from 1 to 50 mM for C1–C4 and from 0.1 to 5 mM for C5. Total solids and volatile solids were determined according to standard methods (APHA-AWWA-WPCF, 1975).
Laboratory Reactors 4.5-L reactors with active volumes of 3.5 L were used for the experiments. The reactor design was as previously described (Angelidaki and Ahring, 1993). Prefiltration tests with a rotating filter were performed on a mesophilic reactor, fed 3 times per day, with an average organic loading of 2.55 g VS/L.d (15 d HRT). Temperature was kept at 35°C. Table I.
Prefiltration Tests A maximal flux capacity of the rotating filter (with a pore size of 60 m) was tested at varying temperature and total
Manure composition for biogas reactors used for testing pre-filter and sensor. Mesophilic reactor experiments
Component TS VS Acetate Propionate Butyrate Valerate Total VFA Carbonate pH
56
g/L g/L g/L g/L g/L g/L g/L g Acetate/L g CO2/L
Thermophilic reactor experiments
Raw cow manure
Influent 60% manure
Reactor manure
Influent 70% manure
Reactor manure
Full-scale reactor manure
87.4 63.7 6.34 2.50 1.47 0.59 10.91 9.72 3.75 7.09
52.7 38.4 3.82 1.51 0.89 0.36 6.57 5.85 2.26 7.09
37.1 27.5 0.07 0.0 0.0 0.0 0.07 0.07 7.75 7.41
76.6 60.2 5.3 1.88 1.14 0.37 8.72 7.84 5 7.41
55 39.5 0.57 0.19 0.02 0.06 0.84 0.77 12 7.6
48.5 30 1.14 0.21 0.04 0.01 1.40 1.38 12.10 8.01
BIOTECHNOLOGY AND BIOENGINEERING, VOL. 82, NO. 1, APRIL 5, 2003
Figure 3. Lab-scale reactor set-up including prefiltration unit. (A) Influent of heating/cooling water to control temperature; (B) rotating filter placed inside reactor; (C) liquid manometer; (D) sampling point for control extraction by syringe; (E) sample outlet for large sample amounts, e.g., calibration of flow rate; (F) sample tube for reactor samples.
solids concentrations. The filter was mounted in a laboratory reactor as previously described (Fig. 3). The reactor contained digested manure (Table I) from a mesophilic reactor with a TS of 35 g/L. The temperature of the reactor was controlled using the reactor heating control. Rotating speed was controlled using a Heidolph RZR 2050 stirrer. A peristaltic pump controlled flow from the filter. A liquid manometer measured the suction vacuum on the filter. Differential pressure varied from 0.05 to 0.15 atm, with increasing differences at higher fluxes. Maximum flux was estimated as the flux obtained just prior to filter constipation, observed as a rapid and irreversible pressure drop. Total solids concentrations were increased by adding raw undiluted manure to the reactor while subtracting an equal amount of volume from the filtered stream, thereby increasing the TS content up to 106 g/L. Comparison of the VFA content in reactor samples and filtered manure was made to validate VFA reproducibility using the 60-m filter pore size. Samples were taken from the mesophilic reactor while conducting pulse additions of manure to obtain higher VFA variation. Ultrafiltration Tests Ultrafiltration tests were conducted on a set-up similar to Figure 2 [without instrument (1), (10)–(18) in Fig. 2]. The ultrafiltration unit consisted of an A/G-UFP-100-E-4A membrane (A/G Technology Corporation), membrane area 420 cm2, average pore size 100,000 NMWC. The flux was measured by temporarily bypassing the permeate to a measure beaker. Permeate samples for VFA analyses were taken from a specially designed septum (hold-up volume less than
0.1 mL). The membrane cartridge volume exceeded the membrane flux that results in a non-instant recovery of VFA concentrations in the ultrafiltered medium. The VFA concentration recovered in the ultrafiltered medium will slowly approach the actual VFA concentration in the media recirculating inside the membrane. Breakthrough of the membrane is defined as the percentage of the VFA concentration recovered in the ultrafiltered medium compared to the VFA concentration in the prefiltered medium recirculation inside the membrane. Breakthrough tests were conducted on a batch of prefiltered manure previously frozen to minimize possible conversion of VFA during the test. Injecting a pulse of acetate momentarily increased the acetate level in the manure. The effect of fouling was monitored by using the complete set-up (Fig. 2) over a longer period of time on a full-scale biogas plant. The system ran continuously and sampling was conducted within 15-min intervals. Each interval included an active pump recycling on the membrane of 6 min resulting in a membrane “use-time” of 9 h and 36 min every day. Fouling would only build up during the 6 min of pumping. If the flux decreased to less than 15 L/(m2 ⭈ h), the membrane was washed with hot water (60°C) for 1 h, followed by back-flushing of the membrane. Full-Scale Test The complete system and the sample preparation system were tested on a full-scale biogas plant (Snertinge Biogas Plant, Denmark). Samples were taken from the recirculation loop servicing two reactors operating in the temperature range from 37–45°C with a reactor TS content ranging from 44 to 48 g/L. Average manure characteristics are shown in Table I. The system was used and tested regardless of temperature and TS changes that occurred as part of the normal operation. During the test, prefiltered and ultrafiltered media were not returned to the sampling point (as opposed to that shown in Fig. 2). Instead, the medium was sent to the influent tank of the plant. Membrane breakthrough was tested for each sample by back-flushing the outer membrane cartridge with 20 mL of distilled water [(9) in Fig. 2], ensuring that VFA from the previous sample would not significantly influence the VFA measurement of the new sample. This enabled estimation of the recovery factor in the ultrafiltered sample, which will decrease with decreasing flux. More than 1000 samples were taken and analyzed automatically during a period of 2 months. Control samples from reactor, the prefiltered and ultrafiltered media were taken periodically during the testing of the sensor system. RESULTS AND DISCUSSION Prefiltration Tests Preliminary test with a fixed filter (dead-end filtration) showed clogging within seconds and would therefore re-
PIND ET AL.: A NEW VFA SENSOR TECHNIQUE FOR ANAEROBIC REACTORS
57
quire a continuous cleaning of the filter. Backwashing was evaluated to be unsuitable for this dead-end filtration. For crossflow filtration, both tubular and plate filtration would require crossflow velocities higher then 1 m/s. This would require unreasonably high recirculation rates in the range of m3/h. Therefore, it was decided to experiment with a reverse implementation of crossflow: moving the filter surface instead of the liquid. Rotating a tubular filter inside the media showed high flux capabilities with very small dimensions. By having the filter freely submerged in the medium, particles larger than the pore size employed could not clog the filter. The pore size of 60 m, was chosen based on commercial availability of filters with a pore size lower then 100 m still having a high pore area per filter area. This unique filter construction occupies only 12–20 cm3 (length: 4–6 cm and diameter: 2–3 cm) inside the reactor system and can be placed in laboratory reactors and even standard piping in full-scale plants. Pumping and pressure control is applied on permeate (not on the medium), thus reducing pumping and hold-up volumes considerably (less than 20 mL). It was clear to us that a certain velocity barrier had to be overcome before a more or less linear relationship between tangential speed and flux capacity was obtained (Fig. 4). This linear relationship can be modelled by a proportionality factor between the tangential speed of the filter surface v2 and the linear flow velocity through the filter surface v3. Adding the two vectored speeds and multiplying with a factor ␣, one should obtain the resulting cross flow velocity v1 of the medium. The factor ␣ is assumed to be dependent on the composition and viscosity of the medium. v1 = ␣ ⭈ (v2 + v3)
(1)
Like the ␣ factor, the linear proportional factor between v2 and v3 is assumed to be a function of particle composition and viscosity of the medium. The 60-m pore size filter could supply 2000 L/(m3 ⭈ h) at 1 m/s at normal reactor conditions: TS lower than 50 g/L and temperatures higher than 30°C, corresponding to approximately 80 mL/min with an area of 25 cm2. It was expected that the hold-up volume of an ultrafiltration unit would be 50–70 mL. Using a flow of 80–100 mL/min prefiltered material a 99% breakthrough could be obtained within 5 min, provided that this hold-up volume was completely mixed, and lower than 1 min at plug-flow conditions. This was considered an acceptable time delay compared to the typical GC analysis time of 10–15 min. It was, therefore, decided to use a pumping rate of 80 mL/min and rotating rate of 1000 rpm during a 3-month test on the mesophilic reactor. An average run time of 2 h per day, 5 d per week was employed to see whether the reactor process was disturbed or showed a decreased biogas yield. The average biogas yield of 218 mL CH4/g VS was compared to an average biogas yield of 215 mL CH4/g VS from a control reactor. The difference of 1.4% was comparable to the hold-up volume in the pumping loop of the filter compared to the reactor volume of 3.5 L. However the difference lies within the accuracy of ± 5% and it was concluded that the use of the rotating filter would not cause extra biological stress on the laboratory reactors. The prefiltered manure had a TS content of 24–25 g/L when the reactor was running at normal conditions. The biogas potential of the prefiltered manure was one third of the reactor content (data not shown). An increase in the TS of the prefiltrate was observed when the reactor TS was artificially increased up to 106 g TS/L, but it never exceeded 35 g/L. Testing of the consistency between the VFA concentra-
Figure 4. Maximal flux capacity of a 60-m pore size rotating filter in digested manure with a TS concentration varying from 35 to 106 g/L as a function of the tangential speed and temperature varying from 7 to 38°C. Differential pressure on the filter was in the range of 0.05 to 0.15 atm.
58
BIOTECHNOLOGY AND BIOENGINEERING, VOL. 82, NO. 1, APRIL 5, 2003
tions of the prefiltered manure and the VFA concentration measured off-line is shown in Figure 5. The average deviation of + 2.85% on the filtered samples is not considered significant since it is based on a few samples obtained in the concentration range between 1 and 5 mM VFA. These samples were obtained by adding a day load of manure to the reactor only once per day. Representative sampling from the reactor is quite difficult under these conditions because of the large hold-up volume (10 mL) in the sample extraction-tube placed inside the reactor [(F) in Fig. 3]. Later, controls from full-scale test showed a very good and consistent correlation of reactor samples and prefiltered samples with a factor of [prefiltrate] ⳱ 0.9922 ⭈ [reactor samples] and a regression factor of 0.9964 (R2) (data not shown). The prefilter showed high reliability during the laboratory and full-scale tests. The filter facilitates sampling, both manually and automatically, from biological systems that exhibit relatively high concentrations of particulate matter. The filter could be applicable not only in biogas reactors but for sampling of many types of biological systems. Ultrafiltration Tests A membrane unit with an average pore size of 100,000 NMWC and a total area of 420 cm2 was chosen on the basis of screening tests. Screening tests with prefiltered manure showed no significant difference in the fluxes for pore sizes ranging from 100,000 to 500,000 (12–15 L/(m2 ⭈ h) at a differential pressure of 1 atm, at room temperature. Membrane flux dependency of pressure had the same characteristics as previous described by Ross et al. (1990). Flux of the applied membrane increased linearly with pressure increases until 1.1 atm. It was decided to use a recirculation pump [Fig. 2 (5)] with a small hold-up volume of 8 mL, which could supply 2.5 L/min at approximately 1 atm. This resulted in a theoretical crossflow rate of 1.2 m/s using water.
Figure 5. Comparison of specific VFA concentrations measured in a lab-scale mesophilic reactor. Reactor samples and samples taken from the recirculation loop of the rotating filter. The line is the best linear fit of the data.
Breakthrough of the membrane could be modelled mathematically by assuming a hold-up volume of 20 mL and a plug-flow volume of 5 mL of the membrane cartridge as shown in Figure 6. On that basis, sufficient breakthrough (99.4%) could be obtained within 5 minutes when the flux exceeded 28 L/(m2 ⭈ h). Ninety-five percent breakthrough could be obtained with a flux of 18 L/(m2 ⭈ h). Note that the breakthrough only corresponds to the relative change in concentration between two measurements. A relative change of 10% between two measurements would be 0.5% off-track with a 95% breakthrough. When the membrane cartridge is flushed/emptied between each sampling, as during the full-scale tests, the time-dependent breakthrough can be observed. For the complete VFA sensor set-up a recirculation period of 6 minutes for the membrane, and a minimal flux of 15 L/(m2 ⭈ h) (minimum 95% theoretical breakthrough) was applied. The sensor system had a samplingtime of 15 minutes (recirculation time on the membrane was less than six fifteenths of the actual exposure time to prefiltered manure). Membrane flux was observed over a period of 2 months testing (see Fig. 7) and showed a satisfying flux if it was cleaned periodically every 15–18 h of use (minimal 38–45 h exposure) or approximately every 200 samples. A declining recovery of the membrane flux was observed within the first 64 h of use. Only hot tap water was used as cleaning agent during the first 64 h. At 64.5 h (indicated by the arrow in Fig. 7) back-flushing of the membrane at point (9) in Figure 2 was introduced during the cleaning procedure and the recovery increased to almost 95%. The flux also increased with increasing medium temperature and even a slight temperature increase resulted in significant flux increase (data not shown). Membrane fouling increased with increasing TS content in the media. The increase in flux with increased temperature and decreased TS content were in agreement with previously reported membrane characteristics of anaerobic systems (Ross et al., 1990). High concentrations of CO2 and CH4 in the reactor samples (because of high pressure at the sampling point) decreased membrane flux when the gas was released at the lower pressure in the sample filtration system. This problem could be avoided by
Figure 6. Breakthrough tests of membrane filter. Low flux 5.3 L/(m2 ⭈ h), high flux 32.5 L/(m2 ⭈ h). All data could be optimally simulated (line) by assuming a 20 mL ideally mixed hold-up volume and a 5 mL plug flow delay.
PIND ET AL.: A NEW VFA SENSOR TECHNIQUE FOR ANAEROBIC REACTORS
59
Figure 7. Membrane flux test, during continuous testing of the VFA sensor system at a full-scale biogas plant. The flux is illustrated as a function of accumulated time where the membrane had been in use. The membrane was periodically cleaned with hot water (1 h) when insufficient flux was observed. Arrow indicates when high pressure back-flushing of the membrane was induced as part of the cleaning.
placing a degassing unit prior to the membrane recirculation loop or by sampling at another point with the pressure closer to atmospheric pressure. The same ultra-membrane was used for more than 1000 samples during the 2 months fullscale testing and additional 1000 samples during 70-d labscale reactor testing before replacement was found necessary. Therefore, it was concluded that the membrane employed could provide sufficient flux if periodical cleaning was employed. Cleaning was controlled manually during the present tests, but could easily be done automatically. Sample Preparation and GC Analysis After prefiltration and ultrafiltration the sample has to be transferred to a GC by techniques insensitive to the presence of gas (carbonate), inorganic, and organic components in the sample. Data obtained from the full-scale tests showed a fixed and constant mixing of sample and 1% (w/v) phosphoric acid throughout the test period (Fig. 8a). A comparison of the same data with controls taken from the sampling point showed that the sensor system had a recovery of 98.14% compared to the manually obtained sample (Fig. 8b), based on best fit of the data. This is consistent with the theoretical expected 95–99.5% breakthrough in the ultramembrane where a similar recovery of 98.03% was found as the best fit (with 6 min recirculation before sampling). The recovery would statistically have been higher if the membrane had not been flushed for each sample. Comparison of VFA concentrations in samples taken directly form the reactor and samples taken from the prefiltrate and ultrafilter showed similar good linear data fit as in Figure 8 (data not shown). Mixing the sample with phosphoric acid immediately caused CO2 stripping. The waste pump (12) and flushing pump (13) had to be configured with 3 times higher flow than (11) and (10) to prevent pressure build-up and overflow because of CO2 stripping (numbers referred to in Fig. 2). Likewise, the mixed sample had to stand in the vial or flush cell for 20–30 s before CO2 had stripped off.
60
Figure 8. Comparison of specific VFA concentrations measured at different points of the sensor system during a full-scale test at Snertinge biogas plant. Line is best fit of the data. (a) VFA concentrations measured by the VFA sensor and control samples taken from the ultra-filtered medium; (b) VFA concentrations measured by VFA sensor and VFA measured in reactor samples.
The presence of other organic and inorganic compounds in the sample did not cause fouling of the tubes and sample preparation system. However, salts did precipitate when the samples were acidified, and removal was found necessary from the sample and system. Placing a mini-filter just prior to the flow-cell captured the precipitating salts. Backflushing with a neutralizing dilution of NaOH dissolved the salts, which subsequently were removed from the system. The liner used in the GC captured inorganic components that could not evaporate at the injector temperature employed. This liner was replaced periodically for every 200– 400 samples. No effects on column retention, detector signal, and reproducibility were seen during the full-scale test with more than 1000 samples. Calibration curves showed reproducibility throughout the test. Accuracy below the standard range was tested using diluted standard solutions. All VFA could be measured with an accuracy of ± 2% between 1–50 mM, ± 5% between 0.1–1mM, ± 10% between 0.05–0.1 mM and ± 30% between 0.02–0.05 mM. CONCLUSION A novel rotating filter unit capable of direct (in situ) filtration in anaerobic reactor systems has been developed and shown to work at high concentrations of TS. Flux capacities increased with increasing temperature, rotating speed, and decreasing TS concentations. Optimal running conditions
BIOTECHNOLOGY AND BIOENGINEERING, VOL. 82, NO. 1, APRIL 5, 2003
were a tangential speed higher than 1 m/s and TS concentrations lower than 50 g/L, in the mesophilic to thermophilic temperature range. Complete recovery of dissolved VFA was seen in the filtered media. A suitable unit of 18 cm3 could be placed in a lab-scale reactor of 3.5 L without any biological stress on the system. A commercial ultramembrane unit could be used for filtrating manure with periodical cleaning with hot water and back-flushing. Membrane breakthrough could be obtained within 5 minutes at optimal conditions and was suitable for automatic measurements of VFA in biogas reactors. A sample preparation system of VFA analysis on manure was proven to work on both lab-scale and full-scale biogas reactors. The system could provide reliable data for more than 200 samples without service. Furthermore, the system could run for more than 100 days with periodically cleaning without replacement of any units. Part of this work was presented at the 8th Anaerobic Digestion Meeting in Antwerpen, September 2001. We thank John Hansen, Erik Nielsen, and co-workers at Snertinge Biogas plant and Thomas Andersen and Michael Tranekjer for technical assistance and proofreading of the article.
Reference Ahring BK, Sandberg M, Angelidaki I. 1995. Volatile fatty acids as indicators of process imbalance in anaerobic digestors. Appl Microbiol Biotechnol 43:559–565. Alonso MS. 1992. Influence of the acetate concentration. Biotechnol Lett 14:535–538. Angelidaki I, Ahring BK. 1993. Thermophilic anaerobic digestion of livestock waste: The effect of ammonia. Appl Microbiol Biotechnol 38: 560–564. Angelidaki I, Ellegaard L, Ahring BK. 1992. Compact automated displacement gas metering system for measurement of low gas rate from laboratory fermentors. Biotechnol Bioeng 39:351–353. Angelidaki I, Petersen SP, Ahring BK. 1990. Effects of Lipids on thermophilic anaerobic digestion and reduction of lipid inhibition upon addition of bentonite. Appl Microbiol Biotechnol 33:469–472. APHA-AWWA-WPCF. 1975. Standard methods: For the examination of water and wastewater. Washington, DC: American Public Health Association. Bjo¨rnsson L, Murto M, Henrysson T. 1997. Comparison of parameters for monitoring anaerobic co-digestion of sewage sludge and industrial waste. Proceedings of the 8th International Conference on Anaerobic Digestion. Biotech Bioeng p 79–82. Cobb SA, Hill DT. 1991. Volatile fatty acid relationships in attached growth anaerobic fermenters. Trans ASAE 34:2564–2572. Hickey RF, Switzenbaum MS. 1991. Thermodynamics of volatile fattyacids accumulation in anaerobic digesters subject to increase in hydraulic and organic loading. Res J Water Pollut Control Fed 63: 141–144.
Hill DT. 1982. A comprehensive dynamic model for animal waste methanogenesis. Trans ASAE 25:1374–1380. Hill DT, Bolte JP. 1989. Digester stress as related to iso-butyric and isovaleric acids. Biol Wastes 28:33–37. Hill DT, Holmberg RD. 1988. Long chain volatile fatty acid relationships in anaerobic digestion of swine waste. Biol Wastes 23:195–214. Meyer B, Heinzle E. 1998. Dynamic determination of anaerobic acetate kinetics using membrane mass spectrometry. Biotechnol Bioeng 57: 127–135. Mladenovska Z, Ahring BK. 2000. Growth kinetics of thermophilic Methanosarcina spp. isolated from full-scale biogas plants treating animal manures. FEMS Microbiol Ecol 31:225–229. Mo¨sche M, Jo¨rdening HJ. 1999. Comparison of different models of substrate and product inhibition in anaerobic digestion. Water Res 33: 2545–2554. Mo¨sche M, Jo¨rdening HJ. 1998. Detection of very low saturation constants in anaerobic digestion: Influences of calcium carbonate precipitation and pH. Appl Microbiol Biotechnol 49:793–799. ¨ ztu¨rk M. 1991. Converstion of acetate, propionate, and butyrate to methO ane under thermophilic conditions in batch reactors. Water Res 25: 1509–1513. Pind FP, Angelidaki I, Ahring BK. 1999. The use of VFA measurements as process indicators in anaerobic reactors treating manure. IAWQ. 2nd International Symposium on Anaerobic Digestion of Solid Waste, Barcelona 15–17 June, 1999. p 41–44. Pullammanappallil PC, Svoronos SA, Chynoweth DP, Lyberatos G. 1998. Expert system for control of anaerobic digesters. Biotechnol Bioeng 58:13–22. Renard P, Dochain D, Bastin G, Naveau H, Nyns E-J. 1988. Adaptive control of anerobic digestion process—A pilot-scale application. Biotechnol Bioeng 31:287–294. Renard P, Van Breusegem V, Nguyen M-T, Naveau H, Nyns E-J. 1991. Implementation of an adaptive controller for the startup and steadystate running of a biomethanation process operated in the CSTR mode. Biomass Bioenergy 38:805–812. Ross WR, Barnard JP, Roux J, Williers HA. 1990. Application of ultrafiltration membranes for solids–liquid separation in anaerobic digestion systems: The ADUF process. Water Sa 16:85–91. Rozzi,A, Massone A, Antonelli M. 1997. A VFA measuring biosensor based on nitrate reduction. Water Sci Technol 36:183–189. Ryhiner GB, Heinzle E, Dunn IJ. 1993. Modeling and simulation of anaerobic waste-water treatment and its application to control design— Case whey. Biotechnol Prog 9:332–343. Slater WR, Merigh M, Ricker NL, Labib F, Ferguson JF, Benjamin MM. 1990. A microcomputer-based instrumentation system for anaerobic wastewater treatment process. Water Res 24:121–123. Sørensen AH, Winther-Nielsen M, Ahring BK. 1991. Kinetics of lactate, acetate and propionate in unadapted and lactate-adapted thermophilic, anaerobic sewage sludge: The influence of sludge adaptation for startup of thermophilic UASB-reactors. Appl Microbiol Biotechnol 34: 823–827. Vavilin VA, Lokshina LYa. 1996. Modeling of volatile fatty acids degradation kinetics and evaluation of microorganism activity. Bioresource Technol 57:69–80. Zumbusch PV, Meyer-Jens T, Brunner G, Ma¨rkl H. 1994. On-line monitoring of organid substances with high-pressure liquid chromatography (HPLC) during the anaerobic fermentation of waste-water. Appl Microbiol Biotechnol 42:140–146.
PIND ET AL.: A NEW VFA SENSOR TECHNIQUE FOR ANAEROBIC REACTORS
61
