• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Scale-Up Considerations for a Hollow-Fiber-Membrane Bioreactor Treating Trichloroethylene-Contaminated Water

Posted on: Wednesday, 12 October 2005, 03:01 CDT

By Pressman, Jonathan G; Georgiou, George; Speitel, Gerald E Jr

ABSTRACT:

Scale-up of a hollow-fiber-membrane (HFM) bioreactor treating trichloroethylene- (TCE-) contaminated water via co-metabolism with the memanotroph Methylosinus trichosporium OB3b PP358 was investigated through cost comparisons, bioreactor experiments, and mathematical modeling. Cost comparisons, based on a hypothetical treatment scenario of 568-L/min (150-gpm) flowrate with an influent TCE concentration of 100 g/L, resulted in a configuration of treatment trains with two HFM modules in series and an overall annual cost of $0.36/m^sup 3^ treated. Biological experiments were conducted with short lumen and shell residence times, 0.16 and 0.40 min, respectively, as a result of the cost comparisons. A new variable, specific transformation, was defined for characterizing the cometabolic transformation in continuous-flow systems, and values as large as 38.5 g TCE/mg total suspended solids were sustainable for TCE treatment. Using mathematical modeling, HFM bioreactor system design was investigated, resulting in a five-step system design strategy to facilitate sizing of the unit processes. Water Environ. Res., 77, 533 (2005).

KEYWORDS: hollow-fiber membrane, bioreactor, trichloroethylene, methanotrophs, biodegradation.

doi:10.2175/106143005X67458

Introduction

Contamination of groundwater and soils with chlorinated aliphatic solvents is a widespread problem. One promising approach for treating chlorinated solvents is to destroy them through co- metabolism in aerobic biological processes. However, the difficulty in developing bioreactors for treating chlorinated-solvent waste streams is the method of contacting or transferring the chlorinated solvents to the bacterial population for biodegradation. The hollow- fiber-membrane (HFM) bioreactor is based on a specialized microorganism, Methylosinus trichosporium OB3b PP358 (Phelps et al., 1992), developed in our laboratory and on a new type of bioreactor, a HFM reactor. A HFM reactor provides a protective barrier between the organisms and the contaminated water or air, because only the volatile chemicals can cross the membrane. The reactor also allows a very high degree of control over the biological process, so that maximum biodegradation rates can be obtained.

Research with the HFM bioreactor system has been ongoing for several years (Pressman, 2001; Pressman et al, 1999 and 2000). Previous efforts have dealt with process-engineering issues and technology demonstrations. This report focuses on considerations for system scale-up and competitiveness with existing technologies. To start, a simple cost comparison was conducted to evaluate methods for ensuring that the HFM system is competitive with existing technologies. The results of the cost comparison indicated that higher flowrates were required; thus, high flowrate experiments were conducted to demonstrate the HFM bioreactor system for the destruction of trichloroethylene- (TCE-) contaminated water at flowrates in the range of economic comparison to existing technologies. Finally, a mathematical-modeling analysis of system variables was conducted to develop a basic system design strategy for scale-up.

Materials and Methods

Chemicals. Trichloroethylene (Fisher Scientific, Hampton, New Hampshire) and chloroform (EMD Chemicals, Gibbstown, New Jersey) were both American Chemical Society certified and 99+% pure. Bromodichloromethane (Chem Service, West Chester, Pennsylvania) was 99% pure. Optima-grade methanol (Fisher Scientific) was used as growth substrate for the microorganisms and to prepare stock solutions for gas chromatography. Pesticidegrade n-pentane (Fisher Scientific) was used for extracting chlorinated solvents from water. Streptomycin sulfate, nalidixic acid, and cyclohexamide (Sigma Chemical Co., St. Louis, Missouri) were added to minimize contaminating organisms. Denatured ethyl alcohol (Fisher Scientific) was mixed in a 70/30 volume ratio with water to sterilize the HFM module (HFMM).

Microbial Culture. Methylosinus trichosporiuin OBSb PP358 is a methanotrophic bacterium that has the ability to rapidly cometabolize chlorinated solvents. Strain PP358 is a spontaneous mutant of wild type OB3b, in which soluble methane monooxygenase (sMMO), the enzyme known to actively degrade chlorinated solvents, is constitutively produced in the presence of up to 0.8 mg/L copper (Phelps et al., 1992). OBSb strain PP358 is also resistant to antibiotics (Phelps et al., 1992), allowing 20-mg/L streptomycin sulfate and 20-mg/L nalidixic acid to be added as a means of minimizing contamination by other organisms, and 20-mg/L cyclohexamide to be added to control fungal growth. Methanol was used as the growth substrate in this research. PP358 was grown in nitrate minimal salts medium (Fitch et al., 1996), with methanol at an influent concentration between 0.3 and 0.9% (vol./vol.). Cells were grown in a 2.5-L Bioflow III fermentor (New Brunswick Scientific, Edison, New Jersey) operated as a chemostat with 2.5-L culture volume at 29C, pH 6.90 0.10, dissolved oxygen (DO) (Ingold, Bedford, Massachusetts) 20 to 90% of saturation, 90- to 450-rpm agitation, 400- to 1000-mL/min air feed rate, and a dilution rate of 0.03 h^sup -1^.

Hollow-Fiber-Membrane Module. The HFMM was a LiquiCeI high- efficiency phase contactor (Celgard Limited Liability Company [LLC], Charlotte, North Carolina) using polypropylene fibers housed in a polypropylene shell. The HFMM contained 10 176 fibers with an internal diameter of 240 m, a 0.03-m pore size, and 40% membrane porosity. Membrane pores remained air-filled because of the pore size and hydrophobic nature of polypropylene. Contaminated water flows through the inside (lumen) of the fibers, and a solution containing the methanotrophs flows on the outside (shell) of the fibers. Chlorinated solvents are quite volatile and readily transfer across the membrane from the lumen to the methanotrophs on the shell side. The hydraulics are plug-flow in the lumen and radial cross- flow in the shell (Figure 1). Countercurrent contacting was used in the baffled radial-flow HFMM because studies indicated that the rapid mass transfer could only be realized through this flow scheme.

Figure 1-Hollow-fiber-membrane-bioreactor schematic.

Analytical. Aqueous TCE samples were taken in 25-mL vials and prepared by liquid-liquid extraction using pentane as the extraction fluid and 1 mg/L chloroform as an internal standard. The extract was analyzed by gas chromatography (GC) (electron capture detector, DB- 1701 column [J & W Scientific, Folsom, California], and 50 to 70C) and compared with known standards. Samples for methanol analysis were taken in 1.5-mL Eppendorf centrifuge tubes with 6N hydrochloric acid to inactivate the organisms and suppress methanol use. The tubes were centrifuged for 8 minutes to remove biomass, and the supernatant was analyzed by GC (flame ionization detector, DB-WAX column [J & W Scientific No. 125-7032]) and compared with known standards.

Experimental Equipment and Procedures. A schematic of the experimental apparatus is shown in Figure 1. Distilled water served as the lumen feed. Aqueous solution saturated with TCE was mixed with the lumen feed in a 1200-mL mixing vessel to provide constant delivery concentrations. The TCE-contaminated water was then pumped through the lumen of the HFMM using a peristaltic pump and ultimately directed to a waste sink.

Once the microorganisms reached a steady-state optical density at a wavelength of 600 nm, biological experiments were started by initiating lumen and shell flow through the HFMM. The cell suspension was pumped from the chemostat to the shell of the HFMM. Following the HFMM shell was a 340-mL kinetics reactor installed to measure the pseudo-first-order degradation rate constant during the biological experiment. The kinetics reactor was created , from 15.2 m (50 ft) of coiled 6.4-mm (0.25-in.) stainless-steel tubing to promote plug-flow hydraulics. Previous studies have indicated that pseudo-first-order degradation rates apply at all concentrations studied in these experiments (Aziz et al., 1999). The pseudo-first- order degradation rate constant (k^sub 1^) was then calculated from the change in concentration between the influent and effluent of the kinetics reactor, using ideal plug-flow hydraulics (Pressman et al., 2000). This method allowed k^sub 1^ to be determined at every sampling time.

A polishing reactor was designed to allow for residual TCE degradation to occur before the organisms were returned to the growth chemostat. The polishing reactor was placed directly after the kinetics reactor. Countercurrent flow within the HFMM lumen caused significant TCE to remain in the effluent, thereby creating a need for additional residence time in the polishing reactor (3.2 L) to complete degradation.

During biological experiments, optical density, dissolved oxygen, methanol, pH, and dilution rate were monitored in the chemostat. Samples from the chemostat were plated on plate-count agar, which does not support growth of OB3b, to determine contamination by other organisms. In addition, samples from the HFMM lumen influent, lumen effluent, shell influent, shell effluent, kinetics-reactor effluent, andpolishing-reactor effluent were collected for analysis by GC. Dissolved oxygen was measured throughout the system by taking 30-mL samples for analysis with a YSI dissolved-oxygen meter (model 54A) (Yellow Springs, Ohio).

Membrane and shell mass-transfer correlations from previous modeling efforts remained the same.

The model was used in two ways: (1) to solve for experimental parameters (best-fit), and (2) to ascertain the predictive ability of the model (model predictions). In the best-fit method, the model was fitted to the experimental data by adjusting the overall mass- transfer coefficient (K^sub L^), until the normalized residual error was minimized. Normalized residual error is defined as the shell and lumen sum of the squared difference between experimental- and modeled-effluent concentrations, divided by the respective squared experimentaleffluent concentration. The model was also used in a predictive mode, assessing the ability to accurately simulate bioreactor performance in biotic experiments. The user must input expected conditions, such as flowrates, lumen influent TCE concentration, biomass concentration, and pseudo-first-order biodegradation rate constant. With the entered values and the K^sub L^ calculated from mass-transfer correlations, the model predicts lumen- and shell-effluent concentrations. In biotic experiments, because all the decision variables and parameters are determined independently, the model is truly predictive and can be used as a tool for planning experiments and for scale-up purposes.

Cost Modeling. The HFM bioreactor system costs were modeled in an Excel spreadsheet (Microsoft Corporation, Bellevue, Washington), Capital-cost factors for HFMMs, plug-flow reactors, chemostat, and pumps were obtained from equipment manufacturers and used to generate overall capital costs. Operational-cost factors for methanol and nutrients, heating, maintenance, energy, and labor make up the operation and maintenance costs. These costs were amortized over a 10-year project period and reported as dollars per cubic meter treated to facilitate comparisons.

Results and Discussion

Preliminary Cost Comparisons. Using the HFM bioreactor cost model, a variety of designs were developed to illustrate how the costs might change with various design and operating conditions. A typical design scenario was used to program the cost model. The scenario was based on the treatment of water contaminated with 100 g/ L TCE, and the decision variables and parameters are shown in Table 1. The design uses the Celgard 10 28 HFMM, which is a full-sized commercial module with 224 640 fibers, a 254-mm (10-in.) diameter bundle and a 61.28-cm length. The module and fibers are of similar construction to the baffled radial-flow, laboratory-scale HFMMs used in previous experiments. One significant uncertainty in the analysis is the cost of the HFMMs. Although $10,000 was used for the analysis, the membrane manufacturer, Celgard LLC, has not determined HFMM prices for this application. Therefore, significant differences in overall capital costs could occur based on the actual price charged.

Table 1.-Hollow-fiber-membrane bioreactor, cost-analysis decision variables and parameters.

The results for six separate designs are summarized in Table 2. The differences among the designs are highlighted by italicizing the key decision variable or parameter in each. Costs for air stripping and adsorption systems were previously investigated by Dvorak (1994). Based on his research, conventional treatment, consisting of air stripping and adsorption systems, would cost approximately 0.15 to 0.20 $/m^sup 3^ for the design scenario considered here. Ultimately, it will be necessary to bring the economics of HFMM bioreactor technology down to this level to guarantee its success. At this stage of technology development, it was important to determine if the preliminary design costs of the HFMM bioreactor system were near that of conventional treatment. Specific comments on each design follow:

* Design 1 assumes all modules are arrayed in parallel; the costs indicate that this scenario is the least desirable option.

* Designs 2 through 4 are for two modules in series. Putting two in series substantially reduces the cost because much more water can be processed through each module. The differences in designs 2 through 4 are related to the amount of biodegradation that occurs in the module itself versus that in an external degradation reactor. Thus, design 2 requires a smaller degradation reactor than designs 3 or 4. The cost decreases as more water is processed through the modules, but the amount of biodegradation also decreases because of the shorter fluid-residence time. In the extreme, the modules would serve only as mass-transfer devices as opposed to a mass transfer/ biodegradation reactor. The minimum extent of biodegradation was arbitrarily limited to 10% in these designs. As the biodegradation in the module decreases, a larger external degradation reactor is needed, but these are significantly less expensive than membrane modules.

Table 2-Cost summary of various HFM bioreactor system designs.

* Design 5 is for three modules in series. The benefits of putting three in series seem marginal, especially because the overall cost is the same as for two modules in series.

* Design 6 shows what would happen to the cost of two modules in series if mass transfer was better than that suggested by the correlation in our model (in this case, a mass-transfer coefficient of 1.67 times the model correlation). This highlights the sensitivity of the design to mass transfer. It is important to recognize that some uncertainty exists with this mass-transfer correlation in going to a larger module and also the higher velocities envisioned in these designs.

The most promising designs are those with two modules in series (designs 2 through 4). The number of modules is reduced to a reasonable level, and the costs are reasonably close to the costs of conventional treatment. An experimental approach was devised to simulate these designs in the laboratory-scale HFMMs, as an approximation of the full-scale system. This was accomplished by scaling down the flowrates from the full-scale system to the laboratory-scale system, through the use of equivalent lumen and shell residence times.

Biological Experiments. The HFM bioreactor system was operated under high-flowrate conditions in experiments HF1 and HF2 (Table 3). In experiment HF1a, the flowrates were set to match design 3 (Table 2), resulting in residence times of 0.21 and 0.57 min for the lumen and shell, respectively. After the system was demonstrated to be capable of sustaining these conditions, the residence times were decreased to 0.16 and 0.41 min for the lumen and shell, respectively (experiment HFIb). The flowrates for this set of conditions were chosen to simulate design 4 (Table 2), representing the shortest residence times tested in this research. The conditions of experiment HFIc were essentially the same as HFIb; however, at this point, the experiment was entering process failure, and it was more useful to separate the results for analysis. For experiments HFIa and b, a majority of the TCE was removed from the contaminated lumen stream (59 to 71%), and, of the transferred TCE, nearly all was ultimately biodegraded (92 to 95%). A comparison between the TCE degraded in the HFMM only (26 to 28%) and the HFMM plus the kinetics and polishing reactors (92 to 95%) demonstrates that the polishing reactor is an integral portion of the system. Any successful design of the system would require a properly sized polishing reactor for residual TCE degradation.

Table 3-Summary of HFM bioreactor experiments HF1 and HF2.

The mass-rate data from experiment HFl are shown in Figure 2. The figure shows the mass flowrates of TCE in the system versus operating time. The vertical lines denote times at which conditions were changed, and the residence times are shown on the figure. During the first set of conditions (HFIa), stable operation was . maintained for six days before changing any major system parameters. Following day six, the residence times were decreased, which significantly increased the lumen influent TCE mass flowrate. Two days of operation were observed before the system began to enter process failure. Then, in the last third of the figure, the data from the final few days during process failure are shown. While the two days represented by HFIb do not include enough data to verify steady-state operation, the data from HFIb give a sense of operation at these conditions. Process failure was most likely caused by extensive growth of organisms other than OB3b, as indicated by plate- count-agar plates of the chemostat, but it was not possible to determine whether the increased TCE mass rates in HFIb played any role in inactivating OB3b and allowing other organisms to thrive.

Figure 2-TCE mass flowrates in biological experiment HF1.

Figure 3-Measured TCE pseudo-first-order degradation-rate constant and biomass concentration in experiment HF1.

Examining the difference between the lumen-influent and lumeneffluent data in Figure 2 gives a graphical representation of the mass of TCE that transferred to the shell of the membrane. The polishingreactor data remained near zero during the experiment, indicating that nearly all of the transferred TCE was degraded before the organisms returned to the chemostat. Also, the difference between the shell-effluent and polishing-effluent mass flowrates gives a sense of the mass of TCE undergoing degradation within the polishing reactor.

Figure 3 shows the individual daily data for pseudo-first-order degradation rate constants and biomass concentration. With the exception of one measurement, k^sub 1^ was greater than 1 L mg^sup - 1^d^sup -1^ for over 175 hours. This was the largest sustained value of k^sub 1^ observed throughout this research and \compares closely with the best rates observed by other researchers using OB3b PP358 in continuous-flow systems (Aziz, 1997; Fitch, 1996). The biomass concentration was very stable throughout experiments HFIa and HFIb. However, as the system began to destabilize, the biomass concentration started to decrease even before the degradation rate constant. This was not expected, because in previous investigations, as k^sub 1^ decreased and contaminating organisms took over, the biomass concentration would increase greatly. In this experiment, contaminating organisms, while present, were not able to significantly overtake the reactor. Perhaps the mass of TCE was so large that it became toxic to the contaminating organisms and reduced their growth rates significantly.

Dissolved oxygen and methanol were also monitored in experiment HFl. Because DO transfers in the HFMM in the same manner as the volatile chlorinated solvents, available DO for the organisms was predicted to decrease, based on DO transfer from the shell to the lumen. Furthermore, this lead to a concern over maintaining enough DO for the organisms to completely cometabolize TCE before returning to the chemostat. Anoxic or anaerobic conditions in the polishing reactor would have adverse effects on process operation. While the DO results from experiment HFl demonstrated approximately 2 mg/L DO transferred from the HFMM shell to the lumen, a significant DO concentration (approximately 3 mg/L) remained in the polishing reactor, allowing for continuous TCE degradation.

The influent methanol-feed concentration to the chemostat was 3500 mg/L for experiment HFl. Based on the true yield (0.4 mg biomass/mg methanol) determined by Fitch (1996), and with nearly complete methanol use, the expected biomass concentration would be approximately 1400 mg total suspended solids (TSS)/L. From Figure 3, the biomass concentration during the initial steady-state operation was between 700 and 800 mg TSS/L. This implies that, either influent methanol was not completely used by the organisms for cell growth, or endogenous decay was unusually large. Determining the fate of the methanol within the system is important because of the direct effect that methanol has on overall operating costs. There are two potential fates for methanol besides biological use. The methanol may be either volatilized into the chemostat air supply and swept away in the effluent gases or carried out of the chemostat in the liquid overflow. Because less than 2 mg/L methanol was measured in the chemostat-liquid effluent, the excess methanol was most likely volatilized into the chemostat airflow. This raises a complex issue related to the careful balance between air supply to the chemostat and methanol volatilization that would need to be considered for system scale-up. Finally, because the DO was always of sufficient concentration, this implies that methanol was the limiting nutrient, and adding more methanol would increase the biomass concentration. This theory was tested in experiment HF2, where the influent- chemostat methanol concentration was raised to 4700 mg/L. This resulted in a larger biomass concentration of over 1000 mg TSS/L, as illustrated in Table 4.

Experiment HF2 was designed to further stress the system by operating at the highest flowrates of HFl with an increased concentration (199 g/L for HF2a). The main purpose for this experiment was to increase the specific transformation, which is discussed below. Process failure was once again occurring during HF2b; thus, separating the analysis into the two parts allows a better look at the sustained conditions of HF2a. In HF2a (Table 3), the system was successfully operated for 100 hours under conditions similar to experiment HFIb. Therefore, the cause of process failure in experiment HFl was most likely attributable to contaminating organisms. Comparing HF2a to HFIb shows that slightly more mass transfer occurred when the influent concentration was greater. This would be expected, as the concentration driving force is increased when the concentration is higher, leading to increased mass transfer.

Table 4-Summary of HFM bioreactor modeling for experiments HF1 and HF2.

Modeling of Biological Experiments. The average conditions of each phase of biological experiments HFl and HF2 were modeled using both the best-fit method and mass transfer correlations (Table 4). The average lumen-influent TCE concentrations (Table 3), biomass concentrations, and pseudo-first-order degradation rate constants were used in both cases. In the best-fit method, the mass-transfer coefficient was varied until the normalized residual error was minimized. The normalized residual error was less than 0.024 for all conditions tested, indicating a close fit between the modeled effluent concentrations and the experimental effluent concentrations. Using the mass-transfer correlations resulted in normalized residual errors that were slightly larger, ranging between 0.003 and 0.112. The maximum relative error between lumen predictions and experimental values was 15%. Shell-effluent concentrations were generally overpredicted, with a 28% maximum relative error between the predicted shell effluent and experimental shell effluent observed in experiment HFIb.

Comparing the two methods of modeling, mass-transfer coefficients from the model predictions with correlations were larger than those of best-fit modeling only for experiments HFIb and HFIc. Mass- transfer coefficients for the remainder of experiments were nearly identical for both modeling approaches, indicating that the mass- transfer correlations worked well in simulating performance in the high-flowrate experiments.

Figure 4-Modeling of biological experiment HF2 using measured k^sub 1^ and correlated K^sub L^.

Because of the accuracy of the mass-transfer correlations, the model was operated in the predictive mode, entering the flowrates, influent concentration, biomass concentration, and measured rate constant for each sample of experiment HF2. Concentration versus time data are presented for the modeling results and experimental data (Figure 4). Overall, the model predictions were excellent, especially considering the true predictive simulation of the experiment. The predicted lumen-effluent concentrations matched the experimental data very well. The predicted shell-effluent concentrations were also close to the experimental data, except that the model had some difficulty closely simulating shell-effluent concentrations from earlier times in the experiment. The ability to closely simulate HFM bioreactor performance with a mathematical model compliments laboratory experiments, because modeling can be conducted much more quickly than experiments.

Specific Transformation. Many investigators have found that the oxidation of chlorinated solvents by sMMO causes product toxicity, which is proportional to the amount of chlorinated organic degraded (Alvarez-Cohen and McCarty, 1991b; Fox et al., 1990; Oldenhuis et al, 1991). Thus, the cells have a finite capacity for transforming chlorinated solvents. To account for this limited degradation capacity, transformation capacity (T^sub c^) was defined as the maximum mass of compound that can be degraded per mass of cells before inactivation (Alvarez-Cohen and McCarty, 1991a).

Thus, specific transformation is a measure of the total mass of TCE degraded by the organisms before they are discarded in the waste.

Although specific transformation is similar to transformation capacity, it is important to make the distinction that specific transformation represents the actual conditions organisms are exposed to in a bioreactor, while transformation capacity represents a characteristic of the organisms: the maximum chlorinated-solvent transformation possible before complete inactivation. The Tc values for OB3b PP358 were previously measured (Fitch et al, 1996); Tc ranged between 54 and 87 g TCE/mg TSS for growth conditions similar to those of this research. Because specific transformation is a new decision variable, comparative values are not reported in the literature. An attempt was made to calculate specific transformations from various reports; however, in every case, at least one parameter necessary for the calculation was absent.

In previous research, experiments were specifically designed to operate with low specific transformations to minimize the potential effect that product toxicity could have on steady-state operation. The maximum specific transformation from previous work was 14.5 g TCE/mg TSS (Pressman, 2001). While planning the high-flowrate experiments, specific transformation was quickly recognized as an important decision variable related to HFM bioreactor system design. With limited experimental knowledge on specific transformation, the designs in Table 2 were completed by limiting the maximum value of specific transformation to 30 g TCE/mg TSS. Because of the important role that specific transformation plays in system design and costs, much effort was spent in this phase of the experimental work to define the upper limits of this decision variable, while maintaining stable system operation. Specific transformation in experiment HFIa was 26.2 g TCE/mg TSS. With decreased residence times and increased TCE concentration, experiment HFIb demonstrated operation with a specific transformation of 36.4 g TCE/mg TSS. However, a very limited amount of data was taken before entering process failure (HFIc), and it was not clear whether the process failure was related to the high value of specific transformation. At the same time, contamination of the culture by other organisms was observed, so the contamination also could have caused the process failure.

Experiment HF2 was undertaken to confirm values of specific transformation in the 30-g TCE/mg TSS range and attempt to increase it even more. Experiment HF2a successfully proved that specific transform\ation could be maintained at 38.5 g TCE/mg TSS for 100 hours. Upon increasing the influent TCE concentration further for experiment HF2b to increase the specific transformation even more, contamination once again was detected, the system began to fail, and the average specific transformation for experiment HF2b was only 25.9 g TCE/mg TSS. The highest observed specific transformation for TCE of 38.5 g TCE/mg TSS is nearly one-half of the maximum transformation capacity determined by Fitch (1996). Conservative system designs would certainly set a maximum for specific transformation that is less than the value that causes complete cell inactivation. The values tested in these experiments were probably near the upper range of what should be considered in design.

System-Design Overview. To foster the development of this technology on a commercial scale, a methodology to design full- scale treatment units must be developed. With an appropriate engineering-design strategy, HFM bioreactor treatment systems can; be designed for many different waste sites with different flows and types of contamination. Developing a design strategy is a significant task because of the many decision variables (properties under the designer's control) and parameters (properties not under the designer's control) inherent to this technology.

The many decision variables and parameters can be broken down into two categories, based on the lumen or shell side of the system. On the lumen side, the parameters include the lumen-influent concentration, the lumen-effluent concentration and lumen flowrate. The lumen-influent concentration and flowrate will be given by the site contamination, and the lumen-effluent concentration will be specified by the cleanup requirements. The lumen-HFM volume is a decision variable and will be based on the commercial membrane modules chosen and the number used for a specific treatment scenario. Thus, design of the lumen side of the process is relatively straightforward, but has significant cost implications. Membrane modules comprise a very large fraction of the bioreactor- system capital cost; therefore, the economics are most attractive when the lumen residence time is minimized, so as much contaminated water or air as possible is treated per membrane module.

Most of the decision variables are associated with the shell side of the process and include biomass concentration, shell side flowrate, HFM shell volume, polishing-reactor volume, specific transformation, chemostat volume, chemostat dilution rate, and methanol and oxygen concentrations. The interactions among these decision variables are quite complex, yet need to be understood to optimize the operation and cost of the process. The pseudo-first- order rate constant is the main parameter on the shell side of the process.

Because the mathematical model closely simulates process performance, it was used to investigate the most critical decision variables and parameters over a wide range of possible values to understand the process sensitivity to changes in the decision variables and parameters. Designs consisting of two HFMMs in series were investigated because the initial cost analyses showed this configuration to be the least costly. To facilitate the analysis, the typical design scenario identified in Table 1 was used. Because the initial cost comparison indicated that the total project cost is highly dependent on the membrane capital cost, maximizing the lumen flowrate (or minimizing the lumen residence time) was identified as an important design goal. The resulting analyses were, therefore, conducted with an emphasis on relationships with lumen residence time.

Hollow-Fiber-Membrane-Module Modeling. Figure 5 details the effects of lumen residence time and shell/lumen residence time ratio (θ^sub S^/θ^sub L^) on percent mass transfer and biodegradation for the first HFMM in the treatment train. The performance of the second HFMM is nearly identical (Pressman, 2001). As anticipated, with decreasing lumen residence time, the percent mass transfer decreases because less time is available for the lumen fluid to contact the membrane surface. Little biodegradation occurs in the HFMM at short lumen residence times because the corresponding shell residence times are short. For short lumen residence times the HFMM is a mass-transfer device, and nearly all of the degradation must occur in the polishing reactor. As the lumen residence time increases, and, thus, the shell-residence time also increases, more biodegradation occurs in the shell.

An analogous view of the previous relationship can be seen in Figure 6. When the lumen-influent concentration is 100 g/L, 77.7% mass transfer must be realized by each of the two HFMMs in the treatment train to meet the treatment goal. If a specific lumen residence time is chosen, the residence-time ratio required to meet the treatment goal can be determined from Figure 6. The three additional lines in Figure 6 represent increasing lumen-influent concentrations. To obtain the required increase in mass transfer as the influent concentration increases, the lumen residence time must increase, the residence time ratio must decrease, or a combination of both must be applied, illustrating an important design relationship. The required mass transfer in a treatment train of HFMMs is given when the influent concentration and treatment goal are specified. The designer can then choose combinations of lumen residence time and residence-time ratio that meet the problem specifications. Other decision variables and parameters, such as biomass concentration and pseudo-first-order degradation-rate constant, will shift the lines on Figure 6 in varying directions, but the combination of lumen residence time and residence-time ratio will remain critical to meeting the required percent removal.

Figure 5-Effect of lumen residence time and residence-time ratio on mass transfer and degradation in the HFMM.

Polishing-Reactor Modeling. To make the economics more attractive, all membrane modules in a given treatment stage were assumed to be served by a single polishing reactor. Thus, a design with parallel trains of two HFMMs in series would be served by two polishing reactors. Figure 7 shows the effect of lumen residence time and residence-time ratio on polishing-reactor volume for the entire 568-L/min (150-gpm) project flowrate. Polishing-reactor volume is calculated using ideal-plug-flow hydraulics. The analysis was simplified by only considering the polishing-reactor volume from the first HFMM in each treatment train. The shell/lumen residence time ratio significantly affects the required volume because as the ratio increases, more degradation occurs in the HFMM, and less polishing reactor volume is required. The balance between the relative biodgradation in the HFMM versus the polishing reactor would ultimately be based on an economic analysis identifying the least cost feasible design. An analysis of the polishing-reactor volume for the second HFMM in each treatment train would be similar, except that the required polishing-reactor volume would be different because the concentrations entering the polishing reactor from the second HFMM are lower.

Figure 6-Effect of lumen residence time and lumen-influent concentration on residence-time ratio and percent mass transfer required to meet 5-g/L treatment goal.

Figure 7-Effect of lumen residence time and residence-time ratio on polishing-reactor volume for entire project flow/rate (568 L/min [150 gpm]) associated with first HFMM in each treatment train.

The foregoing analysis represents a somewhat simplified view of the system in that possible substrate use and organism growth and endogenous decay within the HFMM shell and polishing reactor were ignored. The HFMM shell and polishing reactor can be viewed as a recirculating sidestream system attached to the chemostat; the assumption in this analysis is that only TCE co-metabolism occurs in this sidestream. Certainly, the HFMM shell and polishing reactor could be considered in a more sophisticated analysis of the system; however, experience to date does not suggest that this level of complexity in modeling is warranted.

Design Strategy. The above discussion is the first effort to investigate the interactions among the many decision variables and parameters involved with the design of the HFM bioreactor system. With respect to an engineering design strategy, the analysis indicates that many feasible designs exist for any given scenario. Minimizing costs, maintaining system reliability, and providing process redundancy would likely be major factors in the overall design.

The overall goal of the preceding analysis was to understand the interactions of the many decision variables and parameters and develop a system design strategy. The proposed design strategy is as follows:

(1) The design is initiated from the known lumen-influent concentration and desired treatment goal. Based on the previous discussions, two HFMMs in series are assumed to create one treatment train. Calculate the percent removal required across the two HFMMs in series to meet the required treatment goal. The percent removal across each HFMM in a treatment train is equal when equal lumen flowrates are used.

(2) For several shell/lumen residence time ratio values, use the mathematical model to find the lumen residence time and, thus, the shell residence time required to meet the treatment goal. Most likely, several residence-time ratios and lumen residence-time combinations will work. Two other decision variables should be considered at this point: the percent biodegradation in the shell and the biomass concentration. The biomass concentration should be increased above the values used in the experimental research, as this was demonstrated in the modeling section to have significant effects on process sizing. Biomass concentrations in the 2000 m\g/L range should be explored, but experimental verification is required to determine if this can be accomplished. If significant degradation is desired within the HFMM, the percent degradation can be specified to meet at least a certain minimum value, such as 10%. Setting the biomass concentration and percent biodegradation will limit the number of feasible residence-time combinations. Residence-time ratios that achieve a good balance between mass transfer and biodegradation seem to fall into a range between 1.5 and 3. This analysis will yield the flowrate per treatment train, which can then be scaled to the entire project flowrate to determine the number of parallel treatment trains required.

(3) The polishing-reactor volume is calculated using the flow characteristics of the reactor, pseudo-first-order co-metabolism kinetics, and the shell flowrate and effluent concentration determined in step two. As noted above, one polishing reactor per treatment stage is recommended.

(4) The chemostat can then be sized from the specific transformation relationship (eqs 2 and 3). The entire project flowrate should be used to size one large chemostat for all treatment trains. An operating specific transformation and chemostat dilution rate should be chosen. Based on experimental results, specific transformation values of up to 40 μg TCE/mg TSS can be selected for TCE. Generally, specific transformations of up to approximately 50% of the transformation capacity of the chlorinated solvent should be acceptable, although further testing should be performed as confirmation.

(5) Several iterations of the above design steps may be required to determine all of the feasible designs. Cost modeling of the feasible designs would then be used to determine the least cost feasible design.

Figure 8-Effect of specific transformation and biomass concentration on chemostat volume for entire project flowrate (568 L/ min [150 gpm]).

The steps described above will result in complete system design. This design strategy should be considered as the basis for further HFM bioreactor system development, laboratory experimentation, and field-scale demonstration. In other words, the design strategy is a very basic strategy intended to give the HFM bioreactor designer some direction in sizing the unit processes.

Conclusions

Preliminary cost comparisons of various HFM bioreactor designs resulted in costs within a similar range as air stripping with carbon adsorption when flowrates were significantly larger than in previous research and HFMMs were configured in treatment trains consisting of two HFMMs in series. Successful high-flowrate biological experiments proved that the system could be operated in the required flow range, with lumen and shell residence times as short as 0.16 and 0.40 min, respectively. A new variable, specific transformation was defined for characterizing the co-metabolic transformation in continuous-flow systems and was as large as 38.5 g TCE/mg TSS. Furthermore, an inverse relationship between increasing specific transformation and decreasing chemostat volume was shown to control the sizing of the chemostat. Using mathematical modeling, HFM bioreactor system design was investigated, and the interdependency of lumen residence time, residence-time ratio, mass transfer, biodgradation, and the adequate sizing of the polishing reactor and chemostat were confirmed. Finally, the development of a five-step system design strategy will facilitate sizing of the unit processes in future HFM bioreactor efforts.

References

Alvarez-Cohen, L.; McCarty, P. L. (199Ia) A Cometabolic Biotransformation Model for Halogenated Aliphatic Compounds Exhibiting Product Toxicity. Environ. Sd. Technol, 25 (8), 1381- 1387.

Alvarez-Cohen, L.; McCarty, P. L. (199Ib) Effects of Toxicity, Aeration, and Reductant Supply on Trichloroethylene Transformation by a Mixed Methanotrophic Culture. Appl. Environ. MicrobioL, 57 (1), 228-235.

Aziz, C. E.; Georgiou, G.; Speitcl, G. E., Jr. (1999) Cometabolism of Chlorinated Solvents and Binary Chlorinated Solvent Mixtures using M. trichosporium OB3b PP358. Biotechnol. Bioeng., 65 (1), 100-107.

Aziz, C. E. (1997) Cometabolic Degradation of Chlorinated Solvent Mixtures by M. trichosporium OB3b PP358. Doctoral Dissertation, Department of Civil Engineering, University of Texas at Austin.

Dvorak, B. I. (1994) Selection Among Treatment Options for Removing Synthetic Organics from Water. Doctoral Dissertation, Department of Civil Engineering, University of Texas at Austin.

Fitch, M. W. (1996) Trichloroethylene Degradation by Methylosinus Trichosporium OB3b in a Hollow Fiber Membrane Reactor. Doctoral Dissertation, Department of Chemical Engineering, University of Texas at Austin.

Fitch, M. W.; Speitel G. E., Jr.; Georgiou, G. (1996) Degradation of Trichloroethylene by Methanol-Grown Cultures of Methylosinus trichosporium OB3b PP358. Appl. Environ. MicrobioL, 62 (3), 1124- 1128.

Fox, B. G.; Borneman, J. G.; Wackett, L. P.; Lipscomb, J. D. (1990) Haloalkene Oxidation by the Soluble Methane Monooxygenase from Methylosinus trichosporium OB3b: Mechanistic and Environmental Implications. Biochem., 29, 6419-6427.

Oldenhuis, R.; Oedzes, J. Y.; Van der Waarde, J. J.; Janssen, D. B. (1991) Kinetics of Chlorinated Hydrocarbon Degradation by Methylosinus trichosporium OB3b and Toxicity of Trichloroethylene. Appl. Environ. MicrobioL, 57, 7-14.

Phelps, P. A.; Agarwal, S. K.; Speitel, G. E., Jr.; Georgiou, G. (1992) Methylosinus trichosporium OB3b Mutants Having Constitutive Expression of Soluble Methane Monooxygenase in the Presence of High Levels of Copper. Appl. Environ. MicrobioL, 58, 3701-3708.

Pressman, J. G. (2001) Development of a Hollow Fiber Membrane Bioreactor for Cometabolic Degradation of Chlorinated Solvents. Doctoral Dissertation, Department of Civil Engineering, University of Texas at Austin.

Pressman, J. G.; Georgiou, G.; Speitel, G. E., Jr. (1999) Demonstration of Efficient Trichloroethylene Biodegradation in a Hollow-Fiber Membrane Bioreactor. Biotechnol. Bioeng., 62 (6), 681- 692.

Pressman, J. G.; Georgiou, G.; Speitel, G. E., Jr. (2000) A Hollow-Fiber Membrane Bioreactor for the Removal of Trichloroethylene from the Vapor Phase. Biotechnol. Bioeng., 68 (5), 548-556.

Acknowledgments

Credits. The authors gratefully acknowledge support for this research provided by The U.S. Environmental Protection Agency's Science to Achieve Results (STAR) Graduate Fellowship Program (Washington, D.C.) and Celgard LLC (Charlotte, North Carolina) for providing the hollow-fiber-membrane modules.

Authors. Jonathan G. Pressman is currently employed at U.S. Environmental Protection Agency, Cincinnati, Ohio; at the time of this work, he was a doctoral candidate at the University of Texas at Austin. George Georgiou is a professor in the Department of Chemical Engineering at the University of Texas at Austin. Gerald E. Speitel, Jr., is a professor in the Department of Civil Engineering at the University of Texas at Austin. Correspondence should be addressed to Jonathan G. Pressman, U.S. Environmental Protection Agency, National Risk Management Research Laboratory, 26 West Martin Luther King Drive (ms681), Cincinnati, OH 45268; e-mail: pressman.jonathan@ epa.gov.

Submitted for publication October 1, 2001; revised manuscript submitted January 28,2004; accepted/or publication July 15, 2004.

The deadline to submit Discussions of this paper is January 15, 2006.

Copyright Water Environment Federation Sep/Oct 2005

Source: Water Environment Research

More News in this Category