Modeling of a Combined Ultraviolet-Biofilter System to Treat Gaseous Chlorobenzene I: Model Development and Parametric Sensitivity

ABSTRACT

A new type of a combined ultraviolet (UV)-biofilter system for air pollution control is developed. In this paper, two conceptual mathematical submodels of the UV reactor and standalone biofilter are developed. All model parameters have been determined by independent experiments or have been taken from literature. Results from UV and the standalone biofilter submodels are in a good agreement with experimental data. However, the performance of the combined system has significantly deviated from those of the UV or standalone submodels because of the stimulating effects of UV irradiation products on the subsequent biofilter performance. A modified model that considers the stimulating effects has agreed well with experimental data over a wide range of operating conditions. Further analysis of the primary parametric sensitivity of the model has shown that inlet chlorobenzene concentrations, gas empty-bed residence time in the UV reactor, and light intensity are important operating conditions.

IMPLICATIONS

This study focuses on the model development of a combined UV-biofilter system to treat gaseous chlorobenzene. The results of this study are useful in understanding the effects of UV pretreatment on the performance of the subsequent biofilter. It can also help predict chlorobenzene removal performance in combined UV-biofilter systems.

INTRODUCTION

Biofiltration is widely considered a cost-effective technology because of its low operating costs and convenient maintenance.1 However, biofilters can not achieve satisfactory performance in treating some recalcitrant volatile organic compounds (VOCs) given the low biodegradability of these compounds. Several efforts have been proposed to address these challenges.2–4 One of the efforts comes from combination of ultraviolet (UV) photodegradation and biofiltration technology.

Previous studies have mentioned that the combination of UV photodegradation and biofiltration technologies can be successful for treating some recalcitrant gaseous contaminants. Koh et al.5 and Wang et al.6 have shown that UV photodegradation transformed some recalcitrant VOCs into more biodegradable products. Moussavi et al.7 and Wang et al.8,9 produced biofilters with improved performance by incorporating ozone-producing UV lamps for pretreatment. Another study that explored the mechanisms of UV pretreatment on biofilters has shown that ozone produced by UV irradiation might be responsible for the performance enhancement of biofilters.10

Combined UV-biofilter systems are complex, and detailed mathematical modeling is necessary to quantitatively describe the stimulating effects of the UV pretreatment on the subsequent biofilter and in optimizing or scaling up the combined system. In this study, a conceptual model on gaseous chlorobenzene degradation in a combined UV-biofilter system is presented and validated with experimental results. Model parametric sensitivity studies are also discussed in this paper.

MODEL DEVELOPMENT

Submodel of the UV Reactor

In the gas phase, photons are absorbed to initiate photo-chemical reaction with chlorobenzene. Quantum yield (Φ) is defined as follows (eq 1)11:

(1)

The removal rate constant k (sec⁻¹) can be defined as (eq 2)

(2)

On the basis of the Beer-Lambert law, the kinetic model for the photochemical reaction can be deduced as11

(3)

The flow pattern inside of the UV reactor used in this study can be considered an ideal plug flow. Thus, the following equation is given:

(4)

From eqs. 3 and 4,

(5)

Submodel of the Biofilter

In the gas phase, chlorobenzene is transported from the bulk gas stream into the water surface. The differential mass balance for chlorobenzene in the gas phase is (eq 6)

(6)

where c _g is the chlorobenzene concentration in the gas phase, J _x is the chlorobenzene flux in the water phase, x is the height coordinate, and u is the superficial velocity inside of the biofilter.

At the gas-water interface, chlorobenzene concentrations in gas and liquid phases can be defined according to Henry's law equilibrium (eq 7),

(7)

Because no chlorobenzene biodegradation occurs in the water phase (i.e., chlorobenzene is degraded in the biofilm), the mass transport can be described according to Fick's first law,

(8)

where z is the depth to the biofilm surface.

The chlorobenzene flux and chlorobenzene concentration profile can be expressed by truly closed-form analytical solutions. The differential mass balance in the biofilm is

(9)

A boundary condition corresponds to no flux into the attachment surface (eq 10).

(10)

An analytical solution to eq 9 is available when concentrations at both boundaries of the biofilm are known. Combining eqs 8 and 9 gives

(11)

There are two simplifications for the submodel. The first is that the concentration at attachment surface approached zero (eq 12).

(12)

The liquid phase is quite thin, given the application of water.12 Therefore, in this study, it was reasonable to disregard water layer thickness. If such is the case, the concentrations in the gas phase and biofilm could be defined as (13)

(13)

Finally, eq 11 yields (eq 14)

(14)

Parameter Determination

Average light intensity (I _ave) inside of the reactor is calculated by the Line Source with Spherical Emission (LSSE) model (eq 15).13

(15)

Φ is calculated by using eq 1 at various inlet chlorobenzene concentrations. The experimental relationship between Φ and inlet concentrations (c _A0) can be achieved as shown in Figure 1. Results have shown that a higher Φ could be achieved at the lower c _A0. Therefore, eq 5 can be expressed as (eq 16)

Figure 1. Relationship simulations between Φ and c _A0.

(16)

Other parameters in UV and biofilter submodels are shown in Tables 1 and 2, respectively.

Table 1. A list of symbols and parameters and their values for UV submodel

Symbol	Parameter	Numerical Value	Methods and Source
b	Optical path	0.04 m	Measured in this study
c A0	UV inlet chlorobenzene concentration	100–4000 mg m^−3	GC analyzed in this study
c A1	UV outlet chlorobenzene concentration	100–3000 mg m^−3	GC analyzed in this study
E 1	Energy of one photon	7.8 10^−19 J	Reference 16
H	Reactor height	0.45 m	Measured in this study
H	Planck's constant	6.62 × 10^−34 m^2 hr^−1	Reference 16
I ave	Average light intensity	38 W m^−3	Equation 15
L	Lamp length	0.44 m	Measured in this study
N A	Number of chlorobenzene molecule
N	Avogadro constant	6.023 × 10^23 m^2 hr^−1	Reference 16
P i	Emission energy fraction at wavelength i	0.75	Lamp parameter
Q	Gas flow rate	0.1–0.5 m^3 hr^−1	Measured in this study
R	Reactor internal radius	0.04 m	Measured in this study
r 1	Lamp radius	0.001 m	Measured in this study
R	Radial coordinate
S L	Light source strength per unit length	52.3 W m^−1	Calculation in this study
V UV	UV reactor volume	2.2 × 10^−3 m^3	Measured in this study
X	Number of absorbed photons
Z	Rectangular coordinate
τUV	Gas empty-bed residence time in the UV reactor	0–81 sec	Calculation in this study
Φ	Quantum yield		Calculation in this study
μi	Absorbency per unit optical thickness	0.0894 m^−1	Reference 11
ϵA	Mole extinction coefficients	7.4 × 10^3 L mol^−1 cm^−1	Reference 17

Table 2. A list of symbols and parameters and their values for biofilter submodel

Symbol	Parameter	Numerical Value	Methods and Source
A BF	Cross-sectional area of biofilter	1.13 × 10^−2 m^2	Measured in this study
a	Specific area of the filter bed	860 m^2 m^−3 (UV-BF) 820 m^2 m^−3 (BF)	Calculation in this study
c A1	Biofilter inlet chlorobenzene concentration	100–3000 mg m^−3	Measured in this study
c A2	Biofilter outlet chlorobenzene concentration	100–3000 mg m^−3	Measured in this study
c S	Chlorobenzene concentrations at the biofilm/liquid interface
c f	Chlorobenzene concentrations in the biofilm
D	Diffusion coefficient of chlorobenzene in water	2.44 × 10^−6 m^2 hr^−1	Reference 18
D f	Diffusion coefficient of chlorobenzene in biofilm	1.97 × 10^−7 m^2 hr^−1	Reference 18
H′	Distribution coefficient of chlorobenzene	0.15	Reference 19
K s	Half-saturation constant	1.5 × 10^3 mg m^−3	Reference 20
L′	Thickness of the water layer over the biofilm
L f	Thickness of biofilm	0.15 mm (UV-BF) 0.2 mm (BF)	Measured in this study
L	Height of biofilter	0.3 m	Measured in this study
Q	Gas flow rate	0.1–0.5 m^3 hr^−1	Measured in this study
V BF	Biofilter volume	3.4 × 10^−3 m^3	Measured in this study
νm	Maximum substrate utilization rate	0.85 hr^−1	Reference 20
X f	Microorganism concentration	6.7 × 10^4 g m^−3 (UV-BF) 7.2 × 10^4 g m^−3 (BF)	Measured in this study
τBF	Gas empty-bed residence time in the biofilter	0–122 sec	Calculation in this study
∈o	Initial void fraction of the filter bed	0.59 (UV-BF) 0.53 (BF)	Calculation in this study
Notes: BF = biofilter.

MATERIALS AND METHODS

Experimental Setup and Operation

Combined UV-biofilter and control biofilter processes were established, as shown in Figure 2.12,14 The UV reactor was made of stainless steel with an effective volume of 2.3 L. An ozone-producing low-pressure mercury vapor lamp was inserted into the airtight reactor. Both biofilters were identical in configuration with heights of 45 cm and internal diameters of 12 cm. The packing was made of nonbiodegradable bamboos (a depth of 30 cm), which could be easily acquired.

Figure 2. Setup of the experiment system.

Experiments were conducted for 280 days under various operating conditions, including varying c _A0 and gas flow rates (Table 3). Biofilters were supplied with approximately 80 mL of nutrient solution every 1 hr. Leachate from the filter bed was discharged directly without recirculation. No pH control was integrated with the experimental system. Temperature was maintained relatively constant at 20–30 °C.

Table 3. Operating conditions of the UV-biofilter system

		Operating Conditions
Run Number	Running Days	UV Lamp (on/off)	τuv (sec)	τBF (sec)	Inlet Concentration (mg m^−3)
Run 1	0–30	Off	0	122, 41, 24	250 ± 150
Run 2	31–47	On	81	122	1000 ± 500
Run 3	48–59	On	40	61	1000 ± 500
Run 4	60–80	On	27	41	1000 ± 500
Run 5	81–95	On	20	30	1000 ± 500
Run 6	96–108	On	16	24	1000 ± 500
Run 7	109–120	On	27	41	250 ± 150
Run 8	121–146	On	27	41	800 ± 200
Run 9	147–180	On	27	41	1500 ± 500
Run 10	181–194	On	27	41	400 ± 150
Run 11	195–207	On	27	41	2500 ± 300
Run 12	208–228	On	40	122	800 ± 300
Run 13	229–238	On	27	122	800 ± 300
Run 14	239–280	On	20	122	800 ± 300

Analytical Methods

Concentrations of chlorobenzene were analyzed by a gas chromatograph (Shimadazu, GC-14B) with an flameionization detector (FID) detector and a capillary column (Ulbon HR-1, 0.25 mm × 30 m).

The ozone content in the gas was measured using the gas detection tube method. The 100-mL sample was injected into a tube containing detecting agents at a flow rate of 50 mL min⁻¹. The concentration of ozone was read on-site on the basis of the color change of the detecting agents (Beijing LaoBaoSuo Science and Technology Company).

RESULTS AND DISCUSSION

Comparison of UV Submodel Simulations and Experimental Results

Figure 3 shows the UV submodel simulation and UV reactor experimental data at different c _A0 and gas flow rates. The model prediction was in good agreement with experimental data. Differences beyond the experimental error bars were observed between the model and experimental reactor at lower flow rates. However, for a model without any fitted parameter, small deviations might arise.

Figure 3. Model simulations (lines) and experimental data (symbols) of chlorobenzene removal efficiencies for (a) c _A0 changes and (b) gas flow rate variations.

Figure 3 presents a comparison of the model with the experimental data obtained when the c _A0 in the UV reactor varied from 100 to 2800 mg m⁻³ at an I _ave of 38 W m⁻³ and a flow rate of 0.1 m³ hr⁻¹. The model appropriately predicted the experimental data, corresponding to a linear decrease in chlorobenzene removal efficiency with increasing c _A0. The model also accurately predicted removal efficiencies at various flow rates. Once again, the comparison of the model and the experimental data has revealed that the UV submodel describes the results well.

Comparison of Biofilter Submodel Simulations and Experimental Results

Figure 4 illustrates chlorobenzene removal efficiencies for a 280-day simulation of the standalone biofilter (no UV pretreatment). Simulation results for the removal efficiencies were consistent with experimental performance.

Figure 4. Model simulations (lines) and experimental data (symbols) of chlorobenzene removal efficiencies of the control biofilter (no UV treatment).

In the first 100 days of operation, the residence time inside of the standalone biofilter decreased with an increases in gas flow rate, resulting in a slow decline of removal efficiencies. The model simulation also clearly demonstrated the gradual decline in removal efficiency. On day 100 in the simulation, a removal efficiency of approximately 10% was observed. From days 180 to 194, remarkable climbing removal efficiency was observed, and this might have been caused by c _A0 reduction. These findings were consistent with predicted results. In other operating conditions, the biofilter submodel yielded relatively accurate predictions of the standalone biofilter performance.

Stimulating Effects of UV Pretreatment on Subsequent Biofilter Performance

A previous study has reported that some photodegradation products (e.g., chlorophenol, phenol, and low-molecular-weight acids) and the byproduct of ozone can be produced by UV irradiation.7 These products are soluble and biodegradable, resulting in low pH of the filter bed, which threatens the biofilter performance. In addition, ozone has been proven to reduce the extra-cellular polymer substance (EPS) content of biofilm in a biofilter.10 Reducing EPS content can influence the mass transfer in the biofilm, resulting in a different diffusion coefficient of chlorobenzene (D _f) compared with that in a control biofilter. Ozone and other photodegradation products also play key roles in the microbial metabolic activity of biofilms and changed the structure of the microbial communities.10 These differences can alter some model parameters, such as the maximum substrate utilization rate and half-saturation constant.

Some parameters in the biofilter submodel of the combined UV-biofilter system were different from those of the standalone biofilter. Accordingly, significant deviations were observed between the combined UV-biofilter system experiments and predictions based on the two submodels (Figure 5).

Figure 5. Relationship between ozone concentration and ψ.

To accurately simulate the performance of the combined UV-biofilter system, a new parameter (i.e., stimulating factor [ψ]) was defined to determine the stimulating effects of the ozone on subsequent biofilter performance.15

(17)

where η_exp is the experimental removal efficiency measured by the inlet and outlet concentrations and is η_the is the prediction data based on the two submodels.

Some factors (e.g., gas flow rate or light intensity) can affect ozone concentrations at the UV outlet. A series of independent experiments was conducted to calculate the average ψ at according to eq 17. The relationship between ψ and the measured ozone concentrations at the outlet of the UV reactor in these experiments was also analyzed (Figure 5).

As shown in Figure 5, an empirical formula similar to the Monod model was acquired.

(18)

According to Figure 5 and eq 18, higher ozone concentrations would result in higher ψ. Further increases in ozone concentration led to a plateau, an indication of the stimulating limitation of the ozone.

Thus, the final predicted results of the combined system could be obtained as (eq 19)

(19)

Model Simulation of Steady-State Performance of the Combined UV-Biofilter System

Modified model simulations (including ψ) were compared with combined model simulations (excluding ψ) and experimental results. Some parameters in the modified model and combined models were determined independently (Table 2).

Figure 6 shows the model simulations of outlet concentration, removal efficiencies, and removal rates and experimental data at different operating conditions. The modified model predictions were in better agreement with experimental data compared with the combined model predictions, showing the stimulating effects of UV pretreatment on subsequent biofilter performance. One explanation for the discrepancy between the two models could be that the ozone produced by UV irradiation has stimulated the performance of the biofilter.10

Figure 6. Modified model simulations (solid lines), combined model simulation (dashed line), and experimental data (symbols) of the combined UV-biofilter system performance at various operating conditions: (a) outlet concentration, (b) removal efficiency, and (c) removal rate.

Parametric Sensitivity

The sensitivity of the primary parameter in the combined UV-biofilter model used to control gaseous chlorobenzene was examined. To show the parametric sensitivity, the relative sensitive index (RSI) was defined as (eq 20)

(20)

where P ₀ and P are the initial and final values of one parameter, respectively, and y _i0 and y _i are the initial and final function values, respectively.

In this study, the sensitivity of chlorobenzene removal to the model parameters was examined using the following conditions: c _A0 of 1000 mg m⁻³; empty-bed residence time in the UV reactor and biofilter of 27 sec (τ_UV) and 41 sec (τ_BF), respectively; and an I _ave of 38 W m⁻³. The sensitivity of chlorobenzene removal efficiency of the above-mentioned operating parameters was studied by varying the values of operating parameters (P ₀) to 1.2 times (P = 1.2P ₀) of the initial values.

The final sensitivity results of c _A0, τ_UV, τ_BF, and I _ave were 0.34, −0.12, −0.03, and −0.15, respectively. Results showed that c _A0, τ_UV, and I _ave played important roles in the overall performance of the combined system, whereas τ_BF did not.

CONCLUSIONS

A conceptual mathematical model was developed for a combined UV-biofilter system to treat gaseous chlorobenzene. All model parameters were taken from literature or determined by independent experiments; thus, the model was fully predictive. Standalone UV and biofilter submodels were in a good agreement with experimental data. However, the performance of the combined model showed significant deviations from either of the two sub-models. The stimulating effects of ozone were considered, and the modified model was in good agreement with experimental data over a wide range of operating conditions. Further analysis of the primary parametric sensitivity of the model showed that c _A0, τ_UV, and I _ave were important operating conditions.

ACKNOWLEDGMENTS

This study was supported by the National Natural Science Foundation of China (Grant No. 50708049).