th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011
Modeling of Biofiltration process for VC Removal from an Air Stream and Optimization of Model Parameters by Genetic Algorithm S.H. Esmaeili Faraj, Y. Amini, M. Nasr Esfahany Corresponding Author’s Address: Department of Chemical Engineering, Isfahan University of Technology, Isfahan, 8415683111, Iran.
Corresponding Author’s E-mail:
[email protected]
Abstract In this article, Modeling of Biofiltration process for VC Removal is presented. The modified Ottengraf mathematical modeling was used for modeling process. With determining process kinetic parameters and some of the special model parameters, experimental results fitted by model equation. equation For optimization this parameters used genetic algor algorithm. ithm. The final parameters by curve fitting model parameters for experimental results compare with the genetic algorithm method. Both of the results indicated thatt the modified Ottengraf model could conform to a final experimental data of Biotrickling filter very well. Key words: Biofiltration, Vinyl Chloride, Modelling,, Sensitivity Analysis, Genetic Algorithm
1- Introduction Biofiltration is an out-coming coming technique based on the capability of some microorganisms to degrade organic matter for their their own metabolism. In such bioreactors, waste gas flows throughout a packing material covered by an active biomass. At the beginning, biofilters were settled to remove unpleasant and noxious odours from composting and wastewater treatment plants, but their effectiveness has been further demonstrated in VOCs emission control from many industrial applications. Many types of equipment have been designed for biological VOCs oxidation and they essentially differ in the system used to supply water to the biomass. Biological Biological oxidation can reach 98% removal efficiency for very biodegradable compounds and it successfully operates with high flow rates and low pollutant concentrations [1]. Microorganisms consume nutrients for the synthesis of lipids, proteins and polysaccharides, which are found in the cellular matter. Biotrickling filters consist consist of a bed of rough media such as crushed trap rock, granite, activated carbon, limestone, clinkers, wood slats, plastic tubes,
Modeling of Biofiltration process for VC ……
corrugated plastic sections, hard coal, or other material over which gas and water are distributed and contacted. Contaminated ted air flows over the contact media on which a biofilm develops. Pollutants in the waste gas are transported into the biofilm, where biodegradation takes place. In many cases, biotrickling filters are more efficient than conventional biofilters for treating treati contaminated air streams, and in spite of their higher operational and investment costs, they are often preferred [2-4]. The different biofiltration designs basically differ because of the system to provide water inside the bioreactor. Three main equipments can be identified. Conventional biofilters (BFs) consist in a packing material covered with biomass with no continuous mobile liquid phase; inlet gas stream is humidified in an pre pre-humidification humidification unit before reaching the reactor. In Biotrickling filters (BTFs), a water stream trickles throughout throughout the packing, keeping biomass wet. Bioscrubbers (BSs) are constituted by an absorption unit and a reaction unit, where biomass is suspended into the liquid phase [5]. Biotrickling filters (BTFs) are characterized by a continuous aqueous phase trickling tric throughout the reactor bed. Even if working principles are the same compared with conventional biofilters, the presence of a trickling liquid imposes some different design conditions. Trickling liquid increases the risk of bed compaction. For this reason, reason, the packing is normally constituted by inert or synthetic material and the control of the pH, nutrients, presence of toxics is allowed by the analysis of the trickling solution which is normally recirculated. Trickling liquid is also a good mean to remove emove toxic or acidifying by-products by from inside the bed [5]. Since removal VC was presented with using of Biofiltration process for the first time, representing a model could predict an acceptable result of process might be very useful as far as possible.. Also, in this article Modeling of Biofiltration process for VC Removal is presented. Therefore, optimization present parameters used to the genetic algorithm too. 2- Method of Experiments As shown in figure 1, the biotrickling filter employed in this study udy configured gas and liquid streams as downflow. The biotrickling filter was made from transparent Plexiglas with an inner diameter of 100 mm. The column had a height of 0.5 m and contained 0.3 m material bed. Sampling port was placed at the column ends and middle of the bed for sampling the gas and packing media, respectively. Circulating liquid provided moisture and required nutrients, and regulated pH in the bed. Parallel stream of circulating liquid were fed to the beds by a centrifugal pump. Air flow was provided by a fan while VC was provided from filled tanks and mixed with air stream in a T junction. The mixed gas flow was regulated by a rotameter. Feed of the biotrickling filter was entered the top of the column and ultimately was exited at the bo bottom of the column (Figure 1).
th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011
Fig 1. Schematic manufacture biofilter for VC Removal from an Air Stream
The performance of the biotrickling filter was evaluated by the following performance parameters: a) VC inlet load (L) is the mass of VC fed to the biofilter biofilter per unit time per unit volume of the bed, (g.m−3.h−1): (1) L b) Removal Efficiency (RE) is defined as the fraction of pollutant (VC) degraded by the biotrickling filter, (%): ,
RE
,
,
,
100
(2)
c) Elimination Capacity (EC) is the amount of pollutant (VC) degraded degraded per unit time per unit −3 −1 volume of bed material, (g.m .h ):
EC (3) d) Empty Bed Residence Time (EBRT) is the time required by the gas to cross the empty column, (hr): EBRT (4) ,
,
Where in these equations, QV is volumetric gas flow rate, (m3.h−1), V is the volume of the packed bed, (m3) and C is the VC concentration in the gas stream, (g.m−3). Subscripts “in” and “out” are referred to the inlet and outlet of the packed bed. EC is always less than L and equal to L for an ideal case when 100% removal efficiency is achieved.
3- Modeling Ottengraf’s model considers the different phenomena ruling biofilter performance: mass transfer and biological reaction. A schematic of the model is represented in figure (2). At low inlet concentrations, the driving force ruling the mass transfer is limited. Therefore, the amount of pollutant which passes into the liquid phase is moderate and, as pollutant comes in contact with the biomass, it is completely degraded. In these con conditions, ditions, diffusion is the rate determining step. With higher gas concentrations, mass transfer is conversely promoted. The amount of pollutant transferred in the aqueous phase is greater and biomass could not be able to completely degrade this amount. In such uch conditions, the reaction limits the process rate. Ottengraf proposed some equations to represent what occurs in the water film in these two opposite situations.
Modeling of Biofiltration process for VC ……
Hypothesis 1. Stationary state is supposed 2. Biological kinetic is zero-order zero in respect of the substrate. Oxygen is always in excess and does not affect kinetics. Biofilm thickness is negligible in respect of the diameter of the carrier particles and its value is constant along the biofilter. 3. Two phases are considered: gas phase and (water/biofilm) phase. Pollutant diffusion in (water/biofilm) phase follows Ficks’s law. 4. Fluid interface equilibrium can be represented by Henry’s law. 5. Gas stream is a plug flow, with no axial dispersion. 6. One pollutant only is considered.
Figure 2: Biophysical biofilm model for Ottengraf. Profile 1 is related to the reaction limitation area, while profile 2 to the diffusion limitation area [3].
Pollutant concentration in the gas phase can be expressed by the following expression: dC U NA dh
(5)
Where Ug is the superficial gas velocity [m/h], h is the reactor height [m], N is the flux of substrate from the gas to the liquid [g/(m2 .h)] and As is the specific surface area [m2 /m3 ]. Mass balance in the the water/biofilm can be writ written as follows: ! d C" (6) D ! k& 0 dx Where D is the diffusion coefficient [m2 /h], x is the direction perpendicular to the gas-liquid gas interface and k0 the zero-order order constant [g/(m3 .h)]. Ottengraf’s model individuates two different phenomena, ruling and determining the rate of the biofiltration process. At low load values, diffusion is the rate determining step and, in such conditions, the elimination capacity is given by the following equation:
th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011
EC'"
L (1
)1
!
k&D V : = > A : 2m QL
(7)
where the index dl stands for diffusion di limiting. Otherwise, at high loads, the removal of the pollutant is mainly influenced by the biological reaction and the elimination capacity is load load-independent: (8) EC*" EC+,- A k & δ But, having the use of one equation only that can continuously continuously connect the different expression of ECdl and ECrl can be very useful for biofiltration design. The following equation can satisfy this condition: EC C'" EC+,EC*" EC+,- . (9) L B 1 . ? ∗A L
transition between reaction and diffusion limitation occurs: For where L* is the load at which the transition * LL*, all the second term on the right side becomes zero and therefore EC EC*" . Parameter p is calculated by fitting of the experimental data. Its value specifies the rate at which the passage between the two different limiting conditions occurs. With some arithmetical steps and using the definition of L and EC, it it is also possible to write efficiency and Cg,out as a function of Cg,in ! C ,67 Q V k D C1 D1 A E & F G V Q 2mC67 2 L A k δ . B & 1 K C ,67 (10) 1.H ∗ I C J RE 0 C ,67 Q V ! V k&D F G A k& δ C67 C1 D1 A E Q 2mC67 2QA k & δ L (11) C345 C67 1 . K B 1 V K C67 1.? ∗A C 0 J where C* is the inlet concentration at which load is equal to the L*, at constant flow rate and volume. 3-2- Sensitivity Analysis Sensitivity analysis is very useful to evaluate how parameters parameters affect the outputs of a mathematical model. It can be also used to evaluate the quality of a model and to verify if there is a good agreement between the physics of the process and the model itself. The main parameters to be considered are: As , EBRT ,δ , k0 , L* and P
Modeling of Biofiltration process for VC ……
3-3- Genetic Algorithm Because of modeling parameters fitting used to the genetic algorithm. The objective function used for the optimization that equal with a mean square error (MSE) that follow as: Z
MNOPQPRS TUV U WXY [\]
Vdef g
!
(12)
In order to use genetic algorithm for solving the given mathematical models at first, the chromosome of the problem should be determined. determined The chromosome for the problem includes five continuous genes (Di) which are formed randomly random between zero to one, that are as the followings: (13) chromosome cD] , D! , Dh , Di , Dj k The obtained genes are normalized ones. Therefore in order to measure the objective function by using maximum and minimum of parameters, the values of the parameters are obtained. Here, in order to choose the pairs, random weighted selection method is applied, which is based on the cost of the chromosomes. For recombination of the chromosomes, extrapolation method with two cutting points is used in which two random chromosomes from the parents are selected and then combined together with a random weight. In Fig. Fig.٣ the flowchart of using genetic algorithm to solve the models is given.
Fig.3: Flowchart of Genetic Algorithm
th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011
4- Discussion ssion and conclusion 4-1- Conclusion of modified Ottengraf model In the table 4 available value parameters for modeling, represented five unknown parameter initial and final values too. Table 1: system and model value parameters evaluate of curve fitting on the experimental data
parameter
symbol
Fixed parameters Gas flow rate Bed volume Diffusivity in water Air/water partition coefficient Fitted parameters Specified surface area Biofilm thickness Zero order kinetic constant Critical mass loading rate Exponent
Q V D m
As δ k0 L* p
Values Initial set
unit
Final set
m3/h m3 m2/h g/g
0.444 0.0023 3.32×10-6 0.54
1600 3.747×10-4 1 2.148 1
m2/m3 m g/m3.h g/m3.h -
1605 8.468×10-9 1 2.153 1.001
0.022
Outlet VC concentration, (g/m3)
0.021
0.02
0.019
0.018
0.017
0.016 0.02
Modified ottengraf model Experimental data
0.021
0.022
0.023 0.024 Inlet VC concentration, (g/m3)
0.025
0.026
0.027
Fig 4: The modified Ottengraf model fitting figure on the experimental data of Biofiltration VC
4-2- Conclusion of sensitive analysis For the present system behavior and the determine amount of model sensitive to the change of value sensitive doing analysis parameters. In the table 2 was showed the list of using parameters and using values for sensitive analysis. In the t fig 5 until 7 presented the data conclusion of the sensitive analysis.
Modeling of Biofiltration process for VC …… Table 2: The values of model parameters that used for sensitive analysis
parameter
As EBRT k0 δ L* p
-10%
1280 0.0043 0.8 6.77 1.722 0.8008
1440 0.0048 0.9 7.62 1.938 0.9009
unit
0.023
0.023
0.022
0.022
0.021
0.021
0.018
0.017
0.017
0.016
0.016
0.022
1760 0.0058 1.1 9.31 2.368 1.1011
1920 0.0064 1.2 10.16 2.584 1.2012
m2/m3 h g/m3.h ×10-9 m g/m3.h -
0.019
0.018
0.021
+20%
0.02
0.019
0.015 0.02
+10%
Biomass thickness increasing
k 0 increasing Outlet VC,(g/m3)
0.02 Outlet VC,(g/m3)
-20%
Values Main value 1600 0.0053 1 8.468 2.153 1.001
0.023 0.024 Inlet VC,(g/m 3)
0.025
0.026
0.015 0.02
0.027
0.021
0.022
0.023
0.024
0.025
0.026
0.027
0.026
0.027
Inle t VC,(g/m3)
(b)
(a)
Fig 5: The effect a)δ and b)k0 on the output VC concentration centration value of biofilter 0.022
0.022
0.021
0.021
L* increasing
p decreasing 0.02 Outlet VC, (g/m3)
Outlet VC, (g/m3)
0.02
0.019
0.019
0.018
0.018
0.017
0.017
0.016 0.02
0.021
0.022
0.023
0.024
Inle t VC, (g/m3)
(b)
0.025
0.026
0.027
0.016 0.02
0.021
0.022
0.023
0.024
0.025
Inlet VC, (g/m3)
(a)
Fig 6: The effect a) p and b) L* on the output VC concentration value of biofilter
th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011 0.023
0.022
0.022 0.021 0.021 0.02 Outlet VC, (g/m3)
Outlet VC, (g/m3)
0.02 As increasing 0.019
0.019
0.018 0.018 0.017 EBRT=0.0064 EBRT=0.0058 EBRT=0.0053 EBRT=0.0048 EBRT=0.0043
0.017 0.016
0.015 0.02
0.021
0.022
0.023 0.024 Inlet VC, (g/m3)
0.025
0.026
0.027
0.016 0.02
(b)
0.021
0.022
0.023 0.024 Inle t VC, (g/m3)
0.025
0.026
0.027
(a)
Fig 7: The effect a) EBRT and b) As on the output VC concentration value of biofilter
4-3- Conclusion of optimization tion by genetic algorithm By using of data conclusion as sensitive analysis and conclusion of curve fitting, to choosing constraints for model parameters: 1500 m A m 2500 0 m k & m 10 (14) 10 ]& m δ m 10 j 0mpm3 0 m L m 10 The conclusion of genetic algorithm was present as follow: As=1669 k0=0.78 δ=5.3×10 ×10-9 p=1.8 L=1.7
(15)
In Fig. 7 cost function versus generation is presented. As shown, cost functions get to their the smallest value after some recombination.
Modeling of Biofiltration process for VC ……
0.8
Best Fitness Function Mean Fitness Function
3
Fitness Function(g/m )
0.7 0.6
6 x 10
-3
0.5 4
0.4 2
0.3 0 4
5
6
7
0.2 0.1 0 0
5
10
15
20
25
Generation Fig. 7: Cost function versus Generation
As we have five parameters to be optimized, the initial population is increased from 500 to 5000, to ensure the search in all domains. Fig.8 shows the best cost cost versus initial population. It is clear the number of 3500 for the initial population is optimum. For lower values, it does not search the entire domain of optimization that is due to the low number of the population. For greater initial population, extraa recombination in chromosomes causes an increase in the cost. In order to get the best mutation coefficient, for the initial population of 3500 (best initial population), mutation is changed from 0.001 to 0.05.. Fig. 9 shows the best cost versus mutation ccoefficient. As it can be seen, 0.02 is the optimal value for mutation coefficient x 10
-6
1.58
3
Fitness Function(g/m )
1.6
1.56 1.54 1.52 1.5 1.48 500
1000
1500
2000
2500
3000
3500
Initial Population Fig.8: Best cost versus initial population
4000
4500
5000
th
The 7 International Chemical Engineering Congress & Exhibition (IChEC 2011)
Kish, Iran, 21-24 November, 2011
1.62
x 10
-6
3
Fitness Function(g/m )
1.6 1.58 1.56 1.54 1.52 1.5 1.48 0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
Mutation Fig.5: Best cost versus mutation factor
In the present method, the value of conclusion by genetic algorithm was optimal and having the smallest error percent modeling, Therefore; by using of this optimal values could be experiments the output CV of Biotrickling filters versus input concentration. The result of this action was presented in the figure 10. 0.022
Outlet VC concentration, (g/m3)
0.021
0.02
0.019
0.018
0.017
0.016 0.02
Model optimized by GA Exprimental data
0.021
0.022
0.023 0.024 Inl et VC concentrati on, (g/m3)
0.025
0.026
0.027
Fig 10: The modified Ottengraf model figure that optimization parameters by genetic algorithm
Modeling of Biofiltration process for VC ……
5- Conclusion In this research by use of a Biotrickling filter for biodegradation of VC. VC. The modify Ottengraf model with a zero order kinetic equation was fitted on experimen experimental data. Sensitivity analysis was applied for measure the effect of parameters. Then model parameters was optimized by genetic algorithm and the modified Ottengraf model drawn by using optimal parameters. At last in this research, 2 results was gain: 1- The modify Ottengraf model conform to experimental conclusion very easy 2- When the parameters of this model was optimized, can predict the behavior of system very well. References 1- Wang, L.K., Pereira, N.C., and Hung, Y.T., “Biological Treatment Processes” , Humana Press, 2009. 2- Iranpour, R., Cox, H.H.J., Deshusses, M.A., & Schroeder, E.D., “ Literature review of air pollution control biofilters and biotrickling filters for odor and VOC removal”, environmental progress, 24(3), 254-267, 267, 2005. 3- Shareefdeen, Z., and Singh, A., “Biotechnology for Odor and Air Pollution Control” , SpringerSpringer Verlag Berlin Heidelberg, 2005. 4- Vedova, L.D., “Biofilteration of Industrial Waste Gases in Trickling Trickling-Bed Bed Bioreactors” , Ph. D. Dissertation, Universit`a degli Studi di Padova, Italy, It 2008. 5- Lim, K.H., Lee, E.J., “Biofilter Modeling for Waste Air Treatment: Comparisns of Inherent Charecteristics of biofilter models” , Korean J. Chem. Eng., Vol. 20, No. 2, 315-327, 315 2003. 6- Liang, Y., and et al., “Longterm results of ammonia removal and an transformation by biofilteration”,journal of hazardous materials, 2000, B80, 259-269. 259 7- Cox, H.H.J., & Deshusses, M.A., “Effect of Starvation on the Performance and Re Re-acclimation of Biotrickling Filters for Air Pollution Control” , Environ. Sci. Technol., Vol. 36, 3069-3073, 3069 2002. 8- Seignez, C., Atti, A., Adler, N., and Peringer, P., “Effect of Biotrickling filter Opereting Parameters on Chlorobenzenes Degradation” , Journal of Environmental Engineering, Vol. 128, No. 4, 360-366, 2002. .J., “Biofilter Modeling for Waste Air Treatment: Comparisons of Inherent 9- Lim, K.H., and Lee, E.J., Charecteristics of biofilter models” , Korean J. Chem. Eng., Vol. 20, No. 2, 315-327, 315 2003. 10- Ardjmand. M., Safekordi, A., and Farjadfar, S., “Simulation of biofilter used for removal of air contaminants (ethanil)”, Int. J. Environ. Sci. Tech., Vol. 2, No. 1, 69 69-82, 82, 2005.
