Development, experimental validation and dynamic analysis of a general transient biofilter model

Pergamon

Development, experimental validation and dynamic analysis of a general transient biofilter model S. M. Zarook,” Department

of Chemical

A. A. Shaikh

and Z. Ansar

Engineering, King Fahd University Dhahran-31261, Saudi Arabia

(Received

15 April 1996; accepted

of Petroleum

2 September

and Minerals.

1996)

Abstract-In this study, a general transient biofiltration model. which incorporates general mixing phenomena, oxygen limitation effects, adsorption phenomena and general biodegradation reaction kinetics, is developed. Solutions are presented with and without the assumption of pseudo-steady state for the biofilm leading to approximate and general models, respectively. Solutions of the model are presented and validated with experimental transient data of benzene and toluene. Significant improvement in the model prediction is observed in comparison to earlier simplified models. However, the general model predictions seem to be better and it is superior to the approximate model as it does not require any correlations for film thickness or effectiveness factors. Dynamic analysis of the model is performed and compared with experimental data from the literature. Transient behavior during shut-down and restart-up are also well predicted by the model and the transient period does not seem to be long. Model predictions show that the biofilter is able to withstand extreme practical conditions such as random variations in the inlet concentration and gas flow rate. Theoretical analysis shows that the assumption of excess oxygen availability is not a good one, specially at high inlet concentration levels. Sensitivity studies show that mixing in the gas phase is an important phenomenon which should not be neglected and that some parameters need to be accurately estimated. (; 1997 Elsevier Science Ltd. All rights reserved Krywords:

Transient

biofilter

model; dynamic

analysis: biofiltration

of benzene

and toluene

Frisch, 1993; Zarook, 1994) have established biofiltration as an efficient treatment process, However, in order to understand the process better and subsequently develop it into an optimized technology, effective modeling of the process is required. Process modeling is fundamentally important as realistic models can lead to development of reliable design equations, as well as reduce the time and cost of experimentation at the pilot scale. However, until recently. efforts to model this process have been limited to steady-state biofiltration. The earliest and most widely used steady-state biofiltration model was developed by Ottengraf and van den Oever in 1983. This model assumes plug-flow behavior for the gas phase and availability of excess oxygen. Although the authors acknowledge that based on shake-flask experiments the biodegradation kinetics of single VOCs follow the Monod model, they only consider two limiting cases: zero- and first-order kinetics. However, recent studies (Zarook er trl.. 1993; Deshusses and Hamer. 1993: Chang et trl., 1993; Oh et al., 1994; Zarook and Baltzis. kinetics are more *Corresponding author. Tel.: (966)-03-860-4754; fax:(966)- 1994b) have shown that degradation complex than zero- or first-order-type kinetics. 03-860-4234; e-mail: [email protected]. INTRODUCTIOh

The volatile organic compounds (VOCs) emitted from chemical and process industries if left untreated can pose potential health risks in addition to causing severe environmental problems. A novel technology for treatment of off gases is biofiltration. The process is carried out in a biofilter which is basically a packed column containing micro-organisms immobilized on the packing material. These micro-organisms form active biolayers on the surface of the packing material. When the waste gas is forced to rise through the column, the pollutants are transferred to the biolayer where they undergo biological degradation by the immobilized micro-organisms. Proper selection of the microbial culture and of the biofilter size result in a pollutant-free airstream exiting the reactor. A number of experimental studies (e.g. Ottengraf and van den Oever, 1983; Zarook et a/., 1993; Leson et al., 1993; Deshusses and Hamer, 1993; Tonga and

759

760

S.

M. Zarook

In practical applications, and especially for cases of emissions, transient operation is more common due to the fact that the VOC emission level in any plant is unlikely to be constant. Furthermore, biofiltration can be applied to batch processes and thus, even if the emission level is constant, biofilters may be operating in an intermittent mode (e.g. in painting booth facilities). Hence, questions such as to how well can a biofilter respond to variations in volumetric flow rate, concentration, and composition are of paramount importance for commercial application of this technology. The fact that biofilters are most likely to operate under varying load conditions was recognized early by Ottengraf et al. (1983). There are few published studies on the transient performance and response of biofilters. Togna and Frish (1993), in a demonstration study for styrene vapor removal, observed that intermittent operation caused no problem in the performance of biofilters; biofilters inactive over the weekend period achieved normal removal rates within a few hours. Zilli et al. (1993) observed that their biofilter could be inactive for a period of 10 days, and subsequently be restored to pre-interruption performance levels within a day. To describe the transient aspects of biofiltration such as sudden changes in inlet concentration or flowrate, shut-down and restart-up conditions, etc., a theoretical model is essential. Zarook and Baltzis (1994b) have developed a transient model for the biofiltration of a single VOC taking into consideration the effects of adsorption, limiting effects of oxygen and interactive kinetics. Their model, however, is limited to the case of plug flow for the gas phase. The model equations have been further simplified through the use of effectiveness factors. The correlations for effectiveness factors and biofilm thickness were based on the assumption of pseudo-steady state within the biofilm. While all the steady-state and transient biofiltration models (e.g. Ottengraf et al., 1983; Zarook et al., 1993; Zarook and Baltzis, 1994b) are based on the assumptions that substrates are transported into the biofilm through diffusion, the recent transient model of Hodge and Devinny (1995) treats the biofilm and solid as a single phase, thus diffusion in the biofilm and details of the adsorption process are ignored. Furthermore, this model is also based on the assumption of first-order kinetics for biodegradation rate and excess oxygen. In a very recent study, Deshusses et al. (1995a) presented a transient biofiltration model which is also based on the assumptions of excess oxygen (no limitation by oxygen) and plug flow for the gas phase. These assumptions were justified experimentally for the case of methyl ethyl ketone (MEK) and methyl isobutyl ketone (MIBK) mixtures (Deshusses, 1994; Deshusses et al., 1995b, 1996). However, several studies (Zarook and Baltzis, 1993; Zarook, 1994; Zarook and Baltzis, 1994a) show that biofiltration of both hydrophilic and hydrophobic compounds are affected by availability of oxygen. Thus, a general model should incorporate limitations by oxygen. A concept of sorption volume is introduced, in their work, to account for adsorption

et al.

phenomena which are a vital part of transient biofiltration process. The sorption volume is defined as the difference between total volume of water in the bed and biofilm volume. Thus, adsorption into the bare surface of the solid is not considered. Moreover, a constant film thickness of 100 pm was used throughout the biofilter, however, it is now well established (Zarook et al., 1993; Zarook and Baltzis, 1994b) that the active film thickness varies along the biofilter column due to changes in the VOC concentration levels. Despite these drawbacks, the work of Desbusses et al. (1995a) is one of the few detailed works on transient biofiltration. The objective of the present study is to develop, perform dynamic analysis, and experimentally evaluate a general transient biofiltration model which incorporates general mixing phenomena (axial dispersion effects), oxygen limitations, general kinetics and adsorption phenomena. The model equations, in this study, have been solved by two methods. In the first method, the assumption of pseudo-steady state in the biofilm is made to reduce the mathematical complexity of the model as was also done by Zarook and Baltzis (1994b) in an earlier work. In the other method, this assumption is relaxed and model equations have been solved without any simplifications. The model is validated with the experimental transient biofiltration data of benzene and toluene (Zarook, 1994; Zarook and Baltzis, 1994b; Tang et al., 1995). Both substances are classified as primary pollutants by the US EPA. They are present in gasolines (Potter, 1992), and they are also widely used as industrial solvents, as well as feedstocks for synthesis (Reisch, 1992). Comparison between the models, dynamic responses to shock loadings, oxygen limitations studies, and parameter sensitivity analysis are also presented. MODEL

DEVELOPMENT

When deriving a theoretical model describing the behavior of a biofilter bed, one starts from a concept that is very customary in cases where the packing material serves as a support for micro-organisms. For a biological filter bed, it is supposed that the constituent particles are surrounded by a wet biolayer, as schematically illustrated in Fig. 1. When air flows around the particles there is continuous mass transfer between the gas phase and the biolayer. Volatile pollutants present in the waste gas, as well as oxygen, are partially dissolved in the liquid phase of the biolayer and are degraded or consumed by aerobic microbial activity. In this way a concentration gradient is established in the biolayer, which maintains a continuous flow of the component from the gas to the wet biolayer. Due to the small particle diameter (of the order of magnitude of a few millimeters), usually typical in biological filter beds, and a generally low solubility of the compounds to be transfered, the mass transfer resistance in the gas phase can be neglected. Zarook et al. (1993) and Baltzis and Zarook (1993) have introduced detailed models describing biofiltration of single and mixed pollutants under steady-state

General

transient

(a)

biofilter

761

model (b)

I Contaminated

/

Cc) BIOFILM

ADSORPTION

SURFACE

Gas Phase

Fig. I. Schematic representation for the model concepts used in the development of the model: (a) segment of the biofilter; (b) a solid support particle which is partially covered with biofilm; (c) oxygen and VOC transported into the biofilm where diffusion and reaction take place.

conditions. A general transient biofiltration model should incorporate dispersion, diffusion, reaction and adsorption phenomena. Since the biofilter contains a certain moisture content, the term adsorption is used loosely here and it refers to both actual adsorption on the solid as well as absorption of VOCs in the water retained in the packing materials. At steadystate conditions, the adsorption process is in equilibrium and thus, it does not come into play. However, under transient conditions, adsorption process needs to be explicitly accounted for in the model development. Furthermore, most of the past biofiltration models assume plug flow for the gas phase, however, dispersion in a biofilter is more likely since a biofilter is essentially a packed-bed reactor. During start-up of a biofilter unit, the biofilm is in formation, thus the phenomena associated with process start-up cannot easily be modeled. Hence, the transient model de-

veloped in this study is applicable from one set of operating conditions to another, after the biofilm is fully developed. The theoretical model describing the elimination of VOCs in the biofilter bed is based on the following assumptions. Assumptions (1) The biolayer is formed on the exterior surface of the particles. Biomass does not grow in the pores of the particles, and thus no reaction occurs in the pores. Sizes of these pores are too small for microbial growth (Deshusses et al., 1995a). (2) The biolayer is not necessarily formed uniformly around particles. In actuality, there may be patches of biofilm on the solids leaving the bare surface of the solids in direct contact with the airstream.

762

S. M. Zarook et al.

(3) Adsorption of the pollutant on the solid particles occurs only through the direct bare solid/air interface. Adsorption does not occur on the biofilm. There is no series adsorption of the substrate into the solid through the biofilm and no back diffusion of the substrate from the solid into the biofilm. This can be explained by the following points: (a) The concentration encountered in a biofilter is generally very low; (b) the adsorptive capacity of the packing material (e.g. peat) used is very low; (c) the fractional surface area covered by the biofilm is much less than the bare surface. (4) Oxygen is not adsorbed on the solid particles since concentration gradient of oxygen between the gas phase and the solid is almost negligible. (5) The thickness of the biolayer is small relative to the main curvature of the solid particles, and thus planar geometry can be used. (6) The extent of the biofilm patch is much larger than its depth. Hence, the VOC and oxygen transported into the biolayer through the side surfaces of the biofilm patch can be neglected, and diffusion/reaction in the biofilm can be considered in a single direction only. (7) Adsorption is a reversible process and its equilibrium characteristics can be described through the use of adsorption isotherms. (8) The VOC and oxygen at the biolayer/air interface are always in equilibrium as dictated by Henry’s law. The distribution coefficients are the same as if the biolayer was made of water only. (9) The VOC and/or oxygen are depleted in a fraction of the actual biolayer. This fraction is called the effective biolayer or active biofilm thickness (Zarook et al., 1993). (10) Diffusivities of the VOC and oxygen are equal to the diffusivities of the same compounds in water, ccrrected by a factor depending on the biofilm density, according to the expression of Fan et a/. (1990). (11) The biofilm density, defined as the amount of dry biomass per unit volume of biolayer, is constant. (12) There is no accumulation of biomass in the filter bed and thus, the specific biolayer surface area is constant. (13) The biodegradation rate depends on the concentration of the VOC and oxygen. Several of the assumptions listed above are justified in our previous works (Zarook et al., 1993, and Zarook and Baltzis, 1994b). However, some remarks should be made concerning assumptions (11) and (12). The biofilm density may vary along the biofilter column due to changes in biofilm thickness or biolayer surface area, however, its variation is not expected to be significant for a biofilter which is already in operation for a long time. Similarly, it is reported that the accumulation of biomass is very small for such a biofilter (Zarook et al., 1993), however, during start-up period exponential buildup of biomass was observed (Deshusses et a/., 1996). Thus, assumptions (11) and (12) are justified for a biofilter which is in operation

and the model developed in this study may not accurately predict initial start-up period. But one can extend the model developed in this study for predicting the performance of a biofilter for initial start-up conditions by introducing mass balance on the biomass. Deviations from ideal plug flow in a biofilter can be caused by channeling of fluid, due to wide size distribution of particles and lack of adequate means of distributing the feed. Such deviations can be explained through incorporating axial dispersion terms. Thus, the plug-flow assumption for the gas phase in the work of Zarook and Baltzis (1994b) is relaxed and dispersion effects are now introduced. Considering an airstream carrying the pollutant j, the model equations can be written as follows. Mass balances in the biojilm (a) Organic substrate (VOC)

2

2 -$

=f(Xv)Djw

SO).

/Lj(Sj,

(1)

I

(b) Oxygen

(2) with initial and boundary t = 0,

conditions,

h = 0, x = 0, Sk = s

(3) mk

O
t=o,

Sk=--

Ck

o(h)

(4) mk

t=o,

o
(5)

sk=sk,,,(x)

Ck

OU,t,

X=0,

o(h)

(6)

sk=L mk

0

<

t <

7;

h < U,t,

X =

0,

Sk

=

C,(h) -

(7)

mk

t > T; h > 0,

X =

0,

Sk

=

C,(h) -

(8)

mk

t>O;h>O,

x=6,

$=O

(9)

where k refers to either VOC (j) or oxygen (0). Since transients of the biofiltration process last for a very long time (days), while the space time (2) is in the order of minutes, without any loss of accuracy, one can omit conditions (6) and (7), and use condition (8) for any t > 0 rather than for t > T only. Mass balances in the gas phase (a) Organic substrate (VOC)

- (1 - a)A,*kj(Cj - Cf)

General

transient biofilter model

(11) with the following

initial and boundary

conditions:

t = 0, h = 0. CA= CL,.”

(12)

t = 0. 0 < h < If, Ck = C,,,,(h)

(13)

h =o,

l),li$=

t>O,

_U9(Ckl,,

h=H

’

- cklo*) (14)

(7cI=O

115)

?h

where k refers to either VOC (j) or oxygen (0). Conditions (14) and (15) are the well-known Danckwerts’ boundary conditions which account for axial dispersion effects at the inlet and exit of the biofilter. Mass hulance

in the

solid phuse

(1 - “)/ly~

= kj(l

- ‘~)As*(Cj -

Ci*)

(16)

with initial condition

expressions

and udsorption

(17)

isotherms

Solution of the model requires knowledge of kinetic expressions and adsorption isotherms. This model is compared, as will be discussed in a later section, with transient biofiltration data of two hydrophobic solvent vapors; namely, with benzene and toluene. The specific growth rates of the biomass on benzene and toluene were determined through shake flask experiments by Oh et trl. (1993) and Zarook and Baltzis (1994b) as follows. Andrews (or Haldane) kinetics for toluene:

PTsj

pits;) = K, +

sj +

i Monod

3 K,i

A%,).

OE‘THE MODEL

EQU.ATIONS

The transient model constitutes a problem of coupled partial differential equations (l)-(21) in three dimensions: time, length scales in biolayer and biofilter height. The above set form a system of highly non-linear complex equations. We have solved this system by two approaches. In the first approach, the number of partial differential equations is reduced to three instead of five by making the assumption used by Zarook and Baltzis (1994b).This is done to reduce the mathematical complexity of the problem and hence the resulting final model is labeled as the trpprorimtrtc model. In the second approach, the model equations (lW?l) are solved without any simplifications and hence the resulting final model is labeled as the !/entJr.rrl model. Approzirmte

t = 0, h > 0, C, = C,,,(h) Kinetic

In this work, the adsorption parameters of benzene were determined experimentally using a procedure similar to that used earlier in the aforementioned study. Values of the parameters estimated experimentally are given in Table I. The results of this experiment are also presented in Fig. 2. Note that due to the higher hydrophobic property of benzene. the value of kR is little less than that for toluene, as expected. SOLCTIOR

* . t >o.

763

rnodvl

In this approach.

to simplify the solution of the transient model, a quasi-steady-state approximation is made for the biolayer wherein effectiveness factors are introduced by following the approach of Zarook and Baltzis (1994b). The effectiveness factor is defined as the ratio of substrate or oxygen flux into the biofilm to the biodegradation reaction rate if the entire biofilm were exposed to air/biofilm concentration of

(18)

kinetics for benzene:

I’,(S,,) =

&dS,,I.

The functional dependence on oxygen is given by

(19)

of the specific growth rate

H(So) =(K,

so + So)

(20)

The adsorption of toluene on the solid packing material is described by the Freundlich isotherm as given by Zarook and Baltzis (1994b). C,

= kdi(CT)“‘.

(21)

~6’~~~_~~~

~_i______ 8

12

16

C’, , (g - benzene/ m3 air)

Fig. 2. Equilibrium adsorption isotherm of benzene on peat. The curve represents a fitting of the data points to the Freundlich isotherm.

S. M. Zarook et al.

764

either the VOC or oxygen, respectively. These factors are given as follows.

and for Monod kinetics as Y(Cj3cO) =

For VOC (j):

ftxV)DjW

ei= -

The Fruendlich given by

2 = 0. x

CjCO&j&O

(28)

(1 + CjjFj)(l + CO&O)’

isotherm in dimensionless

form is

0

Cf = $j(Cjp)l/nj.

6$CPj(sj. SO)lx=O J

(29)

Initial conditions:

For oxygen:

r = 0, 5 = 0,

e. = -

Z = 0,

0 d Z < 1,

C, = 1, Cji, = Cji,,O(O) (30) C, = ck,O(Z)r

Cjp = Cjp.,(Z). (31)

(23) Boundary conditions:

The use of effectiveness factors, in conjunction with the quasi-steady-state approximation, permits the omission of eqs (lH8), which are related to mass balances on the biofilm. The following are the dimensionless variables and groups used in deriving the model: cjP(l

-

COi CO= Komo

x =

ejHaAzGXVp*

(1 -

UU,CjiYj

P2 =

’

eoHcrA;GXyp* DUgCOi

yOj

a)HA,*kj uu,

’

*j

=

&..(I

~D~~rkd~l’nj’

With the above dimensionless groups, the model equations are reduced to the following system of three differential equations (24H26) along with initial and boundary conditions. Mass balances in the gas phase

acj_ 1 Nj al

1 Xj

PejaZ2

- ii%i!

- PI g(Cj, CO) - X(Cj -

at

azco iaco - ;- - B2gtcjT Pea azz

(32)

t>o,

CO).

dCk=() az

Z=],

General model

Here, the model equations have been solved without any simplifications or quasi-steady-state approximation in the biofilm. Thus, the following additional dimensionless variables and groups are introduced:

CT)

P3=

I

C,jzzo+)

_

(24)

de, -=--

2 =- Pek(CklzCom-

where k refers to either VOC (j) or oxygen (0). Solution of the approximate model requires correlations for film thickness (6) and effectiveness factors (ej, eo). The correlations used in the work of Zarook and Baltzis (1994b) were obtained by fitting the film thickness and effectiveness factors to the solution of the model under quasi-steady-state approximation in the biofilm. Although this procedure simplifies the mathematical complexity, it requires elaborate calculations prior to solution of the model. Furthermore, the correlations are not general and they are dependent on the VOC under study. Hence, the general model, without any simplifications, is solved as described in the next section.

u)P~

Cji 0

Pl =

5 > 0, Z = 0,

(25)

Djwff~&.ffXv)Kj 9

VUgCji6

~owH~&f(XvWo

84=

DllgCOiG

Mass balance in the solid phase

ac,

43 =

DowuK$WvJ; d4

!l

where g(cj, co) is given by for Andrews (or Haldane) kinetics as S(cj, CO) =

- CjCO&j&O

(1 + Cj&j + ~~&~y)(l + Co&O)

(27)

=

;yi%:

01

01

9

Introducing the above dimensionless variables and parameters, the model can be reduced to the following system of differential equations (34), (35), (41) and (42) along with their initial and boundary conditions (36H40).

165

General transient biofilter model Mass

halancrs

in the hiofilm

The initial and boundary 5 = 0,

c - 0, c-

where ;‘1 and ;‘2 are constants which normally have values of about 0.7 and 0.5, respectively. Considering the bed as an assemblage of randomly oriented cylin-

conditions

Z = 0, I) = 0. s, = Q,

0 < 0 < I, s, = S,,,(O)

; > 0, 0 = 0,

z > 0, s, = 8kCk(Z)

<>O,f1=1,

z>o,s=o

(39) (40)

Numerical

are:

0 < z < 1, 0 = 0, s, = &S&Z)

< = 0.

drical pores suggests 7, = l/42, which is close to the experimental values derived from dispersion measurements for gases at low Reynolds number (Ruthven, 1984). A typical biofilter operates at a low Reynolds number of about 0.2-0.5 (Hodge and Devinny, 1995). The bed tortuosity is related to voidage as ;‘, = 0.45 + 0.55~. Once an average radius of particles the (R,), ?I and ;j2 are known, one can estimate dispersion coefficient Dj from eq. (43). All the other parameters used in solving the model equations are listed in Table 1.

r’u

(36) (37) (38)

where k refers to either VOC (j) or oxygen (0). Mass

balances

in the gas phase

Note that the mass balance in the solid phase [eq. (26)] and initial and boundary conditions [eqs (30)+33) required for solving the mass balance equations in the gas and solid phases are the same for both approximate and general models. Estimation

of’model

parameters

Solution of these two models requires parameter values. Most of the parameters needed to solve the models can be obtained from the previous works (Zarook, 1994; Zarook and Baltzis, 1994a, b). As stated previously, the adsorption parameters for benzene have been experimentally found, in this work, by using the same procedure described by Zarook and Baltzis (1994b). The mass transfer coefficient of benzene from gas to solid phase is assumed to be the same as that for toluene. The film thickness and effectiveness factor correlations obtained from the same study were used to solve the approximate model. Thus, the correlations used were 6j = 1.5Cj + 33.4 and ek = O.O3C, + 0.2 for film thickness and effectiveness factors, respectively. In the case of the general model, a value of the film thickness is not necessary as the model itself determines the value when either VOC or oxygen gets completely consumed. This procedure of estimating the effective film thickness is discussed in more detail by Zarook et a/. (1993). The values for the gas-phase dispersion coefficients for the compounds benzene, toluene and oxygen were estimated using the following correlation which is, in general, valid for flow through porous packed-beds (Ruthven, 1984): Di = ;‘, Dj,

+ ;‘z2Rpu,/v

(43)

method

In simulation problems related to such type of mathematical formulations, the numerical solution scheme of orthogonal collocation has been found to be quite effective. In essence, the orthogonal collocation replaces the spatial derivatives by the so-called collocation matrices and hence the set of partial differential equations is reduced to a set of simultaneous ordinary differential equations which can then be solved by a variety of standard numerical schemes available for the solution of ordinary differential equations. The resulting partial differential equations for the approximate model equations (24t(33) and the general model (26) (34)+42) are reduced to ordinary differential equations by orthogonal collocation. Ten collocation points were used for discretizing the column height Z (from Z = 0 to Z = 1) for both models. In the case of the general model, six points were used to discretize 0, the biofilm depth (from 1) = 0 to l), in addition to ten collocation points for the column height Z. The resulting set of simultaneous ordinary differential equations, 30 for the approximate model and I50 for the general model, were then solved using the subroutine DIVPRK of the International Mathematical Subroutine Library (IMSL). RESULTS AND DISCUSSIONS

For validating the models, the experimental data of Zarook and Baltzis (1994a, b), and Tang et al. (1995) are used. Figure 3 shows comparisons between the predicted concentration profiles of three models, namely the plug-flow model of Zarook and Baltzis (1994133,approximate and general models of this work against the toluene transient biofiltration data. At one-third height indicated by the group (a), it is clear that both the general and approximate models predict better than the plug-flow model. However, the predictions by the general model are better. At the exit of the column indicated by the group (b), data are scattered, thus it is difficult to give any conclusion from this graph. However, all model predictions are not widely different. Both models were compared with experimentally evaluated (Zarook and Baltzis. 1994a)

766

S. M. Zarook et al. Table 1. Parameter values used for solving the approximate and general model equations Parameter

Value

Units

A A;;

40.0 23.3 275.0 0.0895 0.0792 0.2132 1.04 X lo-9 1.03 x lo-9 2.41 x lO-9 0.195 6.04 x 10-3 8.68 x 1o-6 2.25 X lo-5 78.94 0.26 12.22 11.03 0.23 34.4 0.27 1.04 0.93 100.0 0.708 0.336 0.341 0.708 0.3 1.50 0.68 0.3 4.28 x 10’

m-’ m-’ grn-j cm2sm1 cm* SK’ cm2s-’ m2s-’ m’s_’ m’s_’

co<

D BA

DTA DOA DBW DTW

f& b k, kdll kdT KII-

Ko KB KT mlJ

mo mT n(Toluene) n(Benzene) X” Y, YOB YOT Y, a * Q prl ” PP

mh-’ g/g---particle g/g-particle gmm3 gme3 grnm3 gmm3

Ref. 1

I 1 3 3 3 1

1 1 1 I

‘!

Present study

1

1 1 2 Present study kg me3

1 1

h-’ h-’ grn-j

Note: (1) Zarook and Baltzis (1994a); (2) Zarook and Baltzis (1994b); (3) Hsieh et al. (1993).

steady-state removal rate which is defined as the mass of VOC consumed per volume of packing per time. In almost all cases, the approximate and general model predict better than the plug-flow model, and the general model predicts better than the approximate model. Hence, results from only the general model are used in the analysis of subsequent sections. In the case of toluene as discussed, the improvement is not significant, however this is not the case with benzene. Table 2 shows the space time, measured inlet and exit concentration of benzene at steady-state conditions. The predicted exit concentration profile by the general model is improved significantly as compared to the previous model of Zarook and Baltzis (1994a). Figures 4(a) and (b) again show the experimental and model (general) predicted concentration profiles for benzene in the column for two different operating conditions. As can be seen, the general model predictions are in good agreement with the experimental results. However, Fig. 4(a) shows that the initial period of four days is not well described by the general model. Experimental data reported in Fig. 4(a) were

obtained prior to those of Fig. 4(b), thus the initial adsorption process as well as the process of formation of biofilm in the initial period may be the reasons for this long transient period. Hence, the general model developed in this study may not give reasonable predictions of the initial start-up period, however, it can be used for predicting the transient performance from one set of operating conditions to another. Shock loading effects Figures 5 and 6 compare the model predictions against the recently reported experimental data of Tang et al. (1995). They report shock loading studies with the inlet toluene concentration and gas velocity in biofilters with three types of packing materials: chaff/compost, diatomaceous earth/compost and granular activated carbon/compost. Although the experimental results are interesting, a theoretical model was not presented. In our study, the experimental data of the chaff/compost biofilter were chosen. Since the parameters for chaff/compost packing material were not available, the same parameters as for

161

General transient biofilter model peat/perlite packing, except for the biolayer surface area, were used to predict the biofilter performance by the general model. The value of the biolayer surface area, was obtained by fitting the data reported in Fig. 5. In fitting, an attempt was made to describe all the data points. Notice that there is an abrupt increase in the exit concentration when the inlet concentration is increased to 3.33 from 1.87 gmm3. This may be due to desorption of toluene. This part of the data is not well described by the model. Otherwise, with a single fitted parameter value of 88 m2 mm3 for AsT, our model is able to predict well the dynamic response of the biofilter to a sudden change in the inlet concentration. This parameter value was unchanged and was used in the model to describe the dynamic response to a sudden change in the inlet gas flow rate. Figure 6 shows that

the model is able to describe, at least qualitatively, the dynamic response reasonably well considering the fact that the parameters, especially the mass transfer coefficient and isotherm parameters were different. If these parameters were experimentally determined for chaff/compost packing and used in the solution of the

Cai = 0.43 g m-3

??

Exp.(7 - 17 June, 93)

(Zarook & Baltzis, 1994a)

T = 4.5 min

P-----l I 0

Plug Flow Model (Zarook& Baltzis, 1994b Approximate Model General Model

--___

0

(a)

__---. // _______ jl= .

.

?? Experimental Data (Zamok & B&is, 19941

-J

0

2

I

I

I

4

6

a

10

Time (days)

1.0 ,

(a)

/

‘-----.

I

,?.8

-

T = 4.1 min

/I li l’

,;

!

0

(b)

C,, = 0.28 g me3

.-__--

0

0

4

I

~,,,_____!“!~-__~-----.

,’

1

?? Exp.(S - 12 July, 93)

i

2

3

4

(Zarook & Baltzis, 1994a)

5

Time (days)

Fig. 3. Transient biofiltration data of toluene vapor when r = 6.3 min and Cri = 2.81 g m - ’ are compared against the data predicted by three models, namely plug-flow model (Zarook and Baltzis, 1994b), approximate model and the general model. Curve (a) at one-third height and curve (b) at the exit of the biofilter.

0.0 l. 0

I 1

/ 2

I 3

I 4

I 5

6

Time (days)

Fig. 4. Transient biofiltration data of exit concentration of benzene (Zarook and Baltzis, 1994a) is compared against the model predictions.

Table 2. Comparison of biofilter models against experimental data of benzene (Zarook and Baltzis, 1994a) under steady-state conditions

t (min)

Ci”,,, (gme3)

c (exp) OU’ (gmm3)

C out (model)* (g m3)

4.1

0.28

0.19

0.16

- 15.8

0.18

4.5

0.43

0.23

0.25

8.7

0.24

4.3

4.7 2.7

0.56 0.13

0.2 1 0.09

0.31 0.09

47.6 0.0

0.29 0.09

38.1 0.0

2.7 4.1

0.12 0.07

0.08 0.04

0.09 0.04

- 11.1 0.0

0.08 0.04

0.0 0.0

Error (X)

*Plug-flow model of Zarook and Baltzis (1994a). ‘General model of present study.

c DUL (model)’ (g mm3)

Error W) - 5.3

768

S. M. Zarook

et al.

biofilter is not affected by random perturbations considerably. Figure 8 shows response of the biofilter to shutdown and restart-up conditions. For simulating data,

0.6 1

C,,= 1.65g m-' : T -~7.7min

0 48

96

144 192 240 268 336 384 432 480

1

104

L_

00

0

(a)

0.1 -

0.0

,

2.0

F

0.5

_-L-L-L.

10

5

15

528 576

20

25

30

J

0.0 35

( 5)

Time

Time (hr)

Fig. 5. Transient response to abrupt change in the inlet toluene concentration. Concentration is varied from 0.051 to 1.87 and then to 3.32 gmm3. Gas velocity uq and residence time T were kept at 0.67 cm s- ‘, and 2.5 min. respectively.

(b)

5.4

5.2

: f

E

1.0

5.0

II

0.9

;

4.8

0.8 Inlet Cont. -0.7 ‘E

0 ___

0.0

Exit Conc.(Tang et al., 1995) Exit Conc.(modeI)

%6 5 ‘50.5

Y 0

5

I

I

10

15

Time

.-

20

25

1.

4.6 30

( 5)

Fig. 7. Transient response of the exit toluene concentration to the random perturbations of (a) inlet concentration and (b) gas velocity

EO.4 s 0.3

1.0 0.9

0.1 0.0

r----

up=0.67 cm/s

0.8 -

0

24

48

72

96

120 144 168 192 216 240 264 288 Time (hr)

-0.7

Inlet Cont. 0

-

‘E %6

Exit Conc.(Tang et al., 1995)

~

Exit Concentration (model)

96

144

-

Fig. 6. Transient response to abrupt change in the gas flow rate. Inlet concentration is varied from 0.86 to 0.84 and then to 0.86 g m 3 and the residence time is kept at 2.5 min.

model, then the predictions would probably be much better. In practical applications, an emission level is likely to vary with time. One of the most common types of such variations will be due to random fluctuations of gas velocity and inlet concentration. It is interesting to examine the dynamic response of a biofilter to such variations. Response curves of such a study are given in Figs 7(a) and (b), which show that the biofilter is able to respond very well to the random variations in the inlet concentration and gas velocity, respectively. In both cases, the steady-state performance of the

0.0 0

48

192

240

288

336

384

Time (hr)

Fig. 8. Transient response of the exit toluene concentration to shut down and restart of the biofilter. Inlet concentration and superficial velocity before shut down and after restart were 0.89 gme3 and 0.67 rns-‘.

General

transient biofilter model

inlet concentration and gas superficial velocity used before shut-down and after restart-up were about 0.89 g m- 3 and 0.67 m s- ‘, respectively. Model predictions show that the biofilter is able to respond to previous steady-state conditions within a day after a six-day shut-down period. Thus, both model predictions and experimental data of Tang et al. (1995) show that activity of the biofilter can be restored within one to two days. Oxygen limitations

Figures 9(a) and (b) show the concentration profiles of toluene and oxygen in the biofilm at a column height of 0.25 m and at different inlet concentrations. It can be seen that at an inlet concentration of 0.62 g m _ 3, toluene gets depleted at a biofilm depth of around 35 pm [Fig. 9(a)]. But interestingly, as the inlet concentration is increased to 6.2 gmm3, oxygen gets depleted at a depth of around 12 pm indicating oxygen limitation at higher substrate concentrations. This confirms the findings by Zarook and Baltzis (1993) that for hydrophilic solvents oxygen limits the process even at low inlet VOC concentrations.

769

Numerical results also show that at moderately high inlet toluene concentrations and large residence times, the exit oxygen concentration was low, indicating greater consumption of oxygen in the biofilter. Figure 10 shows predictions of the exit concentration profile along the biofilter column by the general model for inlet toluene concentration of 2.81 grn-’ and residence time of 6.3 min. As can be seen, the exit oxygen dimensionless concentration falls sharply and remains constant at a value of about 0.6. Thus, oxygen level drops from 21 to 13% in the polluted air. The transient model of Deshusses et al. (1995) is limited to conditions where oxygen limitation does not occur, as in the case of biofiltration of methyl ethyl ketone (MEK) and methyl isobutyl ketone (MIBK). Even for this system, at high inlet concentrations oxygen limitation may occur. Furthermore, it is the depletion of oxygen or VOC that determines the active biofilm thickness. All of these results imply that oxygen needs to be explicitly taken into account in the modeling of biofiltration process under steady-state or transient conditions. Sensitivity studies

8 I--

l

(a)

Toluene oxygen

0

5

10

15

20

25

30

A thorough investigation of parameter sensitivity of the model was performed. Some of the important results are discussed in this section. Figure 11 shows that the effect of PeT on the toluene concentration profile in the column. The symbol j’is the relative parameter value of PeT, i.e. a large value off implies higher Peclet number as compared to the predicted values calculated using the parameters given in Table 1. A high value for Per denotes plug flow and low value indicates well-mixed flow in the biofilter. It can be seen that as the value of .f’decreases, the concentration profile flattens, indicating a transition from plug-flow behavior to mixed flow.

35

Blofilm Depth 6 (pm)

-

20.4 lz

Toluene Oxygen

01

0

1

I

5

10

Toluene

I

I

1

,

I

15

20

25

30

35

I

Biofilm Depth 6 (Nm)

Fig. 9. Concentration profiles of toluene and oxygen in the biolilm. (a) Toluene gets depleted in the biofilm at low inlet concentrations to the biofilter (CTi = 0.62 gme3). (b) Oxygen gets depleted in the biofilm at high inlet toluene concentration to the biofilier (CTi = 6.2 gmm3). For both cases, T = 2.7 min at 2 = 0.41.

-i 0.2

rp

o.ol”, 0

,~~ --.-

IO

20 Time

315

( 5)

Fig. 10. Transient concentration profiles of oxygen and toluene; inlet conditions C,; = 2.81 gm-” and t = 6.3 min.

et al.

S. M. Zarook

770

I_

‘$

_ -------

\

f=O.Ol f=O.l f=0.3 f=0.6 f=l f=5

1

\

‘0 ~0.6 P b s 50.4

0.2

0.2

0.0

,

0.0 0.0

0.2

0.4 Height

0.6

0.8

1 .O

0.0

02

I

I

I

0.4

0.6

0.8

(2 )

Fig. Il. Etfect of Pe, on the toluene concentration the column.

10

Height(Z ) profile in

A decrease in PeT implies better mixing in the gas phase leading to higher rates of mass transfer and better conversions. These conditions prevail more at the inlet of the column where the concentration levels are maximum. Similarly, the general model seems to be very sensitive to the value of Peclet number of oxygen. It can be seen in Fig. 12 that as the value of Pea is reduced by a factor of 10, the concentration profile along the column is flat with a higher exit toluene concentration. These results indicate that prior knowledge of the mixing pattern is very important in model predictions. Thus, the plug-flow assumption of previous works (Ottengraf and van den Oever, 1993; Zarook et al., 1993; Zarook and Baltzis, and 1994b; Deshusses et ul., 1995) is questionable mixing pattern in the gas phase still needs to be experimentally verified through residence time distribution (RTD) studies. Figure 13 shows the effects of dimensionless parameters pi, lj2, Ej and co on the exit toluene concentration. As can be seen, the model is very sensitive to these parameters for low relative values, less than one, after which the effect is not significant. The parameter pi can be interpreted as the ratio of the amount of substrate diffused into the biofilm to the degree of convection in the gas phase. Larger values of 8, or B2 imply greater mass transport into the biofilm resulting in higher conversions, Large values of cj imply lower volatility of the compound, i.e. the lower the volatility of the compound, the greater the diffusion and reaction in the biofilm. The figure also shows the effect of ao on the exit concentration. As ao increases the exit concentration falls. This happens sharply at very low relative parameter values. This is expected as co is proportional to the inlet oxygen concentration which has a drastic effect on the conversion. Figure 14 shows the effect of the biofilm parameters 4,, cj2 and & on the exit concentration of toluene. As

Fig. 12. Effect of Pe, on the concentration

profile

in the

biofilter.

0.8

‘OO.6 .-5 E 5 E 2 0.4

5

w

0.2

I 1

2

3

4

5

Relattve Parameter Value Fig. 13. Effect of dimensionless parameter p,, /13, cj, and E,, on the exit toluene concentration. Relative parameter value I refers to condition of an actual experiment.

can be seen, the model is very sensitive to all these parameters. The exit concentration increases with increase in 4i. A large value of 4i indicates higher diffusion of toluene into the biofilm. This leads to reaction limitation in the biofilm resulting from substrate inhibition due to large toluene concentration and less availability of oxygen. Large values of d2, on the other hand, imply greater reaction in the biofilm and less convection in the gas phase. This result could be easily expected due to high rate of reaction and large residence time. Greater diffusion of oxygen in the biofilm results in greater conversion and this explains the behavior of the exit concentration for the

771

General transient biofilter model

0.6

0.0 1

2

3

4

5

RelativeParameter Value Fig. 14. Effect of 4,, +z and & on the exit toluene concentration.

parameter 43. These results indicate that except for the parameter 41 all other parameters, at high relative parameter values (> l), are not sensitive, thus accurate determination of parameter $1 is more important than the others. However, at low relative values (< l), the model is sensitive to all of these parameters. CONCLUSIONS

A thorough literature review on the modeling of transient biofiltration process shows that the theoretical studies are very limited; in fact only three models (Zarook and Baltzis, 1994b; Hodge and Devinny, 1995; Deshusses et al., 1995a) are available. These models are based on restrictive assumptions, hence in the present study a general model which incorporates general mixing phenomena, oxygen limitation aspects, adsorption and general reaction kinetics is developed. The model has been solved by two approaches (approximate and general) and validated with experimental data of previous works. The approximate model requires correlations for film thickness and effectiveness factors as does the plug-flow model of Zarook and Baltzis (1994b); these correlations are based on the assumption of pseudo-steady state in the biofilm. Although the approximate model is superior to the plug-flow model of Zarook and Baltzis (1994b) as it includes mixing phenomena, it is still inferior compared to the general model. Furthermore, the results show that predictions by the general model are improved significantly. Dynamic analysis of the model is compared with the experimental data obtained from the literature. Both experimental data and model predictions have shown that transient conditions during shut-down and restart-up are not very long. Moreover, the theoretical predictions of responses to random variation in the operating conditions show that the biofilter is able to withstand

extreme conditions commonly encountered in practical applications. Thus, these results show that the model is able to qualitatively describe the transient aspects of the biofiltration process very well. Studies on oxygen limitation effects show that the assumption of excess oxygen availability is not a good one, specially at high inlet concentration levels. Sensitivity studies show that accurate estimation of some parameters is more important than others. Particularly, these studies show that mixing phenomena is very important, thus the plug-flow assumption of previous works is still in question. It is therefore advisable that an experimental RTD analysis be performed to test the mixing patterns in biofilters. Such an experimental investigation is in progress and will be reported in a forthcoming contribution. Furthermore, it should be acknowledged that the model developed in this study is highly complex, limited to single VOC removal and for a biofilter which is already in operation. The complexity of the model is inevitable since several aspects of biofiltration process, namely inhibitory kinetics (Andrews, 1968), general mixing phenomena, limitations by oxygen, variation in the biofilm thickness, etc. are incorporated. Furthermore, the biofiltration process involves three phases, namely gas, liquid and solid, thus, mass balances of VOCs and oxygen should be considered in all phases. However, for specific cases this model can be simplified. For example, if one neglects the effect of oxygen limitations or use plug flow for the gas phase, complexity of the model will be substantially reduced. In such limiting cases, the model will not be general and will be limited to the compound or biofilter under study. Biofiltration of many VOCs is limited by the availability of oxygen and recent experimental findings suggest that there is significant amount of dispersion in biofilters. Biofiltration is a complex process, thus, the complexity involved in modeling of this process is not easily avoidable. A good quality transient biofiltration experimental data is necessary for further validation of this model as well as for a model to describe biofiltration of mixed VOCs. Extension of this model for biofiltration of VOC mixtures is not direct and simple as it involves, in addition to processes described above, interactions between VOCs, balances for all VOCs and oxygen and multicomponent adsorption as well. Such experimental and theoretical study is in the scope of our future research. Acknowledgement The authors are grateful to King Fahd University of Petroleum and Minerals for the support of this work, through the funded project # CHE/BIOFILTER/l80. NOTATION AT

Asj

total surface area available for biolayer formation and adsorption per unit volume of biofilter, m- ’ biolayer surface area per unit volume of reactor, for VOC,, m-l

772

S. M. Zarook et al.

cj Cj* cj,O Cji cji,

0

CjP

cjP,

0

co

C 0.0 cOi

coi.

0

cj

q

cjj,

CO

Djw

Dj

DjA

ej

e0

f(X") h H kj k, Ki

K,j

Ko

mj m0

nj

concentration of substance j in the air at a position h along the biofilter, g m-3 equilibrium pollutant j concentration in the gas phase, gm-3 ValUeOfCjatt=0,gm-3 value of Cj at h = 0, gmm3 value of Cj at h = 0 and t = 0, gmm3 concentration of substance j on the solid particle, g of pollutant j-adsorbed/g particle value of Cj, at t = 0, y of pollutant j-adsorbed/g particle oxygen concentration in the air at a position h along the biofilter, gmm3 valueofCoatt=0,gm-3 oxygen concentration in the air at the inlet of the biofilter, g m- 3 oxygen concentration at h =0 and t=0,gm-3 dimensionless concentration of pollutant j in the air defined as Cj = CjICji dimensionless equilibrium concentration of pollutant ,j defined as CT = CT JC,, dimensionless concentration of substance ,j on the solid particle defined as (l - u)Ppcjp/ucji dimensionless concentration of oxygen in the air defined as Co = Co/Coi diffusion coefficient of substrates in water, m2 h-r dispersion coefficient of substrates in air, m2hm1 diffusion coefficient of substrates in air, m2h-’ effectiveness factor based on pollutant j, defined by eq. (22) effectiveness factor based on oxygen, defined by cq. (23) ratio of diffusivity of a compound in the biofilm to that in water position in the column, m; h = 0 at the entrance, h = H at the exit total height of the biofilter bed, m mass transfer coefficient between the gas and the solid particle, m h- 1 Freundlich isotherm parameter for the organic substrate j constant in the specific growth rate expression of a culture growing on compound j, gm-” inhibition constant in the specific growth rate expression of a culture growing on compound j grne3 constant in the specific growth rate expression of a culture, expressing the effect of oxygen, g me3 distribution coefficient for the substance j/water system distribution coefficient for the oxygen-inair/water system Freundlich isotherm parameter for substrate ,j removal

Pej Pea RP sj

Sj,o ST

So

So.0

t ug Xv X Yj Yoj

Z

Peclet number of substrate ,j defined as I.+, H/Dj V Peclet number of oxygen defined as ug H/Do radius of the particle, m concentration of pollutant ,j at a position x in the biolayer at a point h along the column, g m 3 ValUeOfSjat t=O,gm-’ dimensionless concentration of substance j at a point 0 in the biolayer; defined as Sj(e)/Sj(o = 0) oxygen concentration at a position x in the biolayer, at a point h along the column, grn-” value of So at t = 0, gmm3 time, h superficial air velocity in the biofilter, rnh-’ biofilm density, g-dry cell mm3 position in the biolayer, m yield coefficient of a culture on VOCj, g-biomass g- l-compound j yield coefficient of a culture on oxygen, gbiomass g ’ -oxygen, when VOCj is the carbon source dimensionless position in the biofilter

( = h/H) Greek letters fraction of total surface area available biofilm formation dimensionless group defined [ejx6A,*XV Hpf/( YjUqCjiO)] dimensionless group defined a2 ci

for

Ceoz6As*XVH~Lfl(Yoj~gCoI~)1 B3

dimensionless

group defined Cji u)I group defined

CDjwHr @.f(XV)Kj/‘(~a, 84

dimensionless

6 6*

CDowHxAs*f(XV)Kol(Gu,Coiu)l inverse dimensionless inhibition constant defined as 7 = Kj/K,j effective biolayer thickness, m actual biolayer thickness, m dimensionless Henry’s coefficient for j defined [CjJ(mjKj)] dimensionless Henry’s coefficient for oxygen

cj

CO

0 Pi

5 PP

defined CCoi/(moKo)l dimensionless position in the biolayer defined (x/6) specific growth rate, h r, given by equations (18) and (19) constant in the specific growth rate expression, h- ’ dimensionless time defined (u,t/H) density of the solid particles, gme3 space-time, h porosity of the biofilter bed dimensionless group defined

CDjwHf(X~)/(Ug62)1 dimensionless

group defined

CHXrp(i+I(u, YiKj)I

General transient biofilter dimensionless

group defined

c~ow~f(x”Y(~g~z)l dimensionless

group defined

CHX,Cli*l(u, YOjKOj)I dimensionless group defined Ckj(l - a)AbH/(uq 011 dimensionless group defined i l/C,;i) [ UC,i/( 1 - V)pp kdj] ““‘} Special subscript j = B, T compounds are benzene and toluene, respectively k= B, compounds are benzene, toluene and oxyT. 0 gen, respectively REFERENCES

Andrews, J. F. (1968) A mathematical model for the continuous culture of microorganisms utilizing inhibitory substrates. BiotechnoL Bioengng 10, 707. Baltzis, B. C. and Zarook, S. M. (1993) Modeling and preliminary design criteria for packed-bed biofilters, pp. l-16. In Proceedings ofthe 86th A&WMA Meeting, Paper no. 93-TP-52A.03. Air & Waste Management Association, Denver, CO. Chang, M.-K., Voice, T. C. and Criddle, C. S. (1993) Kinetics of competitive inhibition and cometabolism in the biodegradation of benzene, toluene, and p-xylene by two Pseudomonas isolates. Biotechnol. Bioengng 41, 1057. Deshusses, M. A. (1994) Biodegradation of mixtures of ketone vapors in biofilters for the treatment of waste air. Ph.D. dissertation, Swiss Federal Institute of Technology, Zurich, Switzerland. Deshusses, M. A. and Dunn, I. J. (1993) Modeling experiments on the kinetics of mixed-solvent removal from waste gas in a biofilter. In Proceedings of the 6th European Congress on Biotechnology, Florence, 13-l 7 June. Deshusses, M. A. and Hamer, G. (1993) The removal of volatile ketone mixtures from air in biofilters. Bioprocess. Engng 9, 141. Deshusses, M. A., Hamer, G. and Dunn, I. J. (1995a) Behavior of biofilters for waste air biotreatment 1. Dynamic model development. Environ. Sci. Techno/. 29, 1048. Deshusses, M. A. Hamer, G. and Dunn, I. J. (1995b) Behavior of biofilters for waste air biotreatment 2. Experiment evaluation of a dynamic model. Environ. Sci. Technol. 29, 1059. Deshusses, M. A., Hamer, G. and Dunn, I. J. (1996) Transient-state behavior of biofilters removing mixtures of vapors of MEK and MIBK from air. Biotechnoi. Bioengng 49, 587. Fan, L.-S., Leyva-Ramoq R. Wisecarver, K. D. and Zehner, B. J. (1990) Diffusion of phenol through a biofilm grown on activated carbon particles in a draft-tube three-phase fluidized-bed bioreactor. Biotechnol Bioengng 35, 279. Hodge, D. and Devinny, J. (1995) Modeling removal of air contaminants by biofiltration. J. Environ. engng 121,21. Hsieh, C., Kyoung, S. and Stenstrom, M. K. (1993) Estimating emissions of 20 VOCs I. Surface aeration J. Environ. engng 119, 1077.

model

113

Leson, G., Hodge, D. S. Tabatabai, F. and Winer, A. M. (1993) Biofilter demonstration projects for the control of ethanol emissions. 86th Annual Meeting of the Air 8~ Waste Management Association, Paper No. 93-WP-52C.04, Denver, CO, June 13-18. Monod, J. (1942) Recherches sur /a Croissancr des Cultures Bacteriennes, Hermann et Cie, Paris Oh, Y. S., Zarook, S. M. Baltzis, B. C. and Bartha, R. (1994) Interactions between benzene, toluene, and p-xylene (BTX) during their biodegradation”, Biotechnol. Bioengng 44, 533. Ottengraf, S. P. P. (1986) Exhaust gas purification. In Biotechnology, ed. W. Shonborn, Vol. 8, pp. 425-452. VCH Verlagsgesellschaft, Weinheim, Germany. Ottengraf, S. P. P. (1987) Biological systems for waste gas elimination. Trends BiotechnoL 5, 132. Ottengraf, S. P. P. and van den Oever, A. H. C. (1983) Kinetics of organic compound removal from waste gases with a biological filter. BiotechnoL Biorwgmg 25, 3089. Potter, T. L. (1992) Fingerprinting petroleum products: Unleaded gasolines. In Petroleum Contaminated Soils, eds P. T. Kostecki and E. J. Calabrese, Vol. 3, pp. 83-92. Lewis Publishers, Chelsea, MI. Reisch, M. S. (1992) Top 50 chemicals production stagnated last year. Chem. Engng News 70, 16. Ruthven, D. M. (1984) Principles of Adsorption and Adsorption Processes. Wiley, New York. Tang, H. Hwang, S. and Hwang, S. (1995) Dynamics of toluene degradation in biofilters. Hal. Waste Haz. Mat/s 12, 207. Tong, C. C. and Fan, L.-S. (1988) Concentration multiplicity in a draft tube fluidized-bed bioreactor involving two limiting substrates. Biotechnol. Bioengng 31, 24. Tonga, A. P. and Frisch, S. (1993) Field-pilot study of styrene biodegradation using biofiltration., in the proceedings of the 86th Annual Meeting of the Air & Waste Management Association, Paper No. 93WP-52C.03, Denver, CO, June 13-18. Utgikar, V., Govind, R., Shan, Y., Safferman, S. and Brenner, R. C. (1991) Biodegradation of volatile organic chemicals in a biofilter. In Emerging Technologies in Hazardous Waste Management II, eds D. W. Tedder and F. G. Pokland, pp. 233-260. ACS Symposium Series 468, Washington, DC, U.S.A. Zarook, S. M. (1994) Engineering analysis of a packed-bed biofilter for removal of VOC (volatile organic compound) emissions. Ph.D. dissertation, New Jersey Institute of Technology, Newark, New Jersey. Zarook S. M. and Baltzis, B. C. (1994a) Biological removal of hydrophobic solvent vapors from airstreams. In Advances in Bioprocess Engineering, eds E. Galindo and 0. T. Ramirez, pp. 397404. Kluwer Academic Publishers, Dordrecht, The Netherlands. Zarook, S. M. and Baltzis, B. C. (1994b) Biofiltration of toluene vapor under steady-state and transient conditions: theory and experimental results. Chcm. Engng Sci. 49,4347. Zarook S. M., Baltzis, B. C., Oh, Y.-S. and Bartha, R. (1993) Biofiltration of methanol vapor. Biotechnol. Bioengng 41, 512. Zilli, M., Converti, A., Lodi, A., Del Borghi, M. and Ferraiolo, G. (1993) Phenol removal from waste gases with a biological filter by Pseudomonas Putida. Biotechnol. Bioengng 41, 693.