Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract: A mathematical model is developed to predict continuous adsorption-regeneration cycling of volatile organic compounds
共VOCs兲 on activated carbon fiber cloth 共ACFC兲 at the indoor VOC concentration levels. The adsorption-regeneration model incorporates
both the adsorption equilibrium and mass transfer fundamentals. It assumes local equilibrium between the gas-phase and the solid-phase,
and axially dispersed-flow, film transfer, and intraparticle transport by surface and pore diffusion. Successful agreement between model
simulations and experimental data was obtained and the kinetic properties of the adsorption/regeneration cycling on the ACFC were
characterized. For the adsorption process, the film transfer is the dominant factor for mass transfer at low flow rates 共45–184 L/min兲, and
the intraparticle mass transfer rate controls over the gas-phase rate as the flow rates increase. The regeneration concentration profiles are
most sensitive to the adsorption isotherms at the temperatures of interest, especially as desorption is initiated. The surface diffusivity also
contributes to the shape of the regeneration profile: the tailing of desorption profile shifts up with the increase of surface diffusivity.
DOI: 10.1061/共ASCE兲EE.1943-7870.0000110
CE Database subject headings: Adsorption; Indoor air pollution; Activated carbon; Fibers; Organic compounds.
The dispersed-flow pore surface diffusion model 共DFPSDM兲 for + 3Stg关C̄ p,d,i共r̄ = 1,z̄,T兲 − C̄d,i共z̄,T兲兴
GAC adsorption developed by Friedman 共1984兲 is a starting point 共5兲
for the development of ACFC adsorption-regeneration model.
Since carbon fibers are woven into cloth, it is assumed that a Clean air stream passes through the ACFC fixed-bed during re-
Downloaded from ascelibrary.org by MANIPAL INST OF TECHNOLOGY on 07/19/20. Copyright ASCE. For personal use only; all rights reserved.
single bundle of activated carbon fibers is equivalent to spherical generation, therefore the associated initial and boundary equa-
particle. Therefore, the diameter of the single bundle for ACFC tions for gas-phase mass balance change to
materials is equivalent to the particle diameter. The adsorption
model incorporates both the adsorption equilibrium and mass C̄d,i共0 艋 z̄ 艋 1,T = 0兲 = C̄ave,a,i 共6兲
transfer. The mass balance including gas-phase and intraparticle
phase mass transfer, and adsorption equilibrium relationship are
1 C̄d,i共z̄ = 0+,T兲
incorporated into the model. C̄d,i共z̄ = 0+,T 艌 0兲 = 共7兲
The main assumptions made for adsorption-regeneration Pe z̄
model are: 共1兲 dispersed-flow conditions exist in the bed; 共2兲 mass
flux at the exterior surface of the adsorbent particle is described 2C̄d,i共z̄ = 1,T兲 C̄d,i共z̄,T = 0兲
by linear driving forces 共film transfer兲; 共3兲 intraparticle mass flux = 0 and =0 共8兲
is described by surface and pore diffusion; 共4兲 adsorption equilib- z̄ T z̄
rium of individual compounds can be represented by the Freun-
dlich isotherm equation; 共5兲 both density distribution and
resistance on ACFCs are uniform; 共6兲 the system is isothermal for Intraparticle Mass Balance for
VOCs; and 共7兲 temperature distribution is uniform across the sur- Adsorption-Regeneration Process
face area of carbon cloth, and no temperature gradient exists
along the radial direction of the carbon cloth due to the short bed Adsorption
length. The last assumption is valid because the surface tempera- Using dimensionless terms, the intraparticle mass balances for
ture of the ACFC remained at 150⫾ 5 ° C along the entire carbon adsorption is as follows:
cloth during constant power application. Based on these model
mechanisms and assumptions, model equations are derived for Dgt,a,i Ȳ a,i共r̄,z̄,T兲
T
再冋
multicomponents and then are converted to dimensionless equa- Dgt,a,i + 1
tions which are presented herein.
=
1
r̄2 r̄
r̄2 共Eds,a,i + D · Ed p,a,i兲 ·
Ȳ a,i共r̄,z̄,T兲
r̄
册
冎
Gas-Phase Mass Balance for Adsorption-Regeneration
Process C̄ p,a,i共r̄,z̄,T兲
+ 共1 − D兲 · Ed p,a,i · 共9兲
r̄
Adsorption
The gas-phase mass balance for adsorption in dimensionless form The virgin ACFC was free of VOCs before the first adsorption
can be written as cycle started. A “clean bed” condition is used as the initial con-
dition for the first adsorption cycle. The associated initial and
boundary conditions are as follows:
1 C̄a,i共z̄,T兲 C̄a,i共z̄,T兲 1 2C̄a,i共z̄,T兲
=− +
Dgt,a,i + 1 T z̄ Pe z̄2 Ȳ a,i共0 艋 r̄ 艋 1,0 艋 z̄ 艋 1,T = 0兲 = 0 共10兲
= 3Stg关C̄a,i共z̄,T兲 − C̄ p,a,i共r̄ = 1,z̄,T兲兴 共1兲
Ȳ a,i共r̄ = 0,0 艋 z̄ 艋 1,T 艌 0兲 = 0 共11兲
where subscript a = adsorption and p = particle. r̄
The associated initial and boundary conditions for adsorption
冕
in the first cycle are as follows: 1
1
Ȳ a,i共r̄,z̄,T兲r̄2dr̄ = Stg关C̄a,i共z̄,T兲 − C̄ p,a,i共r̄ = 1,z̄,T兲兴
Dgt,a,i + 1 T
C̄a.i共0 艋 z̄ 艋 1,T = 0兲 = 0 共2兲 0
共12兲
In order to evaluate the performance of ACFC materials after
1 C̄a,i共z̄ = 0+,T兲
C̄a,i共z̄ = 0+,T 艌 0兲 = 1 − 共3兲 a long-term operation, the model was developed to predict the
Pe z̄ adsorption/regeneration profiles for multiple adsorption/
regeneration cycles. Heating ACFC and preloading on reused
ACFCs result in variations of initial conditions for the subsequent
2C̄a,i共z̄ = 1,T兲 a,i共z̄,T = 0兲 processes. The initial conditions of the intraparticle phase mass
= 0 and =0 共4兲
z̄ T z̄ balance for the adsorber packed with used ACFCs are following:
冋 册
The Biot number is the ratio of gas-phase mass transfer rate to
p intraparticle phase mass transfer rate. The definition of the Biot
Ȳ ave,d,i Kd,i共C0,d,i兲1/nd,i + C0,d,i
number is shown as
= 共14兲
p
Ka,i共C0,a,i兲 1/na,i
+ C0,a,i k f D p共1 − 兲
Bis = 共19兲
2DsDg
Downloaded from ascelibrary.org by MANIPAL INST OF TECHNOLOGY on 07/19/20. Copyright ASCE. For personal use only; all rights reserved.
Regeneration
The intraparticle mass balance for regeneration has the same ex- Stanton Number (St)
pression as adsorption, while the subscript d stands for desorption The Stanton number is the ratio of film transfer mass transfer rate
process to mass transport rate by advection. As the Stanton number in-
creases, the rate of transport by advection controls over film trans-
Dgt,d,i Ȳ d,i共r̄,z̄,T兲 fer transport. As the Stanton number decreases, the rate by film
transfer controls. The mathematical expression for Stanton num-
T
再冋
Dgt,d,i + 1
册
ber is
1 Ȳ d,i共r̄,z̄,T兲
= r̄2 共Eds,d,i + D · Ed p,d,i兲 · k f 共1 − 兲
r̄ r̄
2
r̄ 共20兲
冎
St =
R
C̄ p,d,i共r̄,z̄,T兲
+ 共1 − D兲 · Ed p,d,i · 共15兲
r̄ Diffusion Modulus (Ed)
The diffusion modulus is the ratio of intraparticle phase mass rate
As a sequential process of adsorption, the initial conditions with
to the rate of mass transport by advection. The diffusion modulus
uniformly adsorbed mass in adsorbent were considered for the
based on the surface diffusivity is defined below
intraparticle phase mass balance. The associated initial conditions
for the intraparticle phase mass balance can be written as
LdsDg
Eds = 共21兲
Y ave,a,i v sR 2
Ȳ d,i共0 艋 r̄ 艋 1,0 艋 z̄ 艋 1,T = 0兲 = 共16兲
Y ave,d,i where L = length of the fixed bed.
The dimensionless DFPSDM partial differential equations de-
An alternative expression for Eq. 共17兲 is
scribed above were converted into ordinary differential equations
共ODEs兲 using the orthogonal collocation method. The set of
Ȳ d,i共0 艋 r̄ 艋 1,0 艋 z̄ 艋 1,T = 0兲 ODEs was solved using the backward differentiation method for-
=
冋
Ȳ ave,a,i Ka,i共C0,a,i兲1/na,i +
p
C0,a,i 册 共17兲
mula, also called Gear’s stiff method. Analytical solutions have
been derived for describing dispersed-flow conditions in the
fixed-bed by assuming steady state operation and infinite adsorp-
p tion capacity. Analytical solutions have been derived for describ-
Kd,i共C0,d,i兲1/nd,i + C0,d,i
ing dispersed-flow conditions in the fixed-bed by assuming steady
state operation and infinite adsorption capacity by Friedman
In order to solve the gas-phase and intraparticle phase mass bal-
共1984兲 and the numerical solutions for adsorption model has been
ance, the following nonlinear coupling equation is used to de-
verified by comparing with the analytical solutions 共Friedman
scribe the multicomponent equilibrium between gas-phase and
1984兲. The solutions for desorption model were verified using the
intraparticle phase during adsorption and regeneration processes:
same method.
Both the adsorption and regeneration models were coupled to
p
Ȳ i共r̄,z̄,T兲Y e,i − C̄ p,i共r̄,z̄,T兲Co,i continuously simulate multiple adsorption/regeneration cycles,
a and compiled in FORTRAN code. In addition to performing cal-
冋 册
C̄ p,i共r̄,z̄,T兲 = m
p culations to determine the adsorption/regeneration profiles, the
C̄o,i 兺
j=1
Y j共r̄,z̄,T兲Y e,j −
a
C̄ p,j共r̄,z̄,T兲C̄o,j model also provides output for tracking the adsorbed/desorbed
mass during each cycle by integrating the adsorption/regeneration
冤 再冋 册冎 冥
concentration profiles for each cycle. These calculations provide
m
p ni predictions of long term operation over multiple cycles.
兺 nk Ȳ k共r̄,z̄,T兲Y e,k − C̄ p,j共r̄,z̄,T兲Co,i
k=1 a
g 共18兲
n iK i
Parameter Determination
Parameter Correlations
Adsorption and Desorption Experiments
Isotherm Parameters
Downloaded from ascelibrary.org by MANIPAL INST OF TECHNOLOGY on 07/19/20. Copyright ASCE. For personal use only; all rights reserved.
Fixed-bed experiments were conducted with one layer ACFC Langmuir equation is valid for homogeneous surface; however,
共11.4 cm⫻ 11.4 cm, 1.4 g兲 共Calgon Corporation, Type FM10兲 for activated carbon is extremely heterogeneous. Freundlich equation
model calibration and verification. The properties of ACFC are is valid for heterogeneous and can describe equilibrium better for
given in Table 1. Toluene was selected for kinetic experiments adsorption on ACFC.
and modeling since toluene is one of the most commonly en- Therefore, the isotherm parameters for the adsorption-
countered VOCs in indoor environment and exhibit relatively regeneration processes were determined by fitting the isotherm
high concentrations 共Scholz and Santl 1999; Lee et al. 2001, experimental data to the Freundlich isotherm equation
2002兲. Fig. 1 shows the experimental schematic diagram. A high
degree of reproducibility of the concentration profiles for adsorp- q共r,t兲 = KiC p,i共r,t兲1/ni 共22兲
tion and regeneration processes was demonstrated by repeating
adsorption-regeneration experiments on other swaths of the same
ACFC. Film Transfer
The continuous adsorption-desorption cycling was conducted Currently no correlation exists for determining film transfer coef-
with the inlet toluene concentration of around 100 ppb to demon- ficient for short bed adsorber. The film transfer coefficients for
strate the functionality of the process in an uninterrupted con- adsorption are obtained by fitting model to experimental data.
trolled scenario. An 80 L Tedlar bag was filled with high purity Regeneration processes have a Reynolds number of 0.045 due
air provided by a Whatman Zero Air Generator 共Whatman Inc., to the low flow rate of 0.5 L/min. No empirical correlation is
Mass.兲 and injected with a known amount of liquid toluene. available to calculate the film transfer coefficient for such condi-
The resulting air stream containing a high concentration toluene tions with low Reynolds numbers. The correlation of Petrovic and
was pumped to mix with a purified air stream, creating an Thodos is used to determine the film transfer coefficient during
adsorbate-laden air stream with a constant concentration of desorption. This correlation is suitable for fixed-bed reactors and
around 100 ppbv. Since water vapor impacts on adsorption were low Reynolds number region 共3 ⬍ R ⬍ 2 , 000兲
not of a major interest of this study, the relative humidity was
controlled at 20–22% for all the adsorption. The adsorption ex- 0.357 Dg
kf = 共R兲0.64共Sc兲0.33 共23兲
periments were conducted at the superficial velocities ranging dp
from 9.6 to 18.3 m/min, which are used by most indoor air treat-
ment systems that employ adsorption operate adsorption devices. Surface Diffusivity
Therefore, adsorption experiments were performed at two indoor Diffusivity of VOCs along the surface of ACFCs is a function of
air flow rates 共92 and 184 L/min兲. Influent and effluent concen- temperature, concentration, and pore size distribution and it in-
trations were measured by a gas chromatography with automatic creases with increasing adsorbate concentration 共Lordgooei et al.
sampler 共PE Photovac, Waltham, Mass.兲. The adsorption curve 1998; Lordgooei et al. 2001兲. Currently no correlations have been
was obtained by monitoring effluent concentration as a function developed to determine surface diffusivity by accounting for tem-
of time. perature, concentration, and properties of ACFC adsorbents. The
The ACFC was regenerated thermally by passing alternating surface diffusion coefficient, Ds 共cm2 / s兲, can be obtained by re-
electrical current through the fabric once the 23 h adsorption was lating the surface diffusion flux to the pore diffusion flux by a
completed. The electrical current was provided by Powerstat vari- correlation factor, the surface to pore diffusion flux ratio
able autotransformer 共Type 3PN116C, The Superior Electric, 共SPDFR兲. It results in the following correlation which is used to
Co.兲. True RMS Multimeter 共Fluke, Inc.兲 was used to measure calculate the surface diffusion coefficient 共Sontheimer et al.
voltage during applying Joule Effect. Thermocouples 共Omega 1988兲:
Inc., Type I兲 were positioned on the surface of the ACFC to
DgpC0
Ds = ⫻ SPDFR 共24兲
aKC1/n
0
Gas Chromatography
Inlet Flow Sampling Port
The DFPSDM model was fit to the experimental data to deter-
mine the SPDFR for both adsorption and regeneration processes.
Tedlar Bags Pumps
Adsorber
persion coefficient when they developed a model to simulate profile and model prediction for limonene at the flow rate of 45
VOCs adsorption on ACFCs, in which effective diffusivity and L/min, respectively. The predicted profile for limonene demon-
dispersion were incorporated strates good closure with the experimental data. Deviations from
Downloaded from ascelibrary.org by MANIPAL INST OF TECHNOLOGY on 07/19/20. Copyright ASCE. For personal use only; all rights reserved.
sensitive to film transfer and the tail of adsorption profile is sen- 0.4
sitive to surface diffusion, it is feasible to fit two parameters at the 0.3
same time. The influence of pore and surface diffusivities can be 0.2
neglected for the adsorption at low flow rate of 45 L/min and the 0.1
temperature of 20° C. As expected, the model is not sensitive to 0.0
pore diffusion at the three flow rates. 0 5 10 15 20 25
The sensitivity analysis of regeneration process to adsorption Time (hour)
capacity was also conducted. The initial desorbed concentration Adsorption Flow Rate =45L/min
increases as the adsorption capacity decreases, which is consistent
with the experiment observation. The sensitivity of regeneration Fig. 2. Model simulation of adsorption breakthrough profiles for
model to the surface diffusion exhibits the same pattern as the toluene and limonene
adsorption model. The tailing of the concentration profiles shifts
up with the increase of surface diffusivities. The impacts of pore
1.00
diffusivity and film transfer are negligible compared to the two
0.95
factors mentioned above.
0.90
0.85
0.75
0.70 Cycle#1-Experimental Data
Adsorption Model Validation for Single Run at a Flow
0.65 Cycle#2-Experimental Data
Rate of 45 L/min
0.60 Cycle#1-Model simulation
Model was fit to the adsorption profile 共C0 / C兲 for toluene using 0.55 Cycle#2-Model Simulation
least-squares fitting. The fitted model was used to predict column 0.50
data for limonene. The limonene Freundlich isotherm parameters 0 5 10 15 20 25
were predicted by the D-R equation using toluene as a reference Time (hour)
compound 共Cal et al. 1994; Das et al. 2004; Crittenden et al. Adsorption Flow Rate =184 L/min
2004兲. The isotherm parameters and the physical properties for
toluene and limonene are shown in Table 2. Fig. 3. Comparison of model fitted data and experimental data for
Fig. 2 shows the model calibration for toluene breakthrough adsorption at flow rate of 184 L/min
0.50
0 5 10 15 20 25
Time (hour)
Adsorption Flow Rate =184 L/min
Fig. 5. Model simulations and experimental data for continuous de-
Fig. 4. Comparison of model predictions and experimental data for sorption 共loading flow rate= 184 L / min兲
adsorption at the flow rate of 184 L/min
face diffusivity was constant for those cycles on the used carbon
cloth. Fig. 4 shows model predictions of the adsorption profiles The diffusion modulus reflects the same results. As Eds increases,
for 3rd and 4th cycles along with the experimentally measured the mass transport by advection becomes controlling. As Eds de-
adsorption profiles, which confirms the validity of the model. The creases, the intraparticle phase mass transport rate becomes con-
concentration profiles demonstrate a similar trend, but the profiles trolling at the flow rate of 184 L/min.
slightly shift up as more adsorption/desorption cycles were per-
formed. This is because more ACFC surface was covered as more
toluene passed through the fixed bed. The available adsorption Desorption Model Validation
capacity of ACFC reduced when increasing the amount of adsorp-
Fig. 5 shows model simulations for regenerations conducted at
tion from zero coverage to complete coverage.
the temperature of 150⫾ 5 ° C after the adsorption process was
Model simulations find that the driving forces from the gas-
phase to the exterior surface of the adsorbent particle become performed at the flow rate of 184 L/min. The model simulations
greater when the bed contact time is longer. During the continu- were fit to the experimental data for Cycles #1 and #2. The pre-
ous adsorption cycles at various flow rates, both gas diffusivities dicted profiles for Cycles #3 and #4 demonstrate a similar trend of
and film transfer coefficients show an increasing trend as the flow the measured curves, but the discrepancy between the model
rates decrease, which agrees with the observation that the 23 h simulation and experimental measurements were observed. The
removal efficiency is higher at lower flow rate. Table 3 presents deviations from the experimental data can be explained by the
the dimensionless groups for a better understanding of controlling following reasons. The desorbed concentrations were most sensi-
mass transfer mechanism for adsorption onto the ACFC. For tive to the adsorption capacity when desorption was initiated.
large Biot numbers 共greater than 30兲, the intrapaticle-phase mass During experimentation voltages were manually adjusted at the
transfer is much slower than the gas-phase rate and therefore beginning of the regeneration to achieve the temperature of inter-
controls the rate of adsorption. For small Biot numbers 共less est, hence voltages cannot be increased at the uniform rates for
than 0.5兲 the gas-phase rate controls over the intraparticle rate each single experiment. In fact each experiment had a different
共Sontheimer et al. 1988兲. The increasing Biot numbers with the rate of temperature increase and the initial increased temperatures
increases in gas flow rates shows the importance of the intrapar- varied within the range of 共1 ⫾ 6.5%兲 of 150° C due to the varia-
ticle phase mass transfer rate as the controlling factor for the tions of the applied voltages. The assumption of the constant sur-
adsorption at the higher flow rate. The value of Biot number face diffusivity for each cycle is not valid because the variations
which is only 0.44 共less than 0.5兲 at the flow rate of 45 L/min in solid-phase concentrations on the ACFC influence the surface
indicates the gas-phase rate controls over the intraparticle rate. diffusivity. The model was run again to fit the four cycles. Fig. 6
shows the best-fitted desorption profiles for the four cycles and
Table 4 presents the fitted kinetic parameters for these cycles.
Table 3. Dimensionless Groups for Adsorption Processes at Various
Loading Flow Rates
After the first-time heating of the ACFC, the fitted adsorption
capacities for Cycles #2–#4 vary within 10% and the surface
Flow rate 共L/min兲 diffusivity vary within 35%.
Dimensionless
group 184 92 45 Two regeneration cycles conducted at the temperatures of 180
and 200° C were fitted and the results are presented in Fig. 7.
Re 32.4 16.2 7.9
Regeneration is more favorable at the higher regeneration tem-
St 0.20 0.23 1.10
perature in that the adsorption potential is lower. Table 5 summa-
Bis 26.4 4.85 0.44
rizes the dimensionless groups for the regenerations. The Biot
Eds 0.006 0.057 2.494
numbers ranging from 246.4 at the temperature of 180° C to
Edp 0.73⫻ 10−7 0.15⫻ 10−7 0.30⫻ 10−6
781.9 at 200° C indicate that the intraparticle mass transfer is the
Sc 1.78 1.78 1.78
dominant factor from the kinetic perspective.
Model Simulation-Cycle #2
Model Simulation-Cycle#3 R 0.045 0.045 0.045
160.5 Model Simulation-Cycle#4
St 156.88 156.88 156.88
120.5 Bis 409.9 781.91 246.42
Eds 0.3827 0.20 0.64
80.5
Sc 1.82 1.82 1.82
Downloaded from ascelibrary.org by MANIPAL INST OF TECHNOLOGY on 07/19/20. Copyright ASCE. For personal use only; all rights reserved.
40.5
0.5
flow rates increase. With regard to regeneration, the desorbed
0.0 0.2 0.4 0.6 0.8 1.0
Time (hour)
concentration profiles are most sensitive to the adsorption iso-
Regeneration Flow Rate =0.5L/min therm at the temperature of interest, especially as desorption was
initiated. While the tailing of desorption profile shifts up with the
Fig. 6. Best-fitted model results and experimental data for continu- increase of surface diffusivity. Both adsorption and regeneration
ous desorption 共loading flow rate= 184 L / min兲 processes are not sensitive to dispersion coefficients.
Conclusions Discussions
An adsorption model coupled with a regeneration model was de- Currently no empirical correlations give satisfactory estimation
veloped to describe the mechanistic processes of using an ACFC for the surface diffusivity on the short ACFC bed. Surface diffu-
adsorber to remove VOCs from indoor air. Model simulations sivity is found to be an important kinetic parameter for both the
were conducted to predict the concentration profiles of a continu- adsorption and regeneration processes. The model is able to pre-
ous adsorption/regeneration cycling on the ACFC system. Good dict the adsorption process using the constant surface diffusivity
agreement between model simulations and experimental data was throughout the continuous cycles. However, the model predic-
obtained and the kinetic properties of the adsorption/regeneration tions using the constant surface diffusivity had deviations from
process on ACFC were characterized. For adsorption process, the the measured desorbed concentrations. The deviations are caused
film transfer is the dominant factor for mass transfer at low flow by complicated influencing factors, such as the possible modifi-
rates, and the intraparticle rate controls over gas-phase rate as the cations of the ACFC properties after heating, and temperature
variations; especially the dependency of surface diffusivity on the
solid-phase concentrations is a main reason. The deviations are
Table 4. Best-Fitted Surface Diffusion Coefficients and Adsorption caused by complicated influencing factors, such as the possible
Potentials for Regeneration modifications of the ACFC properties after heating, and tempera-
Cycle Ds ture variations; especially the dependency of surface diffusivity
number 共cm2 / s兲 K 关共g / g兲共L / g兲1/n兴 on the solid-phase concentrations is a main reason. This needs to
be validated through additional experimentations and character-
1 1.00⫻ 10−10 198.3
izations and consequently the correlation for surface diffusivity
2 6.50⫻ 10−10 119.0
can be obtained.
3 1.00⫻ 10−9 112.0
4 1.20⫻ 10−9 111.0
Note: Loading flow rate= 184 L / min, temperature= 150⫾ 10° C.
Notation
180
Regeneration Concentration (ppm)