&EPA
United States Office of Air Quality EPA-450/3-88-012
Environmental Protection Planning and Standards June 1988
Agency Research Triangle Park NC 27711
Carbon Adsorption
for Control of
VOC Emissions:
Theory and Full Scale
System Performance
-------
Carbon Adsorption for Control of VOC Emissions:
Theory and Full Scale System Performance
Emission Standards Division
PROPERTY OF
EPA LIBRARY, RTP.NC
U. S. ENVIRONMENTAL PROTECTION AGENCY
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, NC 27711
August 1988
-------
TABLE OF CONTENTS
Section Page
1 INTRODUCTION 1-1
2 SUMMARY AND CONCLUSIONS 2-1
3 THEORY AND PERFORMANCE OF CONVENTIONAL FIXED BED
CARBON ADSORBERS 3-1
3.1 Mechanism of Adsorption and Desorption 3-1
3.2 Full Scale Adsorption Systems 3-13
'3.2.1 System Overview 3-13
3.2.2 Full Scale System Design Considerations 3-15
3.3 Carbon Adsorber Long- and Short-Term Efficiency 3-22
3.3.1 Calculation of Carbon Adsorber Efficiency 3-23
3.3.2 Variability of Short-Term Removal Efficiency 3-25
3.3.3 Relationship of Outlet Concentration and
Efficiency 3-25
3.4 Effect of Operating Variables on Adsorber Performance 3-29
3.4.1 Temperature 3-29
3.4.2 Concentration 3-33
3.4.3 Humidity 3-34
3.4.4 Volumetric Flowrate 3-38
3.4.5 Bed Fouling 3-41
3.4.6 Channeling 3-43
3.5 Deliberate Changes from Initial Design
Operating Conditions 3-43
3.5.1 Adsorbate 3-45
3.5.2 Steaming Conditions 3-45
3.6 Performance Information on Industrial Adsorbers 3-46
3.6.1 Data Sources 3-47
3.6.2 Removal Efficiency Data for Performance Test 3-52
3.6.3 Continuous Removal Efficiency Data 3-57
3.7 Conclusions Regarding Carbon Adsorber Performance 3-60
11
-------
TABLE OF CONTENTS, Continued
Section Page
4 CARBON ADSORPTION SYSTEM AT COMMENTER'S FACILITY 4-1
5 REFERENCES 5-1
-------
LIST OF TABLES
Tables Page
3-1 Reported Bed Lives for Various Solvent Blends 3-44
3-2 Performance Test Data for Carbon Adsorption Systems 3-48
3-3 Performance Test Data for Carbon Adsorption Systems on a
Per Bed Basis 3-50
IV
-------
LIST OF FIGURES
Figure Page
.3-1 Representation of Carbon Pellets/Particles in Carbon Bed 3-3
3-2 Representation of Pores in Activated Carbon Particle 3-4
3-3 Mechanism of Adsorption 3-6
3-4 Simp! ified Representation of Carbon Capacity 3-8
3-5 Available Working Capacity as a Function of Distance
Through Bed for an Operating Bed 3-9
3-6 Simplified Pore Representation of Capacity as a Function
of Distance Through Bed 3-11
3-7 Vapor stream Concentration as a Function of Distance
Through Bed 3-12
3-8 Carbon Adsorber System Process Flow Diagram 3-14
3-9 Steam Consumption Versus Working Capacities 3-18
3-10 Adsorbate Concentration in Bed After Steaming Countercurrently
as a Function of Distance Through Bed 3-19
3-11 Adsorption/Desorption Cycles in a 2 Bed System 3-21
3-12 Determination of Carbon Adsorber Removal Efficiency 3-24
3-13 Typical Carbon Adsorption Breakthrough Curve 3-26
3-14 Carbon Capacity vs. Temperature at Constant Pressure 3-31
3-15 Carbon Adsorber Removal Efficiency with Varying Bed
Temperature for a Complete Adsorption Cycle 3-32
3-16 GTR Test Number 6 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency 3-35
3-17 GTR Test Number 5 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency 3-36
3-18 Effect of Relative Humidity on Working Capacity 3-37
3-19 Effect of Variation in Volumetric Flowrate on the Shape
of the Breakthrough Curve 3-39
-------
LIST OF FIGURES, Continued
Figure Page
3-20 Carbon Adsorber Removal Efficiency with Varying Flowrate
for a Complete Adsorption Cycle 3-40
3-21 Carbon Adsorber Removal Efficiency as a Function of Bed
Age for a Ketone Containing System 3-42
3-22 Inlet/Outlet Concentration Curve for Test 3 3-54
3-23 GTR Test Number 5 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency 3-58
3-24 GTR Test Number 6 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency 3-59
3-25a Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #1 3-61
3-25b Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #1 3-61
3-25c Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #2 3-62
3-25d Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #2 3-62
3-25e Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #3 3-63
3-25f Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #3 3-63
3-25g Inlet and Outlet Concentration Versus Cycle Time for
Adsorber f 4 3-64
3-25H Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #4 3-64
3-25i Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #5 3-65
3-25J Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #5 3-65
vi
-------
LIST OF FIGURES, Continued
Figure Page
3-25k Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #6 3-66
3-251 Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #6 3-&6
vii
-------
1. INTRODUCTION
This report presents the results of an investigation into the performance
and operation of vapor phase carbon adsorption systems. This investigation
was initiated as a result of comments received by the U. S. Environmental
Protection Agency (EPA). These comments were in reference to the draft new
source performance standards (NSPS) for control of VOC emissions from the
manufacture of magnetic tape. The commenter challenged EPA's supporting data
for the proposed performance requirements for carbon adsorption systems and
the costs for operating and maintaining carbon adsorption systems at the
required performance level. Specifically, the commenter contended that the
95 percent efficiency requirement is not achievable on a continuous basis due
to the inherent variability of carbon adsorption systems. They also stated
that ketones, which are commonly used solvents in magnetic tape manufacture,
reduce the performance of carbon adsorbers and increase system variability and
shorten bed life which increases the cost of using carbon adsorption.
In order to respond to these comments, the EPA requested additional
information from manufacturers and users of carbon adsorber systems to further
investigate system performance and costs. The EPA also again reviewed
information obtained from previous studies by the Agency. This report
summarizes the results of this study.
This report is organized as follows. Section 2 presents the conclusions
of this study. Section 3 presents a description of the vapor phase adsorption
process, discusses impacts of changes in inlet vent stream characteristics on
adsorber performance, and presents supporting test data. Section 4 presents a
description of the carbon adsorber system which the commenter used as a basis
for developing their comments, and a discussion of the design and operation of
that system.
1-1
-------
2. SUMMARY AND CONCLUSIONS
This report was specifically designed to address comments submitted
concerning the proposed new source performance standards for magnetic tape
manufacture. These comments may be summarized as follows:
t One commenter questioned whether any supporting data exist for the
Agency's position that 95 percent VOC removal efficiency can be
achieved continuously by carbon adsorption systems over all
averaging periods, including short-term periods. The commenter
submitted information indicating that 24-hour averages of efficiency
of his adsorption system vary dramatically from day to day. Days
when the average efficiency was above 95 percent were followed
quickly by days with an average efficiency of less than 90 percent.
It was this information that caused the commenter to question the
Agency's decision to determine compliance and assess adsorber
operation and maintenance based on short-term measurements of
adsorber performance.
The same commenter also stated that the evaluation of carbon
adsorption to control VOC emissions has not adequately addressed the
problems associated with the use of ketones by the magnetic tape
industry. The commenter submitted data that, in the commenter's
opinion, demonstrated reduced adsorber efficiency caused by the use
of ketones. The commenter also implied that the variability in
carbon adsorber performance is greater when ketones are present in
the solvent laden air stream and that ketones shorten the useful
life of the carbon in adsorption systems, resulting in greater cost
impacts attributable to the NSPS than indicated by the cost analysis
carried out by EPA prior to proposal.
In order to address these comments, information was requested from a
number of sources. These included both magnetic tape manufacturers and other
types of coating operations using carbon adsorbers to control VOC emissions.
A meeting was also held with representatives of a major supplier of activated
carbon to obtain their perspective on proper adsorber system design and
operation based on their long term experience in this field. In addition, two
sites were visited to obtain first hand information on the operation of carbon
adsorber systems. Emission test data from 15 tests performed for the Emission
Standards Division of EPA and the Office of Research and Development were also
reviewed to provide additional substantiation of adsorber performance and to
attempt to compare long- and short-term removal efficiency.
2-1
-------
Based on the data collected for this study, the following conclusions
have been made.
t When the carbon adsorber system is properly designed and operated,
the adsorption/desorption process is predictable and reproducible
from cycle to cycle.
0 For well designed and operated carbon adsorber systems, continuous
VOC removal efficiencies of over 95 percent are achievable over
long- and short-term periods for a variety of solvents, including
mixtures that contain ketones such as methyl ethyl ketone (MEK) and
cyclohexanone. Several plants have been shown to continuously
achieve removal efficiencies of 97-99 percent.
t All facilities identified by this study which had removal
efficiencies below 95 percent had identifiable operational problems
which contributed to their poor performance. All of the operational
problems identified were correctable.
A carbon adsorption system must be designed based on the following
process parameters: 1) particular solvent or solvent blend being
recovered, 2) solvent load, 3) vent stream flowrate, and 4) vent
stream temperature. Assuming the initial design provides sufficient
capacity to account for normal daily process variations and the
adsorber is properly operated, the performance of a carbon adsorber
system will be essentially constant from cycle to cycle, and
long-term and short-term efficiency will be the same.
t If a carbon adsorber bed is left on-line after breakthrough,
adsorber efficiency will be significantly reduced and may also
become much more variable. A key to maintaining high continuous
removal efficiencies is to detect breakthrough and bring a fresh bed
on-line. There should always be a fresh bed available if the system
is properly designed and operated.
Because continuous high removal efficiency can be achieved by a
properly designed and operated carbon adsorber system, short-term
performance testing and monitoring requirements are appropriate as
long as the complete system cycle is included in the test or
monitoring period.
A consensus of carbon suppliers, carbon adsorption system vendors,
and carbon adsorption system operators indicates that when
cyclohexanone is adsorbed, it exothermally reacts on the carbon
surface to form higher molecular weight products which cannot be
removed by normal steam desorption. The subsequent build up of
these compounds results in a steady decrease in the adsorptive
capacity of the carbon. This loss in adsorptive capacity decreases
the time which a carbon bed can remain on-line before breakthrough
occurs. When the adsorption cycle time approaches the time required
2-2
-------
for off-line bed regeneration and proper cool down, the carbon must
be replaced. If the carbon is not replaced the point will be
reached where the operator will be forced to either: 1) leave the
bed on-line past breakthrough in order to properly regenerate the
off-line bed or 2) switch to the regenerating bed before it has been
adequately steamed and cooled. The result in either case will be a
dramatic drop in removal efficiency.
The carbon adsorber system of which the commenter reported as having
highly variable and frequently low removal efficiencies is
significantly under-designed for the actual solvent loading it is
required to control. This results in the system being operated a
significant portion of the time after breakthrough has occurred. As
a result, the efficiency of this system is extremely sensitive to
variations in the process conditions and therefore exhibits
significant variations in efficiency from day-to-day and
cycle to cycle. The variations also result in significantly reduced
long-term efficiency. If the system was operated within the design
limits, then the reduced efficiency and efficiency variation should
not occur.
As a carbon bed ages and its total adsorptive capacity gradually
decreases due to fouling, the working capacity can in some cases be
maintained at the desired level by increasing the steam flow during
desorption. This will increase steam costs. The decision of
whether to use higher steam flow or replace the carbon is based on
the cost of additional steam versus the cost of new carbon.
A key parameter to maintaining continuously high removal efficiency
is replacing the carbon well before fouling reduces the adsorptive
capcity. More frequent carbon replacement results in higher
annualized carbon costs, but also prevents reaching the point where
the adsorber performance falls below the design value.
2-3
-------
3. THEORY AND PERFORMANCE OF CONVENTIONAL FIXED BED CARBON ADSORBERS
This section presents a general description of carbon adsorbers used to
remove volatile organic compounds (VOC) from a gaseous process stream. The
purpose of this presentation is to provide information necessary to draw
conclusions concerning the following:
The ability of carbon adsorbers to remove 95 percent or more of the
VOC in a process stream on both a short-term and long-term basis;
the effect of various operating parameters such as solvent type, bed
age, inlet stream temperature, concentration, flowrate, and
regeneration steam flow on adsorber performance.
The information presented here was obtained from industry and vendor responses
to information requests from EPA, evaluation of carbon adsorber emission test
data gathered for this and previous EPA studies, and a meeting with a major
vendor of activated carbon and carbon adsorption/incineration systems.
The first part of this section is a discussion of the basic theory of
carbon adsorption (Section 3.1). Section 3.2 describes the design
considerations for full scale systems, Section 3.3 presents a discussion of
how adsorber efficiency is calculated, and Section 3.4 presents the impact of
operating variables on system performance. Section 3.5 presents emission test
data to substantiate the performance of carbon adsorbers. Finally Section 3.5
presents conclusions concerning carbon adsorber performance.
3.1 MECHANISM OF ADSORPTION AND DESORPTION
This section presents a detailed description of the mechanisms of carbon
adsorption and desorption. To describe the mechanisms involved, a simplified
approach using a single bed of activated carbon is developed. The principles
involved in the single bed system are then applied to describe the operation
of a typical carbon adsorber system.
For gas phase carbon adsorption applications, the adsorber system does
not actually recover the VOC. It is used to transfer the VOC from a medium
where it is difficult to recover (the vent stream gas), to a more
concentrated form in a different medium (usually steam) where the VOC can be
more easily recovered. This transfer occurs in two steps. The first is the
3-1
-------
adsorption step where the VOC (adsorbate) is adsorbed onto the surface of
activated carbon (adsorbent). The second step is where the adsorbate is
removed from the carbon (desorption) and recovered for reuse. Both of these
steps are equally important in the overall process.
The adsorption process can be either physical or chemical. In physical
adsorption the organic molecule is held to the surface by weak van der Waal
type forces or intermolecular cohesion. The chemical nature of the adsorbed
gas remains unchanged, thus the process is readily reversible. Regenerative
pollution control equipment requires the adsorption process be physical. In
chemical adsorption, electrons are exchanged thus chemically bonding the
molecule to the surface of the carbon particle. Chemical adsorption is not
readily reversible and, therefore, is not suitable for the regenerative
adsorber systems used in air pollution control applications.
Figure 3-1 presents a series of exploded views which describe the
subsystems which make up a carbon adsorber bed. A carbon bed is comprised of
carbon pellets. The pellets are made up of carbon particles which have been
sintered together. The carbon used in adsorption is made by a two step
process. In the first, material from various sources such as coconut shells,
petroleum products, wood and coal is carbonized by heating it in the absence
of air until all organic compounds except the carbon are volatilized. Then
using high temperature steam, air, or carbon dioxide, the carbon is made
4
porous or activated. Depending upon the extent of this process and the
original source, the carbon can be made to fit the use for which it is
desired.
The pore structure within a carbon particle is illustrated in Figure 3-2.
The external surface area of a carbon particle is a few square meters per
gram; however, within the pores the available surface area is hundreds of
square meters per gram.
The pores within the carbon are classified according to their size.
Large pores (greater than 2,000 nanometers in diameter) are called macropores
and smaller pores (less than 200 nanometers) are called micropores.6 Pores
with diameters between these ranges are called transitional pores.7
Micropores are where the majority of the adsorption occurs so it is desirable
Q
to have a large amount of the pore space in this form.
3-2
-------
Carbon Pellets in
Cross Section of Carbon Bed
;'-V;V^r^-«,7A^\--ll-Xr?^^
:Vx^-V>-^v\v:^y^
's'^-r.^-'l:-'^^^^^^
,\>." \\''.\l ,-> _> ,/,- \ "v / * 7 ; VN ' T' J-/X r-' N '-*,' " i-v/OC-' -V'V.'» ',s -I »' IN -V,- >
^^.'-.-T'-'X'-^tL.s'.'i7v"%lr:V'^Mi'r'^:,cx<'>.~.7^^V-7!'4-\v.-*kL
Carbon Particle
cr
o
Figure 3-1. Representation of Carbon Pellets/Particles in Carbon Bed
3-3
-------
Adsorbale
Adsorbent
A/1
Figure 3-2. Representation of Pores in Activated Carbon Particle
DC
O
-------
The adsorption process begins with the mechanical movement of the vent
stream through the carbon bed, which brings the organic molecules into contact
with the carbon pellets. The remainder of the adsorption process consists of
three steps as illustrated in Figure 3-3. The adsorbate must first diffuse
into the carbon pellet to the surface of the carbon particle. Next, the
adsorbate molecule must diffuse from the surface into the pores within the
carbon particle. The extent of the diffusion within the pores is dependent on
the size of the molecules and the pore structure of the carbon. Diffusion
into the larger pores occurs fairly rapidly, but as the pore diameters become
smaller the diffusing molecule strikes the walls and sticks for short periods
g
of time. This diffusion process continues until the molecule reaches a
location where it no longer has sufficient energy to escape the forces which
hold it to the pore wall. This usually occurs where the pore diameter is not
more than approximately twice the diameter of the adsorbate molecule.
The adsorption process continues until the amount of adsorbate on the
carbon reaches a thermodynamic equilibrium with the adsorbate in the gas
phase. The thermodynamic equilibrium is a function of the carbon type,
temperature of the carbon and adsorbate, and the adsorbate partial pressure
(concentration) in the vent stream. The amount of adsorbate a particular
carbon can hold is called the equilibrium capacity.
As previously mentioned, the purpose of the carbon adsorber is to
actually transfer the adsorbate from the gas stream to a medium where it can
more easily be recovered or disposed of. Therefore, at some point the
adsorbate must be removed from the carbon. This process is called desorption
or regeneration. Desorption is accomplished by shifting the thermodynamic
equilibrium established during the adsorption step. There are three ways to
shift the equilibrium: 1) increase the temperature, which is usually brought
about by the addition of steam, 2) reduce the pressure of the atmosphere
surrounding the carbon, and 3) reduce the concentration in the gas stream
outside the carbon to a value less than the concentration inside the carbon.
In most air pollution control applications, increasing the temperature is used
for desorption.
During desorption some adsorbed molecules are not removed. The reason is
that to remove all the adsorbate requires sufficient'time for the adsorbate
3-5
-------
Step 1: diffusion to
adsorbent surface
Step 2: migration into
pores of adsorbent
Step 3: monolayer
buildup of adsorbate
Adsorbate molecules
CO
Ot
cc
cu
Figure 3-3. Mechanism of Adsorption
o
oo
CO
-------
molecules to diffuse out of the carbon particle, and that the temperature be
high enough to cause all the adsorbate to desorb. However, the energy cost to
accomplish this is higher than the cost to leave some adsorbate in the carbon
and use a larger amount of carbon to achieve the desired system performance.
Adsorbate remaining in the carbon after desorption is called the heel.
The amount of heel is a function of the desorption time and temperature.
Increasing adsorption time and/or temperature will reduce the heel. In virgin
carbon, a stable heel is established after two to three adsorption/ desorption
cycles.
The process discussed above is summarized by the simplified
representation of a carbon pore shown in Figure 3-4. As shown, the pore has
three different volumes: the equilibrium capacity, working capacity, and the
12
heel. As previously discussed, the equilibrium capacity is a function of
the carbon type, bed temperature, and the partial pressure of the adsorbate in
the vent stream. It represents the maximum amount of adsorbate which can be
adsorbed by the carbon when it is at equilibrium with the surrounding
conditions. The heel represents the adsorbate which remains in the pore after
desorption. It is a function of the particular carbon, the adsorbates in the
vent stream, and the steaming conditions.
The practical application of the adsorption process to a full size carbon
bed is illustrated in Figure 3-5. In this figure, the solvent laden air (SLA)
flows from left to right. As shown, there are three zones in the bed labeled
saturated, mass transfer, and fresh. The saturated zone is located at the
entrance to the bed and represents the carbon which has already adsorbed its
working capacity of adsorbate. The saturated carbon is at thermodynamic
equilibrium with the incoming vent stream. Therefore, no net mass transfer
occurs in this zone. The mass transfer zone (MTZ) is the section of the
carbon bed where the adsorbate is removed from the carrier stream. The carbon
in this zone is at various degrees of saturation, but is still able to adsorb
some adsorbate. For a typical system, the mass transfer occurs within a
section approximately three inches in depth. The fresh zone is downstream
of the mass transfer zone and represents the region of the bed where no new
adsorbate has passed since the last regeneration. This zone still has all its
working capacity (i.e., equilibrium capacity minus the heel) available.
3-7
-------
A = Residual or Heel (located deep in the pore and difficult to dislodge)
B = Working Capacity (bounded by the heel remaining from previous cycles and portion of pore
too large in diameter to retain organic)
C = Equilibrium Capacity
Figure 3-4. Simplified Representation of Carbon Capacity12
3-8
-------
SLA Flow
100%
% Of Total Equllbrlum
Capacity Available
UJ
(O
"
Saturated -»
Zone
' .^
/
Mass
Transfer Zone
Fresh
Zone
Heel
Remaining
Working
Capacity
Distance Through Bed
Figure 3-5. Available Working Capacity as a Function of Distance
Through Bed for an Operating Bed
8
-------
During operation the mass transfer zone moves down the bed in the direction of
flow. Breakthrough occurs when the mass transfer zone first reaches the bed
outlet. The breakthrough point is characterized by the beginning of a sharp
increase in the outlet concentration. The available adsorption time for a
specific bed before breakthrough occurs is a function of the amount of carbon
present, its working capacity, and the concentration and mass flowrate of
adsorbate.
Figure 3-6 shows a simplified representation of the carbon pores in each
of the three zones. Pores A through E represent typical pores at different
locations in the bed. Pore A, which is at the front of the bed, is already
completely saturated while Pore E, which has not been exposed to adsorbate
during this cycle, still retains its entire working capacity. Pores B, C, and
D which are located in the MTZ, depict various degrees of saturation. A cross
section of the bed perpendicular to the air flow will reveal pores at similar
levels of saturation. In the MTZ, pore B has been exposed to the adsorbate
for the longest period of time and is nearly saturated while the MTZ has just
reached pore D which still retains most of its adsorptive capacity. As the
adsorption cycle continues and more adsorbate enters the bed, the mass
transfer zone will continue to move through the bed.
Figure 3-7 depicts the VOC concentration within the carrier gas as a
function of axial distance down the bed. Since equilibrium has been reached
with the incoming adsorbate in the region prior to the mass transfer zone, the
vapor stream concentration is equal to the inlet concentration. Within the
mass transfer zone, the concentration of the vapor stream drops off because
the organic is being adsorbed into the pores.
Theoretically, the concentration in the third zone should be zero.14
However, a small amount of adsorbate is typically present. This is a result
of two factors:
1. A small amount of SLA may pass through the adsorber without actually
contacting the carbon.
2. Due to the low concentration of adsorbate in the vent stream in the
last few inches of the bed, the heel remaining from the previous
cycle will slowly desorb.
3-10
-------
SLA Flow
Saturated Zone
A
Saturated Zone
MTZ
BCD
Fresh Zone
E
X
V
MTZ
B
almost all
working capacity
consumed
Fresh Zone
working
capacity
consumed
\
half of working
capacity consumed
* all working
capacity available
almost all working
capacity available
Figure 3-6. Simplified Pore Representation of Capacity as a Function of
Distance Through Bed
3-11
-------
SLA Flow i
Inlet Concentration
ro
Vapor Stream
Concentration
(ppm)
V
~\
\
Outlet
Concentration
1 1 *
-] Mass U- Exit
Transfer Zone Concentration
Distance Through Bed
Figure 3-7. Vapor Stream Concentration as a Function of
Distance Through Bed
(C
8
-------
Test results on full scale systems have shown outlet concentrations as low
as 0.5 ppm. This outlet concentration can be minimized by proper system
design, as discussed in the next section.
The breakthrough curve, which is the outlet concentration as a function
of time, is a mirror image of the concentration profile in the mass transfer
zone. As the mass transfer zone reaches the end of the bed, the outlet
concentration rises. This will continue until the outlet equals the inlet
concentration.
3.2 FULL SCALE ADSORPTION SYSTEMS
This section describes full scale adsorption system design and
operation. The basic mechanisms were previously described in Section 3.1.
Section 3.2.1 presents an overview of the adsorber system. Section 3.2.2
discusses specific design considerations for a full scale system.
3.2.1 System Overview
The process flow diagram for a typical two bed carbon adsorber system
is shown in Figure 3-8. The adsorber system can be broken down into three
separate sections; pretreatment, carbon adsorber, and recovery/waste
treatment. The vent stream containing the adsorbate enters the adsorption
system via the pretreatment section. If the vent stream is above the
maximum design temperature it is reduced within the pretreatment section,
usually with a heat exchanger. In addition, a filter is included in the
pretreatment section to remove any particulate present in the vent stream.
From the pretreatment section, the vent stream enters the adsorber.
Figure 3-8 depicts a two bed adsorber system. In order to provide
continuous emission control, at least two adsorber beds are needed so that
one is on-line while the other is regenerated. Adsorber systems with three or
more beds are operated similarly. During operation, the organic-laden vent
stream passes through the on-line bed for a predetermined time period or
until breakthrough occurs. The on-line bed is then taken off-line for
regeneration (desorption) and the other bed is brought on-line.
Regeneration of the off-line bed is usually accomplished by passing
steam through the bed countercurrent to the direction of vent stream flow.
The steam which is injected into the bed serves several purposes; 1) it
3-13
-------
Vent Stream
Filler Blower
Heat Exchanger
(optional)
Ambient
Air Intake
(or Cooling/Drying
(H
Filter Blower
Steam
Carbon
Absorber
I
Carbon
Absorber
Exhaust
Vent
I
Aqueous Phase to
Disposal or
Treatment
Decanter
Condenser
Organic Phase to Recovery
Figure 3-8. Carbon Adsorber System Process Flow Diagram
-------
provides the energy to raise and hold the bed at an elevated temperature,
2) it provides the energy required to desorb the adsorbate from the carbon,
and 3) it carries the desorbed adsorbate from the bed. The steam is condensed
and then decanted. There are two liquid phases present in the decanter, the
aqueous and organic. The organic phase is generally recovered for reuse. The
aqueous phase is either disposed of or, if the level of organics is high,
treated prior to disposal. After the desorption step, the bed is sometimes
dried using heated air. However, this is not required in most cases because
removing water from the carbon usually has little effect on the adsorption
process. In fact, the moisture left on the bed can be beneficial because it
acts as a heat sink during the adsorption process.
Finally, the regenerated bed is then cooled by passing ambient air
through it. In a well designed system both cooling and drying are performed
with the air flow countercurrent to the direction of flow when the adsorber is
on-line. The air exiting the regenerating bed is directed through the on-line
bed to remove any trace adsorbate.
3.2.2 Full Scale System Design Considerations
Section 3.2.1 discussed the overall adsorption system. This section
focuses on the design of the adsorber section itself. Both the physical
system design and the system control and operation during adsorption and
desorption are important in order to achieve high removal efficiencies on a
continuous basis.
The design of full scale carbon adsorption systems begins with a
determination of the inlet stream characteristics. The characteristics
which may be important are:
t Specific compound(s) present;
flowrate and temperature (range and average);
adsorbate concentration (range and average; and
relative humidity.
Any commercial activated carbon should be capable of providing acceptable
performance if the system is designed based on that particular carbon.
However, selecting a carbon which has a majority of micropores which are
3-15
-------
smaller than approximately twice the diameter of the adsorbate molecules, will
result in the greatest adsorptive forces.
Once the carbon has been selected, the required bed area is calculated
based on the desired superficial velocity. For a specified flowrate, the bed
area determines the superficial velocity of the vent stream through the bed.
The lower limit of superficial velocity is 20 ft/min to insure proper air
18
distribution. The upper limit is usually 100 ft/min. This upper limit is
to keep bed pressure drops within the discharge head capacities of the types
of fans used in these applications, and to avoid excessively high system power
costs. Typical superficial velocities are based on vendor experience and the
results of pilot scale testing and will usually be between 50 and 100 ft/min.
19
Generally, carbon adsorber bed depths range from 1.5 to 3.0 feet. A
bed depth of at least 1.5 feet is used to insure that the bed is substantially
20
deeper than the MTZ, which is normally three inches deep. If the MTZ is
longer than the bed, breakthrough will occur almost immediately. The maximum
bed depth of three feet is based on keeping system pressure drop within
reasonable limits.
Within the constraints discussed above determination of the bed depth
becomes a function of the volume of carbon required for one adsorption cycle.
The minimum volume of carbon is determined by the solvent mass loading, the
carbon's working capacity, carbon density, and the desired available
adsorption time. The solvent mass loading and carbon density are fixed by the
stream being treated and the choices of carbon, respectively. The working
capacity and available adsorption time are interrelated and are determined by
the particular carbon, design temperature, adsorbate concentration, specific
compounds present, superficial velocity, and regeneration parameters. The
available adsorption time as a minimum must be greater than the time required
to regenerate (steam and cool) the off-line bed(s).
If the adsorbate contains multiple organic compounds, interactions
between those compounds must also be considered in the estimation of working
capacity. More strongly adsorbed compounds displace the less strongly
21
adsorbed and push them through the bed. This creates a wave front of the
lower molecular weight compounds (which tend to be the compounds less strongly
3-16
-------
adsorbed) at the front of the MTZ. The phenomenon must be accounted for in
estimating the system design working capacity to insure that breakthrough of
any of the compounds does not occur.
At this point it is necessary to determine the working capacity and
specify an adsorption time to determine the carbon volume. Empirical data
from pilot scale testing are usually required to accurately determine working
capacity for a specified set of inlet conditions, superficial velocity, and
desorption steam flow. However, the desorption steam flow selected, in turn,
affects the working capacity and the minimum required adsorption time. The
economic trade-offs of system capital costs versus steam costs will determine
what set of regeneration conditions will result in the lowest annual co^s.
Steaming requirements are set as part of the initial system design. The
longer the bed steaming time the greater the amount of adsorbate removed, and
therefore the smaller the amount of removable heel remaining. As previously
discussed, the working capacity of a carbon bed, which is the amount of
adsorbate the bed can remove during an adsorption cycle, is the difference
between the heel and the equilibrium capacity. Therefore, the longer the bed
is steamed, the greater the available working capacity. An example of the
relationship of working capacity versus steam consumption for three compounds
22
is shown in Figure 3-9. The shape of this curve is similar for most
compounds. However, specific values of working capacity versus steam flow
vary from compound to compound. The curve usually begins to flatten out at
some steam consumption. Increasing steam use beyond the point where the curve
begins to flatten out will result in only a small increase in working
capacity.
In well designed systems the bed is steamed countercurrent to the
23
direction of flow during adsorption. This will help minimize the adsorbate
emitted at the adsorber outlet prior to breakthrough. Figure 3-10, which is a
plot of the adsorbate concentration left on the bed after steaming as a
function of axial distance through the bed, illustrates why this is true.
After steaming, the concentration of adsorbate (i.e., the amount of heel which
remains) is lower at the end of the bed where the steam enters. When the
adsorber is brought on-line, the lower amount of heel where the SLA exits the
bed means less adsorbate is available to desorb. Also, having more working
3-17
-------
00
o
*
o 2
IB O
O.Q.
£ o
<* (0
C CO
TJ
C
I
80
2 4 6 8 10
Pounds of Steam per Pound of Desorbed Hydrocarbons
Figure 3-9. Steam Consumption versus Working Capacities
-------
Direction of Adsorption Air Flow
Direction of Steam Flow
Adsorbate
Left on
Bed After
Steaming
Heel
Distance Through Bed
Figure 3-10. Adsorbate Concentration in Bed After Steaming Countercurrently
as a Function of Distance Through Bed
-------
capacity available at the bed exit helps prevent momentary increases in outlet
concentration as a result of changes in inlet conditions caused by process
upsets. If the bed is steamed cocurrent to the direction of flow during
adsorption, the reverse heel profile exists and a higher outlet concentrations
will result.
Another consideration in adsorber design is fouling. Fouling occurs when
compounds are present in the vent stream which will not desorb from the bed.
These compounds can be solid particles, high molecular weight compounds, or
compounds which chemically react on the surface of the carbon (such as some
ketones). Regardless of the source, bed fouling gradually reduces the carbon
adsorption capacity.
There are two methods to compensate for fouling. One is to increase
volume of carbon beyond the minimum required to achieve the desired adsorption
time. The second is to gradually increase the amount of steam used to
regenerate the bed. Increasing the steam used in regeneration reduces the
heel, which helps maintain sufficient working capacity. A combination of
these methods can also be used.
A typical adsorption/desorption cycling arrangement for a two bed
adsorber system is shown in Figure 3-11. For the purpose of discussion,
illustrative times are shown on the figure corresponding to operational
aspects of the system. The sequence begins with bed 1 coming on-line as bed 2
goes off-line at t For the example shown, adsorption lasts 90 minutes, the
steaming time is fixed at 30 minutes, and the cooling/drying time is also
30 minutes. The off-line bed has 30 minutes during which it is on standby.
In this example, the 30 minutes of standby time allows the operator to
compensate any daily variations in vent stream conditions and bed fouling
without having to leave a bed on-line after breakthrough. It is important
that a bed not be left on-line after breakthrough because that will
significantly reduce the overall removal efficiency for that cycle.
Two types of trigger mechanisms are used for controlling the adsorption/
desorption cycles: continuous monitors and timers. Continuous monitors take
a bed off-line when a specified outlet concentration is reached. Timers cycle
the bed at a specified time. A combination may also be used. One advantage
to using continuous monitors is that they allow the beds to remain on-line
3-20
-------
Bed #1
Adsorption
Regeneration
Steaming I Drying 1 Stand-By
Adsorption
0 30
Time
(mln)
i
60
1
90
1
120
1
150
160
i
210
240 27
Bed
Regeneration
Steaming Drying i Stand-By
Adsorption
Regeneration
Steaming i Drying , Stand-By
ro
0 3°
Time
(mln)
r
60
._. , .
90
| |
120 150 160 210 240 27
Figure 3-11. Adsorption/Desorption Cycles in a 2 Bed System
-------
until breakthrough, thus fully utilizing their capacity during each cycle.
This is not the case for a timer based system, because to properly guard
against breakthrough allowances must be made for variations in the
breakthrough time due to changes in the inlet stream characteristics.
Although continuous monitors allow for the use of more of the available
adsorption capacity than timers, do, timers can be used in many situations as
the trigger mechanism. They are especially appropriate for adsorbates which
do not foul the bed or where inlet stream characteristics are very stable. If
a timer is used, continuous monitors or a periodic sampling program should be
used to adjust the adsorption times as necessary. Deviations in operating
conditions do not affect properly designed systems which use timers unless the
conditions exceed the range of the design specifications. If this is allowed
to occur, a bed may be kept on-line after breakthrough has occurred, This
would result in a significantly reduced removal efficiency.
A final, and important, consideration in system design is prevention of
channeling. Channeling occurs when a portion of the SLA bypasses the bed, or
a certain section of the bed receives a greater portion of the flow than other
sections. The inlet of the vessel must be designed to achieve proper
distribution of the SLA so that it does not impinge on a portion of the bed at
high velocity. The potential for channeling can be minimized by the use of
distribution baffles. It is also important to achieve proper distribution of
the regeneration steam. If steam is not well distributed the steam flow can
also cause channels to form in°the bed. Also, poor steam distribution will
result in some portion of the bed not being properly regenerated.
Proper design can minimize the potential for channeling. However,
maintenance of the distribution baffles and steam distribution system should
be performed during scheduled system shutdowns or whenever an increase is
detected in the adsorber outlet VOC concentration which is significant enough
to result in a removal efficiency below the minimum design level.
3.3 CARBON ADSORBER LONG- AND SHORT-TERM EFFICIENCY
This section discusses the relationship of long and short-term carbon
adsorber efficiency. Section 3.3.1 discusses the calculation of instantaneous
3-22
-------
versus cycle efficiency. Section 3.3.2 discusses the variability of
short-term efficiency. Section 3.3.3 discusses the relationship of outlet
concentration and efficiency.
3.3.1 Calculation of Carbon Adsorber Efficiency
In order to discuss carbon adsorber long- and short-term efficiency, a
short discussion of the relationship of efficiency to the inlet and outlet
concentrations over time is necessary.
The inlet and outlet concentration as a function of time for a single
adsorption cycle of a typical carbon adsorber is shown in Figure 3-12. The
outlet concentration curve is also called the breakthrough curve. For the
example shown, the inlet concentration is C. and the outlet concentration is
C . The adsorber was brought on-line at tQ and taken off-line at t~ when the
outlet concentration reached some predetermined set point concentration level.
At any point in time the instantaneous removal efficiency (IRE) for this
adsorber is determined as the difference between the inlet and outlet
concentration divided by the inlet concentration. At time t, the
instantaneous removal efficiency is:
IRE
1Kb,
The overall removal efficiency (ORE) at any given time during the cycle is
determined by the difference between the areas under the inlet and outlet
curves divided by the area under the outlet curve. For the adsorber shown
in Figures 3-12 at time t, the overall removal efficiency is:
ORE, - Area (ADEF) - Area (BCEF)
1 Area (ADEF)
As an example, in the magnetic tape manufacturing industry a typical inlet
concentration might be 3000 ppmv and a typical outlet concentration might be
24
30 ppmv or less prior to breakthrough. To achieve a 95 percent removal
efficiency over the period of an adsorption cycle a system with an inlet
concentration of 3,000 ppm must have a time weighted average outlet
concentration of 150 ppm or less. At an outlet concentration of 30 ppmv the
removal efficiency is 99 percent for most of the cycle. Therefore, when
breakthrough occurs the outlet concentration can rise above the 150 ppm point
3-23
-------
c, -
ro
C
o
-------
without the overall removal efficiency going below of the required 95 percent.
Therefore, for a system designed and operated to maintain a certain minimum
instantaneous removal efficiency, the cycle efficiency will be higher than the
instantaneous efficiency. At one site the cycling set point for their system
is 2 hours or whenever the instantaneous removal efficiency reaches
95 percent. The result is an overall removal efficiency greater than
99 percent.25
3.3.2 Variability of Short-Term Removal Efficiency
A significant issue raised is the variability of short-term carbon
adsorber removal efficiency. Assuming that inlet stream characteristics never
vary, and the adsorber is always operated the same way, cycle efficiencies
26
should be almost identical. The only change expected would be a gradual
decrease in carbon working capacity due to bed aging. However, the change in
performance from one cycle to the next due to bed aging will be insignificant.
However, in actual applications, inlet stream characteristics such as
concentration, temperature, and flowrate may vary. Also, the operator may
make deliberate changes in the solvent being adsorbed, or in system operation.
If for a well designed and operated system it can be shown that changes in
inlet stream characteristics or operation do not significantly affect
cycle-to-cycle efficiency, then the short-term removal efficiency can be
expected to be essentially constant for industrial applications. These
evaluations are shown in Section 3-4.
3.3.3 Relationship of Outlet Concentration and Efficiency
A typical plot of inlet and outlet concentrations versus time for a
carbon adsorber was previously presented in Figure 3-12. Figure 3-13 presents
a similar curve for the outlet concentration only. In this example Y
represents the outlet concentration at the beginning of the cycle. At the
breakthrough time the outlet concentration begins a sharp increase. At this
time a fresh bed should be put on-line, and the other bed which has just
broken through regenerated.
The dashed line represents what happens if the operator does not remove
the bed which has broken through from service. The outlet concentration will
increase until it equals the adsorber inlet concentration, which in this
example is X.
3-25
-------
Point where bed should
be taken off-line.
X-
ncentra
jg
o
A8
\
^^^
Outlet concentration curve
1 which results if bed is not
' taken off-line at breakthrough.
' brtaktkroHgH
Time
'saturation
Figure 3-13. Typical Carbon Adsorption Breakthrough Curve
3-26
-------
Also shown in this figure are arrows labeled "A" and "B". These arrows
represent the two possible shifts in the breakthrough curve which could occur.
The shift labeled "A" indicates a decrease in the adsorption cycle time prior
to breakthrough. A shift of this type normally should have little effect on
the efficiency of the adsorber, because the bed can be taken off-line for
regeneration prior to the release of any significant emissions. A slight
efficiency reduction over multiple cycles will occur because breakthrough will
occur more frequently. The magnitude of the efficiency change, however, will
be very small.
A shift of the A type will significantly affect efficiency if a bed is
left on-line after breakthrough. There are three possible reasons for a bed
to be left on-line after breakthrough.
1. The operator may not be aware of how quickly the concentration rises
after breakthrough and the resulting deleterious effect on the
efficiency of his adsorption system.
2. The operator may have no way of knowing when breakthrough occurs
(suitable analytical instruments have not been installed).
3. The operator may not have a replacement bed properly desorbed and
cooled, ready for service.
The first two reasons are operational problems and easily overcome. The
third should not occur if the system was designed properly and is being
operated within the design specifications, and the carbon is replaced when
necessary.
If beds are left on-line after breakthrough removal efficiencies will
also become much more variable. As an example, assume that the adsorber
system discussed in Section 3.3.1 has a three hour adsorption time at which
point breakthrough occurs, a 3,000 ppmv inlet concentration, and a 30 ppmv
outlet concentration. Due to a process change the inlet concentration of VOC
now increases to 3,500 ppmv occasionally and this variability was not
accounted for in the system design. In this case, breakthrough would now
occur approximately 26 minutes earlier. If the operator does not take the bed
off-line at this time, the outlet concentration increases very quickly to
3,500 ppm, and for that cycle the adsorber efficiency will be reduced to
approximately 85 percent. If the inlet concentration varies from 3,000 to
3-27
-------
3,500 ppm on a daily basis, then the adsorber removal efficiency will also
vary from 99 to 85 percent on a daily basis. This example demonstrates that
if the system is not designed to account for normal process variations,
efficiency will both be reduced, and become much more variable than is
normally the case with well designed and operated systems.
The shift "B" indicates an increase in the baseline outlet concentration
prior to breakthrough. A shift of this type could result in a significant
decrease in the removal efficiency achievable by the adsorber depending on the
inlet concentration and the magnitude of the outlet concentration change.
For the purpose of this analysis, it is important to understand the two
potential shifts in the breakthrough curve relative to the adsorption
mechanism itself. Assuming a constant adsorbate loading rate, a shift of type
"A" indicates a change in the working capacity of the carbon for a given
adsorbate. The working capacity is a function of fouling and the equilibrium
conditions (i.e., temperature, pressure, and partial pressure of the
adsorbate) for a particular set of operating conditions (i.e., steaming time,
temperature, and duration). Therefore, changes in the equilibrium conditions
which effect the working capacity lead to type "A" shifts in the breakthrough
curve.
A shift of type "B" indicates one of two possibilities: 1) A portion of
the inlet stream has bypassed the bed by either short circuiting or channeling
(as previously discussed, channeling can be avoided with proper design and
maintenance). 2) A greater amount of heel is present in the last few inches
of the carbon bed. As stated previously, the amount of heel is a function of
the conditions which are established at the end of the steaming cycle. The
amount of heel related to the steaming time, temperature, and flow.
Each of the potential operational variables for a carbon adsorber is
evaluated in the next section relative to its ability to shift the
breakthrough curve of Figure 3-13 in either the "A" or "B" direction. Proper
operation practices necessary to prevent degradation of the adsorber system
are also discussed where appropriate. The results from this evaluation are
then used to determine the effect on the removal efficiency which is
3-28
-------
achievable by the system. In presenting this discussion, data and information
available from industry, vendors and emission test reports are used where
available.
3.4 EFFECT OF OPERATING VARIABLES ON ADSORBER PERFORMANCE
The objective of this section is to determine if the normal expected
day-to-day process variations would be expected to necessarily cause daily
variations in carbon adsorber performance. Also assessed are impacts of
deliberate process changes, such as a change in solvent, on adsorber
performance. Potential daily operating variables include the operating
temperature, inlet adsorbate concentration, humidity, volumetric flowrate, and
bed fouling. Changes from the initial design operating conditions include the
adsorbate types(s), and steaming conditions. Channeling will also be
discussed.
Each of the daily normal operational variables is evaluated relative to
its effect on the breakthrough curve from a typical carbon adsorber bed. For
the purpose of this discussion, the assumption is made that any affect on the
performance of a single bed may be taken as representative of the effect on
the overall adsorber system's performance.
3.4.1 Temperature
The operating temperature of an adsorber can be affected in three ways:
changes in the inlet stream temperature, exothermic chemical reactions taking
place inside the adsorber, or failure of the cooling step after regeneration.
Changes in the inlet stream's temperature lead to changes in the adsorber
operating temperature. Changes in the inlet solvent loading can change the
rate of heat generation due to the heat of adsorption. Heat can also be
generated within the system from chemical reactions taking place on the bed.
Ketones in particular, have been identified by several studies as particularly
27
reactive compounds. The problem is usually not serious, however, unless the
concentration of adsorbate is extremely high, the gas flowrate through the
carbon is relatively low, and the carbon is dry and contains no heel.
Each of the possible scenarios given above results in a variation in the
temperature at which the adsorption process takes place. Therefore, the
effect of temperature on the breakthrough curve must be evaluated. As
3-29
-------
previously discussed, the two possible shift directions "A" and "B" can be
assessed by studying the effect of temperature on the working capacity and the
heel, respectively. As shown in Figure 3-14, the relationship between carbon
capacity and temperature indicates that as the temperature within the bed
increases, the adsorptive capacity of the carbon decreases. Thus, as the
temperature increases, the working capacity of the carbon also decreases.
Therefore, a shift in the breakthrough curve is to the left or to shorter
adsorption times. A shift in this direction has no effect on the achievable
removal efficiency but does require a change in the cycle time to compensate
for the shift.
Changes in operating temperature should not cause a B shift in the
breakthrough curve. This is because the outlet concentration at the beginning
of the cycle is primarily a function of the heel remaining in the last few
inches of the bed. The amount of heel is established by the bed steaming
conditions during desorption. Only if the temperature of the carbon in the
adsorber rises to values close to those during steaming is there a chance the
removable heel will desorb and subsequently decrease the achievable removal
efficiency.
Temperature fluctuations in the inlet stream can be essentially
eliminated with installation of a heat exchanger upstream of the carbon
adsorber. A properly designed system will not permit the inlet temperature to
exceed the maximum design temperature.
To illustrate the insensitivity of carbon adsorber efficiency to minor
changes in the bed temperature, the bed temperature and the corresponding
instantaneous removal efficiencies for an operating adsorber system are
28
presented in Figure 3-15. As shown, the carbon bed temperature varies from
60 to 90°F during the adsorption cycle while the corresponding removal
efficiencies remain well above 99 percent. The outlet VOC concentration
oq
remained constant at approximately 20 ppm.
As discussed in the section describing the effect of changes in the
adsorbates, ketones are known to exothermically polymerize on the carbon bed.
A system designed for ketones must assure the air flow through the bed is
sufficient to remove the heat of reaction to insure the bed temperature is not
be significantly affected.
3-30
-------
I
GJ
o
(0
ex
CO
O
c
o
.o
00
O
Temperature
ft
en
Figure 3-14. Carbon Capacity vs. Temperature at Constant Pressure
-------
to
I
to
ro.
60
10
-100
Bed Temperature
Efficiency
3)
(6
o
m
-^«
o"
-------
A properly designed and operated system will limit unacceptable heat
buildup due to reactions in the in any of the following ways:
t Thorough steam desorption and cool down,
t use of the maximum superficial velocity to aid in heat removal,
avoiding prolonged adsorption periods, and
humidity control or even use of liquid water to act as a heat sink.
When a system is designed to handle a ketone-bearing stream, bed temperature
is normally monitored to detect hot spots and initiate protective action for
the carbon bed.
3.4.2 Concentration
The concentration of organics in the inlet stream may vary because of
process changes. Short-term variations are those which occur within a given
cycle while long term variations may last over several cycles. Changes can
occur as equipment or product lines are either brought on or taken off-line.
For the purpose of this discussion, the flowrate through the bed is
assumed to remain constant. Therefore, when the concentration increases, the
loading rate to the adsorber increases.
Increasing the concentration will increase the working capacity of the
carbon. However, the working capacity increase will not be large enough to
completely offset the increase in mass loading. Therefore, the net effect
will be a breakthrough curve shift in the A direction. The effect of
variations in inlet concentration on the outlet concentration prior to
breakthrough should be negligible. As stated previously, the outlet
concentration is a function of the heel in the last few inches of the bed that
remains after regeneration. Because the inlet stream reaches equilibrium with
the carbon within the mass transfer zone, the amount of heel at the adsorber
outlet is independent of inlet concentration. Therefore, short-term
variations in the inlet concentration will not cause a B shift in the
breakthrough curve.
To illustrate the independence of the outlet concentration on short-term
variations in the inlet concentration, the inlet and outlet concentration and
corresponding removal efficiency for an operating adsorber system are
3-33
-------
presented in Figure 3-16. (GTR Test #6) As shown in Figure 3-16 the inlet
concentration varies continuously over the six hour period shown while the
corresponding outlet concentrations and removal efficiencies show little
variation. For inlet variations between 200 and 550 ppm the outlet
concentration varies only from 5 to 15 ppm and the corresponding removal
efficiency varies from 95 to 99 percent.
Figure 3-17 also presents inlet concentration, outlet concentration, and
the corresponding removal efficiency for a similar performance test conducted
on the same adsorber system discussed above. As shown in Figure 3-17, the
outlet concentration (0 to 5 ppm) remains relatively constant for the entire
test period, although the inlet concentration varies from 40 ppm to 880 ppm.
Figure 3-17 also shows the removal efficiency over the test period. During
the majority of the test period the removal efficiency was well above
95 percent. However, when the inlet concentration dropped below 50 ppm, the
removal efficiency was also significantly reduced. This is as expected
because carbon adsorbers are essentially constant outlet devices, so a large
decrease in inlet concentration will reduce short-term removal efficiency.
Because the outlet concentration remains constant throughout an
adsorption cycle, large variations in the inlet concentration wilT result in
corresponding variations in removal efficiency. However, if the bed is
properly regenerated, the outlet concentration can be set at a level where
greater than 95 percent removal is achieved for the entire range of inlet
concentrations. In addition, in many applications of carbon adsorption, a
reduction in inlet concentration is the result of equipment (such as coating
lines) being shut down. By diverting or shutting off the air flow from idle
equipment, inlet concentrations can be maintained at higher levels required to
ensure the desired removal efficiency.
3.4.3 Humidity
Working capacity as a function of steam consumption is shown for relative
humidities of 50 and 100 percent, in Figure 3-18.33 As shown, relative
humidity does not significantly affect working capacity. This is generally
the case for adsorbate concentrations greater than 1,000 ppm.34 Therefore,
there should be only a slight change in the breakthrough time associated with
variations in relative humidity in this case.
3-34
-------
^ soo -
a
a
c
o
400 -
o
O
«
ZOO
C
10 OO I 1 OO 12 OO 13OO
15 OO 16 OO
Inlet Concentration vs. Tlm«
?
a
a
c
o
4*
+*
U
c
0
U
«*
**
o
i?
v^
(j
^
u
o
>
o
E
K
100 -
9O
SO -
70 -
SO -
SO -
40 -
so -
20 -
10 -
0 -
f 1-,
^_j-J
9-OO 10:00 1100 12:OO 13:OO MOO IS:OO 18OO
OutM Concentration
-------
E
a
a
o
o
c
O
I
900
900 -
700 -
MO -
nr
_r
1000 " °°
Tim*
Inlet Concentration vs. Time
12 OO
a
a
c
o
u
o
u
20 -
1 -OH
O
9:00
10:00
11:00
!2:OO
Tim.
Outlet Concantritlon *. Time
9:00
10:00
11:00
12:00
Removal Efficiency vs. Time
Figure 3-17. GTR Test Number 5 Continuous inlet/outlet VOC concentration data
and removal efficiency.
3-36
-------
15
Co
I
CO
C
O
£
(3
O
.Q
8
75
'c
00
00
CO
.0
O.
id
O
CD
C
O
10
EPA Stream B
Toluene - 600 ppm
MEK -1,000 ppm
Cyclohexanone - 400 ppm
Adsorption - 75 fpm
50% Relative Humidity
100% Relative Humidity
1.0 1.5
Ibs. Steam/lbs. Carbon
2.0
-------
Below adsorbate concentrations of 1,000 ppm water begins to compete with
adsorbate for the available adsorption sites and the bed working capacity for
that adsorbate is then affected. In this case, some type of dehumidification
system upstream of the bed or dilution with ambient air may be required.
Relative humidity has no effect on the amount of heel which is retained
within the carbon pore. Therefore, there is no B shift in the breakthrough
curve and no subsequent change in the achievable removal efficiency on a
long- or short-term basis.
High relative humidities are present in most operating systems regardless
of the vent stream conditions because of the water remaining on the bed after
steaming. As shown in Figure 3-18, the working capacity gained by reducing
the humidity is small. In this case, reducing steam humidity would probably
not be cost-effective. In addition, the water content in the bed provides a
heat sink valuable in controlling bed temperature.
3.4.4 Volumetric Flowrate
The superficial bed velocity for a system changes as the volumetric flow
to the system changes. The primary effect is to change the width of the mass
transfer zone within the bed. As the superficial velocity increases, the
width of the mass transfer zone also increases because the individual carbon
pellets are exposed to the adsorbate for a shorter period of time, thus the
quantity removed at a given point decreases. The effect of a wider mass
transfer zone on the shape of the breakthrough curve is shown in the top of
Figure 3-19. As shown in the bottom figure, the time prior to breakthrough is
shortened by increases in volumetric flowrate because of the wider MTZ.
To illustrate the independence of carbon adsorber removal efficiency to
short-term variations in the volumetric flowrate, the flowrate and
corresponding instantaneous removal efficiencies for an operating adsorber
system are presented in Figure 3-20. As shown, the flowrate varies randomly
during the entire adsorption cycle while the corresponding removal
efficiencies show little variation. For flowrate variations from 45,000 to
25,000 scfm the removal efficiency varies less than 0.5 percent with all
efficiencies being well above 99 percent.
Since variations in the volumetric flowrate do not affect the amount of
heel on the bed at a given time, there is no B shift in the outlet
3-38
-------
a
c
g
I
Q>
O
C
O
O
CO
2
c/5
o
a.
ca
Constant Loading
Increasing Flowrate
Distance Through Bed
c
g
O
C
o
o
O
Constant Loading
Increasing Flowrate
Time
a
3
Figure 3-19. Effect of Variation in Volumetric Flowrate on the Shape
of the Breakthrough Curve
3-39
-------
100
o
CO
CO
2
o
Flowrate
Removal Efficiency
90
60 80 100
Time, minutes
120
140
160
Figure 3-20. Carbon Adsorber Removal Efficiency with Varying Flowrate for a
Complete Adsorption Cycle
(t
3
-------
concentration prior to breakthrough, due to either long- or short-term
variations in the flowrate. Consequently, there is no effect on the long- or
short-term removal efficiency achievable by the system.
3.4.5 Bed Fouling
This section discusses the effect of bed fouling an adsorber removal
efficiency. The causes of bed fouling were previously discussed in
Section 3.2.2. Bed fouling gradually decreases working capacity by tying up
the active adsorption sites in the micropores or blocking the pores which
prevent adsorbate molecules from entering. Because the capacity of the system
is decreased, the time prior to breakthrough is shortened. As discussed
previously, this has no effect on an adsorber's removal efficiency until the
shortened length of the adsorption cycle begins to conflict with the
regeneration time. At this point the carbon should be replaced.
Fouling will not affect the outlet concentration prior to breakthrough.
Therefore there will be no B shift in the breakthrough curve. The reason is
that fouling will not affect the amount of heel left in the bed.
As previously discussed, fouling does gradually reduce bed working
capacity. In some cases the steam flow and/or temperature can be increased to
reduce the hee.l and therefore increase working capacity the bed ages.
However, as previously shown in Figure 3-9, the point will be reached where
increasing the steam flow will have little beneficial effect on working
capacity. Therefore, even if the system is well designed and operated, a
point will be approached where there is insufficient time to regenerate the
off-line bed before the on-line bed reaches breakthrough. At this time,
carbon will need to be replaced.
Figure 3-21 presents overall removal efficiency plotted as function of
carbon bed age for a system on a vent stream containing cyclohexanone,
tetrahydrofuran, methyl ethylketone (MEK), and toluene. Both cyclohexanone
and MEK are known to cause bed fouling. As shown, there is only a slight
decline in the removal efficiency from 99.6 percent for the newest bed to 99.4
percent for the oldest bed.
Though fouling of the carbon bed has no affect on the efficiency of an
adsorber system, it does reduce bed life, which in turn increases the annual
operating cost of the system. The fouling rate is affected by numerous
3-41
-------
100
99-
98-
co
I
U
$
"o
£!
H-
Ul
96
95
Legend
Average Composition of
Recovered Solvent (Wt %)
Cyclohexanone 44
THF 23
MEK 19
Toluene 14
12 16
Bed Age, weeks
20
24
28
Figure 3-21. Carbon Adsorber Removal Efficiency as a Function of Bed Age for a
Ketone Containing System
tc.
CM
(O
00
r^
o
-------
factors, but the adsorbate characteristics can be considered the most
dramatic. The effect of several solvent blends on bed life is shown in
Table 3-1. For the toluene and isopropylacetate (IPA) application shown, the
beds have not yet been changed after six years of operation and no shortening
of the adsorption time prior to breakthrough has been detected. For the
toluene and hexane application shown, the bed life is reported to be 10 years.
Lifetime removal efficiencies are also shown in Table 3-1. High removal
efficiencies are shown even for streams containing high concentrations of
cyclohexanone, a known fouling agent. Facility A has a 99.4 percent removal
efficiency on a six bed adsorption system. The bed life for this facility is
significantly less than the other facilities shown. The overall removal
efficiency at Facility A reflects the aggregate for a system with beds at
various stages of life. The solvent recovered at this facility includes
approximately 44 percent by weight cyclohexanone and 19 percent methyl ethyl
ketone. Based on the overall system removal efficiency for this system of
99.4 percent, it can be concluded that all the beds in the system are
achieving well above 95 percent removal.
3.4.6 Channeling
As discussed in Section 3.3, a carbon adsorber system should be designed
with adequate flow baffles and proper steam distribution to prevent
channeling. If channeling does occur, it will cause elevation of the
background outlet concentration over a cycle, or a gradual increase during the
cycle. For systems with VOC monitors, these increases will be readily
apparent. If the amount of channeling is small, the system may still be able
to retain the required removal efficiency. If significant channeling occurs,
then adsorber removal efficiency would be significantly degraded.
In a well designed system, channeling need not occur. From the
perspective of the ability of a carbon adsorber to meet a specific regulatory
removal requirement channeling is actually a malfunction of the system, rather
than a factor causing inherent variability in short-term efficiency.
3.5 DELIBERATE CHANGES FROM INITIAL DESIGN OPERATING CONDITIONS
This section discusses the effect of changing the adsorbate type(s) and
steaming conditions. To assess the impact of each change on the future
3-43
-------
TABLE 3-1. REPORTED BED LIVES FOR VARIOUS SOLVENT BLENDS
38-41
Facility
A
B
C
D
Solvent blend
44% Cyclohexanone0
14% Methyl Ethyl Ketone
23% Tetrahydrofuran
19% Toluene
50% Toluene
50% Isopropyl Acetate
95% Toluene
5% Hexane
Methyl Ethyl Ketone
Reported
bed life
_d
>6 Years
10 Years
5 Years
Reported removal
efficiency (%)
99.4
98a
99. 5b
99.6
Value reported by company. No data was provided to verify.
Estimation: Assumed average inlet loading was mid range in design
specification range, and outlet loading was the reported value.
cActual solvent blends at this facility vary. The values are typical of the
total solvent recovered daily.
The specific bed life at this facility is confidential business information.
However, it is significantly lower than the other values shown. This reduced
bed life is believed to be at least partially due to the presence of
cyclohexanone and methyl ethyl ketone.
3-44
-------
operation and performance of the adsorber system, the logic associated with
the initial specification of each design parameter is discussed.
3.5.1 Adsorbate
The concentration and type of organic are key factors in the design of a
carbon adsorption system. The adsorption characteristics of each compound are
assessed by using their physical properties data, such as: polarity;
refractive index; boiling point; molecular weight; and solubility in water.
Nonpolar compounds and compounds with high refractive indices tend to be
j «
adsorbed more readily. High vapor pressure/low boiling point adsorbates and
low molecular weight compounds adsorb less readily. Compounds with
molecular weights greater than 142 adsorb readily but are difficult to
44
desorb.
If the adsorbate is water soluble, water left as condensate in the bed
45
after steaming and cooling can contain adsorbate. When the adsorber is
brought on-line, the water and adsorbate will evaporate from the bed during
the first part of the adsorption cycle, slightly increasing the initial outlet
concentration for a brief time until the concentration falls rapidly to a
normal baseline value.
The properties and adsorption characteristics affect both the design and
operating conditions. If the feed stream is changed, the adsorber system must
be re-evaluated. If it can accommodate the new feed, there will be no effect
on the achievable removal efficiency; although on-line adsorption time and
steaming requirements may need to be changed. If timers are used as the
trigger mechanisms, the new working capacity of the beds must be determined.
Using this working capacity and a maximum inlet loading, the appropriate new
adsorption time can be determined so that the timers can be reset for the new
operating conditions.
Changing the adsorbate can also affect the desorption cycle. The
relationship between steam usage and working capacity was previously shown for
three different compounds in Figure 3-9. As can be seen, if the adsorbate
blend is changed, the optimum steam requirements may also change.
3.5.2 Steaming Conditions
As previously discussed, steaming requirements are determined as part of
the initial system design. Variables which must be considered are the
3-45
-------
steaming temperature, duration, and rate. Generally, steam temperature is
fixed with a given plant. For this reason, the effect of temperature is not
discussed. The amount of steam required is determined by the required working
capacity. Once the initial design is set, as long as the amount of steam used
per desorption cycle remains constant, the available working capacity will
remain constant assuming no fouling or other degradation of the carbon bed.
In actual application, however, the carbon's total absorption capacity
gradually decreases over time due to fouling. If the operator desires to
maintain the same breakthrough time, steam use per desorption cycle must be
gradually increased. (Alternately, if sufficient standby time is available,
the length of the adsorption cycle can be gradually decreased as previously
discussed.) At some point the amount of steam required per desorption cycle
becomes so great that either there is insufficient time to complete desorption
before breakthrough of the on-line bed, or the cost of steam becomes too
great. At this point the carbon must be replaced.
Although steaming amount is important in the desorption process, duration
is also a consideration. In order to remove the adsorbate, sufficient time at
the steaming temperature is required. This is to allow for diffusion of the
adsorbate out of the pores and out of the carbon particle. Without sufficient
time, increasing the flow of steam will not remove the adsorbate from deep
within the pores of the carbon.
3.6 PERFORMANCE INFORMATION ON INDUSTRIAL ADSORBERS
Available data concerning the ability of carbon adsorber systems to
achieve 95 percent removal efficiencies are summarized in this-section. Data
from performance tests sponsored by EPA's Office of Research and Development
(ORD), three test programs sponsored by EPA's Emissions Standards Division
(ESD), and industry are used to support the conclusions reached. For each of
the tests, the design parameters, operating conditions during testing, and
test information and results are given. A comparison between design and
operating information is then used to evaluate if a given system was operated
within design limits during testing.
Data sources along with the site codes and test numbering scheme used in
the presentation are discussed in Section 3.6.1. In Section 3..6.Z the
3-46
-------
adsorber system operation is discussed with respect to the removal
efficiencies achieved at a given site. Specifically, poor system operating
procedures which were identified are explained. Finally, in Section 3.6 the
conclusions reached regarding adsorber performance are presented.
3.6.1 Data Sources
A summary of the available carbon adsorber emissions test results is
given in Table 3-2. Average overall removal efficiencies, design and actual
operating conditions, and testing information are given for each of 12 sites.
Emissions test data were available for 11 of the 12 sites. The twelfth
reported efficiencies, but provided no supporting data. For each of the tests
1-15, a unique test number is reported. Repeat tests were done at 4 of the
twelve sites. Table 3-3 presents individual average bed removal efficiencies
for several of these sites.
Tests 1-10 were performance tests performed as part of an EPA/ORD study
of carbon adsorber performance in various industries. The manufacturing
processes included are rubberized fabric, magnetic tape, flexible packaging,
and rotogravure printing. Test 6 of this study was not presented for reasons
discussed in Section 3.6.2. Six of the 10 tests were conducted in early 1982.
The four follow-up tests were performed one to two years after the initial
test. Tests 11-13 were performed as part of EPA/ESD performance evaluations
47-49
at three specific sites. Results from Tests 14 and 15 were provided by
industry.50'51
Tests 1-13 were conducted in accordance with approved EPA methods. Inlet
and outlet concentrations were measured semicontinuously with flame ionization
detector total hydrocarbon instrumentation as described in EPA Method 25A.
Volumetric flow rates were measured according to EPA Methods 1 and 2. All
tests were verified by EPA-specified quality assurance/quality control
procedures.
Tests 14 and 15 were performed in accordance with EPA Methods 1,2,
and 25. Method 25 differs from Method 25A in that integrated bag samples are
taken and the concentration of the gas within the bag is used to determine the
removal efficiency over the sampling period. Although this method does not
give the semicontinuous data provided by Method 25A, it provides a means for
accurately assessing the average removal efficiency over the sampling period.
3-47
-------
TABLE 3-2. PERFORMANCE TEST DATA FOR CARBON ADSORPTION SYSTEMS
CO
Design Parameters
Test
No.
1
2
3
4
5
7
8
9
10
11
12
13
14
15
Flowrate
Site (SCFM)
A 11,200
B 12,700
B 12,700
C 23,000
D 22,000
E 11,100
E 11,100
F 80,000
F 80,000
G NR
H NR
I 75,000
J 10,000 -
83,700
K 24,000
L 28,000
Adsorbate Steam
Components Loading Temperature Flow Flowrate
In Inlet stream (Ib/hr) (F) (Ib/hr) (SCFM)
MEK - 60X
Toluene - 40
THF - SOX
Toluene - SOX
THF - SOX
Toluene - SOX
THF, Toluene, MEK,
MIBK, Cyclohexanone
Hexane - 100X
Toluene - 100X
Toluene - 100X
Toluene - 60X
IPAC - 40X
Toluene - 60X
IPAC - 40X
NR
NR
NR
MEK - 20-50X
Cyclohexanone - 20- SOX
THF - 5-25X
Toluene - 5-25X
MEK - 100X
Toluene - 95X
Hexane - 5X
575
140
140
600
1,300
204
204
810
810
NR
NR
NR
180 -
2,800
1,200
860 -
1,070
95 2,375 11,400
90 900 9,800
90 900 9,500
85 2,400 19,800
95 4,130 17,700
100 1,170 9,100
100 1,170 7,800
120 1,180 33,400
120 1,180 33,900
NR NR 8,400
NR NR 48,800
NR NR 61 , 200
<200 .12,000 60,000-
70,000
NR 1,750 24,000
120 3,000 28,000
Conditions During Test
Average
Adsorbate Steam Bed Removal
Components In Loading Temperature Flow Age Efficiency
Inlet Stream* (Ib/hr) (F) (Ib/hr) (years) (X)
b
MEK - 100X 284 94 2,410 0.4 84.9
THF - SOX 195° 74 600 2 99.7
Toluene = SOX
d
THF - 75X 140 90 860 3.8 95.3
Toluene - 25X
THF, Toluene, MEK, 1,260° 81 2,900 0.4 94. 8e
MIBK, Cyclohexanone
Hexane - 100X 355 89 3,200 3 99.1
Toluene - 100X 249° 104 3,000 5 97.6
Toluene - 100X 101 132 2,800 6.5 94.68/h
Toluene - 60X 929° 91 3,400 0.2 97.5
IPAC - 40X
Toluene - 60X 892° 67 3,300 1.6 97.8
IPAC - 4 OX
MEK - 95X 3,550 NR NR NR 98.9
MIBK, Toluene - 5X
Toluene - 100X 980 94 11,000 NR 98^
Toluene - 30X 1,279 100 NR NR 95.8
Xylene - 4X
Lactol Spirits - 66X
THF - 5,410 (Ib
per day) NR 120 12,000 Variable 99.4
Toluene - 3,210
MEK - 4,480
Cyclohexanone - 10,500
L
MEK - 100X 402 83 1,750 2 99.6
NR NR 120 3,000 N/A 99.5
-------
TABLE 3-2. PERFORMANCE TEST DATA FOR CARBON ADSORPTION SYSTEMS (Continued)
'HEX - Methyl ethyl ketone > THF - Tetrahydrofuran; MIBK = Methyl Isobutyl ketone; IPAC = Isopropyl acetate.
The low removal efficiency was a result of adsorbate blend being different from system design specifications, cocurrent steaming,
cooling with adsorbate laden air, and malfunction In condenser system.
Solvent loading was above design specification during test.
Kemoval efficiency was reduced due to reduction In desorptlon steam temperature, change In solvents, and elevated adsorber Inlet
temperature .
Removal efficiency was likely reduced due to high adsocbate loading causing premature system breakthrough.
Inlet stream temperature was above design specifications during test.
Data from bed 2 were not included in average because a steam valve leak caused a severe reduction In bed capacity.
Overall system removal efficiency was reduced due to the inlet stream temperature being above design.
HR - Data not reported.
Removal efficiency la time weighted according to length of time Individual lines were operating.
Inlet adsorbate loading expressed as 67 Ib C /hr. To convert to loading in terms of Ib MEK/hr need to multiply by 6 (MW MEK - 72/MU
iS-12>.
Not applicable. These data are not from a specific performance test. Removal efficiency was calculated from Inlet and outlet
concentrations reported by the facility.
m
Continuous average. Including distillation steam.
Th« average of the removal efficiency of the Individual beds.
ID
-------
TABLE 3-3. PERFORMANCE TEST DATA FOR CARBON ADSORPTION SYSTEMS
ON A PER BED BASIS
53-56
Test number
1
2
3
4
5
7
Bed designation
A
B
1
2
3
1
2
3
1A
IB
IB
IB
2A
2A
2B
2B
3A
3A
3A
3B
3B
1
2
2
2
3
A
B
C
Data collection
time period
NAb
NA
NA
NA
NA
NA '
NA
NA
705 min.
826 min.
382 min.
280 min.
247 min.
279 min.
105 min.
247 min.
128 min.
271 min.
647 min.
898 min.
781 min.
451 min.
388 min.
507 min.
228 min.
295 min.
NA
NA
NA
Removal
efficiency, %
87. 2a
78. 9a
99.6
99.8
99.8
97.3
92.7
95.9
94.0
94.0
94. 3r ,
89.2c'a
92.9
91.3
88.9d>e
95.5
96.2
97.7
96.6
98.0
96.2
99.4
99.3
99.0
99. 4e
98. 9f
97.0
97.6
98.3
(continued)
3-50
-------
TABLE 3-3. (Continued)
Test number Bed designation
8 A
B
C
9 1-1
1-1
1-1
1-1
1-1
1-1
1-2
1-2
1-2
1-2
1-3
1-3
1-3
10 1-1
1-2
1-3
11 1
1
1
1
1
3
3
3
3
3
Data collection
time period
NA
NA
NA
7.6 hrs.
9.1 hrs.
8.6 hrs.
8.1 hrs.
5.5 hrs.
6.8 hrs.
8.5
9.4
8.2
6.0
8.1
6.4
7.4
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
Removal
efficiency,. %
91.4
91.4
97.8
98.0
97.4
97.8
96.6
97.3
97.6
97.9
96.8
97.0
97.6
97.8
97.5
98.1
97.9
98.1
97.5
99.8
95.5
97.7
97.8
98.9
99.4
96.4
98.5
98.8
98.7
aData are from one cycle; overall system efficiency was 84.9 percent.
bNA = Not available.
C0nly one coating line in operation.
Collected during startup of one of the lines.
eCoating process unsteady during this period.
Not representative of normal operation; system returning to steady-state
after all beds on adsorption.
3-51
-------
Site L supplied no detailed information so only the reported removal
52
efficiency is presented.
3.6.2 Removal Efficiency Data for Performance Tests
The overall removal efficiencies presented in Table 3-2 range from
84.9 to 99.7 percent for a variety of adsorbates including MEK, toluene, THF,
MIBK, cyclohexanone, hexane, and IPAC. On a per bed basis, the range of
removal efficiencies is 52.3 to 99.8 percent. As shown, the bed ages
associated with the adsorbers tested range from 3 to 78 months.
The overall removal efficiency of 84.9 percent reported for Test 1 is the
result of both operating outside of the original design range and poor
operation during the test. The system was designed to recover an adsorbate
blend consisting of 60 percent methyl ethyl ketone and 40 percent toluene. At
the time of the test, the organic feed was 100 percent methyl ethyl ketone.
As discussed in Section 3.5.1, switching adsorbate blends can have a
detrimental effect on removal efficiency. In addition, several operating
practices at this site may have contributed to the reduced control efficiency.
These include cocurrent steaming of the bed, cocurrent cooling of the bed with
the adsorbate laden stream, and operating with a malfunctioning steam
condenser.
Cocurrent steaming leaves more residual solvent at the bed outlet than
countercurrent steaming, thus increasing the outlet concentration when the bed
is brought on-line. Cooling the bed with the adsorbate laden vent stream
further aggravates this problem because it allows adsorbate laden air to enter
the bed when the system working capacity is at its lowest. This allows
adsorbate to spread down the bed much further than if the system is operated
correctly. For reasons which have been discussed in Section 3.4.1 this caused
the system to be more sensitive to variations in the operating temperature.
The final problem associated with this system was a malfunction in the
condenser system. The system was not cooling the desorbate stream
sufficiently. Since the steam from the condenser was recycled to the on-line
bed, unusually high amounts of solvent were allowed to enter the bed from the
recycle stream. This additional solvent loading led to premature
breakthrough.
3-52
-------
In summary, these problems would indicate that this system was not well
designed and was poorly operated. Its removal efficiency is not
representative of a well designed and operated system. It should be noted
that several of the problems would have been discovered as part of normal
operation if the system had used continuous outlet monitors.
The average overall removal efficiency presented for Test 2 is
99.7 percent.
The individual bed efficiencies range from 99.6 to 99.8
percent. The solvent blend for this system was 50 percent toluene and
50 percent tetrahydrofuran, and all of the operating conditions were within
design specifications with the exception of the loading rate. The design
specification was 140 Ib/hr, but the actual loading rate was 195 Ib/hr.
Test 3 is a follow-up test at site B. For this test, the average
system overall removal efficiency was 95.3 percent with the individual bed
efficiencies ranging from 92.7 to 97.3 percent. All of the beds had lower
removal efficiencies than in the initial test, Test 2. The reduced removal
efficiency during the second test was attributed to the following in the
test report:
Increased carbon age;
t lower regeneration steam temperature;
t higher SLA inlet temperature; and
change in solvent formulation to 75 tetrahydrofuran and 25 percent
toluene.
58
Figure 3-22 presents a typical outlet concentration curve for this test.
During the initial test (Test 2), the typical outlet concentration had an
initial spike and then decreased to approximately 1 ppm for the remainder
adsorption cycle. However, during Test 3, the outlet concentration was much
higher and an upward trend indicating the beginning of breakthrough can be
59
seen at the end of the cycle. This result would be expected due to the
factors shown above. If the desorption and adsorption cycles had been
adjusted to account for the changes in SLA inlet temperature, solvent
3-53
-------
Average Ctn =1180ppmv
Average Gout = 25 ppmv
I
0)
c
0>
a
o
CO <^»
<0
o
i
O
O
20 30 40
Time (min.)
in
n
Figure 3-22. Inlet/Outlet Concentration Curve for Test 3
3-54
-------
composition, and regeneration steam temperature, and carbon age then the
performance during Test 3 should have been similar to the Test 2 performance.
However, even though this system did show reduced efficiency in Test 2
the average system efficiency was over 95 percent, which would be sufficient
to meet the 95 percent removal efficiency requirement in the proposed magnetic
tapes regulation.
The adsorbate blend concentrations for Test 4 are not specified, but the
blend included tetrahydrofuran, toluene, methyl ethyl ketone, methyl isobutyl
ketone, and cyclohexanone. The average overall removal efficiency for Test 4
is 94.8 percent. Individual bed removal efficiencies ranged between 89.2 and
98.0 percent. The average loading for the test period was 1,260 Ib/hr which
is over twice the design level of 600 Ib/hr. If the adsorption time had been
shortened to account for the increased loading (as discussed in Section 3.4),
removal efficiencies for all beds would have been higher and as discussed in
Section 3.3.3, the variability would be less.
Hexane is the only adsorbate at Test 5. The overall removal efficiency
shown for this test period is 99.1 percent with individual bed efficiencies
ranging between 98.9 and 99.4 percent. All of the operating conditions were
within design specifications.
The follow up test at this site was Test 6. In this test, an extremely
low removal efficiency was achieved due to a leaking steam valve. Excursions
in the outlet concentration were shown to coincide with the steaming cycle for
the off-line bed. In this system, the steam flow is cocurrent, in the same
direction as the air flow during adsorption. Consequently, the steam leak
allowed the solvent laden steam from the off-line bed to enter the outlet
stream of the on line beds. This resulted in false readings at the adsorber
outlet. For this reason no data from this test were included in this report.
The results shown for Test 7 are for a single component system in which
toluene is recovered. The overall removal efficiency for this test is
97.6 percent even through both the inlet temperature and loading rate were
slightly above design specifications. The individual bed removal efficiencies
were between 97.0 and 98.3 percent.
The follow up to Test 7 is Test 8. Toluene was also the only adsorbate
for this test. The removal efficiency for bed 2 is not included in the
3-55
-------
average removal efficiency of 94.6 percent which is shown for this test.
(Table 3-2) A steam valve leak allowed steam to leak into bed 2 raising the
temperature significantly, and severely reducing the working capacity. As in
Test 6, the outlet concentration peaks were, shown to coincide with the
desorption period of the off-line bed. Since the data for this bed are not
representative, they were not included. The lower removal efficiency for the
other two beds is the result of two related problems. A malfunctioning inlet
air cooler allowed the inlet temperature to rise 32°F above the design maximum
and also a timer occasionally malfunctioned. Once again, the use of
continuous outlet monitors in the operation of this system would have helped
to uncover the malfunction in operation.
Tests 9 and 10 were both conducted at site F with an adsorbate blend of
60 percent toluene and 40 percent isopropyl acetate. The overall removal
efficiencies for these two tests were 97.5 percent and 97.8 percent,
respectively. The individual bed removal efficiencies were 96.8 to
98.1 percent. These removal efficiencies agree well with what would be
expected for the two sets of operating conditions. In both tests, the inlet
adsorbate loading was above the design specifications of 810 Ib/hr. All of
the other operating conditions were within design specifications during both
tests.
No design parameters or individual bed removal efficiencies were
available for Tests 11, 12, and 13. Therefore, it was not possible to assess
the system operation in terms of design. The adsorbates for Test 11 were
methyl ethyl ketone, methyl isobutyl ketone, and toluene. The overall removal
efficiency shown for Test 11 is 98.9 percent. For Test 12 the overall removal
efficiency was 98 percent for a adsorbate of 100 percent toluene. The
adsorbate mixture for Test 13 was 30 percent toluene, 4 percent xylene, and 66
percent lactol spirits, and the overall removal efficiency is 95.8 percent.
The overall removal efficiency shown for Test 14 is 99.4 percent. The
average composition of the recovered VOC at this site is as follows:
44 percent cyclohexanone, 23 percent tetrahydrofuran, 19 percent methyl ethyl
ketone, and 14 percent toluene. All of the operating conditions shown for
this test are within the design limits. In Test 15, the adsorbate is
3-56
-------
100 percent methyl ethyl ketone and the overall removal efficiency is
99.6 percent. None of the operating conditions at this site were outside of
the design conditions reported.
No actual test data are available for site L, but the overall removal
efficiency was reported by the company 99.5 percent. The adsorbate blend at
this site was 95 percent toluene and 5 percent hexane.
The performance test data shown in this section generally show removal
efficiencies above 95 percent. For cases where the removal efficiency was
below 95 percent, correctable problems were identified which were the cause of
the lower removal efficiencies. It should be noted that these performance
test data are fairly short duration ranging from less than 2 hours up to
15 hours. If the time periods of startup and system malfunctions are not
considered, the removal efficiencies are fairly consistent with little
variability from bed to bed.
3.6.3 Continuous Removal Efficiency Data
Continuous efficiency data are available from two sites. These data are
presented to show a short-term efficiency variability. The data encompass a
relatively broad range of solvent blends, adsorbate loadings, flowrates, and
inlet temperatures.
Figures 3-23 and 3-24 present continuous inlet and outlet concentration
and removal efficiency data versus time for two test runs from site G in
Table 3-2. As shown in the figures, the inlet concentrations vary
significantly throughout the respective testing periods. However, the outlet
concentrations remain fairly consistent regardless of the inlet
concentrations, and are almost always less than 10 ppm. The removal
efficiencies are also fairly consistent and are generally above 95 percent.
The only time the removal efficiency is below 95 percent is when the inlet
concentration falls below about 50 ppm. This is expected since as previously
discussed, the outlet concentration is independent of the inlet concentration.
Therefore, if the inlet concentration is allowed to fall below the design
value, the instantaneous removal efficiency can also decrease below design
levels.
3-57
-------
TOO -
900 -
400 -
c
o
u
I
1OO
o
»00
20
10
*:00
10:00 >"»
IViw
InM Concontratton v*. Tim*
1200
12:00
OutM Coneontrauon «. Tim*
11:00
12:00
Figure 3-23.
CMeloncy «. Tbn*
GTR Test Number 5 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency
3-58
-------
^> 900
E
a
a
5 *»
o
u
«
200
c
I
o
u
O
u
O
0 -
M -
4O -
JO -
M -
1O -
0
100
K «
3 '
& w
1 i'
' <
w -
n
*OO 10 OO MOO UOO 13-00 MOO 15 OO If 00
InM ConewMrdlon rt. T\m*
*OO IttOO 11:00 12:00 I MO I«OO 15:00 1«:OO
Tta»
OMM Concentration «. Thm
Figure 3-24.
MO IO-OO 11^0 U-OO 13:OO I*OO UOO If-OO
Tim.
GTR Test Number 6 Continuous Inlet/Outlet VOC Concentration
Data and Removal Efficiency
3-59
-------
Inlet and outlet concentrations and instantaneous and cumulative removal
efficiencies for six adsorber beds at site J are presented in Figures 3-25a
through 3-25J. The continuous data shown were obtained at the request of EPA
and are typical of normal facility operation. As previously discussed, the
solvent blend at this facility ranges from 20-25 weight percent cyclohexanone
and 20-50 weight percent methyl ethyl ketone (MEK). Both cyclohexanone and
MEK have been identified as chemicals which react on carbon to cause fouling.
The data are continuous monitor readouts for each of the six carbon beds which
comprise the complete system. The inlet concentrations vary between about
2,000 ppm and 3,000 ppm, with adsorbers #1 and #2 having the highest inlet
concentrations. The continuous outlet concentrations from all 6 adsorbers are
below 50 ppm. Of particular note is the fact that both the instantaneous
removal efficiency and the cumulative efficiency over the entire monitoring
period for all adsorbers are above 98.5 percent.
The instantaneous removal efficiencies of the newest bed ranged from
99.3 to 99.8 percent. The efficiency of the oldest bed ranged from 99.9 to
98.5 percent. One reason the removal efficiencies of the beds stay well above
95 percent is that the beds are changed frequently. This facility could
operate the beds until their removal efficiency has reached 95 percent,
but has chosen not be do so to avoid operation problems which could cause
overall removal efficiency to fall below the desired value.
The data from these two sites indicate that the 95 percent removal
efficiency can be maintained continuously if the carbon absorber is properly
designed, operated within its design specifications, and well maintained.
3.7 CONCLUSIONS REGARDING CARBON ADSORBER PERFORMANCE
The data presented in Section 3.6 demonstrate that properly designed and
operated carbon adsorption systems can achieve 95 percent removal on a
continuous basis. This removal efficiency is shown for numerous solvent
blends and bed ages. Greater than 95 percent removal efficiency is shown for
streams that contain mixtures of ketones that include cyclohexanone, and for
which claims have been made that 95 percent is not achievable using carbon
3-60
-------
E
a.
O
u
o
u
c
u
u
400
350 -
300 -1
250 -
200 -
150 -
100 -
50 -
20 40 60
H Inlet (xl/10)
80 100
TIME (min)
120 140
O Outlet
160
180
Figure 3-25a. Inlet and Outlet Concentration Versus Cycle Time for
Adsorber II
100
99.9
99.8
99.7
99.6
99.S
99.4
99.3
99.2
99.1
99
98.9
»a.a
98.7
98.6
98.3
98.4
98.3
98.2
98.1
98
T~
20
I 1
40
Instantaneous
60
I
80
I
100
I
120
I
140
I
160
I
180
TIME (min)
Cummutativ*
Figure 3-25b. Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #1
3-61
-------
a
a
^^
O
u
u
O
u
400
350 -
300
250 -
200 -
150 -
100
50 -
180
Figure
8
a
\^
s
5
o
3-25c. Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #2
99.9
180
Instantaneous
Figure 3-25d.
Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #2
3-62
-------
E
a
a
s^
o
P
w
u
o
u
c
u
5
u
300
280 -
260 -
240 -
220 -
200 -
180 -<
160 -
140 -
120 -
100 -
80 -
60 -
40 -
20 -
0
20 40 60 80
TIME (min)
Q lnl«t (xl/10)
100
120
Outlet
140
160
Figure 3-25e. Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #3
160
ln«tonton«ou*
TIME (min)
Cummulativ*
Figure 3-25f. Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #3
3-63
-------
E
a
a
s^
o
o
o
o
u
u.
u
260
240 -
220 -
200 -
180 -
160 -
140 -
120 -
100 -
80 -
60 -
40 -
20 -
20
40 60
tatat (xl/10)
80 100
TIME (min)
120
140 160
Outtet
200
Figure 3-25g. Inlet and Outlet Concentration Versus Cycle Time for
Adsorber # 4
99.9
200
Cummulativ*
Figure 3-25h. Instantaneous and Cumulative Efficiency Versus Cycle Time
or Adsorber #4
3-64
-------
300
280 -
260 -
24O -
220 -
200 -
ISO -
16O -
14O -
120 -
100 -
80 -
60 -
4O -
20 -
0
a^B-B-e-a.
T
40
I
60
20
Inlet (xl/10)
1 (
80
TIME (min)
I
100
I
120
140
I
160
180
Outlet
U
5
o
Ul
Figure 3-251. Inlet and Outlet Concentration Versus Cycle Time for
Adsorber #5
too
99.9 -
99.8 -
99.7 -
99.6 -
99.S -
99.4 -
99.3 -
99.2 -
99.1 -
99 -
98.9 -
98.8 -
98.7 -
98.6 -
98.5 -
98.4 -
98.3 -i
98.2 -
98.1 -
98
JO O O O O
20
1 1 1 T
4O 60
1 1 1 1 1 1 1
80 100 120 140
~\ T
160
180
Instantaneous
TIME (min)
Cummulativ*
Figure 3-25J. Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber #5
3-65
-------
E
a
a
^
o
o
I
o
260 -
240 -
220 -
200 -
180 -
16O -
140 -
120 -
100 -
80 -
60 -
40 -
20 -
0
o o o D o a ex.
o a D D a o
20 40 60 80 100
TIME (min)
D lnl«t (xl/10)
120
Outtat
140
160
Figure
3-25k. Inlet and Outlet Concentration Versus Cycle Time
for Adsorber #6.
100
u
5
u
III
Cl
Instantaneous
TIME (min)
Cummulativo
Figure 3-251. Instantaneous and Cumulative Efficiency Versus Cycle Time
for Adsorber 16.
3-66
-------
adsorption. In every case where the removal efficiency was less than
95 percent, correctable and easily identifiable operational problems were
responsible for lower removal efficiencies.
The key to achieving 95 percent removal is proper design and operation
of the adsorption system. If this is done, maintaining a removal efficiency
of 95 percent becomes only a matter of cost where the economic trade-offs
come in the form of steam cost versus carbon replacement costs. The carbon
must be steamed sufficiently to desorb the adsorbate, but excessive steam
use raises the operating costs. The adsorption time must be sufficiently
long to allow regeneration of the other bed(s). This will require
replacement of the carbon when its working capacity gets too low.
In Section 3.4 it was shown that if a system is designed for a full range
of operating conditions, operated correctly, and the carbon is replaced
before its working capacity has been reduced to the point where beds are
operated after breakthrough, the short-term removal efficiency should not
vary significantly. Based on the information and data presented here, it
can be concluded that a removal efficiency of 95 percent or greater is
continuously achievable.
3-67
-------
4. CARBON ADSORPTION SYSTEM AT COMMENTER'S FACILITY
This section presents an analysis of the carbon adsorption system located
at the commenter's facility. The information presented contains data for
which the commenter has made a claim of confidentiality. This information is
located in the confidential files of the Director, Emission Standards
Division, Office of Air Quality Planning and Standards, U. S. Environmental
Protection Agency, Research Triangle Park, North Carolina. This information
is confidential, pending final determination by the Administrator, and is not
available for public inspection.
4-1
-------
5. REFERENCES
1. U. S. Environmental Protection Agency, APTI Course 415: Control of
Gaseous Emissions. EPA-450/2-81-005. Research Triangle Park,
North Carolina. December 1981. p. 5-4.
2. Reference 1.
3. Reference 1, p. 5-6.
4. Reference 1, p. 5-13.
5. Reference 1, p. 5-2.
6. Reference 1, p. 5-14.
7. Reference 1, p. 5-14.
8. Reference 1, p. 5-14.
9. Reference 1, p. 5-3.
10. Reference 1, p. 5-14.
11. Memorandum from Barnett, K. W., Radian Corporation to Carbon Adsorber/
Condensation Project File. February 29, 1988. Meeting Minutes
EPA/Calgon Carbon Corporation, p. 8.
12. Reference 11, p. 7.
13. Reference 11, p. 14.
14. Reference 11, p. 18.
15. Crane, G., Carbon Adsorption for VOC Control. U. S. Environmental
Protection Agency. January 1982. p. 18.
16. Reference 11, pp. 17-18.
17. Reference 1, p. 5-14.
18. Reference 1, p. 5-20.
19. Reference 11, p. 14.
20. Reference 11, p. 14.
21. Reference 11, p. 12.
22. Reference 11, p. A-8.
5-1
-------
23. Reference 11, p. 18.
24. Magnetic tape Manufacturing Industry - Background Information for
Proposed Standards. U. S. Enviormental Protection Agency. Research
Triangle Park, North Caronlina. Publication No. EPA-450/3-85-029a.
December 1985. Appendix C.
25. Memorandum from Grumpier, D., U. S. Environmental Protection Agency to
Berry, J., U. S. Environmental Protection Agency. April 21, 1988.
Plant Visit to Ampex Corporation, p. 2.
26. Reference 11, p. 14.
27. Miller, K. J., C. R. Noddings, and R. C. Nattkemper. 3M Company. St.
Paul, Minnesota. (Presented at Air Pollution Control Association
Meeting. New York, New York. June 21-26, 1987.) p. 4.
28. Memorandum from May, P., Radian Corporation, to file. May 23, 1988.
Ampex Data Analysis, p. 14.
29. Reference 28, p. 14.
30. Reference 11, pp. 14, 17.
31. U. S. Environmental Protection Agency, Industrial Surface Coating
Emission Test Report: General Tire and Rubber Company; Reading,
Massachusetts; Test Series 2. EPA/EMB-80-VNC-1B. Research Triangle
Park, North Carolina. July 1982. Appendix B.I.
32. Reference 31.
33. Reference 11, p. 10.
34. Letter from Shuliger, W., Calgon Carbon Corporation to Wyatt, S. R.,
EPA. November 11, 1987. Response to information request.
35. Reference 28, p. 5.
36. Reference 11, p. 19.
37. Reference 28, p. 14.
38. Reference 28, p. 13.
39. Letter and attachments from Sapovich, M., Diversitech General, to
Farmer, J., U. S. Environmental Protection Agency. January 12, 1988.
Response to Section 114 information request.
5-2
-------
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
Letter and attachments from Jackson, P., Dayco Products, Inc., to Farmer,
J. U. S. Environmental Protection Agency. December 12, 1987. Response
to Section 114 information request.
Letter and attachments from Johnson, W., R. J. Reynolds Tobacco (Archer),
to Farmer, J., U. S. Environmental Protection Agency. December 21, 1987.
Response to Section 114 information request.
Reference
Reference
Reference
Reference
11,
11,
11,
11,
P-
P-
P.
P-
4.
4.
8.
10.
Blacksmith, J. R., T. P. Nelson, and J. L. Randall (Radian
Corporation). Full Scale Carbon Adsorption Applications Study.
Prepared for U. S. Environmental Protection Agency. Cincinnati, Ohio.
Contract No. 68-03-3038. May 1984. 211 p.
Feairheller, W. R. (Monsanto Research Corporation). Graphic Arts
Emissions Test Report: Meredith Burda, Lynchburg, Virginia. Prepared
for U. S. Environmental Protection Agency. Research Triangle Park,
North Carolina. Publication No. 79-GRA-l. March 1979. 30 p.
Feairheller, W. R. (Monsanto Research Corporation). Graphic Arts
Emission Test Report: Texas Color Printers, Dallas, Texas. Prepared
for U. S. Environmental Protection Agency. Research Triangle Park,
North Carolina. Publication No. 79-GRA-3. August, 1979.
Reference 31.
Letter and attachments from Gilbert, R., Ampex to Farmer, J. R., U. S.
Environmental Protection Agency. December 16, 1987. Response to
Section 114 questionnaire.
Reference 39.
Reference 40.
Reference 46.
Full Scale Carbon Adsorption Applications Study. Plant 3. Radian
Corporation (Prepared for the U. S. Environmental Protection Agency).
Cincinnati, Ohio. August 19, 1982. Draft Report.
Full Scale Carbon Adsorption Application Study. Plant 4. Radian
Corporation (Prepared for the U. S. Environmental Protection Agency).
Cincinnati, Ohio. February 25, 1983.
5-3
-------
56. Full Scale Carbon Adsortpion Application Study. Plant 6. Radian
Corporation (Prepared for the U. S. Environmental Protection Agency)
Cincinnati, Ohio. October 29, 1982.
57. Reference 46, p. 108.
58. Reference 46, p. 109.
59. Reference 46, p. 110.
60. Reference 24, pp. 3-4.
5-4
------- |