-------
the National Academy of Sciences in 1971. In preparing this
report, the Academy made a concerted effort to evaluate the
world literature on the subject and distill the best scientific
knowledge available on the biological effects, of fluorides.
This report concludes: "Current knowledge indicates that air-
borne fluoride presents no direct hazard to man, except in
industrial exposure. However, through the commercial, aesthetic,
and ecologic functions=of plants, fluoride in the environment
may indirectly influence man's health and well being." After
considering the available information on fluorides, the Administrator
has concluded that, even though present evidence indicates that
fluorides in the range of ambient concentrations encountered under
worst conditions do not damage human health through inhalation,
they do present a serious risk to public welfare- and warrant
control. Fluoride emissions affect public welfare not only through
their effects on aesthetic values, but also through a decrease in
the economic value of crops which are damaged by exposure to
fluorides and through adverse effects on the health of animals
ingesting vegetation which has accumulated excessive amounts of
fluorides.
Ij]As used in the Clean Air Act, the term "effects on welfare"
induces, but is not limited to, ". . . effects on soils, water,
crops, vegetation, man-made materials, animals, wildlife, weather,
visibility, and climate, damage to and deterioration of property,
and hazards to transportation, as well as effects on economic values
and on personal comfort and well being." [See section 302(h)
42 U.S.C. 1857h(h) as amended.]
xviii
-------
Private citizens and citizens' groups have actively sought
means to alleviate fluoride damage. One citizens' group, Center
for Science in the Public Interest, has written the Agency
describing at great length the need for fluoride regulations.
A number of lawsuits have been initiated which are concerned with
fluoride effects on agricultural products, and at least 20
citizens' suits have been filed against aluminum plants that emit
fluorides.
The Administrator's decision to control fluoride emissions at
the national level was based on the following:
1; 'The present national ambient air quality standards for
particulate matter, standing alone, would not provide
adequate welfare protection against the effects of
fluoride for two reasons: (a) fluorides are emitted
as both particulate matter and gases, and (b) since
the ambient standard is for "non-specific" particulate
matter, compliance with that standard would not ensure
fluoride concentrations sufficiently low to prevent
damage.
2. Although many states have adopted fluoride control
regulations, major sources of fluoride emissions exist
in several states with no fluoride regulations.
3. A uniform national standard of performance for new
sources would discourage movement of major fluoride
emitters to states with no fluoride regulations.
xix
-------
4. Primary aluminum reduction plants, one of the major
sources of fluoride emissions, are commonly located
near major waterways that comprise borders between
states. The potential for interstate conflict concerning
'control of emissions from such plants has prompted
Federal investigations in the past, and in at least one
case a state has requested initiation of abatement
conference proceedings under section 115 of the Act
[42 U.S.C. 1857d].
An EPA report entitled "Preferred Standards Path Report for
Fluorides" (November 1972) contains a detailed discussion of the
advantages and disadvantages of each regulatory option provided
to the Administrator under the Act to control fluoride emissions
on a national level.- In general, the Administrator concluded
that fluorides should be regulated under section 111 of the Act
for the following reasons:
1. In contrast with the problems presented by the six
pollutants for which national ambient air quality
standards have been promulgated, the fluoride problem
is highly localized in the vicinity of major point
sources in agricultural areas and is not complicated
by the presence of numerous mobile sources. Promulgating
a national ambient air quality standard for fluorides
77 A copy of this report is. available for inspection during
normaT business hours at the Freedom of Information Center,
Environmental Protection Agency, 401 M Street, S.W., Washington, D. C.
xx
-------
under section 109 would require states to submit
implementation plans to.attain a'nd maintain such
standards. Because of the complex problems involved
in relating emissions to ambient levels, most plans
would include regulations based on best demonstrated
control technology. The same result can be accom-
plished more directly and efficiently through the
promulgation of standards of performance.
Adopting national standards of performance would be
more compatible with existing state regulations than
adopting ambient air quality standards.
Since accumulation of fluorides during chronic exposure
to low-level ambient concentrations may result in
fluoride levels detrimental to either vegetation or to
the health of animals consuming the vegetation, an
ambient standard for fluorides may not in fact ensure
prevention of adverse welfare effects.
An ambient fluoride standard stringent enough to ensure
complete protection against any welfare effects might
require closure of major sources of fluoride emissions.
A more practical, and feasible approach is to minimize
fluoride damage through best demonstrated control
technology; i.e., by regulating fluoride emissions
under section 111.
xxi
-------
5. The National Academy of Sciences report indicates that
because fluorides present no direct hazard to human
health, the provisions of section 112 for controlling
fluorides as a hazardous air pollutant could not be used.
Promulgation of the proposed standards of performance for
fluorides will affect existing sources as explained in section F
of this preface. Of particular note is that states will be
required to establish standards for the control of fluorides
from existing sources under section m(d) of the Act. The
resulting control may not be as stringent as that required by
the standards of performance for new sources. As indicated
previously, regulations prescribing procedures for control of
existing sources under section lll(d) will be proposed as Subpart
B of 40 CFR Part 60.
xxii
-------
TABLE OF CONTENTS
Section Page
Summary of Proposed Standards 1
Description of Process 2
Emissions and Methods of Control ...... 9
Rati onal e for Proposed Standards 14
' Selection of Pollutants for Control ,14
- : Selection of Units for the Standard 15
Selection of Samp!ing and Analytical Techniques 15
Discussion 20
A. Determination of Affected Facilities 20
B. Determination of Best Control Techniques for
Alumi num Producti on 21
C. Data Base - Primary Emissions 28
D. Data Base - Secondary Emissions 33
E. Data Base - Carbon Anode Bake Plant Emissions 37
F. Cost Analysis of Alternative Fluoride Control Systems .. 41
6. Economic Analysis of Proposed Standards 64
H. Summary 94
References y/
Technical Report Data Sheet , . 99
XXIII
-------
-------
PRIMARY ALUMINUM REDUCTION PLANTS
SUMMARY OF PROPOSED STANDARDS
Standards of performance are being proposed for new primary aluminum reduction
plants. The proposed standards would limit emissions of total fluorides and
visible emissions from potroom(s) which house primary aluminum reduction cells
and from anode bake plants. The entire plant is the affected facility.
The standards" apply at the point(s) where emissions are discharged, from the
air pollution control system or from the affected facility if
untreated by an air pollution control system.
The proposed standards would limit emissions to the atmosphere as follows: <
Total Fluorides
No more than 1 kg of total fluorides per metric ton of aluminum.-(or-
al uminum equivalent) produced (2.00 Ibs/ton) from the primary aluminum
reduction plant, including the carbon anode bake plant.
Visible Emissions
1. Less than 10 percent opacity from the potroom.
2. Less than 20 percent opacity from the anode bake plant
-------
DESCRIPTION OF PROCESS
All aluminum production in the United States' is by electrolytic reduction
of alumina (AlgCL). Alumina, itself an intermediate product, is produced
from bauxite, a naturally occurring ore of hydrated oxides of aluminum.
Major world sources of bauxite are South America and Australia.
Figure 1 presents a schematic flow sheet of the aluminum reduction process.
Alumina is shipped to the primary reduction plant where it is electrically
reduced to aluminum and oxygen. This reduction process is carried out in
shallow rectangular cells (pots) made of carbon-lined steel with consumable
carbon blocks which,are suspended above and extend down into the pot
(Figure 2). The pots and carbon blocks are connected electrically to serve
as cathodes and anodes, respectively, for the electrolytical process.
Cryolite, a double fluoride salt of sodium and aluminum (NaJVIFg), serves as
both an electrolyte and a solvent for alumina. Alumina is added to and
dissolves in the molten cryolite. The cells are heated and operated
between 950° and 1,000°C with heat generated by the electrical resistance
between the electrodes. During the reduction process, aluminum ions
migrate to the cathode where they are reduced to aluminum. Because of its
heavier weight, the aluminum remains as a molten metal layer underneath
the cryolite. Oxygen ions migrate to and react with carbon in the anode
to form carbon dioxide and carbon monoxide which continually evolve from
the cell.
Alumina and cryolite are periodically added to the pot to replenish
material which is removed or consumed during normal operation. Periodically,
the molten aluminum is siphoned or "tapped" from beneath the cryolite bath
-------
CO
CO
CD
O
o
o
-4-»
o
T3
£
_^
<
uu
-------
UJ
Q£
CC
=D
O
f-
O
LLJ
S=
Q
0)
o
CM
0)
13
O)
L
4
-------
and moved to holding furnaces in the casting area. The product aluminum is
held in the molten state until it is cast into billets to await further
processing.
Three different types of cells are used for the production of aluminum: the
vertical stud Soderberg (VSS), the horizontal stud Soderberg (HSS), and the
prebake (PB). Schematic diagrams of these cells are shown in Figures 3, 4,
and 5. These cells differ primarily in their physical configuration, to wit,
the provisions for introducing the electrical current across the cryolite
bath. Although they require more power, the Soderberg systems were acclaimed
initially because they obviated the need for a separate facility to
manufacture anodes. Soderberg cells permit the consumable anode to be baked
in situ. A mixture of ground coke and coal tar pitch is periodically added
to the top of the electrode. Heat from the process drives off the lower
boiling organics and fuses the new material to the old electrode. Partially
because of the problems with the volatile pitch which condenses in the duct-
work and the control device, and partially because of the problems inherent
in simultaneously controlling fluorides and organic emissions, any previous
economic advantage of the Soderberg systems is diminishing and the trend
appears to be toward the PB cell. As can be seen from Figures 3 and 4, the
major difference between the two types of Soderberg cells insofar as the
process is concerned is the manner in which the pins which carry the current
are inserted into the anode.
The PB cell uses an anode that is precast. Since the anode is consumed
during normal operation, old anode remnants or "butts" are replaced
periodically with new anodes. The old remnants are removed from the cell,
-------
I-CD
O
-------
o
o
o
CD
O
D)
L_
CD
O
03
a
O
"to
o
'o.
D)
ii.
-------
ffi
-------
cleaned, ground, mixed with new coke, and blended together with coal tar
pitch in an anode prebake plant. The mixture is weighed, then solidified by
slowly baking in a furnace for about 30 days.
Although somewhat academic from the process aspect, the type of stud used has
a major effect on fugitive emissions that escape the cell as we shall see
later. This is partially because air volume through the hooding system varies
between cell -types, from 4,000 to 8,000 scfm on HSS and PB and from 400 to 600
on VSS.
EMISSIONS AND METHODS OF CONTROL
Several types of pollutants are emitted during the production of primary
aluminum. The major emission source is the reduction cell. Another source
is the anode baking facility.
An uncontrolled primary aluminum plant can emit 40 to 60 pounds of fluoride
(F~) per ton of aluminum produced.1 A poorly controlled primary aluminum
plant can release 15 pounds of F~ per ton of aluminum. Such installations
are likely to be equipped with inefficient capture systems (hoods) on the
reduction cells and inefficient water scrubbers. A 600-ton-per day (TPD)
plant so equipped would emit 9*000 pounds of F~ each day. Plant capacities
in the United States range from 100 to 750 tons of aluminum per day. In
most primary aluminum plants, those emissions that escape the hoods (thereby
escaping the primary control system) exit directly through the roof of the
building to the atmosphere (Figure 6). Such "uncontrolled" secondary
emissions can be several times as large as those which pass through the
primary control system. An overall control efficiency of 95 to 97 percent
9.
-------
LL
10
-------
of the fluorides generated in the potroom will be required to meet the proposed
performance standard. The proposed standard for fluorides will also result in
efficient control of both particulates and organics.
An uncontrolled plant can be a significant'source of particulates by emitting
112 pounds of particulates per ton of aluminum produced or over 33 tons each
day.1 A poorly controlled plant may release as much as 40 pounds of particulates
per ton of aluminum produced/ Such installations frequently attempt
to control emissions with relatively inefficient water scrubbers. A 600-TPD
plant so equipped would emit 12 tons of particulates each day. Particulates
possibly originate in two ways: simple entrapment in the vent system during
periodic additions of alumina and cryolite, and condensation of material
vaporized from the molten bath and carbon anodes. It is estimated that 25
percent of the weight of particulates can be complex fluoride compounds such
as cryolite (Na3A!F6), aluminum fluoride (A1F3)5 calcium fluoride (CaF2), and
chio'lite (Na5Al3F4).1 These can be divided into two categories: water
soluble and water insoluble. (CaF2 is the primary water-insoluble fluoride.)
Measurements of particulate emissions and over 40 hours of visible emission
readings .recorded by EPA indicated that dry control systems or wet scrubbers
in series with wet electrostatic precipitators provide the best control for
particulates. These systems had opacity readings of 10 percent or less at
all times.
The high temperature of the cell causes emissions of organics (tar-fog) from
the anodes at the cells. This fume is not effectively controlled by water
scrubbers and forms the bluish haze characteristic of aluminum plants. This
haze is quite visible from HSS and VSS cells. (Although most of these fumes
-------
are sometimes burned at VSS cells, the burners are not always maintained in
proper operating condition.) The PB has the least visible organic fume
because the low-boiling organics have already been driven off during the
baking process at the anode bake plant. Both dry control systems and wet
electrostatic precipitators appear to provide good control of organic fumes.
It has been reported that large amounts of carbon monoxide (CO) are generated
at the reduction cell. Measurements by EPA show roughly less than 1 percent
by volume in the exit gas streams. Some <
occurs when the hot CO gases contact air.
by volume in the exit gas streams. Some combustion to carbon dioxide (C02)
S02 emissions are the product of sulfur contamination in the organics from
which the anodes are formed. Anode bake plants emitted from 5 to 47 parts
per million (ppm) and VSS and HSS plants, up to 80 ppm based on limited EPA
tests. (One source reports up to 200 ppm S02 can be emitted.3) The sulfur
content of the coke now ranges from 2.5 to 5.0 percent, equivalent to S0?
emissions of 7 to 14 tons each day from a 600-TPD plant.
Results of a limited number of samples indicate NO emissions from primary
X
aluminum plants are very low, about 5 ppm.
Fugitive dust and visible emissions from ancillary operations such as
production of anodes and handling of raw materials can be controlled by
installation of suitable control devices. Historically, cyclones, baghouses,
and pneumatic handling systems have been justified based on the value of
recovered materials.
12
-------
Several State and local regulations limit fluoride emissions from primary
aluminum plants. Some base the restrictions on production rate, others on
the ambient air concentration of fluoride or its concentration in surrounding
vegetation. The most stringent State regulation has dual limits of an annual
average of 1.00 pound of total fluoride per ton of aluminum produced and a
monthly average of 1.3 pounds (the monthly average is based on the average
of three emission measurements which are required per month). This regulation
was developed from emission data collected in early 1973 at one prebake plant
that combined a new dry primary control system with an existing wet secondary
control system. Although the sampling and analytical techniques used are not
known, EPA's Method 13 was not used, so results and standards are not
directly comparable. Also it should be recognized that a standard based
upon averaging over relatively long periods is less stringent than a
numerically equivalent standard established as a nonexceedable limit.
Unquestionably, emissions as low as required by this State standard
may be achievable. An examination of the available fluoride emissions from
primary aluminum pi ants.show that about 75 percent of the allowable emissions
will exit from the roof without being exposed to any control device.
Certainly the installation of even a poor secondary control device could
reduce the total emissions. However, it is the Administrator's judgment
that the overall Federal standard of 2.0 pounds for the potroom will indeed
require the best demonstrated technology, considering cost, within the
intent of the Clean Air Act.
13
-------
RATIONALE FOR PROPOSED STANDARDS
Sel ectiotyof Poljutants for Control
Information gained from the study by EPA titled, "Air Pollution Control in the
Primary Aluminum Industry," (EPA No. 450/3-73-004 A&B and PB-224-282/AS)
indicated that fluorides and particulates are the principal pollutants from
primary aluminum plants. During this investigation, some information was also
collected on emissions of sulfur oxides, nitrogen oxides, and carbon monoxide.
A standard for control of sulfur oxides is not now being considered because
control technology has not been demonstrated in this industry. Primary aluminum
reduction plants could become a significant source of SOg emissions partially
because of a trend toward the use of higher sulfur raw materials for the
manufacture of anodes. Nitrogen oxide emissions were found to be insignificant.
The available data on carbon monoxide emissions indicated that these emissions
were also insignificant.
Documented evidence has shown that fluorides emitted by industrial plants are
responsible for damage to commercially grown flowers, fruits, and vegetables.
Fluorides in low concentrations can also be absorbed by grasses and plants.
They can then cause fluorosis in animals that feed upon such forage. This
disease distorts bone development, retards growth, mottles teeth and adversely
affects general health. ' For these reasonSj fluor1des were selected for
control. Subsequent source tests have shown that if fluorides are well-
controlled, the resulting control of particulates and organics will also be
good.
14
-------
Selection of Units for the Standard
Although both concentration and mass units were considered for the standard,
mass units of pounds of total fluorides (versus water-soluble or gaseous
fluorides) per ton of aluminum produced are recommended for the following
* reasons: ''
' 1. Sampling techniques which permit segregation of particulate and
gaseous fluorides have not been standardized and are not widely
accepted.
2.^ Sampling techniques for water-soluble fluorides have also been
suspect on occasion. Some accepted insoluble fluoride compounds
-. may convert to a soluble form in the sampling train if the
samples are held too long before analysis.
3. To control the emission of total fluorides, the source must control
both particulate and gaseous fluorides. The standard thus indirectly
controls particulate emissions.
4. A standard based on concentration units would be inconsistent
because of variations in the volume of ventilation air used by
various plants and the large fluctuations within a single plant.
5. Aluminum production rates are relatively steady.
6. Product aluminum is weighed as it is removed from the potline';
therefore, accurate production rates are available.
Selection of Sampling and Analytical Techniques
Where possible, sampling and analytical procedures were used which conformed
to Methods 5, 6, and 7, determination of particulate, sulfur dioxide, and
15
-------
nitrogen dioxide emissions from stationary sources, respectively (described in
the Appendix to the December 23, 1971, Federal Register, Volume 36, Number 247).
Measurements of oxygen (02), carbon dioxide (CCL), and carbon monoxide (CO) were
conducted with an Orsat analyzer.
Samples of emissions from primary control systems for fluoride analysis were
collected isokinetically with the sampling train described in Method 5 and
traversed in accordance with Method 1, "Sample and Velocity Traverses for
Stationary Sources."
Standard methods for measurement of fluoride emissions were not available in
the aluminum industry when EPA began its emission test program. EPA determined
that the basic sampling train used with EPA Method 5 could be used to collect
samples of fluoride emissions. Several minor modifications of the train were
tried during the initial tests. At the same time, a similar program of fluoride
measurement was being conducted by EPA in the fertilizer industry. EPA's
Method 13 was developed from the experience gained in these measurement programs.
Although basically the same as Method 5, Method 13 incorporates some options or
variations to improve sampling for fluorides, whereas Method 5 is designed for
particulate sampling. Results of the earlier tests, although not sampled in
strict accordance with Method 13 as 1t finally developed, are comparable.
The analytical method recommended for analysis of samples for fluoride is the
SPADNS Zirconium Lake Method. It was chosen after several analytical methods
were studied by EPA. It has proven accurate and reliable for years by
governmental and industry sources. Samples from EPA's emission test program
were analyzed by this or comparable methods.
16
-------
When emission samples upstream and downstream of a control device could not be
taken simultaneously, sequential sampling was accomplished as quickly as
possible. Length of sampling times varied from 2 to 24 hours.
Special sampling techniques were required to measure emissions of exit gases
where we could neither traverse nor sample isokinetically. These special
sampling techniques were used to sample secondary emissions from roofs or
monitors. Some plants utilized a control system to reduce these emissions,
others did not. Traverse sampling was not practical because of the unusual
configuration of the area to be covered. Also, isokinetic sampling was
compromised to maximize the collection efficiency of the pollutant (fluoride)
in the impinger section of the EPA train. Isokinetic sampling would sub-
stantially reduce the impinger collection efficiency because of low gas
velocities at the sampling location.
Before selecting a method, the Agency held discussions with each company on its
sampling techniques for secondary emissions. As a .result, sampling at a
constant rate was selected. In addition, since the gas velocity was low and
reasonably constant, a single point in the gas stream was sampled. This
single-point method is not unreasonable since about 70 percent of the .partic-
ulates are below 5 microns in diameter and behave almost like a gas. (Some .
data indicate that up to 60 percent of particulates released at the cell could
be less than 3 microns in size.)1 A very important criterion'was to ensure
the sample rate permitted maximum efficiency of the impinger section of the
sampling train. The samples were -collected at a rate of 1 cubic foot per
minute (ft3/min) close to the center of the discharge of the gas stream to
minimize ductwork or other interferences. The sampling points for secondary
17
-------
systems were usually above the cells controlled by the primary control system.
When possible, the primary and secondary systems were sampled simultaneously.
Gas flows from the building were either measured by the company with EPA
observing or the average gas flows provided by the company were used.
A second and more elaborate method of sampling emissions from the monitor was
provided by one company who uses this system at several of their plants.
Continuous samples from a multipoint sampling system in the roof monitor
(Figure 7) were drawn through intake nozzles to a common manifold and
discharged through a stack. Velocity into the nozzles corresponded to the
:-24-hour average velocity through the monitor. Isokinetic sampling was then
performed at the stack to determine the concentration of total fluorides. Gas
flow through the roof monitor was simultaneously measured at many points with
anemometers. These anemometers, spaced in the monitor area, were connected
to a computer which provided a gas velocity reading every few minutes. This
reading permitted a gross measurement of gas flow during the sampling period.
The company has provided data which indicate excellent correlation between
this sampling method and elaborate manual techniques required for sampling
the monitors. Certainly, results of this type of test on a secondary system
should be more representative than those from the single point sample.
Samples were analyzed for both water-soluble and water-insoluble fluorides.
The water-soluble fluorides were determined by the SPADNS Zirconium Lake
Method 12, after the sample was first distilled with sulfuric acid. Water
insoluble fluorides were determined by the SPADNS method after the sample
had been fused with NaOH. These are both standard fluoride analytical
techniques used for many years by industrial and governmental laboratories.
18
-------
e
03
-------
1
Discussion
A. Determination of Affected Facilities
Information initially available for use in the development of standards of
performance for the primary aluminum industry resulted from a study by EPA.
The study had been in process for over a year prior to initiation of the
program to develop standards of performance and was primarily concerned with
emissions and control techniques of the United States primary aluminum
industry. It utilized a survey of the industry by questionnaire, a literature
search, and measurements of emissions from select primary aluminum plants. The
study provided information concerning the history, trends, industrial statis-
tics, processes, emissions, economics, and emission control technology and
procedures of the primary aluminum industry.
After reviewing this work, EPA consulted representatives of several State
agencies and manufacturers of control equipment. Assimilation of all
information confirmed that the reduction cell and anode prebake furnace are
the major sources of pollution at primary aluminum plants, and that the primary
pollutants are fluorides and particulates.
Efficient removal of fluoride from a gas stream is relatively easy. Unfortu-
nately, a significant portion of the gaseous emissions from a cell can escape
capture by the hoods. These fluoride-bearing gases then bypass the primary
collection system. To properly determine the total emissions from a primary
aluminum plant, emissions must be measured as they exit both the building (or
potroom) and the control device. Consequently, it was not possible to use the
cell as an affected facility, and the entire building had to be so designated.
20
-------
All emissions from the anode bake plant exit through the control device so the
entire plant was selected as a second affected facility. Figure 8 presents a
schematic of the two affected facilities.
B. Determination of Best Control Techniques for Aluminum Production
A discussion of best control for potrooms of primary aluminum plants must
consider the two routes by which emissions exhaust to the atmosphere: those
captured by the hood which subsequently pass through the primary control device
(which we will refer to as "primary emissions") and those which elude the hood
system and exit the building through the roof monitors ('-'secondary" emissions).
Most plants do not utilize a control device to reduce secondary emissions.
Table 1 affirms the importance to both wet and dry types of primary collection
devices of good capture of emissions by'the primary hood collection system.
Notice the average fluoride removal efficiency of all primary control devices
was 99.5 percent. Not unexpected is the much lower efficiency of the low-
energy spray-screen scrubbers sometimes used to control secondary emissions.
Collectively, the two averaged only 76.6 percent removal efficiency.
It is obvious that the "best demonstrated control" of fluorides is heavily,
if not totally, dependent on use of a hood which is highly efficient at
capturing fluoride emissions and directing them to the primary control device
rather than permitting them to escape and be scrubbed by the much less
efficient secondary system.
As mentioned in "Description of the Process," the physical characteristics
of each of the three types of cells place various limitations on the design
of the hooding for the primary collection device. At one extreme is the
21
-------
D-
U_
22
-------
CO
S2
o
o
ID
LU
o:
CO
<
o *
LU "-1-1
O <~*"
LU
LI-
LI-
LJ_
m 2: £-
S S
CO "^ LJ
3D CO 1
_J CQ
1 1 ^ 1
III
1 * T
QL.
LU
1
ID
O
LU
1
"ff^ti
t
So
LJ_
Q.
>
1
1 ^
t ^
(_)
! Q^
i s>_
O P-
1
1
Q_
r-.
Os"
i i
CD*
CD
oo"
LU
OO
a
.
i
LL-
OQ
Q_
LU
*S* ,rv
Q_
CO
CO
/^A
CD
Os
CD
CD"
R
vO
CO
c^O
=D Q!
f>X
0 LU
CO QX
Q_
co
CO
1 1 t
Os
OO
Os
Os
CD
CD*
sO
CD
V^5
co
Q
LU
DQ
d
1
Ll_
CO
CO
i
LU
O^\
L^^
Os
Os
LU
LU - .^
o o °3
^C ^
CK CQ JT"1
i i i ^ ^
^>
<< o"
^
s
C£
Q_
m
CO
r*^
v^^
CD
LfN
OO
Csj
LU
LU
0
CO
^^
&L
Q_
CO
CO
CO
Q
CD
J2
S
Csj
CD
c H
1 1
LU
^
1
\~.
^
Q_
CO
CO
CO
LU
vQ
^J
(*^^
*
LU
CD
C ^
-------
VSS cell. Since a substantial area of the surface of the molten bath is out-
side of the skirt of the hood (Figure 3), the capture efficiency of the hood
for this area is poor. Fortunately, the molten surface is usually covered by
a crust of cryolite and regularly replenished with a layer of alumina. The
latter will adsorb fluorides that otherwise would escape. However, periodically
areas of the molten bath are exposed, such as immediately after the crust is
broken to permit addition of alumina. Because these breaks are outside of the
hood, much of the subsequent emissions escape unattended to the secondary system.
Although the length of time between breaking the crust and the addition of new
alumina is largely a function of operating procedure, it can be as long as
half an hour. This delay could be significantly reduced by training and
closer supervision.
The PB and HSS cells can be completely hooded as shown in Figures 4 and 5.
The collection efficiencies of such hoods have been estimated at 97 to 99
percent. Still higher capture efficiencies have been elusive because the
hoods must be opened on a scheduled basis to perform various "cell work,"
such as pin or anode changes, raw material additions, crust breaking, tapping,
and any other operation that requires access to the interior of the cell.
Some companies in the industry have recognized the advantages of containing
fluoride emissions in a primary system rather than attempting to install
or improve the collection efficiency of a 'secondary system. They have made
changes which significantly affect total emissions.
Historically, the doors on HSS hoods extend the full length of both
sides of the cell (15 to 36 feet). As a result, when the operator opens
the door of the cell, he exposes a side the complete length of the cell.
24
-------
Most emission control systems are inadequate to provide sufficient draft to
capture all the emissions under these circumstances. A PB hood can be
segmented so that the area of the cell exposed during working can be minimized
and capture gas velocity maintained proportionately higher. Also one company
using the prebake process has alleviated the draft problem somewhat by using
throttling valves in the ductwork of all cells so that the capture gas velocity
can be doubled before one or more of the segmented doors are opened. These
methods of improving capture are resulting in fewer overall emissions. Modern
plants are assisting air pollution control by closely controlling the process
variables of temperature, chemistry, voltage, alumina and cryolite additions,
and automatic crust breaking without opening the hood.
Companies have also become more aware of the effect of poor maintenance of
cells and hoods on emissions. As equipment ages, hoods and doors are bent,
broken, misplaced, and even partially melted. The increase in open area
can render a marginal draft system completely ineffectual.
After the emissions are captured by the primary system, control is not
difficult. Plants may achieve good control with either of two systems.
"Dry" systems, which sometimes incorporate cyclones upstream of the
control device, take advantage of the strong affinity of alumina for
fluoride. Figure 9 shows two types of dry systems. After the effluent
passes through it, the alumina is fed to the cells, thereby providing a
closed-loop recycle for fluorides. Dry control systems have been
developed for .all three types of cell. Well-controlled "wet" primary
systems, such as shown in Figure 10, utilize a high-efficiency wet scrubber
followed by a wet electrostatic precipitator. These too can be used on
all three types of cell.
25
-------
FLUID-BED
DRY SCRUBBER
ROOF
EMISSIONS
FAN
CELL GAS
ALUMINA
ROOF
EMISSIONS
Figure 9. Fluid-bed dry scrubber (top); injected alumina dry scrubber (bottorr^,
26
-------
CO
o
to
>*
to
o
o
to
to
CD
>,
i_
co
o
o
CD
to
T3
Ctf
CO
_E
O-
o
JD
U-
o
§
O)
27
-------
Those emissions that elude the primary collector are much more likely to
escape to the atmosphere. Potrooms, which house the cells, are designed to
encourage convection currents of ventilation air to sweep the building.
This is accomplished by constructing openings in the floor or by providing
louvers along the lower walls of the potroom or both. As a result, secondary
emissions are diluted by the 30,000 to 60,000 cubic feet of ventilation air
per cell which passes through the roof monitor each minute. Strong cross-
winds can even carry emissions through the floor or wall openings, causing
them to bypass the secondary control system. One operator of a new VSS cell
plant has minimized this problem by closing the sides of the potroom and using
forced ventilation. Forced ventilation will minimize the effect of outside
weather conditions, yet provide for the health, safety, and comfort of
operating personnel.
C. Data Base - Primary Emissions
Preliminary investigations revealed the locations of several reportedly well-
controlled plants. Eight were visited and information was obtained on the
process and control equipment. Six plants were selected to be tested; these
included all three types of reduction cells and both wet and dry air pollution
control systems. The other two plants had no stacks suitable for source
measurements. Two of the six plants tested were later retested a second time
to confirm or supplement previous data. Inspections of two primary aluminum
reduction plants in Europe, one a PB, the other a VSS, revealed their primary
control systems are comparable to those in the best-controlled plants in the
United States. Both plants use dry control systems. More recently, data were
submitted to EPA on a PB plant in the United States using a dry primary control
system and a wet secondary control system.
28
-------
Figure 11 presents the results of all measurements of emissions from primary
control systems. One of the first plants sampled by EPA as part of the
original study?was Plant A, a VSS plant. This original test indicated
emissions of 1.0 to 1.4 pounds of total fluoride per ton of aluminum produced
(Ib TF/TAP). The company extensively modified its control system.by adding a
wet electrostatic precipitator downstream of the "bubbler" scrubber. Three
measurements reveal emissions averaged 0.012 Ib TF/TAP after the modification
(A,), More recently, three measurements by EPA on this same control system
averaged 0.016 Ib TF/TAP (A2). One difference between the second and third
series of measurements conducted at this plant was an increase in sampling
time from about 6 to 24 hours per measurement to confirm this factor did not
affect the data.
Data submitted by the company to a State agency on total emissions of fluoride
averaged about 0.024 Ib TF/TAP (A3). Maximum emissions from the primary system
were 0.039 Ib TF/TAP.9 Calculations based on analysis of inlet concentrations
at Plant A indicate the primary control system is 99.9 percent efficient and
the secondary control system is about 73 percent. Reduced efficiency is the
common shortcoming of low energy wet secondary systems on a low concentration
gas stream.
The average of three measurements by EPA of a primary control system on an HSS
plant reveal emissions from a scrubber in series with a wet electrostatic
precipitator averaged 0.41 Ib TF/TAP (B). Information provided by the company
2
indicates an average of 0.50 Ib TF/TAP (B.,) for the primary control system.
Results of four measurements by EPA of a primary control system show emissions
from a PB plant which uses a fluid-bed dry scrubber averaged 0.15 Ib TF/TAP (C),
29
-------
J..U
0.80
' ALUMINUM PRODUCED
p p o
Jsj !t» en
c=> o o
0
i
s 0.10
I 0.08
CO
CO
gj 0.06
LU
0
§ 0.04
u.
_l
H-
0.02
n ni
PLANT
CELL TYPE
CONTROL
EQUIPMENT
- I I I I I I I I ^ I I I -
, . II,
KEY R ;Ci j^j
f[l| AVERAGE o ' f ^ , i
U ^^ ! 1 « oco
0 v I | 1
EPA TEST 'o 1 ' °
METHOD c j j
COMPANY uLi) | |
- TEST METHOD !j ocxx.
o
,°, o -
| [ LEGEND
~~ ,8j ST- SPRAY TOWER ~
" M BS- "BUBBLER" SCRUBBER
8 P - WET ELECTROSTATIC PRECIPITATOR
FBDS- FLUID-BED DRY SCRUBBER, ~
c IADS - INJECTED ALUMINUM DRY SCRUBBER
/ > ' '
t^ J 1 1 1 1 1 1 1 1 1 1
AI A£ AS B BI C D DI E F G H
VSS VSS VSS HSS HSS PB PB PB PB VSS PB PB
BS-P BS-P BS-P ST-P ST-P FBDS FBDS FBDS FBDS FBDS IADS IADS
Figure 11. Primary emissions from the primary aluminum reduction industry;
30
-------
Results of six measurements by EPA of a primary control system indicate
emissions from another PB plant with a fluid-bed dry scrubber averaged 0.87
Ib TF/TAP (D). Although measurements ranged from 0.26 to 1.70 Ib TF/TAP, two
high measurements of 1.70 Ib are suspect. They are not representative based
on both the company's normal operating parameters and results of their own
measurements. Furthermore, these two measurements were obtained during two
long runs on the same day. During the last of the two, the operator identified
and corrected a malfunction in the control device.
The other four measurements shown for Plant D represent tests before and after
the malfunction and indicate average emissions of 0.49 Ib TF/TAP. Three additional
measurements (D,) on another day showed average emissions of 0.44 Ib TF/TAP.
Fourteen data points provided by the company which owns Plants C and D, for a
1 year period, indicate emissions from primary fluid-bed dry scrubbers averaged
2
0.51 Ib TF/TAP. These are represented by (E) in Figure 11.
This company, which also manufactures the control device, published a paper in
The Journal of the Air Pollution Control Association, Volume 21, No. 8, August
1971, which shows emissions from three PB plants which use dry control systems.
These data, which show total fluoride emissions, are presented here as Table 2.
Company officials indicate that the high figures in Table 2 are now rare due to
improved operating experience gained in the more than 2 years since this data
was published.
Data presented to a State agency in 1973 by one PB plant, indicate emissions
from a primary injected dry control system over a 6-month period averaged
0.31 Ib TF/TAP (H).9
31
-------
CO
CO
LU
CO
CO
LU
CO
cxi
CQ
QQ
CO
co
CJ) - QQ
UU g _J
CO S
a:
a_
CO
OO
o_
CXI
if\
CD
C3
ct:
CXI
CXJ
D_
CD
LU
LTv
CSJ
LT\
LfN
vO
CXJ
CXI
CXI
ON
32
-------
One plant In Europe which uses VSS cells reported emissions of 0.04 1b TF/TAP
(F) from a dry primary control system. Another European plant which uses PB
cells and a dry primary control system reports average emissions as 0.74 Ib
TF/TAP (G).
From the summary information presented in Figure 11 and Table 2, it is obvious
that a primary control system can achieve levels of control of 0.5 Ib TF/TAP
or less. The high values shown were resolved by one operator who added a wet
precipitator to modify the control system. Another discovered an upset in the
control device which was immediately resolved.
Sampling techniques used by companies furnishing other data were not known but
it is assumed that a sampling technique relatively common to the industry was
used. ......
D. Data Base - Secondary Emissions
. Results of all available measurements of secondary emissions are presented in
Figure 12. Measurements taken from the discharge of a spray-screen scrubbing
device .indicate emissions averaged 1.65 Ib TF/TAP (A), 0.80 "ib TF/TAP (A]) and
2.02 Ib TF/TAP (AJ. This is especially significant when you remember that
primary emissions from the plant average only 0.02-0.03 Ib TF/TAP after
modifications (A, and A2). The operator of this plant has submitted informa-
tion to the State agency which showed secondary emissions averaged 1.37 Ib
TF/TAP (A3), and to EPA which showed the average of 32 tests over a longer
period as 1.52 Ib TF/TAP (A4).
Plant B presents the results of four measurements taken by EPA at the roof
monitor of a plant which has no secondary control device. Emissions averaged
33
-------
o n
o.u
2.8
26
*.«u
2.4
2.2
c
§ 2.0
e/»~
J?
£ 1.8
1
"J 1.6
a
cc
o
3 1.4
n
_i
-------
2.10 Ib TF/TAP (B). The company provided data to the State agency which
g
revealed secondary emissions from all the potrooms to average 1.73 Ib TF/TAP.
The figures for secondary emissions merit some discussion because they vary
so much. One, if not the most significant, variable which affects the
emissions through the monitor is the capture efficiency of the hood. The
potroom, which EPA sampled, was one of the plant's newest and contained the
latest in hooded cells. Secondary emissions from this building as measured
by the operator ranged from 0.91 to 1.68 Ib TF/TAP with an average of 1.15
Ib TF/TAP (B,).2 The differences between results of tests by EPA and the operator
do not seem unreasonable. The operator's data were obtained by simultaneous 24-hour
measurements at several points over many months.
In contrast, only one point in the monitor was measured by EPA. Furthermore,
all measurements, except one, were conducted during the day: 12 to 13 hours
for three measurements and 8 hours for one measurement. This single sample
point is assumed to represent the total emissions from the 14 cells controlled
by the primary control system. Realistically, however, these secondary
emissions through the monitor are also affected by the adjacent cells,
t
especially the other 14 cells on the other side of the building which share
the same monitor area. Other causes for variation result from working +ho
cells. Hood doors do not need to be opened for long periods for this purpose,
but other activities may require longer exposure of the cell bed. Records
kept by EPA and the operator of all activities to the 28 cells during the
sampling periods reveal:
1. The lowest emission rate was indicated by a 13-hour sampling from
Saturday afternoon to Sunday morning when the cells are rela-
tively unattended.
35
-------
2. The highest emission rate was indicated by an 8-hour measurement
from early morning to afternoon during a weekday when activities
for all cells are at their peak. In fact, there were pin
changing operations directly below the sampling point which take
considerable time.
It seems reasonable to conclude that:
If multipoint 24-hour sampling were performed, the extremes obtained during
EPA tests should be dampened and more in line with the operator's results.
The results of four measurements of secondary emissions from a PB plant which
has no secondary control device averaged 1.26 Ib TF/TAP (C). The company
indicated the cell hooding in this potroom was not their best compared to
another plant within the company. Measurements of this secondary system were
taken only to provide companion data to the primary system. The operator's
best primary control device is connected to these hoods.
The result of two measurements of uncontrolled secondary emissions from another
PB plant averaged 1.20 Ib TF/TAP (D).
o
The data for (E) represent results of emission measurements from an uncontrolled
secondary system and were provided by the PB plant operator. A high value of
4.4 Ib TF/TAP has been deleted from consideration because the primary collection
system or process was not operating properly, also, this value is
not statistically supported by the other measurements. The average emission
rate without the 4.4 value is 1.14 Ib TF/TAP.
One European operator reports emission rates from uncontrolled roof monitors
as 1.18 Ib
operation.
as 1.18 Ib TF/TAP (F) based on samples taken during 18 months of continuous
36
-------
Data presented to a State agency in 1973 by one PB plant indicate emissions
from a wet sec
Ib TF/TAP (G).
from a wet secondary control system over a period of 6 months averaged 0.66
9
Although examination of Figure 12 would indicate measurements of secondary
emissions ranged from 0.4 to 2.9 Ib TF/TAP, more explanation is essential to
understanding the data. All results of measurements by EPA, with the
exception of Plants C and D, were obtained with a single-point sample under
nonisokinetic conditions. Plants C and D utilized the elaborate multipoint
sampling and flow measurement devices shown in Figure 7. Consequently, it
was only on the latter two plants that sampling could be accomplished in
accordance with accepted practices. Therefore, more confidence can be placed
in those two plants which averaged less than 1.2 Ib TF/TAP. These levels were
achieved at both plants even without a secondary control device.
From the summary of results as presented in Figure 12, it is obvious that good
hood capture efficiency can limit secondary emissions to 1.5 Ib TF/TAP or less
without a secondary control device.
It should be pointed out, howeve'r, that this degree of control has not
been demonstrated for the VSS cell using only the primary control system.' A
new VSS system may require a secondary control system to achieve the level of
emissions required by the standard of performance.
E. DateTBase - Carbon Anode Bake Plant Emissions
The source of fluoride emissioris from the carbon anode bake plant is the anode
remnants or "butts" returned from the primary aluminum plant. Although typical
operating procedures call for removal of fused cryolite from the surface of the
37
-------
butts before they are ground and mixed with hot coal-tar pitch, the absence of
quality control or supervisory emphasis was obvious during EPA plant visits.
We speculate that the recommended level of control may be achievable by better
cleaning of anode remnants.
Two dorrestic carbon anode bake plants were measured. The control equipment
on either was designed specifically for fluorides. Each plant uses an
electrostatic precipitator (ESP) to control organic and particulate emissions.
Before entering the ESP, the gases from the anode plant are treated in a wet
conditioner to improve the effectiveness of collection. Although some
incidental fluoride removal probably occurs, the wet conditioner is designed
to control gas temperature. Results of measurements from the first plant
reveal average emissions of 0.90 pound of total fluoride per ton of carbon
anodes produced (Ib TF/TCAP), as shown in Figure 13, (A). As all the carbon:
anode production is consumed outside, the emission levels can be converted
to their equivalent aluminum production. This conversion indicates average
emissions of 0.45 Ib TF/TAP. Results from the .second anode bake plant reveal
average emissions of 1.25 Ib TF/TCAP (B). This plant sells anodes to other
»
aluminum reduction plants; therefore, no direct correlation with actual
aluminum production was possible. However, based on an estimate commonly
used of 1,000 pounds of anode consumed per ton of aluminum produced, average
emissions were calculated as 0.63 Ib TF/TAP.
It appears that the best controlled carbon anode bake plant is in Norway.
It uses an electrostatic precipitator, a venturi scrubber, and a spray tower,
scrubber in series. Arrangements could not be made to measure this plant.
38
-------
1.3
o
UJ 19
0 L'L
o
o
os
Q_
UJ
0
o , ,
z. 1.1
-------
Using data from studies of standards of performance for the phosphate fertilizer
industry, we conservatively projected that emissions from anode plants can be
controlled with 90 percent efficiency. Calculations based on the highest emission
rate from the two domestic plants (1.33 Ib TF/CAP) indicate a bake plant could
achieve 0.12 Ib TF/TCAP.
-40-
-------
F. Cost Analysis of Alternative-Fluoride Control Systems
1. Introduction
The purpose of this section is to report the expected capital and annual
costs for the control devices necessary to'meet the proposed standard of
performance. Generally two sizes of model plants are analyzed to show how
the control cost changes with size. However, in the primary aluminum
industry the control cost, per ton of aluminum production is not substan-
tially affected by the size of the plant. This is due to the physical
arrangement of the potlines and the control systems.
The basis for coating the control systems was to consider
control device modules (i.e., several aluminum reduction cells
ducted.together and vented to a common control device located
in the :area between the potrooms). Thus, to control a
larger plant additional modules are added. The use of this method reduces
the usual economics of scale associated with control of larger plants.
The 'capital costs for each of the models discussed in the next section
are reported in $/ton of annual capacity. The annual costs are reported
in $/ton of aluminum that would be produced at full capacity rather than
in $/ton of actual aluminum production. Historically the ratio of pro-
duction to capacity has varied widely, as shown in Figure 14. Therefore,
$/ton of capacity is a more stable number which can easily be converted
to $/ton of production for any given production/capacity ratio.
Numerous combinations of control devices have and could be used to con-
trol emissions from aluminum smelters. Each system has a different cost
and control effectiveness. Normally the control systems with higher
41
-------
z
o
ce
O
x
O
1
0.9
0.8
0.7
0.6
O.S
OX
I _ L__' _ I _ 1 _ I __ ! _ ! _ L__l _ !
I
_ __ _ _ __ _ __ _ __
1943 1950 1952 1954 1956 1958 1960 1962 1964 1966 1968 1970
YEAR
DATA SOURCE: A.l M.MIHU.'A STATISTICS, 19i9 THE /.lUMINL'M ASSOCIATION
Figure 14. Excess capacity U.S. primary aluminum industry.
42
-------
efficiency cost more. For two of the cell types in this industry, the
expected cost-effectiveness relationship does not hold. In fact, the
most effective control systems for the pre-bake and vertical-stud Soder-
berg cells are relatively low-cost systems due to credits from recovered
material. 'This effect can be seen-on Figures 15 and 16, wh.ich display the net
annual control costs vs. control effectiveness for several control systems.
Thus, the proposed standard which is based on best demonstrated technology
is also the "best" from a control cost standpoint, that is, a lower cost
system.
The control cost for the horizontal stud Soderberg cells will be higher
since the complex wet control system necessary to meet the 'proposed
standard does not yield any'credits for recovered material. In this
*'eetse the usual relationship between cost and-effectiveness is shown in
" Figure 17.
2. Model Plants
"B *' '
a. Prebake Cells
Since prebake cells can be tightly hooded, the proposed standard can
be met by a good primary control system for the gases collected from
;,-«the. reduction cells. -With a good primary control system, no secondary
control devices should be required on the potroom roof vents.
One of the most attractive control systems (the fluidized-bed dry
, =*. f r '
scrubber--FBDS) for prebake cells uses a fluidized bed of alumina to
absorb the gaseous fluorides and a baghouse to trap the particulate
emissions and any entrained alumina. A second similar system (the
43
-------
.a
s
CO
o
CO
_J
z.
=C
LU
1.0
.8
.6
.4
.2
_
2
8 11
1 ' 4 10 5 7
0. iu a ,
9
-12 6
0
\ 1 1 1 I
468
TOTAL OUTLET EMISSIONS, lb F/ton Al
10
Primary Control Equipment
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
Fluid Bed Dry Scrubber
Miltiple Cyclones plus Venturi Scrubber
Mi Hi pie Cyclones plus High Pressure Spray Screen
Dry Electrostatic Precipitator plus Spray Tower
Multiple Cyclones plus Cross Flow Packed Bed Scrubber
Multiple Cyclones plus Spray Tower
Spray Tower
Multiple Cyclones plus Dry Electrostatic Precipitator plus
Spray Tower
Multiple Cyclones, plus Vertical Flow Packed Bed Scrubber
Multiple Cyclones, plus Floating Ball Wet Scrubber
Chamber Scrubber
Injected Alumina Dry Scrubber
Figure 15. Cost/effectiveness plot for prebake process.
44
-------
- T-8
S:
r
-t»-
GO
O
O
_l
«=c
B i.o
. 2
1,9
3,9 .4,9
5,9
- J6,9
2 9
0.8-9
i
t
3 .4
5
'.6
7
_
2
i: 8 | I 1
2468
TOTAL OUTLET EMISISONS, lb F/ton Al
Primary Control Equipment
1. Spray Tower plus Wet Electrostatic Precipitator
2. Fluid-Bed Dry Scrubber
3. Dry Electrostatic Precipitator Plus Spray Tower
4. Multiple Cyclones plus Venturi Scrubber
5. Multiple Cyclones plus High-Pressure Spray Screens
6. Multiple Cyclones plus Cross Flow Packed Bed Scrubber
7. Spray Tower
8. Injected Alumina Dry Scrubber
9. Spray Screen - Secondary Control
Figure 16. Cost effectiveness plot for VSS process.
45
-------
3.0 -
2.0
«=c
.a
oo
o
o
1.0
3
*
1 I I 1
2468
TOTAL OUTLET EMISSIONS, 1b F/ton Al
1. Cross Flow Packed Bed Scrubber plus Wet Electrostatic
Precipitator
2. Spray Tower plus Wet Electrostatic Precipitator
3. Cross Flow Packed Bed Scrubber
4. Spray Tower
5. Floating Ball Wet Scrubber
Figure 17. Cost effectiveness plot for HSS process.
46
-------
injected alumina dry scrubberIADS) based on the same principle
injects the alumina into the gas stream where adsorption of the
fluorides takes place. The spent alumina is then fed to the reduc-
tion cell. One of the attractive features of the dry systems is that
the captured alumina and fluorides are returned to the cells, thus
reducing input material costs.
Table 3 shows the capital and annual costs for control of prebake
cells^ The costs for control of the vertical and horizontal stud
Soderberg cells are also shown for easy comparison. Included in the
capital cost are the primary collection system (hoods and ducts;, the
fans and other auxiliary equipment, the collection device, and water
treatment facilities if required. All control costs are given in
terms of 1972 dollars. Current costs are approximately 20 percent greater.
Since the carbon anodes used in the prebake cells are made in a
separate operation, the anode baking furnace emissions must be added
to the reduction cell emissions in order to determine the total emis-
sions which are covered by the proposed standard. Table 4 presents
the range of control costs for the anode baking furnace. The low end
of the range is based on a control system consisting of a precooler,
dry electrostatic precipitator, and a wet scrubber. The high end of
the range is based on using a wet scrubber followed by a wet electro-
static precipitator. A water treatment system is included in the
costs. Table 5 shows a summary of the control cost range for the three
cell types.
47
-------
Table 3. COST OF POTLINE CONTROLS FOR ALUMINUM REDUCTION SMELTERS9
Cell Type
Control Equipment13
Capital Cost ($/Ton)
Annual Cost ($/Ton)
Operating and
Maintenance
Depreciation 8%
Administrative
Overhead 5%
Property Tax,
Insurance 2%
15%
Interest 8%
Royal tyd
Gross Annual Cost
Credits (Alumina @
$0. 032/1 b, and
Fluoride @ $0.25/
Ib
Net Annual Cost
($/ton)
(*/lb)
Prebake
1°-FBDS
67C
5.57
10.02
1
5.36
.33
21.28
(10.54)
10.74
0.54
1°-IADS
59
4.35
8.80
4.70
17.85
(10,54)
7.31
0.37
Vertical
Stud Soderberg
1°-FBDS
2°-SS
95C
9.70
14.31
7:64
.33
31.98
(9.19)
22.79
1.14
1°-ST+WESP
2°-SS
117
11.69
17.49
9.32
--
38.50
38.50
1.93
Horizontal
Stud
Soderberg
1°-ST+WESP
' 193
~
11.89
28.91
15.46
__
56.26,
56.26
2.81
Singmaster & Breyer, Air Pollution Control in the Primary Aluminum Industry, July
23, 1973, under Contract No. CPA 70-21 for the Environmental Protection Agency.
Updated to 1972 dollars.
bFBDS - Fluidized Bed Dry Scrubber
IADS - Injected Alumina Dry Scrubber
ST - Spray Tower
WESP - Wet Electrostatic Precipitator
SS - Spray Screen
In addition a $100,000 one-time fee is charged per company for this design.
Correspondence with Mr. Holmes, Manager, Badin (N.C.) Works, ALCOA.
1° = primary control system
2° = secondary control system
48
-------
Table 4. CONTROL COSTS FOR PREBAKE ANODE BAKING FURNACES
Control Equipment
Capital Cost ($/Ton)
Annual Cost U/lb)
PC+DESP+WS
6
0.088
or WS+WESP*
-12
- 0.20
PC - Preeooler
WS - Wet Scrubber
DESP - Dry Electrostatic Precipitator
WESP - Wet Electrostatic Precipitator
49
-------
Table 5. SUMMARY OF CONTROL COSTS FOR PRIMARY ALUMINUM PLANTS
Costs
Capital Cost ($/Ton)
Annual Cost U/lb)
PB (Reduction Cells Plus
Anode Baking Furnace)
65 to 79
0.45 to 0.74
VSS
95 to 117
1.14 to 1.93
HSS
193
2.82
50
-------
b. Vertical Stud Soderberg Cells
The physical arrangement of the vertical stud Soderberg (VSS) cell
makes complete collection of the gases difficult. The control costs
reported for the VSS cells in Tables 3 and 5 are based on the assump-
tion that a secondary control system on the potroom roof vents will
be necessary to meet the proposed standard. '
The primary control systems reported are the FBDS and spray tower ;
plus wet electrostatic precipitator combination (ST + WESP). The
secondary control system is a spray screen device (SS). Although no
VSS cells are presently controlled with the IADS system, there does
not seem to be any reason why they couldn't be controlled with the
IADS. The costs for a primary IADS system would be less than the FBDS
system by the same proportion that it is for the prebake cell.
The present hood design for VSS cells allows enough gas to escape
into the potroom so that the total emissions from a plant with only -
a primary control system would exceed the proposed standard by a
factor of 2 to 5. The costs reported here assume that a secondary
system will be required. Since the secondary system is expensive
and relatively inefficient, some manufacturers may elect to improve
the hooding to raise the collection efficiency so that a primary con-
trol system alone will meet the proposed standard. This alternative
could result in lower costs than for the system reported here, but
it has yet to be proven. Table 6 shows the breakdown of control cost
between the primary and secondary control systems. Obviously, there
is considerable cost incentive to improve the hooding so that a primary
system alone could meet the proposed standard.
51
-------
Table 6. CONTROL COSTS FOR VERTICAL STUD SODERBERGS3
Control Equipment
Capital Cost ($/Ton)
Annual Cost ($/Ton)
Operating and Maintenance
Depreciation 8%
Administrative Over-
head 5%
Property Tax,
Insurance 2%
15%
Interest 8%
Royalty0
Gross Annual Cost
Credits (Alumina @
$0. 032/1 b, flouride
0 $0.25/lb)
Net Annual Cost
($/ton)
(*/lb)
Control System
Primary
ST+WESP
53
4.28
7.91
4.21
16.40
16.40
0.82
Secondary
SS
64
7.41
9.58
5.11
22.10
22.10
1.11
1
Total
117
11.69
17.49
9.32
38.50
---
38.50
1.93
I Control System 2
Primary
FBDS
31b
2.29
4.73
2.53
0.33
9.88
[9.19)
0.69
0.03
Secondary
SS
64
7.41
9.58
5.11
22.10
22.10
i.n
Total
95
9.70
14.31
7.64
0.33
31.98
(9.19)
22.79
1.14
aSingmaster and Breyer, Air Pollution Control in the Primary Aluminum Industry,
July 23, 1973, under Contract CPA 70-21 for the Environmental Protection
Agency. Updated to 1972 dollars.
In addition, a $100,000 one-time fee is charged per company for this design.
f+
Correspondence with Mr. Holmes, Manager, Badin, N. C. Works, ALCOA.
52
-------
c. Horizontal Stud Soderberg Cells
The horizontal stud Soderberg cells might be hooded well enough to
meet the proposed standard by the installation of a good primary
system alone. The control system reported in Tables 3 and 5 is a wet
scrubber plus a wet electrostatic precipitator combination. This is
the only system that has been demonstrated at this time. However, it
is an expensive system with no credits for recovery of fluorides or
alumina. The hydrocarbon tars given off in this process have pre-
vented the use of the dry systems up to this time. Research is being
done to determine the technical and economic feasibility of using the
dry systems for this type of cell. Indications are that if the dry
systems can be used, the costs will be significantly reduced for
control of the horizontal cells.
3. Monitoring System Cost
In order to accurately measure the emissions escaping through the roof
vents, a representative sample must be taken. One way to accomplish this.
is to install a permanent sample collection system. The system consists of
a sampling manifold along the roof vent which is aucted to an exhaust fan
and stack. A composite sample of the roof vent gas can then be measured
at the stack. Several anemometers are installed along the roof vent to
measure the gas velocity. The estimated capital cost of this system in
a pot room is $8500. For a plant producing 100,000 tons pe^ year, the
total cost would be about $34,000 (assuming four potrooms). The annual
cost for this system is about $8400 excluding thb " ,x» cost for collecting
c'.d analyzing the samrles. The annual cost per pound of alunr.ncti is about
53
-------
4. Control Costs to Meet Existing State Standards
Table 7 shows the estimated average emisisons, control level, and con-
trol costs for existing plants as of 1971'. In the absence of a new
source performance standard, new aluminum plants would probably invest
at least as much as the 1971 average for control equipment. As shown
by the cost ranges in Table 7, the costs of controlling some prebake and
VSS cells to levels required by existing State standards are now near
the costs estimated for the proposed standard.
A "high side" approximation of the added cost due to the new source per-
formance standard can be determined by the difference between the costs
in Tables 5 and 7. These results are shown in Table 8. The added cost
for preba-ke plants-is relatively small.. The added cost for VSS is greater,
and the added cost for the HSS is substantial.
The differential control cost would increase the control efficiency to
approximately 96-97 percent from the 1971 industry average of about 74
percent. The emission rate would be reduced from the 1971 industry average
of about 12 pounds per ton to 2.0 pounds per ton.
5. Cost Effectiveness of Secondary Control Systems
The preceding discussion of*control costs is based on the assumption that
the proposed standard can be met by the use of good hooding and a high-
efficiency primary control system on the prebake and HSS cells. Because of
the problem of complete hooding around the VSS cells, the VSS cell plants
will probably require a secondary control system. Even with good hooding
54
-------
Table 7. 1971 PRIMARY ALUMINUM INDUSTRY COSTS,
CONTROL LEVEL, AND EMISSION RATEa
Cell Type
Prebake
Vertical
Stud
Soderberg
Horizontal
Stud
Soderberg
Total
Industry
Capacity
(1000 Tons)
3020
601
1033
4654
Capital Cost
($/ton) ,
50.10.
(28-100r
60.50
(32-109)
49.00
52.20
Annual Cost
(*/lb)
0.54
(0.19-1.23)
UOO
(0.48-1.84)
0.84
0.67
Overall Control
Efficiency (85)
73.9
82.6
70.3
74.2 .-
Emission Rate
(lb F/ton Al)
12.0
8.0
13.6
11.8
a
Singmaster and Breyer, Air Pollution Control in the Primary Aluminum Industry
July 23, 1973, under Contract CPA 70-21 for the Environmental Protection Agency.
bNumbers in parentheses indicate the range of costs reported for existing plants.
55
-------
Table 8. RANGE OF APPROXIMATE COST DIFFERENCES BETWEEN THE PROPOSED
STANDARD OF PERFORMANCE AND EXISTING STATE STANDARDS
Costs
A Capital Cost
($/ton)
A Annual Cost .
U/lb)
Prebake
14.90 to 28.90
(0.09) to 0,20
VSS
34.50 to 56.50
0.15 to 0.93
HSS
144.0
1.98
-------
a small amount of emissions will escape capture at the cell. Secondary
control systems on the roof could be installed to remove some of these
emissions. The following discussion is based on uncontrolled secondary
emissions of 1.5 pounds per ton of aluminum escaping from the best
primary control systems for PB and HSS cell plants.
The cost effectiveness of adding secondary controls can be illustrated
by looking at two types of secondary emission control. The first type
would involve the installation of a spray screen as the secondary control
system on the PB and HSS cell plants. To meet the same degree of control,
the VSS plant would probably have to upgrade their secondary control sys-
tem to that described for the second type. The spray screen is the least
expensive secondary control system, and it is estimated to achieve about
a 35 percent reduction of the 1.5 pounds per ton of secondary emissions
for PB and HSS cell'piants.
the second type of secondary control would involve the installation of a
cross flow packed bed (CFPB) scrubber on the PB and HSS plants. It is
questionable whether the VSS plant could achieve this degree of control.
The CFPB scrubber is highly efficient in removing gaseous fluorides, but
it is not as efficient in removing small particulates. A combined removal
efficiency for the 1.5 pounds per ton secondary emissions of about 60
percent has been estimated with this device. The higher cost and effi-
ciency for the CFPB scrubber are compared with those for the spray screen
and the proposed standard in Table 9. The overall efficiencies are based
on a primary control system of 96 percent efficiency and uncontrolled
secondary emissions of 1.5 pounds per ton of aluminum.
57
-------
CO
CO
LU
CO
CO
Ord
10 I
CJ
ore
cu
00 O
re
re
r- CU
CU CO U
CO > T-
o o <*-
S- O 4-
CO
cu
CU r^
r CO O
O > !-
S- O H-
CO O M-
§° E«
r O 4-> O
1- oo E
co re
0)
i- 03
O E
0 S-
C S-
O CO
CJ 4->
re
b3
||
C re
t- 3
cu
XI CI r
o -T3 c:
t- O O
oo o o
oo en co
EC:"
CO O C
CO
co -a cu
-a co s-
1- CO CJ-
r oo re
a. E
o oo co
i re re co
0) i-
0) >, C
E: Q.I
o o i >,
s- QJ s-
CO S-
(U O
3 cu
Q.
T3 >»
CO +->
u cu
o o
re co
re CO
£?g*:
c S- ca
a> o Q-
o
u -a
co s- cu
re x:
s. re co
re a. co
a.
S? >>
O QJ
r- a. E T- a.
-a i-
c: cu
QJ <»-
p- co >, M- co
s--a
0) 0)
(U
QJ
0)
«
S- !-
CU
S- O r
O CO T-
o re x:
o a.
o
cu cu
oo .c
XI 0
S- T3 i s-
O S- CU +J
co re o c-
T3 o
CQ C CO o
o- re co
CJ CO
cu
re -a x:,
cu +> co
£""
O CU
£*- >
CU CO
CO
CO O
9 <->
-------
As shown in Table 9, the cost of control rises sharply with each incre-
mpnt of control efficiency, designated in the table as Proposed Stan-
dard (96 percent), Control Level #2 (97 percent), and Control Level #3
(98 percent). For example, the annual control cost for the prebuke cell
more than doubles in going from the proposed standard to Control Level
#2 and goes up by a factor of 3 to 4 between the proposed standard and
Control Level #3.
The cost effectiveness of the various control systems in terms of cost
per pound of fluoride emissions captured is shown in Tables 10-12. For
the prebake process shown in Table 10, the industry is currently spending
$50 of capital per ton of aluminum capacity in order to capture 74 percent
of the fluoride emissions from the process. This is equivalent to $1.35
per pound of fluoride captured. In order to achieve a control level of
96 percent, the industry would have to spend a total of $73 of capital
per ton of aluminum capacity. Therefore, the average cost for the entire
48 pounds of fluoride that could be captured per ton of aluminum capacity
amounts to $1.52. This is equivalent to $1.35 for the first 37 pounds of
fluoride captured and $2.09 for the next 11 pounds of fluoride captured.
It can be seen that a large gap exists between the capital costs required
for fluoride control at the proposed standard level (96 percent) and
Control Level #2 (97 percent). Whereas the average capital cost per pound
of fluoride removed only increases from $1.52/lb to $2.45/lb, the cost of
capturing the additional 0.5 pound of fluoride amounts to $92.00/lb of
fluoride. The operating costs for the fluoride control systems behave
in a similar manner. It costs an average of $32.00/Tb of fluoride
59
-------
i I CU
CO S-
.-O a.
CO X» CO
CJ
CJ
i- +J-i- £=
CU E Q- O
> O CO I
~ CJ CJ .
CO
cu
' -d 3
co S-
&-+>!-
cu c: a. o
> o as\
", O CJ '
BtL
CO
CO
«*
o oo 10
io CM «*
in CM LD oo
ff) If) ^ if)
CM OO
O OO CT1 LO
- VO 1
00
o o to o>
1^ CO CO CO
CO
cu
CU
t? S- CM OO
-P to =tt==fe
t/) T3
ZI C i i
-O CO >
t-H CO O) i 3
c/>- »->
o -a a.
o -a
CU
CD O
S- 4-> VI -Q
CU C O r
> O CJ *-^
". O -t*
(A 1
CJ S- -
cu o
, ST
o
CU i r-:
en o
CO S- 4->
i. 4-> 10 J3
CU C O i
p cj-
co
CU T3
-a cu
r- S-
o «->
3 Q.
i co
U- CJ
Oi
s- a>
c cu
en in o o
CM i O UO
o o CM r^
oo co
o to o o
«* oo o to
10 o oo i
o o o o
cr> <
> a>
CM CM LO OO
O O '
o LO LO 10
«d- CM CM
IO VO «* i
»
O O r CM
CT»
. oo
O O 10 CT»
T*~CQ CQ CQ
co «*«*
cu
> .
cu
_i
t? S- CM CO
!-> CO =«==«=
w> -a
3 c: r r
-a co cu cu
c !->>>
HH CO CU CD
cn-o
c cu
i- u> o o
4-> O S- S-
r- O
X S- O O
UJ Q- O CJ
CU
cu
cu
CJ
cu
in
to
CQ
to
60
-------
captured to go from 48.0 pounds to 48.5 pounds of fluoride captured,
whereas the cost of capturing the first 37.0 pounds of fluoride is
$0.29/lb and the cost of capturing the next 11 pounds of fluoride is
$0.15/lb.
Table 11 shows that a similar gap exists between the proposed standard
level and Control Level #2 for the vertical stud Soderberg process.
Capital requirements average $4.29 per pound of fluoride captured at
97 percent control versus $2.21 per pound of fluoride captured at 96
percent control, but the cost of capturing the last 0.5 pound of fluoride
costs $204.00/lb versus $1.45/lb for the first 41.5 pounds of fluoride
captured and $7.08 for the next 6.5 pounds of fluoride captured. Operating
costs amount to $26.00/lb of fluoride captured for the last 0.5 pound
captured versus $0.48/lb for the first 41.5 pounds captured and $1.63 for
the next 6.5 pounds of fluoride captured.
Table 12, which measures the cost effectiveness of various control levels
on the horizontal stud Soderberg process, again shows a large gap between
the costs incurred at a control level of 96 percent versus a control
level of 97 percent.
61
-------
to
co
a.-a -t->
o5 -o a.
o O 03 t
=C CJ CJ '
re
4-> .
a>
UJ
CO
O
CO
co
uj
re s- a
o a)
&=»
O Q-
-
CD r r i
o> o 03 «=c
tO S- ^
S- +> -r- C
Q. O
O)
.
> O 03 I
CJ>
CO
CO »-<
UJ
CO
UJ 11
> s:
HH UJ
o
UJ
u_
re
CO T3
T3 CO
i~ i-
S- 3
O -I-5
3 Q.
i re
U. CJ
UJ
CO
8
cu
CO
a:
s-
LO CD LO
i oo oo
O)
cu
S- S- CM
-l-> 03 =f»=