Date:	March 6, 2013

Subject:           Development of Background Information on Proposed
Area Source Emissions  

                        Limit - Determination of GACT for Wool
Fiberglass Furnaces at Area Sources

	EPA Contract No. EP-D-11-084; EPA Work Assignment No. 1-07

	RTI Project No. 0213199.001.07

From: 		Cindy Hancy

		Dave Reeves

To: 		Susan Fairchild, EPA/OAQPS/SPPD/MMG

I.  Introduction  

Section 112(k) of the Clean Air Act requires the U.S. Environmental
Protection Agency (EPA) to develop standards for area sources which
account for 90 percent of the emissions in urban areas of the 33 urban
hazardous air pollutants (HAP) listed in EPA’s Integrated Urban Air
Toxics Strategy.  These area source standards can require emission
control levels that are equivalent to generally achievable control
technology (GACT) for affected sources.  For the development of area
source standards, GACT is meant to include methods, practices and
techniques which are commercially available and appropriate for
application by the sources in the category considering economic impacts
and the technical capabilities of the firms to operate and maintain the
emissions control systems.

The purpose of this memorandum is to describe how GACT was determined
for glass-melting furnaces, which is the affected source for the
proposed National Emission Standards for Hazardous Air Pollutants
(NESHAP) for Wool Fiberglass Manufacturing Area Sources.  In the
development of the residual risk and technology review of the major
source NESHAP for wool fiberglass manufacturing (Subpart NNN), the
industry trade association (NAIMA) and EPA sent information requests to
all major and area source wool fiberglass facilities and had several
meetings with industry regarding current practices at wool fiberglass
plants. Because of the large number of emission test reports and other
documents reviewed, and the fact that most documents included
information claimed as confidential business information (CBI), the
specific references for the test data are not listed in this memorandum.
 However, a summary of the test reports and all of the other documents
that were used in the analysis of GACT, other than those that were
claimed to be CBI, can be found in the docket. 

II. Glass-melting Furnaces

Glass-melting furnaces vary according to size, design, fuel source, raw
material mixtures, melting times and methods of heat application.
Glass-melting furnaces can be electric steel shell, cold top electric,
air-gas, oxyfuel, or combinations of gas and electric. There are two
types of electric furnaces currently used in the wool fiberglass
industry: electric steel shell and cold top electric. There are two
types of gas-fired glass-melting furnaces: air-gas, and oxy-fuel.  There
may also be combinations of gas fired and electric furnaces (known as
‘electric boost’) in operation. About 80 percent of the
glass-melting furnaces used in the wool fiberglass industry are electric
(e.g., steel shell or cold-top) and about 20 percent are gas-fired
(e.g., air gas or oxy-fuel).

⁰ F.

⁰ F. inside the furnace) than air gas furnaces (2400⁰ F), whose
temperatures are reduced by the dilution of oxygen in the outside air. 

III.  Existing Glass-melting Furnace Emission Controls  

The type of control device used to control particulate matter (PM)
emissions varies by furnace type. Electric steel shell furnaces use a
baghouse and/or fabric filter; cold top electric furnaces use batch
wetting as the method of PM control. Air-gas furnaces currently use a
combination of electrostatic precipitators (ESP) and baghouses; oxyfuel
furnaces use an ESP only (with one oxyfuel furnace using a scrubber
along with an ESP). Figure 1 shows PM emission rates by control device
for all furnaces in the wool fiberglass manufacturing source category
that submitted PM data.

IV.  Glass-Melting Furnace Emissions  

Chromium Emissions

Wool fiberglass Manufacturing is being listed as an area source category
based on emissions of chromium compounds.  In the development of the
residual risk and technology review of Subpart NNN (for major sources),
the industry trade association (NAIMA) and EPA received emissions
information from the majority of all major and area source wool
fiberglass facilities. The EPA evaluated chromium emissions data for
major and area source glass-melting furnaces and analyzed the
operational differences of the different types of glass-melting
furnaces.  The EPA found that electric furnaces (electric steel shell
and cold-top electric) do not have the potential to emit chromium at the
levels of gas-fired furnaces, and that chromium emissions from electric
furnaces are about 2 orders of magnitude lower than the chromium
emissions from gas-fired furnaces as shown in Figure 2. 

Figure 1. Furnace PM Emission Rate by Control Device

Figure 2.  Chromium Compounds Emission Rates by Furnace Type

This higher chromium emission potential for gas-fired glass-melting
furnaces is due to differences in design and operation of gas-fired
glass-melting furnaces compared to electric glass-melting furnaces. 
Although all glass-melting furnaces are constructed using chromium
refractories at and below the line of contact defined by the refractory
wall and the molten glass within the glass-melting furnace (the
glass/metal line), oxy-fuel and some air gas glass-melting furnaces have
other glass-melting furnace parts constructed using chromium
refractories, such as the crown, sidewalls, front wall, back wall, and
forehearth. The use of chromium refractories above the melt line is
necessary to obtain the desired furnace life and reduce the necessity
for hot repairs of the furnace. When the hot, corrosive and reactive
gases of a gas-fired glass-melting furnace come in contact with the high
chromium refractories lining the area above the glass melt in high
temperature glass-melting furnaces, the chromium is available to be
oxidized and converted into its hexavalent form. 

◦ to 4,500◦F, while the temperatures in electric glass-melting
furnaces are less than 300⁰ F. Due to their higher operating
temperatures, gas-fired glass-melting furnaces are constructed using
chromium refractories at various parts of the glass-melting furnace that
are above the molten glass, including the crown. Chromium refractories
may contain over 90% chrome bearing compounds, specifically, chomite ore
(as Cr2O3) and chromic oxide (as Cr3O4). The use of chromium
refractories above the glass melt line and corrosivity due to thermal
and chemical degradation of the furnace interior contribute to the
greater potential for gas-fired furnaces to chromium compounds. 

Studies indicate that, under normal industrial temperatures and
oxidizing conditions, trivalent chromium, which is present in the
refractory, oxidizes readily to its hexavalent state.1 

Additionally, while the degradation of the glass-melting furnace
refractory indicates increasing chromium emissions, that process does
not necessarily follow a normal and predictable pattern. The degradation
of refractories within the glass-melting furnace is a function of
numerous factors, including temperature, time, stress, and the composite
effects of aging and creep response. These processes are highly
nonlinear, so the traditional equations that assume steady-state
deformation rates are not appropriate.2 We have also found that as the
refractories of the gas-fired glass-melting furnaces degrade, the
chromium of those refractories at and above the metal/glass line is
emitted as particulate to the outside air. Chromium from the
refractories below the metal/glass line is absorbed into the molten
glass and becomes vitrified with the other raw minerals. Industry
commented that refractory loss from degradation of the refractory walls
in use is approximately 20,000 pounds of refractory annually.3 When that
loss occurs below the glass melt line, it is believed to be incorporated
into the molten glass. When that loss occurs above the glass melt line,
it is emitted from the furnace. While existing air pollution controls
achieve reductions in the total PM, a greater fraction of that PM
consists of chromium in the gas fired furnaces than exists in the
electric furnaces, and have been measured from gas-fired furnaces at
over 500 pounds per year. Therefore, additional chromium reductions for
gas-fired furnaces in this particular industry are needed.

EPA has learned that if a source of reasonably priced industrial-grade
oxygen is available, the oxyfuel glass-melting furnace is the design
favored for use by glass manufacturers due to the glass-melting
furnace’s low NOx emissions, low energy demands per volume output of
glass, and extended furnace life span. The low NOx emissions of an
oxy-fuel glass-melting furnace result from the fact that no air (which
contains nitrogen) is introduced into the high temperature zone above
the glass melt. Instead, the oxy-fuel glass-melting furnace design mixes
the natural gas fuel with pure oxygen for combustion, thus reducing NOx
emissions. 

In the 1999 Wool Fiberglass Manufacturing NESHAP (subpart NNN), EPA
promulgated emission limits for PM (0.5 lb PM per ton of glass pulled)
as a surrogate for metal HAP emitted from glass-melting furnaces. The
rationale for this surrogacy is that the metal HAP are contained (in
trace amounts) in the PM, and that the PM control devices remove metal
HAP with the same efficiency as they remove total PM.  Therefore, a
limit on PM also creates equivalent control of the metal HAP.

IV.  Technical Basis For Setting Emissions Limits For Both PM And
Chromium for Gas Fired Furnaces

We continue to believe that PM is an appropriate surrogate for metal HAP
for electric glass-melting furnaces at wool fiberglass manufacturing
facilities. The EPA may use a surrogate to regulate HAP if there is
reasonable basis to do so.  We have used PM as a surrogate to regulate
hazardous pollutants because: “(1) HAP metals are invariably present .
. . in PM;” (2) “PM Control technology indiscriminately captures HAP
metals along with other particulates;” and (3) “PM control is the
only means by which facilities achieve reduction in HAP metal
emissions.” National Lime v. EPA, 233 F.3d at 639.  We have also
always reasoned that “[i]f HAP metals are invariably present in ...PM,
then even if the ratio of metals to PM is small and variable, or simply
unknown, PM is a reasonable surrogate for the metals-assuming ... that
PM control technology indiscriminately captures HAP metals along with
other particulates.” Id. Additionally, we believe that PM controls are
still effective for chromium emissions at the same efficiency as total
PM and other metal HAP.  However, we believe that in the case of
gas-fired furnaces a separate chromium limit is necessary in order to
insure control of chromium emissions.

In general, the PM emissions from glass-melting furnaces result from the
entrainment of small particles of the feed in the exhaust gas.  The feed
materials going into the furnace contain trace amounts of chromium and
other HAP metals.  Therefore, the resulting PM also contains HAP metals
(including chromium) in trace amounts.  However, there is also a second
particulate source which is the refractory above the glass melt line of
the furnace. Due to the chemical and thermal properties of the gases, a
corrosive environment exists within a gas-fired furnace. This causes
erosion of the refractory lining of the furnace, which releases
chromium-bearing refractory particles into the furnace exhaust gas as
PM. 

All electric furnaces are lined with chromium refractories at and below
the glass melt line. The chromium refractory in an electric furnace
extends up to two inches above the melt. Gas-fired furnaces may
construct the entire furnace (crown, sidewalls, front and back walls,
forehearth, etc.) from chromium refractories. In the case of gas-fired
furnaces, the refractory does not contain chromium in trace amounts,
rather chromium compounds make up a significant portion of the
refractory composition above the glass melt line.  This results in
gas-fired furnaces emitting particulate that contains chromium in larger
amounts than that of electric furnaces. Based on our dataset, chromium
constitutes an average of 0.96 percent of PM emissions for gas-fired
furnaces, which is 13 times higher than the average for electric
furnaces (0.07 percent of PM emissions are chromium). 

As an example, if a furnace is emitting 100 pounds per year (lb/yr) of
uncontrolled PM and has a 99.9 percent efficient control device; assume
that total metal HAP are about 1 percent of the uncontrolled PM, and
chromium makes up 10 percent of the metal HAP.  Controlled PM emissions
would be 0.10 lb/yr of PM, and controlled chromium emissions would be
0.0001 lb/yr. However, if the 100 lb/yr of PM came from the chrome
refractory located above the glass melt line, the chromium content might
be as high as 50 percent of the PM.  This would result in the same level
of controlled PM emissions, 0.10 lb/yr, but controlled chromium
emissions of 0.05 lb/yr. This is a 500 fold increase in chromium
emissions but no increase in total PM emissions. 

Because chromium compounds are a significant component of the refractory
used above the glass melt line, a greater potential for chromium
emissions exists, which is not associated with a commensurate increase
in emissions of other HAP metals.  Historical test data show there can
be a delay in the increase of chromium emissions due to the wear of the
furnace refractory. Table 1 shows measured total chromium emissions over
one campaign of a single oxyfuel furnace. The emissions testing
conducted in 2010 for the furnace summarized in Table 1 showed that 93
percent of the chromium was in the hexavalent state (i.e., Cr+6).
Because the increase in chromium emissions may occur as a result of
numerous variables, and may be closely related to specific furnace
design and maintenance elements, we believe it is appropriate to set a
chromium limit in addition to the PM limit.

Table 1. Summary of Chromium Emissions from 2004 – 2010

Year	Chromium Emissions from Highest Emitting Glass-Melting Furnace 

at permitted production rate, pounds per year

2004	<5

2005	30

2008	114

2010	540



Additionally, emissions testing conducted at wool fiberglass furnaces
raise pollutant-specific issues.  Emissions test data for glass-melting
furnaces at wool fiberglass manufacturing facilities indicated that the
phenomena discussed above can result in emissions levels of chromium up
to 50 lb/year, and in one case as high as 540 lb/year from gas-fired
furnaces, while emissions from electric furnaces at similar PM emissions
levels are below 5 lb/year. Table 2 shows a comparison of the average
and maximum total chromium emissions between electric and gas-fired
glass-melting furnaces. Chromium has been identified as one of the urban
air toxics based on its prevalence and toxicity in urban areas. 

Table 2. Average Annual Total Chromium Emissions for Glass-melting
Furnaces1

	Average  Chromium Emissions per Furnace (lb/yr)	Maximum  Chromium
Emissions from a Furnace (lb/yr)

Electric Furnaces 	0.38	1.7

Gas-fired Furnaces	42	540

Based on furnaces for which PM and chromium emissions testing was
submitted

In lieu of setting both a PM and chromium limit, we considered setting a
chromium limit alone. However, for all the reasons noted above (i.e. the
fact that the PM controls are still effective, but the chromium content
of the particulate is increased) we believe it would be difficult to
improve the efficiency of the PM controls to meet a chromium limit based
on GACT. 

We also note that the control technology that would be required to meet
the proposed chromium emissions limit is a packed bed sodium hydroxide
(NaOH) scrubber. However, this type of control is not effective in
reducing PM emissions. Therefore, in establishing only a chromium limit,
PM emissions (and thus emissions of other HAP metals) would be left
essentially uncontrolled.

V. Determination of GACT for Glass-Melting Furnaces

In order to set PM and chromium emission control levels that are
equivalent to GACT, the EPA evaluated the methods, practices and
techniques that are commercially available for gas-fired glass-melting
furnaces. As shown in Figure 1, most gas-fired furnaces use an ESP
and/or baghouse to control PM and chromium emissions.  The basic
approach to determining GACT was to evaluate the available data on
controlled emissions from gas-fired glass- melting furnaces and then
determine the control levels that appear to be generally available for
reducing emissions of chromium and PM from gas-fired furnaces.  Test
data from facility emission test reports were compiled into a
spreadsheet and sorted by pollutant and furnace type. Figure 2 shows the
emission rate test data that were provided by facilities for chromium
compounds. 

EPA reviewed these data and then used production data for each furnace
to estimate the annual PM and chromium compound emissions. There are no
differences between furnaces located at area sources and major sources.
The reason is that the differences between major and area sources relate
to emissions from the binding process, not to emissions or emissions
controls on the furnaces. Therefore we looked to the entire data set to
develop options for GACT.

In the case of PM we evaluated several options as potential GACT. The
first option was 0.33 pounds PM per ton of glass pulled This are the
same PM limit proposed for major source furnaces as a result of
technology review performed under 112(d)(6). Given the fact that major
and area source furnaces are no different, this was a logical option to
evaluate and we believe represents the least stringent option we should
consider. This level would result in no costs or emission reductions
because all furnaces currently meet this limit. However, this limit
would codify the current level of PM emissions and prevent any future PM
emissions increase.

We also looked evaluated potentially lower limits as GACT. These were
0.2, 0.1, and 0.05 lb PM/ton, and at a level of control equal to MACT of
0.0093 lb/ton glass for existing sources and 0.0018 lb/ton glass for new
sources. At these levels 5 to 12 furnaces would have to invest in
additional add-on PM emissions control. We estimate these costs would
exceed $50,000 per ton of PM reduction for the first three options. In
the case of the MACT floor levels of control, the cost of PM reductions
would be $480,000 per ton PM removal ($200,000 per furnace for a total
of $2,400,000 for 12 furnaces that currently do not meet the limit to
remove five tons of PM annually). These cost effectiveness values are
significantly higher than any the Agency has accepted as reasonable for
PM. 

For chromium the first GACT option evaluated was 0.00006 lb/ton glass
pulled. As was the case with the PM limit, this is the same chromium
limit proposed for major source furnaces as a result of technology
review performed under 112(d)(6). There are three area source gas-fired
furnaces that currently do not meet the proposed GACT for chromium.
However, data are available for industries with similar control
requirements that demonstrate that there are effective chromium control
technologies available. We searched other industries for controls that
would remove chromium and found that a sodium hydroxide (NaOH) scrubber
is used in both high temperature metallurgical industries and in the
chromium electroplating industry for removal of hexavalent chromium.
Based on the effectiveness of this technology on to two different types
of exhaust gas streams, we believe this control technology is
transferable to wool fiberglass furnaces. Though there are currently no
NaOH scrubbers applied in the wool fiberglass industry, there is
currently one gas-fired furnace equipped with a PM control followed by a
wet scrubber for SO2 control. This is directly analogous to using a NaOH
wet scrubber downstream of the PM controls to achieve additional
chromium removal. 

Assuming that the facilities not currently meeting the proposed chromium
emission limit opted to use the NaOH scrubbers to achieve compliance,
the cost of the proposed chromium emissions limit is $7,600 per pound of
chromium. This is a reasonable cost given that chromium is an urban air
toxic and that a significant portion of the chromium emitted from
gas-fired glass-melting furnaces is hexavalent chromium, which is
extremely toxic and carcinogenic even in low amounts. We note that we
found $11,000 per pound chromium removed to be a reasonable cost in the
final Chromium Electroplating RTR rulemaking, where we regulated
chromium compounds (77 FR 59220, September 19, 2012). 

For chromium we evaluated a MACT floor level emissions limit of 0.000022
pounds (lb) of chromium compounds per ton of glass pulled (0.022 lb per
thousand tons glass pulled) for new and existing sources. In this case
of the MACT floor levels of control, the cost per pound of chromium
removed would be $34,400 annually ($250,000 for 10 furnaces that
currently do not meet the limit to remove 75 pounds of chromium
compounds annually). Therefore, the cost-effectiveness of achieving the
chromium emissions levels would be about five times higher than the
first option GACT standard which we are proposing. These cost
effectiveness values are significantly higher than any the Agency has
accepted as reasonable for chromium. We also considered some level
between 0.00006 and 0.000022, but we noted that any level below 0.00006
would incur similar costs and emissions reductions. 

In summary, we found the cost effectiveness of the proposed GACT limits
to be reasonable. We evaluated more stringent emissions limits that
those proposed, but the overall emission levels would not be
significantly reduced (because most of the area sources are relatively
low emitters), and the cost and cost-effectiveness of achieving these
emissions would be significantly higher than the cost of proposed GACT
standards without significant added environmental benefit. In addition,
and more significantly, setting more stringent limit that those being
proposed this would result in a situation where area source gas-fired
glass-melting furnaces would be subject to stricter regulations than the
major source gas-fired glass-melting furnaces, which would not be
appropriate given that there are no discernible differences between area
and major source furnaces.

References: 

D

£

©

h

h

h

h

h

h

h

h

Metallurgical and Materials Transactions, 1998. 

Wool Fiberglass Insulation Manufacturing. Background Information for
Proposed Standards. December 1983.  

Minutes of EPA Meeting with Representatives of the Wool Fiberglass
Industry and NAIMA. August 31, 2011.

Notes from Steffan Johnson on Caustic Scrubbing of Hexavalent Chromium
Emissions from Industrial Sources. October 26, 2011.

 Information obtained during site visits in December 2012 indicated that
differences in furnace designs, placement of refractory products,
composition and manufacturing specifications of refractory products,
differences in installation of refractory products, and composition of
refractory patches used in “hot” repairs all may influence chromium
emissions from gas-fired furnaces. 

 NaOH Scrubber Information. Telephone discussion and emails between
vendors, companies and EPA. Steffan Johnson, Measurement Policy Group.
USEPA/OAQPS/SPPD. 

  PAGE   \* MERGEFORMAT  1 

