www.erg.

com

MEMORANDUM

TO: Andrew Bouchard, U.S. EPA/OAQPS/SPPD – EPA Office of Air Quality

Planning and Standards

FROM: Eastern Research Group, Inc.

DATE: June 2019

SUBJECT: Assessment of Work Practice Standards for Ethylene Cracking Furnace Decoking
Operations Located in the Ethylene Production Source Category

1.0 INTRODUCTION
This memorandum addresses work practice standards under CAA sections 112(d)(2) and
(3) and CAA section 112(h) (due to the removal of the SSM exemption) for ethylene cracking
furnace shutdowns related to decoking operations. This memorandum reviews control options
and estimates the costs and environmental impacts for the options identified.

2.0 BACKGROUND
The Environmental Protection Agency (EPA) promulgated the final Generic Maximum
Achievable Control Technology (GMACT) standards on June 29, 1999; however these standards
did not include national emission standards for hazardous air pollutants (NESHAP) for the
ethylene production source category. 1 The current NESHAP for the ethylene production source
category was proposed to be included in the GMACT standards on December 6, 2000 (65 FR
76408) and was promulgated on July 12, 2002 (67 FR 46258). 2 As promulgated in 2002 and
further amended on April 13, 2005 (70 FR 19266), the NESHAP includes maximum achievable
control technology (MACT) standards that apply to affected sources of hazardous air pollutant
(HAP) emissions at ethylene production facilities.

The ethylene production MACT standards of 40 CFR part 63, subparts XX and YY,
regulate HAP emissions from an ethylene production unit at major sources. The ethylene
production unit is defined as a chemical manufacturing process unit in which ethylene and/or
propylene are produced by separation from petroleum refining process streams or by subjecting
hydrocarbons to high temperatures in the presence of steam. 3

1
The original proposed GMACT included NESHAP for four small source categories: acetal resins production,
acrylic and modacrylic fiber production, hydrogen fluoride production, and polycarbonate(s) production.
2
Codified at 40 CFR part 63, subparts XX and YY.
3
The ethylene production unit includes the separation of ethylene and/or propylene from associated streams such
as product made from compounds composed of four carbon atoms (C4), pyrolysis gasoline, and pyrolysis fuel
oil. Ethylene production does not include the manufacture of synthetic organic chemical manufacturing industry

1
 As of January 1, 2017, there were 26 ethylene production facilities that are major sources
of HAP emissions in operation that are subject to the ethylene production MACT standards. We
are also aware that several of these facilities are in the process of expanding their operations; and
that new facilities are being constructed. Thus, based upon our review of the anticipated growth
for the ethylene production source category, we anticipate that a total of 31 ethylene production
facilities are major sources of HAP emissions that will ultimately be subject to the residual risk
and technology review (RTR) rulemaking. Appendix A of the memorandum entitled “Review of
the RACT/BACT/LAER Clearinghouse Database for the Ethylene Production Source Category”
in Docket ID No. EPA-HQ-OAR-2017-0357 provides a list of facilities located in the United
States that are considered to be major sources of HAP and part of the ethylene production source
category (ERG, 2018a).

Section 112 of the CAA requires the EPA to establish technology-based standards for
listed source categories as NESHAP. These technology-based standards are often referred to as
MACT standards. At the time of promulgation of the ethylene production MACT standards, the
EPA determined that the decoking of an ethylene cracking furnace (i.e., the coke removal
activity that takes place inside the radiant tubes, discussed in section 4.2) is a shutdown activity
and is required to be addressed in a facility’s startup, shutdown, and malfunction plan. However,
in its 2008 decision in Sierra Club v. EPA, 551 F.3d 1019 (D.C. Cir. 2008), the United States
Court of Appeals for the District of Columbia Circuit vacated portions of two provisions in the
EPA’s CAA section 112 regulations governing the emissions of HAP during periods of SSM.
Specifically, the Court vacated the SSM exemption contained in 40 CFR 63.6(f)(1) and 40 CFR
63.6(h)(1), holding that under section 302(k) of the CAA, emissions standards or limitations
must be continuous in nature and that the SSM exemption violates the CAA’s requirement that
some section 112 standards apply continuously. Therefore, in this memorandum, we conduct a
review of alternative standards that could apply to ethylene cracking furnace shutdowns related
to decoking operations under CAA section 112(h) due to the removal of the SSM exemption.

3.0 SOURCES OF AVAILABLE INFORMATION

To identify alternative standards that could apply during shutdowns related to decoking
operations, we consulted facility responses to the EPA’s CAA section 114 request 4 and other
literature.

(SOCMI) chemicals such as the production of butadiene from the C4 stream and aromatics from pyrolysis
gasoline. For purposes of discussion in this memorandum, the term ethylene will be used to describe the source
category and the associated process unit equipment even though other products, such as propylene, may be
produced in addition to and in greater or lesser quantities than ethylene.
4
In July 2014, the EPA issued a request, pursuant to CAA section 114, to collect information from ethylene
production facilities. Available information was requested regarding process equipment, control technologies,
point and fugitive emissions, and other aspects of facility operations. Companies completed the survey for their
facilities and submitted responses (and follow-up responses) to the EPA between November 2014 and September
2015. Additionally, the EPA requested that the facilities provide historical monitoring and compliance data,
conduct emissions source sampling for certain pollutants, and perform stack tests for certain pollutants in 2016
for heat exchange systems and ethylene cracking furnaces under both normal operation as well as during
decoking operations. The results of these tests were submitted to the EPA in the fall of 2016 and early 2017. For
more information regarding the section 114 request, please refer to the memorandum entitled “Data Received
From Information Collection Request for the Ethylene Production Source Category,” (ERG, 2018b).

2
 This exercise revealed alternative standards that could be applied to ethylene cracking
furnace shutdowns related to decoking operations located in the ethylene production source
category, which are discussed further in section 5.0.

4.0 DESCRIPTION OF CRACKING FURNACE OPERATIONS AND EMISSION

RELEASES

4.1 Normal Operations

The ethylene production process involves converting large hydrocarbon molecules into
smaller molecules through a process referred to as “thermal cracking.” This takes place in the
ethylene cracking furnace.

An ethylene cracking furnace is designed to subject certain hydrocarbon feedstocks (i.e.,

ethane, propane, butane, naptha, or gas oils) to high temperatures in the presence of steam to
“crack” the feedstock (break the feedstock molecules apart). Figure 1 provides a diagram of a
typical ethylene cracking furnace.

Figure 1. Ethylene Cracking Furnace Diagram (normal operations)

The feedstock travels through the furnace via piping (or tubing) and never comes into
direct contact with the fuel being burned in the furnace firebox. It is delivered to the piping in its
gas or liquid state (e.g., the heavier feedstocks are delivered in liquid form). The feedstock is first
passed through the top portion of the furnace (called the “convection” section) for preheating;
steam is added to the feedstock only after the feedstock has travelled through a portion of the
piping (the steam is called ‘diluted steam’ because it acts as a diluting agent to lower the partial

3
pressure of the feedstock and keep the feedstock molecules from recombining once broken
apart). The feedstock/steam mixture then passes through the bottom of the furnace firebox where
the “cracking” occurs (inside radiant tubes; the piping that is positioned vertically inside the
furnace firebox is typically referred to as “radiant tubes”). Effluent cracked gas from each
furnace is passed through one or more shell and tube heat exchangers and aggregated into a
cracked gas header via a system of transfer line valves prior to downstream operations.

4.2 Decoking Operations

As feedstock passes through the radiant tubes, a layer of carbon (i.e., coke) builds up on
the interior of the tubing over time, forming a physical restrictive barrier inside the tubing.
Because of this build up, the tubing will gradually get hotter during the cracking process (i.e., the
temperature of the tubing typically increases by 3 to 4 °F per day even with a constant firebox
temperature because the coke acts as an insulator on the tubing). When the temperature of the
tubing gets to about 2,000 °F, the furnace must be taken out of production so the coke can be
removed (i.e., burned out) of the tubing (this is known as decoking). These coke deposits are so
severe that the cracking furnace must be taken off-line and “decoked” about every 20 to 60 days
(McKimpson, 2006).

Prior to decoking, the fuel firing rate of the ethylene cracking furnace is reduced, and the
hydrocarbon feedstock is stopped, leaving steam as the only stream being sent through the piping
(including radiant tubes). During this time the piping continues to be purged of any remaining
feedstock using steam, and this purge stream is sent downstream to the cracked gas header. After
all feedstock is purged from the radiant tubing (this is determined by examining a sample from a
bleed line located just prior to the downstream operations), the furnace is isolated from the
process, the radiant tubing effluent is directed to the decoke header and either diverted to a large
cyclone separation device called a “decoking pot” (see Figure 2) or back into the firebox, and
oxygen (i.e., air) and steam is gradually added to the inside of the radiant tubing until the coke
ignites. The coke burn is then typically monitored via the radiant tube outlet temperature (also
referred to as coil outlet temperature) and/or by monitoring carbon dioxide (CO2) levels in the
decoke header stack, where adjustments to the air and/or steam are made based on these
parameters to ensure good control of the coke burn. Once the burn is complete (e.g., a decrease
in CO2 levels indicates that combustion of the coke inside the radiant tubing is slowing or
stopping; eventually, this indicates when there is no longer any coke build up inside the radiant
tubing (this is also checked manually)), the furnace can be placed back into normal operation.

4
 Figure 2. Ethylene Cracking Furnace Diagram (decoking operations)
(Titan Engineering, Inc. 2012)

4.3 Decoking Operations Emission Release Points

Depending on design configuration, there are often (but not always as explained below)
two emission release points associated with ethylene cracking furnace decoking emissions:

• Furnace firebox stack: If a decoking pot is not used during decoking operations, then
during decoking operations, the coke combustion emissions and steam from inside the
radiant tubing is routed directly back into the furnace firebox near the burners, and
the furnace firebox stack will include a combination of fuel and coke combustion
emissions along with steam from the radiant tubing.

• Decoking pot stack: If a decoking pot is used during decoking operations (see Figure
2 for this type of configuration), then the combination of coke combustion emissions
and steam from inside the radiant tubing are routed through the decoking pot stack
(and fuel combustion emissions from the gas being combusted in the furnace firebox
are routed through the furnace firebox stack). The decoking pot stack is not equipped

5
 with APCDs for HAP, and it vents directly to the atmosphere. 5 The bottoms, or
separated coke fines, that separate from the gases and fall out of the bottom of decoke
pot are either: sold as fuel, sent to a sump where a lift station transports it to a
WWTP, or disposed of as a non-hazardous material. Most ethylene cracking furnaces
(i.e., ~70 percent based on CAA section 114 responses) have a decoking pot stack
configuration.

5.0 WORK PRACTICE STANDARDS FOR ETHYLENE CRACKING FURNACE

SHUTDOWNS RELATED TO DECOKING OPERATIONS
The emissions stream generated from decoking operations (i.e., the combination of coke
combustion emissions and excess additional steam added to the radiant tubing) is very dilute
with a high moisture content (e.g., generally >95% water) and the EPA, as part of its CAA
section 114 request, required companies perform testing for HAP from this emissions source at
certain ethylene cracking furnaces. A minimum of three decoking cycles were required to be
tested; and emissions data were obtained for a minimum of three valid test runs equally spaced
over each decoking cycle. The test data collected from industry confirm HAP emissions occur
during decoking operations, although more than half of the data are below detection limit (BDL).
Specifically, the test methods employed for organic HAP that are considered non-polycyclic
aromatic hydrocarbons (non-PAHs) are non-isokinetic in nature, and the majority (i.e., 88%) of
non-PAH organic HAP were found to be BDL.

The test methods employed for PAHs and metal HAP have isokinetic sampling
requirements. Isokinetic sampling is required for any method where compounds may exist in a
particle or aerosol phase in order to collect a representative sample with respect to a flow
weighted average concentration and particle or aerosol size distribution. Without an appropriate
isokinetic sample collected to represent the flow weighted average concentration of these
particular classes of pollutants, the data may be biased and unreliable for compliance
demonstrations. While the CAA section 114 test data shows that PAHs and metal HAP are
emitted, the majority of the test runs do not meet the requirements in the isokinetic sampling test
methods used which require samples to be within +/-10% of isokinetic. The EPA required the
emissions of PAHs and metal HAP from the decoking pots to be tested for semiqualitative
purposes; however, the EPA was aware that it would be extremely difficult for facilities to meet
the 10% isokinetic requirement of the sampling methods during the majority of a decoke cycle.
In order to pull a sample in an isokinetic manner, the tester must have knowledge about the large
components of the gas stream such as moisture, oxygen, and carbon dioxide. When a gas stream
is nearly pure moisture (e.g., >90% moisture), even slight deviations in the assumed moisture can
cause large changes in the flow through the sampling nozzle, which is controlled through dry gas
measurements. For example, an assumed gas stream moisture content of 97% with a true gas
stream moisture content of 98% would cause the isokinetic rate to be off by around 30%. The
same margin of error in moisture assumption at 10%-20% gas stream moisture content (normal
combustion levels) would only cause the isokinetic rate to be off by a couple of percent. This
thin margin of error for moisture assumption makes it difficult to achieve required isokinetic

5
Although uncommon, the agency is aware of one facility where coke combustion gases are routed through a
decoking pot stack and then routed directly back into the furnace firebox before being released to the
atmosphere.

6
rates at these high moisture conditions. Because it is technically impracticable to achieve
representative and precise emissions measurement samples for PAHs and metal HAP for
decoking operations, work practice standards are appropriate. 6

Owners and operators of ethylene cracking furnaces already have an incentive to

minimize decoking events (i.e., owners and operators desire to reduce coke formation inside the
radiant tubes). Decoking operations are undesirable for the following reasons:

• High Operating and Maintenance Costs:

o The coke deposits of a few millimeters to centimeters in thickness lead to poor
heat transfer. To retain the same process temperature and hence the same cracking
conversion, owners and operators must raise the cracking furnace firebox
temperature continuously (e.g., burn more fuel), which often leads to more rapid
coke formation. (Haiyong, 2002)
o Shortened life of the radiant tubes because of the constant thermal cycling of the
tubes. (Haiyong, 2002)
o Excessively high temperature during decoking cycles is assumed to be the most
important cause of radiant tube failure. (Tari, 2009)

• Loss of Ethylene Production:

o The coke build-up also increases the pressure drop, which results in lower
ethylene yield. (Haiyong, 2002)
o With time, the accumulation of coke forces the operator to shut down the cracking
furnace due to temperature or pressure drop. The furnace is, therefore, taken
offline for coke removal (decoking). (Haiyong, 2002)

Moreover, according to startup, shutdown, and malfunction plans submitted to the EPA
in response to the CAA section 114 request (many of these plans are considered CBI; therefore,
are not provided in this memorandum); and, according to discussions with stakeholders (see
Appendices A, B, C, and E), owners and operators conduct one or more of the following work
practices:

• To minimize coke formation during normal operations thus prolonging radiant tube
life (and minimizing coke combustion emissions):
o Apply a coating material to the interior of the radiant tube to protect against coke
formation inside the radiant tubes.
o Conduct a daily firebox inspection during normal operations and address any
burners that are impinging on the radiant tubes.
o Document in furnace operating procedures, steps to expedite air removal
following coke removal (e.g., steam purge). Determination of the completion of
decoking operations and the corresponding air removal should be verified and
recorded. Air is allowed during decoking operations.

6
CAA section 112(h)(1) states that the Administrator may prescribe a work practice standard or other
requirements, consistent with the provisions of CAA sections 112(d) or (f), in those cases where, in the judgment
of the Administrator, it is not feasible to enforce an emission standard. CAA section 112(h)(2)(B) further defines
the term “not feasible” in this context to apply when “the application of measurement technology to a particular
class of sources is not practicable due to technological and economic limitations.”

7
 • To minimize coke combustion emissions during decoking operations:
o Monitor radiant tube outlet temperature (also referred to as coil outlet
temperature) during decoking operations.
o Monitor CO2 at radiant tube outlet during decoking operations.
• To minimize the introduction of additional sources of HAP during the isolation of the
decoking operations from the ethylene production process: 7
o Document in an isolation plan, that the owner or operator has confirmed that prior
to routing the decoke effluent to the decoking header, the furnace is properly
isolated from the process where additional sources of emissions from other
normally operating cracking furnaces cannot be introduced; and prior to returning
the furnace to normal operation, ensuring that the furnace is properly isolated
from the decoking header where additional sources of emissions (e.g., cracked
gas) from that furnace that is now back normally operating cannot be released
through the decoke stack. If poor isolation is identified, then the owner or
operator will make repairs as soon as practical.

Based on this information, we evaluated what the best performers are doing with respect
to work practice standards for minimizing emissions from decoking operations of ethylene
cracking furnaces. These work practices consist of requiring owners and operators to:

• conduct daily inspections of the ethylene cracking furnace firebox burners for flame
impingement and repair any such flame impingements as described in (a) below, to
minimize coke formation (and hence reduce coke combustion emissions by reducing
the need to commence decoking of the radiant tube(s) in each ethylene cracking
furnace as frequently);
• use at least two of control measures described in (b) through (e) below, to minimize
coke combustion emissions from the decoking of the radiant tube(s) in each ethylene
cracking furnace; and
• conduct ethylene cracking furnace isolation valve inspections, both prior to decoking
the radiant tube and prior to returning the ethylene cracking furnace to normal
operations, to minimize the introduction of additional sources of HAP during the
isolation of the decoking operations from the ethylene production process.

a) During normal operations, inspect the firebox burners daily and repair all burners that
are impinging on the radiant tube(s) as soon as practical, but not later than 1 calendar
day after the flame impingement is found. An inspection may include, but is not
limited to: visual inspection of the radiant tube(s) for localized bright spots (this may
be confirmed with a temperature gun), use of luminescent powders injected into the
burner to illuminate the flame pattern, or identifying continued localized coke build-
up that causes short runtimes between decoking cycles. A repair may include, but is
not limited to: taking the burner out of service, replacing the burner, adjusting the
alignment of the burner, adjusting burner configuration, making burner air

7
During a 2013 investigation (see Appendix D), TCEQ staff documented that a facility had released more than
800 tons of volatile organic compounds (including more than 20 tons of 1,3-butadiene) to the atmosphere
through a decoking pot because two motor operated valves remained partially open following a decoking cycle,
allowing loss of process gases during normal operation.

8
 corrections, repairing a malfunction of the fuel liquid removal equipment, or adding
insulation around the radiant tube(s).
b) During decoking operations, continuously monitor (or use a gas detection tube every
hour to monitor) the CO2 concentration at the radiant tube(s) outlet for indication that
the coke combustion in the ethylene cracking furnace radiant tube(s) is complete. The
owner or operator must immediately initiate procedures to stop the decoking cycle
once the CO2 concentration at the radiant tube(s) outlet consistently reaches a level
that indicates combustion of coke inside the radiant tube(s) is slowing or stopping.
c) During decoking operations, monitor the temperature at the radiant tube(s) outlet to
ensure the coke combustion occurring inside the radiant tube(s) is not so aggressive
(i.e., too hot) that it damages either the radiant tube(s) or ethylene cracking furnace
isolation valve(s). The owner or operator must immediately initiate procedures to
reduce the temperature at the radiant tube(s) outlet once the temperature reaches a
level that indicates combustion of coke inside the radiant tube(s) is too aggressive.
d) After decoking, but before returning the ethylene cracking furnace back to normal
operations, purge the radiant tube(s) with steam and verify that all air is removed.
e) After decoking, but before returning the ethylene cracking furnace back to normal
operations, apply a coating material to the interior of the radiant tube(s) to protect
against coke formation inside the radiant tubes during normal operation.

The work practice standards described above will foster less frequent decoking events by
reducing coke formation, minimize emissions during coke combustion, and will ensure
additional sources of HAP are not introduced when the ethylene cracking furnace is both isolated
from the process for decoking and is brought back online for normal operation after decoking.
Thus, given that these work practice standards cover all facets of emissions generation potential
from decoking operations, we believe the option we presented represents what the best
performers are doing to control HAP emissions from decoking operations.

6.0 COST AND ENVIRONMENTAL IMPACTS

In this section, we evaluate the cost and environmental impacts for the ethylene cracking furnace
decoking operations work practice standards that we identified in section 5.0.
6.1 Emission Reductions
We assumed that implementing the work practice standards would result in half of the
ethylene cracking furnaces being able to conduct one less decoking event per year, and
calculated the corresponding HAP emission reductions. Based on the CAA section 114
responses, the average decoking event is 40 hours and there are an average of 10 decoking events
per furnace per year. Facilities would have two general options during the 40 hour period during
which a furnace is no longer conducting decoking operations: (1) not operate the furnace (and
conduct maintenance activities, for example), or (2) continue to operate the furnace under normal
operations in order to produce more ethylene and/or propylene. Emission reductions are
calculated for both options and are spread evenly among all facilities (i.e., rather than half the
facilities) to calculate an overall average emissions reduction for the source category. This was
done by assuming all furnaces reduce their emissions from decoking operations by 0.5 decoke
cycles per year (rather than randomly assuming half of the furnaces reduce their emissions from
one decoke cycle) given that we don’t have sufficient information to know which furnaces would

9
specifically be impacted by these work practices. Under option 1, emission reductions are
attributable to the reduction in emissions from decoking operations only, whereas for option 2,
we subtract the emission reductions estimated from option 1 from the additional emissions that
would result from the fuel combustion activities that would occur in the ethylene cracking
furnace firebox during normal operations.

Source category average hourly HAP emission rates for ethylene cracking furnace normal
operations and decoking operations were estimated during the RTR Modeling File development. 8
Separate emission rates were provided for decoking through a decoking pot and decoking back
into the firebox, and we calculated a weighted average emission rate for decoking operations.
The weighted average emission rate accounts for the fact that 70% of ethylene cracking furnaces
use decoking pots and 30% decoke back into the firebox. The hourly HAP emission rates are
summarized in Table 1 and the calculated emission reductions are presented in Table 2.

Table 1. Average Hourly HAP Emission Rate for Ethylene Cracking

Furnace Normal and Decoking Operations (lb/hr per furnace)
Cracking Furnace Normal Cracking Furnace Decoking
Operations Operations
1.4 1.8

Table 2. HAP Emission Reductions Due to Implementing Work Practice Standards

HAP Emission HAP Emission
Cracking Reductions - Do Not Reductions - Conduct
Facility
Furnaces Operate the Furnace Normal Operations
(Option 1) (tpy) (Option 2) (tpy)
Chevron Phillips Chemical Company L.P. Cedar Bayou
22 0.39 0.081
Facility
Chevron Phillips Chemical Company L.P. Port Arthur
16 0.28 0.059
Facility
Chevron Phillips Chemical Sweeny Complex 25 0.44 0.093
Dow Chemical Texas Operations 23 0.41 0.085
The Dow Chemical Company Louisiana Operations 38 0.67 0.14
Union Carbide Corp. - Olefins 1 and 2 Plants 21 0.37 0.078
Eastman Chemical Company, Texas Operations 19(c) 0.34 0.070
Equistar Chemicals, LP: Channelview Complex 34 0.60 0.13
Equistar Chemicals, LP: Clinton Complex 13 0.23 0.048
Equistar Chemicals, LP: Corpus Christi Complex 15 0.27 0.056
Equistar Chemicals, LP: La Porte Complex 11 0.20 0.041
Equistar Chemicals, LP: Morris Complex 14 0.25 0.052
ExxonMobil Baton Rouge Chemical Plant 10 0.18 0.037
ExxonMobil Baytown Olefins Plant 27(d) 0.48 0.100
ExxonMobil Beaumont Chemical Plant (BMCP) 14 0.25 0.052
Formosa Plastics Corporation, Texas 31(e) 0.55 0.12
Sasol Chemicals (USA) LLC 15(d) 0.27 0.056
Shell Chemical LP -- Deer Park Chemical Plant 12 0.21 0.044
Shell Chemical LP -- Norco Manufacturing Complex 24 0.43 0.089
Westlake Polymers, LLC - Ethylene Manufacturing
13 0.23 0.048
Complex

8
See Table 6 within Appendix 1 of the memorandum entitled “Residual Risk Assessment for the Ethylene
Production Source Category in Support of the 2019 Risk and Technology Review Proposed Rule” in Docket ID
No. EPA-HQ-OAR-2017-0357.

10
 Table 2. HAP Emission Reductions Due to Implementing Work Practice Standards
HAP Emission HAP Emission
Cracking Reductions - Do Not Reductions - Conduct
Facility
Furnaces Operate the Furnace Normal Operations
(Option 1) (tpy) (Option 2) (tpy)
Westlake Vinyls Inc. 6 0.11 0.022
BASF Fina Petrochemicals - BASF Total Petrochemicals (a) 10(c) 0.18 0.037
DuPont - Sabine River Works(a) 10(c) 0.18 0.037
Flint Hills Resources - Port Arthur Chemicals(a) 9(c) 0.16 0.033
Huntsman Corporation - Port Neches Plant(a) 9(c) 0.16 0.033
INEOS - Chocolate Bayou Plant (Olefins Business Unit)(a) 17(c) 0.30 0.063
Formosa Petrochemical Corporation, St James Parish, LA(b) 28(c) 0.50 0.10
Indorama Ventures Olefins LLC, Lake Charles, LA(b) 8(c) 0.14 0.030
LACC LLC, Lake Charles, LA(b) 7(c) 0.12 0.026
Occidental Chemical Corporation, Ingleside, TX(b) 5(c) 0.089 0.019
Shell Chemical Appalachia LLC, Monaca, PA(b) 7(c) 0.12 0.026
TOTAL 513 9.1 1.9
(a) Existing facility that did not receive 114.
(b) New facility to be constructed.
(c) Based on air permit or air permit application.
(d) Includes 8 new furnaces not reported in section 114 response due to a recent expansion.
(e) Includes 9 new furnaces not reported in section 114 response due to a recent expansion.

Based on the analysis presented in Table 2 above, we project that anywhere from 1.9 -
9.1 tons/yr of HAP will be reduced from implementation of the work practice standards for
decoking operations. Given that most facilities would likely elect to use the additional time to
make more product, option 2 presented in Table 2 above is likely most reflective of the actual
HAP emissions reductions that will be realized (i.e., 1.9 tons/yr HAP).

6.2 Costs
This section provides the costs associated with the work practice standards discussed in
section 5.0.

As discussed in section 5.0, daily firebox inspections for flame impingement in ethylene
cracking furnaces is a practice that most ethylene facilities are already implementing. However,
we are uncertain of the frequency each ethylene facility conducts these inspections. As such, we
assumed 75% of the facilities (23) perform the inspections daily, 12.5% of the facilities (4) do
the inspections every other day, and the remaining 12.5% of the facilities (4) conduct the
inspections on a weekly basis, and divided these costs evenly among all facilities given the
uncertainties in our data with respect to which facilities fall into which bins. Because all ethylene
facilities perform firebox inspections, we also assumed all ethylene facilities already perform
repairs when flame impingements are detected and that this cost of repair would be an
operational cost that would be incurred by all facilities as part routine maintenance and
equipment replacement (e.g., burner replacement). In addition, we assumed that each inspection
takes 5 minutes per ethylene cracking furnace at a rate of $42.46 per hour. 9 Using these
assumptions, the average cost for conducting a firebox inspection was determined to be $218.50.
Because we are unsure of the frequency that each facility conducts firebox inspections, we

9
Labor rate as reported by U.S. Department of Labor, Bureau of Labor Statistics. Labor rate for May 2016 for
general manufacturing industry sector: 2-digit NAICS code 31 through 33
(https://www.bls.gov/oes/current/oes_nat.htm accessed in February 2017).

11
applied the average cost to all ethylene cracking furnaces across the industry. Table 2 provides
the number of ethylene cracking furnaces at each facility. For all 513 cracking furnaces, we
estimated a total cost for conducting daily firebox inspections to be $112,095.

For the requirement that owners or operators use at least two of the work practices
identified as work practices (b) through (e) in section 5.0 to minimize coke combustion
emissions from the decoking of the radiant tube(s) in each ethylene cracking furnace, we
assumed facilities will not incur any additional expense to implement this portion of the work
practice standard. We assumed that facilities would choose the lowest cost option, and that all
facilities are already using a combination of at least two of the work practice options that would
minimize coke formation during normal operations and minimize emissions during decoking
operations.

With regard to isolation valve inspections, we assumed that the majority of ethylene
facilities are already conducting some form of ethylene cracking furnace isolation valve
inspection prior to decoking the radiant tube(s) (to confirm that the radiant tube(s) being decoked
is completely isolated from the ethylene production process) and prior to returning the ethylene
cracking furnace to normal operations (to confirm that the radiant tube(s) that was decoked is
completely isolated from the decoking pot or furnace firebox). However, given the uncertainties
surrounding the frequency with which these inspections are conducted at a given facility for a
given furnace, to be conservative, we estimated cost impacts for this particular work practice for
most facilities in the source category. We also assumed, for cost impact purposes, that based on a
2003 enforcement case (see Appendix D), Shell Chemical LP owned ethylene facilities are
already conducting these inspections. Using the information received in response to the EPA’s
CAA section 114 request (see section 3.0), we determined that an ethylene cracking furnace is
decoked on average 10 times per year. As such, we assumed each furnace would require 20
isolation valve inspections per year (before and after each decoking event). In addition, we
assumed that each inspection takes 5 minutes per ethylene cracking furnace at a rate of $42.46
per hour.9 We also assumed an isolation valve repair cost of $212 per year per facility (based on
5 hours of repair time per year). Using these assumptions (and the number of ethylene cracking
furnaces at each facility provided in Table 2), the total annual cost for conducting isolation valve
inspections (and performing repairs) was determined to be $39,200.

Table 3 summarizes the costs to implement the work practice standards for ethylene
cracking furnaces during decoking operations.

Table 3. Work Practice Standards Costs

Daily Burner Isolation Valve
Inspection Inspection Isolation Valve Total
Annual Cost Annual Cost Repair Annual Cost
Facility ($/yr) ($/yr) Cost ($/yr) ($/yr)
Chevron Phillips Chemical Company L.P. Cedar
4,807 1,557 212 6,576
Bayou Facility
Chevron Phillips Chemical Company L.P. Port Arthur
3,496 1,132 212 4,841
Facility
Chevron Phillips Chemical Sweeny Complex 5,463 1,769 212 7,444
Dow Chemical Texas Operations 5,026 1,628 212 6,866
The Dow Chemical Company Louisiana Operations 8,303 2,689 212 11,205
Union Carbide Corp. - Olefins 1 and 2 Plants 4,589 1,486 212 6,287
Eastman Chemical Company, Texas Operations 4,152 1,345 212 5,709

12
 Table 3. Work Practice Standards Costs
Daily Burner Isolation Valve
Inspection Inspection Isolation Valve Total
Annual Cost Annual Cost Repair Annual Cost
Facility ($/yr) ($/yr) Cost ($/yr) ($/yr)
Equistar Chemicals, LP: Channelview Complex 7,429 2,406 212 10,048
Equistar Chemicals, LP: Clinton Complex 2,841 920 212 3,973
Equistar Chemicals, LP: Corpus Christi Complex 3,278 1,062 212 4,551
Equistar Chemicals, LP: La Porte Complex 2,404 778 212 3,394
Equistar Chemicals, LP: Morris Complex 3,059 991 212 4,262
ExxonMobil Baton Rouge Chemical Plant 2,185 708 212 3,105
ExxonMobil Baytown Olefins Plant 5,900 1,911 212 8,023
ExxonMobil Beaumont Chemical Plant (BMCP) 3,059 991 212 4,262
Formosa Plastics Corporation, Texas 6,774 2,194 212 9,180
Sasol Chemicals (USA) LLC 3,278 1,062 212 4,551
Shell Chemical LP -- Deer Park Chemical Plant 2,622 0 0 2,622
Shell Chemical LP -- Norco Manufacturing Complex 5,244 0 0 5,244
Westlake Polymers, LLC - Ethylene Manufacturing
2,841 920 212 3,973
Complex
Westlake Vinyls Inc. 1,311 425 212 1,948
BASF Fina Petrochemicals - BASF Total
2,185 708 212 3,105
Petrochemicals (a)
DuPont - Sabine River Works(a) 2,185 708 212 3,105
Flint Hills Resources - Port Arthur Chemicals(a) 1,967 637 212 2,816
Huntsman Corporation - Port Neches Plant(a) 1,967 637 212 2,816
INEOS - Chocolate Bayou Plant (Olefins Business
3,715 1,203 212 5,130
Unit)(a)
Formosa Petrochemical Corporation, St James Parish,
6,118 1,981 212 8,312
LA(b)
Indorama Ventures Olefins LLC, Lake Charles, LA(b) 1,748 566 212 2,527
LACC LLC, Lake Charles, LA(b) 1,530 495 212 2,237
Occidental Chemical Corporation, Ingleside, TX(b) 1,093 354 212 1,659
Shell Chemical Appalachia LLC, Monaca, PA(b) 1,530 0 0 1,530
TOTAL 112,095 33,260 5,944 151,300
(a) Existing facility that did not receive 114.
(b) New facility to be constructed.

7.0 REFERENCES
Haiyong. 2002. Et.al. Coke formation in steam crackers for ethylene production. Chemical
Engineering and Processing 41. p199–214. EPA Docket No. EPA-HQ-OAR-2017-0357.
EPA. 2019. Residual Risk Assessment for the Ethylene Production Source Category in Support
of the 2019 Risk and Technology Review Proposed Rule. Prepared for the U.S. EPA Office
of Air Quality Planning and Standards, Research Triangle Park, NC. EPA Docket No. EPA-
HQ-OAR-2017-0357.
ERG. 2018a. Review of the RACT/BACT/LAER Clearinghouse Database for the Ethylene
Production Source Category. EPA Docket No. EPA-HQ-OAR-2017-0357.
ERG. 2018b. Data Received From Information Collection Request for the Ethylene Production
Source Category. EPA Docket No. EPA-HQ-OAR-2017-0357.

13
McKimpson. 2006. High-Performance, Oxide-Dispersion-Strengthened Tubes for Production of
Ethylene and Other Industrial Chemicals. EPA Docket No. EPA-HQ-OAR-2017-0357.
Tari. 2009. Failure Analysis of Ethylene Cracking Tube. Journal of Failure Analysis and
Prevention. DOI 10.1007/s11668-009-9259-5. EPA-HQ-OAR-2017-0357.
Titan Engineering, Inc. 2012. Permit Application Additional Furnace Project. Prepared for
INEOS USA LLC. Project No. 412-15. EPA-HQ-OAR-2017-0357.

14
 Appendix A

Agenda and Handout for February 27, 2018 Meeting Between EPA, American
Chemistry Council (ACC), and American Fuel & Petrochemical
Manufacturers (AFPM)
 Agenda for ACC/AFPM EPA RTP Meeting
February 27, 2018
1:30 P.M. ET (12:30 CT) – 3:30 P.M. ET (2:30 CT)
Phone Line: 919-541-1958

1) Ethylene MACT Proposal Timeline

a. Ask EPA for projections/progress updates on their timeline for the proposal
2) Risk Modeling Files
a. Ask EPA about availability of input files
3) Furnaces
a. Work Practice Standards Presentation
b. Isokinetic/Sample Size Variance
4) Atmospheric PRDs
a. Discuss data on ACC spreadsheet
b. RMP/PSM Applicability
c. Additional Information to support work practices and related exemptions?
5) Flares
a. Discuss data on ACC spreadsheet
6) Other Updates
a. Process Vents: Maintenance vent 10% LEL / 50 lbs criteria
i. Other updates?
b. Heat Exchange: Mentioned modified El Paso previously
i. Other updates?
c. Fenceline Monitoring: Any updates?
d. Miscellaneous: Tanks, Equipment Leaks, Transfer Racks, Wastewater
7) Next Meeting
a. Schedule date for March meeting

A-1
 ACC/AFPM EPA RTP Meeting
Recommended Work Practices Outline
February 27, 2018
1:30 P.M. ET (12:30 CT) – 3:30 P.M. ET (2:30 CT)
Phone Line: 919-541-1958

1) Category: Burner Operation Work Practice (1)

a) Recommendation: “The owner or operator shall document in the furnace operating
procedures the correct burner configuration allowing for burner maintenance for
operation and the steps required to adjust air to ensure combusting burners with fuel have
adequate air.”

2) Category: Burner Operation Work Practice (2): Impingement

a) Recommendation: “The owner or operator shall conduct a daily firebox inspection and
address any burners that are impinging on the radiant tubes.”

3) Category: Fuel Gas Knock-Out Pot Work Practice

a) Recommendation: “The owner or operator shall conduct a daily inspection to verify the
fuel gas liquid knock-out pot, coalescer or other liquid removal equipment is functioning.
If the equipment is not functioning, it should be repaired as soon as practical.”

4) Category: Burner Health (BPH MACT Req.)

a) Recommendation: “When a furnace is offline and open for entry, inspect the accessible
burners, and clean or replace burner components.”

5) Category: Process Isolation from Decoke Pot

a) Recommendation: “Prior to routing the decoke effluent to the decoke pot the furnace
must be isolated from the process. Prior to returning the furnace to cracking operation
the furnace must be isolated from the decoke pot. The owner/operator shall document an
isolation plan.”

6) Category: Air Removal at End of Decoke

a) Recommendation: “The operating procedure shall include steps to expedite air removal
following coke removal. Air is allowed during startup and shutdown operation.”

A-2
 Appendix B

Agenda and Handout for April 10, 2018 Meeting Between EPA, American
Chemistry Council (ACC), and American Fuel & Petrochemical
Manufacturers (AFPM)
 Agenda for ACC/AFPM EPA RTP Meeting
April 10, 2018
12:30 P.M. ET (11:30 CT) – 2:00 P.M. ET (1:00 CT)
Phone Line: 919-541-1591

1) Ethylene MACT Proposal Timeline

a. Ask EPA for projections/progress updates on their timeline for the proposal
2) EPA Questions
a. Tube Analysis
b. Decoke to the Firebox—location of emissions injected into firebox
3) Furnaces
a. Control Technology Discussion
b. Work Practice Standards Presentation and Discussion
4) Atmospheric PRDs
a. Discussion and Explanation of Certain Performers
5) Flares
a. ACC Flare Analysis
b. Update and Discussion on Potential Requirements
6) Other Updates
a. Storage Tanks
b. Heat Exchange Systems
c. Fenceline Monitoring
d. Miscellaneous: Process Vents, Equipment Leaks, Transfer Racks, Wastewater
7) Next Meeting
a. Timeframe for May meeting

B-1
 April 10, 2018
Furnace Work Practices

Note: black text was the verbiage submitted to EPA in Feb. 2018 meeting. red text is proposed
revision. Revisions were made to include additional "enforceability" per EPA's request in Feb.
2018 meeting or to clarify prior submitted work practice language.

1. "The owner or operator shall document in the furnace operating procedures the correct burner
configuration allowing for burner maintenance for operation and the steps required to adjust
air to ensure combusting burners with fuel have adequate air. After adjustments are made,
optimize total emissions of carbon monoxide consistent with the manufacturer's
specifications, if available, and to any NOx requirement to which the unit is subject.
Measurements may be made using a portable CO analyzer or CO CEMS."

2. "The owner or operator shall conduct a daily firebox inspection and address any burners that
are impinging on the radiant tubes."

3. "The owner or operator shall conduct a daily inspection to verify the fuel gas liquid knockout
pot, coalescer or other liquid removal equipment is functioning. If the equipment is not
functioning, it should be repaired as soon as practical."

4. "During the next scheduled shutdown, when a furnace is offline and open for entry, inspect
the accessible burners, and clean or replace burner components as necessary."

5. "Prior to routing the decoke effluent to the decoke pot the furnace must be isolated from the
process. Prior to returning the furnace to cracking operation the furnace must be isolated
from the decoke pot. The owner or operator shall document an isolation plan. The isolation
confirmation should be recorded. If poor isolation is identified, the owner or operator shall
make repairs as soon as practical."

6. "The operating procedure shall include steps to expedite air removal following coke removal.
Determination of decoke completion and corresponding air removal should be verified and
recorded. Air is allowed during decoking, startup, and shutdown operation."

B-2
 Appendix C

Technical Document on Ethylene Cracking Furnace Test Data and Radiant

Tube Flame Impingement Provided to EPA on July 27, 2018 From American
Chemistry Council (ACC)
 ETHYLENE CRACKING FURNACES: INSUFFICIENCY OF EXISTING STACK TEST DATA
SUPPORTS A WORK PRACTICE APPROACH BASED ON FLAME IMPINGEMENT

The U.S. Environmental Protection Agency (EPA) is undertaking the risk and technology
review (RTR) rulemaking process for the Ethylene Production source category following the
2005 amendments to the Clean Air Act (CAA or the Act) Section 112(d) National Emissions
Standards for Hazardous Air Pollutants (NESHAP or MACT) for Source Categories: Generic
Maximum Achievable Control Technology Standards, 40 C.F.R. Part 63, Subpart YY. 1 Often
referred to as the “Generic MACT,” Subpart YY addresses several source categories, ethylene
production among them.2 In conjunction with the RTR process, EPA is currently evaluating
whether to amend the Ethylene Production MACT to address certain emission points that were
determined not appropriate for regulation in the original MACT due in part to their expected low
Hazardous Air Pollutants (HAP) emission levels.3

This White Paper addresses EPA’s CAA Section 112 authority to revisit the
establishment of a “no control” MACT floor for ethylene cracking furnaces in the Ethylene
Production source category and to establish a work practice standard for Particulate Matter (PM)
and/or metal HAP under Section 112(h), notwithstanding the stack testing data generated
through an Information Collection Request (ICR) issued to a subset of sources in the category.
To analyze available options, it is important to understand the nature and the impact of the data
gathered to date. Options available to EPA may include:

(1) determining not to revise the standard at all because the original MACT’s “no
control” floor was not challenged;
(2) deferring any decision to revise the “no control” floor until the next periodic
technology review under CAA Section 112(d)(6);
(3) issuing a work practice standard now pursuant to a Section 112(d)/(h) analysis; or
(4) issuing a work practice pursuant to CAA Section 112(d)(6)’s technology review.

EPA’s rulemaking process to promulgate MACT requirements, including for the Ethylene
Production source category, begins by analyzing whether a numerical emission limit can be set.
The U.S. Court of Appeals for the District of Columbia Circuit (D.C. Circuit) has recognized that
“[t]he CAA expresses a clear preference for MACT emission standards.”4 Section 112(d)(2) of
the CAA requires that:

1
National Emission Standards for Hazardous Air Pollutants: Generic Maximum Achievable Control Technology;
Proposed Rule, 65 Fed. Reg. 76,408, 76,431 (Dec. 6, 2000).
2
40 C.F.R. § 63.1100(a), Table 1, Row 5. The Generic MACT was initially promulgated in 1999, 64 Fed. Reg.
34,863 (June 29, 1999), and was amended to add standards for the Ethylene Production category in 2002, 67 Fed.
Reg. 46,258 (July 12, 2002), which were subsequently clarified in 2005 (Ethylene Production MACT). 70 Fed. Reg.
19,266 (Apr. 13, 2005). These standards were not challenged, despite the fact that potential challengers were on
notice of the U.S. Court of Appeals for the District of Columbia Circuit’s precedent in Nat’l Lime Ass’n v. EPA, 233
F.3d 625, 633-34 (D.C. Cir. 2000), invalidating no emissions control MACT floors under the Act.
3
National Emission Standards for Hazardous Air Pollutants: Generic Maximum Achievable Control Technology;
Proposed Rule, 65 Fed. Reg. 76,408, 76,431 (Dec. 6, 2000).
4
U.S. Sugar Corp. v. EPA, 830 F.3d 579, 622-23 (D.C. Cir. 2016).
 Emissions standards promulgated under this subsection . . . shall require the
maximum degree of reduction in emissions of the hazardous air pollutants . . . that
the Administrator, taking into consideration the cost of achieving such emission
reduction, and any non-air quality health and environmental impacts and energy
requirements, determines is achievable.5

For existing sources, such standards may not be less stringent than the “the average emission
limitation achieved by the best performing 12% of the existing sources (for which the
Administrator has emissions information).”6 Alternatively, where EPA determines it is “not
feasible . . . to prescribe or enforce an emission standard,” EPA may instead “promulgate a
design, equipment, work practice, or operational standard, or combination thereof.”7 Where
EPA has identified a source of HAP emissions, “some section 112 standard must apply
continuously,” whether that standard is a numeric emissions standard under CAA Section
112(d)(2)-(3) or a work practice standard under CAA Section 112(h).8

Numeric Emissions Limit: EPA can reasonably determine not to set an emission standard for
“cracking” furnaces in the Ethylene Production source category under CAA Section 112(d)
due to data limitations.

EPA can reasonably conclude that it cannot establish cracking furnace numerical limits
under CAA Section 112(d)(2)-(3) because the emissions dataset it possesses is both too limited
and too variable to support establishment of an emission limit that could be defended in court or,

5
42 U.S.C. § 7412(d)(2).
6
42 U.S.C. § 7412(d)(3). The terms “emission limitation” and “emission standard” are both defined as a
requirement “which limits the quantity, rate, or concentration of emissions of air pollutants on a continuous basis,
including any requirement relating to the operation or maintenance of a source to assure continuous emission
reduction, and any design, equipment, work practice or operational standard promulgated under this chapter.” 42
U.S.C. § 7602(k).
7
42 U.S.C. § 7412(h)(1).
8
Sierra Club v. EPA, 551 F.3d 1019, 1028 (D.C. Cir. 2008); see also Nat’l Lime, 233 F.3d at 633-34 (EPA may not
set a “no control” MACT floor for HAP emissions from a source). In some prior instances, EPA has opted not to fill
a “gap” existing in the original MACT during the RTR rulemaking process. See, e.g., National Emission Standards
for Hazardous Air Pollutants for Chemical Recovery Combustion Sources at Kraft, Soda, Sulfite, and Stand-Alone
Semichemical Pulp Mills; Final Rule, 82 Fed. Reg. 47,328, 47,335 (Oct. 11, 2017) (declining to set standards for
previously unregulated HAP emissions due to the condensed timeframe required to meet the court ordered
rulemaking deadline). EPA could opt to do so again for cracking furnaces, specifically because of the insufficiency
of the data for particular HAP. EPA received a January 14, 2009 petition for rulemaking for numerous source
categories, including Ethylene Production standards under the Generic MACT, requesting EPA promulgate
standards for HAP emissions from cracking furnaces. Letter from James S. Pew, Earthjustice & John D. Walke,
NRDC to Stephen L. Johnson, Adm’r, EPA (Jan. 14, 2009). Any challenges to the “no control” floor for these units
should have been raised in the context of the original NESHAP in 2002, particularly given that Nat’l Lime was
issued in 2000. EPA’s court-ordered deadline to complete the RTR by March 2020 does not include a requirement
to fill this gap. Cal. Communities Against Toxics v. McCarthy, No. 15-0512 (D.D.C. Mar. 13, 2017).
EPA could also decline to revisit the “no control” floor for cracking furnaces based on a finding that the emissions
are de minimis, and thus are not actually a gap. EPA could exercise its de minimis authority under Section 112 not
to address certain sources or HAP and that a gap is permissible. The facts here are strong in that although the metals
and particulate matter emissions from cracking furnaces exist and due to developments in sampling methodology are
now detectable, they are very low and are not a risk driver for the source category.

2|Page
at a minimum, that the quality and quantity of the dataset is such that EPA at least has the
discretion not to base a CAA Section 112(d)(2)-(3) standard on it.

The reasons that EPA could be successful in this approach are outlined in detail in
Attachment 1, Overview of Ethylene NESHAP Furnace On-Line Test Data and Analysis for PM
and Metals. First, the dataset is small. Survey data were collected from only 19 (out of 518)
furnaces in the source category, including seven individual test runs from each furnace during
on-line operation.9 Second, the data are highly variable. EPA has previously contended with
small or variable datasets by using an upper prediction limit (UPL) adjustment calculation to
establish numeric emissions standards and this methodology has been upheld by the D.C.
Circuit.10 The cracking furnace dataset, however, has too great a degree of variation to be
accounted for with a UPL analysis.

The average furnace data across the 19 units tested exhibits a significant range for
emissions of Filterable PM (222 μg/dscm – 3327 μg/dscm, with a standard deviation as a percent
of the average of 55), Chromium (2 μg/dscm – 45 μg/dscm, with a standard deviation as a
percent of the average of 53), and Nickel (2 μg/dscm – 45 μg/dscm, with a standard deviation as
a percent of the average of 69). Significantly, in many cases, the standard deviation of the
average emission value for each furnace is larger than the average itself, thereby demonstrating
the significant variability between average furnace emissions. Further indicating the
questionable reliability of the data in terms of accuracy and representativeness, data variability
occurred among measurements at individual furnaces as well, with the standard deviation as a
percent of the average ranging from 17 to 133 for Filterable PM (emissions range of 16 μg/dscm
– 14,000 μg/dscm), from 2.3 to 48 for Chromium (emissions range of 0.05 μg/dscm – 30
μg/dscm), and 9 to 164 for Nickel (emissions range of 0.13 μg/dscm – 71 μg/dscm). A UPL
floor-setting analysis would result in a significant proportion (>40%) of the furnaces being above
the floor, without correlations with the emissions or the ability to distinguish the “best
performers” in terms of control, design, or operation. Notably, EPA determined in the
underlying MACT rulemaking that none of the furnaces were controlling HAP emissions using
add-on control devices and that remains true today. No significant correlation with radiant tube
age on an absolute (or percent of life) was identified, with R11 factors of 0.18 (absolute tube age)
and 0.24 (tube age as a percent of life) and R212 factors of <0.1. Additionally, no significant
correlation with chromium, nickel, or PM emissions has been identified for controllable process

9
The survey dataset is comprised of emissions and operating data provided on a confidential basis by nine
companies, representing only 19 ethylene production furnaces. While this represents a majority of the ICR
responses, it does not include them all. In a few cases, respondents provided corrections to the data they had
submitted that may or may not be incorporated in EPA’s database. It is believed that the survey dataset represents a
substantial portion of the data collected by EPA through the ICRs but that does not mean it is representative of the
operations of the more than 500 sources in the source category.
10
See Nat’l Ass’n of Clean Water Agencies v. EPA, 734 F.3d 1115 (D.C. Cir. 2013); U.S. Sugar Corp. v. EPA, 830
F.3d 579 (D.C. Cir. 2016); and Sierra Club v. EPA, No. 15-1487, __ F.3d __, 2018 WL 3320862 (D.C. Cir. July 6,
2018) (Boiler MACT).
11
R, the correlation coefficient, measures the strength and direction of a linear relationship between two variables on
a scatterplot. The value of R is always between +1 and –1. The closer R is to +1 or -1, the more closely the two
variables are related. If R is close to 0, it means there is no relationship between the variables.
12
R2 or R-squared is a statistical measure of how close the data are to the fitted regression line. It is also known as
the coefficient of determination, or the coefficient of multiple determination for multiple regression. The closer to
1.0, the better the fit of the regression line.

3|Page
variables, including excess oxygen or firebox temperature, with R values of 0.39 and 0.03 and
R2 values of 0.15 and 0.0 for excess oxygen and firebox temperature, respectively.

Average Emissions Data and Correlation Factors

Parameter Absolute Radiant Tube Percent of Radiant Tube Age

Age R/R2
R/R2
PM, lb./MMBTU 0.10/0.01 0.14/0.02
Cr, lb./MMBTU -0.15/0.02 -0.19/0.03
Ni, lb./MMBTU -0.01/0.00 0.03/0.00
PM, µg/dscm 0.18/0.03 0.24/0.06
Cr, µg/dscm -0.18/0.03 -0.17/0.03
Ni, µg/dscm 0.02/0.00 0.14/0.02

Similarly, other tools available to EPA in some other contexts, such as subcategorization,
present challenges, also due to the high degree of data variability here. 13 The cracking furnace
dataset exhibits extreme variation not only within the source category, but even among various
runs at the same furnace, including at some of the units that reported the lowest test values. To
date, it has not been possible to establish a correlation between any particular characteristics or
processes in place at various furnaces that would explain the test outputs. For instance, the feed
material for cracking furnaces does not contain metals, nor are the metals emissions and
variations therein clearly correlated with or readily explained by other factors at the furnaces.
Accordingly, EPA would struggle to provide a basis to address the data variation through
subcategorization.

A MACT floor must be no less stringent “than the average emission limitation achieved
by the best performing 12% of the existing sources (for which the Administrator has emissions
information),” for cracking furnaces.14 Here, the available data from stack testing is so variable
that EPA can reasonably conclude that it cannot identify which sources that reported test results
are the “best performers.” For instance, a single furnace in the survey dataset reports emissions
test values among the lowest and the highest reported for all furnaces tested. This is due in part
to the fact that data variation, as described above, occurs among individual test runs from the
same furnace. Specifically, a single furnace may report some of the lowest stack tested values
for half of the test runs, yet still be among the poorest of performers in its remaining documented
test runs. EPA has previously articulated that a MACT floor cannot be set where emission levels
are not replicable by the best performing sources because they are not being “achieved” by the
those sources.15 Thus, EPA can reasonably determine that the data it possesses are not reflective
13
42 U.S.C. § 7412(d)(1).
14
42 U.S.C. § 7412(d)(3) (emphases added).
15
National Emission Standards for Hazardous Air Pollutants: Final Standards for Hazardous Air Pollutants for
Hazardous Waste Combustors (Phase I Final Replacement Standards and Phase II); Final Rule, 70 Fed. Reg.
59,402, 59,461-62 (Oct. 12, 2005). EPA granted reconsideration of the 2005 Hazardous Waste Combustors (HWC)
rule and specifically reconsidered certain standards in light of the Brick MACT decision, including standards for
dioxin/furan emissions from solid fuel boilers and certain liquid fuel boilers. National Emission Standards for
Hazardous Air Pollutants: Standards for Hazardous Waste Combustors: Reconsideration; Final Rule, 73 Fed. Reg.
64,068, 64,089 (Oct. 28, 2008); see also National Emission Standards for Hazardous Air Pollutants: Standards for

4|Page
of the “best performing” 12% of sources, and EPA can reasonably conclude that at this time the
data are inadequate and insufficient to establish a numeric emissions standard as a MACT floor
in accordance with CAA Section 112(d)(3) or to go “beyond-the-floor” under Section
112(d)(2).16

If EPA concludes, as it reasonably can here, that the dataset is insufficient to set a CAA
Section 112(d)(2)-(3) standard at this time, its judgment should be afforded an “extreme degree
of deference” by a reviewing court in the event of a challenge. 17 In U.S. Sugar, industry
petitioners challenged EPA’s adoption of emission standards for major boilers using carbon
monoxide (CO) as a surrogate for organic HAP emissions, arguing it was not feasible to set a
numeric emission standard due to numerous non-detects in the data.18 The D.C. Circuit rejected
the argument and deferred to EPA’s scientific conclusion regarding the sufficiency of its data to
set numeric standards.19 On the same reasoning that EPA should receive deference when it
determines the dataset is sufficient for standard-setting (e.g., by taking into account below
detection limit (BDL) results), a court should defer to an adequately-supported EPA
determination that the available dataset is insufficient to set a numeric emission limit. Here, EPA
could rely on the extreme variability at a single unit and the inability to explain the variations in
test results for similar operations or to determine what sources are the “best performers.”

Authority for Work Practice: It is appropriate to establish a work practice standard for certain
HAP emitted from cracking furnaces under CAA Section 112(h), in lieu of an emissions
standard, because the data deficiencies make it “infeasible” to prescribe or enforce a numeric
emissions standard.

Where EPA determines it is “not feasible . . . to prescribe or enforce an emission

standard,” CAA Section 112(h) empowers EPA to establish work practice standards instead. 20
The statute defines “not feasible to prescribe or enforce an emission standard” to include

Hazardous Waste Combustors; Solicitation of Comment on Legal Analysis, 72 Fed. Reg. 54,875, 54,886 (Sept. 27,
2007) (notice of intent not to defend certain standards in light of Brick MACT decision). The D.C. Circuit later
granted EPA’s motion for voluntary remand, without vacatur, of all standards in both the 2005 HWC Rule and the
2008 HWC Reconsideration Rule. Order, Sierra Club v. EPA et al., No. 05-1441, Doc. No. 1201131 (D.C. Cir.
Aug. 14, 2009).
16
Emissions standards promulgated under CAA Section 112(d)(2) “shall require the maximum degree of reduction
in emissions . . . that the Administrator . . . determines is achievable . . . through application of measures, processes,
methods, systems or techniques.” 42 U.S.C. § 7412(d)(2). For cracking furnaces, EPA has been unable to identify a
correlation between any of these things and emissions. EPA has proclaimed a longstanding principle in other CAA
contexts that both emissions and control measures must be quantifiable. See State Implementation Plans; General
Preamble for the Implementation of Title I of the Clean Air Act Amendments of 1990, 57 Fed. Reg. 13,498, 13,567
(Apr. 16, 1992) (describing the general principles in the State Implementation Plan context).
17
U.S. Sugar Corp., 830 F.3d at 623 (quoting Kennecott Greens Creek Mining Co. v. Mine Safety & Health Admin.,
476 F.3d 946, 954-55 (D.C. Cir. 2007).
18
Id.
19
Id.
20
42 U.S.C. § 7412(h)(1). The D.C. Circuit has held that “only if measuring emission levels is technologically or
economically impracticable,” does Section 112(h) “allow[ ] EPA to substitute work practice standards for emission
floors.” Sierra Club, 479 F.3d at 884 (Brick MACT); see also U.S. Sugar, 830 F.3d at 622-23 (“The CAA expresses
a clear preference for MACT emission standards and limits the EPA’s ability to fashion more flexible work-practice
standards. . . . To set a work-practice standard for these emissions, in fact, the EPA would need to find that it is
infeasible to set a numeric standard for a particular HAP. Id. § 7412(h)(1).”).

5|Page
situations in which “the application of measurement methodology to a particular class of sources
is not practicable due to technological and economic limitations.”21 The “application of
measurement methodology” language can be reasonably interpreted in a manner that ensures that
the D.C. Circuit’s mandate for continuous emission standards to be satisfied.22 Thus, in a
situation where the data EPA possesses do not support standard-setting under CAA Section
112(d)(2)-(3) (as here given the variability and inability to correlate results with any relevant
factor of operation), CAA Section 112(h)’s language can be reasonably interpreted to allow a
work practice to ensure continuously applicable standards for cracking furnaces.

EPA’s authority to set work practice standards under Section 112(h) “centrally relies on
‘the judgment of the Administrator’ regarding when an emissions standard is not feasible, and
. . . whether the standard is appropriately stringent to meet the statute’s objectives.” 23 The D.C.
Circuit has also found that “[t]hat judgment of course must be guided by permissible factors.”24
One such “permissible factor” that the D.C. Circuit has recognized is “known technological
limitations on the use of control devices during the volatile conditions that characterize
startup.”25 Similarly here, evidence suggests that baghouses or other add-on control devices are
not feasible and, at the very least, are not cost-effective, to control emissions from cracking
furnaces.

The phrase “application of measurement methodology” under CAA Section 112(h)(2)(B)

can reasonably be read to include the need for meaningful data measurements, rather than just
the physical ability to collect test data. The “measurements” available for cracking furnaces are
not meaningful for several reasons, among them that many furnaces share a stack and those
furnaces may be operating in different modes at one time.26 Determining emissions from a
single furnace would necessitate shutting down all other furnaces that share a common stack in
order to run a stack test, at a tremendous cost to the facility. Setting a CO limit as a means to
ensure good combustion, and thus low organic HAPs emissions, would not make sense if it were
measured across multiple furnaces. Similarly, setting numeric limits for metals or PM would not
make sense for these cases, absent an emissions averaging provision and perhaps not even
then. Likewise, some facilities have multiple furnaces and other combustion units that share a
common heat recovery system, which vents out of a common stack. Often, there are no locations
on the duct work for installation of adequate ports between each individual furnace and the heat

21
42 U.S.C. § 7412(h)(2)(B). Section 112(h)(2)(A) provides another situation where emission standards are “not
feasible to prescribe or enforce,” where EPA determines that “a hazardous air pollutant or pollutants cannot be
emitted through a conveyance designed and constructed to emit or capture such pollutant, or that any requirement
for, or use of, such a conveyance would be inconsistent with any Federal, State or local law.” 42 U.S.C.
§ 7412(h)(2)(A).
22
See Sierra Club, 551 F.3d at 1028.
23
Sierra Club v. EPA, 884 F.3d 1185, 1204 (D.C. Cir. 2018) (Industrial Boiler MACT) (quoting 42 U.S.C.
§ 7412(h)(1)).
24
Id.
25
Id. (citing U.S. Sugar, 830 F.3d at 664-65).
26
For example, one furnace could be in cracking mode, another in decoking mode and another in hot-steam standby.
There also are companies that have one or more ethylene furnaces and other combustion units (regulated under other
NESHAP) tying into a waste heat boiler and then to one stack.

6|Page
recovery equipment.27 Moreover, some heat recovery equipment is gas-fired, meaning that an
emissions measurement taken from the common stack would not be representative of the
cracking furnace emissions.28 Where measurements collected are so variable as to prevent
identification of the best performing sources and lack any correlation to possible control
measures to render an emissions standard achievable, it is appropriate for EPA to determine that
it is not feasible to prescribe or enforce a numeric emission standard until such time as
measurement techniques are developed that can support such correlations.

This approach finds support in prior rulemakings, in which EPA has cited a lack of
reliable emissions data in determining that an emission standard is infeasible. EPA offered one
articulation of this insufficiently meaningful data concept in the RTR rulemakings for the Wool
Fiberglass and Primary Aluminum source categories, where EPA relied on the high proportion (a
majority) of test results below detection levels to justify promulgating work practice standards. 29
EPA explained its rationale particularly well in the Primary Aluminum rule, stating that it “does
not regard measurements which are unreliable, non-duplicable, and nonreplicable to be
practicable. Simply put, the CAA simply does not compel promulgation of numerical emission
standards that are too unreliable to be meaningful.”30

To give effect to this principle, the phrase “application of measurement methodology” in

CAA Section 112(h)(2)(B) should be interpreted to incorporate the need for meaningful data
measurements. In the Primary Aluminum and Wool Fiberglass instances noted above, EPA
determined it lacked “meaningful” data where a significant proportion of the available data were
below detection limits. So too here, where the dataset for cracking furnaces lacks sufficient
meaning because it is highly variable, i.e., the variability of test results among the sources and
between different test runs at the same source precludes setting a numeric emission standard.
The D.C. Circuit has stated that EPA is afforded an “extreme degree of deference” in
determining whether data are sufficient to promulgate an emission standard.31 That deference
could similarly insulate EPA from challenges by environmental groups if EPA reasonably
decides to promulgate a work practice based on a finding that a numeric emission standard is not
feasible due to inadequate data.

27
Such sample ports must comply with the requirements of 40 C.F.R. Part 60 App’x A-1, Method 1, Section 11.1.1
(eight duct diameters downstream and two diameters upstream of a flow disturbance or the alternative of at least two
diameters downstream and a half diameter upstream of a flow disturbance).
28
A further complication is that the fired waste heat boilers are subject to the Boiler and Process Heater MACT. 40
C.F.R. Part 63, Subpart DDDDD—National Emission Standards for Hazardous Air Pollutants for Major Sources:
Industrial, Commercial, and Institutional Boilers and Process Heaters.
29
See National Emissions Standards for Hazardous Air Pollutants for Mineral Wool Production and Wool
Fiberglass Manufacturing; Final Rule, 80 Fed. Reg. 45,280 (July 29, 2015); National Emission Standards for
Hazardous Air Pollutants for Primary Aluminum Reduction Plants; Final Rule, 80 Fed. Reg. 62,390 (Oct. 15,
2015).
30
80 Fed. Reg. at 62,406.
31
U.S. Sugar, 830 F.3d at 363 (citation omitted).

7|Page
Appropriate Work Practice: Sound engineering supports flame impingement management as
an appropriate work practice under CAA Section 112(h).

Once EPA determines it is appropriate to set a work practice standard in lieu of a numeric
emissions limit for cracking furnaces, EPA could reasonably promulgate a work practice
standard requiring flame impingement checks. Attachment 2, Summary of Literature Search:
Achievement of Lower Furnace Stack Metal Emissions via Reduction of Burner Flame
Impingement on Radiant Tubes, describes the sound engineering basis for EPA to conclude that
flame impingement checks/response is a reasonable work practice standard to address metals and
particulate matter emissions from cracking furnaces. In brief, the document explains that furnace
stack metals emissions originate from radiant tubes within the steam cracking furnace. The
radiant tubes are made of heat resistant steels with high nickel and chromium content. Some
metal oxidation can occur at temperatures required for steam cracking operations, which may
increase as temperatures rise. As a result of the metal oxidation process, chromium scales can
form at the surface of the radiant tube. Traces of the metals can then travel out of the furnace
stack along with the furnace flue gas. Flame impingement increases the radiant tube metal
temperature and the velocity of the gas hitting the radiant tubes; both of which would increase
metals emissions. Managing flame impingement, therefore, would act to mitigate such
emissions. Photographs illustrating the presence of flame impingement are provided below:

Flame Impingement Figure 1. Baking Soda Used to Visualize Flame

Flame Impingement Figure 2. Color Change on Tubes Indicating Flame Impingement

8|Page
 Based on the above, EPA could adopt a work practice standard requiring cracking
furnace operators to perform visual verification and management of flame impingement on the
radiant tubes to minimize higher temperatures during cracking, thus controlling the PM and
metals emissions from the furnace stack. Periodic visual inspections would be conducted to
detect radiant tube hotspots and fallen refractory along with burner inspections to verify proper
functioning of relevant aspects of the burner. Attachment 3 outlines a draft approach to the work
practice standard for consideration.

Attachments

1. Overview of Ethylene NESHAP Furnace On-Line Test Data and Analysis for PM and Metals
2. Summary of Literature Search: Achievement of Lower Furnace Stack Metal Emissions via
Reduction of Burner Flame Impingement on Radiant Tubes
3. Flame Impingement Work Practice Proposal

9|Page
 Attachment 1

Overview of Ethylene NESHAP Furnace On-line Test Data and

Analysis for PM and Metals

The Ethylene National Emissions Standards for Hazardous Air Pollutants (NESHAP or MACT)
source category includes 518 furnaces at 31 facilities. In response to the Ethylene NESHAP
Information Collection Request (EMACT ICR), ethylene furnace emissions and operating data
were submitted to EPA. The emissions and operating data were subsequently provided to
AECOM and Fresh Air Consulting on a confidential basis by nine companies, representing 19
ethylene production furnaces32. That information was supplemented with additional furnace and
operating information through confidential surveys. The data were analyzed to determine
average emissions and emission factors for chromium, nickel, filterable particulate matter (PM)
and other pollutants. Data quality was also evaluated by calculating the standard deviations
associated with intra and inter-furnace averages. Significantly, in many cases the standard
deviation of the 7-test average emission values for each furnace was equal to or larger than the
average itself, thereby demonstrating the significant variability between furnace emissions
measurements. Attempts were also made to identify correlations between pollutants, and
between pollutants and operating parameters, but no consistent, strong correlations were found.
The emissions data analysis discussed here related to the emission results is limited to the
Ethylene NESHAP ICR test results when the furnaces were in “cracking” or “on-line” operation
(i.e., when feed is heated inside the furnace tubes in the presence of steam to make ethylene and
ethylene byproducts). There are no federal or state requirements for periodic or ongoing
chromium, nickel, or filterable PM tests, nor continuous monitoring of PM emissions
demonstrated for this source category or emissions at this level.
For the 19 furnaces tested, all were cracking gas feeds and the furnace burners were fueled with
natural gas-like fuels mixed with byproduct hydrogen33. Since the ethylene production reaction
is contained in the furnace tubes, the furnace stack emissions reflect the result of combusting the
fuel gas.
Ethylene furnaces have a range of sizes, typically characterized by the maximum firing rate.
Thus, emissions are typically reported on a µg/dscm (corrected to 3% oxygen) or lb/million BTU
(MMBTU) basis, as well as a lb/hr basis, to adjust for furnace size. For the 19 furnaces in the
on-line analysis, the firing rates averaged 85% of maximum firing.

32
While this represents a majority of the ICR responses, it does not include them all. In a few cases respondents
provided corrections to the data they had submitted, that may or may not also be incorporated in the EPA database.
33
For all operations studied, furnace burners were fueled with gaseous fuels, typically natural gas or plant fuel gas
supplemented with natural gas that was blended with hydrogen, and ethylene process byproduct. The hydrogen
content of ethylene furnace fuel can range from very little to as much as 80 mole percent, depending on plant
hydrogen balances and operating situations. Even where normal hydrogen contents are low, high hydrogen
situations can occur during unusual operations. Most of the hydrocarbon portion of furnace fuel gas is typically
methane. Thus, furnace stack emissions would be expected to normally reflect fuel gas and hydrogen combustion.
The on-line test results were all representative of furnaces feeding gas feed and include two
furnaces feeding propane and six feeding ethane/propane mix. The other furnaces were feeding
ethane. In all cases furnace burners were fueled with natural gas or plant fuel gas supplemented
with natural gas and/or hydrogen. Overall furnace averages were calculated as the average of the
average results for each furnace. For the on-line furnace stack analyses, seven replicate samples
were obtained, and the individual furnace averages was calculated from those seven sets of
results.
In the MACT standard setting process, the Agency must determine the “level of control”
currently achieved by the best-controlled similar source (for new source standards) or the
average of the best-performing sources (for existing source standards). The Agency uses the
Upper Prediction Limit (UPL)34 to both evaluate average emissions performance and estimate
variability to determine the level of control sources currently achieve.
To determine the MACT floor, an average concentration value is calculated for each source.
Average values include any data reported as non-detect. Sources are then ranked from lowest to
highest based on their average concentration. Next, the best performing sources are identified
based on the lowest average concentration values. For the case of Ethylene Furnace on-line PM
emissions, the MACT floor would be based on the three best performing sources with the lowest
average emission results (12% of the tested furnaces). The individual test runs for the three best
performing sources are considered the MACT floor dataset. The distribution of the dataset is
determined (e.g., normal, lognormal, etc.) and the appropriate equation is used to calculate the
UPL based on the top performers.
Applying the UPL method to the top three best performers’ results in a calculated MACT floor
of 860 µg/dscm (adjusted to 3% O2). This floor is below the average of the overall furnace test
results, which is highly variable as represented by a wide range and high standard deviation. The
overall furnace average PM emissions is 1055 µg/dscm with an average standard deviation of the
19 furnace averages results of 716 µg/dscm (55% of the average value). The average furnace
PM values range from 222 µg/dscm to 3327 µg/dscm with an even wider range on an individual
basis, as discussed below. This large standard deviation results in little confidence that the
calculated average value is a valid representation of the PM emissions.
Overall Furnace Average Data Analysis, Filterable PM (µg/dscm)

+/-1 Standard Deviation Overall Furnace Average Furnace Averages MACT Floor

34
For additional information on EPA’s use of the UPL refer to the memorandum entitled “Use of the Upper
Prediction Limit for Calculating MACT Floors,” from Jeff Telander, Office of Air Quality Planning and Standards
(OAQPS), EPA, to Docket ID No. EPA-HQ-OAR-2013-0291 (Sept. 9, 2014), Docket No. EPA-HQ-OAR-2013-
0290-0217.

2|Page
Each individual furnace completed a seven run test to calculate the furnace average filterable PM
emissions. The individual values ranged from 16 to 14000 µg/dscm for the full dataset with an
overall average of 1050 µg/dscm and individual furnace average standard deviations ranging
from 106 µg/dscm to 4440 µg/dscm (17% to 133% of the furnace average value). This results in
a majority of the tested furnaces (11 out of 19) having a wider range in the tested results then the
calculated 860 µg/dscm MACT floor. Because of the large variability shown by the standard
deviations there can be little confidence that the average is a valid representation of the PM
emissions from ethylene furnaces.
For example, one of the furnaces in the middle of the dataset had an average of 821 µg/dscm
with a standard deviation of 882 µg/dscm and a range in its seven run, steady state results from
153 µg/dscm to 2780 µg/dscm. In order to predict the future measurement for this furnace on a
90% confidence level, the range of expected measurements is between 765 µg/dscm and 1335
µg/dscm. These results raise considerable question as to what the emissions actually are, and it
is unclear how a reliable compliance demonstration could be performed in the face of this large
variability.
Furthermore, the on-line PM emission test results are only a snapshot of the filterable PM
performance of the sources when held at steady state test conditions and may not be
representative of all normal operations.
Finally, “the level of control” for the lowest emitting sources identified in the one-time seven-run
test, cannot be differentiated among the other tested sources in the dataset. The lowest emitting
units do not have means to control PM to a particular emission level; and given the variation in
the average filterable PM emissions on both a furnace average and individual furnace basis
(regarding both the broad range of results and high standard deviation), even a best performer
may find results outside of any MACT floor standard during future tests.
No differentiating factors have been identified that explain differences in filterable PM or metals
emissions for Ethylene NESHAP furnace sources. Neither radiant coil life nor percent of life,
radiant coil surface area, control operation (i.e. controllable process variables), furnace fuel fired,
nor radiant coil tube metallurgy explain the differences.
The calculated correlation factor (R)35 and the coefficient of determination (R2)36 for the best fit
straight line for chromium, nickel and on-line PM emissions were evaluated versus tube age for
the data set (µg/dscm basis). No significant correlation with radiant tube age on an absolute (or
percent of life) basis was observed for any of the emission parameters evaluated, with R factors
of 0.18 (absolute tube age), 0.24 (tube age as a percent of life) and R2 factors of <0.1.
Additionally, no significant correlation with chromium, nickel, or PM emissions has been
identified for controllable process variables including excess oxygen or firebox temperature with
R values of 0.39 and 0.03 and R2 values of 0.15 and 0.0 for excess oxygen and firebox
temperature, respectively.

35
R, the correlation coefficient, measures the strength and direction of a linear relationship between two variables on
a scatterplot. The value of R is always between +1 and –1. The closer R is to +1 or -1, the more closely the two
variables are related. If R is close to 0, it means there is no relationship between the variables.
36
R2 or R-squared is a statistical measure of how close the data are to the fitted regression line. It is also known as
the coefficient of determination, or the coefficient of multiple determination for multiple regression. The closer to
1.0, the better the fit of the regression line.

3|Page
 Average Emissions Data and Correlation Factors

Parameter Absolute Radiant Tube Age Percent of Radiant Tube Age

R/R2 R/R2
PM, lb./MMBTU 0.10/0.01 0.14/0.02
Cr, lb./MMBTU -0.15/0.02 -0.19/0.03
Ni, lb./MMBTU -0.01/0.00 0.03/0.00
PM, µg/dscm 0.18/0.03 0.24/0.06
Cr, µg/dscm -0.18/0.03 -0.17/0.03
Ni, µg/dscm 0.02/0.00 0.14/0.02

Correlation Factors for Cr on Absolute Tube Life Basis

Given that these sources do not have add on control devices for PM, there is no assurance that
the best performers can achieve, in the future the emission level reported in the tested dataset.
Further, the test data do not reflect these sources’ variability, and the variability is largely
unquantifiable given the uncertainties regarding control mechanisms. Hence, that reported
emission level is not being “achieved” for the purpose of establishing a floor.

4|Page
 Attachment 2

Summary of Literature Search: Achievement of Lower Furnace Stack Metal

Emissions via Reduction of Burner Flame Impingement on Radiant Tubes

Flame impingement is a known mechanism for generation of metals emissions from radiant
tubes in Steam Cracking furnaces. Therefore, monitoring for flame impingement and
interrupting it when it occurs will reduce emissions of metals from these furnaces. Because
Steam Cracking furnaces do not exhibit other sources or known mechanisms for generating
metals emissions, flame impingement monitoring and management is the best known mechanism
to directionally manage and minimize these emissions.
Radiant tubes in Steam Cracking furnaces are made of heat resistant steels with high nickel and
high chromium content. At Steam Cracking temperatures (often in the range of [890°C-1200°C])
metal oxidation is expected. As temperatures increase metal oxidation increases 37,38. For
chromium this mechanism is:
4 2
𝐶𝑟 + 𝑂2 ↔ 𝐶𝑟2 𝑂3
3 3
The metal oxidation forms chromia scales at the tube surface. Potential mechanisms for
chromium depletion are then:
1. Erosion of the oxide layer in moving gas streams;
2. Spalling of thick scaling layers due to thermal cycling;
3. Degradation of the oxide scale as a result of progressive chromium depletion.
Through these mechanisms, traces of these metals can travel, both in a steady state as well as
non-steady state, with the furnace flue gas out of the furnace stack. Thus, furnace stack metal
emissions are due to the radiant tube oxidation, and subsequent oxide removal, on the outside
surface of the tubes.
Flame impingement both increases the gas velocity hitting the radiant tubes, as well as increases
the radiant tube metal temperature. Literature results demonstrated exposure to higher velocity
gas streams was found to reduce the Cr2O3 oxide layer39. Typical burner tip velocities can be in
the range of 100-300 ft/s. Exposing the radiant tubes directly to these velocities due to flame
impingement is more likely to erode the oxide layer and increase metal emissions. Flame
impingement also increases the radiant tube metal temperatures at the impingement location.
This increases removal of chromium from the radiant tubes via the third mechanism described
above, degradation. Volatilization of Cr2O3 is not expected at Steam Cracking temperatures, but
the following mechanism can create volatile chromium:
3
𝐶𝑟2 𝑂3(𝑠) + 𝑂2(𝑔) ↔ 2𝐶𝑟𝑂3(𝑔)
2

37
D. Caplan, A. Harvey & M. Cohen, “Oxidation of Chromium at 890°C-1200°C,” Corrosion Science, 1963, Vol.3
at 161-175.
38
B. Ebbinghuas, “Thermodynamics of Gas Phase Chromium Species: The Chromium Oxides, the Chromium
Oxyhydroxides, and Volatility Calculations in Waste Incineration Processes,” Combustion and Flame, 1993, Vol.
93 at 119-137.
39
J. Wright & J. Leader, “The Effects of Gas Velocity and of Temperature on the Oxidative Response of Selected
Sheet Superalloys,” ASME, Presented at the Gas Turbine and Aeroengine Congress, 1988.
Literature data shows that chromium depletion increases at higher temperatures due to this
mechanism40.
Flame impingement on radiant tubes increases the gas velocity hitting the metal oxide layer,
increasing erosion and degradation, and, ultimately metal emissions out of the furnace stack.
Flame impingement also increases the radiant tube metal temperature increasing the
volatilization of the metal surface, increasing metal emissions out of the furnace stack. These
results are supported both by fundamentals and literature experimental data. Though this paper
describes the specific mechanisms for Cr oxide emissions, the mechanisms described apply to all
metals present in radiant coil metallurgy. This can include nickel, iron, aluminum, tungsten,
niobium, manganese, titanium, etc.

40
D. Caplan & M.Cohen, “The Volatilization of Chromium Oxide,” Journal of the Electrochemical Society, 1961,
Vol 108, at 438-442.

Page | 2
 Attachment 3

Flame Impingement Work Practice Proposal

Flame impingement causes:

1) Localized high temperatures on the radiant tube with temperature cycling causing higher
oxidation rate.
2) Localized high coking at the high temperature regions resulting in short runtimes between
decokes. It may also cause a shutdown to mechanically remove the coke on the inside of
the tube.
3) Overheating, which can reduce metal strength and cause radiant coil failure.
4) Increased radiant tube carburization, which can lead to brittle fracture.
Industry would be incentivized to apply and improve flame impingement prevention work
practices to avoid downtimes and improve operation and reliability41.
Flame impingement may be caused by the following:
1) Burner fouling.
2) Burner component failure.
3) Refractory material falling on burners.
4) Air maldistribution throughout the firebox.
5) Radiant coil distortion bending the coil into the path of the flame.
Flame impingement may be detected by:
1) Visual inspection of the radiant coil for localized bright spots. This may be confirmed
with a temperature gun.
2) Use of luminescent powders injected into the burner to illuminate the flame pattern.
3) Continued localized coking causing short runtimes between decokes.
Recommended work practice:
The owner or operator shall conduct a firebox inspection once per shift to detect radiant tube
hotspots and to detect fallen refractory. The owner or operator shall conduct a daily inspection
to confirm the correct burner configuration and adequate burner air per operating procedures
(allowing for burner maintenance) and to verify the fuel gas liquid knock-out pot, coalescer or
other liquid removal equipment, if present, are functioning properly.
Upon positive detection of a hot spot or fallen insulation on burners or the malfunction of the
above noted equipment, the operator must take the necessary options to repair as soon as
practical. The recommended action plan should be documented by the owner and operator.
Furnace operators should be trained on the identification of, the required recording of positive
detection and necessary actions to mitigate hot spots.

41
S. Inamdar, R. Stegemann. “Improve Furnace Availability and Control Downtime.” AIChE Paper Number: 95a.
Prepared for Presentation at the 2014 Spring National Meeting. New Orleands, LA, March 30 - April 3, 2014.
Possible actions may include:
1) The confirmation of a high temperature with a temperature detection device.
2) Isolation of malfunctioning burners until burner repair, replacement or cleaning or
refractory removal.
3) Use of luminescent powders injected into the burner to illuminate the flame pattern so
that it may be corrected.
4) Protection from the burner with insulation around the radiant tube.
5) Repair of malfunctioning fuel liquid removal equipment.
6) Burner configuration and/or burner air corrections.

Page | 2
 Appendix D

Enforcement Matter
Case No. 48396 (Shell Chemical LP)
 Appendix E

Presentation Provided to EPA on December 14, 2017 From American

Chemistry Council (ACC) and American Fuel & Petrochemical
Manufacturers (AFPM)
EPA Steam Cracking
Technical Primer
 Discussion Outline
• Steam Cracking Overview
• Furnace Configuration
• Online Emissions
• Fuel Types
• Combustion Control
• Description of Emissions
• Measurement Error and Normalization Methods
• Decoke Emissions
• Types of Coke
• General Decoking Steps/Process
• Description of Decoking Control Devices
• Description of Emissions
• Measurement Error and Normalization Methods
 Steam Cracking Furnace
Three main sections of a pyrolysis furnace:
Convection 1. Convection Section
Section • Preheats and vaporizes the feed by absorbing waste
Flue Gas heat from the radiant box
Quench Section ID
(Oil or Water) Fan • Mixes the hydrocarbon with dilution steam
HC FEED
Dilution Steam • Recovers heat for steam generation
Transfer Line to
Primary Fractionator
2. Radiant Section
or Quench Tower Radiant • Uses heat to crack apart carbon and hydrogen bonds
Transfer Line Section
Exchanger or
to form double bonded olefins.
Quench Header • Coke (carbon) is a byproduct left inside the radiant
tubes
• Radiant tubes are typically composed of Ni, Cr, and Fe
Firebox
3. Quench Section
• Arrests the reaction by cooling the effluent with
Gas Burners quench oil or water
• Heat is used to generate steam and drive turbines
Steam Cracking Furnaces Are Highly Efficient
Fuel Types
• Industry fuel gas compositions vary and are dependent on the process and
commercial agreements
• Hydrogen fuel content varies from 0-85 mol%
• Primary components CH4 and H2
• Start-Up on Natural Gas
High Combustion Efficiency
Flue Gas

• Steam cracker furnaces have high fuel ID Fan/

Louvre
combustion efficiency because: HC
Dilution Feed
• Fuel is clean Steam
• Excess air is controlled O2 Analyzer CONTROLLED

• Firebox temperatures are high

• Regulated NOx and CO emission
• Can be hundreds of burners per furnace
• Uneven combustion results in production loss Wall Burners
• Maximum ethylene production drives towards
healthy, even heat distribution
• Drives emphasis to keep fuel and burners clean
• Confirmation of burner health could be done Can be either
Floor or both
with periodic burner health checks using CO as
Burners
health confirmation
• Same as BPH MACT
On-Line Emission Sources
• Organics formation is impacted by:
• Burner Air:Fuel ratio
• Burner fuel composition– Function of process configuration
• Burner condition
• Process leaks – Stack opacity is historical indicator
• Cr & Ni oxides form from radiant material oxidation during on-line and decoking operation
• Radiant material is composed of Fe, Ni, and Cr
• Can’t control so must account for high variation
• ACC Average Data: Online Operation vs. Decoking
Firing Total PAH* LB/HR Cr LB/HR Ni LB/HR *Includes naphthalene and 2-MN. Approx.
50% non-detects were assumed at DL or RL
Online Design 1.6E-4 1.1E-3 4.9E-4 (conservative).
Decoke Firing 25% of Design 3.9E-4 4.5E-4 1.9E-4

• Increase in decoking PAH emissions is consistent with literature and expected at lower temperature
• Cr and Ni decrease in LB/HR emissions less than firing reduction
• Although decoking temperature is cooler it is also higher in excess air
• Furnace Size Normalization Recommendation: LB/MBTU
Compound Method Errors: On-Line
• Testing error makes it difficult to obtain repeatable and reliable results, therefore
quantification of emissions must answer:
• How accurately does the measured emission rate represent the stream?
• How much precision is there with stream flow measurement?
• Furnace Emissions from Same Size and Configuration:
• Similar design and operating conditions should produce the same emissions even at a different site
Example of actual error range using normalized emission (LB/MBTU) from different companies:
Company 1 Furnace Stack Online Company 2 Furnace During Decoke
120% 120%
Potential Sources
of Error 100% 100%
Dilution Factor
Error 80% 80%
Furnace 1
Flow Measurement Furnace 1
Error
60% 60% Furance 2
Furnace 2
Furnace 3
40% 40%
Differing
Detection Limits
20% 20%

0% 0%
*Total PAH only ADL
*PAH Cr Ni *PAH Cr Ni
Large error bands come from detection limit and other process measurement error
Coke formation dependent on feed type:
• Catalytic
• Radiant tube surface Fe and Ni catalyze coke
• Filamentous structure containing the metal
• In all feed types but more dominant in gas (ethane) cracking
• Current set of data is from mostly catalytic coke from ethane and propane feedstock
• Pyrolytic
• Byproduct of thermal cracking process
• Amorphous structure
• In all feed types. Becomes more dominate as feed MW rises
• Condensative
• High MW products condense on the cool quench surfaces
• Liquid feed types
• Decoking process varies for the different cokes types
Dataset does not represent emissions from some commonly used feedstocks.
The standard should not limit feedstock flexibility.
 When to Decoke and for How Long
• Periodically coke must be removed
• Frequency depends on many factors
• Decokes are initiated when
• the radiant tube metal temperature reaches design
• pressure drop causes flow maldistribution and loss of feed
• reduced quench exchanger cooling causes external piping to reach design
temp.
• maintenance or shutdown is required
• Runlength between decokes and decoke times vary with
technology and feed type
General Decoking Procedure
• Keep steam, remove feed and reduce firing Called Clean

• Transfer the furnace effluent from the Controlled Burn Clean-Up

recovery section to the atmospheric decoke
drum/firebox
• Confirm no hydrocarbon leaking
• Slowly add air to a safe concentration and
hold until the coil is clean
• This is a controlled exothermic burn
• Manual check to ensure coil is clean
• Remove air and switch effluent back to
recovery system
• Confirm no leaking Coil Outlet Temp Air Conc.

To determine emissions/decoke the 3 samples should be averaged over the decoke time
HAP standard between air in and out captures decoke emissions
 Steam Cracking – Coke Removal Controls
2. Convection
• Decoke Drum Transfer- Line to Section

Flue Gas
• Knocks out large particulates Primary Fractionator
or Quench Tower 4. Quench ID Fan
• Steam, air, combustion products, and small Section HC Feed

particulates go overhead Dilution Steam

3. Radiant
Section
• Decoke to Firebox
TLE or
Quench
Decoke
Header
• Routes decoke effluent to firebox
Drum

• Small particulates likely to combust while Decoke to

large particulates can make it out of the Firebox 1. Firebox

stack
Gas Burners

Decoke Drum + Furnace Stack During Decoke ~ Furnace Stack During Decoke to the Firebox
ACC Decoking Control Comparison
• Decoking testing conditions more
challenging than online Decoke Drum + Stack Decoke to Firebox
1.0E-03 2.5E-01
• High and changing moisture,
temperature, and velocities and 9.0E-04

coke 8.0E-04 2.0E-01

LB/MBTU (PAH/Cr/Ni)
• Variability of ~30% in measuring 7.0E-04

LB/MBTU (PM)
the same flow rate 6.0E-04 1.5E-01

• Measured blanks of PAH’s were

5.0E-04

same order of magnitude as 4.0E-04 1.0E-01

sample measurements 3.0E-04

• On average blanks ~13% of sample 2.0E-04 5.0E-02

• Compounding estimated error

1.0E-04

range shows control device 0.0E+00

Total PAH Cr Ni PM
0.0E+00

emissions are similar

• Decoke Stack +/- 43%
• Furnace Stack +/- 13% Control Devices Result in Similar Emissions
PAH Measurement Example
PAH Blank ng Avg Measured ng Blank/Measured
2-Methylnaphthalene 534 2684.7 19.9%
Acenaphthylene 19 390.0 4.9%
Anthracene 32.9 227.6 14.5%
Fluoranthene 48 220.1 21.8%
Fluorene 49 401.1 12.2%
Naphthalene 733 9281.4 7.9%
Phenanthene 228 1210.7 18.8%
Pyrene 58 291.0 19.9%
Σ19 PAH’s 1951.8 15107.8 12.9%

• Measured sample catch and blank catch

• Highest 8 measured PAH’s shown
• Measured sample uses 5 x blank if sample:blank <5
• Blank as a percentage of average measured value represents “accuracy” of the
measurements
ACC Avg. Cr and Ni Trends Over Course of Decoke
Industry Cr Emissions Industry Ni Emissions
8.00E-04 1.40E-03

7.00E-04 1.20E-03
6.00E-04
1.00E-03
5.00E-04

LB/MBTU
LB/MBTU

8.00E-04
4.00E-04
6.00E-04
3.00E-04 Cr in embedded in coke and continues to Ni is embedded in coke and
4.00E-04
2.00E-04 oxidize when the tube surface becomes declines as coke is removed
1.00E-04 exposed 2.00E-04

0.00E+00 0.00E+00
1 2 3 1 2 3
Drum + Stack Drum + Stack
Industry PM Emissions
3.00E-01

2.50E-01

2.00E-01
LB/MBTU

1.50E-01

1.00E-01

5.00E-02 Generally tracks coke removal

0.00E+00
1 2 3
Drum + Stack
Decoke Drum Isolation Prevents HAP Emissions
• All 3 runs measured HAP from the same decoke drum from sister furnaces
• + Emissions from organic HAP Transfer- Line to
Primary Fractionator
or Quench Tower
Normalized Total HAP Out of Common Decoke Drum
120%
Poor Isolation of
100% Process

80% Decoke
Drum
60%

40%

20%

0%

ΔP High ΔP Low ΔP High

Best Performance: Prevent process leakage into decoke drum

 Normalization Method
• Options:
• Coke in coil – No standard method for estimating
• Gasified coke – Measurement for CO and CO2 had high variability and the two control methods
make gasification rates very different
• Decoke Firing – Generally an indicator of furnace size and consequently coke volume
• Decoke Air Rate – Generally an indicator of furnace size and consequently coke volume. Don’t
want to incentivize higher air rate.

• Recommended normalization method Avg LB/(Decoke MBTU/hr)

• (Avg lb/hr*decoke hr)/(Mbtu/hr)
Summary
• HAP Sources:
• Combustion byproducts of fuel or coke
• Tube metallurgy oxidation
• Independent of decoke control device
• Process Leaks – These are controllable

• For gas crackers, decoke control devices aren’t notably different with respect to
emissions.
• Requirements should allow for all common feeds
• Process isolation away from the decoke drum is best performance in HAP reduction

• Online HAP Reduction:

• Requirements should allow for large fuel gas composition range
• Conduct periodic burner health checks to ensure complete combustion and confirmation
with CO measurement
Backup
Introduction to an Olefins Plant- The Basics
PROCESS GAS H2
COMPRESSION

C1 FUEL

C2=
C5--
C3=
C4's
and
C5's
HYDROCARBON COLD ENDS
FEED SEPARATION
BZ/SCN AND
CRYOGENIC
PROCESSING

SCGO
STEAM

STEAM TAR
CRACKING HOT ENDS
FURNACE FRACTIONATION
Measuring Decoke Flow
• Looked at when decoke flow rates and temperatures were constant
• Expect the measured testing flow to be constant
• Instead see ~30% variation

Pitot Tube Measurements of Constant Flow

3.0

2.5

2.0
Sqrt (ΔP)

1.5 Variation

1.0

0.5

0.0
Traverse Points
Decoke Stack Online Furnace Stack Decoking Furnace Stack
Effects on Emissions Rates
• Potential for compounding error could significantly vary results
HAP Emission Variation
140%
13%

120%
29%
• Potential range for a given
emission rate 100%

• Only including estimated 29%

80%
flow and measurement
uncertainty 13%
60%

40%
HAP Bias Low Flow Bias Low Measured Flow Bias High HAP Bias High