Page 1
ORIGINAL PAPER
A reduced reaction mechanism for the simulation in ethyleneflare combustion
Helen H. Lou • Christopher B. Martin • Daniel Chen •
Xianchang Li • Kyuen Li • Hitesh Vaid • Anjan Tula Kumar •
Kanwar Devesh Singh • Doyle P. Bean Jr.
Received: 6 January 2011 / Accepted: 31 May 2011
� Springer-Verlag 2011
Abstract Industrial ethylene flares are considered to be a
probable major source of volatile organic compounds
(VOCs) such as formaldehyde. VOCs are chemicals that
are responsible for the formation of other atmospheric
pollutants like ozone. Due to the difficulty and cost of field
measurements, on-line monitoring is not practical and other
methods must be employed. Current methodologies for
calculating speciated and total VOC emissions from flaring
activities generally apply a simple mass reduction to the
VOC species sent to the flare that does not consider the
production of incomplete combustion or other intermedi-
ates. There arises a need of a speciation study for the
inspection of these flare for their emission. However, most
of the detailed kinetic mechanisms for the speciation study
of flaring events are too complex, consist of large number
of reactions and species, and also are computationally
expensive. A reduced mechanism will thus be desirable
for improving computational efficiency. In this study, a
reduced mechanism for simulating ethylene flare combus-
tion is presented. By retaining the important features of the
detailed mechanism in the form of elementary reactions,
and satisfying the species constraint of commercial CFD
packages, the reduced mechanism, thereby, is useful for
speciation study of flaring event.
Keywords Combustion � Reduced mechanism �Ethylene flare
Introduction
According to the US Environmental Protection Agency
(EPA), industrial operations dispose of 100 million tons of
pollutants into the atmosphere every year. The flaring
emissions during chemical plant operations generate huge
amounts of CO, CO2, NOx, volatile organic compounds
(VOC) such as formaldehyde and highly reactive VOC
(HRVOC) (defined in Texas air quality regulation as eth-
ylene, propylene, isomers of butene and 1,3-butadiene).
VOCs mixed with NOx have been identified associating
with high concentrations of ozone observed in Texas
Houston/Galveston area, which violates the National
Ambient Air Quality Standards (NAAQS) for ozone (Nam
et al. 2008; EPA 2010). Industrial flare emissions from
ethylene plants are considered as a probable major source
of HRVOCs (USEPA 2010).
Results from the most recent Texas Air Quality Model
Simulation of Houston study indicated that neither the
current HRVOC emission inventory (EI) nor adding more
HRVOC in EI could explain the high ozone scenarios in
the Houston–Galveston Area (Byun et al. 2007; Jiang and
Fast 2004). These results also indicated that the models
were very limited in their ability to predict radical inter-
mediates and formaldehyde (CH2O or HCHO), which
could be a great radical source if it was produced as a
primary source from flaring (Byun et al. 2007). However,
no HCHO has been reported as a primary emission from
H. H. Lou (&) � D. Chen � K. Li � H. Vaid �A. T. Kumar � K. D. Singh � D. P. Bean Jr.
Dan F. Smith Department of Chemical Engineering,
Lamar University, Beaumont, TX 77710, USA
e-mail: [email protected]
C. B. Martin
Department of Chemistry and Physics, Lamar University,
Beaumont, TX 77710, USA
X. Li
Department of Mechanical Engineering, Lamar University,
Beaumont, TX 77710, USA
123
Clean Techn Environ Policy
DOI 10.1007/s10098-011-0394-9
Page 2
flaring. All of the VOC emissions in the inventories are
assumed to be reactive organic gas (ROG), and a emissions
processing software, SMOKE (Simon et al. 2007) was used
to convert the VOC to total organic gas (TOG) before
converting the emissions into speciation for the air quality
models (TCEQ 2007).
Research has been performed on the underrated emis-
sions in experimental flares (Kostiuk et al. 2000; USEPA
1983; URS corporation 2004; Castineira and Edgar 2008).
These scientific findings by the TexAQS2000 study
(NOAA Aeronomy Laboratory 2003) have shown that
ambient VOC emissions sampled by aircraft did not coin-
cide with the annual emissions inventory. Consequently, it
was inferred that some VOC sources were in the emission
inventory or may be recognized as sources, but were sig-
nificantly under reported.
Flare emissions may vary based on actual flare operation
and there may be more variables that affect flare operation
than were identified in the previous studies. Thus, it is
desirable to be able to determine speciated emissions and
combustion efficiency during actual operation (Peters and
Kee 1987). Field measurement tests have also been done
using techniques such as Passive Fourier Transform Infrared
(PFTIR) and Resonance-Enhanced Multiphoton Ionization
(REMPI). PFTIR was used to measure emissions and com-
bustion efficiency from an actual flare, a study sponsored by
TCEQ (Simon et al. 2007). Due to the problems such as
minimum detection limit and accuracy of the system setup,
development of a practical and reliable routine detection of
VOC’s through field measurement is still under study. Field
measurements are also difficult and costly, which contrib-
utes to the current lack of field deployment. Therefore, a
reliable and practical alternative, such as a computational
model for flare speciation, is greatly needed.
Computational methods employed to predict the amount
of under-estimated VOCs, like formaldehyde, require
solving the governing chemical transport equations with
detailed kinetic mechanisms. Detailed chemical kinetic
mechanisms for some of the fuels are available, but they
contain tens to hundreds of species and hundreds to thou-
sands of reactions. Hence, solving equations of hundreds of
species with such complicated mechanisms coupled with
continuity, momentum, energy, radiation, and gravity
equations in CFD grid becomes computationally expensive
and in most cases not feasible. At the same time, the
capacity of the commercial CFD packages like ANSYS
FLUENT 6.3.26 is limited and cannot handle more than 50
species in the simulation. Therefore, there is a need for the
reduction of the detailed kinetic mechanism in a way that
the number of species falls below the allowable limit of
commercial CFD packages. While still retaining important
features of the detailed mechanism, thus making the
chemistry less restrictive and also improving the
computational efficiency. The reduced mechanism pro-
posed in this article can be used further for modeling eth-
ylene flare combustion process in a follow up study.
A review on available methods for the reduction
of detailed chemical kinetic mechanism
Modeling turbulent combustion requires expensive com-
putational resources and thus there is a need of reduced
mechanisms. There are several ways to reduce detailed
mechanism. Among them, the two major methods are the
skeletal reduction and time scale analysis. Skeletal mech-
anisms are on the same form as detailed mechanisms with
the standard Arrhenius elementary reactions. Skeletal
reduction is achieved with different methods including
sensitivity analysis, principal component analysis, lump-
ing, genetic algorithms, optimization, and adaptive reduc-
tion. On the other hand, time scale reduction is based on the
quasi steady-state approximation method (QSSA) (Peters
1988) and the partial equilibrium (PE) method. Other
methods based on QSSA have been reported in several
literatures (Androulakis 2000; Bhattacharjee et al. 2003;
Petzold and Zhu 1999; Chen 1988).
Research efforts have been conducted in both fields and
can be found in literatures (Simon et al. 2007; Brink and
Kilpinen 1999). Work was also done to reduce reaction
mechanism for combustion applications (Peters and Rogg
1993). These techniques reduce the number of reactions too
much by forming global reactions, which may not be able to
take advantage of the involved elementary reactions.
Another drawback is that these techniques may give reduced
mechanism in codes which are difficult to manage due to the
difference in programming language interface. Some of the
software packages for automatic mechanism reduction are
discussed here briefly, which are listed in the following
sections.
KINALC
KINALC is a postprocessor to CHEMKIN (Tomlin et al.
1992) based simulation programs developed by the Com-
bustion Simulations group at the University of Leeds
(Tomlin et al. 1997). It basically conducts three types of
analysis: processing sensitivity analysis results, extraction
of information from reaction rates, and stoichiometry, and
provides information about redundant and QSS species.
Species that are the least connected to the important spe-
cies can be eliminated from the mechanism. KINALC
facilitates the accurate selection of QSSA species by pre-
dicting the instantaneous error of QSSA species (URS
Corporation 2004).
H. H. Lou et al.
123
Page 3
RIOT (Range Identification and Optimization Toolkit)
RIOT is a software tool to reduce the number of species and/
or reactions in a reaction mechanism, while maintaining user
specified tolerances on the accuracy of the reduced mecha-
nism (Schuchardt et al. 2005). RIOT is very handy at per-
forming an initial reduction of very large models, but it lacks
the ability of working with irreversible reactions, which is
the case with most combustion mechanisms.
Computational Singular Perturbation (CSP)
CSP developed by Lam and Goussis is based on a metic-
ulous timescale analysis, where fast and slow subspaces of
the chemical source term are defined and a reduced
mechanism with less stiffness is formulated. The CSP
method constructs reduced mechanisms consisting of
‘‘virtual global steps,’’ with rates as linear combinations of
the rates of the elementary reactions that form the detailed
mechanism (Neophytou et al. 2004).
CARM (Computer Assisted Reduction Mechanism)
CARM is an interactive code used for automatic generation of
reduced chemistry following the procedure developed by Chen
(NOAA Aeronomy Laboratory 2003). The base of CARM
code is the general idea of quasi-steady-state (QSS) assump-
tion for reducing the complex reaction mechanism. It identifies
the quasi-steady-state species by the criterion in Eq. 1:
xpk � xc
k
��
��
max xpk
����; xc
k
����
� �� d; ð1Þ
where the production rate, xpk ; and consumption rate, xc
k; of
the kth species are given by Eqs. 2 and 3, respectively.
xpk ¼
XI
i¼1
max mki; 0ð Þqi ð2Þ
xck ¼
XI
i¼1
min mki; 0ð Þqi: ð3Þ
The relative error d is usually far less than 1 and set
between 10-2 and 10-3. mki is the stoichiometric coefficient
of the kth species in the ith reaction, qi is the rate of
progress variable for the ith reaction step, and I is the total
number of reaction steps. CARM is used with CHEMKIN
and other flame codes to develop and test reduced
mechanisms. Once the reduced mechanism is built,
CARM automatically generates the source code in
FORTRAN, which can be used in CHEMKIN or can be
used as an individual mechanism (Chen et al. 1997).
In the present work, a new methodology based on the idea
of skeletal mechanism and the basic facts of mass action
kinetics was developed for the formation of reduced mech-
anism. This methodology utilizes simulation result of
CHEMKIN PSR reactor, which is a zero-dimensional
kinetic solver. The simulation result consists of information
on species concentration, residence time, temperature, and
pressure. The main advantage of this methodology lies in the
fact that it preserves the important reaction pathway in their
elementary form and produces reduced reaction mechanism
in standard CHEMKIN format and hence it is easy to use
without any user programming. The reaction mechanism file
needed for CHEMKIN was a combined mechanism formed
using two widely used chemical kinetic mechanisms, i.e.,
GRI-3.0 (Smith et al. 2000) and USC (Wang and Laskin
1998). The use of two mechanisms to form a combined
mechanism has been detailed later in this study.
Introductory information about the reduction
procedure
Two widely used mechanism, GRI 3.0 and USC (75 species),
are available for the CFD simulation of flaring. The GRI-
Mech 3.0 performs well for an extensive range of combus-
tion conditions, which has been evaluated and shown on their
website. The USC mechanism consisting of 75 species is a
comprehensive kinetic model for representing ethylene and
acetylene combustion. It has been evaluated for predicting
combustion properties of both C2 and C3 fuels. However, due
to some of their limitation, as described below, none of them
were used independently for the present work.
1 GRI-3.0 mechanism (with 53 species) was developed
and optimized for the combustion of methane and not
ethylene. A few aspects of natural gas combustion
chemistry are not described by GRI-Mech 3.0; these
include soot formation and the chemistry involved in
selective non-catalytic reduction of NO. The latter may
be important in natural gas reburning at lower
temperatures.
2 USC mechanism (containing 75 species) was optimized
for ethylene combustion reactions, but the absence of
NOx producing species in the mechanism was not
desirable.
To overcome this problem, both of the reaction mech-
anisms were combined so as to yield a mechanism which
could satisfy all the above-mentioned criteria. Consider-
ation of the NOx producing species contained in the GRI
mechanism gives more satisfactory results for the com-
bustion of ethylene than the USC mechanism alone. The
combined GRI-USC mechanism consists of 93 species;
however, to run the mechanism for CFD modeling, the
CFD simulation package, FLUENT (Lu and Law 2005),
can only handle a maximum of 50 species. Thus, the
A reduced reaction mechanism for the simulation in ethylene flare combustion
123
Page 4
number of species in the combined GRI-USC mechanism
has to be reduced from a total of 93 to at least 50 species,
while still maintaining the accuracy in the predictability of
the reduced mechanism.
To reduce the number of species from the combined
mechanism, a new algorithm, which is illustrated in Fig. 1,
was developed for screening out those species not critical
to the accuracy of the reduced mechanism. The algorithm
basically consists of three major steps, as shown in Fig. 2.
The major steps are time-based sensitivity analysis,
examination of species in the reaction pathway, and pro-
gress variable analysis. These steps are interlinked through
an iterative process as explained in this section. Sensitivity
analysis was conducted using solution file containing data
for exit mole fraction of species at a range of different
residence time. The solution file was obtained by solving
the mechanism in a PSR reactor model of CHEMKIN at
conditions specified in Table 1.
Outline for the reduction procedure
This section presents the outline of the basic algorithm
described in Fig. 1. The methodology involved the
inspection of species based on the following factors, in
order to determine the criticality of a particular species and
its fate in the reduced mechanism. The factors considered
for the evaluation of species are discussed below in detail.
Step 1: Time-based sensitivity analysis of each species
was performed at a total of four different residence times
for each intermediate reduction. These residence times are
typical in an open flare.
i. 2.00E-05 s
ii. 9.00E-05 s
iii. 5.00E-01 s
iv. 1.00E?00 s
The general trend of the species concentration over time
was analyzed, and the maximum mole fraction of each
species was used to identify the critical species. The
highest mole fraction was thus used as one of the factor for
the identification of critical species.
USC + GRI Species
Contains nitrogen?
MMF > 500 ppb
# of rxns 20
MMF > 25 ppb
Keep
Remove
MMF > 100 ppb
# of rxns 8
# of rxns 10
Keep
# of rxns 17
Any HCHO rxns?
Rate constant in slowest quartile
Remove
Legends
No
Yes
Flow
MMF Maximum mass
Fig. 1 Decision tree for identifying the critical species
Time based sensitivity analysis
Reaction Pathway Analysis
Reaction rate constant analysis
Combined mechanism consisting of 93 species
CHEMKIN4.1.1 PSR
reactor
Output file containing mole fraction of
species at the exit of the PSR reactor
Reduced reaction
mechanism
Fig. 2 Overview of the algorithm
H. H. Lou et al.
123
Page 5
The species with extremely low values of maximum mole
fraction were considered least critical and removed from the
combined mechanism in each of the subsequent steps.
Step 2: The total number of reactions, in which a par-
ticular species is involved, were listed and counted. The
importance of this can be seen in Eq. 4, which presents the
net molar reaction rate of a species. Species which were
involved in a large number of reactions were considered
CRITICAL due to the possibility of their relatively higher
net molar rate of production. Species with a moderate
number of reactions were further analyzed, while the
remaining with few reactions (e.g., 1 or 2 non-HCHO
reactions) was removed.
net molar reaction rate of species
¼Xnr
i¼1
rate of production: ð4Þ
In mathematical form, it is expressed as:
rrk ¼Xnr
i¼1
aiqi; ð5Þ
where ai is the stoichiometric coefficient of the species in
the reaction and qi is the progress variable containing the
concentration terms.
The numbers of reactions corresponding to each species
that also involved HCHO were counted. Species involved
in a very low number of HCHO reactions were considered
least critical in the reaction pathway of formaldehyde and
therefore were removed. Those species with high or mod-
erate number of HCHO reactions were kept in the reduced
mechanism. The species with an intermediate number of
HCHO reactions were further analyzed in Step 3.
Step 3: Values of the rate constant: Forward rate con-
stant of each reaction (involving the species under con-
sideration and HCHO) was calculated and analyzed by
assuming the bath temperature to be 2100 K. The assumed
bath temperature is approximately equal to the temperature
at the center of the flare. If the species under consideration
involve mostly the reactions that have low values of rate
constant and also have low concentration trend as dis-
cussed above, the particular species were removed. For-
ward rate constant and concentration of species can be
linked to progress variable as shown in Eq. 6.
Table 1 Input conditions for the CHEMKIN PSR reactor
Inlet fuel and oxidizer temperature 500 K
Equivalence ratio of fuel to oxidizer 1.0
Reactor temperature 1700 K
Table 2 List of species involved in the mechanisms
Mechanism Number
of species
Species list
Mechanism-1 93 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
HO2, C, C2O, CH4, C2H, CH2*, H2CC, HCN, HCCOH, CH2OH, CH2CHO, N, HNCO, H2O2, HCNO, C4H2,
C3H3, C3H2, CH3OH, NCO, NH, CH3CHO, pC3H4, aC3H4, N2O, CH3O, C2H5, aC3H5, NH2, CN, NNH,
C2H3CHO, HNO, CH3CO, NH3, C4H4, i-C4H3, HOCN, NO2, H2C4O, C4H, C3H6, n-C4H3, HCNN, C2H6,
CH3CCH2, H2CN, CH3CHCH, cC3H4, C6H2, nC3H7, C4H6, i-C4H5, n-C4H5, l-C6H4, iC3H7, C4H7, C6H3,
c-C6H4, C5H5, C3H8, A1, C6H5O, C6H5OH, C5H6, C5H4O, C5H4OH, C5H5O, C3H7
Mechanism-2 80 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
HO2, C, C2O, CH4, C2H, CH2*, H2CC, HCN, HCCOH, CH2OH, CH2CHO, N, HNCO, H2O2, HCNO, C4H2,
C3H3, C3H2, CH3OH, NCO, NH, CH3CHO, pC3H4, aC3H4, N2O, CH3O, C2H5, aC3H5, NH2, CN, NNH,
C2H3CHO, HNO, CH3CO, NH3, C4H4, i-C4H3, HOCN, NO2, H2C4O, C4H, C3H6, n-C4H3, HCNN, C2H6,
CH3CCH2, CH3CHCH, cC3H4, C6H2, C4H6, i-C4H5, l-C6H4, iC3H7, C4H7, C3H8, A1, C6H5OH, C5H5O
Mechanism-3 70 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
HO2, C, C2O, CH4, C2H, CH2*, H2CC, HCN, HCCOH, CH2OH, CH2CHO, N, HNCO, H2O2, HCNO, C4H2,
C3H3, C3H2, CH3OH, NCO, NH, CH3CHO, pC3H4, aC3H4, N2O, CH3O, C2H5, aC3H5, NH2, CN, NNH,
C2H3CHO, HNO, CH3CO, NH3, C4H4, i-C4H3, HOCN, NO2, H2C4O, C3H6, n-C4H3, HCNN, C2H6,
CH3CHCH, C6H2, C4H6, C5H5O
Mechanism-4 60 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
HO2, C, C2O, CH4, CH2*, H2CC, HCN, HCCOH, CH2OH, CH2CHO, N, HNCO, H2O2, HCNO, C4H2, C3H3,
C3H2, CH3OH, NCO, NH, CH3CHO, pC3H4, aC3H4, N2O, CH3O, C2H5, aC3H5, NH2, CN, NNH, C2H3CHO,
HNO, NH3, C4H4, HOCN, NO2, C3H6, HCNN, C2H6, C5H5O
Mechanism-5 50 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
HO2, CH4, CH2*, H2CC, HCN, CH2OH, CH2CHO, N, HNCO, HCNO, C4H2, NCO, NH, pC3H4, aC3H4, N2O,
CH3O, C2H5, aC3H5, NH2, CN, NNH, HNO, NH3, HOCN, NO2, C3H6, HCNN, C2H6, C5H5O
Mechanism-6 40 N2, H2O, CO, CO2, O2, OH, H2, H, O, C2H4, C2H2, CH2CO, CH3, C2H3, CH2O, NO, HCO, HCCO, CH2, CH,
CH4, H2CC, HCN, CH2OH, CH2CHO, N, C4H2, NCO, NH, N2O, CH3O, C2H5, aC3H5, NH2, NNH, HNO,
NH3, NO2, C3H6, C2H6
A reduced reaction mechanism for the simulation in ethylene flare combustion
123
Page 6
Table 3 Exit gas mole fraction
at the residence time
of 2.00E-05 s
CHEMKIN mole fraction results
Mechanism 1 2 3 4 5
# Species 93 70 60 50 40
N2 0.6978 0.6978 0.6979 0.6981 0.6982
AR 0.00E?00 0.00E?00 0.00E?00 0.00E?00 0.00E?00
C2H2 3.16E-03 3.16E-03 3.15E-03 3.40E-03 3.72E-03
CH2CHO 3.88E-05 3.89E-05 3.93E-05 3.87E-05 3.22E-05
CH2CO 1.66E-03 1.66E-03 1.68E-03 1.61E-03 1.43E-03
CH2O 4.94E-04 4.94E-04 4.94E-04 4.80E-04 2.17E-04
CO 7.60E-02 7.59E-02 7.59E-02 7.50E-02 7.46E-02
CO2 2.05E-02 2.05E-02 2.04E-02 2.13E-02 2.18E-02
H 3.74E-03 3.74E-03 3.74E-03 3.87E-03 3.87E-03
H2 2.19E-02 2.19E-02 2.19E-02 2.15E-02 2.15E-02
H2CC 2.76E-05 2.76E-05 2.76E-05 2.85E-05 2.06E-05
H2O 7.58E-02 7.58E-02 7.57E-02 7.62E-02 7.62E-02
HNO 6.52E-09 6.19E-09 4.71E-09 5.68E-09 1.68E-09
N2O 3.15E-08 3.15E-08 3.15E-08 3.12E-08 1.76E-08
NO 3.43E-06 3.72E-06 4.13E-06 4.93E-06 2.14E-06
O 1.03E-03 1.03E-03 1.03E-03 1.04E-03 1.12E-03
O2 8.59E-02 8.59E-02 8.59E-02 8.53E-02 8.56E-02
OH 2.68E-03 2.68E-03 2.68E-03 2.85E-03 2.93E-03
Table 4 Exit gas mole fraction
at the residence time of
1.00E?00 s
* in CH2* shows that it is
singlet methylene when
compared to CH2, which is
triplet methylene
CHEMKIN mole fraction results
Mechanism 1 2 3 4 5
# Species 93 80 70 60 50
N2 0.7364 0.7364 0.7364 0.7364 0.7362
AR 0.00E?00 0.00E?00 0.00E?00 0.00E?00 0.00E?00
C2H2 5.62E-06 5.61E-06 5.6E-06 5.6E-06 4.85E-06
C2H3 2.66E-08 2.65E-08 2.65E-08 2.65E-08 2.61E-08
CH2* 8.96E-10 8.96E-10 8.96E-10 8.99E-10 1.17E-09
CH2CHO 2.44E-10 2.45E-10 2.45E-10 2.45E-10 2.59E-10
CH2CO 4.89E-07 4.88E-07 4.88E-07 4.90E-07 5.36E-07
CH2O 2.30E-07 2.30E-07 2.30E-07 2.31E-07 2.59E-07
CO 3.67E-03 3.68E-03 3.68E-03 3.68E-03 3.93E-03
CO2 1.27E-01 1.27E-01 1.27E-01 1.27E-01 1.27E-01
H 8.85E-05 8.87E-05 8.87E-05 8.89E-05 1.04E-04
H2 1.07E-03 1.07E-03 1.07E-03 1.07E-03 1.15E-03
H2CC 2.21E-10 2.21E-10 2.21E-10 2.21E-10 2.26E-10
H2O 0.1291 0.1291 0.1291 0.1291 0.129
HNO 3.77E-09 3.78E-09 3.60E-09 3.12E-09 3.81E-09
HO2 3.78E-07 3.80E-07 3.80E-07 3.80E-07 3.16E-07
N 3.78E-10 3.79E-10 3.66E-10 2.15E-10 2.59E-10
N2O 8.27E-08 8.28E-08 8.25E-08 8.03E-08 8.45E-08
NO 4.66E-05 4.66E-05 4.44E-05 5.48E-05 6.22E-05
O 3.68E-05 3.69E-05 3.69E-05 3.70E-05 4.37E-05
O2 2.24E-03 2.24E-03 2.24E-03 2.24E-03 2.38E-03
OH 5.51E-04 5.51E-04 5.51E-04 5.52E-04 6.01E-04
H. H. Lou et al.
123
Page 7
qi ¼ kfi � Xi; ð6Þ
where kfi is the forward rate constant and Xi is the con-
centration of species i.
The above equation shows that (qi) can be used as a
substitute for the combination of the forward rate constant
and concentration. As a result of it, we can use (qi) as an
index for identifying critical species. The forward rate
constants kfi
� �
of individual reactions are calculated in a
spreadsheet using Arrhenius equation as given below.
kfi ¼ AiTbi exp
�Ei
RcT
� �
; ð7Þ
where Ai, bi, and Ei are pre-exponential factor, temperature
exponent, and activation energy, respectively, which are all
supplied in the reaction mechanism file for all the
reactions.
In Step 1, the species falling below a specific maximum
mole fraction limit were removed to trim the mechanism.
This step was followed by a manual inspection of species
for their involvement in the number of reactions and also in
reactions linked, directly or indirectly, with pollutant of
interest. Eventually, the intermediate mechanism was
analyzed for progress variable (qi), which takes into
account the rate constant and the maximum concentration.
The intermediate mechanism was then re-run in the
CHEMKIN PSR reactor using the same set of condition as
supplied for the combined mechanism. This was done in
order to check the accuracy of exit mole fraction predic-
tion, as compared to the combined mechanism. If the
results were within acceptable tolerance, the intermediate
mechanism was sent for further reduction. Intermediate
reduced mechanisms were formed after completion of
iteration of the reduction procedure. Consequently, several
reduced mechanisms were obtained from the reduction
procedure. After each step of species removal, the resulting
mechanism was re-run in CHEMKIN PSR reactor at the
specified condition shown in Table 1. This method allows
the effect of each species removal to be analyzed.
Results
This section discusses the effect of species removal at each
step on the accuracy of the reduced mechanism. Several
mechanisms other than the combined 93 species mecha-
nism were produced using the reduction algorithm. The
lists of species of some of the mechanisms are shown in
Table 2.
Each mechanism was simulated in CHEMKIN PSR
reactor with the same set of condition as mentioned above,
but the residence time was varied. The results (exit gas
mole fraction) of each mechanism were then compared to
the original 93 species mechanism, to check its accuracy.
Tables 3 and 4 show the exit gas mole fraction of some
critical species at residence time of 2.00E-05 and
1.00E?00 s, respectively. It is clear that the accuracy of
the reduced mechanism decreases with each trim. Bulk
species such as N2 and O2 show good agreement, but the
errors are increasing among intermediate species such as
CH2O and NO. Desired accuracy of the mechanism will be
subjective to the system of interest. The targeted end use of
the reduced mechanism in this study is the speciation study
of the flaring system. Such study requires prediction of
bulk, intermediate, and radical species with sufficient
accuracy. Hence, the mechanism with 50 species was
selected due to its highest accuracy among the mechanism
which meets the computational restriction. Other systems
with only flow dynamics or heat generation modeling can
opt for the mechanism with 40 species.
The accuracy of the reduced mechanism is demonstrated
by the absolute error in the mole fraction of each species
analyzed. As shown in Table 5, the highest error in the exit
gas mole fraction is in the order of E-09. For the most
critical species HCHO, the mole fraction value has an error
Table 5 Errors in the reduced mechanism as compared to the ori-
ginal mechanism
Highest errors in reduced mechanism mole fractions*
S. no. Compound Full mechanism
(93 species)
Mechanism
(50 species)
Absolute
error
1 C3H8 3.90E-11 4.08E-09 4.04E-09
2 C3H3 1.14E-07 4.28E-07 3.14E-07
3 CH3CHO 3.88E-07 7.27E-07 3.40E-07
4 CH3O 9.73E-08 4.11E-08 -5.63E-08
5 HNCO 4.26E-07 2.31E-07 -1.94E-07
6 NH2 2.83E-08 1.54E-08 -1.29E-08
7 i-C4H3 1.63E-08 8.90E-09 -7.38E-09
8 CH 5.23E-06 7.27E-06 2.04E-06
9 HCNN 8.29E-09 1.12E-08 2.95E-09
10 C3H6 3.09E-07 4.17E-07 1.08E-07
11 pC3H4 6.83E-07 9.16E-07 2.33E-07
12 HCN 2.90E-06 3.77E-06 8.75E-07
13 aC3H4 3.83E-07 4.98E-07 1.15E-07
14 C4H4 3.31E-08 2.48E-08 -8.21E-09
15 n-C4H3 3.70E-09 2.78E-09 -9.14E-10
16 NO 2.67E-05 3.28E-05 6.19E-06
17 N 1.78E-07 2.15E-07 3.72E-08
18 CH2OH 1.40E-06 1.12E-06 -2.81E-07
19 aC3H5 1.31E-07 1.57E-07 2.60E-08
20 NCO 4.39E-08 5.19E-08 8.01E-09
… … … … …44 CH2O 7.16E-05 7.08E-05 1%
*Residence time = 9E-05 s, T = 500 K, % stoich., air = 100%
A reduced reaction mechanism for the simulation in ethylene flare combustion
123
Page 8
0.00E+00
5.00E-06
1.00E-05
1.50E-05
2.00E-05
2.50E-05
3.00E-05
3.50E-05
4.00E-05
4.50E-05
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Mol
e F
ract
ion
Residence Time (seconds)
93 species
80 species
70 species
60 species
50 species
HCHO (500K, 100% Air)
Fig. 3 Comparison of HCHO
mole fraction versus residence
time for all the mechanism
0.00E+00
1.00E-05
2.00E-05
3.00E-05
4.00E-05
5.00E-05
6.00E-05
7.00E-05
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Mol
e F
ract
ion
Residence Time (seconds)
93 species
80 species
70 species
60 species
50 species
NO(500K, 100% Air)
Fig. 4 Comparison of NO mole
fraction versus residence time
for all the mechanism
H. H. Lou et al.
123
Page 9
0.00E+00
5.00E-07
1.00E-06
1.50E-06
2.00E-06
2.50E-06
3.00E-06
3.50E-06
4.00E-06
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Mol
e F
ract
ion
Residence Time (seconds)
93 species
80 species
60 species
50 species
70 species
HCN(500K, 100% Air)
Fig. 5 Comparison of HCN
mole fraction versus residence
time for all the mechanism
0.00E+00
5.00E-06
1.00E-05
1.50E-05
2.00E-05
2.50E-05
3.00E-05
3.50E-05
4.00E-05
4.50E-05
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Mol
e F
ract
ion
Residence Time (seconds)
93 species
80 species
70 species
60 species
50 species
CH2CHO (500K, 100% Air)
Fig. 6 Comparison of
CH2CHO mole fraction versus
residence time for all the
mechanism
A reduced reaction mechanism for the simulation in ethylene flare combustion
123
Page 10
percentage of as low as 1%. Therefore, it can be concluded
that this mechanism is fairly accurate when considered for
the prediction of formaldehyde formation during the flaring
event involving ethylene combustion.
To further investigate the accuracy of the mechanism,
some of the species besides HCHO, i.e. NO, HCN,
CH2CHO, and CH3CHO, were also analyzed. This analysis
was conducted by comparing the mole fraction of each
species at different residence time for the combined and all
other reduced mechanisms. The comparison of the results
is illustrated in Figs. 3, 4, 5, 6, and 7 for species HCHO,
NO, HCN, CH2CHO, and CH3CHO, respectively. For
instance, Fig. 3 shows the variation of exit gas mole frac-
tion for HCHO at different residence time, and Table 6
shows the data points used in Fig. 3. The five different
curves in Fig. 3 represent different reaction mechanisms
with different numbers of species. All the curves show
good agreement with the 93 species reaction mechanism at
most of the residence times. This clearly shows that even
though the number of species was gradually reduced from
93 to 50, the reduced mechanism agrees with the original
combined mechanism.
Figure 4 shows a similar type of curve for NO mole
fraction versus residence time, calculated using different
mechanisms. Again, all of the curves are in close agree-
ment with each other. It should be noted that the highest
error noted here is in the order of 1.00E-05, which is very
low. Similar observations can be made for other species.
Comparison with other methods
The reduction procedure presented here has some unique
features and advantages, as compared to other methods. It
has the advantage of handling reversible reaction when
compared to RIOT. The reaction mechanism can be
reduced targeting a set of particular species. For example,
simulation which needs to evaluate radical emission for
ozone formation will need to focus reduction of mechanism
around VOCs such as formaldehyde and NOx.
Also the formation of global reactions steps, as with
CARM and CSP, reduces the system capability of taking
advantages of elementary reaction steps and hence may
result in undershoot or overshoot in prediction of radical
species. This undershoot phenomena was observed with a
0.00E+00
2.00E-06
4.00E-06
6.00E-06
8.00E-06
1.00E-05
1.20E-05
1.00E-05 1.00E-04 1.00E-03 1.00E-02 1.00E-01 1.00E+00
Mol
e F
ract
ion
Residence Time (seconds)
93 species
80 species
70 species
60 species
50 species
CH3CHO (500K, 100% Air)
Fig. 7 Comparison of
CH3CHO mole fraction versus
residence time for all the
mechanism
Table 6 Comparison of HCHO mole fraction versus residence time
for all the mechanisms
Residence time (s) 2.00E-05 9.00E-05 5.00E-01 1.00E?00
93 species 4.94E-04 7.16E-05 1.58E-06 4.89E-07
80 species 4.94E-04 7.16E-05 9.32E-07 4.88E-07
70 species 4.94E-04 7.16E-05 1.58E-06 4.88E-07
60 species 4.94E-04 7.15E-05 1.58E-06 4.90E-07
50 species 4.80E-04 7.08E-05 5.36E-07 5.36E-07
H. H. Lou et al.
123
Page 11
50 species reduced mechanism for ethylene, formed using
CARM. The formation of global reaction steps resulted in
under predicting of radical species such as formaldehyde
and NO by at least 14%. This undershoot was just under
1% for current method as shown in Table 5.
Conclusion
A general reaction mechanism reduction methodology was
developed so that a reduced mechanism can be formed
from any given detailed mechanism with different condi-
tions of interest. A reduced reaction mechanism of ethylene
was formed to overcome the computational limit of com-
mercial software. Numerical experiments demonstrate that
it is in full agreement with the combined mechanism. The
reduced reaction mechanism will enable speciation study
of HVOC’s such as formaldehyde during ethylene flaring
event through CFD simulation.
Acknowledgments The authors gratefully acknowledge the finan-
cial support from TARC (Texas Air Research Center), TCEQ (Texas
Commission on Environmental Quality), AQRP (Air Quality
Research Project) Project No.10-022 and SEP (State Environmental
Policy) Agreement No. 2009-009.
References
Androulakis IP (2000) Kinetic mechanism reduction based on an
integer programming approach. AIChE J 46(2):361–371
Bhattacharjee B, Schwer DA, Barton PI, Green WH (2003)
Optimally-reduced kinetic models: reaction elimination in
large-scale kinetic mechanisms. Combust Flame 135(3):191–208
Brink A, Kilpinen P (1999) Development of reduced mechanisms
applicable to CFD modeling of NOx emissions from biomass
combustion. A´
bo Akademi Process Chemistry Group, Finland
Byun W, Kim S, Kim SB (2007) Evaluation of air quality models for
the simulation of a high ozone episode in the Houston
metropolitan area. Atmos Environ 41:837–853
Castineira D, Edgar EF (2008) Computational fluid dynamics for
simulation of wind-tunnel experiments on flare combustion
systems. Energy Fuels 22:1698–1706
Chen JY (1988) A general procedure for constructing reduced
reaction mechanisms with given independent reactions. Combust
Sci Tech 57:89–94
Chen JY (1997) Development of reduced mechanisms for numerical
modeling of turbulent combustion. Workshop on Numerical
Aspects of Reduction in Chemical Kinetics, France
EPA, National Ambient Air Quality Standards (2010) http://www.
epa.gov/air/criteria.html. Accessed 30 December 2010
Jiang G, Fast JD (2004) Modeling the effects of VOC and NOx
emission sources on ozone formation in Houston during the
TexAQS 2000 field campaign. Atmos Environ 38(30):5071–
5085
Kostiuk L, Johnson M, Thomas G (2000) University of Alberta Flare
Research Project Interim Report
Lu TF, Law CK (2005) A directed relation graph method for
mechanism reduction. Proc Combust Inst 30:1333–1341
Nam J, Webster M, Kimura Y, Jeffries H, Vizuete W, Allen DT
(2008) Reductions in ozone concentrations due to controls on
variability in industrial flare emissions in Houston, Texas. Atmos
Environ 42:4198–4211
Neophytou MK, Goussis DA, van Loon M, Mastorakos E (2004)
Reduced chemical mechanisms for atmospheric pollution using
computational singular perturbation analysis. Atmos Environ
38(22):3661–3673
NOAA Aeronomy Laboratory (2003) Texas Commission on envi-
ronmental quality. Texas 2000 Air Quality Study. Phase II
Analysis of NOAA Data, Final Report
Peters N (1988) Systematic reduction of flame kinetics—principles
and details. In: 11th International Colloquium on Dynamics of
Explosions and Reactive Systems, Warsaw, Poland
Peters N, Kee RJ (1987) The computation of stretched laminar
methane-air diffusion flames using a reduced four-step mecha-
nism. Combust Flame 68:17
Peters N, Rogg B (1993) Reduced reaction mechanisms for applica-
tions in combustion systems, vol 15. Springer, Berlin
Petzold L, Zhu WJ (1999) Model reduction for chemical kinetics: an
optimization approach. AIChE J 45(4):869–886
Schuchardt K, Oluwole O, Pitz W, Rahn LA, Green WH, Leahy D,
Pancerella C, Sjoberg M, Dec J (2005) New approaches for
collaborative sharing of chemical model data and analysis tools.
U.S. Sections of the Combustion Institute, Philadelphia, PA
Simon H, Kimura Y, Allen D (2007) Texas commission on
environmental quality—processing a model-ready emissions
inventory for primary PM. www.tceq.state.tx.us/…/air/texaqs/
workshop/20070529/2007.05.31/Morning-2/3-Model-Ready_EI-
PM-Simon.pdf-2007-06-12. Accessed 03 November 2010
Smith GP, Golden GM, Frenklach M, Moriarty NW, Eiteneer B,
Goldenberg M, Bowman T, Hanson RK, Song S, Gardiner WC,
Lissianski VV, Qin Z (2000) http://www.me.berkeley.edu/gri_
mech/. Accessed 03 October 2010
Texas Commission on Environmental Quality (2007) Emissions
modeling-2.1 emissions modeling overview. http://www.tceq.
state.tx.us/assets/public/implementation/air/sip/haze/CHAPTE_
2.pdf. Accessed 03 August 2010
Tomlin AS, Pilling MJ, Turanyi T, Merkin JH, Brindley J (1992)
Mechanism reduction for the oscillatory oxidation of hydrogen:
sensitivity and quasi-steady-state analysis. Combust Flame
91:107–130
Tomlin AS, Turanyi T, Pilling MJ (1997) Mathematical tools for the
construction, investigation and reduction of combustion mech-
anisms. In: Pilling MJ (ed) Low temperature combustion and
autoignition. Elsevier, Amsterdam, pp 293–437
URS Corporation (2004) Passive FTIR phase I testing of simulated
and controlled flare systems. Final report, for Texas Commission
on Environmental Quality and University of Houston
U.S. Environmental Protection Agency (1983) Flare Efficiency Study,
EPA-600/2-83-052, Cincinnati, OH
U.S. Environmental Protection Agency (2010) http://www.epa.gov/
air/oaqps/permjmp.html. Accessed 29 December 2010
Wang H, Laskin A (1998) A comprehensive kinetic model of ethylene
and acetylene oxidation at high temperatures. Combustion
Kinetics Laboratory, Document, Internal report
A reduced reaction mechanism for the simulation in ethylene flare combustion
123