Professional Documents
Culture Documents
To cite this article: Zhaoyu Luo , Max Plomer , Tianfeng Lu , Sibendu Som & Douglas E. Longman
(2012) A reduced mechanism for biodiesel surrogates with low temperature chemistry for
compression ignition engine applications, Combustion Theory and Modelling, 16:2, 369-385, DOI:
10.1080/13647830.2011.631034
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Downloaded by [University of Leeds] at 13:07 12 November 2014
Combustion Theory and Modelling
Vol. 16, No. 2, 2012, 369–385
1. Introduction
Biodiesel is a promising alternative fuel that can be produced from renewable sources,
such as vegetable oil, animal fat, and waste cooking oil. It can be used in existing diesel
engines with reduced CO and unburned hydrocarbons (HC) emission without significant
changes to their design [1]. Biodiesel primarily consists of fatty acid methyl esters (FAME),
which feature the ester functional group and long carbon chains with a certain level of
unsaturation. The chemical kinetics of biodiesel combustion is highly complex due to its
large molecular size, and the detailed mechanisms for biodiesel can consist of over 3000
species and 10000 reactions [2–4]. Therefore, it is computationally expensive to perform
compression ignition (CI) engine simulations with detailed mechanisms for biodiesel,
∗
Corresponding author. Email: luozy@engr.uconn.edu
particularly when the negative temperature coefficient (NTC) range is involved. However,
predictive numerical simulations with realistic chemistry are still promising in reducing
the overall cost in engine design if the detailed mechanisms can be substantially reduced
to small sizes without significant loss of modelling accuracy. The technique of mechanism
reduction is thereby required to develop reduced mechanisms that can be used for practical
engine simulations.
Mechanism reduction has been extensively studied over the last few decades, and a
lot of reduction methods have been developed as reviewed in reference [5]. One of the
major reduction methods is skeletal reduction, which eliminates unimportant species and
reactions from detailed mechanisms. Unimportant species and reactions can be identified
with the methods based on sensitivity analysis [6,7], principle component analysis [8,9],
and Jacobian analysis and computational singular perturbation [10–12]. Skeletal reduction
can also be based on reaction rate and species flux analyses, e.g. detailed reduction [13],
Downloaded by [University of Leeds] at 13:07 12 November 2014
directed relation graph (DRG) [14], DRG with error propagation (DRGEP) [15], and the
path flux analysis [16]. These methods do not involve Jacobian matrix evaluation and
factorisation and typically feature low reduction costs. In particular, the DRG method
features a linear reduction time, such that it is highly efficient to reduce extremely large
mechanisms such as those for biodiesel. DRG was recently extended to accommodate expert
knowledge on chemical kinetics to develop skeletal mechanisms with higher accuracy
compared to the original DRG method [17]. It is noted that the DRG-based methods
can be combined with sensitivity analysis to further reduce the skeletal mechanisms, e.g.
using DRG-aided sensitivity analysis (DRGASA) [18,19] and DRGEP with sensitivity
analysis (DRGEPSA) [20]. The skeletal reduction can be followed by such methods as
lumping [21–25], in particular isomer lumping for large hydrocarbons [26,27], time scale
analysis such as quasi steady state approximations (QSSA) [28–32] and partial equilibrium
approximations [33,34], computational singular perturbation (CSP) [35–37], intrinsic low-
dimensional manifold (ILDM) [38–40], pre-image curve (PIC) [41,42], and tabulation
[43–45].
Despite the large number of methods available for mechanism reduction, research on
mechanism reduction for biodiesel is rather limited, particularly when low temperature
chemistry is involved. This is primarily because of the extreme sizes and complexities of
the biodiesel mechanisms, which render the reduction itself difficult and time consuming.
In previous works, methyl butanoate was selected as a biodiesel surrogate since it can
provide valuable insight into the combustion of methyl esters. A reduced mechanism with
low-temperature chemistry for methyl butanoate (MB) was obtained by Brakora et al. and
validated under biodiesel-fuelled engine conditions [46]. The chain length of MB is never-
theless short compared with that of real biodiesel thus it is not able to accurately represent
the physicochemical properties of biodiesel. Most notably, the lack of NTC behaviour ob-
served in experimental and kinetic studies [47] significantly reduces the applicability of MB
as a biodiesel surrogate. In addition, recent modelling studies by Som and Longman [48]
have shown that MB cannot accurately predict the flame lift-off and emission characteristics
of biodiesel combustion.
Methyl decanoate (MD) was later recognised as a more viable surrogate to biodiesel
fuels since it features both a long carbon chain and the ester group. MD reacts similarly
to biodiesel derived from rapeseed-based methyl esters (RME). A comprehensive and
accurate skeletal mechanism for MD with low-T chemistry was derived using DRG for
1-D flame analysis by Sarathy et al. [49]. However, this mechanism consists of more than
600 species and is not suitable for 3-D engine simulations. A mixture of MD, methyl-9-
decanoate (MD9D) and n-heptane was then selected as the biodiesel surrogate to better
Combustion Theory and Modelling 371
represent the biodiesel properties in terms of oxidation of unsaturated carbon bonds and
adjustment of ignition delays [3]. Skeletal mechanisms for the surrogate mixture were
recently developed for high temperature applications using DRG-based methods [50,51].
These skeletal mechanisms consist of approximately 120 species, but they are not suitable
for CI engine simulations that involve low temperature ignition near the NTC region.
In the present study, an integrated reduction approach that includes DRG, DRGASA
and isomer lumping is employed to obtain a skeletal mechanism for biodiesel surrogate with
low temperature kinetics. These methods have been shown highly effective in developing
skeletal mechanisms of biodiesel surrogates in the previous works [50,51]. The method of
DRGASA was stretched to the limit to test almost all the species in the mechanism such that
the smallest skeletal mechanism can be obtained for 3-D engine simulations. The validity
of the resulting skeletal mechanism was further shown by comparing the engine simulation
results with the experimental data under CI engine conditions.
Downloaded by [University of Leeds] at 13:07 12 November 2014
2. Methodologies
2.1. Mechanism reduction
The starting mechanism in the present reduction was developed by the Lawrence Livermore
National Laboratory (LLNL) [3]. The mechanism is for surrogate mixtures of MD, MD9D
and n-heptane and consists of 3299 species and 10806 elementary reactions. The tri-
component surrogate mixture allows flexibility in matching the fuel’s physical properties in
terms of oxygen mass fraction and molecular weight, and important combustion properties
such as ignition delay and flame lift-off distance in engine simulations by fine-tuning the
fuel composition. The reduction of the detailed mechanism was performed based on a large
set of reaction states sampled within the parameter space of pressures from 1 to 100 atm,
equivalence ratios from 0.5 to 2.0, initial temperatures from 700 to 1800 K for auto-ignition,
and inlet temperature of 300 K for extinction in perfectly stirred reactors (PSR). Note that
the NTC region that is important for engine ignition was covered in the reduction. Jet stirred
reactors (JSR) with diluted mixtures and intermediate temperatures were also included in
the sampling. The fuel mixture consists of 25% MD, 25% MD9D and 50% n-heptane by
mole. It is noted that the fuel mixture composition in the present study is slightly different
from previous work [51]. Since the reduction was performed using DRG-based methods, it
has been shown in reference [51] that the resulting skeletal mechanisms for fuel mixtures
usually feature good extensibility predicting ignition of mixtures with various compositions.
The worst case error tolerance was set to be 40%. It was found that such an error tolerance
was comparable to the overall uncertainty of the detailed mechanism, as such there is no
significant loss in chemical fidelity through the skeletal reduction [51]. It will be further
shown in the next section that the resulting skeletal mechanism performs well compared
with the detailed mechanism and also in predicting the experiments in some cases.
A DRG method that was improved for robust reduction of mechanisms with large
numbers of isomers was employed as the first step in the reduction. The DRG method
resolves species coupling by introducing a parameter rAB that indicates the relative error
induced to species A if species B is removed from the detailed mechanism [14]. An
error tolerance can be specified for the DRG reduction such that species B is considered
important to species A if rAB is larger than the error tolerance at any reaction state covered
in the parameter space. As discussed in reference [51], the extent of reduction with the
original DRG method is limited by the large number of isomers in detailed mechanisms of
large hydrocarbons, while the improved DRG method removed this restriction, such that
372 Z. Luo et al.
a reasonable large reduction error tolerance, say 20∼40%, can be specified without the
elimination of important species. The improved DRG method also features high efficiency
and has been implemented for parallel computation. It takes only minutes to process tens
of thousands of sampled reaction states using a small PC cluster. More detailed discussion
of the improved DRG method can be found in reference [51].
H radical was selected as the starting species in the present reduction since it is important
for hydrocarbon oxidation under almost all the conditions. After several iterations of the
DRG method and by allowing the worst-case reduction error of approximately 30%, a
skeletal mechanism with 664 species and 2672 reactions was first obtained. Note that
this skeletal mechanism is comparable in size to that for pure MD in reference [49],
but substantially larger than that for high temperature combustion of the tri-component
surrogate mixtures in reference [51].
The 664-species skeletal mechanism was then reduced with isomer lumping [26],
Downloaded by [University of Leeds] at 13:07 12 November 2014
which groups the isomers with similar thermal and transport properties to further reduce
the mechanism size. In the present study, it was nevertheless found that only a small number
of isomers can be lumped in the 664-species skeletal mechanism when low temperature
chemistry is involved, in contrast to the large extent of reduction achieved in the previously
developed high temperature skeletal mechanism [51]. This observation implies that the
isomer concentrations may be less correlated at lower temperatures. As a result, only a
slightly smaller skeletal mechanism with 641 species and 2670 elementary reactions was
obtained after isomer lumping. As a last measure to obtain a substantially smaller skeletal
mechanism, DRGASA therefore faces a challenge aiming to further reduce the mechanism
by a factor of 5 in size, since only reduction extents by factors of 2∼3 had been achieved
in most previous works with DRGASA.
As discussed in references [18–19], the simulation error on ignition delay and extinction
residence time in PSR, which was induced by eliminating each individual species, was first
estimated in DRGASA. The estimated errors were then sorted in ascending order for
sequential sensitivity analysis for species elimination. The species with estimated errors
larger than, say, 0.5 are considered important. The elimination of such species may result
in difficulties in numerical convergence, and subsequently long computation time in the
global sensitivity analyses. Therefore, these species were usually excluded from sensitivity
analyses in previous studies with DRGASA. However, since the final size of the mechanism,
rather than the reduction cost, is the primary concern in the present work, almost every
species will be included in the global sensitivity analyses in the following reduction to
ensure that the resulting mechanism is minimal in size.
By specifying a worst-case relative error of 40% for ignition delay and extinction time
in PSR in DRGASA, a skeletal mechanism with 123 species and 394 elementary reactions
was eventually obtained. The reduction was performed in parallel with 200 CPU cores
and took approximately two weeks in wall-clock time. The present DRGASA is more
time-consuming than those performed previously because of the large number of species
tested.
gas-phase flow field is described using the Favre-Averaged Navier-Stokes equations in con-
junction with the Renormalisation Group (RNG) k-ε turbulence model, which includes
source terms for the effects of dispersed phase on gas-phase turbulence. These equations
are solved using a finite volume solver. The details of these models can be found in previous
publications [48,55,56], hence only a brief description is provided here.
Fuel injection is simulated using the blob injection model. Following the injection,
Kelvin Helmholtz (KH) and Rayleigh Taylor (RT) models are used to predict the primary
and secondary breakup of the computational parcels [57,58]. A breakup length is used
within which the KH model is used to predict the primary breakup. Beyond the breakup
length, the KH and RT models compete in breaking up the droplets. Droplet collisions
Auto-ignition PSR
Downloaded by [University of Leeds] at 13:07 12 November 2014
1
10 1600
10
0 Biodiesel - Air
1500
φ = 0.5
Ignition Delay Time, s
-1
10 Temperature, K
1400 p = 1 atm
10
-2 p = 1 atm
10 1300
-3
10 10
10
-4 1200
100 φ = 0.5
-5
10 1100
-6 100
10 1000
-7
10 -4 -3 -2 -1 0
0.4 0.6 0.8 1.0 1.2 1.4 10 10 10 10 10
0
10
2400
-1 100
10 2200
φ=1
Ignition Delay Time, s
-2 10
Temperature, K
10 p = 1 atm 2000
-3 10 1800 p = 1 atm
10
-4 1600
10
100 φ=1
-5
1400
10 Detailed
Skeletal 1200
-6
10
1000
-7
10 10
-6
10
-5
10
-4
10
-3 -2
10 10
-1
10
0
0.4 0.6 0.8 1.0 1.2 1.4
0
10
1800
-1
10 φ=2
1700
Ignition Delay Time, s
-2 1600
Temperature, K
10
p = 1 atm
1500
-3
10 10 1400 1
-4
10 1300 10
100 1200
10
-5
φ=2
1100 p = 100 atm
-6
10 1000
10
-7 900
-6 -5 -4 -3 -2 -1 0
0.4 0.6 0.8 1.0 1.2 1.4 10 10 10 10 10 10 10
1000/T, K
-1 Residence Time, s
Figure 1. Comparison of the 123-species skeletal mechanism with the detailed mechanism for
biodiesel-air, a) ignition delays, and b) extinction temperature profiles in PSR.
374 Z. Luo et al.
are based on the no time counter algorithm [59]. Once collision occurs, the outcomes of
the collision are predicted as bouncing, stretching, reflexively separating, or coalescing
[60]. A droplet evaporation model based on the Frossling correlation [54] is used in the
present simulations. A dynamic drag model is also used postulating that the drag coefficient
depends upon the shape of the droplet, which can vary between a sphere and a disk [61]. The
effects of turbulence on the droplet are also included, using a stochastic turbulent dispersion
model. Kinetic modelling in CONVERGE is performed using the SAGE chemical kinetic
solver [62] and is directly coupled with the gas-phase calculations using a well-stirred
reactor model. The soot mass production within a computation cell is determined from a
single-step competition between formation and oxidation rates of C2 H2 species based on
the Hiroyasu model [54], which has been extensively used in engine-modelling literature.
CONVERGE uses an innovative, modified cut-cell Cartesian method for grid genera-
tion. The grid is generated internally to the code at runtime. For all cases, the base grid size
Downloaded by [University of Leeds] at 13:07 12 November 2014
was fixed at 4 mm. In order to resolve the flow near the injector, a fixed grid embedding is
employed such that the minimum grid size is 0.25 mm. Apart from this region, it is rather
0
10
-1 Biodiesel-Air
10 CO
(a)
Species Mole Fraction
-2 n-C7H16
10
-3 MD
10
-4
10
-5
P = 1 atm
10 C2H2
φ=1
-6
T0 = 1000K
10 Lines: Detailed
Symbols: Skeletal
-7
10 -4 -3 -2 -1 0
10 10 10 10 10
Residence Time, s
2800
(b) Biodiesel-Air
2400
Temperature, K
p = 1 atm
φ=1
2000
T0 = 1800K
1600
1400K
1200
1000K
800 700 K
-12 -10 -8 -6 -4 -2 0 2
10 10 10 10 10 10 10 10
Residence Time, s
Figure 2. Comparison of (a) mole fractions for major species at T0 = 1000 K and (b) temperature
profiles at different initial temperatures, in a constant pressure auto-ignition process, calculated with
the detailed and the 123-species skeletal mechanisms, respectively.
Combustion Theory and Modelling 375
difficult to determine a priori where a refined grid is needed. Hence, four levels of adaptive
mesh refinement are employed for the velocity field. In order to match the combustion
chamber geometry used in the experimental study [63], a cubical geometry of 108 mm on
each side is generated. The temperature-dependent fuel properties of soy biodiesel, such
as density, kinematic viscosity, surface tension, vapour pressure, heat of vaporisation, and
specific heat, were obtained from Ra et al. [64]. It is noted that the effect of the distillation
curve for soy biodiesel was not accounted for in the present study. Nevertheless, the boiling
point for soy-diesel was measured [65] and used as an input for the simulations.
homogenous applications including auto-ignition and PSR. Figure 1a shows the calculated
ignition delay time of the mixture comprised of 25% MD, 25% MD9D, and 50% n-heptane
in mole as a function of the initial temperature at different pressures and equivalence ratios.
It is seen that the skeletal mechanism accurately mimics the detailed mechanism for most of
the displayed conditions, from the NTC region for engine ignition to the high temperature
region. Figure 1b further compares the temperature profiles in PSR for biodiesel-air at
various pressures and equivalence ratios. It is seen that the skeletal mechanism again
shows close agreements with the detailed mechanism for branches above the upper turning
points, which are widely regarded as the extinction states for steady state combustion.
Some moderately large discrepancies were observed in the lower branch of the temperature
profiles, which are nevertheless less relevant to stable combustion systems. Note that the
reaction states on the lower branches of PSR were not included in the sample space for the
reduction.
To further demonstrate the fidelity of the skeletal mechanism, species and temperature
profiles in auto-ignition are scrutinised in Figure 2. Figure 2a shows the concentrations
PSR
-1
10
Extinction Residence Time, s
φ = 0.5
-3
10
1.5
-4
10
1
-5
10
1 10 100
Pressure, atm
Figure 3. Extinction residence time of PSR as a function of pressure for biodiesel-air mixture at
various equivalence ratios, calculated with the detailed and the 123-species skeletal mechanisms,
respectively.
376 Z. Luo et al.
JSR
-2
10
O2
Species Mole Fraction
Lines: Detailed
n-C7H16 CO2 Symbols: Skeletal
-4
10
φ = 0.5
p = 1 atm
τ=1s
-6
10
-3
10
-4
10
φ=2
-5
10
-6
n-C7H16
10
-7
10
800 900 1000 1100 1200 1300
Temperature, K
Figure 4. Calculated species profiles in JSR as a function of temperature for the biodiesel oxidation
in diluted air (99% N2 in mole) under atmospheric pressure with fixed residence time of 1 s, calculated
with the detailed and the 123-species skeletal mechanisms, respectively, for a) equivalence ratio φ =
0.5, and b) φ = 2.0.
Combustion Theory and Modelling 377
concentrations can be accurately predicted by the skeletal mechanism considering that they
were not included in DRGASA.
that the simulation results from both mechanisms agree with the measured datasets fairly
JSR
-1
10
O2
Species Mole Fraction
-2
10
CO
-3
10 CO2
C2H4
-4
10
φ = 0.5
Lines: Simulations
Symbols: Experiments
-5
10
800 850 900 950 1000 1050
-2
CO
10
Species Mole Fraction
-3
10 CO2
O2
-4
10 φ=1 C2H4
DashLine: Skeletal
SolidLine: Detailed
-5
10
800 900 1000 1100 1200
Temperature, K
Figure 5. Species concentrations in JSR as a function of temperature for RME (diluted with N2 )
at pressure of 10 atm and residence time of 1 s. Symbols: experimental data with initial fuel mole
fraction of 0.05% [66], solid lines: values calculated with the detailed mechanism, with the same fuel
mole fraction of surrogate mixture (25% MD + 25% MD9D + 50% n-heptane) as the experiment,
dashline: values calculated with the 123-species skeletal mechanism.
378 Z. Luo et al.
-2
10
MD - O2 -Ar
-4
10
Skeletal
Detailed
Experiments
-5
10
0.74 0.76 0.78 0.80 0.82 0.84 0.86
Downloaded by [University of Leeds] at 13:07 12 November 2014
-1
1000/T, K
Figure 6. Comparison of simulated and measured ignition delay times as a function of the initial
temperature for MD-O2 -Ar mixture at pressure of 7 atm, equivalence ratio of 0.09 and fuel mole
fraction of 1005 ppm. The measurement data were obtained from reference [67]. The calculation was
performed with the detailed and the 123-species skeletal mechanisms, respectively.
well in trend under most of the displayed conditions. In some cases the skeletal mechanism
even shows a better prediction than the detailed mechanism, such as the O2 profile at the
stoichiometric condition. It is noted that some rather large discrepancies exist between some
simulation results and the measured values, such as the CO2 profile at fuel-lean conditions.
The 123-species skeletal mechanism and the detailed mechanism were then validated
against experimental measurements of ignition delay for MD oxidation. The experiment
was performed by Haylett et al. and the data were obtained from reference [67]. Figure
6 compares the calculated ignition delay time for the MD-oxygen-argon mixture with the
measurements at various initial temperatures under an extremely fuel-lean condition (φ =
70
Laminar Flame Speed, cm/s
MD-Air
60
50
p = 1 atm
40 T0 = 403 K
30 Simulations
Experiments
20
0.6 0.8 1.0 1.2 1.4 1.6
Equivalence Ratio
Figure 7. Laminar flame speed as a function of the equivalence ratio for MD-air mixture at atmo-
spheric pressure and initial temperature of 403 K. Symbols: experimental measurements [68], lines:
calculated values with the 123-species skeletal mechanism.
Combustion Theory and Modelling 379
60
50
Penetration, mm
Vapour
40
30
Liquid
20
10 Simulations
Experiments
0
0 400 800 1200
Downloaded by [University of Leeds] at 13:07 12 November 2014
Time, µ s
Figure 8. Measured [65] and predicted liquid spray and vapour penetrations vs. time for biodiesel
fuel.
0.09). It is seen in Figure 6 that the simulations with both mechanisms agree in trend with
the experiments, while the detailed mechanism gives overall shorter ignition delays.
The 123-species skeletal mechanism was next applied to 1-D laminar premixed flames.
While it is difficult to validate the skeletal mechanism with the extremely large detailed
mechanism in flames, comparison with experimental flame measurement can be readily
performed. Figure 7 shows the calculated laminar flame speeds in comparison with the
experimental measurements in reference [68] for MD-air mixture at atmospheric pressure.
The calculated results match tightly with the experiments for fuel rich cases, while under-
prediction by about 7 cm/s is observed at fuel-lean conditions.
Table 1. Test conditions for combustion experiments at Sandia National Laboratories [65].
Parameter Quantity
Figure 9. Validation of flame lift-off length against the OH-chemiluminescence data from reference
[65]. The average equivalence ratio at flame lift-off location is also indicated.
Figure 10. Comparison of OH mole-fraction contours from simulations against the OH-
chemiluminescence data from reference [65], at different instances during the combustion process.
Combustion Theory and Modelling 381
Downloaded by [University of Leeds] at 13:07 12 November 2014
Figure 11. Comparison of soot mole-fraction contours from simulations against the soot data using
LII from reference [65] at different instances during the combustion process.
the axial and radial directions respectively. It is seen that the lift-off length is over predicted
by about 25% under these conditions, but the average equivalence ratio at lift-off length
is fairly well captured in the simulations, with a calculated result of 1.92 compared to the
experimentally measured value of 2.08. The measured ignition delay is 0.396 ms, while
the simulated value is about 0.5 ms, again showing a 25% over-prediction. The width and
length of the flame is also well captured by the simulations. Since it has been observed
in Figure 8 that the model predicted the mixing process quite well, the over-prediction of
flame lift-off length and ignition delay in the present study could be primarily due to the
uncertainities in the detailed mechanism or the reduction error in the skeletal mechanism.
Figures 10 and 11 then present the validations of simulated OH and soot contours
against the chemiluminescence and Laser Induced Incandescence (LII) data, respectively,
at different times during the combustion event. It is seen in Figures 10 and 11 that the
experimental OH and soot contours are well predicted by the simulations in terms of
Downloaded by [University of Leeds] at 13:07 12 November 2014
location and occurrence during the combustion event. It is noted that although the lift-off
length in Figure 9 was over-predicted by 25% using the 123-species skeletal mechanism,
such a discrepancy is quite encouraging considering that the detailed and the skeletal
mechanisms were not tuned for the experimental conditions. Future studies will involve
further improvement of the detailed or reduced mechanisms to better predict ignition and
flame lift-off under CI engine conditions, by updating the rate constants of MD, MD9D or
n-heptane sub-mechanisms.
4. Concluding remarks
A small and comprehensive skeletal mechanism with 123-species for a biodiesel surrogate
mixture, consisting of MD, MD9D and n-heptane, was developed using the DRG-based
methods for biodiesel combustion under a wide range of conditions including low temper-
ature ignition in the NTC region. The large extent of reduction was achieved by stretching
the DRGASA method to include almost every species in the sensitivity analysis.
Validations of the skeletal mechanism were performed against the detailed mechanism
in 0-D applications, e.g. global parameters in auto-ignition and extinction in PSR, species
concentrations in auto-ignition and JSR. The skeletal mechanism was further validated
with the experimentally measured species concentrations for RME in JSR, ignition delays
of MD in aerosol shock tubes, and 1-D laminar premixed flame that is a diffusive system.
The skeletal mechanism was shown to be reasonably accurate, particularly considering the
significant reduction in mechanism size by a factor of 27.
Validation was also performed for 3-D turbulent spray combustion under CI engine
conditions in comparison with the experimental data from Sandia National Laboratories.
The mechanism is demonstrated to be very versatile and robust since it performed satis-
factorily in predicting the ignition delay and flame lift-off length, as well as the OH and
soot concentration profiles. It is expected that with further updates in the MD, MD9D or
n-heptane sub-mechanisms, the skeletal mechanism may show a better agreement with
experimental measurements, which merits further studies.
Acknowledgements
The work at University of Connecticut was supported by the National Science Foundation under Grant
0904771. Any opinions, findings, and conclusions or recommendations expressed in this material are
those of the authors and do not necessarily reflect the views of the National Science Foundation.
Combustion Theory and Modelling 383
The submitted manuscript has been created by UChicago Argonne, LLC, operator of Argonne
National Laboratory (Argonne). Argonne, a US Department of Energy Office of Science Laboratory,
is operated under Contract No. DE-AC02–06CH11357. The US Government retains for itself, and
others acting on its behalf, a paid-up, nonexclusive, irrevocable worldwide license in said article to
reproduce, prepare derivative works, distribute copies to the public, and perform publicly and display
publicly, by or on behalf of the Government.
The authors thank Dr. Lyle Pickett from Sandia National Laboratories for sharing the flame lift-off
data and many insightful discussions.
References
[1] A. Deepak, S. Sinha, and K.A. Agarwal, Experimental investigation of control of NOx emissions
in biodiesel-fueled compression ignition engine, Renew Energ. 31 (2006), pp. 2356–2369.
[2] O. Herbinet, W.J. Pitz, and C.K. Westbrook, Detailed chemical kinetic oxidation mechanism for
a biodiesel surrogate, Combust. Flame 154 (2008), pp. 507–528.
[3] O. Herbinet, W.J. Pitz, and C.K. Westbrook, Detailed chemical kinetic mechanism for the
Downloaded by [University of Leeds] at 13:07 12 November 2014
oxidation of biodiesel fuels blend surrogate, Combust. Flame 157 (2010), pp. 893–908.
[4] C.K. Westbrook, C.V. Naik, O. Herbinet, W.J. Pitz, M. Mehl, S.M. Sarathy, and H.J. Curran,
Detailed chemical kinetic reaction mechanisms for soy and rapeseed biodiesel fuel, Combust.
Flame 158 (2011), pp. 742–755.
[5] T.F. Lu and C.K. Law, Toward accommodating realistic fuel chemistry in large-scale computa-
tions, Prog. Energy Combust. Sci. 35 (2009), pp. 192–215.
[6] A.S. Tomlin, M.J. Pilling, T. Turanyi, J.H. Merkin, and J. Brindley, Mechanism reduction for
the oscillatory oxidation of hydrogen – sensitivity and quasi-steady-state analyses, Combust.
Flame 91 (1992), pp. 107–130.
[7] T. Turanyi, Sensitivity analysis of complex kinetic systems – tools and applications, J. Math.
Chem. 5 (1990), pp. 203–248.
[8] S. Vajda, P. Valko, and T. Turanyi, Principal component analysis of kinetic-models, Int. J. Chem.
Kinet. 17 (1985), pp. 55–81.
[9] S. Vajda and T. Turanyi, Principal component analysis for reducing the Edelson-Field-Noyes
model of the Belousov-Zhabotinsky reaction, J. Phys. Chem. 90 (1986), pp. 1664–1670.
[10] T. Turanyi, Reduction of large reaction-mechanisms, New J. Chem. 14 (1990), pp. 795–803.
[11] A. Massias, D. Diamantis, E. Mastorakos, and D.A. Goussis, An algorithm for the construction
of global reduced mechanisms with CSP data, Combust. Flame 117 (1999), pp. 685–708.
[12] M. Valorani, F. Creta, D.A. Goussis, J.C. Lee, and H.N. Najm, An automatic procedure for
the simplification of chemical kinetic mechanisms based on CSP, Combust. Flame 146 (2006),
pp. 29–51.
[13] H. Wang and M. Frenklach, Detailed reduction of reaction-mechanisms for flame modeling,
Combust. Flame 87 (1991), pp. 365–370.
[14] T.F. Lu and C.K. Law, A directed relation graph method for mechanism reduction, Proc.
Combust. Inst. 30 (2005), pp. 1333–1341.
[15] P. Pepiot-Desjardins and H. Pitsch, An efficient error-propagation-based reduction method for
large chemical kinetic mechanisms, Combust. Flame 154 (2008), pp. 67–81.
[16] W. Sun, Z. Chen, X. Gou, and Y. Ju, A path flux analysis method for the reduction of detailed
chemical kinetic mechanisms, Combust. Flame 157 (2010), pp. 1298–1307.
[17] T. Lu, P. Max, Z. Luo, S.M. Sarathy, W.J. Pitz, S. Som, and D.E. Longman, Directed relation
graph with expert knowledge for skeletal mechanism reduction, 7th US National Combustion
Meeting, Atlanta, GA, 2011.
[18] R. Sankaran, E.R. Hawkes, J.H. Chen, T. Lu, and C.K. Law, Structure of a spatially developing
turbulent lean methane-air Bunsen flame, Proc. Combust. Inst. 31 (2007), pp. 1291–1298.
[19] X.L. Zheng, T.F. Lu, and C.K. Law, Experimental counterflow ignition temperatures and reac-
tion mechanisms of 1,3-butadiene, Proc. Combust. Inst. 31 (2007), pp. 367–375.
[20] K. Niemeyer, C. Sung, and M. Raju, Skeletal mechanism generation for surrogate fuels using
directed relation graph with error propagation and sensitivity analysis, Combust. Flame 157
(2010), pp. 1760–1770.
[21] S.S. Ahmed, F. Mauss, G. Moreac, and T. Zeuch, A comprehensive and compact n-heptane
oxidation model derived using chemical lumping, Phys. Chem. Chem. Phys. 9 (2009), pp. 1107–
1126.
384 Z. Luo et al.
[22] G.Y. Li and H. Rabitz, A general-analysis of exact lumping in chemical-kinetics, Chem. Eng.
Sci. 44 (1989), pp. 1413–1430.
[23] G.Y. Li and H. Rabitz, A general-analysis of approximate lumping in chemical-kinetics, Chem.
Eng. Sci. 45 (1990), pp. 977–1002.
[24] A.S. Tomlin, G.Y. Li, H. Rabitz, and J. Toth, A general-analysis of approximate nonlinear
lumping in chemical-kinetics.2. Constrained lumping, J. Chem. Phys. 101 (1994), pp. 1188–
1201.
[25] E. Ranzi, M. Dente, A. Goldaniga, G. Bozzano, and T. Faravelli, Lumping procedures in detailed
kinetic modeling of gasification, pyrolysis, partial oxidation and combustion of hydrocarbon
mixtures, Prog. Energy Combust. Sci. 27 (2001), pp. 99–139.
[26] T.F. Lu and C.K. Law, Strategies for mechanism reduction for large hydrocarbons: n-heptane,
Combust. Flame 154 (2008), pp. 153–163.
[27] P. Pepiot-Desjardins and H. Pitsch, An automatic chemical lumping method for the reduction of
large chemical kinetic mechanisms, Combust. Theor. Model. 12 (2008), pp. 1089–1108.
[28] T.F. Lu and C.K. Law, Systematic approach to obtain analytic solutions of quasi steady state
species in reduced mechanisms, J. Phys. Chem. A 110 (2006), pp. 13202–13208.
Downloaded by [University of Leeds] at 13:07 12 November 2014
[29] T.F. Lu and C.K. Law, A criterion based on computational singular perturbation for the iden-
tification of quasi steady state species: A reduced mechanism for methane oxidation with NO
chemistry, Combust. Flame 154 (2008), pp. 761–774.
[30] C.J. Sung, C.K. Law, and J.Y. Chen, Augmented reduced mechanisms for NO emission in
methane oxidation, Combust. Flame 125 (2001), pp. 906–919.
[31] Z.Y. Ren and S.B. Pope, Entropy production and element conservation in the quasi-steady-state
approximation, Combust. Flame 137 (2004), pp. 251–254.
[32] N. Peters, Numerical and asymptotic analysis of systematically reduced reaction schemes for
hydrocarbon flames, Lect. Notes Phys. 241 (1985), pp. 90–109.
[33] M. Bodenstein, Eine theorie der photochemischen reaktionsgeschwindigkeiten, Z. Phys. Chem.
85 (1913), pp. 329–397.
[34] L.K. Underhill and D.L. Chapman, The interaction of chlorine and hydrogen, J. Chem. Soc.
Trans. 103 (1913), pp. 496–508.
[35] S.H. Lam, Using CSP to understand complex chemical-kinetics, Combust. Sci. Technol. 89
(1993), pp. 375–404.
[36] S.H. Lam and D.A. Goussis, Reduced chemistry modeling in reacting flows, Int. J. Chem. Kinet.
26 (1994), pp. 461–486.
[37] S.H. Lam, Reduced chemistry-diffusion coupling, Combust. Sci. Technol. 179 (2007), pp. 767–
786.
[38] U. Maas and S.B. Pope, Simplifying chemical-kinetics - intrinsic low-dimensional manifolds in
composition space, Combust. Flame 88 (1992), pp. 239–264.
[39] H. Bongers, J.A.V. Oijen, and L.P.H. De Goey, Intrinsic low-dimensional manifold method
extended with diffusion, Proc. Combust. Inst. 29 (2003), pp.1371–1378.
[40] Z.Y. Ren and S.B. Pope, The use of slow manifolds in reactive flows, Combust. Flame 147
(2006), pp. 243–261.
[41] Z.Y. Ren and S.B. Pope, Species reconstruction using pre-image curves, Proc. Combust. Inst.
30 (2005), pp. 1293–1300.
[42] Z.Y. Ren, S.B. Pope, A. Vladimirsky, and J.M. Guckenheimer, The ICE-PIC method for the
dimension reduction of chemical kinetics, J. Chem. Phys. 124 (2006), 114111.
[43] S.B. Pope, Computationally efficient implementation of combustion chemistry using in situ
adaptive tabulation, Combust. Theor. Model. 1 (1997), pp. 41–63.
[44] J. Lee, H. Najm, S. Lefantzi, J. Ray, M. Frenklach, M. Valorani, and D. Goussis, A CSP
and tabulation based adaptive chemistry model, Combust. Theor. Model. 11 (2007), pp. 73–
102.
[45] S. Tonse, N. Moriarty, N. Brown, and M. Frenklach, PRISM: Piecewise reusable implementation
of solution mapping. An economical strategy for chemical kinetics, Israel. J. Chem. 39 (1998),
pp. 97–106.
[46] J.L. Brakora, Y. Ra, R.D. Reitz, J. McFarlane, and C.S. Daw, Development and validation of a
reduced reaction mechanism for biodiesel-fueled engine simulation, SAE Int. J. Fuels Lubr. 1
(2008), pp. 675–702.
[47] T. Vaughn, M. Hammill, M. Harris, and A.J. Marchese, Ignition delay of bio-ester fuel droplets.
Powertrain & Fluid Systems Conference and Exhibition, Toronto, ON, Canada, 2006.
Combustion Theory and Modelling 385
[48] S. Som and D.E. Longman, Numerical study comparing the combustion and emission charac-
teristics of biodiesel to petrodiesel, Energ. Fuel. 25 (2011), pp. 1373–1386.
[49] S.M. Sarathy, M.J. Thomson, W.J. Pitz, and T. Lu, An experimental and kinetic modeling study
of methyl decanoate combustion, Proc. Combust. Inst. 33 (2011), pp. 399–405.
[50] K. Seshadri, T.F. Lu, O. Herbinet, S.B. Humer, U. Niemann, W.J. Pitz, R. Seiser, and C.K.
Law, Experimental and kinetic modeling study of extinction and ignition of methyldecanoate in
laminar non-premixed flows, Proc. Combust. Inst. 32 (2009), pp. 1067–1074.
[51] Z. Luo, T. Lu, M.J. Maciaszek, S. Som, and D.E. Longman, A reduced mechanism for high
temperature oxidation of biodiesel surrogates, Energ. Fuel. 24 (2010), pp. 6283–6293.
[52] K.J. Richards, P.K. Senecal, and E. Pomraning, CONVERGETM (Version 1.2) manual, Conver-
gent Science Inc., Middleton, WI, 2008.
[53] P.K. Senecal, K.J. Richards, E. Pomraning, T. Yang, M.Z. Dai, R.M. McDavid, M.A. Patterson,
S. Hou, and T. Shethaji, A new parallel cut-cell cartesian CFD code for rapid grid generation
applied to in-cylinder diesel engine simulations. SAE World Congress & Exhibition, Detroit,
MI, 2007.
[54] S. Som, Development and validation of spray models for investigating diesel engine combustion
Downloaded by [University of Leeds] at 13:07 12 November 2014