Article

pubs.acs.org/EF

Development and Validation of a Reduced Chemical Kinetic

Mechanism for Computational Fluid Dynamics Simulations of
Natural Gas/Diesel Dual-Fuel Engines
Andrew Hockett,† Greg Hampson,‡ and Anthony J. Marchese*,†
†
Department of Mechanical Engineering, Colorado State University, Fort Collins, Colorado 80523, United States
‡
Woodward Inc., Loveland, Colorado 80538, United States
*
S Supporting Information

ABSTRACT: A reduced chemical kinetic mechanism consisting of 141 species and 709 reactions has been constructed to
simulate the combustion of both natural gas and diesel fuels in a dual-fuel engine. Natural gas is modeled as a mixture of
methane, ethane, and propane, while the diesel fuel is modeled as n-heptane. The new reduced mechanism combines reduced
versions of a detailed n-heptane mechanism and a detailed methane through n-pentane mechanism, each of which was reduced
using a direct relation graph method. The reduced dual-fuel mechanism is validated against ignition delay computations with full
detailed mechanisms, adiabatic homogeneous charge compression ignition simulations with full detailed mechanisms,
experimental premixed laminar flame speeds of CH4/O2/He mixtures at 40 and 60 atm, ignition delay and lift-off length from a
diesel spray experiment in a constant-volume chamber, and finally against dual-fuel engine experiments using multidimensional
computational fluid dynamics simulations. The engine simulations were performed for direct comparison against natural gas/
diesel dual-fuel engine experiments at varying injection timings, engine loads, substitution percentages, and natural gas
compositions. An engine experiment with additional propane was used to induce engine knock for the purpose of validating the
reduced mechanism’s ability to predict natural gas autoignition. The results show that the newly reduced mechanism accurately
reproduces the chemical kinetic behavior of the detailed mechanism, including the laminar flame speed at high pressure, the
ignition delay and lift-off length in the diesel spray experiment, and the pressure and heat release rate in the engine experiments.
In the experiment where engine knock occurred, the model predicts the phasing and magnitude of a sudden acceleration in the
combustion rate and reproduces the observed high-frequency pressure oscillations.

1. INTRODUCTION path toward meeting new GHG emissions regulations with

The incentive to power internal combustion engines with
natural gas continues to increase because of both the economic
and emissions advantages of natural gas in comparison with
traditional gasoline and diesel fuels. Natural gas is expected to
remain less expensive than petroleum fuels on an energy-
equivalent basis through 2040.1 For many engine applications,
the fuel cost savings is sufficient to offset the initial capital
investment of converting equipment to operate on natural gas.
In addition, lean premixed combustion of natural gas avoids the
rich regions within diesel sprays where soot is formed.
Furthermore, the U.S. Environmental Protection Agency
(EPA) has initiated regulations on greenhouse gas (GHG)
emissions and fuel efficiency standards for light-duty vehicles by
2018 and heavy-duty vehicles by 2025.2,3 Because natural gas is
composed primarily of methane, natural gas can offer an
approximate 25% reduction in carbon dioxide emissions
compared with diesel fuel due to the higher hydrogen to
carbon ratio and higher heating value.4 However, the reduction
in total GHG emissions can be compromised if the natural gas
does not burn to completion and is exhausted as unburned
hydrocarbon (UHC) emissions, because methane has a global
warming potential that is 25 times greater than that of carbon
dioxide over a 100 year period.5 Therefore, if UHC emissions
can be minimized, then natural gas engines can offer a possible
significant fuel cost savings.
Diesel engines can be readily converted to natural gas/diesel
dual-fuel engines by fumigating the intake manifold with natural
gas and using a small diesel injection as an ignition source.
Ignition of the diesel fuel subsequently initiates combustion of
the premixed natural gas/air mixture near the diesel flame. If
the natural gas to air ratio is within the flammability limits, then
a propagating premixed flame will consume the remaining
natural gas in the cylinder.
The amount of natural gas that is substituted for diesel is
known as the substitution percentage, z, and is defined on an
energy basis using the lower heating values of the fuels.
Depending on the speed and load, the substitution percentage
can be limited by a variety of constraints, including excessive
pressure rise rate, engine knock, preignition, injector tip
temperature, nitric oxide (NOX) emissions, and UHC
emissions due to incomplete combustion of natural gas.6 A
greater understanding of the in-cylinder physics can lead to
novel engine technologies that can expand the natural gas
substitution limits and minimize incomplete combustion of
natural gas. To this end, multidimensional computational
Received: November 10, 2015
Revised: January 23, 2016
Published: January 26, 2016

© 2016 American Chemical Society 2414 DOI: 10.1021/acs.energyfuels.5b02655

Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
Table 1. Conditions Used To Reduce the Base Mechanisms Using DRGEPSA
detailed base mechanism LLNL n-heptane15,16
target species n-C7H16, O2, OH, HO2, CH4, C2H6, C3H8,
He, Ar, N2
fuel composition n-C7H16 = 1.0
(mole fraction)
temperature (K) 700, 750, 800, 850, 900, 950, 1000, 1050,
1100, 1150, 1200
pressure (bar) 10, 30, 80
equivalence ratio, φ 0.2, 0.5, 1.0, 1.5, 2.0
ignition delay error (%) 0.1
sensitivity analysis fraction 0.75
modeling is often employed to gain valuable information about
in-cylinder physics and aid in the development of new engine
designs and combustion strategies.
Prior to the present study, a sufficiently accurate and
compact chemical kinetic mechanism to enable accurate
simulation of natural gas/diesel dual-fuel combustion was not
available. In particular, no reduced dual-fuel mechanism was
available that accounted for the variation in chemical reactivity
of natural gas due to variation in the chemical composition.
Niemann et al.7 used pure methane as a natural gas surrogate in
a multidimensional model for natural gas/diesel reactivity-
controlled compression ignition (RCCI) under the assumption
that a natural gas mixture with high methane content should
behave similarly to pure methane. However, Bourque et al.8
found that the ignition delay periods and flame speeds for
typical pipeline natural gas compositions with small concen-
trations of heavier hydrocarbons differ substantially from those
for pure methane. Rahimi et al.9 presented a 76-species reduced
chemical kinetic mechanism for natural gas/diesel homoge-
neous charge compression ignition (HCCI) engines. This
mechanism was developed by combining a reduced n-heptane
mechanism with the GRI-Mech 3.0 natural gas mechanism10
and then using a genetic optimization algorithm to adjust the
rate constants for three different engine operating conditions
consisting of different natural gas substitution percentage, total
equivalence ratio, exhaust gas recirculation (EGR), and
compression ratio. However, only one natural gas composition
was validated in the study by Rahimi et al., and therefore, a rate
constant optimization and validation study would likely be
necessary before modeling of a different natural gas
composition. Furthermore, the mechanism of Rahimi et al.
was validated for HCCI simulations using a five-zone model
and not for sufficiently resolved multidimensional computa-
tional fluid dynamics (CFD) simulations of premixed
propagating natural gas flames or non-premixed diesel spray
flames, which are present in conventional dual-fuel engines.
The objective of the present study was to develop a reduced
chemical kinetic mechanism from highly detailed base
mechanisms for use in multidimensional CFD simulations of
natural gas/diesel dual-fuel engines where premixed propagat-
ing natural gas flames and non-premixed diesel spray flames are
present. The reduced dual-fuel mechanism was validated
against ignition delay computations with full detailed
mechanisms, adiabatic zero-dimensional HCCI simulations
with full detailed mechanisms, experimental premixed laminar
flame speeds of CH4/O2/He mixtures at 40 and 60 atm,
ignition delay and lift-off length from a diesel spray experiment
in a constant-volume chamber, and finally against dual-fuel
engine experiments using multidimensional CFD simulations
performed using the commercial CONVERGE code. The
Article
NUIG natural gas14 NUIG natural gas14
CH4, C2H6, C3H8, O2, OH, HO2, CO2, CH4, C2H6, C3H8, O2, OH, HO2, CO2,
H2O, Ar, He, N2 H2O, Ar, He, N2
CH4 = 0.85, C2H6 = 0.1, C3H8 = 0.05 CH4 = 0.93, C2H6 = 0.02, C3H8 = 0.05
750, 800, 850, 900, 950, 1000, 1050, 1100, 750, 800, 850, 900, 950, 1000, 1050, 1100,
1150, 1200 1150, 1200
10, 30, 80 10, 30, 80
0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4 0.2, 0.4, 0.6, 0.8, 1.0, 1.2, 1.4
0.5 0.5
0.75 0.75
newly developed mechanism did not require any rate constant
tuning and successfully reproduced in-cylinder pressure
measurements from dual-fuel engine experiments conducted
under multiple operating conditions, including varying
substitution percentage, injection timing, engine load, two
different natural gas compositions, and a case with additional
propane to induce engine knock. Moreover, the mechanism is
sufficiently compact that three-dimensional Reynolds-averaged
Navier−Stokes (RANS)-based engine simulations can be
conducted within a reasonable computational time (less than
24 h).
2. CHEMICAL MECHANISM REDUCTION METHOD
The new reduced dual-fuel chemical mechanism was constructed from
previously published detailed mechanisms. Diesel fuel was modeled
using n-heptane because of their similar cetane numbers and heating
values.11 Natural gas was modeled using mixtures of methane, ethane,
and propane as surrogates. While pipeline-quality natural gas is
predominantly methane, its reactivity is greatly influenced by the
presence of ethane and propane, which are typically present at low
concentrations (<10% by volume combined) but have much higher
reactivities than methane. Larger hydrocarbons such as butane and
pentane were not taken into account in the reduced mechanism
because the assumption was made that their concentrations in the
pipeline natural gas mixture being modeled would be low enough not
to significantly impact the combustion behavior.12,13
A reduced methane/ethane/propane mechanism was formulated
from the detailed methane through n-pentane mechanism (293
species, 1588 reactions) of Healy et al.14 at the National University of
Ireland Galway, hereafter in this paper denoted as the NUIG
mechanism. This detailed natural gas mechanism had previously
been validated against experimental ignition delay data from rapid
compression machine and shock tube experiments at temperatures
between 630 and 1550 K and pressures up to 30 bar under lean,
stoichiometric, and rich conditions. A reduced n-heptane mechanism
was formulated from the detailed n-heptane mechanism (654 species,
2827 reactions) of Curran et al.15,16 at Lawrence Livermore National
Laboratory (LLNL), hereafter denoted as the LLNL mechanism. The
detailed LLNL mechanism had previously been validated against
experimental ignition delay data from rapid compression machine and
shock tube experiments at temperatures between 550 and 1700 K,
pressures between 1 and 42 atm, and equivalence ratios from 0.3 to
1.5.
Both the NUIG and LLNL detailed mechanisms were developed by
many of the same authors and therefore employ similar values for most
of the reaction rate constants. Also, the LLNL n-heptane mechanism
includes many of the natural gas reaction pathways present in the
NUIG mechanism because higher hydrocarbons break down into
smaller hydrocarbon species that are important in natural gas
chemistry. However, not all of the reactions and rate constants for
smaller hydrocarbon species are identical in the NUIG natural gas
mechanism and the LLNL n-heptane mechanism. Without a rigorous
validation of the LLNL n-heptane mechanism against experimental
data for natural gas mixtures, the choice was made to reduce the two
2415 DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Figure 1. Procedure used to reduce the detailed mechanisms and combine the reduced mechanisms.

detailed mechanisms separately and then combine the reduced 3. ENGINE EXPERIMENTAL PROCEDURE
mechanisms. Natural gas/diesel dual-fuel engine experiments were used to validate
The detailed base mechanisms were reduced using the direct
the CSU141 mechanism. Engine experiments were conducted using a
relation graph with error propagation and sensitivity analysis four-cylinder light-duty General Motors (GM) 1.9 L, common rail,
(DRGEPSA) method,17 which is an automatic mechanism reduction turbocharged diesel engine. The specifications of the engine are listed
algorithm included within the CONVERGE software package. The in Table 2, and the diesel injector specifications are listed in Table 3.
NUIG mechanism was reduced on the basis of ignition delay
calculations on two different natural gas mixtures in air as described in
Table 1. The target species included helium and argon to retain the Table 2. Engine Specifications
third-body efficiency information because these inert species are often
cylinder arrangement four inline
used in laminar flame experiments at high pressure. The 85% methane
displacement volume 1.9 L
composition was chosen to be similar to the composition measured in
the engine experiments. This reduction resulted in a 49-species skeletal bore 82 mm
mechanism for methane, ethane, and propane mixtures. A preliminary stroke 90.4 mm
study of this reduced mechanism found that the ignition delay for pure connecting rod length 161 mm
propane in air did not agree well with the detailed NUIG mechanism. compression ratio 17.5
Therefore, another reduction was performed for a composition with injection system Bosch common rail
higher propane content than ethane content, which resulted in a 46- valves per cylinder four DOHC
species skeletal natural gas mechanism. Upon comparison of the 46- rated power 110 kW (147.5 hp) @ 4000 rpm
and 49-species mechanisms, there were four species that were not in rated torque 315 N m (232 ft-lbf) @ 2000 rpm
the initial 49-species mechanism: C3H6OOH1-3, C3H6OOH1-3O2,
IC3H7O2, and C3KET13. These additional species are important in
the low-temperature branching pathways for propane. Consequently, Table 3. Diesel Injector Specifications
the 46-species skeletal mechanism yielded much better agreement with name Bosch CRIP 2-MI
the detailed mechanism for ignition delay of pure propane. Ignition
no. of nozzles 7
delay plots for pure methane, ethane, and propane in air are given in
the Supporting Information. Next, the DRGEPSA method was used to nozzle orifice diameter 141 μm
reduce the LLNL n-heptane mechanism to a skeletal mechanism included spray angle 148°
containing 131 species and 656 reactions. The three reduced discharge coefficient 0.86
mechanisms were then combined to form the final reduced chemical
kinetic mechanism for mixtures of methane, ethane, propane, and n- Engine experiments were conducted with the two different natural gas
heptane containing 141 species and 709 reactions, which is hereafter compositions listed in Table 4, which are called the low-reactivity
denoted as the CSU141 mechanism.
An illustration of the reduction procedure is given in Figure 1. Any Table 4. Measured Natural Gas Compositions at Two
duplicate reactions in the three reduced mechanisms were given the Different Facilities
rate constant values from the reduced natural gas mechanisms to
ensure that engine knocking from natural gas autoignition is accurately species low-reactivity natural gas high-reactivity natural gas
represented. Care was taken to find and add reactions from the CH4 0.9461 0.8812
detailed base mechanisms that were not in any of the individual
C2H6 0.0285 0.0823
reduced mechanisms but whose species were all present after the
C3H8 0.0031 0.0075
reduced mechanisms were combined. For laminar flame speed
calculations, the natural gas transport data from Bourque et al.8 were CO2 0.0125 0.0167
combined with the transport data from the LLNL n-heptane N2 0.0093 0.0123
mechanism, and any duplicate species were given the transport data
from the natural gas data file. The CSU141 mechanism files (species,
reactions, thermodynamic data, and transport data, in CHEMKIN natural gas and the high-reactivity natural gas. The high-reactivity
format) are included in the Supporting Information. natural gas mixture had noticeably higher levels of ethane and propane.
Ignition delay and zero-dimensional HCCI engine calculations were It should be noted that both of the natural gas compositions were
performed using CHEMKIN. The ignition delay period was defined by municipal natural gas, which underscores the wide range of natural gas
the maximum temperature derivative corresponding to the largest reactivities of “pipeline-quality” natural gas.
temperature increase. In this paper, the CSU141 mechanism is not For the experiments described herein, the engine was modified to
compared against experimental ignition delay data because a heat loss allow multiport injection of natural gas. This modification was
model would need to be added in the calculations and this would accomplished by building a natural gas accumulator with attached
introduce additional error, making it difficult to assess the accuracy of Woodward natural gas injectors and connecting tubing from each
the reduction method. The reader should refer to the detailed injector to the intake manifold runners, such that the gas was injected
mechanisms for comparisons with experimental ignition delay periods perpendicular to the air flow direction. In a similar fashion, a propane
and then consider the percent error between the detailed mechanism injector was connected to cylinder 2 using a T connector in the tube
and the reduced mechanism. line for natural gas. The propane injection system was used to provide

2416 DOI: 10.1021/acs.energyfuels.5b02655

Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Table 5. Experimental Operating Conditions

fuel case name D25G17 D12G22 D11G8 D11G18 D12G22P2
engine speed (rpm) 2200 2200 2000 2000 2200
injected diesel per cylinder (mg) 25.4 12.1 ± 3 10.6 10.6 11.7 ± 3
natural gas per cylinder (mg) 16.5 21.8 8.0 17.9 22.1
propane per cylinder (mg) 0 0 0 0 2.3
natural gas substitution, z (%) 43.3 67.9 47.6 67.1 71.9
natural gas composition lower-reactivity lower-reactivity higher-reactivity higher-reactivity lower-reactivity
diesel injection pressure (bar) 700 700 700 700 700
injection duration (crank angle deg) 13.4 6.2 5.9 5.9 5.3
electronic SOI ( deg ATDC) −8.8 −13 −8 −14 −12 −12 −10
boost pressure (kPa abs) 169.3 167.2 168.2 163.6 120 120 163
manifold temperature (°C) 43.6 43.5 44.6 44.4 39.6 38.6 48.2
net IMEP (bar) 15.0 15.6 10.9 10.7 7.0 11.7 11.7
natural gas equivalence ratio 0.357 0.362 0.479 0.512 0.25 0.55 0.600
total equivalence ratio 0.845 0.856 0.712 0.765 0.55 0.74 0.824

a means of varying the natural gas reactivity to examine the effect of
natural gas reactivity on engine knock.
The engine and injectors were controlled by a Woodward engine
control module (ECM), which enabled individual cylinder control of
the diesel and natural gas injection timing and duration. The diesel fuel
consumption was measured using Micro Motion CMF025 Coriolis
mass flow meters on the supply and return of the fuel tank, which have
an accuracy of ±0.1% of the measured rate. The natural gas
consumption was measured using a Micro Motion CMF010 Coriolis
mass flow meter, which has an accuracy of ±0.35% of the measured
rate. The air mass flow rate was measured with an automotive hot wire
anemometer, which has an accuracy of ±3%. In-cylinder pressure
transducers (Kistler 6058A) were installed in the glow plug hole of
each cylinder, and pressure data were acquired on a National
Instruments data acquisition system. Crank angle data were acquired
with a resolution of 0.25 crank angle degree. Apparent heat release
rates (AHRRs) were calculated on the basis of 100-cycle-averaged
pressure data assuming a constant ratio of specific heats of 1.3. A five-
point moving average was applied to the AHRR profiles.
In order to accurately model the breakup and vaporization of diesel
sprays, it is important to obtain a profile of the injected mass flow rate
throughout the duration of the injection, which is specific to the
injector being modeled. As described by Hockett,18 a rate of injection
meter was used to measure the rate shape, the injection duration, and
the delay period between the applied electric current and the start of
injection. These rate of injection data were gathered for each of the
four injectors used in the engine experiments. These rate of injection
experiments also provided data on the injected mass as a function of
electronic pulse width and injection pressure, which were used to set
the injected mass in the engine simulations.
The experimental operating conditions are listed in Table 5. The
nomenclature for referring to the different operating conditions in
Table 5 uses a “D” to designate diesel followed by the mass of diesel
injected per cylinder per cycle in units of milligrams rounded to the
nearest integer. In a similar manner, “G” signifies the mass of natural
gas in milligrams, and “P” signifies the mass of propane in milligrams.
Start of injection (SOI) timing sweeps were performed for two
different fuel combinations at 2200 rpm and the highest boost pressure
attainable at this engine speed. The D25G17 SOI sweep had a
moderate substitution rate of 43% natural gas by energy, and the
D12G22 sweep had a higher substitution rate of 68%. Additionally, a
load sweep at 2000 rpm and a lower boost pressure was performed by
increasing the mass of natural gas while keeping the injected diesel
mass constant. The load sweep used the higher-reactivity natural gas
composition while the SOI sweeps used the lower-reactivity natural
gas composition, as described in Table 4. Additionally, in the
D12G22P2 experiment, an additional 2 mg of propane was added to a
case from the D12G22 SOI sweep to induce engine knocking and
assess whether the model could predict natural gas autoignition.
Therefore, these experiments encompass a wide range of substitution
percentages, engine loads, natural gas equivalence ratios, natural gas
reactivities, and top dead center (TDC) conditions for validation of
the reduced mechanism.
4. COMPUTATIONAL METHOD FOR
MULTIDIMENSIONAL ENGINE SIMULATIONS
Computational modeling of the in-cylinder physics was performed
using the CONVERGE code.19 All of the combustion simulations
presented herein began at intake valve closure (IVC) and used a one-
seventh sector of the cylinder geometry wherein each sector included
one of the injector’s seven nozzles, as depicted in Figure 2a. To
Figure 2. (a) One-seventh sector geometry used for engine
combustion simulations. (b) Cylinder and port geometry from the
GM 1.9 L engine used to obtain the velocity and turbulence fields at
IVC for use in the sector simulations. The cylinder liner is hidden to
show the piston geometry.
initialize the velocity and turbulence fields at IVC, a preliminary
simulation of the intake stroke was performed utilizing the geometry of
the full cylinder with valves and intake and exhaust ports, as depicted
in Figure 2b. The results of this intake stroke simulation were then
used to map the velocity and turbulence fields in the sector
simulations. A single-cylinder GT-POWER model was used to find
the trapped mass at IVC, the exhaust residual mass, the IVC pressure,
and the IVC bulk temperature. The sector simulations assumed
homogeneous composition, temperature, and pressure of the gas at
IVC. While the trapped mass was kept fixed, the IVC pressure and
temperature used in the sector simulations were slightly adjusted (±5
kPa) from the values provided by GT-POWER in order to obtain
closer agreement with the experimental pressure during compression.
In some cases, small adjustments to the piston and head wall
temperatures (±20 K) were necessary to match the TDC pressure.
CONVERGE uses a fixed and uniform Cartesian grid and a cut cell
method at the boundaries. The simulations used adaptive mesh
refinement (AMR) based on temperature, velocity, and n-heptane

2417 DOI: 10.1021/acs.energyfuels.5b02655

Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
Figure 3. Comparison of ignition delay plots for n-heptane/air mixtures obtained using the CSU141 reduced mechanism and the detailed LLNL
mechanism15,16 for (a) very lean (φ = 0.2), (b) lean (φ = 0.5), (c) stoichiometric (φ = 1.0), and (d) rich (φ = 2.0) mixtures.
mass fraction with a minimum cell size of 0.25 mm, as recommended
by Senecal et al.11 Fixed embedding of 0.25 mm cells was used in the
near-nozzle region. The base cell size had a cell width of 4 mm. The
transport equations were solved using a cell-centered finite-volume
method with second-order central differencing for the surface fluxes.
The gas-phase transport properties were based on the viscosity and
thermal conductivity for air. The combustion chemistry was modeled
using the CSU141 mechanism presented herein. To reduce the
computational cost, a multizone chemistry model was employed in
which the chemistries for cells similar in temperature, progress
equivalence ratio, and mass fraction of n-heptane were solved
simultaneously. The n-heptane criterion is necessary in dual-fuel
simulations to prevent cells with similar equivalence ratio but different
fuel types from being grouped together. In the simulation of a case in
which engine knock occurred, the hydroperoxyl radical (HO2) was
added as a dimension in the multizone chemistry strategy.
The assumption of a well-stirred reactor in each cell was used. No
subgrid model for turbulence-chemistry interactions (TCI) was used
to correct for commutation error between the average cell temperature
and average cell reaction rate. However, Pomraning et al.20 found that
the well-stirred reactor assumption with a RANS turbulence model
and grid refinement using AMR can successfully achieve grid
convergence for premixed and non-premixed combustion at a cell
width of 0.25 mm. They found that the converged solution agreed well
with experimental measurements of premixed and non-premixed
flames. Pomraning et al. concluded that in a sufficiently resolved
RANS simulation of typical flow conditions in an engine (i.e.,
Reynolds number), the magnitude of the TCI commutation error from
the well-stirred-reactor assumption must not be significant in
comparison with other sources of uncertainty if the goal is to capture
global pressure and heat release rates. The RANS turbulence model
used was the renormalization group (RNG) κ−ε model.21 The
turbulent heat transfer model from Amsden22 was used for the wall
boundaries.
2418
Article
The spray model utilized in this study was of the Lagrangian drop−
Eulerian fluid type. The liquid phase was modeled as a single
component with the thermodynamic properties of diesel fuel obtained
from the CONVERGE liquids property library (denoted as
“DIESEL2”),19 which are the recommended properties from
Convergent Science, Inc. The breakup of the spray parcels was
modeled using the modified KH−RT droplet breakup mechanism,
which does not use a defined breakup length, as described in the
CONVERGE theory manual.19 A more comprehensive description of
the spray model is provided in the Supporting Information.
The spray model was validated against liquid and vapor penetration
measurements from experiments done by the Engine Combustion
Network (ECN),23 which consisted of a vaporizing but nonreacting
spray injected into a constant-volume chamber under typical TDC
temperature and pressure conditions. Specific details about the ECN
data sets used and comparisons of the simulated and experimental
liquid and vapor penetration can be found in the Supporting
Information.
5. RESULTS AND DISCUSSION
5.1. n-Heptane Ignition Delay Calculations. Figure 3
shows comparisons of adiabatic zero-dimensional ignition delay
periods obtained using the detailed LLNL n-heptane
mechanism and the CSU141 reduced dual-fuel mechanism.
Comparisons are made for n-heptane/air mixtures at pressures
of 10, 30, and 80 bar, temperatures between 700 and 1300 K,
and equivalence ratios between φ = 0.2 (very lean) and φ = 2.0
(rich). These temperatures and pressures are relevant to diesel
sprays under engine TDC conditions. The reduced CSU141
mechanism shows excellent agreement with the detailed base
mechanism across all of the conditions presented in Figure 3
and in particular captures the negative temperature coefficient
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Figure 4. Comparisons of natural gas/air ignition delay calculations using the reduced CSU141 mechanism and the detailed NUIG mechanism.14
Results for the 90/6.7/3.3 mixture are shown in the left column for (a) very lean (φ = 0.2), (b) lean (φ = 0.5), and (c) stoichiometric (φ = 1.0)
mixtures. Results for the 70/20/10 mixture are shown in the right column for (d) very lean (φ = 0.2), (e) lean (φ = 0.5), and (f) stoichiometric (φ =
1.0) mixtures.

(NTC) region resulting from the two-stage ignition process for
n-heptane. Figure 3d shows excellent agreement for rich
mixtures between 700 and 1000 K, which are conditions
relevant to diesel spray ignition. The maximum deviation for
the φ = 2.0, 80 bar results occurs at 1100 K, where the reduced
mechanism predicts a 0.02 ms longer ignition delay period,
which translates to 0.25 crank angle degree at 2000 rpm. It
should be noted that the ignition delay of diesel jets in dual-fuel
engines is not controlled solely by chemical kinetics and that
turbulent mixing rates are also important.
In addition to conventional dual-fuel combustion, early-
injection schemes such as RCCI require accurate modeling of
the chemical kinetics of diesel fuel over a wider range of
equivalence ratios. Fuel/air mixtures in RCCI engines can be
anywhere between slightly rich to very lean and begin at
temperatures lower than peak compression temperatures.
Figure 3a−c shows excellent agreement for CSU141 at lower
temperatures for very lean to stoichiometric conditions.
2419
5.2. Natural Gas Ignition Delay Calculations. Compar-
isons of ignition delays obtained using the detailed NUIG
mechanism and the reduced CSU141 mechanism for two
different natural gas/air mixtures consisting of methane, ethane,
propane, and air are shown in Figure 4. The 90/6.7/3.3 mixture
consisted of 90 mol % methane, 6.7 mol % ethane, and 3.3 mol
% propane. The 70/20/10 mixture consisted of 70 mol %
methane, 20 mol % ethane, and 10 mol% propane. The ignition
delay plots show that the reduced CSU141 mechanism captures
with excellent accuracy the behavior of the detailed mechanism
for both mixtures across the range of pressures, temperatures,
and equivalence ratios. As discussed earlier, experimental data
are not included in these plots because of the need to model
the heat transfer in ignition delay experiments from a rapid
compression machine. The error in ignition delay for the
reduced mechanism relative to the detailed natural gas
mechanism is less than 0.1% for all of the conditions presented,
and therefore, it should be expected that the reduced
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Figure 5. Adiabatic zero-dimensional HCCI engine simulation comparisons between the detailed LLNL mechanism15,16 and the reduced CSU141
mechanism for varying initial temperatures of n-heptane/air mixtures at (a) φ = 0.5 and (b) φ = 2.0.

Figure 6. Zero-dimensional adiabatic HCCI engine simulation comparisons between the detailed NUIG mechanism14 and the reduced CSU141
mechanism for varying initial temperatures at φ = 0.5 for (a) the 90/6.7/3.3 natural gas mixture in air and (b) the 70/20/10 natural gas mixture in
air.

mechanism can perform as well as the detailed mechanism in
comparisons with experimental data. While the NUIG
mechanism was only validated with experimental data up to
30 atm,24 the 80 bar calculations were of interest because
typical turbocharged light-duty diesel engines can have TDC
compression pressures near 80 bar. Therefore, under the
assumption that the NUIG mechanism is accurate at diesel
TDC pressures, the CSU141 mechanism should be able to
predict autoignition in natural gas/diesel dual-fuel engine
simulations.
5.3. Adiabatic HCCI Simulations. The ignition delay
calculations presented thus far were conducted at constant
volume and fixed initial temperature/pressure. The perform-
ance of the mechanism was further validated for the effects of
varying temperature and pressure from the compression and
expansion strokes in an engine by performing zero-dimensional
adiabatic HCCI engine simulations using CHEMKIN. The
engine geometry for these calculations was identical to that of
the experimental engine described previously. The engine
speed was chosen to be 2000 rpm, which is close to the speed
used in the engine experiments. Figure 5 shows computational
2420
comparisons between the detailed LLNL n-heptane mechanism
and the CSU141 mechanism for n-heptane/air mixtures at φ =
0.5 and φ = 2.0. For these computations, the initial pressure
was fixed at 1.7 bar and the initial temperature was varied. The
CSU141 and LLNL mechanisms exhibited excellent agreement
in terms of start of combustion timing for the low-temperature
heat release, high-temperature ignition, and pressure rise. The
agreement tends to diverge slightly as the start of combustion
retards past TDC. However, the results in Figure 5 strongly
suggest that the CSU141 mechanism can predict start of
combustion behavior of diesel sprays and autoignition in HCCI
or RCCI engines to the same level of accuracy as the detailed
LLNL n-heptane mechanism.
Figure 6a shows calculated HCCI pressure curves for the 90/
6.7/3.3 natural gas mixture in air at φ = 0.5 for both the
detailed NUIG mechanism and the reduced CSU141
mechanism for various initial temperatures at a fixed initial
pressure of 1.7 bar. Figure 6b shows a similar comparison for
the 70/20/10 natural gas mixture. It can be seen that the
reduced mechanism captures the ignition behavior of the
detailed mechanism for these natural gas mixtures. At lower
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
initial temperature, when the start of combustion begins to
retard past TDC, the reduced mechanism begins to predict
earlier start of combustion than the detailed mechanism.
However, the results in Figure 6 strongly suggest that the
CSU141 mechanism can predict the end gas autoignition of
natural gas to the same level of accuracy as the detailed NUIG
mechanism.
5.4. Laminar Flame Speed Calculations of Methane at
High Pressures. High-pressure methane laminar flame speed
measurements using spherically propagating flames in a
constant-volume chamber were conducted by Rozenchan et
al.25 at 40 and 60 atm. These high-pressure experiments were
conducted with helium as the inert gas and with a reduced
oxygen concentration to prevent flow instabilities and maintain
laminar behavior. Specifically, laminar flame speed measure-
ments in CH4/O2/He mixtures at various equivalence ratios
were conducted at 40 atm using an oxidizing gas consisting of
17% O2 and 83% He by volume and at 60 atm using an
oxidizing gas with 15% O2 and 85% He by volume. For these
experiments, the unburned gas temperature was held fixed at
298 K.
Figure 7 compares the experimental and calculated results
using the reduced mechanism at both 40 and 60 atm. In the 40
Figure 7. Comparisons of experimental laminar flame speeds with the
results of CSU141 computations for CH4/O2/He mixtures at (a) 40
atm and (b) 60 atm. The experimental data are from Rozenchan et
al.25 The indicated % O2 refers to the molar composition in the O2/He
oxidizer gas.
atm case, the CSU141 mechanism agrees with experiments to
within 6% for lean to stoichiometric conditions and slightly
overpredicts the flame speed for rich conditions. For the 60 atm
case, the reduced CSU141 mechanism overpredicts the flame
speed across the range of equivalence ratios investigated, but
the deviation diminishes as the mixture becomes leaner. The
relative error in the flame speed for the leanest condition, φ =
0.85, is only 8%, and that for φ = 0.9 is 16%. Martinez et al.26
compared the results from a variety of natural gas mechanisms
with experimental laminar flame speeds for methane/air
mixtures at 20 atm and different equivalence ratios. At φ =
0.9 they found a widespread error in the predictions, with most
of the mechanisms overpredicting the flame speed by 20% to
100%. Because the agreement between the simulations and
experiments for methane laminar flame speed at high pressure
had less than 20% error, CSU141 is as effective as any of the
natural gas mechanisms evaluated by Martinez et al. In addition,
2421
Article
the highly turbulent flame speeds experienced in a diesel dual-
fuel engine can often be more dependent on the turbulence
intensity than the laminar flame speed, which allows greater
error in laminar flame speed predictions to be tolerated.
Because the natural gas/air mixtures in dual-fuel engines are
usually lean and contain mostly methane, the results in Figure 7
are encouraging and suggest that the reduced CSU141
mechanism can be used to predict flame speeds for lean
natural gas mixtures at high pressures in dual-fuel engines.
5.5. Combusting Diesel Spray. The ECN spray data-
base23 includes a set of data for a diesel spray at an injection
pressure of 1400 bar into a chamber filled with 21% oxygen at a
nominal temperature of 1000 K and a density of 30 kg/m3,
which are typical TDC conditions for high-compression-ratio
diesel engines. This data set was used to validate the combined
capabilities of the spray, turbulence, and reduced chemical
kinetic mechanism to simulate reacting diesel sprays. Table 6
Table 6. Experimental Conditions for the ECN Combusting
Diesel Spray Used for Model Validation23
fuel diesel #2
injection pressure (bar) 1400
injected mass (mg) 14.3
injection duration (ms) 5.1
nominal ambient temperature (K) 1000
nominal ambient density (kg/m3) 30
oxygen concentration (vol %) 21
nozzle orifice diameter (mm) 0.100
discharge coefficient 0.8
fuel temperature (K) 436
lists some of the important experimental conditions for this
spray, while more information can be found on the ECN Web
site.23 The experimental ignition delay and flame lift-off length
are used as benchmarks to validate the simulations. The
experimental ignition delay is based on the time between the
start of injection and the time of rapid pressure rise. In the
simulations, the ignition delay is the elapsed time when the
temperature derivative reaches a local maximum that
corresponds to the largest temperature rise. The experimental
lift-off length is the axial distance from the nozzle to the flame
as measured by OH chemiluminescence. The simulated lift-off
length is indicated by the location where the OH mass fraction
becomes larger than 2% of the maximum in the domain. The
simulated lift-off length was measured at 2 ms, which is well
after steady-state behavior was established. For these
simulations, the liquid properties of diesel were represented
using the properties of n-tetradecane because it was found that
the use of n-tetradecane had a 1 mm improvement in lift-off
length compared with the use of “DIESEL2” from the
CONVERGE liquids property library.
Figure 8 shows the simulation results using a plane through
the center of the spray colored by the simulated mass fraction
of OH; the lift-off length is also indicated. Table 7 presents a
comparison of the simulated and experimental ignition delays
and lift-off lengths. The simulated ignition delay is within 14%
of the experimental value. When the ignition delay periods are
converted to crank angle degrees at an engine speed of 2000
rpm, the simulated ignition delay is longer than the
experimental one by only 0.4 crank angle degree, which is
acceptable for engine simulations. The simulated lift-off length
is within 20% of the experimental value, with a difference of 2
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
Figure 8. Simulation results for OH mass fraction at 2 ms. The lift-off
length is defined as the first location of 2% of the maximum OH mass
fraction.
Table 7. Comparison of Experimental and Simulated
Ignition Delays and Lift-off Lengths
ignition delay (ms) lift-off length (mm)
experiment 0.24 9.5
simulation 0.273 11.4
mm. The errors in the ignition delay and lift-off length obtained
in this study are comparable to the levels of error achieved by
other authors. For instance, Bajaj et al.27 compared simulations
and optical spray experiments of reacting n-heptane sprays
across a wide range of ambient conditions and generally
achieved lift-off length agreement within 25% and ignition delay
agreement within 30%. The error in the ignition delay and lift-
off length are the result of an accumulation of multiple error
sources, including the accuracy of the detailed mechanism,
turbulence-chemistry interactions omitted by assuming a well-
stirred reactor in each cell, solving only for the ensemble-
averaged flow via the RANS turbulence model, and the method
used to define the experimental values.27 The agreement
between model and experiment demonstrates that the coupling
of the reduced chemistry, turbulence, and spray models can
provide reasonable predictions of non-premixed diesel jet
flames that are useful for engine modeling.
5.6. Multidimensional Engine Simulations. The exper-
imental and computational pressures and heat release rates for
the D25G17 and D12G22 SOI timing sweeps are compared in
Figure 9a and Figure 9b, respectively. Each simulation took
approximately 12 h to compute using 48 cores. As shown in
Figure 9, excellent agreement between the model and
experiment was obtained for the start of combustion phasing,
Figure 9. Comparisons of experimental and computational pressures and AHRRs for (a) the D25G17 SOI sweep and (b) the D12G22 SOI sweep.
2422
Article
pressure rise rate, peak pressure, combustion duration, and
shape of the heat release rate as the injection timing changed. A
quantitative summary of the errors from the engine simulations
is given at the end of this section. There is some overprediction
of the AHRR during the premixed ignition of the spray, as
indicated by the peak in the AHRR at the start of combustion.
For dual-fuel combustion, the first peak in the AHRR curve
includes heat release from the premixed burn of the diesel
vapors and entrained natural gas in the head of the spray. The
slight discrepancy between the model and experiment for the
AHRR during spray ignition could be the result of using a
single-component fuel surrogate for the diesel chemistry, using
a single-component liquid surrogate for the vaporization model,
error in the measurement method for the injector rate shape,
experimental uncertainty in the mass of diesel fuel injected, and
using a sector simulation that assumes perfectly symmetrical
behavior for each spray. Overall, the model effectively captures
the general trends in pressure and AHRR behavior for the two
injection timings. The model predictions are particularly
noteworthy given the fact that no adjustments (i.e., tuning)
of the model constants was done between different injection
timing cases. The only differences between the input files other
than the injection timing included accounting for small
variations in manifold pressure and the ingested air mass, as
these parameters varied slightly in the experiments as the
injection timing was advanced.
In-cylinder temperature plots in the plane of the spray axis
from the D25G17 simulation with −13° SOI are shown in
Figure 10. From Figure 10a it can be seen that the ignition of
the spray in dual-fuel combustion is similar to traditional diesel
spray ignition, where a premixed autoignition reaction zone is
observed near the head of the spray plume. The premixed
autoignition zone grows around the spray and transitions into a
non-premixed flame by approximately −1° after TDC
(ATDC), as shown in Figure 10b. The non-premixed spray
flame then initiates a premixed propagating natural gas flame, as
shown by the growth of the green area, which indicates a flame
temperature of approximately 2000 K, in Figure 10c−f. It
should be noted that the non-premixed diesel flame continues
to burn during natural gas flame propagation and therefore a
dual-fuel model must be able to simultaneously simulate the
important physics in both non-premixed and premixed flames.
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
Figure 10. Time development of the temperature in the plane of the spray axis for the D25G17, −13° ATDC SOI case. Time proceeds
sequentiallyin going from (a) to (f).
The images in Figure 10 demonstrate that the essential physical
phenomena of the diesel spray ignition, the non-premixed
diesel jet flame, and the premixed propagating natural gas flame
necessary for reproducing the global AHRR have been
effectively captured via the coupling of the reduced chemical
kinetic mechanism, the RANS turbulence model, and the spray
models described herein.
Figure 11 presents comparisons between the experiments
and simulations for the natural gas load sweep using a fixed
Figure 11. Comparisons of experimental and computational pressures
and AHRRs for the dual-fuel load sweep using a constant diesel
injection of 11 mg.
boost pressure and fixed diesel injection mass and timing. This
load sweep therefore provides validation with variation in the
natural gas equivalence ratio. As previously explained, these
experiments were conducted with a higher-reactivity natural gas
mixture because of the greater ethane and propane contents.
Additionally, the boost pressure used for the load sweep was
appreciably lower than that used in the SOI sweeps. The
simulations were able to capture the start of combustion timing,
the pressure rise rate, the shape and magnitude of the AHRR
profile, and the combustion duration. Again, some over-
prediction of the heat release rates during spray ignition was
2423
Article
observed, which can be attributed to the same sources of error
discussed above for the SOI timing sweep. The peak pressure
was somewhat underpredicted, and as a result, the expansion
pressure was also lower. Examination of the experimental
AHRR for the D11G18 case near the timing of peak pressure
shows a small increase in the AHRR that is not captured by the
simulations. This discrepancy could be due to sources of error
that are not specifically related to the accuracy of the chemistry
in the reduced chemical mechanism, such as uncertainties in the
air and fuel mass flow measurements, the assumptions of a
RANS-based turbulence model, constant wall temperatures,
and homogeneous composition and temperature at IVC, and
the symmetry assumption from using a sector simulation. In
general, the agreement in Figure 11 demonstrates the ability of
the reduced mechanism to predict the combustion behavior
responsible for the observed AHRR curves at two widely
different natural gas equivalence ratios, natural gas substitution
percentages, and total fuel masses. In particular, the CSU141
mechanism is able to capture the observed AHRR curves for a
natural gas mixture with higher reactivity than the previous
mixture used in the SOI sweeps.
Comparisons between 100-cycle-averaged experimental
engine data and volume-averaged simulation results for the
case with additional propane, D12G22P2, are shown in Figure
12a. The experimental AHRR curve shows a sudden
acceleration in combustion rate that occurs just before peak
pressure. The simulation also captures the timing of this spike
in the AHRR, but the magnitude is somewhat higher, leading to
a higher peak pressure and consequently a higher pressure
during expansion. Figure 12b compares the experimental
pressure from a single cycle and the simulated pressure from
a single cell at the same location as the pressure transducer tip.
The single-cycle experimental pressure shows fluctuations that
can be classified as engine knock. The single-cell simulation
pressure shows pressure fluctuations with similar frequency and
amplitude beginning when the fast combustion event occurs.
The agreement between the AHRR profiles for the propane
addition experiment shows that the CSU141 mechanism can be
useful in explaining the chemical kinetic processes that cause
knocking in dual fuel engines. Figure 13 presents three-
dimensional images of the engine sector showing the
development of a 2000 K isosurface, which is indicative of
the boundary between the unburned and burned regions. At 1°
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Figure 12. Comparisons between experiment and simulation for the additional propane case where engine knocking was observed: (a) 100-cycle-
averaged experimental pressure and volume-averaged simulation pressure; (b) single-cycle experimental pressure and simulation pressure at a single
cell with the same location as the pressure transducer tip.

Figure 13. Development of a 2000 K temperature isosurface with a local heat release rate clip plane showing three flames in the bottom of the bowl
region for the propane addition case.

ATDC (Figure 13a), the burned region has enveloped the
spray axis, and the non-premixed diesel spray flame has been
established. A portion of the temperature isosurface has crossed
the periodic boundary and therefore represents the burned
region from the adjacent diesel spray moving into the sector.
The swirl motion causes the burned region to pass through
only one of the periodic boundaries. By 3° ATDC (Figure
13b), the burned region has grown in a direction normal to the
spray axis, which is representative of a propagating natural gas
flame. Also, the burned region has moved downward along the
bowl wall as a result of the impact of the diesel jet. The growing
2424
isosurface in the squish region is due to natural gas flame
propagation. Figure 13b shows that the premixed flame from
the adjacent spray has merged with the flame from this sector,
creating a cusplike crease where the two flames meet. In the
6.5° ATDC image (Figure 13c), a planar cross section plot of
the local heat release rate has been added to show the location
of the reaction zones in the bottom of the bowl in relation to
the temperature isosurface. To provide a clearer view of the
local heat release planar cross section, the 2000 K isosurface has
been removed in Figure 13d, which depicts three flames. The
flame along the bowl wall is the non-premixed diesel flame
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

Figure 14. Plots of local heat release rates in the bowl region during the uncontrolled fast combustion event for the D12G22P2 propane addition
case.

spreading down into the bottom of the bowl, and the other two Table 8. Summary of Error in Combustion Metrics between
are due to the merging premixed natural gas flames from Engine Experiments and Simulations
adjacent spray axes. All three are expanding in size as they travel
CA10− peak pressure
into the bottom of the bowl. CA10 CA50 90 pressure rise rate
The growth of these three flames during the uncontrolled error error error error error
fast combustion event are explained using the planar cross case name (deg) (deg) (deg) (bar) (bar/deg)
sections of local heat release rates shown in Figure 14. As the D12G22, SOI = −10° 0.41 0.03 1.10 0.06 0.77
flames move toward each other (Figure 14a,b), volumetric heat D12G22, SOI = −14° 0.15 0.11 1.60 2.63 0.56
release begins to occur throughout the volume of gas in front of D25G17, SOI = −8.8° 0.50 0.37 1.12 0.12 0.05
and between the natural gas flames that are in the process of D25G17, SOI = −13° 0.13 0.14 4.75 0.22 1.09
merging together. Volumetric heat release does not occur near D11G8 0.25 0.39 4.00 1.70 0.97
the walls of the bowl because of heat loss to the walls. A quarter D11G18 0.13 0.23 1.47 4.43 0.69
of a degree later (Figure 14c), the rate of heat release D12G22P2 0.14 0.12 2.51 6.62 5.52
throughout the volume in front of the merging natural gas
flames has increased considerably, indicating volumetric larger errors in combustion duration were the result of a small
ignition. This volumetric heat release occurs at 8.50° ATDC, disagreement in the heat release rate profile at the tail end of
which is the same timing as the peak in the AHRR from the the combustion duration. The peak pressure was matched
simulation. At 8.75° ATDC (Figure 14d), this volume is no within 5 bar except for the D12G22P2 propane addition
longer producing heat and the flames have traveled to the walls, experiment, where a slight overprediction of the magnitude of
where they are slowed by heat transfer. Therefore, the the fast combustion event led to a greater increase in pressure.
simulation shows that the uncontrolled fast combustion This fast combustion also led to greater pressure rise rate error,
observed in the propane addition experiment is likely due to which was otherwise within 1.1 bar/deg. The level of agreement
volumetric ignition of natural gas in front of merging natural between the engine experiments and simulations shown in
gas flames originating from adjacent diesel sprays. Table 8 demonstrates that the reduced mechanism is a useful
A summary of the agreement between the combustion model for dual-fuel engine simulations.
metrics obtained from the engine experiments and simulations
is given in Table 8. The ignition timing, as indicated by the 6. CONCLUSIONS
crank angle at which 10% of the total integrated heat release has A reduced chemical kinetic mechanism consisting of 141
been reached (CA10), shows that the simulation was within 0.5 species and 709 reactions has been formulated for natural gas/
crank angle degree for all of the operating conditions diesel dual-fuel combustion wherein n-heptane is used to
investigated. The simulations were able to predict the 50% represent diesel fuel and a mixture of methane, ethane, and
heat release timing (CA50) within 0.5 crank angle degree. The propane represents natural gas. The final reduced mechanism is
CA10−90 combustion duration, which is the crank angle a combination of reduced versions of the detailed n-heptane
difference between the 10% and 90% heat release timings, mechanism of Curran et al.15,16 and the detailed methane
shows agreement within 5 crank angle degrees. Cases with through n-pentane mechanism of Healy et al.14 Ignition delay
2425 DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels
plots for n-heptane showed excellent agreement between the
reduced CSU141 mechanism and the detailed LLNL
mechanism for pressures from 10 to 80 bar, temperatures
from 700 to 1400 K, and equivalence ratios from φ = 0.2 (very
lean) to φ = 2.0 (rich). The natural gas chemistry was validated
using ignition delay plots for two different natural gas
compositions with varying ethane and propane contents, and
excellent agreement with the detailed NUIG mechanism was
achieved for lean to stoichiometric conditions and pressures
between 10 and 80 bar. However, the detailed mechanisms
have only been validated using experimental data up to 30 bar,
and therefore, the agreement at 80 bar assumes that the rate
constants in the detailed mechanisms are accurate at higher
pressures. Additionally, the reduced mechanism achieved
excellent agreement with the detailed mechanisms for initial
temperature sweeps of adiabatic HCCI simulations for both n-
heptane and two natural gas mixtures. Furthermore, the
CSU141 mechanism was able to replicate experimental laminar
flame speeds for lean CH4/He/O2 mixtures at 40 and 60 atm to
within 16% relative error. Finally, the n-heptane chemistry in
the CSU141 mechanism was able to predict the experimental
ignition delay and lift-off length from a reacting diesel spray in a
constant-volume chamber at engine TDC conditions to within
14% and 20% error, respectively.
Comparisons between natural gas/diesel dual-fuel engine
experiments and multidimensional CFD simulations showed
excellent pressure and AHRR agreement for two start of
injection timing sweeps with different substitution percentages,
a natural gas load sweep at fixed diesel injection, and an engine
knocking case using additional propane to increase the mixture
reactivity. The simulations sufficiently captured the start of
combustion timing, the pressure rise rate, the shape and
magnitude of the heat release rate profile, and the combustion
duration. In the engine knocking case, the simulation was able
to predict the timing and magnitude of a sudden acceleration in
combustion rate as well as the high-frequency pressure
oscillations. In-cylinder images from the simulation indicate
that this fast combustion was due to volumetric heat release in
the bottom of the bowl where propagating natural gas flames
originating from adjacent spray axes merge together.
Some of the simulations showed a small discrepancy in peak
pressure and expansion pressure, but this could be due to error
sources other than the chemical kinetics. Such sources of error
include uncertainties in the air and fuel mass flow rates, solving
for the ensemble-averaged flow using a RANS turbulence
model, assuming homogeneous composition and temperature
at IVC, and the sector geometry assumption that the
combustion behavior is perfectly symmetric with respect to
the spray axis. Despite these sources of error, the agreement
achieved from coupling of the reduced mechanism with the
spray models and the RANS turbulence model demonstrates
that the reduced mechanism is a useful tool for capturing the
important chemical kinetics in diesel ignition, natural gas flame
propagation, and natural gas autoignition occurring in dual-fuel
engines. It particular, the reduced mechanism is able to
accurately model varying natural gas reactivity without relying
on rate constant tuning.
Future validation work with natural gas/diesel RCCI engine
experiments is necessary in order to confirm whether the
CSU141 mechanism is capable of modeling premixed auto-
ignition of large reactivity gradients where the diesel/natural
gas mixtures can vary from slightly rich to very lean. In addition,
it should be reiterated that the CSU141 mechanism makes the
2426
Article
assumption that the concentrations of hydrocarbons heavier
than propane in the natural gas mixture are low enough that
their effect on the ignition delay and flame speed is negligible.
Future work is necessary in order to develop a reduced dual-
fuel mechanism that includes the chemical kinetics of heavier
hydrocarbons, such as butane, for dual-fuel applications that
require operation with a low-methane-number natural gas.
■
*
ASSOCIATED CONTENT
S Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/acs.energy-
fuels.5b02655.
Ignition delay data for methane, ethane, and propane and
additional details about validation of the spray model
using experimental data from the ECN (PDF)
CHEMKIN-formatted files for the reduced CSU141
mechanism (species, reactions, thermodynamic data, and
transport data) (ZIP)
■
AUTHOR INFORMATION
Corresponding Author
*E-mail: marchese@colostate.edu.
Notes
The authors declare no competing financial interest.
■ ACKNOWLEDGMENTS
The authors thank Woodward Inc. for funding this project and
providing access to their engine test facility. Particular
appreciation is extended to Jason Barta for help in conducting
the engine experiments. The authors are grateful to Convergent
Science, Inc. for their continuous technical support with the
engine simulations.
■
REFERENCES
(1) Annual Energy Outlook 2014 with Projections to 2040; Report
DOE/EIA-0383(2014); U.S. Energy Information Administration:
Washington, DC, 2014.
(2) EPA and NHTSA Adopt First-Ever Program To Reduce Greenhouse
Gas Emissions and Improve Fuel Efficiency of Medium- and Heavy-Duty
Vehicles; Regulatory Announcement EPA-420-F-11-031; U.S. Environ-
mental Protection Agency: Washington, DC, 2011.
(3) EPA and NHTSA Set Standards To Reduce Greenhouse Gases and
Improve Fuel Economy for Model Years 2017−2025 Cars and Light
Trucks; Regulatory Announcement EPA-420-F-12-051; U.S. Environ-
mental Protection Agency: Washington, DC, 2012.
(4) Serrano, D.; Bertrand, L. Exploring the Potential of Dual Fuel
Diesel-CNG Combustion for Passenger Car Engine. Lect. Notes Electr.
Eng. 2013, 191, 139−153.
(5) Fourth Assessment Report: Climate Change 2007: Working Group I:
The Physical Science Basis, Section 2.10.2: Direct Global Warming
Potentials; Intergovernmental Panel on Climate Change: Geneva,
2007; Table 2.14.
(6) Konigsson, F. Advancing the Limits of Dual Fuel Combustion.
Licentiate Thesis, Department of Machine Design, Royal Institute of
Technology, Stockholm, Sweden, 2012.
(7) Nieman, D.; Dempsey, A.; Reitz, R. D. Heavy-Duty RCCI
Operation Using Natural Gas and Diesel. SAE Technol. Pap. Ser. 2012,
5, 270−285.
(8) Bourque, G.; Healy, D.; Curran, H. J.; Zinner, C.; Kalitan, D.; de
Vries, J.; Aul, C.; Petersen, E. Ignition and Flame Speed Kinetics of
Two Natural Gas Blends with High Levels of Heavier Hydrocarbons.
Proc. ASME Turbo Expo 2008, 3, 1051.
DOI: 10.1021/acs.energyfuels.5b02655
Energy Fuels 2016, 30, 2414−2427
Energy & Fuels Article

(9) Rahimi, A.; Fatehifar, E.; Saray, R. K. Development of an

optimized chemical kinetic mechanism for homogeneous charge
compression ignition combustion of a fuel blend of n-heptane and
natural gas using a genetic algorithm. Proc. Inst. Mech. Eng., Part D
2010, 224, 1141−1159.
(10) Smith, G. P.; Golden, D. M.; Frenklach, M.; Moriarty, N. W.;
Eiteneer, B.; Goldenberg, M.; Bowman, C. T.; Hanson, R. K.; Song, S.;
Gardiner, W. C.; Lissianski, V. V.; Qin, Z. GRI-Mech Home Page.
http://www.me.berkeley.edu/gri_mech/ (accessed Nov 10, 2015).
(11) Senecal, P.; Pomraning, E.; Richards, K.; Som, S. Grid-
convergent spray models for internal combustion engine CFD
simulations. Proc. ASME ICEF 2012, 697−710.
(12) Dagaut, P. On the kinetics of hydrocarbons oxidation from
natural gas to kerosene and diesel fuel. Phys. Chem. Chem. Phys. 2002,
4, 2079−2094.
(13) Turbiez, A.; El Bakali, A.; Pauwels, J. F.; Rida, A.; Meunier, P.
Experimental study of a low pressure stoichiometric premixed
methane, methane/ethane, methane/ethane/propane and synthetic
natural gas flames. Fuel 2004, 83, 933−941.
(14) Healy, D.; Kalitan, D.; Aul, C.; Petersen, E.; Bourque, G.;
Curran, H. J. Oxidation of C1-C5 Alkane Quinternary Natural Gas
Mixtures at High Pressures. Energy Fuels 2010, 24, 1521−1528.
(15) Curran, H. J.; Gaffuri, P.; Pitz, W. J.; Westbrook, C. K. A
Comprehensive Modeling Study of n-Heptane Oxidation. Combust.
Flame 1998, 114, 149−177.
(16) Curran, H. J.; Gaffuri, P.; Pitz, W. J.; Westbrook, C. K. A
Comprehensive Modeling Study of iso-Octane Oxidation. Combust.
Flame 2002, 129, 253−280.
(17) Niemeyer, K. E.; Sung, C.-J.; Raju, M. P. Skeletal mechanism
generation for surrogate fuels using directed relation graph with error
propagation and sensitivity analysis. Combust. Flame 2010, 157, 1760−
1770.
(18) Hockett, A. A Computational and Experimental Study on
Combustion Processes in Natural Gas/Diesel Dual Fuel Engines.
Ph.D. Dissertation, Colorado State University, Fort Collins, CO, 2015.
(19) CONVERGE 2.2.0 Theory Manual; Convergent Science Inc.:
Madison, WI, 2014.
(20) Pomraning, E.; Richards, K.; Senecal, P. Modeling Turbulent
Combustion Using a RANS Model, Detailed Chemistry, and Adaptive
Mesh Refinement. SAE Tech. Pap. Ser. 2014, DOI: 10.4271/2014-01-
1116.
(21) Han, Z.; Reitz, R. D. Turbulence Modeling of Internal
Combustion Engines Using RNG k-ε models. Combust. Sci. Technol.
1995, 106, 267−295.
(22) Amsden, A. KIVA 3-V: A Block Structured KIVA Program for
Engines with Vertical or Canted Valves; Report LA-13313-MS; Los
Alamos National Laboratory: Los Alamos, NM, 1997.
(23) Sandia National Laboratories Engine Combustion Network.
Diesel Spray Combustion Home Page. http://www.sandia.gov/ecn/
dieselSprayCombustion.php (accessed in 2014 and 2015).
(24) Healy, D.; Curran, H. J.; Simmie, J. M.; Kalitan, D. M.; Zinner, J.
M.; Barrett, A. B.; Petersen, E. L.; Bourque, G. Methane/ethane/
propane mixture oxidation at high pressures and at high, intermediate,
and low temperatures. Combust. Flame 2008, 155, 441−448.
(25) Rozenchan, G.; Zhu, D. L.; Law, C. K.; Tse, S. D. Outward
Propagation, Burning Velocities, and Chemical Effects of Methane
Flames up to 60 atm. Proc. Combust. Inst. 2002, 29, 1461−1469.
(26) Martinez-Morett, D.; Tozzi, L.; Marchese, A. J. A Reduced
Chemical Kinetic Mechanism for CFD Simulations of High BMEP,
Lean-Burn Natural Gas Engines. Proc. ASME ICES 2012, 61−71.
(27) Bajaj, C.; Ameen, M.; Abraham, J. Evaluation of an Unsteady
Flamelet Progress Variable Model for Autoignition and Flame Lift-Off
in Diesel Jets. Combust. Sci. Technol. 2013, 185, 454−472.

2427 DOI: 10.1021/acs.energyfuels.5b02655

Energy Fuels 2016, 30, 2414−2427