Fuel 305 (2021) 121522

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Optimization of a diesel/natural gas dual fuel engine under different diesel

substitution ratios
Jie Liu a, b, *, Qiang Guo a, b, Jiayong Guo a, b, Fan Wang a, b
a
Department of Power Mechanical Engineering, Beijing Jiaotong University, Beijing 100044, PR China
b
Beijing Key Laboratory of New Energy Vehicle Powertrain Technology, Beijing 100044, PR China

A R T I C L E I N F O A B S T R A C T

Keywords: Dual fuel engines are attracting more and more attentions due to their low emissions and high efficiency
Optimization compared with traditional single fuel engines. However, complex parameters coupling remains challengeable
Genetic algorithm issues in dual fuel engine optimization. Here, the objective of this study is to optimize the diesel fuel ratio and the
Natural gas
combustion parameters of a diesel/natural gas dual fuel engine by coupling the genetic algorithm with a 3D CFD
Dual fuel
engine combustion simulation. A total number of 9 parameters were optimized simultaneously. The results
indicate that the optimized diesel injection timing is gradually postponed with the increase of the pilot diesel
quantity. The indicated specific energy consumption (ISEC) can be achieved when the fuel ratio is in the range of
0.2 to 0.3 and injection timing is in the range of − 30 to − 15 ◦ CA ATDC. The open bowl geometry is chosen in all
the Pareto front solutions, which indicates the advantages of this chamber type for the dual fuel combustion
mode. The compression ratios, exhaust gas recirculation (EGR) rates are gradually concentrated to several
optimal values, while the swirl ratios are widely distributed in the range from 0.5 to 1.2. Furthermore, the
temporal and spatial distribution of the cylinder temperature, NOx and CH4 emissions analysis showed that the
higher diesel fuel ratio will generate a widely distributed high temperature region, which is beneficial for the
methane oxidation. In addition, lower EGR rate and higher diesel injection quantity are beneficial to methane
oxidation. This work illustrates a new optimization method for dual fuel engine complex parameters.

It is of great significance to explore the injection parameters of

1. Introduction diesel/natural gas dual fuel engines, the low reactivity natural gas fuel
was introduced through the port injection to form a premixed mixture,
With the increasing pollution of the environment and the depletion and the ignition timing is controlled by direct injection of the high
of oil resources, how to improve the thermal efficiency of internal reactivity diesel fuel. Yousefi et al. [11] investigated the impacts of
combustion engines while reducing pollutant emissions has been diesel injection timing on the performance of a heavy-duty natural gas/
attracting more and more attention. Using advanced combustion tech­ diesel dual fuel engine. The results showed that the peak cylinder
nology with alternative fuels, such as natural gas [1] and oxygenated pressure, thermal efficiency, and NOx emission all increased with
fuels [2], can simultaneously reduce NOx and Soot emissions, and advancing diesel injection timings under all the load-speed conditions.
improve thermal efficiency. However, there are still plenty of challenges The research of Wang et al. [12] indicated that a two-stage auto-ignition
in using advanced combustion modes in single-fuel engines, such as mode can be achieved by significantly advancing the diesel injection
difficulty in controlling the combustion phase [3] combustion instability timing. The results were also verified by W Kim [13]. In his study, an
[4], narrow operating range [5] and massive emissions of pollutants earlier injection timing created a rapid influx of propagating flame,
[2,6]. Dual fuel compression ignition combustion modes, such as while effectively suppressing the formation of intense soot radiation
Reactive Controlled Compression Ignition (RCCI) [7,8], and dual fuel through longer ignition delay. Yousefi et al. [14] also investigated the
combustion [9,10], are considered to be more competitive to overcome effect of single and double pulses injection strategies of pilot diesel on
the above shortcomings of the single fuel advanced combustion mode. the combustion and emission performance of a heavy-duty natural gas/
Therefore, since the superiority of dual fuel diesel engines, it is prefer­ diesel dual fuel engine. The results showed that split injection mode
able to optimize a solution for diesel engine parameters. considerably increased the peak in-cylinder pressure and the indicated

* Corresponding author at: Department of Power Mechanical Engineering, Beijing Jiaotong University, Beijing 100044, PR China.
E-mail address: ljie@bjtu.edu.cn (J. Liu).

https://doi.org/10.1016/j.fuel.2021.121522
Received 6 May 2021; Received in revised form 29 June 2021; Accepted 18 July 2021
Available online 26 July 2021
0016-2361/© 2021 Elsevier Ltd. All rights reserved.
J. Liu et al. Fuel 305 (2021) 121522

Nomenclature HACA Hydrogen Extraction/Carbon addition

EDC Eddy-Dissipation Concept
Abbreviations ISFC Indicated Specific Fuel Consumption
ISEC Indicated Specific Energy Consumption IVC Intake Valve Closure
EGR Exhaust Gas Recirculation EVO Exhaust Valve Open
RCCI Reactivity Controlled Compression Ignition FR Fuel Ratio
FID Flame Ionization Detector ATDC After Top Dead Center
THC Total Hydrocarbon TDC Top Dead Center
CNG Compressed Natural Gas RD Radius Difference
CFD Computational Fluid Dynamics IT Injection timing
KH-RT Kelvin-Helmholtz instability theory and Rayleigh-Taylor SA Spray angle
instability theory CR Compression Ratio
PaSR Partially Stirred Reactor SR Swirl ratio

thermal efficiency increased (on average) by 7% compared to that of a

single injection. Huang et al. [15] studied the effect of multiple in­ Table 1
jections on the low-load combustion and emission characteristics of Specifications of the dual fuel engine.
natural gas/diesel dual fuel engines. The results show that when the first Engine parameters Specifications
injection timing of diesel is advanced, the indicated thermal efficiency
Bore × Stroke 126 × 130 mm
and NOx emission firstly increase and then decrease, and methane
Cylinder number 6
emission firstly decreases and then increases. Shu et al. [16] confirmed Displacement 9.726 L
that the peak cylinder increased and the NOx emission ascended as the Rated power/speed 247 kW/2200 rpm
spray angle increased from 60◦ to 140◦ , and an opposite trend would be Compression ratio 17.0
Number of injector nozzle holes 7
shown with further increase the spray angle. The computational result of
Spray angle 146◦
Lee et al. [17] demonstrated that a 25% smaller diameter nozzle with an Exhaust valve opening timing 131 ◦ CA ATDC
advanced rate-of-injection exhibited a 3.4% increase in the gross indi­ Inlet valve closing timing 146 ◦ CA ATDC
cated mean effective pressure. However, most of the studies are mainly
focusing on the one or two injection parameters, the comprehensive
parameters of diesel engine are seldom investigated. solutions can meet the EURO VI emission limit. Salek et al. [25] con­
In compressed natural gas/diesel dual fuel combustion mode, emis­ ducted multi-objective optimization of the engine performance and
sions and performance characteristics of the engines are significantly emissions for a dual fuel engine. The optimized dual fuel combustion
affected by the diesel substitution rate. Vávra et al. suggested [18] that scheme showed lower emissions, higher thermal efficiency including the
the higher the natural gas replacement ratio, the earlier the fuel injec­ engine equivalent brake specific fuel consumption.
tion timing was needed. Zheng [19] revealed that regardless of the From what has been discussed above, although there are plenty of
nozzle parameters, EGR rate and injection parameters, the natural gas studies focusing on the optimization of the combustion parameters of
consumption rate decreased with the increase of the equivalence ratio. the diesel/natural gas dual fuel engines, the combined optimization of
Lee et al. [20]. investigated the effect of various natural gas substitution injection timing, diesel substitution ratio and combustion chamber ge­
ratios on the combustion and Soot emission characteristics in an optical ometry has been rarely carried out. The main objective of this study is to
diesel engine. The results showed that with the increase of the diesel find the optimized diesel fuel injection timings and diesel fuel ratios
replacement rate, the peak pressure and heat release rate decreased, the under various combustion chamber geometries. The artificial intelli­
ignition delay period was prolonged, and the premixed flame distribu­ gence genetic algorithm, a new method, coupled with KIVA-3V was used
tion was more extensive. Geng, L. et al. [21] suggested that in terms of to solve the multi-objective optimization problem. A total number of 9
exhaust emissions, with increasing diesel replacement rate, the soot parameters were optimized simultaneously, which includes diesel in­
emissions decreased while the NOx emissions and the HC emissions jection timings, diesel fuel ratios, operating parameters and combustion
increased. Nevertheless, there are few studies on simultaneous optimi­ bowl geometries.
zation of fuel substitution ratio and multi-parameter combustion in dual
fuel engines. 2. Experimental setup
Since there are too many control parameters in the dual fuel engines,
multi-objective optimization of the injection parameters and operating In this section, the specifications of the test engine are introduced
parameters need to be carried out to obtain higher thermal efficiency first. After that, the equipment and test methods for the measurement of
and lower exhaust emissions. The influence of diesel injection strategy engine exhaust emissions are given. Finally, the characteristics of the
on natural gas/diesel dual fuel combustion was studied by using micro- fuels used in the experiment are briefly introduced.
genetic algorithm (micro-GA) optimization [22]. The results showed
that for the double injection strategy, the thermal efficiency of more 2.1. Engine specification
than 45% could be obtained under the constraints of peak pressure, peak
pressure rise rate and NOx emission. Liu et al. [23] employed the genetic The experiment was conducted on a modified six-cylinder Weichai
algorithm NSGA-II coupled with the KIVA-3V code to optimize the WP10 diesel engine. Natural gas was introduced by adding an extra
diesel/natural gas dual-fuel engine with multiple objectives. The results natural gas supply system to the original diesel engine and mixed with
showed that the straight combustion chamber was effective to reduce air through a mixer installed on the intake manifold. The diesel fuel
CH4 emission and improve fuel economy. After that, the genetic algo­ supply system of the original engine has been retained to account for the
rithm was used to optimize the injection parameters, operating param­ diesel fuel injection. An additional dual fuel engine control unit was
eters and combustion system of the diesel/natural gas dual fuel engine used to control the diesel and natural gas injection strategies including
simultaneously [24]. The results showed that almost all the optimized the injection timing, injection duration and so on. Table 1 gives the

2
J. Liu et al. Fuel 305 (2021) 121522

Table 2 technical parameters of the dual fuel engine.

Properties of the test fuels.
CNG 0# diesel 2.2. Instrumental introduction
Methane 96.51% v/ Cetane number 52.6
v During the experiment, the Horiba-Schenck HT350 AC dynamometer
Ethane 1.2% v/v Density 833.7 kg/ was used to measure the power and torque of the engine. The cylinder
m3 pressure was obtained through a piezoresistive pressure sensor installed
Propane 0.18% v/v Lower heat value 42.74 MJ/ in the first cylinder of the engine coupled with a Kistler 5018 charge
kg
Butane 0.04% v/v Stoichiometric air–fuel 14.5 kg/kg
amplifier. The engine emissions were measured by using the Horiba
ratio MEXA7100DEGR exhaust gas analyzer. The flame ionization detector
Lower heat value 50.9 MJ/ (FID) was employed to detect the total hydrocarbon (THC) and methane
kg emission. The selective combustion technology was used in FID to
Stoichiometric air–fuel 16.88 kg/
measure the CH4 emission. The mass flow rates of diesel and natural gas
ratio kg
were measured by the ONOSOKKO FZ2100 and BROOKS mass flow
meters, respectively.

Table 3
2.3. Properties of the test fuels
KIVA-3V computing model.
Process Model Reference The diesel and natural gas as fuel used in the experiment were ob­
Droplet breakups KH-RT breakup model [27] tained from local gas stations in Beijing city. Table 2 lists the detailed
Droplet collision and O’Rourke collision model [28] properties of these two fuels.
coalescence
Turbulence RNG k-ε [29]
Ignition and combustion PaSR model [30]
3. CFD calculation methodology
The Soot formation and Hiroyasu model and Nagle/ [31]
oxidation process Strickland-Constable model. In this section, the models selected in the modified KIVA-3V code for
NOx generation Extended Zel’dovich NOx mechanism [33]. dual fuel combustion simulation are introduced firstly. After that, the
coupling method of the genetic algorithm NSGA-III and KIVA-3V code is
also introduced. Furthermore, some parameter definitions are

Fig. 1. The flow chart of the optimization process between NSGA-III and KIVA-3V.

3
J. Liu et al. Fuel 305 (2021) 121522

presented. multi-objective optimization problems [34]. Basically, the selection

mechanism of NSGA-III has undergone major changes compared with
3.1. Models used in the KIVA code the previous generation, which uses the reference point methodology to
find a set of preferred Pareto-optimal solutions. The NSGA-III algorithm
The main models used in the KIVA code are listed in Table 3. Diesel is coupled with the KIVA-3V code to carry out the optimization of the
fuel was represented by a mixture of 70% n-heptane and 30% toluene, dual fuel combustion process.
both expressed as volume fractions. Natural gas was given by a mixture Fig. 1 shows the overall coupling method of the NSGA-III, KIVA-3V
of methane, ethane, propane and butane. In addition, there were 81 and automatic mesh generation algorithm. Firstly, the NSGA-III program
elements and 421 elementary reactions in the diesel/natural gas blended randomly generates 9 parameters based on the range selection of the
fuel chemical reaction mechanism [26]. The KH-RT (Kelvin-Helmholtz variables. Then, the diesel injection timing, spray angle, swirl ratio, EGR
instability theory and Rayleigh-Taylor instability theory) breakup model rate, diesel fuel ratio and other parameters are sent to the KIVA-3V
was used to simulate the primary and secondary breakup processes of program. At the same time, the geometry parameters and compression
the spray droplets [27]. The collision and coalescence process between ratio of the combustion chamber are sent to the automatic grid gener­
the droplets were solved by using the O’Rourke collision model [28]. ation program to generate the combustion chamber grid. Subsequently,
The RNG k-ε model was used to describe the turbulence flow in the the KIVA-3V code conducts the calculation of the combustion process of
cylinder [29]. The combustion process was simulated by using the PaSR the dual fuel engine according to the newly generated initial conditions
(Partially Stirred Reactor) combustion model, which was widely used in and the combustion chamber grid. After that, the calculated engine
the combustion simulation of diesel engines [30]. In order to account for emissions and fuel economy data are sent to the NSGA-III program. In
the premixed combustion of natural gas, the micro-mixing time in the the next step, the NSGA-III program mixes the newly calculated results
original PaSR model was significantly reduced. The soot formation and with the previous optimal solutions and selects a set of the new optimal
oxidation process were simulated by using the Hiroyasu model and solutions, thereby forming the Pareto-optimal front of the current gen­
Nagle/Strickland-Constable model [31]. In addition, the growth of the eration. After that, the genetic algorithm will create the next generation
soot surface adopted the HACA (Hydrogen Extraction/Carbon Addition) of the parameters through the crossover and mutation methods and send
mechanism [32]. The formation of NOx was simulated by the extended them to the KIVA-3V and auto mesh generation program for next gen­
Zel’dovich NOx mechanism [33]. eration calculation. This cycle will continue until the maximum number
The source term of chemical reaction in the PaSR model can be of generations is reached. The optimal solution set is on the last Pareto-
simplified as follows, optimal front.
The gross ISEC rate is calculated based on the following equations:
c1 − c0
= kfm (c1 ) (1)
τ ISEC = ISFC × 42.74 MJ/kg (5)
In the equation, c0 represents the component at the inlet, c1 repre­
B Equivalent⋅disesel⋅consumption⋅rate
sents the component at the reactor outlet. τ is the time step, k is a ratio ISFC = = (6)
P (w0 + wc + we ) ∗ rpm/120
which is determined by the time scale τchem of chemical reaction
characteristics. where the B is the equivalent diesel consumption rate. The consumption
τchem rate of natural gas was transformed to the diesel consumption rate based
k= (2)
τchem + τmix on the same lower heat value. The total equivalent diesel consumption
rate was calculated based on the diesel fuel ratio FR.
Chemical reaction time scales can be defined in a number of ways. In
The w0 is the work integrated from the intake valve close to the
this paper, the chemical reaction time scale is defined based on the time
exhaust valve open period.
that the reference component cr participates in the reaction (the refer­
ence component has the shortest lifetime), as shown in Eq. (3). ∫ EVO
( ) w0 = PdV (7)
c c c c IVC
τchem = − r = min − 1 , − 2 , ⋅⋅⋅, − Ns (3)
fr f1 f2 fNs wc is the work done before the intake valve closes during the
compression process from the bottom dead center to intake valve close
τmix represents the characteristic time of the mixing process between
timing.
the unreacted fresh mixture and the flammable gas in the PaSR reaction
model. In this paper, the Taylor time scale is used to calculate mixing wc = (VIVC − VBDC ) ∗ PIVC (8)
time. For turbulence model κ − ε, it can be simplified into Formula (4):
we is the available work done after the exhaust valve opens.
κ
τmix = Cmix (4) we = (VBDC − VEVO ) ∗ (PIVC + PEVO ) ∗ 0.5 (9)
ε
The work for pumping loss was not considered in this study. So the
where Cmix is a constant. In order to account for the premixed natural gas
ISEC is called gross ISEC.
combustion process, the constant Cmix is modified to be a small number.
The definition of the diesel fuel ratio is as follows:
However, there is no specific modification for the flame quenching
process. For the quenching area, the temperature is too low, or the re­ HLD BD
FR = (10)
actants are too little or too much, these will lead to a low chemical re­ HLD BDO + HLG BGO
action, which partly reflects the phenomenon of quenching. The
quenching is mainly a response for the unburned methane, the calcu­ where FR is the diesel fuel ratio, HLD and HLG are the lower heating
lation methane emission result agreed well with the experimental data. values of diesel and natural gas, respectively. BDO and BGO are the
consumption rates of diesel and natural gas in the original engine. BD is
3.2. Optimization procedure the diesel fuel quantity in the current case.
During the calculations, the total fuel energy is kept unchanged, that
It is impossible to obtain a single solution to satisfy all the objectives is the total inlet energy is equal to (HLD*BDO + HLG*BGO). The diesel fuel
at the same time because of the trade-off relationship among several quantity is calculated based on the total fuel energy and diesel fuel ratio
objectives in engineering problems. In recent years, an evolutionary (FR). In addition, the quantity of natural gas is based on the total fuel
optimization algorithm of NSGA-III has been proposed and used to solve energy and natural gas substation ratio, which is 1 minus FR.

4
J. Liu et al. Fuel 305 (2021) 121522

Table 4 size is reduced to 0.15 mm or less, the difference between the cylinder
The selected variables and ranges for the optimization. pressures becomes very small. Therefore, in order to save the calculation
Variables Minimum Maximum Original time and maintain the calculation accuracy, the combustion process of a
dual fuel engine is simulated by using a 0.15 mm grid size and 60-degree
Injection timing/ CA ATDC
◦
− 40.0 0.0 − 7.5
Spray cone angle/◦ 90.0 160.0 146.0 sector calculation area. The comparison between the cylinder pressures
Height of central pip (H1)/cm 0.5 1.525 1.525 and heat release rates predicted by the calculation and the experimental
Bottom radius (R1)/cm 3.3 4.8 3.69 data is shown in Fig. 4 (a). The calculated results are in good agreement
Throat radius (R2)/cm 3.5 4.5 3.765 with the experimental data. In addition, the predicted NO and CH4
Swirl ratio 0.5 2.0 1.4
Compression ratio 15.0 19.0 17.0
emissions also agree well with the experimental data, as shown in Fig. 4
Diesel fuel ratio 0.05 0.6 0.2 (b). The spray and evaporation models were validated by the experiment
EGR rate 0.0 30% 0.0 date taken from the Sandia ECN database. As the n-heptane was taken as
the main component of the diesel component in this study, the vapor and
liquid penetration of the n-heptane was used for the validation. It is
As the NSGA-III program has the capability of dealing with many
shown from Fig. 4 (c) that the fuel spray penetrations can be well pre­
objective optimization problems, the optimization objectives are
dicted by the models using in this study [2,6,35]. Therefore, the above
selected as the ISEC, NO, CH4 and Soot emissions. The optimization
spray and combustion sub-models, chemical reaction mechanism and
parameters include the diesel fuel ratio, diesel fuel injection timing and
the computational grid size will be used in the combustion optimization
spray cone angle, engine operating parameters and combustion chamber
of the dual fuel engine.
profile parameters. Table 4 lists the ranges selected for these 9 variables
during the optimization process.
The definition of the spray cone angle and the geometry parameters 4.2. Optimization results of emissions and ISEC
of the combustion chamber are shown in the following Fig. 2:
During the optimization process, five constraints are used to exclude Fig. 5 shows all the calculation results of the optimization process. It
irrational situations. The peak cylinder pressure, maximum pressure rise can be seen from the figure that the ISEC, NO and CH4 emissions are
rate and acceptable ISEC are limited to less than 20 MPa, 2 MPa/◦ CA and widely distributed in the whole space. Therefore, the selection of the
10.685 MJ/kW⋅h, respectively. In addition, during the selection of parameters during the optimization process is also widely distributed in
optimal parameters, it is also necessary to exclude the misfires and each setting range. At the same time, the Pareto front is located in the
abnormal termination cases in the calculation process [23]. front of all the results, which indicates that the selection mechanism of
the genetic algorithm can be used in multi-objective optimization
4. Results and discussions

In this section, the grid size sensitivity analysis is carried out firstly.
Then, the selected computational fluid dynamics models and combus­
tion models are verified by experimental data. After that, the effects of
different parameters on the energy consumption rate and emissions are
studied by using the response surface analysis. In addition, the param­
eter distributions on the Pareto front solutions are analyzed. Finally, the
cylinder pressures, heat release rates and spatial distributions of the
temperature and pollutant concentrations in the combustion process
with different diesel fuel ratios are analyzed.

4.1. Validation of the calculation method

Firstly, the effect of the grid cell size sensitivity on the combustion
process of the diesel/natural gas dual fuel engine is carried out. The
calculated cylinder pressures are compared in Fig. 3 using three different
Fig. 3. The calculated cylinder pressures comparison of different mesh sizes.
grid sizes, which are 0.10, 0.15 and 0.2 mm, respectively. When the grid

Fig. 2. The definition of the spray cone angle and the structural parameters of the combustion chamber.

5
J. Liu et al. Fuel 305 (2021) 121522

Fig. 4. The validation of the calculation and experiment: (a) pressures and the heat release rates (b) NO and CH4 emissions (c) the penetration length of vapor
and liquid.

Fig. 6. The Soot vs NO emissions of all population and Pareto front.

remarkably distributed in the region with low Soot and NO emissions,

Fig. 5. The calculation results of the optimization process.
while some of the Pareto front solutions have significantly high Soot or
NO emissions. In general, there is a trade-off relationship between Soot
problems. and NO emissions. The increase of Soot emission will lead to the
The Soot versus NO emissions of all the calculation cases and Pareto- decrease of NO emission, vice versa.
optimal front solutions are shown in Fig. 6. It is shown that as most of the The ISEC versus CH4 emission of all the calculation results and Pareto
fuel is natural gas consumed by premixed combustion, the Soot emission optimal front solutions are shown in Fig. 7. It exhibits that the CH4
is relatively low, most of which is below 0.1 g/kW⋅h. However, the NO emission and ISEC maintain a relatively consistent variation trend. The
emission is distributed in a wider range, with the peak value exceeding increase of CH4 emission will cause the ascending of the ISEC, and vice
30 g/kW⋅h. Furthermore, most of the Pareto front solutions are

6
J. Liu et al. Fuel 305 (2021) 121522

on the performance of the dual fuel engine, the response surfaces of the
ISEC, NO and CH4 emissions with respect to diesel fuel ratio versus in­
jection timing are shown in Fig. 8. For the ISEC contour map, the lower
ISEC region is located in the middle part of the selected diesel fuel ratio
and injection timing range, where the fuel ratio is in the range of 0.2 to
0.3 and injection timing is in the range of − 30 to − 15 ◦ CA ATDC. When
the fuel ratio is too low, less than 0.2, or the injection timing is too early
or too late, all of which will lead to the increase of ISEC. In addition,
when the fuel ratio is very low, less than or equal to 0.05, the ISEC will
also increase significantly. Furthermore, when the fuel ratio is high, the
excessive postpone of the injection timing, less than − 5 ◦ CA ATDC, will
also cause a significant increase in ISEC. Therefore, an ISEC limit line
can be found in the response surface diagram of the ISEC, as shown by
the bold dashed line in Fig. 8 (a).
For the NO emission response surface plot, it is shown that advancing
the fuel injection timing will increase NO emission, and vice versa. In
Fig. 7. The evolutions of the ISEC and CH4 emission on the Pareto front. addition, when the fuel ratio is large, the NO emission is relatively high,
as shown in Fig. 8 (b). There is a NO limit line in the region of high fuel
versa. The main reason is that the CH4 emission reflects the combustion ratio and early injection timing region.
efficiency of the dual fuel engine. Higher CH4 emission will reduce the The trend of the methane emission response surface is almost
combustion efficiency and result in an increase of the ISEC. Therefore, in opposite to that of the NO emission, and earlier injection timing can
order to obtain a lower ISEC, the premixed natural gas must be suffi­ significantly reduce the methane emission. In addition, the limit line of
ciently depleted. In addition, it is also displayed in the figure that some methane is located in the region where the fuel ratio is too low and the
ISECs of the Pareto front solutions are relatively high, which are even injection time is too late. Furthermore, when the fuel ratio is too low, the
more than 8.55 MJ/kW⋅h. However, almost all Pareto front solutions are methane emission will be significantly increased.
located in the lowest position of the ISEC under the same CH4 emission
conditions.
4.4. The optimized parameters on the Pareto front

4.3. Response surface of the calculation result The diesel fuel ratio versus injection timing on the Pareto-optimal
front of the final generations is shown in Fig. 9. It is shown that when
In order to further explore the effect of the fuel injection parameters the fuel ratio increases, the optimized injection timing is postponed and

Fig. 8. The response surfaces of diesel ratio VS. injection timing: (a) ISEC (b) NO emission (c) CH4 emission.

7
J. Liu et al. Fuel 305 (2021) 121522

differences are distributed in two bands, which are around − 0.4 mm and
− 1.0 mm, respectively. In addition, the compression ratios are also
converged to two values, which are 17.6 and 18.8, respectively.
The Pareto front solutions employ three EGR rates, which are around
30, 15 and 5, respectively, indicating the high, medium and low pro­
portion, as shown in Fig. 11. However, the swirl ratios are widely
distributed in the range of 0.5 to 1.2.

4.5. Effect of diesel fuel ratio on the combustion and emission

characteristics

In order to further analyze the pilot diesel fuel ratio on the com­
bustion and emission characteristics of the dual fuel engine, four cases
with different pilot diesel fuel ratios are selected for the comparison,
which are 0.053, 0.187, 0.291, 0.354, respectively. The simulation pa­
rameters and calculation results of the four conditions are given in
Fig. 9. The diesel fuel ratio VS. injection timing in the Last Generations of Table 5.
Pareto-optimal front. The radius differences of these four cases are around 0.5 cm or 1.0
cm, and the injection time is gradually delayed with the increase of the
fuel ratio. The spray cone angle is about 70◦ , the compression ratios are
about 17.7 and 18.8, and the swirl ratio is about 1.0. The EGR ratio is
lower for the small fuel ratio cases and higher for the large fuel ratio
condition.
The comparison of the pressure and heat release rate curves under
these four conditions is shown in Fig. 12. It can be seen from the figure
that although the injection timing of case 1 is the earliest, the peak
cylinder pressure is the lowest among these four conditions. Because
case 1 has the lowest amount of pilot diesel. Although the ignition
timing is the most advanced, the heat release rate at the initial stage is
extremely small, as shown in the heat release rate diagram Fig. 12 (b).
Furthermore, the main combustion stage of the premixed natural gas at
the later stage is the most backward of these four conditions, which
results in the lowest cylinder pressure. In comparison with case 1, case 2
has a larger quantity of pilot fuel, which leads to a higher heat release
rate in the initial stage of the diesel combustion. In addition, the ignition
Fig. 10. The radius difference VS. compression ratio in the Last Generations of timing of the premixed natural gas is the earliest, so the peak cylinder
Pareto-optimal front. pressure is the highest.
For case 3 and case 4, their injection timings are relatively consistent,
and the higher compression ratio is compensated with a higher EGR rate,
so their cylinder pressure curves are very close to each other. However,
the peak heat release rate of case 4 is higher than that of case 3, because
of its larger amount of pilot fuel.
In order to further investigate the pollutant generation process under
different pilot diesel fuel ratios, the temperature and pollutants distri­
bution in these four cases are given below.
The distribution of the temperature field is given in Fig. 13. It is
shown that the high temperature reaction zone already appears in case 1
and case 2 at 15 ◦ CA ATDC, mainly because the injection timings of these
two conditions are earlier than the other two cases. However, the high
temperature region of case 1 is mainly distributed in a certain region
downstream of the diesel jet plume, due to the small quantity of the pilot
fuel. However, case 2 has a larger injection mass and higher spray mo­
mentum, and when the fuel jet impinges with the edge of the combustion
chamber, it will move along the wall toward the top and bottom of the
combustion chamber. At the same time, affected by the swirl ratio, the
Fig. 11. The EGR rate VS. swirl ratio timing in the Last Generations of Pareto-
optimal front.
fuel jet moves circumferentially in the combustion chamber. Therefore,
the high temperature area of case 2 is widely distributed. When the
premixed natural gas in case 1 is ignited by the high temperature diesel
gradually approaches TDC. In addition, the injection timings are in the
combustion spot, the flame propagates in the premixed gas and its shape
range of − 20 to − 15 ◦ CA ATDC for the large fuel ratio conditions.
is close to spherical. However, under the other three conditions, after
The relationship between the radius difference and compression
premixed natural gas is ignited, part of the diesel fuel is not completely
ratio on the Pareto front is given in Fig. 10. The radius difference is
burned, which results in the coexistence of premixed flame propagation
defined as the maximum bottom radius (R1) minus the throat radius
and diffusion combustion. At the same time, a much higher temperature
(R2). It can be seen from the figure that all the radius differences are less
zone is formed in the diesel diffusion combustion region. In addition,
than zero, which indicates that the combustion chamber of all the Pareto
due to the decrease of the premixed equivalence ratio from case 1 to case
front solutions is the open bowl geometry. Furthermore, the radius
4, the premixed flame temperature decreases gradually.

8
J. Liu et al. Fuel 305 (2021) 121522

Table 5
The comparison of the parameters, emissions and ISEC under the four selected cases.
Case FR RD IT SA CR SR EGR rate NO CH4 ISEC
NO. % cm ◦
CA ◦
% g/kW⋅h g/kW⋅h MJ/kW⋅h

1 0.053 − 0.92 − 31.6 71.9 17.6 1.03 4.8 6.14 25.6 8.91
2 0.187 − 0.98 − 25.0 61.4 18.9 1.00 3.3 16.71 0.001 7.48
3 0.291 − 0.54 − 19.4 68.9 18.8 0.82 28.8 3.42 2.16 7.85
4 0.354 − 0.54 − 19.0 74.8 17.7 1.12 16.1 8.38 0.06 7.62

Fig. 12. The comparison under the four selected cases: (a) Cylinder pressures (b) Heat release rate.

Fig. 13. The comparison of temperature distribution under the four selected cases.

For the NO distribution area in Fig. 14., the formation of NO is is, the diesel diffusion combustion region. For case 1, although the diesel
mainly related to temperature, oxygen concentration and reaction time, injection quantity is the smallest, due to the highest equivalence ratio of
so NO emission is mainly generated in the high temperature region, that the premixed natural gas-air mixture, it also leads to a moderate NO

9
J. Liu et al. Fuel 305 (2021) 121522

Fig. 14. The comparison of NO emission distribution under the four selected cases.

Fig. 15. The comparison of CH4 emission distribution of temperature distribution under the four selected cases.

generation region. The generation region of NO was similar to the high- Furthermore, since case 2 has a large swirl ratio and kinetic energy, it
temperature region and mainly focused on the downstream of the diesel spreads to the periphery of the combustion chamber after the fuel jet hits
jet plume in a certain region. The early ignition timing and maximum the edge of the combustion chamber, resulting in the most extensive NO
compression ratio result in the highest NO generation in case 2. generating area. The late injection time of case 3 and case 4 resulted in a

10
J. Liu et al. Fuel 305 (2021) 121522

late ignition time and a smaller area of NO generation. The NO gener­ References
ation region mostly moves along the wall toward the top of the com­
bustion chamber. Furthermore, case 3 has the highest EGR rate, which [1] Oh S, Park C, Kim S, Kim Y, Choi Y, Kim C. Natural gas–ammonia dual-fuel
combustion in spark-ignited engine with various air–fuel ratios and split ratios of
leads to the lowest level of NO production. ammonia under part load condition. Fuel 2021;290:120095. https://doi.org/
The distribution of methane is shown in Fig. 15. Although the igni­ 10.1016/j.fuel.2020.120095.
tion timing of case 1 is the earliest, the flame propagation speed in the [2] Chen H, et al. NOx emission of biodiesel compared to diesel: Higher or lower? Appl
Therm Eng 2018;137:584–93.
premixed natural gas-air mixture is relatively low, which leads to the [3] Wang Z, Zhang F, Xia Ye, Wang D, Xu Y, Du G. Combustion phase of a diesel/
existence of a large amount of unburned methane in the piston squish, natural gas dual fuel engine under various pilot diesel injection timings. Fuel 2021;
piston crevice and fuel injector area. For case 3, some unburned methane 289:119869. https://doi.org/10.1016/j.fuel.2020.119869.
[4] Qiao Z, Lv Yu, Hickey J-P. Single-phase instability of intermediate flamelet states in
still exists in the piston crevice area. The main reason is that the high high-pressure combustion. Fuel 2021;288:119736. https://doi.org/10.1016/j.
proportion of EGR reduces the reactivity of the premixed gas, and the fuel.2020.119736.
high surface to volume ratio makes the temperature in this region [5] Uyumaz A, Aydoğan B, Calam A, Aksoy F, Yılmaz E. The effects of diisopropyl ether
on combustion, performance, emissions and operating range in a HCCI engine. Fuel
relatively lower. For case 2 and case 4, the lower EGR rate and higher
2020;265:116919. https://doi.org/10.1016/j.fuel.2019.116919.
injection quantity are beneficial to methane oxidation, so the methane [6] Chen H, Su X, He J, Xie B. Investigation on combustion and emission characteristics
emission is the lowest. of a common rail diesel engine fueled with diesel/n-pentanol/methanol blends.
Energy 2019;167:297–311.
[7] Mohammadian A, Chehrmonavari H, Kakaee A, Paykani A. Effect of injection
5. Conclusion: strategies on a single-fuel RCCI combustion fueled with isobutanol/isobutanol +
DTBP blends. Fuel 2020;278:118219. https://doi.org/10.1016/j.
In this study, the NSGAIII was coupled with the KIVA-3V code for the fuel.2020.118219.
[8] Wang Hu, DelVescovo D, Yao M, Reitz RD. Numerical study of RCCI and HCCI
optimization of the diesel fuel ratio and combustion parameters of a dual combustion processes using gasoline, diesel, iso-butanol and DTBP cetane
fuel engine. The four objectives are optimized simultaneously by improver. SAE Int J Engines 2015;8(2):831–45.
selecting different combinations of nine parameters. The main conclu­ [9] Li W, Liu Z, Wang Z, Dou H. Experimental and theoretical analysis of effects of
atomic, diatomic and polyatomic inert gases in air and EGR on mixture properties,
sions are summarized as follows: combustion, thermal efficiency and NOx emissions of a pilot-ignited NG engine.
Energy Convers Manage 2015;105:1082–95.
1. The lowest ISEC region is located in the middle part of the diesel fuel [10] Koli SR, Hanumantha Rao YV. Combustion analysis of single cylinder diesel engine
operated on dual fuel mode. Mater Today: Proc 2021. https://doi.org/10.1016/j.
ratio and injection timing region, where the fuel ratio is in the range matpr.2021.03.389.
of 0.2 to 0.3 and injection timing is in the range of − 30 to − 15 ◦ CA [11] Yousefi A, Guo H, Birouk M. Effect of diesel injection timing on the combustion of
ATDC. Advancing the fuel injection timing will increase NO emis­ natural gas/diesel dual-fuel engine at low-high load and low-high speed
conditions. Fuel 2019;235:838–46.
sion. In addition, earlier injection timing can significantly reduce
[12] Wang Z, Zhao Z, Wang D, Tan M, Han Y, Liu Z, et al. Impact of pilot diesel ignition
methane emission. mode on combustion and emissions characteristics of a diesel/natural gas dual fuel
2. For the optimized solutions, the diesel injection timing is gradually heavy-duty engine. Fuel 2016;167:248–56.
postponed with the increase of pilot fuel quantity. All the Pareto [13] Kim W, Park C, Bae C. Characterization of combustion process and emissions in a
natural gas/diesel dual-fuel compression-ignition engine. Fuel 2021;291:120043.
front solutions choose the open bowl geometry. In addition, the https://doi.org/10.1016/j.fuel.2020.120043.
compression ratio is also converged to 17.6 and 18.8, respectively. [14] Yousefi A, Guo H, Birouk M. An experimental and numerical study on diesel
Three EGR rates are selected in the Pareto front, which are 30, 15 and injection split of a natural gas/diesel dual-fuel engine at a low engine load. Fuel
2018;212:332–46.
5, respectively. However, the swirl ratio is widely distributed in the [15] Huang H, Zhu Z, Chen Y, Chen Y, Lv D, Zhu J, et al. Experimental and numerical
range of 0.5–1.2. study of multiple injection effects on combustion and emission characteristics of
3. When the pilot diesel fuel quantity is small, less than 0.2, the high natural gas–diesel dual-fuel engine. Energy Convers Manage 2019;183:84–96.
[16] Shu J, Fu J, Liu J, Ma Y, Wang S, Deng B, et al. Effects of injector spray angle on
temperature region is mainly concentrated in a certain region combustion and emissions characteristics of a natural gas (NG)-diesel dual fuel
downstream of the jet plume. However, a larger injection mass will engine based on CFD coupled with reduced chemical kinetic model. Appl Energy
promote the fluid momentum, which leads to the widely distributed 2019;233-234:182–95.
[17] Lee S, Kim C, Lee S, Lee J, Kim J. Diesel injector nozzle optimization for high CNG
high temperature region. substitution in a dual-fuel heavy-duty diesel engine. Fuel 2020;262:116607.
4. The flame propagation speed in the premixed natural gas-air mixture https://doi.org/10.1016/j.fuel.2019.116607.
is relatively slow, which leads to the existence of a large amount of [18] Vávra J, Bortel I, Takáts M, Diviš M. Emissions and performance of diesel–natural
gas dual-fuel engine operated with stoichiometric mixture. Fuel 2017;208:722–33.
unburned methane in the piston squish, piston crevice and fuel
[19] Zheng J, Wang J, Zhao Z, Wang D, Huang Z. Effect of equivalence ratio on
injector area. The lower EGR rate and higher injection quantity are combustion and emissions of a dual-fuel natural gas engine ignited with diesel.
beneficial to methane oxidation. Appl Therm Eng 2019;146:738–51.
[20] Lee C-F, Pang Y, Wu H, Nithyanandan K, Liu F. An optical investigation of
substitution rates on natural gas/diesel dual-fuel combustion in a diesel engine.
CRediT authorship contribution statement Appl Energy 2020;261:114455. https://doi.org/10.1016/j.apenergy.2019.114455.
[21] Geng L, Bi L, Li Qi, Chen H, Xie Y. Experimental study on spray characteristics,
Jie Liu: Conceptualization, Methodology, Software. Qiang Guo: combustion stability, and emission performance of a CRDI diesel engine operated
with biodiesel–ethanol blends. Energy Rep 2021;7:904–15.
Data curation, Writing - review & editing, Formal analysis. Jiayong [22] Wu Z, Han Z. Micro-GA optimization analysis of the effect of diesel injection
Guo: Writing - review & editing, Supervision. Fan Wang: Software, strategy on natural gas-diesel dual-fuel combustion. Fuel 2020;259:116288.
Validation. https://doi.org/10.1016/j.fuel.2019.116288.
[23] Liu J, Wang J, Zhao H. Optimization of the injection parameters and combustion
chamber geometries of a diesel/natural gas RCCI engine. Energy 2018;164:837–52.
Declaration of Competing Interest [24] Chen Y, et al. Verifying the wall-flow-guided assumption of the lateral swirl
combustion system in DI diesel engines. Fuel 2020;266:117079.
[25] Salek F, Babaie M, Hosseini SV, Bég OA. Multi-objective optimization of the engine
The authors declare that they have no known competing financial performance and emissions for a hydrogen/gasoline dual-fuel engine equipped
interests or personal relationships that could have appeared to influence with the port water injection system. Int J Hydrogen Energy 2021;46(17):
the work reported in this paper. 10535–47.
[26] Mattarelli E, Rinaldini CA, Golovitchev VI. CFD-3D analysis of a light duty dual
fuel (diesel/natural gas) combustion engine. Energy Proc 2014;45:929–37.
Acknowledgements [27] Beale JC, Reitz RD. Modeling spray atomization with the Kelvin-Helmholtz/
Rayleigh Taylor hybrid model. Atomization Sprays 1999.
[28] Reitz RD, Munnannur A. Comprehensive collision model for multidimensional
This study is supported by the Fundamental Research Funds for the
engine spray computations. Atomization Sprays 2009;19(7):597–619.
Central Universities (NO. 2021JBM020) and the National Key Research [29] Smith LM, Woodruff SL. Renormalization-Group Analysis of Turbulence. Annu Rev
and Development Program of China (NO. 2017YFB0103500). Fluid Mech 1998;30(1):275–310.

11
J. Liu et al. Fuel 305 (2021) 121522

[30] Golovitchev VI, et al. 3-D diesel spray simulations using a new detailed chemistry [34] Deb K, Jain H. An evolutionary many-objective optimization algorithm using
turbulent combustion model. Sae Technical Papers 2000. reference-point-based nondominated sorting approach, Part I: solving problems
[31] Chen W, Shuai S, Wang J. A soot formation embedded reduced reaction mechanism with box constraints. IEEE Trans Evol Comput 2014;18(4):577–601.
for diesel surrogate fuel. Fuel 2009;88(10):1927–36. [35] Bajaj C, Abraham J, Pickett LM. The role of vaporization in determining transient
[32] Xu F, Lin K-C, Faeth GM. Soot formation in laminar premixed methane/oxygen diesel spray structure. In Proceedings of the 23rd Annual Conference on Liquid
flames at atmospheric pressure. Combust Flame 1998;115(1-2):195–209. Atomization and Spray Systems, Ventura, CA; 2011.
[33] Yang J, et al. Numerical Analysis of NOx Formation Trends in Biodiesel
Combustion using Dynamic -T Parametric Maps. 2011 JSAE/SAE International
Powertrains, Fuels & Lubricants Meeting, Kyoto (Japan), Aug 29-Sep 2; 2011.

12