natural gas RCCI engine

Energy 164 (2018) 837e852

Contents lists available at ScienceDirect

Energy journal homepage: www.elsevier.com/locate/energy

Optimization of the injection parameters and combustion chamber geometries of a diesel/natural gas RCCI engine Jie Liu a, b, *, Junle Wang a, b, Hongbo Zhao a, b a b

Department of Power Mechanical Engineering, Beijing Jiaotong University, Beijing 100044, PR China 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

Article history: Received 28 April 2018 Received in revised form 5 September 2018 Accepted 8 September 2018 Available online 9 September 2018

This study aims at finding the favorable combinations of the diesel injection parameters and combustion chamber shape of a diesel/natural gas dual fuel engine to achieve lower fuel consumption and pollution emissions. The genetic algorithm NSGA-II coupled with the KIVA-3V code was employed for the multiobjective optimizations. The results show that the straight combustion chamber is effective to reduce the CH4 emission and improve the fuel economy, and the indicated thermal efficiency reaches 50.2% with an injection timing of
16.45  CA ATDC. The NO emissions are the lowest when the reentrant-type combustion chamber is used, while CH4 emissions and ISFC are higher than the other two types of the combustion chamber. When the injection timing is far away from the top dead center, the NO emissions, CH4 emissions and ISFC are very close to each other for the three types of the combustion chamber. Slightly narrow spray angle is favorable to improve the performance of the dual fuel engine. © 2018 Elsevier Ltd. All rights reserved.

Keywords: Genetic algorithm Dual fuel Natural gas Combustion chamber

1. Introduction The diesel engine has the highest thermal efficiency due to its very high compression ratio, which makes the operating cost lower than any other practical internal combustion engine. This important feature makes it the most preferred engine especially for the heavy-duty vehicles. However, the NOx and Soot emissions from diesel engines are very high due to its combustion characteristics. Many researches have been done to investigate the method of reducing the NOx and soot simultaneously in diesel engines. New combustion concepts have been proposed for diesel engine to meet the increasingly stringent emission regulations, such as Homogeneous Charge Compression Ignition (HCCI), Low Temperature Combustion (LTC) and Premixed Charge Compression Ignition (PCCI) [1e3]. Among them, the Reactivity Controlled Compression Ignition (RCCI) concept is the most promising method to be applied for diesel engine [4]. For the RCCI combustion strategy, the low reactivity fuel was introduced from port injection to form a homogeneous mixture in the cylinder, and the high cetane number (CN) fuel was injected directly into the cylinder to control the combustion phasing and

* Corresponding author. Department of Power Mechanical Engineering, Beijing Jiaotong University, Beijing 100044, PR China. E-mail address: [email protected] (J. Liu). https://doi.org/10.1016/j.energy.2018.09.064 0360-5442/© 2018 Elsevier Ltd. All rights reserved.

duration. High octane number (ON) fuel is favorable for the premixed fuel, because it has higher resistance to the spontaneous ignition, which is benefit to extend the upper load limit of the dual fuel engine. Therefore, natural gas with high ON is the best choice for dual fuel combustion application. Yousefi et al. studied the effect of diesel injection timings on the combustion performance and emissions of a heavy duty natural gas/diesel dual-fuel engine at 25% engine load. Both experimental and numerical results revealed that advancing the injection timing up to 30  CA BTDC increases the maximum in-cylinder pressure [5]. Rahnama et al. investigated the effect of reformer gas (syngas) composition on the performance and exhaust emissions properties of a natural gas/diesel RCCI engine at low loads numerically. The results indicated that reformer gas addition could enhance the combustion efficiency and decrease CO emission [6,7]. Wang et al. investigated the diesel injection timing on diesel/natural gas dual fuel ignition mode. With advancing diesel injection timing, engine combustion and emissions characteristics, including cylinder pressure, cylinder temperature, heat release rate, start of combustion (SOC), ignition delay, combustion duration, crank angle of 50% heat release (CA50), nitrogen oxides (NOx) and total hydrocarbon (THC), show completely different variation trends in different ignition modes [8]. Papagiannakis et al. revealed that for the examined test engine operating under constant natural gas/diesel mass ratio, a restricted increase in the diesel fuel injection timing

838

J. Liu et al. / Energy 164 (2018) 837e852

could be a promising solution for engine efficiency improvement and CO emission mitigation [9]. Poorghasemi et al. investigated the effects of several parameters, including the premixed ratio of NG, diesel fuel fraction in first and second injection pulses, first and second start of injection timing, injection pressure and the spray angle on the engine performance and emission characteristics. The results indicated that these parameters have significant effects on the light duty RCCI engine performance and engine out emissions [10]. Li et al. showed that the increase of time-sequenced coefficient and heat release rate-balanced coefficient can decrease HC (hydrocarbon) emissions and improve the BTE (brake thermal efficiency) significantly [11]. Song et al. showed that using lowpressure direct injection (LPDI) systems could be an alternative for current high-pressure common rail injection systems, which would significantly reduce the system cost. And larger advanced injection timing was used to realize low temperature combustion and achieve long ignition delay in order to counteract the negative impact of relatively poor atomization quality caused by the low injection pressure [12]. The study of Yang et al. indicated that the particle number concentration (in particular to the ultrafine particle) could be reduced dramatically with increased pilot injection pressure and the percentage energy substitution as well as engine loads [13]. The combustion chamber geometry plays a critical role in the combustion process of the diesel engines. The combustion of a heavy-duty diesel engine was optimized by coupling a multidimensional computational fluid dynamics (CFD) code with genetic algorithm (GA). At each load, a comprehensive optimization of the operating parameters was conducted in order to simultaneously minimize ISFC (indicated specific fuel consumption), NOx (nitrogen oxides) and soot emissions [14]. Jung et al. investigated the fundamentals of dual-fuel combustion and the effects of intake valve closure (IVC) changes in dual-fuel mode using a 1D engine simulation. It is shown that IVC could increase combustion efficiency and affect NOx emissions by controlling the Air/Fuel ratio [15]. The optimization processes of a diesel engine fueled with DME were performed by Park et al. [16], which is based on a microgenetic algorithm with a population number of five for each generation. In addition, the computational mesh was generated by an auto-mesh generator, which was able to produce a computational mesh based on the given parameters, such as cup depth and Beizer curve definitions. The optimization of the combustion chamber geometry for natural gas engines was carried out by Wang et al. using the multi-objective non-dominated sorting genetic algorithm II (NSGA-II) coupled with Kriging-based meta-model [17]. To generate the various combustion chamber geometries, the bowl outline is parameterized by the two cubic Bezier curves while keeping the compression ratio constant. With the optimization, the HC and CO emissions are reduced by 56.47% and 33.55%, respectively. The piston bowl geometry and the operating conditions of a dual-fuel engine were optimized by Lee et al. [18]. As a result of optimization, a 9% improvement in the gross indicated specific fuel consumption and a simultaneous decrease of the overall NOx and soot emissions was achieved. The baseline case has a re-entrant shape, while the optimized case has a shallow shape and a narrower spray angle. The orthogonal design method and multiobjective NLPQL algorithm were employed by Chen et al. to optimize the combustion chamber of DI diesel engine [19]. The combination of the Computational Fluid Dynamics (CFD) modeling and the statistical Design of Experiments (DOE) technique, known as Response Surface Method (RSM) was utilized by Benajes et al. for optimizing the combustion system of Compression Ignition (CI) engines [20]. The genetic algorithm (GA) method was used by Kim et al. to optimize the operating conditions of diesel and dimethylether (DME) fuel in a diesel engine [21]. And different variables of

the operating conditions were determined for the analysis of the optimization conditions. The optimization study of Kavuri et al. showed that an optimum CR of 13.1 with a bowl geometry that has two distinctive regions benefit the low load and high load operating conditions, respectively [22]. Results also showed that a narrow spray angle for diesel fuel and a wide spray angle for gasoline would be necessary to target the two different regions in the bowl. So far, few researches focus on the effect of combinations of the diesel injection parameters and combustion chamber shape on the performance and exhaust gas emissions of the duel fuel natural gas engine. The objectives of the present study are to optimize the combination of injection parameters and the combustion chamber geometries of a diesel/natural gas dual fuel engine to achieve low fuel consumption and low pollution emissions. The genetic algorithm coupled with the KIVA-3V code was employed for the multiobjective optimizations. Two pilot injection parameters (pilot diesel spray angle and start of injection) and three combustion chamber parameters (central pip height, throat radius and maximum bottom radius of the combustion chamber) were optimized simultaneously. In addition, an in-house developed auto mesh generator was used to generate the combustion chamber mesh automatically in order to meet the requirements of the genetic algorithm. 2. Experimental apparatus and setup In this study, a modified WEICHAI WP10 heavy-duty diesel engine was used for the experiments. The exhaust emissions were analyzed by the Horiba MEXA7100DEGR analyzer. The diesel and natural gas fuels were obtained from the local distribution network in Beijing City. 2.1. Test engine The experiments were carried out on a WEICHAI WP10 heavyduty diesel engine, which is a six-cylinder turbo charged engine with common rail system. The summary of the engine's specification can be found in Table 1. The injection timing and quantity of the diesel and the natural gas were controlled by an in-house developed dual-fuel control unit. 2.2. Instrumentation The Dynamometer used in this study is the Horiba-Schenck HT350 AC Transient Dynamometer with relevant signal acquisition and data logging module. Cylinder pressure signal was acquired by a mass-produced piezoresistive cylinder pressure sensor in conjunction with a charge amplifier (Kistler 5018). The Horiba MEXA7100DEGR analyzer was used for the measurement of the engine out emissions. Nitrogen oxides (NOx)

Table 1 Summary of engine specification. Bore  Stroke Number of cylinders Displacement Maximum torque/speed Rated power/speed Compression ratio Number of Injector nozzle holes Injector nozzle spray angle Exhaust valve closing timing Exhaust valve opening timing Inlet valve closing timing Inlet valve opening timing

126  130 mm 6 9.726 L 1250 N$m/1200～1600 rpm 247kW/2200 rpm 17.0 7 146 21 CA ATDC 131 CA ATDC 146 CA BTDC 20 CA BTDC

J. Liu et al. / Energy 164 (2018) 837e852

emission was analyzed with the chemiluminescent detector (CLD), total hydrocarbon (THC) was detected by the flame ionization detector (FID) and carbon monoxide (CO) was measured by the nondispersive infrared analyzer (NDIR), respectively. The diesel fuel mass flow rate was measured by an ONOSOKKO FZ2100 coriolis mass flow meter. The CNG flow rate was measured by a BROOKS gas flow meter. 2.3. Test fuels The engine was fueled with the commercial 0# diesel fuel and natural gas obtained from the local distribution network of Beijing City. The detailed specifications of these two fuels are listed in Table 2. 3. CFD package and sub-models In this study, the KIVA-3V code was used to simulate the combustion and emission characteristics of a diesel/natural gas RCCI engine. The compound combustion model was developed by using a turbulent flame propagation model coupled with the original PSR combustion model. The Non-dominated Sorting Genetic Algorithm NSGA-II was used for the optimization calculations. An in-house auto mesh generation code was developed in order to meet the needs of automatic optimization algorithm.

839

butane. A well-validated reduced reaction mechanism for diesel/ natural gas oxidation, with 81 species and 421 elementary reactions with associated rate coefficient expressions, was used in this study [26]. Soot formation and oxidation were taken care of by Hiroyasu soot formation and Nagle/Strickland-Constable oxidation models [27]. Furthermore, HACA/surface reactions were included in the mechanism. The extended Zel'dovich NOx formation mechanism was implemented in the fuel oxidation mechanism to simulate the prompt NO formation [25].

3.2. Combustion model The turbulent flame propagation model was developed and it was coupled with the partially premixed reactor combustion model to account for the premixed natural gas combustion and spray combustion process of the diesel, respectively. The dual fuel combustion model can be expressed by the following mass balance equation, in which the spray source term is omitted:


  · c1 · c2 vcm cm þ V$ðcm uÞ ¼ V$ rDV þ rm þ rm vt r

(1)

Where the chemical sources of the turbulent flame propagation model and the partially premixed reactor model are represented by · c1

· c2

rm and rm , and cm is the mass fraction of specie m. 3.1. Sub-models of the KIVA code

With the introduction of the characteristic time of the different combustion models, it follows that

The spray and combustion processes were simulated with the KIVA-3V code, which is based on Lagrangian-Drop Eulerian-Fluid (LDEF) approach. The injected liquid fuel was treated as Lagrangian parcels, which were introduced with the blob injection model, and the surrounding gas phase was spatially discretized into cells. The primary break-up of the spray droplet was simulated by the KelvinHelmholtz (KH) instability theory, which stands for stripping of liquid drops from the blob bulk through shear stresses. And the secondary break-up process occurred based on the combination effects of the Kelvin- Helmholtz (KH) and Rayleigh-Taylor (RT) instabilities [23]. The droplet collision process was resolved by the O'Rourke model and the radius-of-influence (ROI) method was used to reduce the grid dependency [24]. The turbulence was calculated based on the RNG k-ε method. And the Turbulence/ chemistry interaction was simulated with the Partially Stirred Reactor combustion model [25]. The diesel fuel was represented by a blend of two hydrocarbons, namely n-heptane and toluene. Both the physical properties and ignition/combustion characteristics of a typical diesel were well represented by a blend of these two components with the ratio of 0.7:0.3. Furthermore, natural gas was represented by several hydrocarbon components, which are methane, ethane, propane and

· c1

cm

tc1

· c2

; rm z

· c2

rm þ rm ¼

cm

tc2

cm

(2)

tc

And the overall characteristic time of the dual fuel combustion mode is introduced as follows:

tc ¼

tc1 tc2 tc1 þ tc2

(3)

This is an accurate “interpolation” between the two combustion models avoiding an inaccuracy of the approach based on the estimation. The laminar flame propagation velocity Sl was calculated by the PREMIX code. And fitted by a polynomial expression. And the reference condition is chosen as: P0 ¼ 50 atm, T0 ¼ 450 K. The details of the TFSC model can be found in Ref. [28], which employs the well-known Bray-Moss method and characterizes the combustion processed through a combustion progress variable c. The conservation equation governing the flame propagation is:

. v ðr~cÞ þ V$ðr~cuÞ ¼ V$½rDV~c þ ru St jV~cj þ rð1
~cÞ tf vt

Table 2 Fuel properties. CNG Methane Ethane Propane Butane Iso-pentane Oxygen Nitrogen Carbon dioxide Lower heat value Stoichiometric air-fuel ratio

· c1

rm z

0# diesel 96.51% v/v 1.2% v/v 0.18% v/v 0.04% v/v 0.01% v/v 0.01% v/v 0.22% v/v 1.81% v/v 50.9 MJ/kg 16.88 kg/kg

Cetane number Density Lower heat value Stoichiometric air-fuel ratio

52.6 833.7 kg/m3 42.74 MJ/kg 14.5 kg/kg

(4)

The subscript u refers to the density of the unburned mixture. St is the turbulent flame propagation velocity, which is assumed to depend on the laminar speed and characteristics of the flow turbulence, e.g., u0  k1=2 [29].

St  u03=4 Sl

1=4

 1=4 l

dl

(5)

The combustion progress variable c is increased monotonically from c ¼ 0 in the unburned gas to c ¼ 1 in the equilibrium products. And it is initiated by the last term in equation (1).

840

J. Liu et al. / Energy 164 (2018) 837e852

3.3. Optimization method based on genetic algorithm Most engineering optimization problems are multi-objective in nature, since normally there are several objectives that must be considered at the same time. The multi-objective evolutionary algorithms (MOEAs), which can be used to obtain a well-converged and well-distributed approximation of the Pareto optimal solutions, have been gaining more and more attention. Different MOEAs have been developed by the researchers. Among them, NSGA-II is one of the most promising methods to find the optimal solutions. NSGA-II is the second generation of the famous “Non-dominated Sorting Genetic Algorithm” based on the work of Prof. Kalyanmoy Deb for solving non-convex and non-smooth single and multi-objective optimization problems [30]. The first nondominated solutions were selected based on the results of the first two generations, and they are set as the first front (rank 1). The other fronts were found based on the solutions besides in the former front, as shown in Fig. 3. After that, the solutions of the new generation were calculated, and they were mixed with the solutions in the former first front forming the new solution set. And the new first front was generated based on the new solution set. The solution with the lower rank is preferred when comparing two solutions with different non-domination ranks. Furthermore, if both solutions are belonging to the same rank, the solution with the longer crowding distance is preferred, where the crowding distance of the solution is the average side-length of the cuboid in its front, as shown in Fig. 4 with a dashed box. Where, f1 and f2 are the objectives of the optimization. In order to optimize the injection and combustion chamber geometry parameters, the KIVA-3V code was coupled with the NSGA-II code. Where the solutions of the KIVA code were provided to the NAGA-II code and the parameters generated by the NSGA-II were sent to the KIVA code. The overall coupling method of these two codes is illustrated in Fig. 5. First, the fuel injection parameters and combustion chamber geometry parameters for the first generation are generated by the NSGA-II code based on the selected parameter range. Then the fuel injection parameters will be passed

to KIVA-3V code as the initial conditions. And the combustion chamber geometry parameters will be passed to the auto mesh generation code for the mesh generation. Second, the 3D combustion process in the dual fuel engine will be calculated by using the new generated injection parameters and new combustion chamber mesh. And the simulation results will be passed to the NSGA-II code. Third, the NSGA-II will evaluate these results, and conduct the selection, crossover, and mutation operation to generate the fuel injection parameters and combustion chamber geometry parameters for the next generation. And these parameters will again be passed to KIVA and auto mesh generation code as the input parameters. This calculation process will be repeated until the maximum generation (40) is achieved. And the optimized solutions can be found in the Pareto front. During the optimization process, the engine speed was fixed at 1320 rpm. The quantities of the diesel and the premixed natural gas fuel were the same as that of the baseline case. Two injection parameters with crucial influence on the engine performance including the injection time and spray angle of the diesel fuel, and three combustion chamber geometry parameters of the central pip height, throat radius and maximum bottom radius were selected as the variables to be optimized by the NSGA-II code. The ranges chosen for the optimization process of these five parameters are listed in Table 3. Three types of combustion chamber, including open combustion chamber, straight combustion chamber and reentrant combustion chamber, can be generated from the selected parameter ranges. Therefore, the favorable combination of the combustion chamber shape and injection parameters can be obtained. In the optimization process, the unreasonable cases were excluded through the introduction of several constraints. The limit of the peak in-cylinder pressure was set as 20 MPa for the robustness consideration, and the maximum pressure rise rate (PRR) was limited under 2 MPa/ CA in order to avoid engine knock. Furthermore, the maximum acceptable indicate specific fuel consumption (ISFC) was set to be 250 g/kW$h in order to prohibit over low thermal efficiency, where the natural gas consumption quantity

Fig. 1. Parameters of the original combustion chamber.

J. Liu et al. / Energy 164 (2018) 837e852

841

Fig. 2. Parameters definition and control points used to generate the chamber geometries.

Fig. 3. Solution rank determination.

was transformed to diesel consumption quantity. The fitness of all the objectives will be set to a small number in the case when one of the constraints was exceeded in order to avoid selecting it as the Pareto front. In addition, the cases with misfire conditions were also excluded in the Pareto front selections [14]. 3.4. Auto mesh generation method In general, the mesh of the combustion chamber was generated

before the calculation of the spray and combustion processes, and it does not need to be changed during the calculation. However, in the genetic algorithm optimization process, there are hundreds of combinations of the geometric parameters, it is impossible to generate the combustion chamber mesh by manual operation in each case. For this reason, an in-house auto mesh generation code was developed, which can adjust and resize the combustion chamber geometry based on the input parameters. The basic idea is to generate the mesh with some specified control points, and the

842

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 4. The cuboid for the crowding distance calcuation.

Fig. 5. The overall computational process of the optimization.

Table 3 Injection and combustion chamber geometry parameters and ranges. parameter

Min.

Max.

Diesel injection timing/ CA BTDC Diesel spray angle/ Central pip height/cm Throat radius/cm Maximum bottom radius/cm


30 90 0.5 3.5 3.3

0.0 160 1.525 4.5 4.8

chamber contour profile is obtained through the lines between these points. These lines may be a straight line or a curve, and they are connected at each control point with the restrictions in order to make the entire contour smooth. Fig. 1 gives the parameters of the original combustion chamber. Fig. 2 shows the parameters definition and control points used to generate the chamber geometries.

In order to increase the computational efficiency, a sector mesh encompassing 1/7th of the combustion chamber with periodic boundaries is used for the spray and combustion simulations. Results are presented from center plane of the sector. The simulations are running from 146 crank angle degrees before top dead center (intake valve close time) to 120 crank angle degrees after top dead center. Chemistry oxidation calculations are called when the cell temperature is over than 700 K. During the optimization process, the compression ratio was maintained the same as the baseline. 4. Results and discussions In this section the dual fuel combustion model will be validated first. Then the optimization results of the spray parameters and combustion chamber geometries will be analyzed. After that, the optimized parameters in the Pareto front will be given, and the

J. Liu et al. / Energy 164 (2018) 837e852

conditions for achieving low pollutant emissions and high combustion efficiency for diesel/natural dual fuel engine are obtained. 4.1. Validation of the dual fuel combustion model

k 1 pdV þ Vdp k
1 k
1

gas dual fuel engine. In the following study, the 0.15 mm cell size will be used for the optimization calculation. 4.2. Parameters optimization results

A study on the grid dependence is first conducted. Three combustion chamber meshes with the size of 0.10, 0.15 and 0.2 mm are used to study the effect of cell size on the combustion characteristics of the dual fuel engine. As shown in Fig. 6, there are almost no difference between the cylinder pressure and the heat release rate calculated by the three meshes. And the all the calculation results are in good agreement with the experimental data. There are some errors in the calculation of the heat release rate, which are mainly from the following aspects: Firstly, an L shape channel was used for the pressure sensor installation, which will generate pressure oscillation for the pressure measurement. Secondly, a 7-points average method was used to smooth the pressure curve, which will introduce some error to the cylinder pressure curve. Thirdly, in the heat release rate calculation process, a simple cylinder pressure based method was used, as shown in equation (6). The heat loss through the wall of the combustion chamber was not included in the calculation. This method will also bring some error to the heat release rate.

dQB ¼

843

(6)

Where QB is the heat generation, p is the cylinder pressure, V is the cylinder volume and k is the adiabatic exponent. The calculated NO, CH4 and CO emissions are compared with the experimental results in Fig. 7. The predicted NO emission is in good agreement with the experimental data. Furthermore, only slightly difference is found between the predicted CH4 and CO emissions and that of the experimental result. The above comparison results indicate that the models selected in this study are capable to simulate the combustion and emission processes in diesel/natural

A total number of 210 cases were calculated during the optimization process. The solutions of all the populations (circles) and the optimized solutions (solid circles) are given in Fig. 8. As there are three objectives to be optimized, only the solutions with nodominated by the other three objectives are chosen as the preferred solutions. The 3D result indicates that the values of the NO emission, CH4 emission and ISFC are widely distributed as they cover a very wide range. The maximum and minimum values of the NO emission, Soot emission and ISFC in the optimized solutions are very close to that of all the populations, which indicates that the optimized solutions considered all the possible optimal conditions. There exists a trade-off relation between the ISFC and NO emissions as shown in Fig. 9. With the decrease of ISFC, an increase trend was found with the NO emissions. The upper left area is the high ISFC region, where the maximum ISFC is around 230.0 g/ kWh, which is obviously higher than the original case with an ISFC of 188.12 g/kWh. However, the NO emissions in this region are significantly lower than that of the original case, which is only 1.75 g/kWh. And two cases in this area are selected in the Pareto front. The downright area is the low ISFC region, the lowest ISFC is about 170.7 g/kWh, which is equivalent to an indicated thermal efficiency of 50.2%. And three cases within this area are chosen in the Pareto front, as shown in the right bottom area in Fig. 9. The trends of the CH4 emissions versus NO emissions are similar with the ISFC versus NO emissions, as shown in Fig. 10. The reduction of CH4 emission will result in an increase of the NO emission. It should be noted that the lowest CH4 emissions of the calculated cases are closed to zero, which means almost all the premixed CH4 was consumed during the combustion process. And these cases are favorable to reduce the fuel consumptions. Because

Fig. 6. Comparisons of the cylinder pressure and heat release rate between calculations and experiment.

844

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 7. Comparisons of the NO, CH4 and CO emissions between calculations and experiment.

Fig. 8. All cases and Pareto solutions.

J. Liu et al. / Energy 164 (2018) 837e852

845

Fig. 9. ISFC and NO emissions of all cases and Pareto solutions.

in these cases, not only the CH4 emissions are the lowest, but also the ISFCs are very low. As shown in Fig. 11, the reduction of CH4 can simultaneously reduce the ISFC. 4.3. Effect of combustion chamber shape on the emissions and ISFC Three types of combustion chamber can be generated based on the difference between the maximum bottom radius and the throat radius. If the maximum bottom radius is equal to the throat radius, a straight combustion chamber is generated. If the throat radius is larger than the maximum bottom radius, it will be an open

Fig. 10. NO and CH4 emissions of all cases and Pareto solutions.

combustion chamber. Otherwise, it is a re-entrant combustion chamber. In this study, the radius difference is expanded from 0 mm to ±0.5 mm for the straight combustion chamber, because it is hard to tell the difference when the radius difference is too small. Fig. 12 gives the difference between the maximum bottom radius and throat radius in the Pareto front at each generation. At the initial stage of the optimization process, both re-entrant combustion chamber and open combustion chamber are dominated in the Pareto front. However, with the increase of the generation, the optimized radius difference of the open combustion chamber getting smaller and smaller. And finally, the open combustion chamber degenerated into the straight combustion chamber at around the 30th generation. After that, only the re-entrant combustion chamber and the straight combustion chamber are appeared in the Pareto front. The effect of combustion chamber types on the emissions and ISFC are shown in Figs. 13e15. With the advance of the diesel injection timing, NO emission is increased with all the three types of the combustion chambers. The NO emission is the lowest when the re-entrant combustion chamber is used, while the NO emission is the highest when the open combustion chamber is used. A reverse trend is found for the CH4 emission. With the advance of the diesel injection timing, CH4 emission is reduced with all the three types of the combustion chambers. The CH4 emission is the highest when the re-entrant combustion chamber is used, while the CH4 emission is the lowest when the open combustion chamber or straight combustion chamber is used. The smallest ISFC is obtained when using the straight combustion with the diesel injection timing is
16.45  CA ATDC. Advancing or retreating the diesel injection timing will both increase the ISFC. It should be noted that when the diesel injection timing is further advanced to around
25.00  CA ATDC, the NO emissions, CH4 emissions and ISFC are very close to

846

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 11. ISFC and CH4 emissions of all cases and Pareto solutions.

Fig. 12. Variations of the radius difference at each generation.

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 13. Effect of the combustion chamber shape on the NO emissions.

Fig. 14. Effect of the combustion chamber shape on the CH4 emissions.

847

848

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 15. Effect of the combustion chamber shape on the ISFC.

each other for the three types of the combustion chamber. The results show that the re-entrant combustion chamber has advantages on the NO emissions compared with the other two combustion chambers. And the open combustion chamber and straight combustion chamber have advantages on the CH4 emissions and ISFC compared with the re-entrant combustion chambers. Moreover, the straight combustion chamber has slightly advantage compared with the open combustion chamber. In the dual fuel engine, the main emissions are NOx and unburned CH4. NOx emissions are mainly generated in the high temperature region (over than 2000 K). ISFC is mainly dependent on the combustion speed (combustion duration) and the combustion efficiency (depend on the THC, which is mainly composed of methane). The natural gas fuel is premixed with the air to form the combustible mixture. The equivalence ratio of this mixture is much less than unity, which will lead to a low combustion temperature and less NOx is formed in the premixed mixture. Therefore, NOx emissions are mainly dependent on the injection strategies of the pilot diesel fuel. HC emission is mostly due to the retention of unburned fuel in the squish region and piston crevices in the cylinder. Where the flame quenching will occur in these regions as the higher surfaceto-volume ratio increase the heat transfer and reduce the mixture temperature. The main method to achieve high thermal efficiency and low HC emission is to burn the fuel in the squish region completely. As the throat radius is less than the maximum bottom radius for the re-entrant combustion chamber, the main combustion processes are confined in the combustion bowl region. However, for the straight combustion chamber or the open combustion chamber, the throat radius is larger, which will promote the squish airflow into the combustion chamber and combustion gas out of the combustion chamber. Therefore, the low temperature mixture

in the squish region will blend with the higher temperature mixture in the combustion bowl region sufficiently and the mixture will burn completely, which result in higher cylinder temperature. In this case, more NOx will be generated. 4.4. The optimized parameters in Pareto front The optimized five parameters in the six Pareto front cases are given in Fig. 16. For the maximum bottom radius and throat radius, case 1 and case 6 are both 4.09 mm and 3.88 mm, respectively. Both of them are the re-entrant combustion chambers, which are similar with the original case. For cases 2 to 5, the maximum bottom radius and the throat radius are approximately equal to each other with three are around 3.63 mm and one is around 3.85 mm. These four combustion chambers are the straight combustion chambers. The central pip heights of the six optimized cases are all smaller than the original case: three cases have a value of 1.38 mm, two cases are around 1.14 mm and one is 0.82 mm. The optimized injection timing is chosen at
16.45  CA ATDC with three cases, one is around the TDC, one is close with the original case and one is far advanced to
24  CA ATDC. Five optimized spray angles are around 116 , the other one is around 136 , and all of them are smaller than the original case whose spay angle is 146 . Therefore, narrow spray angle is preferred to increase the performance of the dual fuel engine. The peak cylinder pressure, peak pressure raise rates, ISFC, CH4 emissions and NO emissions of the optimized cases are shown in Fig. 17. The case with higher peak cylinder pressure has the higher pressure raise rate and higher NO emissions at the same time. However, the ISFC and CH4 emissions have the opposite trend, where the lower value is found with the higher peak cylinder pressure case. That is because the cylinder temperature is higher for

J. Liu et al. / Energy 164 (2018) 837e852

849

Fig. 16. Distributions of design parameters of the optimal cases.

the higher peak cylinder pressure case, which is beneficial for the NO generation and CH4 oxidation. And more work will be done when the cylinder pressure is higher, which can reduce the ISFC. The throat radius is the minimum distance between the combustion chamber edges and the cylinder center around the piston top face, which has a tremendous influence on the squish airflow into the combustion chamber and combustion gas out of the combustion chamber. The maximum bottom radius is the largest radius at the bottom of the combustion chamber, which determiners magnitude of the re-entrant angle, total chamber volume and the compression ratio. The central pip is located at the center of the combustion chamber, where the velocity of the air is low. The central pip allows a higher airflow velocity and better fuel/air mixing in the center of the swirling flow field. The profiles of the combustion chamber and the related parameters of the six optimized cases are shown in Fig. 18. It is shown that the ISFC can be reduced obviously when using the straight combustion chamber compared with using the re-entrant combustion chamber, except for case 3, whose injection timing is near the top dead center. Moreover, the CH4 emissions are significantly lower when using the straight combustion chamber compared with using the re-entrant combustion chamber. In addition, the optimized injection timing for the lowest ISFC case is
16.45  CA ATDC. However, with further advancing of the injection timing to
24  CA ATDC, the ISFC will be increased from 170.72 g/kW$h to 176.15 g/ kW$h. In the next part, the spatial distribution of the temperature and CH4 will be compared with the lowest ISFC case (case 2) and the original case, as they are using the straight combustion chamber and re-entrant combustion chamber, respectively.

4.5. The spatial distributions of the temperature and CH4 mole fraction The spatial distributions of the temperature and CH4 mole

fraction of the original and the optimized case 2 are compared in Fig. 19. The combustion of the premixed natural gas starts earlier in the optimized case 2 compared with the original case, because the diesel fuel injection timing is advanced. At 10  CA ATDC, the high temperature areas spread throughout the combustion chamber and most of the premixed CH4 is consumed in the optimized case 2. However, some low temperature regions are still existed in the original case and plenty of CH4 is not consumed. With the piston going down, the temperature in the cylinder decreases and part of the CH4 cannot be oxidized in the piston gap region, which will lead to high CH4 emissions and high ISFC. 4.6. Comparisons of the flow field in different combustion chambers For the straight combustion chamber, the mixture is compressed into the piston gap at the end of the compression process, as shown in Fig. 20. However, a squish flow directed into the combustion chamber is generated for the re-entrant combustion chamber at the end of the compression process. At the beginning of the expansion process (10  CA ATDC), the mixture in the squish region flows back into the combustion chamber for the straight combustion chamber. In contrast, a turbulent flow back into the piston gap is generated for the re-entrant combustion chamber. At 20  CA ATDC, the flow directions around the piston gap reverse again for both of the combustion chambers. The directions of the vortices generated at the top center and bottom of the combustion chamber are anticlockwise and clockwise, respectively. And the directions of these vortices in the combustion chamber are not changed during the compression and expansion process for both of the combustion chambers. 5. Conclusion In this study, the combinations of the diesel injection

850

J. Liu et al. / Energy 164 (2018) 837e852

Fig. 17. Distributions of NO and CH4 emission, Peak pressure, maximum PRR and ISFC of the optimal cases.

Fig. 18. The optimized combustion chamber geometries and associated parameters.

parameters and combustion chamber shape of a diesel/natural gas dual fuel engine were optimized to achieve lower fuel consumption and pollution emissions. The genetic algorithm NSGA-II coupled with the KIVA-3V code was employed for the multi-objective optimizations. The main findings are as follows: 1. The reduction of CH4 emission can simultaneously reduce the ISFC, while the NO emission will be increased. And the lowest CH4 emissions of the calculated cases are closed to zero, which means almost all the CH4 was consumed during the combustion process. 2. At the initial stage of the optimization process, both re-entrant combustion chamber and open combustion chamber are

dominated in the Pareto front. However, with the increase of the generation, the open combustion chamber degenerated into the straight combustion chamber at around the 30th generations. After that, only the re-entrant combustion chamber and the straight combustion chamber are appeared in the Pareto front. 3. The NO emission is the lowest when the re-entrant combustion chamber is used, while the NO emission is the highest when the open combustion chamber is used. A reverse trend is found for the CH4 emission. When the diesel injection timing is around
25.00  CA ATDC, the NO emissions, CH4 emissions and ISFC are very close to each other for the three types of the combustion chamber.

J. Liu et al. / Energy 164 (2018) 837e852

(a)

(b)

Fig. 19. The distribution of temperature and CH4 mole fraction at selected crank angles.

Fig. 20. The in-cylinder flow field at selected crank angles.

851

852

J. Liu et al. / Energy 164 (2018) 837e852

4. The optimized case with higher peak cylinder pressure has the higher pressure raise rate and higher NO emission at the same time. However, the ISFC and CH4 emissions have the opposite trend, where the lower value is found with the higher peak cylinder pressure case. 5. The ISFC is high when using the re-entrant combustion chamber. And the ISFC can be reduced obviously when using the straight combustion chamber. Moreover, the CH4 emissions are significantly lower when using the straight combustion chamber compared with using the re-entrant combustion chamber. Acknowledgements This study is supported by National Natural Science Foundation of China (NO. 51406007) and The National Key Research and Development Program of China (NO 2017YFB0103500).

[8]

[9]

[10]

[11]

[12]

[13]

[14]

Nomenclature and list of abbreviation HCCI LTC PCCI RCCI CN ON BTDC ATDC SOC THC ISFC IVC CR DME CNG

Homogeneous Charge Compression Ignition Low Temperature Combustion Premixed Charge Compression Ignition Reactivity Controlled Compression Ignition Cetane Number Octane Number Before Top Dead Center After Top Dead Center Start of Combustion Total Hydrocarbon Indicated Specific Fuel Consumption Intake Valve Closure Compression Ratio Dimethyl Ether Compressed Natural Gas

[15]

[16]

[17]

[18]

[19] [20]

[21]

[22]

References [1] Kozarac D, Taritas I, Vuilleumier D, et al. Experimental and numerical analysis of the performance and exhaust gas emissions of a biogas/n-heptane fueled HCCI engine. Energy 2016;115:180e93. [2] Shi L, Xiao W, Li M, et al. Research on the effects of injection strategy on LTC combustion based on two-stage fuel injection. Energy 2017;121:21e31. [3] Jia M, Li Y, Xie M, et al. Numerical evaluation of the potential of late intake valve closing strategy for diesel PCCI (premixed charge compression ignition) engine in a wide speed and load range. Energy 2013;51:203e15. [4] Splitter D, Hanson R, Kokjohn S, Reitz R. Reactivity controlled compression ignition (RCCI) heavy-duty engine operation at mid-and high-loads with conventional and alternative fuels. SAE Paper 2011-01-0363. [5] Yousefi A, Birouk M, Guo H. An experimental and numerical study of the effect of diesel injection timing on natural gas/diesel dual-fuel combustion at low load. Fuel 2017;203:642e57. [6] Rahnama P, Paykani A, Bordbar V, Reitz RD. A numerical study of the effects of reformer gas composition on the combustion and emission characteristics of a natural gas/diesel RCCI engine enriched with reformer gas. Fuel 2017;209: 742e53. [7] Rahnama P, Paykani A, Reitz RD. A numerical study of the effects of using hydrogen, reformer gas and nitrogen on combustion, emissions and load

[23] [24] [25]

[26] [27] [28]

[29]

[30]

limits of a heavy duty natural gas/diesel RCCI engine. Appl Energy 2017;193: 182e98. Wang Z, Zhao Z, Wang D, et al. Impact of pilot diesel ignition mode on combustion and emissions characteristics of a diesel/natural gas dual fuel heavy-duty engine. Fuel 2016;167:248e56. Papagiannakis RG, Krishnan SR, Rakopoulos DC, et al. A combined experimental and theoretical study of diesel fuel injection timing and gaseous fuel/ diesel mass ratio effects on the performance and emissions of natural gasdiesel HDDI engine operating at various loads. Fuel 2017;202:675e87. Poorghasemi K, Saray RK, Ansari E, et al. Effect of diesel injection strategies on natural gas/diesel RCCI combustion characteristics in a light duty diesel engine. Appl Energy 2017;199:430e46. Li W, Liu Z, Wang Z. Experimental and theoretical analysis of the combustion process at low loads of a diesel natural gas dual-fuel engine. Energy 2016;94: 728e41. Song H, Liu C, Li Y, et al. An exploration of utilizing low-pressure diesel injection for natural gas dual-fuel low-temperature combustion. Energy 2018;153:248e55. Yang B, Ning L, Chen W, et al. Parametric investigation the particle number and mass distributions characteristics in a diesel/natural gas dual-fuel engine. Appl Therm Eng 2017;127:402e8. Xu G, Jia M, Li Y, et al. Multi-objective optimization of the combustion of a heavy-duty diesel engine with low temperature combustion under a wide load range: (I) Computational method and optimization results. Energy 2017;126:707e19. Jung J, Song S, Hur KB. Numerical study on the effects of intake valve timing on performance of a natural gas-diesel dual-fuel engine and multi-objective Pareto optimization. Appl Therm Eng 2017;121:604e16. Park S. Optimization of combustion chamber geometry and engine operating conditions for compression ignition engines fueled with dimethyl ether. Fuel 2012;97:61e71. Wang B, Li T, Ge L, Ogawa H. Optimization of combustion chamber geometry for natural gas engines with diesel micro-pilot-induced ignition. Energy Convers Manag 2016;122:552e63. Lee S, Park S. Optimization of the piston bowl geometry and the operating conditions of a gasoline-diesel dual-fuel engine based on a compression ignition engine. Energy 2017;121:433e48. Chen Y, Lv L. The multi-objective optimization of combustion chamber of DI diesel engine by NLPQL algorithm. Appl Therm Eng 2014;73:1332e9. Benajes J, Novella R, Pastor JM, et al. Optimization of the combustion system of a medium duty direct injection diesel engine by combining CFD modeling with experimental validation. Energy Convers Manag 2016;110:212e29. Kim HJ, Park SH. Optimization study on exhaust emissions and fuel consumption in a dimethyl ether (DME) fueled diesel engine. Fuel 2016;182: 541e9. Kavuri C, Kokjohn SL. Computational optimization of a reactivity controlled compression ignition (RCCI) combustion system considering performance at multiple modes simultaneously. Fuel 2017;207:702e18. Beale JC, Reitz RD. Modeling spray atomization with the Kelvin helmholtz/ Rayleigh-taylor hybrid model. Atomization Sprays 1999;9:623e50. Munnannur A, Reitz RD. Comprehensive collision model for multidimensional engine spray computations. Atomization Sprays 2009;19:597e619. pez S Yang J, Golovitchev VI, Redon P, Lo anchez JJ. Numerical analysis of NOx formation trends in biodiesel combustion using dynamic f-t parametric maps. SAE 2011-01-1929. Mattarelli E, Rinaldini CA, Golovitchev VI. CFD-3D analysis of a light duty Dual Fuel (Diesel/Natural Gas) combustion engine. Energy Proced 2014;45:929e37. Chen WM, Shuai SJ, Wang JX. A soot formation embedded reduced reaction mechanism for diesel surrogate fuel. Fuel 2009;88:1927e36. Liu J, Zhang X, Wang T, Zhang J, Wang H. Experimental and numerical study of the pollution formation in a diesel/CNG dual fuel engine. Fuel 2015;159: 418e29. Lipatnikov AN, Chomiak J. Turbulent flame speed and thickness: phenomenology, evaluation, and application in multi-dimensional simulations. Prog Energy Combust Sci 2002;28:1e74. Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective genetic algorithm: NSGA-II. IEEE Trans Evol Comput 2002;6(2):182e97.