1. Introduction
To limit the impact of emissions on human health and the Earth’s environment, the marine emission regulations released by the International Maritime Organization (IMO) are becoming increasingly strict. In order to meet these stricter emission regulations, emission reduction solutions are mandatorily applied on both new-build and existing ships nowadays [
1,
2,
3]. According to the MARPOL (International Convention for the Prevention of Pollution from Ships) Annex VI, since 1 January 2020, all ships have to comply with the utilization of fuels with max. 0.5% sulfur globally. The minimum reduction in carbon intensity per maritime transport work must reach 40% by 2030 compared with 2008, with the target of reaching 70% by 2050. The reduction in greenhouse gas (GHG) emissions from marine shipping must reach at least 50% by 2050 compared with 2008. From 2016, NOx emissions from ships have been limited to 3.4 g/kWh for engines with speeds less than or equal to 130 rpm. This limit gradually decreases with increasing engine speed and reaches only 2 g/kWh for engines with a speed higher than or equal to 2000 rpm [
4]. There are various solutions for reducing emissions from marine engines, but this paper focuses on two aspects: engine technologies and alternative fuels.
Among combustion engines, internal combustion engines (ICEs) with their high thermal efficiency have been widely used in industry. Among ICEs, compression ignition (CI) diesel engines have higher thermal efficiency compared to spark ignition (SI) gasoline engines [
5]. They are therefore currently contributing a very large portion in providing propulsion for propellers as well as generators on ships. However, EGEs such as NOx (nitrogen oxides), soot, and CO
2 (carbon dioxide—one of the strong greenhouse gases—GHGs) produced from the combustion process of such engines are major concerns nowadays.
In ICEs, the combustion of fuel by the O
2 (oxygen) present in the supplied air inside the engine combustion chamber generates engine power, while also producing EGEs. Therefore, engine performance and EGE characteristics are greatly influenced by the fuel type, fuel-supplying method, fuel injection characteristics, fuel–air equivalent ratio, fuel–air mixing quality, combustion chamber geometry, etc. [
6]. Among many kinds of fuels, natural gas (NG), which principally consists of methane (CH
4), has been and will still be widely used in heavy-duty marine engines. It has many advantages, such as low EGEs, low price, no processing, abundant reserves, etc. However, since NG has a high self-ignition temperature (low cetane numbers), it needs an external energy source for ignition, such as a spark plug in SI engines or pilot diesel oil in dual-fuel (DF) engines. If NG is used in the DF engines, pure diesel engines can be modified to an NG–diesel DF engine very easily and with only a low cost [
7,
8,
9]. Numerous studies have been performed to investigate the combustion and EGE characteristics of NG–diesel DF engines [
10]. The detailed properties, as well as the effects of NG on the combustion and EGE characteristics of an NG–diesel DF engine, can also be referred to in our previous studies [
11,
12].
Regarding engine technologies, ME-GI (M-type, Electronically Controlled Gas Injection) engines designed by MAN B&W are one of the state-of-the-art engines in the maritime industry today. In total, more than 200 engines have been ordered and in use since 2015, and the number will rapidly increase in the coming years [
13]. Unlike the DF engines, which use intake port injection for primary gas introduction, ME-GI DF engines use the direct injection method for the primary gas introduction. The gaseous fuel is injected and burned directly, leading to high efficiency and combustion stability according to the diesel cycle principle. With direct gas injection, the injection timing, gas injection rate, and injection pressure can be controlled in a wide range of engine loads and speeds to keep the optimal spraying quality. This is considered to be one of the most important and necessary factors to achieve better combustion efficiency [
14]. Furthermore, the engines designed according to the diesel cycle principle allow the operators to switch smoothly between different fuel modes, i.e., the fuel oil and DF mode [
13].
Through the above analysis, the combination of using cleaner NG fuel and fuel direct injection technology in ME-GI DF engines provides an extremely effective solution for reducing EGEs from ships. To maximize the advantages of ME-GI engines, operators need to have insight into the influence of operating parameters on the engine. However, because the ME-GI engine family is so new, not much research has been done on these engines. In addition, the parameters related to the structure and layout of the fuel injectors in the combustion chamber are confidential documents of companies, so they should always be carefully preserved to avoid technology theft. As a result, it is impossible for engine operators to access these documents. This study aims to analyze the effectiveness of using NG as the primary fuel in ME-GI DF engines and the impacts of some of the main parameters of the fuel injector on the combustion and emission formation of the engine. This research will help engine designers and operators to be aware of the key factors that greatly impact engine power and emissions so that they can design and operate engines more efficiently.
In fuel direct-injection engines, the diffusion combustion phase accounts for most of the engine’s combustion process. In this phase of combustion, the mixing quality between the injected fuel and supplied air inside the engine combustion chamber strongly affects the rate of heat release (RoHR) and the formation of harmful emissions. Ideally, all the injected fuel will be in contact with all the oxygen available in the combustion chamber so that the combustion of fuel can take place as completely as possible. Both the fuel injection characteristics and atomization are the major factors in improving engine power while reducing EGEs [
6]. Regarding the fuel injection characteristics, the injection method (port- or direct-injection), injection timing, injection strategy (single- or multiple-injection), SA, and IP play important roles in the combustion and EGE characteristics of engines. In fuel direct-injection engines, SA and IP have great influences on the combustion and EGE formation of the engine as they determine the fuel injection targeting point in the combustion chamber [
5]. Therefore, studies for SA and IP are very important and will have implications for both engine design and operating engineers.
Regarding the multiple-injection strategy, a series of studies have been carried out and demonstrated the effectiveness of this strategy in reducing soot or even soot and NOx (nitrogen oxides) emissions simultaneously in heavy-duty direct-injection diesel engines [
15,
16,
17,
18,
19]. A systematic analysis of the mechanisms of droplet breakup and mixing with the surrounding fluid of fluid injection, and its relation to vorticity generation and transport, has been reported and can be referred to [
20]. The effectiveness of the double-injection strategy in reducing EGEs applied on a heavy-duty direct-injection diesel marine engine has also been demonstrated and can be referenced in our previous study [
21].
Maghbouli et al. reported in [
22] that the nozzle axial location, which determines IP, considerably affected both the in-cylinder pressure and EGEs of fuel direct-injection engines. Here, a change in the axial location of the fuel nozzle by only a few millimeters led to a considerable change in the engine performance and emission behavior. Mobasheri and Peng reported in [
23] that the narrow-spray-angle injection helped to reduce the NOx and soot emissions without affecting the fuel oil consumption due to improved fuel–air mixture quality. Yoon et al. studied the effect of SA and fuel injection strategy on the combustion and emission characteristics of a direct-injection engine operating with dimethyl ether (DME). Through the study, they found that the in-cylinder peak pressure in the single-injection strategy using narrow-spray-angle injection (60 and 70 degrees) with advanced fuel injection timing was increased compared to that in the wide-spray-angle injection (156 degrees) with later injection timing [
24]. In direct-injection CI diesel engines, the target injection point in the engine’s combustion chamber is very important because it causes wetting of the combustion chamber walls, which causes the formation of unburnt hydrocarbons (UHC). Additionally, the consumption of O
2 in the combustion process also greatly depends on the targeting point [
25,
26,
27]. Furthermore, rapid combustion in fuel-rich regions inside engine combustion chambers causes the deterioration of the combustion efficiency, knocking phenomenon, and high levels of NOx emissions from the engine. Therefore, many researchers have tried to determine the optimal SA [
28,
29,
30,
31]. Fang et al. reported in [
31] that using narrow-angle injection with a 70-degree SA led to higher soot formation due to the deposition of the fuel film on the piston wall. It, however, lowers NOx emissions because of the leaner fuel–air mixture near the piston surface. Shu et al. used the computational fluid dynamic (CFD) method coupled with a chemical kinetic model to investigate the emission characteristics of an NG–diesel DF engine at various SA [
7]. They concluded that NOx emissions increased by 93% when the fuel injector SA increased from 60 to 140 degrees but decreased by 15% when the SA was further increased to 160 degrees. Lim and Min performed a CFD analysis in a diesel engine operating with various SAs in order to reduce soot and minimize wall impingement [
32]. They found that at the SA of 100 degrees, soot was reduced by 41% compared to that in the case of the SA of 70 degrees.
From the above literature review, it is very important to note that although many studies have been carried out on NG–diesel DF engines to investigate the effect of SA on the combustion and emission characteristics of the engine, studies on the effect of SA and IP on NG–diesel ME-GI DF marine engines are very rare. The authors have reviewed a wide variety of literature but still have not found any studies similar to the one in this report.
In this study, the effect of SA and IP on the combustion process and emission characteristics of an NG–diesel ME-GI DF marine engine was investigated by using the CFD method coupled with a reduced chemical kinetic model. The combustion and emission formation occurring inside the engine cylinder were modeled by using the ANSYS Fluent code. The ultimate target of this study was to specify the optimal SA and IP for the engine. The CFD models were validated by the experimental results reported in the engine’s shop test technical data. The study also successfully assigned the optimal SA and IP for the engine to achieve certain emission reductions.
2. Numerical Analysis
The ANSYS Fluent 2019R2 software, with its advanced models, was used to simulate the combustion and emission formations that occur inside the engine cylinder. The working process of the engine was modeled from the Scavenge Port Closing (SPC) to the Exhaust Valve Opening (EVO). The simulation process included three steps: (1) pre-processing, (2) processing, and (3) post-processing. The first step included building the computational domain (Geometry), creating movable computational meshes (Meshing), and setting up the simulation parameters (Setup). The solutions were calculated in the second step. In the third step, the simulation results were analyzed and reported. After obtaining the simulation results, the simulation and experimental results were compared for the CFD models’ validation purposes. This process was repeated until the simulation and experimental results matched so that the numerical simulation results could be exported.
2.1. Combustion Chamber Geometry and Computational Mesh
The object of this study is a 2-stroke heavy-duty ME-GI DF marine engine. This is a DF engine that operates according to the diesel cycle. The engine can be operated smoothly in both diesel and DF modes. In the diesel mode, the engine works with diesel oil only. In the DF mode, diesel oil serves only as a pilot fuel to provide an ignition source, while NG serves as a primary fuel. In ME-GI engines, both the gaseous primary and liquid pilot fuels are directly injected into the engine cylinder by gas and pilot nozzles accordingly. In this study, the effects of SA and IP on the cylinder pressure, temperature, performance, and emission characteristics of the engine in both diesel and DF modes were analyzed.
The CFD analysis is derived from the finite volume method (FVM). This method reduces the degrees of freedom (DoF) from infinite to finite with the help of discretization or meshing. In CFD analysis, a continuous fluid domain is divided into a discrete domain consisting of a finite number of elements called “elements”, which are connected through “nodes”. The governing equations of fluid dynamics (the continuity, momentum, and energy equations) in the computational domain will be calculated at these nodes. This means that the number of equations that need to be solved and thus the calculation time will be proportional to the number of nodes in the computational domain. Therefore, in most cases of CFD analysis, researchers will always try to reduce the number of nodes of the computational domain to be as low as possible while still ensuring the mesh quality criterion. For 3D computational domains that are symmetric (axis or plane), an effective way to reduce the number of nodes is to compute for only one symmetric part of the entire domain. The cutting surfaces of the computational domain will be assigned a “cyclic” or “periodic” boundary condition [
33].
In ICE simulations, if the engine has an axis-symmetric combustion chamber, as typically found in CIE engines since they do not contain a spark plug geometry, it is highly recommended to use only one portion of the entire computational domain for simulation [
34]. In this study, the engine uses 3 gas injectors for NG injection and 3 other pilot injectors for pilot fuel injection, spaced by 120 degrees evenly on the cylinder head. The piston surface is a U-type shape. Due to the axial symmetry of the engine, only a one-third section of the entire 3D model of the engine has been modeled in order to reduce the calculation time. The one-third section of the engine combustion chamber geometry is shown in
Figure 1, while
Table 1 shows the main specifications of the engine.
Figure 2 shows the one-third portion of the entire 3D computational mesh of the combustion chamber at the top dead center (TDC) of the engine. The movement of the piston surface during the calculation was modeled by a dynamic mesh generated by the layering method. Due to the characteristics of the layering method, the computational mesh used in this study is a structured mesh with all elements in the computational domain as hexahedral cells. The movement of the piston surface was controlled by Equation (1). The mesh properties together with mesh quality are shown in
Table 2.
where
is the piston location,
is the crank radius,
is the length of the connecting rod,
is the piston pin offset, while
is the current crank angle degree (CAD).
In this study, the computational mesh is re-meshed according to the movement of the piston based on the dynamic Layering method. During the piston movement, mesh layers will be added or collapsed depending on the movement direction of the piston (downward or upward movement). This means that the number of mesh layers and thus the number of elements will be increased or decreased according to the moving down or up of the piston. In this way, the mesh quality and resolution will be kept almost unchanged during calculation. In this moving action of the piston, only the cylinder liner surface and interior of the computational mesh were re-meshed by the Layering method. The cylinder head surface stayed stable.
In this study, the Stochastic Collision model was used to model the interaction between the fuel droplets and combustion chamber walls. To eliminate the near-wall effects of the piston and cylinder liner surfaces, an inflation with 1.2 mm thickness and 5 inflation layers was applied on the piston and cylinder liner surfaces. Since the target point of fuel injection is only forwards toward the piston and cylinder liner surfaces of the engine, the effects of the near-wall behavior of the cylinder head surface can be neglected. The inflation option therefore only needs to be applied to the piston and cylinder liner surfaces.
2.2. Simulation Cases and Fuel Properties
Eighteen simulations have been carried out for both diesel and DF modes. Specifically, in each operating mode (diesel and DF mode), the IP has been adjusted to 0.01 m, 0.015 m, and 0.02 m distance from the cylinder head surface; in each IP, the SA was adjusted to angles of 40, 45, 50 degrees relative to the vertical direction.
Figure 3 shows the schematic of the SA and IP adjustment in this study.
Table 3 presents the simulation cases for each operating mode. The simulation results were then compared to determine the optimal parameters for the fuel injector, i.e., the SA and IP of the nozzle. In this study, CH
4 was used to represent NG, while C
10H
22 was used to represent diesel oil. The main properties of CH
4 and C
10H
22 are shown in
Table 4.
2.3. CFD Models and Governing Equations
This study used the pressure-based type with absolute velocity formation for the solver. The effect of gravitational acceleration was also included in the simulation. The acceleration value used was 9.81
. Due to the piston’s moving action during the simulation, an unsteady (transient) approach was used. In engine modeling, the time step size is calculated based on the piston movement speed. The time step size value was
s. However, due to the complexity of the fuel combustion process, the time step size of the simulation during this period (12 CADs BTDC to 20 CADs ATDC) was set to be reduced to 16 times to ensure solution convergence. Here, 50 maximum iterations per time step were set. Monitoring the residuals showed that the solution reached convergence after 31 to 35 iterations per time step, and the computation was then skipped to the next time step. The PISO scheme with a value of 1 for both skewness and neighbor corrections was applied. For spatial discretization, the Green-Gauss Node-Based gradient and standard pressure were used. Meanwhile, Second-Order Upwind was applied for density, momentum, TKE, turbulent dissipation rate, nuclei, and pollutant calculations. An absolute criterion of
for convergence criteria was set for continuity, velocity, k, and ε, while a stricter criterion of
was set for energy and pollutant calculations. These convergence criteria are suitable for combustion and emission simulations of internal combustion engines, as presented in [
33].
The standard k-
turbulence model was employed to model the fluid flow turbulence inside the engine cylinder. This model is suitable and widely used to model the turbulence inside the cylinder of ICEs [
37]. The combustion of fuel in the engine cylinder was modeled by the non-premixed combustion model using a reduced chemical mechanism. The reduced chemical reaction was generated using the probability density function (PDF) method in the ANSYS Fluent solver. This model is suitable for simulating non-premixed combustion, where the oxidizers and fuels are introduced into the combustion apparatus separately [
33]. Reaction mechanisms play a very important role in combustion simulations. For a deeper understanding of the reaction mechanism as well as the latest developments in this field, it would be helpful to refer to [
38]. To simulate the spraying of the nozzle, the Discrete Phase Model (DPM) was used. This model offers an effective way to correct the spray velocities and initial droplet diameters caused by the cavitation phenomenon. In this model, the Kelvin–Helmholtz Rayleigh–Taylor (KHRT) sub-model was utilized to model the breakup phenomenon of fuel droplets. Meanwhile, the auto-ignition model was employed to simulate the self-ignition of fuel in the engine cylinder.
The composition PDF transport governing equation is derived from the Navier–Stokes equations as [
39]:
where,
| mean fluid density; |
P = | Favre joint PDF of composition; |
| Favre mean fluid velocity vector; |
| composition space vector; |
| reaction rate for specie k; |
| molecular diffusion flux vector; |
| fluid velocity fluctuation vector. |
In Equation (2), the notation denotes expectations, while is the conditional probability of event A, given that event B occurs. Meanwhile, the two terms on the right-hand side of the equation represent the PDF change due to scalar convection caused by turbulence (turbulent scalar flux) and molecular mixing/diffusion, respectively.
The turbulent scalar flux term is modeled by the gradient-diffusion assumption:
where
and
are the turbulent viscosity and Schmidt number, respectively.
The Extended Zeldovich mechanism was utilized to model the NO formation occurring inside the engine cylinder [
33,
40]. This reaction mechanism consists of 7 species and 3 chemical reactions and can highly accurately predict NO formation over a wide equivalence ratio range. This model is presented in
Appendix A of this article and can also be referenced in detail in our previous publications [
11,
12,
21]. The soot formation during the combustion process of the engine was modeled by the Moss–Brookes model [
33,
41,
42]. The collisions and interactions between the injected fuel droplets and combustion chamber walls were modeled using the Stochastic Collision sub-model. This model is derived from the O’Rourke algorithm [
33,
41,
42]. The CFD models used in this study have been successfully applied and can be found in our previous publication [
21] or references [
33,
40,
41].
Table 5 shows a summary of the CFD numerical models used in this study.
2.4. Boundary and Initial Conditions
The boundary conditions (BCs) at the cylinder liner, cylinder head, and piston surface were defined as impermeable walls. The geometry of the combustion chamber is axially symmetric, so the cutting surfaces of the cutting portion of the computational domain were assigned as cyclic BCs. The simulation was started from the SPC of 30 crank angle degrees (CADs) after the bottom dead center (BDC) to the EVO of 30 CADs before the bottom dead center (BBDC) of the engine. The start of injection (SOI) of the pilot and gaseous fuels were 12 and 10 CAD BTDC, respectively. The injection duration of the pilot and gaseous fuels were 12 and 30 CADs, respectively. The single-injection method with step impulse function was applied for both pilot and gas fuel injections. The injection processes take place from the SOI to the EOI continuously with constant mass flow rates. A schematic of the working process of the engine is shown in
Figure 4. The BC
S and initial conditions for the numerical CFD simulations in this study were obtained from the engine shop test experiments and are summarized in
Table 6.
2.5. Mesh Independence Analysis
Mesh density or mesh resolution plays an important role in mesh quality and therefore in the accuracy of the final CFD result. Additionally, mesh density also affects the calculation time. Therefore, to ensure the accuracy of the final CFD result and the reasonableness of the computation time, a mesh independence analysis was performed. Three simulations using three various mesh densities, including coarse, medium, and fine meshes, were performed to analyze the mesh independence of the solution. The simulations were performed in parallel using an Intel(R) Xeon(R) CPU X5690 @ 3.47 GHz, 16-core, 32-thread processor, 32 GB RAM Windows workstation.
Table 7 shows the properties of the meshes with the respective calculation time in the mesh independence analysis. A comparison between the final CFD results using these three mesh densities is shown in
Figure 5.
The comparison showed the good agreement of the final CFD results between the medium and fine mesh cases. Therefore, the medium mesh was chosen for simulations in this study because it gave accurate CFD results that were the same as the fine mesh but in a shorter calculation time.
2.6. CFD Model Validation
The CFD models used in this study have been validated by comparing the numerical simulation results to the experimental results reported in the engine’s shop test data. The comparisons between the simulation and experimental results for both diesel and DF modes are shown in
Figure 6. The variables used for the CFD models’ validation are in-cylinder peak pressure (
Figure 6a), indicated mean efficiency pressure (IMEP) (
Figure 6b), and NO emission (
Figure 6c). It is obvious that the numerical simulation results and experimental results are in good agreement. The maximum deviation between the simulation and experimental results was only 7% in the case of NO emission comparison in the diesel mode. Meanwhile, the deviations between the remaining cases were less than 1%. These deviations are within the 10% limitation, which is widely accepted in CFD analysis [
43,
44]. The validated CFD models were then utilized for numerical simulations when adjusting the SA and IP of the fuel nozzle in this study.
The model validation demonstrated that the right models were selected among the CFD models for the simulation of the engine. Therefore, these models can in principle be applied to simulate the same phenomena occurring in similar types of engines (i.e., gas–diesel DF engines using the direct-injection method for fuel introduction). In fact, the accuracy of a CFD simulation result depends not only on the models chosen but also on the quality of the mesh. Therefore, when applying these validated models in this study to simulate similar phenomena in similar types of engines, performing a mesh independence analysis of the CFD results is necessary. This was to ensure the suitability of the chosen CFD models for another geometry. To perform a mesh independence analysis, it is recommended to refer to the procedure suggested by Celik [
45].