Abstract

Lean combustion is a promising strategy to increase thermal efficiency in an internal combustion engine, by exploiting a favorable specific heat ratio of the fresh mixture while simultaneously suppressing the heat losses to the cylinder wall. However, unstable ignition and slow flame propagation at fuel-lean conditions lead to large cycle-to-cycle variability and limit the high-efficiency engine operating range. Prechamber ignition is considered an effective concept to extend the lean operating limit, by providing spatially distributed ignition with multiple turbulent flame-jets and enabling a faster combustion rate compared to the conventional spark ignition approach. From a numerical modeling standpoint to date science base and available simulation tools are inadequate to properly understand and predict the combustion processes in prechamber ignited engines. In this paper, conceptually different Reynolds-averaged Navier–Stokes (RANS) combustion models widely adopted in the engine modeling community are used to simulate the ignition and combustion processes in a medium-duty natural gas engine with a prechamber spark-ignition system. A flamelet-based turbulent combustion model, i.e., G-equation, and a multizone well-stirred reactor model are employed for this modeling study. Simulation results are compared with experimental data in terms of in-cylinder pressure and heat release rate. Finally, the analysis of the performance of the two models is carried out to highlight the strengths and limitations of the two evaluated approaches.

1 Introduction

Stringent regulations on fuel economy and pollutants emissions from vehicles continue to drive the automakers to produce more efficient and cleaner powertrains. In theory, the thermal efficiency of an internal combustion engine is defined as a function of compression ratio and specific heat ratio. Specific heat ratios of lean mixtures are close to ideal (air only) mixture properties. In addition, relatively lower combustion temperatures from lean mixtures allow mitigating the heat losses to the cylinder wall. Therefore, the engine thermal efficiency can be improved with lean combustion. However, at lean conditions, the ignition event becomes unstable, and the laminar flame speed decreases, which likely results in large cycle-to-cycle variability. This limits the attainable high-efficiency engine operation. To overcome combustion instability and maximize lean-burn operation, better use of ignition energy is needed.

Prechamber ignition is not a novel concept. The first prechamber engine was invented by Sir Harry Ralph Ricardo in 1918 [1], and the “Lavinia Aktyvatsia Gorenia” (Russian, generally referred to as LAG-ignition) using a low-temperature torch ignition with fuel-rich prechamber mixture was proposed by Goossak Lev Abramovich [2]. Within decades, various kinds of prechamber ignition concepts were developed and proposed, such as the compound vortex controlled combustion by Honda [3] and the premixed charge forced auto-ignition by General Motors [4]. Currently, Prechamber spark-ignition (PCSI) is receiving a renewed interest from the automotive industry due to its capability to extend the lean operating limit by leveraging the multiple turbulent flame/hot jets ignition in the main chamber. The spatially distributed ignition source across the flame/hot jet surface enables faster combustion rates for lean mixtures compared to the conventional spark-ignition (SI) approach. Additionally, PCSI is a suitable concept for a wide range of engine platforms (from light-duty to heavy-duty engines) and fuels (from liquid to gaseous). The PCSI can be classified into unfueled and fueled operations (i.e., unscavenged and scavenged) depend on the fueling strategy. In an unfueled PCSI, the fuel is delivered to the prechamber from the main chamber by pressure differences between the two chambers. In a fueled PCSI, an auxiliary fueling event takes place in the prechamber by an injector or a check-valve, which allows a stoichiometric or fuel-rich mixture formation in the proximity of the spark-plug. For a light-duty gasoline engine platform, the fueled PCSI approach demonstrated the reduction of average brake specific fuel consumption under ultralean conditions by 20–25% compared to benchmark production stoichiometric SI engines [5]. Accordingly, PCSI is regarded as a leading ignition concept with the potential to achieve high thermal efficiency also in medium-duty and heavy-duty natural gas (NG) engines. Shah et al. [6] conducted an experimental study on a heavy-duty NG engine equipped with an unfueled PCSI system and showed that the flame development angle, as well as combustion duration, were shorter in unfueled PCSI than conventional SI under a wide engine operating range with dilution.

Despite the advantages of PCSI with respect to conventional ignition, four major technical barriers to market penetration of PCSI for medium-duty/heavy-duty NG engines were discussed in the 2017 Natural Gas Vehicle Research Workshop [7]. One of the barriers consists of inadequate science base and simulation tools to predict the combustion processes governing PCSI. Computational tools are essential for engineers in the industry to optimize designs for efficiency, pollutant formation, emissions control integration, noise, reliability, and drivability. To date, various turbulence and combustion models based on computational fluid dynamics (CFD) have been used to simulate prechamber combustion. In the Reynolds-averaged Navier–Stokes (RANS) framework, multizone well-stirred reactor (MZ-WSR) model coupled with detailed chemistry [8,9], G-equation model [10,11], and extended coherent flame model (ECFM) [12], were used. Gholamisheeri et al. [9] performed a set of numerical simulations with the MZ-WSR model for a fueled prechamber in a rapid compression machine. The simulation results were able to qualitatively reproduce the experimental burn rate, peak pressure, and ignition delay, albeit with limited quantitative agreements with all chemical reaction mechanisms considered. Xu et al. [10] used the G-equation model to simulate an unscavenged prechamber gas engine and showed overall good agreement between simulation results and experimental data for different engine speed, load, and air/fuel ratio operating conditions. Among the tested turbulent burning velocity correlations, a correlation not considered for the local turbulent length scale condition over-predicted the cylinder pressure significantly. For an engine-like rapid compression-expansion machine, Xu et al. [11] also demonstrated that the G equation model reproduced the measured heat release and mean flame front propagation. On the other hand, based on the large eddy simulation (LES) framework, MZ-WSR [13], ECFM-3Z [14], flamelet generated manifold [15], Weller combustion model [16], and thickened flame model [17] were used. Syrovatka et al. [14] used the LES ECFM-3Z model to investigate the effect of prechamber volume on combustion and performance in an NG engine. The simulation results showed acceptable agreement by matching the measured cyclic variation range. Bolla et al. [15] conducted numerical simulations of an automotive-sized scavenged prechamber mounted in a rapid compression-expansion machine. The LES FGM model slightly under-predicted the experimental peak pressure with the base model but matched the timing of the first jet exiting and jet penetration compared to the experimental data. Malé et al. [17] performed a high-fidelity LES with the TFM model for a scavenged PCSI engine. Despite the absence of experimental validation, numerical results revealed that the combustion in the main chamber starts in a distributed reaction mode before switching to a flame propagation mode. Overall, each of the evaluated models seems to show the potential to match experimental data. However, there is not a clear assessment on which of these models is most suitable to simulate combustion in PCSI medium-duty NG engines.

The theoretical formulation of each model is derived under a set of assumptions, for example, laminar flamelet assumption for either G-equation or ECFM. Such an assumption typically targets a specific type of combustion, e.g., premixed or nonpremixed, so that the simulation result from a model is bounded by its inherent description. On the other hand, tuning of the model constants can disclose the model performance and provide information on whether a specific model is appropriate to investigate a specific application. A previous study by the authors [18] showed that a conventional tuning effort by altering: (1) the reaction multiplier of the MZ-WSR model, (2) b_1 of a turbulent flame speed correlation of the G-equation model, and (3) α of turbulent flame stretch factor of the ECFM, could not be sufficient to deliver a good agreement between simulations and experiments in both the prechamber and main chamber. In addition, it was found that the effect of small-scale turbulence inside the prechamber plays a key role in describing the PCSI combustion process, and multiple regimes of turbulent combustion should be taken into account when modeling PCSI combustion. However, this previous work leveraged closed-cycle simulations and was affected by several uncertainties in initializing the mixture and flow field and defining the boundary conditions. Furthermore, only one excess-air ratio (λ) condition was investigated.

In this paper, simulations are expanded from a closed-cycle geometry to a full engine geometry, and simulation results are evaluated for two operating conditions up to the lean operating limit. To secure a set of reliable boundary conditions, a one-dimensional (1D) model is built and validated with the engine experimental data. For the combustion model evaluation, two conceptually different RANS models that are widely adopted in the engine modeling community, are employed. A flamelet-based combustion model, G-equation, and a homogeneous reactor-type combustion model, MZ-WSR, are used to simulate the ignition and combustion processes in a medium-duty NG engine equipped with a PCSI system and operated under unfueled mode. In the sections, first, the experimental and numerical setup will be described in Sec. 2. Then, the results from the combustion model evaluation will be presented in Sec. 3.1. Finally, the analysis of the PCSI combustion will be discussed in Sec. 3.2.

2 Experimental and Numerical Setup

2.1 Experimental Setup.

The engine experiment is conducted using a single-cylinder, medium-duty NG SI engine at Argonne National Laboratory. The engine has 1.86 L of displacement volume with 130 mm bore and 140 mm stroke, and an initial compression ratio of 11:1 that is slightly lowered by 0.25 due to the prechamber installation, hence resulting in an operating compression ratio of 10.75:1. The prechamber volume is 4.67 cm3, and the ratio of that to the engine cylinder volume at the top dead center (TDC) is 2.45%. Eight nozzles of 1.6 mm diameter are drilled at the tip of the prechamber with an included umbrella angle of 130 deg, and with the umbrella axis parallel to the cylinder vertical axis. The charge in the prechamber is ignited by an M8 spark plug installed on the prechamber head and operated by a capacitive discharge ignition system.

NG is supplied from Nicor Gas (Naperville, IL), and its main components are methane (93.2% v/v), ethane (5.8% v/v), and propane (0.3% v/v), of which detailed composition can be found on the website [19]. The fuel is introduced into the engine intake runner using an electronically controlled gas injector (Clean Air Power, SP-010). Compressed air is routed via an air surge tank and supplied to the intake runner at controlled temperature and pressure. Flow meters for both air and fuel, as well as a λ-analyzer in the emission bench (Horiba MEXA-7100D), are used to measure the air-fuel ratio during the engine run. In this study, as an unfueled-PCSI operation, the prechamber mixture mainly consists of residual burned gas from the previous cycle plus the main chamber fuel/air mixture fed during the compression stroke by the pressure differences.

To acquire the pressure traces in both main and prechamber, two relative high-speed pressure transducers Kistler 6043A) are used and pegged with crank angle resolved intake runner pressure, which is measured by an absolute pressure sensor (Kistler 4049B). Using a combustion analyzer (AVL IndiModule), a total of 300 cycles of pressure data is acquired at 0.1 crank angle degree (CAD) sampling rate under steady-state operation. Table 1 summarizes the engine operating conditions explored in this study, where the λ = 1.65 is the lean operating limit at the given engine speed and intake pressure by keeping the cyclic variability of indicated net effective pressure (COVIMEP) less than 5%.

Table 1

Experimental operating conditions

ParameterUnitValue
Engine speedrpm1200
Net indicated mean effective pressure (nIMEP)bar8.10, 7.56
Excess-air ratio (λ)1.5, 1.65
Spark ignition timingCAD after TDC−17
ParameterUnitValue
Engine speedrpm1200
Net indicated mean effective pressure (nIMEP)bar8.10, 7.56
Excess-air ratio (λ)1.5, 1.65
Spark ignition timingCAD after TDC−17

2.2 Numerical Methodology.

In this study, two RANS combustion models widely used in the engine modeling community and typically implemented in CFD engine codes are used to simulate the ignition and combustion processes in the PCSI NG engine. The turbulent flow is modeled with the renormalization group kε model [20]. For the combustion model evaluation, one MZ-WSR model [21] and a flamelet-based G-equation model [22], are employed. Detailed information on each model can be found in the literature, and only the brief descriptions are provided here.

The MZ-WSR model assumes that each computational cell can be considered as a well-stirred reactor and solves detailed chemistry at a given temperature, pressure, and mixture composition. Here, the model utilizes a multizone method [23] to reduce the computational cost. The gri-mech 3.0 chemical kinetics [24], consisting of 53 species and 325 reactions, is used for the detailed chemistry of NG combustion.

The G-equation model tracks a turbulent premixed flame front represented by a nonreacting isoscalar surface of G(x,t)=G0. To calculate the flame propagation, Ewald and Peters [25] derived the turbulent burning velocity (st) that is valid for both large-scale and small-scale turbulence and is defined as
(1)

where l is integral length scale, lf is laminar flame thickness defined by (λ/cp)/(ρusl), lf,t is turbulent flame brush thickness, u is turbulent intensity, and sl is laminar flame speed, respectively. The laminar flame speed of NG is calculated from a flame database. Using a 1D flame simulation tool with gri-mech 3.0 chemistry, a set of unstretched, adiabatic, laminar premixed flames are solved under a wide range of conditions, 600-1200 K of unburned temperature by 50 K steps-size, 10–60 bar of pressure by 5 bar, 0.9–2.0 of the excess-air ratio by 0.1. These results are tabulated as a three-dimensional database and are looked-up based on the thermochemical conditions in each computational cell when solving the combustion models [26]. The effect of the residual gas fraction on the laminar flame speed is calculated by sl=sl,0×(12.1YRGF). On the other hand, as the G-equation model does not solve the detailed chemistry, the composition in the fully burned region is assumed to be in chemical equilibrium where only the fuel, oxidizer, and major combustion products (CO2, H2O, CO, H2) are resolved.

The combustion is initiated from the spark-plug in the prechamber at the spark timing. Both models describe the ignition as an L-shaped in power, 50 mJ of thermal energy deposition in a 0.5 mm radius of spherical volume placed between the electrode gaps. The MZ-WSR model seamlessly takes the energy deposition and the high-temperature output to calculate the reaction rate, while the G-equation model initializes the G-surface at a cutoff temperature over 3,000 K.

Simulations are performed using CONVERGE [27], a CFD tool that automates the mesh generation process and permits using simple orthogonal grids, locally forced grid embedding, and the adaptive mesh refinement algorithm (AMR). The full engine geometry, including ports and valves of intake and exhaust, is obtained, and the simulation runs for six consecutive cycles where the first cycle is discarded. In this study, a 1D model is built and validated for the given experimental data to provide reliable initial and boundary conditions, which are crank angle resolved pressure and temperature traces at the inlet and outlet boundaries. Figure 1 shows the pressure traces of the cylinder, intake runner, and exhaust runner predicted by the gt-power 1D software. There is a good agreement of the airflow rate between the 1D simulation results and measured data within 1.4% and 1.1% errors for two operating conditions, respectively.

Fig. 1
1D model validation and boundary conditions: (a) in-cylinder pressure for two operating conditions; (b) and (c) for the pressure and temperature of intake and exhaust runners (λ = 1.65 condition shown only)
Fig. 1
1D model validation and boundary conditions: (a) in-cylinder pressure for two operating conditions; (b) and (c) for the pressure and temperature of intake and exhaust runners (λ = 1.65 condition shown only)
Close modal

Figure 2 shows the full engine geometry and the computational grid configuration on a plane through the center of the prechamber. The base grid size is 1 mm and the mesh is refined up to 0.5 mm according to AMR based on gradients of temperature and velocity. Additional fixed embedding is locally set around the spark plug gap to resolve the ignition phase. Both the prechamber region, including the nozzles and the domain occupied by turbulent jets entering the main chamber region, are refined as 0.25 mm.

Fig. 2
Engine surface geometry and computational grid
Fig. 2
Engine surface geometry and computational grid
Close modal

3 Results and Discussion

3.1 Combustion Model Evaluation.

In-cylinder pressure data measured from the engine experiments and heat release rate can quantitatively characterize the combustion process and be the reference to evaluate the performance of each combustion model. Figure 3 compares experimental data and simulation results obtained at two different lean conditions (λ = 1.5 and λ = 1.65), where the MZ-WSR and G-equation models are used as is, without any model constant tuning. Both models fail to match the combustion phasing and the peak pressure value from the experimental data for the prechamber and the main chamber. In addition, as the mixture became leaner, the reduction in the peak pressure of the main chamber from the simulation is larger than what was observed from engine experiments. The simulations predict relatively low combustion rates in the prechamber so that the measured pressure difference build-up between the two chambers is not reproduced. Accordingly, the subsequent combustion in the main chamber is retarded, and the peak pressure is under-estimated by simulations. At the λ = 1.65 condition, by looking at the cylinder pressures of the individual cycle, the MZ-WSR model predicts the larger pressure difference build-up than the G-equation model. For the MZ-WSR model, the maximum pressure difference between two chambers during −15 to 0 CAD is 0.31 MPa, whereas for the G-equation model it is 0.17 MPa. Similarly, the combustion rate in the main chamber by the MZ-WSR model is higher than that by G-equation, where the former one shows a rapid pressure rise after a plateau section between −10 and 0 CAD after firing TDC (aTDCf) period.

Fig. 3
Comparison of cylinder pressure and heat release rate between experimental data and simulation results. Experimental data consists of 300 cycles data and averaged pressure; simulation results with base model configuration include five cycles data and averaged cylinder pressure. Color online.
Fig. 3
Comparison of cylinder pressure and heat release rate between experimental data and simulation results. Experimental data consists of 300 cycles data and averaged pressure; simulation results with base model configuration include five cycles data and averaged cylinder pressure. Color online.
Close modal

Mass fraction burned (MFB) timing and combustion duration can be derived from the heat release rate analysis. To fairly compare the heat release rate between experiments and simulations, the specific heat ratio (γ) is calculated with a temperature-based correlation, and the main chamber temperature is estimated by using the equation of state. Figure 4 shows that both models have doubled durations compared to experimental data in the initial phase (MFB0-2). It is interesting to note that the MZ-WSR model has a comparable duration with experiments after MFB2%, even though the in-cylinder temperature is low due to the under-estimated pressure. In contrast, the G-equation model has prolonged combustion duration at early (MFB0-2) and late (MFB50-90) combustion stages. These results are consistent with previous findings reported in Ref. [18], where the simulation was performed with a close-cycle geometry.

Fig. 4
Comparison of averaged combustion duration from experiments and simulations (base model): λ = 1.5 (box without outline) and λ = 1.65 (box with dash line). Color online.
Fig. 4
Comparison of averaged combustion duration from experiments and simulations (base model): λ = 1.5 (box without outline) and λ = 1.65 (box with dash line). Color online.
Close modal

The model constant tuning is conducted to explore the performance of each model for PCSI simulations. The reaction rate multiplier for the MZ-WSR model is tuned between 1.0 (default value) to 1.4, and b1, b3 constants of the turbulent flame speed correlation in the G-equation model are tuned between 2.0 and 1.0 (default values) to 1.5 and 2.0 for the λ = 1.5 condition, and to 1.5 and 2.5 for the λ = 1.65 condition, respectively. The b1 and b3 constant represents the contribution of large and small-scale turbulence, respectively, that is dominant inside the prechamber. Note that the purpose of the tuning exercise performed in this paper is to assess the model capability, and the level of peak pressure in the main chamber is selected as a target. Further fine-tuning will likely continue to improve the numerical results, but it is out of the scope of this study. Figure 5 shows the comparison of cylinder pressure and heat release rate between the experimental data and simulation results using the tuned models. Overall, the combustion rate of both models can be accelerated to match the measured peak pressure in the main chamber. From the averaged pressure of simulation results, The MZ-WSR model still shows an offset in the combustion phasing inside the prechamber even when it matches the peak pressure in the main chamber. Further increments of the reaction multiplier to advance the prechamber combustion will lead to a significant over-prediction of the main chamber peak pressure. As a reminder, the MZ-WSR model leverages the homogeneous reactor assumption and does not take turbulence into account in the reaction rate formulation. Hence, the lagged combustion phasing in the prechamber even with aggressive model tuning may imply that the effect of turbulence on reaction rate is responsible for adjusting the prechamber combustion phasing. On the other hand, the G-equation model takes both large-scale and small-scale turbulence into account. Indeed, the model well-matches the combustion phasing in the prechamber and the main chamber.

Fig. 5
Comparison of cylinder pressure and heat release rate between experimental data and simulation results. Experimental data consists of 300 cycles data and averaged pressure; simulation results with tuned model configuration include five cycles data and averaged cylinder pressure. Color online.
Fig. 5
Comparison of cylinder pressure and heat release rate between experimental data and simulation results. Experimental data consists of 300 cycles data and averaged pressure; simulation results with tuned model configuration include five cycles data and averaged cylinder pressure. Color online.
Close modal

Combustion duration results from the tuned models quantitatively confirm the above observations, as illustrated in Fig. 6. The MZ-WSR model has a long combustion duration in the initial phase (MFB0-2) and over-estimates the combustion rate in the main chamber for both evaluated conditions. The G-equation model has a slightly longer duration than experimental data in MFB0-2 and agrees well in MFB02-50, but the model results in doubled duration at a later combustion stage. During the MFB50-90, combustion mainly occurs in the squish region. It is likely that the flame-wall interaction is not well captured by the model and thereby results in a prolonged combustion duration. From these simulation results, we can conclude that the G-equation model is able to deliver a better agreement with the experimental data than the MZ-WSR model, in terms of combustion phasing and peak pressure in both chambers. For a fueled PCSI operation, it is expected that the fueling inside the prechamber while keeping the main chamber extremely lean would lead to more complex combustion physics and jet dynamics. The evaluation of available combustion models at those conditions will be the focus of our future work.

Fig. 6
Comparison of averaged combustion duration between experiments and simulations (tuned model): λ = 1.5 (box without outline) and λ = 1.65 (box with dash line). Color online.
Fig. 6
Comparison of averaged combustion duration between experiments and simulations (tuned model): λ = 1.5 (box without outline) and λ = 1.65 (box with dash line). Color online.
Close modal

To summarize the above observations, the timings of peak pressure in both chambers are plotted for experimental data (300 cycles) and simulation results (five cycles) in Fig. 7. In general, be it either from the experiments or the simulations, it is found that a linear correlation exists between the two timings, i.e., faster combustion inside the prechamber leads to higher peak pressure in the main chamber. However, the slope of a trend line from the MZ-WSR model tuning is quite steeper compared to that from the measured data. The MZ-WSR model tuning can accelerate the combustion rate in the main chamber, but it fails to advance the combustion phase in the prechamber at the same time. Conversely, tuning the G-equation model is effective for both operating conditions, even though the base model shows a large offset from the experiments.

Fig. 7
Comparison of timings of peak pressure in the pre- and main chamber between experiments (circle) and simulations (square, base model: filled; tuned model: line)
Fig. 7
Comparison of timings of peak pressure in the pre- and main chamber between experiments (circle) and simulations (square, base model: filled; tuned model: line)
Close modal

3.2 Analysis of Prechamber Spark-Ignition Combustion.

Based on the G-equation model results for the second cycle (i.e., the closest cycle to the average experimental pressure trace) at λ = 1.65 condition, flow and combustion features are investigated to gain an in-depth understanding of the PCSI combustion process, including the turbulent jet dynamics. In Fig. 8, it is shown that, at the time of the spark, the prechamber is mostly filled with the same mixture that is fed from the main chamber, with a small residual gas fraction at the given engine operating condition. During the compression stroke, a strong flow driven by the pressure difference between the two chambers enters into the prechamber through the orifices while producing a small length scale turbulence with high turbulent intensity. The main reason for the generation of small integral length scales comes from the small orifice diameter (d = 1.6 mm), which confines the turbulent flow into structures that are extremely small compared to conventional SI engines. As the flow time scale becomes small, the interaction between turbulence and chemistry plays a key role in the initial combustion stages.

Fig. 8
Mixture distribution and flow field at spark timing from simulation results (second cycle) at the λ = 1.65 condition
Fig. 8
Mixture distribution and flow field at spark timing from simulation results (second cycle) at the λ = 1.65 condition
Close modal

The Damköhler number (Da), which is defined as the flow time scale (l/u) over the chemical time scale (lf/sL), is a useful parameter to evaluate the turbulence-chemistry interaction (TCI) level, so it is used to analyze the combustion process. High Da values mean that the time scale of chemical reactions is shorter than the flow scale, therefore combustion rate is controlled by mixing, whereas low Da values indicate that the combustion rate is governed by chemistry. In Fig. 9, the combustion characteristics are highlighted by the isocontour of G = 0, which represents the mean turbulent flame front, and its surface is colored by Da values. After the spark onset at −17 CAD aTDCf, the flame kernel growth occurs inside the prechamber with moderate Da, and the kernel propagates toward the orifices. Then very high jet velocity and turbulence intensity lead to strong TCI with very low Da along the flame front. However, strong TCI lasts only for a few crank degrees, and fast combustion takes place to consume the main chamber mixture rapidly. Finally, the combustion proceeds on a deflagration mode at later stages. It is worth noting that the duration of MFB01-10 only takes 3.5 CAD, which is extremely fast compared to conventional SI engines in a similar lean environment.

Fig. 9
Combustion process (second cycle) at λ = 1.65 condition visualized with the isosurface of G = 0. Color online.
Fig. 9
Combustion process (second cycle) at λ = 1.65 condition visualized with the isosurface of G = 0. Color online.
Close modal

Figure 10 shows the flow field on a plane through the prechamber axis. The pressure differential pushes the unburned prechamber mixture into the main chamber before the ejection of the flame jets, with very high jet velocity, but still in the subsonic regime level due to high pressure and temperature conditions. The jet momentum as well as the strong turbulent intensity significantly promotes the combustion rate during the initial combustion stage. However, such high momentum does not last more than a few crank angle degrees, due to the small volume of the prechamber that results in a limited amount of burning mass to longer sustain the jets. Instead, the inverse pressure gradient between both chambers is developed by the ongoing combustion in the main chamber, thus a moderate backflow evolves into the prechamber while the jet momentum gradually diminishes around MFB5%.

Fig. 10
Velocity magnitude of the in-cylinder flow field on the plane through the prechamber axis from the second cycle result at λ = 1.65 condition. The isoline shows the G = 0.
Fig. 10
Velocity magnitude of the in-cylinder flow field on the plane through the prechamber axis from the second cycle result at λ = 1.65 condition. The isoline shows the G = 0.
Close modal

Leveraging the G-equation formulation, it is possible to track the turbulent combustion regime on the Borghi–Peters diagram [22]. The ratio of length scale (l/lf) and the ratio of velocity scale (u/sL) is mass-averaged on the flame front where the cell is ahead of G = 0 from −1.5 mm to −0.5 mm. From Fig. 11, it can be seen that most of the combustion processes occur in the thin reaction zone regime for the given operating points. The thin reaction zones regime is bounded by two boundaries of Karlovitz numbers, Ka = 1 and Kaδ = 1, where Ka represents the ratio between a time scale of a chemical reaction and that of Kolmogorov scale eddies as Ka=tF/tη=lf2/η2, and Kaδ is defined as Kaδ=δ2lf2/η2=δ2Ka where δ is assumed to be 0.1lf [22]. In this regime, the Kolmogorov scale eddies cannot enter into the inner reaction layer so that the reaction rate is preserved, while they can affect the preheat zone ahead of the flame. In the engine experiments, the boost pressure for intake air is kept constant so that the turbulent flow properties are almost similar in both conditions. However, as the mixture becomes leaner, the laminar flame speed decreases, and the laminar flame thickness increases. This shifts the entire trajectory slightly toward the broken reaction zone of the Borghi–Peters diagram, still fully remaining within the thin reaction zone. The combustion inside the prechamber proceeds near to the Da = 1 line, and transition into the Da < 1 region is observed at higher lambda when the jet exited from the orifices. After jets enter the main chamber, the flame front starts to interact with the large-scale turbulence as the jet momentum dissipates. Accordingly, the length scale ratio increases while the velocity scale ratio decreases. At a later stage, the trajectory makes a sharp turn with a decreasing length scale around the borderline of the corrugated flamelet regime, which is caused by the small length scale in the squish region.

Fig. 11
Analysis of PCSI combustion process along the turbulent combustion regime diagram using G-equation simulation results. Both length scale ratio and the velocity scale ratio are extracted from each computational cell ahead of the mean turbulent flame front (G = 0) in the range from −1.5 mm to −0.5 mm, and those quantities are mass-averaged over the entire flame surface.
Fig. 11
Analysis of PCSI combustion process along the turbulent combustion regime diagram using G-equation simulation results. Both length scale ratio and the velocity scale ratio are extracted from each computational cell ahead of the mean turbulent flame front (G = 0) in the range from −1.5 mm to −0.5 mm, and those quantities are mass-averaged over the entire flame surface.
Close modal

4 Conclusion

In this study, two RANS combustion models widely adopted in the engine modeling community, namely, MZ-WSR and G-equation, are evaluated to simulate the combustion process in NG PCSI engines. The NG engine runs at unfueled-PCSI under two lean operating points, λ = 1.5 and 1.65, and the cylinder, intake, and exhaust pressures are acquired and used for the assessment of the two models. The cylinder pressure, heat release rate, and combustion duration are leveraged for the model evaluation.

Using the base model configuration, none of the two models can deliver a good agreement with experiments in terms of combustion phasing and peak pressure value in the prechamber and main chamber. In general, the MZ-WSR model predicts faster combustion rates in the main chamber than the G-equation model. By tuning the models, the G-equation model provides better agreement with the experimental data than the MZ-WSR model, for both lean conditions. Model tuning can effectively advance the main chamber combustion phase. However, tuning the chemical reaction rate for the MZ-WSR model fails to match the prechamber combustion phase in the right place. It is because that the small-scale turbulence develops in the prechamber during the compression stroke, which affects the subsequent ignition and combustion processes. Since the effect of turbulence on reaction rate is not considered in the MZ-WSR model, the model cannot adjust the prechamber combustion phasing even with the aggressive tuning. Conversely, the G-equation model is equipped to take both large-scale and small-scale turbulence into account and therefore delivers much better agreement with experiments.

The combustion analysis based on the G-equation model results reveals that TCI plays a crucial role in the combustion process, especially when the jets exit from the orifices. The turbulent jets ejected from the prechamber has very high velocity but within the subsonic regime, and their momentum quickly dissipates and only lasts for a few crank angle due to the small volume (and mass) of the prechamber. This high jet velocity shifts the trajectory toward the broken reaction zone regime. However, most of the combustion lay in the thin reaction zone regime and proceeds toward the corrugated flamelet regime. The multiple combustion regimes with wide-varying turbulent flow properties suggest that a comprehensive turbulent combustion model is needed for simulating PCSI NG engines.

Acknowledgment

The submitted paper has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357. The U.S. Government retains for itself, and others acting on its behalf, a paid-up nonexclusive, irrevocable worldwide license in said article to reproduce, prepare derivative works, distributed copies to the public, and perform publicly and display publicly, by or on behalf of the Government.

This research is funded by DOE's Vehicle Technologies Program, Office of Energy Efficiency and Renewable Energy. The authors would like to express their gratitude to Kevin Stork, program manager at DOE, for his support, and the other DOE Laboratories (NREL, ORNL, and SNL) involved in the technical discussion on high-efficiency PCSI NG engines. In addition, we gratefully acknowledge the computing resources provided on Bebop, a high-performance computing cluster operated by the Laboratory Computing Resource Center at Argonne National Laboratory.

Funding Data

  • Vehicle Technologies Office (Contract No. DE-AC02-06CH11357; Funder ID: 10.13039/100011884).

References

1.
Ricardo
,
H. R.
,
1918
, “
Internal-Combustion Engine
,” US Patent 1,271,942.
2.
Abramovich
,
G. L.
,
1966
, “
Method of Prechamber-Torch Ignition in Internal Combustion Engines
,” US Patent 3,230,939.
3.
Honda Motor Co., Ltd.
,
2020
, “
Introducing the CVCC/1972
,” Honda Motor Co., Ltd., Shizuoka, Japan, accessed Oct. 20, 2020, https://global.honda/heritage/episodes/1972introducingthecvcc.html
4.
Najt
,
P. M.
,
Rask
,
R. B.
, and
Reuss
,
D. L.
,
2003
, “
Dual Mode Engine Combustion Process
,” US Patent 6,595,181.https://patents.google.com/patent/US6595181B2/en
5.
Bunce
,
M.
, and
Blaxill
,
H.
,
2016
, “
Sub-200 g/kWh BSFC on a Light Duty Gasoline Engine
,”
SAE
Paper No. 2016-01-0709.10.4271/2016-01-0709
6.
Shah
,
A.
,
Tunestal
,
P.
, and
Johansson
,
B.
,
2012
, “
Investigation of Performance and Emission Characteristics of a Heavy Duty Natural Gas Engine Operated With Pre-Chamber Spark Plug and Dilution With Excess Air and EGR
,”
SAE Int. J. Engines
,
5
(
4
), pp.
1790
1801
.10.4271/2012-01-1980
7.
National Renewable Energy Laboratory, 2020, “
Natural Gas Vehicle Research Workshop
,” National Renewable Energy Laboratory, accessed Oct. 20, 2020, http://www.nrel.gov/transportation/ngv-research-workshop.html/
8.
Chinnathambi
,
P.
,
Bunce
,
M.
, and
Cruff
,
L.
,
2015
, “
RANS Based Multidimensional Modeling of an Ultra-Lean Burn Pre-Chamber Combustion System With Auxiliary Liquid Gasoline Injection
,”
SAE
Paper No. 2015-01-0386.10.4271/2015-01-0386
9.
Gholamisheeri
,
M.
,
Wichman
,
I. S.
, and
Toulson
,
E.
,
2017
, “
A Study of the Turbulent Jet Flow Field in a Methane Fueled Turbulent Jet Ignition (TJI) System
,”
Combust. Flame
,
183
, pp.
194
206
.10.1016/j.combustflame.2017.05.008
10.
Xu
,
G.
,
Wright
,
Y. M.
,
Schiliro
,
M.
, and
Boulouchos
,
K.
,
2020
, “
Characterization of Combustion in a Gas Engine Ignited Using a Small Un-Scavenged Pre-Chamber
,”
Int. J. Engine Res.
,
21
(
7
), pp.
1085
1106
.10.1177/1468087418798918
11.
Xu
,
G.
,
Kotzagianni
,
M.
,
Kyrtatos
,
P.
,
Wright
,
Y. M.
, and
Boulouchos
,
K.
,
2019
, “
Experimental and Numerical Investigations of the Unscavenged Prechamber Combustion in a Rapid Compression and Expansion Machine Under Engine-Like Conditions
,”
Combust. Flame
,
204
, pp.
68
84
.10.1016/j.combustflame.2019.01.025
12.
Kammel
,
G.
,
Mair
,
F.
,
Zelenka
,
J.
,
Lackner
,
M.
,
Wimmer
,
A.
,
Kogler
,
G.
, and
Bärow
,
E.
,
2019
, “
Simulation Based Predesign and Experimental Validation of a Prechamber Ignited HPDI Gas Combustion Concept
,”
SAE
Paper No. 2019-01-0259.10.4271/2019-01-0259
13.
Gholamisheeri
,
M.
,
Givler
,
S.
, and
Toulson
,
E.
,
2019
, “
Large Eddy Simulation of a Homogeneously Charged Turbulent Jet Ignition System
,”
Int. J. Eng. Res.
,
20
(
2
), pp.
181
193
.10.1177/1468087417742834
14.
Syrovatka
,
Z.
,
Viteck
,
O.
,
Vavra
,
J.
, and
Takats
,
M.
,
2019
, “
Scavenged Pre-Chamber Volume Effect on Gas Engine Performance and Emissions
,”
SAE
Paper No. 2019-01-0258.10.4271/2019-01-0258
15.
Bolla
,
M.
,
Shapiro
,
E.
,
Tiney
,
N.
,
Kyrtatos
,
P.
,
Kotzagianni
,
M.
, and
Boulouchos
,
K.
,
2019
, “
Numerical Simulations of Pre-Chamber Combustion in an Optically Accessible RCEM
,”
SAE
Paper No. 2019-01-0224.10.4271/2019-01-0224
16.
Allison
,
P. M.
,
de Oliveira
,
M.
,
Giusti
,
A.
, and
Mastorakos
,
E.
,
2018
, “
Pre-Chamber Ignition Mechanism: Experiments and Simulations on Turbulent Jet Flame Structure
,”
Fuel
,
230
, pp.
274
281
.10.1016/j.fuel.2018.05.005
17.
Malé
,
Q.
,
Staffelbach
,
G.
,
Vermorel
,
O.
,
Misdariis
,
A.
,
Ravet
,
F.
, and
Poinsot
,
T.
,
2019
, “
Large Eddy Simulation of Pre-Chamber Ignition in an Internal Combustion Engine
,”
Flow Turbul. Combust.
,
103
(
2
), pp.
465
483
.10.1007/s10494-019-00026-y
18.
Kim
,
J.
,
Scarcelli
,
R.
,
Som
,
S.
,
Shah
,
A.
,
Biruduganti
,
M. S.
, and
Longman
,
D. E.
,
2019
, “
Evaluation of Combustion Models for CFD Simulation of Pre-Chamber Ignition in a Natural Gas Engine
,”
Proceedings of the 11th U.S. National Combustion Meeting
, Pasadena, CA, Mar. 24–27, Paper No. 1E18, pp.
1
11
.https://www.osti.gov/biblio/1518482-evaluation-combustion-models-cfd-simulation-pre-chamber-ignition-natural-gas-engine
19.
Nicor Gas, 2018, “
Composition of Natural Gas
,” Nicor Gas, accessed Dec. 3, 2018, https://www.nicorgas.com/business/gas-exchange.html
20.
Yakhot
,
V.
, and
Orszag
,
S. A.
,
1986
, “
Renormalization Group Analysis of Turbulence. I. Basic Theory
,”
J. Sci. Comput.
,
1
(
1
), pp.
3
51
.10.1007/BF01061452
21.
Senecal
,
P.
,
Pomraning
,
E.
,
Richards
,
K. J.
,
Briggs
,
T. E.
,
Choi
,
C. Y.
,
McDavid
,
R. M.
, and
Patterson
,
M. A.
,
2003
, “
Multi-Dimensional Modeling of Direct-Injection Diesel Spray Liquid Length and Flame Lift-Off Length Using CFD and Parallel Detailed Chemistry
,”
SAE
Paper No. 2003-01-1043.10.4271/2003-01-1043
22.
Peters
,
N.
,
2000
,
Turbulent Combustion
,
Cambridge University Press
,
Cambridge, UK
.
23.
Raju
,
M.
,
Wang
,
M.
,
Dai
,
M.
,
Piggott
,
W.
, and
Flowers
,
D.
,
2012
, “
Acceleration of Detailed Chemical Kinetics Using Multi-Zone Modeling for CFD in Internal Combustion Engine Simulations
,”
SAE
Paper No. 2012-01-0135.10.4271/2012-01-0135
24.
Smith
,
G. P.
,
Golden
,
D. M.
,
Frenklach
,
M.
,
Moriarty
,
N. W.
,
Eiteneer
,
B.
,
Goldenberg
,
M.
,
Bowman
,
C. T.
,
Hanson
,
R. K.
,
Song
,
S.
,
Gardiner
,
W. C.
, Jr.
,
Lissianski
,
V. V.
, and
Qin
,
Z.
, “GRI-Mech 3.0,” GRI-Mech, accessed Mar. 25, 2021, http://combustion.berkeley.edu/gri-mech/
25.
Ewald
,
J.
, and
Peters
,
N.
,
2007
, “
On Unsteady Premixed Turbulent Burning Velocity Prediction in Internal Combustion Engines
,”
Proc. Combust. Inst.
,
31
(
2
), pp.
3051
3058
.10.1016/j.proci.2006.07.119
26.
Pal
,
P.
,
Kolodziej
,
C. P.
,
Choi
,
S.
,
Som
,
S.
,
Broatch
,
A.
,
Gomez-Soriano
,
J.
,
Wu
,
Y.
,
Lu
,
T.
, and
Yee
,
C. S.
,
2018
, “
Development of a Virtual CFR Engine Model for Knocking Combustion Analysis
,”
SAE Int. J. Eng.
,
11
(
6
), pp.
1069
1082
.10.4271/2018-01-0187
27.
Convergent Science, Inc.,
2018
,
CONVERGE 2.4 Manual
,
Convergent Science
,
Madison, WI
.