Fuel Injection Optimization for Large-Bore Two-Stroke Natural-Gas Engines

Abstract

Recent methane emissions regulations present a challenge for the large-bore, natural-gas-fueled engines used at over 1700 compression stations across the US. Poor air–fuel mixing in the main combustion chamber of these engines results in low combustion efficiency and the resulting methane emissions. High-pressure fuel injection is believed to be a significant development in improving air–fuel mixing in natural-gas engine combustion chambers. This study aims to determine the sensitivity of in-cylinder mixing to injection pressures using Computational Fluid Dynamics (CFD) simulations, determine the limits of high-pressure fuel injection, and explore high-momentum low-pressure fuel injection. The engine, modeled using Converge Studio for CFD, was a Cooper-Bessemer large-bore, four-cylinder, GMV-4TF spark-ignited natural-gas engine with direct injection. The model was simulated for four sets of configured cases—baseline; ideal mixing; injection pressure variation; and low-pressure, high-momentum injection. The results show that fuel injection at 700 psi and −115 degrees BTDC gives the best in-cylinder mixing and improved mixing, potentially reducing methane emissions by half. The optimal timing for the injection at different injection pressures was determined. The level of mixing in low-pressure fuel-injection systems was also improved by the high-momentum fuel injector design. It was concluded that mixing can be further improved in integral gas compressor engines through fuel injection optimization.

1. Introduction

Natural-gas pipelines run along interstate and intrastate pipelines in the United States, with a total of 1700+ compressor stations along these pipelines [1]. At each of these compression stations, stationary integral compressor engines are used to compress natural gas up to 800–1200 psi and keep the gas moving through the pipelines. There are more than 7000 of these engines scattered across the United States, fueled with 2–3% of the natural gas they compress. They are either lean burn or rich burn, and two-stroke or four-stroke [2]. The integral compressor engines are usually two-stroke, lean burn, large-bore (≥0.3 m), slow-speed (≤500 rpm) engines. Unburned methane escapes through the exhaust and crankcase of large-bore natural-gas-fueled engines due to poor combustion efficiency. Natural gas primarily contains methane, and not all the methane from the engine fuel (natural gas) undergoes complete combustion in the engine cylinder [3]. The inefficient combustion in large-bore engine cylinders has been associated with poor mixing within the cylinder at ignition. There have been efforts to enhance in-cylinder mixing in engines, and improved combustion variability and performance have validated the viability of improved in-cylinder mixing as a retrofit technology for these engines [4]. Work aimed at enhancing mixing in the cylinder of large-bore engines has been performed using experimental and numerical means [5], but most experimental studies relating to mixing have focused on optimizing the mixing control without quantifying the level of in-cylinder mixing [6,7]. The commercial CFD software Converge version 3.0 is utilized for numerical simulations to visualize in-cylinder activity. Converge for CFD is an appropriate tool critical for understanding the complex fluid dynamics of the air–fuel mixture ignited in the cylinder of large-bore natural-gas engines and improving their operation, and has been used in the present study [8].

Combustion inefficiency in the main chambers of large-bore natural-gas engines is the major source of methane emissions associated with these engines [9] and is an indicator of poor performance. Some reasons for poor combustion efficiency include bulk flame quenching, fuel short-circuiting when intake and exhaust ports overlap, air–fuel mixture trapped in the piston top land crevice, and methane stored in the piston ring-pack [10,11,12]. Flame quenching occurs in lean burn natural-gas engines when the bulk charge is extinguished away from the combustion chamber surfaces at quench zones because the air–fuel mixture is very close to the lean combustion limit, and the flame front becomes over-stretched [3,13]. When fuel is injected while the exhaust ports are still open, some unburned fuel can escape as methane emissions due to short-circuiting. Lean burn engines employ combustion at lower combustion temperatures than rich burn engines, leading to incomplete oxidation of the fuel compounds, of which methane is, on average, 90%. Poor air–fuel mixture in the main combustion chamber due to pockets of rich or lean mixtures results in uneven combustion. To improve in-cylinder mixing, high-pressure fuel injection has been explored as a solution [14].

On the basis of previous studies and experiments, including those conducted at Colorado State University (CSU), the potential for high-pressure fuel injection to improve in-cylinder mixing has been established [15]. In a study by Hoffmann et al., the influence of fuel system pressure increases in a gasoline direct ignition (GDi) spark ignition (SI) engine was examined [16]. The results from that study show that hydrocarbon emissions were reduced for each of the three injectors which were examined with increasing fuel pressure. This illustrates the application of high-pressure fuel injection. A different study investigated the process of in-cylinder flow and mixture preparation in a similar high-pressure fuel-injection SI engine [5]. The authors of that study concluded that injection timing significantly influenced the mixture formation and that the in-cylinder flow was intensified because of increased momentum stemming from the high pressure. The present study uses Converge v 3.0 [17] to examine the sensitivity of fuel-injection pressure to mixing in large-bore integral compressor SI engines. The adaptive mesh refinement of Converge provides it with leverage in comparison to the many other commercial software applications used for CFD simulations and ensures more accuracy in the present study on in-cylinder mixing in lean burn large-bore engines. The scope of this study spans establishing a baseline case; investigating the case of uniform heterogeneity to determine optimal methane emissions points; and then varying the injection pressure cases, including flow area considerations for lower pressures. This study aims to determine the limits of high-pressure fuel injection, optimize injection timing for various injection pressures, and improve low-pressure fuel injection through fuel injector design for increased momentum. The results from this study will help as an approach to better combustion efficiency in the lean burn engines used in the natural-gas industry and subsequently reduce methane emissions.

2. Materials and Methods

The Cooper-Bessemer GMV-4TF natural-gas engine was modeled using SolidWorks 2023 for Computer-Aided Drawing (CAD) and imported into Converge for the performance of CFD simulations. It is a slow-speed (300 rpm) 2-stroke cycle engine with a 14 in. (355.6 mm) bore and 14.75 in. (374.7 mm) stroke [2], which is typical of many stationary integral compressor engines used in natural-gas compression stations. The features of this test engine are shown in

Table 1. Test engine technical parameters.

Parameters	Content
Engine type	Two-Stroke
Air-to-fuel ratio (AFR) classification	Lean Burn
Number of cylinders	4
Combustion chambers	Main Combustion Chamber (MCC) and Pre-Combustion Chamber (PCC)
PCC control valves	Mechanical Valves
Fuel	Natural Gas
Fuel-injection configuration	High-Pressure Fuel-Injection Mode
Rated power (bKw)	330
Rated speed (rpm)	300
Piston crown shape	Dome-Shaped

.

For the CFD simulations, four model configurations were used:

• The baseline case;
• The ideal uniform mixing case;
• Various injection-pressure cases (varied during previous experimental studies using the Siemens HPFI valve) at a range of injection-timing values;
• Adjusted valve flow area to increase momentum at low injection pressures.

The results from these configurations served as the deliverables. Each case was configured for high-pressure fuel injection, and parameters such as injection pressure and injection timing were varied to investigate optimal mixing points, while monitoring key parameters such as fuel slip, pollutant emissions, and cylinder pressures. The engine, with its four cylinders (two currently visible), is shown in Figure 1. The external view of the physical fuel valve that was modeled during the simulations is shown in Figure 2. The HPFI valve is mounted on the cylinder head, replacing the original valves previously used with lower pressure limits. The fuel valve was configured for high-pressure fuel injection, but the cylinder (part number: GMVH-11-1B 01300) of the Cooper Bessemer GMV engine [engine manufacturer: Cooper Machinery Services, Houston, TX, USA] and cylinder head used (part number: GMVH 11-6A) were proper to the original equipment manufacturer (OEM) configurations.

To achieve the goals of this study, which involves an in-cylinder investigation, CFD was employed rather than experimental investigations. Using CFD also allowed for the attainment of engine conditions that would otherwise be difficult to reach during experimental testing, such as the uniform mixing scenario [3]. The repeatability and ease of control when it comes to the use of computational models in CFD with respect to this study also presented an advantage [8]. To this end, Converge CFD Studio version 3.0 [17] was used throughout the study. The software’s features, such as the SAGE detailed chemistry solver, adaptive mesh refinement (AMR), and fixed embedding (FE), were applied to each case. Vortex shedding (swirl) is not specifically considered in this study, though the CFD model inherently includes such effects. Only the impact of fuel injection on methane emissions and other engine parameters is evaluated.

2.1. Nominal High-Pressure Fuel Injection CFD Baseline

High-pressure fuel injection has been explored in previous studies and its benefits have been established [4,16]. However, creating a baseline case for the GMV-4TF engine was crucial for this study for effective comparison with the subsequent improved-mixing cases. One cylinder was modeled, representing each of the four cylinders of the GMV-4TF engine. Specifically, cylinder #2 was utilized. Data used for validation, such as piston motion and combustion pressure data, came from cylinder #2. Measurements were obtained from the GMV-4TF engine located at the Powerhouse Energy Campus [18].

The baseline case was split into two sections, namely, the combustion cycle model, and the scavenging cycle model, to reduce computation time. Figure 3 shows the scavenging cycle model, which has all the features of the combustion process, as well as geometry related to the scavenging process such as the intake and exhaust ports, pre-chamber nozzle, PCC fuel inlet, and main chamber fuel-injection valves. The components are listed as the boundaries and classified under regions as shown in Figure 4. The combustion cycle model geometry contains only parts related to the combustion chamber, which are the cylinder head, pre-combustion chamber (PCC), spark plug, piston, and cylinder liner, as seen in Figure 5 and Figure 6. The two baseline case models had a constant NOx value of 0.5 g/bhp-hr, with the OEM PCC included in the computer-aided drawing (CAD) model. In these models, fuel was injected at a nominal value for high-pressure fuel injection: 500 psi, and a nominal injection timing of 115 degrees BTDC. Engines used in the field typically inject fuel at 20–50 psi, and 500 psi is considered high, making this a high-pressure fuel-injection study. The combustion model was validated with the most recent existing experimental combustion data and matched in terms of the location of peak pressures (LoPP). Figure 7 shows the top view of the intake bores, which are in pairs, and the identical exhaust ports. Figure 8 shows a cross-section of the scavenging cycle geometric model when the piston is at the top dead center (TDC) and at the bottom dead center (BDC). The main parameters of the model are shown in

Table 2. CFD baseline geometric model parameters.

Parameters	Dimensions
Bore—in (mm)	14.00 (355.6)
Stroke—in (mm)	14.75 (374.7)
Connecting rod length—in (mm)	35.1 (892)
Crank speed (rpm)	299.8
PCC volume (in3)	3.46
PCC nozzle diameter—in (m)	0.32 (0.008)
Number of intake ports	8
Intake bores height—in (m)	2.83 (0.072)
Intake bores set 1 width—in (m)	2.80 (0.071)
Intake bores set 2 width—in (m)	2.60 (0.066)
Intake bores set 3 width—in (m)	2.48 (0.063)
Intake bores set 4 width—in (m)	2.20 (0.056)
Number of exhaust ports	5
Exhaust ports height—in (m)	4.29 (0.109)
Exhaust ports width—in (m)	2.20 (0.056)

.

The combustion model was tuned to match the experimental data using the engine CFD model configuration [18]. The combustion pressure was tuned by varying the ignition timing. A stable grid of 8 mm in the Cartesian x, y, and z axes was determined for tuning, and a final grid size of 4 mm and 8 mm was used in the combustion and scavenging models, respectively, based on the complexity of each cycle. The Courant–Friedrichs–Lewy (CFL) limits were determined to imitate the engine operation. A variable time-step selection was applied, with the initial, minimum, and maximum time steps being 5 × 10⁻⁷, 5 × 10⁻⁷, and 2 × 10⁻⁵, respectively. The solver was set to use successive over-relaxation (SOR) methods to evaluate momentum, pressure, density, energy, species, turbulent kinetic energy (TKE), dissipation rate, elliptic relaxation function, velocity scale ratio, radiation, and wall distribution equations. SOR was a precursor for the in-built Converge BICGSTAB solver for wall distribution equations. Berkeley’s GRI 3.0 chemical mechanism [19] was also applied to the two models. For the scavenging cycle case setup, there were 4 elements (C, H, O, and N) and 8 major species—O₂, N₂, CO, CO₂, CH₄, C₂H₆, C₃H₈, and H₂O. However, the mechanism contained 5 elements, 573 species, and 375 reactions for the combustion cycle. The natural-gas fuel modeled in this study for both the PCC gas line and main injector was 68.7% methane (CH₄), 18.7% ethane (C₂H₆), 7.2% propane (C₃H₈), 4.9% carbon dioxide (CO₂), and 0.5% Nitrogen (N₂) by mass fraction, based on sample data from the natural-gas supply at the Powerhouse Energy Campus.

2.2. Ideal Uniform In-Cylinder Mixing at Spark

The next case after the baseline was established, focused on forcing the measure of heterogeneity (∅_SD) in the cylinder to be at a value of zero, implying an ideal uniform mixture. This parameter measures the standard deviation of the equivalence ratio through all the spatial points in the cylinder and reports the degree of heterogeneity within the engine’s main chamber. It indicates how much the equivalence ratio at different spatial points in the 3D simulation differs from the average equivalence ratio and is defined as

$${\mathrm{\varnothing }}_{SD}={\displaystyle \frac{\sqrt{{\sum }_i^n{({\mathrm{\varnothing }}_{ave}-{\mathrm{\varnothing }}_i)}^2}}{n-1}}$$

where ∅_ave is the average equivalence ratio of spatial points in the 3D model, ∅_i is the equivalence ratio at each spatial point, n is the number of spatial points, and i indicates each spatial point.

The standard deviation values range from zero (0) to one (1) in the cylinder with values that tend toward one (1) indicating poor mixing and those tending to zero (0) indicating improved mixing. A value of zero therefore indicates perfect mixing. This is an ideal, and cannot be attained by the engine under normal operating conditions. However, it gives a reference point of the lower limit of what is attainable concerning methane emissions via improved mixing. The case was set up using the same conditions as the baseline—the same mass flow rate and other injection conditions, with the only change being the ideal measure of heterogeneity. It was tuned to have the same location of peak pressure, peak pressure, and work per cycle as the baseline. The fuel-injection pressure in this model remained at 500 psi (3.45 MPa), and the injection timing was 115 degrees BTDC. The results from this uniform mixing set-up provided a basis for comparison for subsequent cases but were first compared with the baseline.

2.3. Injection Pressure Variation

Injection fuel pressure variation was conducted to investigate high-pressure fuel injection as a strategy to improve in-cylinder mixing. The fuel-injection pressures for the five (5) CFD cases used for the injection pressure variation were 150 psi, 300 psi, 600 psi, 700 psi, and 800 psi. These are in addition to the 500 psi baseline. Each of these cases was run using different injection-timing modes.

2.3.1. Constant Injection Timing

The first variation of injection pressures was carried out using the same injection timing, 115 degrees BTDC, which is the nominal injection timing for the baseline 500 psi fuel-injection simulation. This was the first run, and results were recorded. However, the injection timing was not optimal for some of the injection pressures, leading to the next set of cases.

2.3.2. Optimal Injection-Timing Investigation

Cases were set up using the matrix shown in

Table 3. Optimal injection-timing investigation simulation matrix.

Injection Pressure (psi)	Injection Timing (Degrees)
150	−145	−135	−125	−115
300	−145	−135	−125	−115
500			−125	−115	−105
600			−125	−115	−110	−105
700			−125	−115	−110	−105
800			−120	−115	−110	−105

to obtain the optimal injection timing for each injection pressure. Each injection pressure was examined at various injection timings to identify the best mixing case, but attention was also paid to cases where fuel slip occurred, and such cases were discarded.

After the investigation was completed, the cases with the optimal fuel-injection timings for each injection pressure were compared. This gave an insight into the trend of mixing within the cylinder as injection pressure increased.

2.3.3. High Momentum Cases for Lower Pressures

The injection duration for the cases examined so far varied inversely with the pressure. This direct relationship between the fuel-injection pressure and the injection duration led to very long injection durations for the 150 psi and 300 psi cases. The 150 psi fuel injection lasted for 66.67 crank angle degrees (37 ms) while the 300 psi fuel injection lasted for 33.33 crank angle degrees (18.5 ms). The nominal fuel injection in these simulations at 500 psi lasted for an injection duration of 20 crank angle degrees (11.1 ms). Considering that the fuel-injection times for the lower pressures, particularly that of 150 psi, are much longer than the nominal fuel injection, the flow area of the fuel-injection valves for the lower pressures was increased, to maintain a constant fuel-injection duration and make comparisons accordingly. For the increased flow area, the new geometry implied an increased momentum for the low-pressure case, but a constant mass of fuel injected per cycle. The changes listed in

Table 4. Dimension comparison for nominal fuel valve design and low pressure increased flow area cases.

Parameter	Nominal Dimension in (mm)	New Dimension in (mm)
Shroud diameter	0.756 (19.2)	0.861 (21.9)
Poppet diameter	0.709 (18.0)	0.709 (18.0)
Maximum valve lift	0.025 (0.64)	0.086 (2.18)

below were developed by applying scale factors to the shroud diameter, poppet diameter, and valve lift for the fuel valve in the engine model. This resulted in the new calculated flow areas detailed in

Table 5. Increased flow areas for low pressure injection.

Parameter	Nominal Flow Area in2 (mm2)	New Flow Area in2 (mm2)
Seat-flow area	0.029 (18.9)	0.097 (62.9)
Shroud-poppet annulus	0.054 (35.1)	0.187 (121)
Port annulus	0.098 (63.0)	0.146 (94.3)

.

3. Results

3.1. Baseline Case

The baseline case was set up in Converge Studio v3.0 using a previously developed geometric model with validated parameters from the GMV-4TF. The geometric model used is shown in Figure 9 and the dimensions are as described in the preceding section of this paper. The baseline combustion pressure trace is shown in Figure 10, combined with experimental data from a previous study [18] focused on the location of peak pressure. The location of peak pressure (LoPP) for the mean experimental cycle was 18.2 degrees ATDC, while that of the baseline simulation was 18.4 degrees ATDC. The maximum cylinder pressures of the baseline simulation and experimental data average were 511.3 and 548.9 psi, respectively. This difference in combustion peak pressure can be attributed to many factors, such as the combustion wall temperature assumption and differences between the chemical kinetic mechanism behavior and actual combustion kinetics. However, the combustion pressure trace is acceptable given the closely matching values of LoPP from the experimental and computational studies.

3.2. Ideal Uniform Mixing Results

The ideal mixing case shown in Figure 11 shows a perfect mixing at spark (−1.5 degrees ATDC). The model used here to visualize events in the engine cylinder during the combustion process was the combustion cycle model. The cylinder pressure trace was tuned to match the baseline regarding the location of peak pressure and peak pressure. Compared to the baseline case, this setup shows better mixing and lower values of heterogeneity, represented by the equivalence ratio standard deviation from spark to the end of the cycle. The most important point to compare fuel–air mixing heterogeneity is at spark. As combustion ensues this parameter is less meaningful.

Figure 12 shows the effect of the uniform mixing at the spark on residual methane in the cylinder. When the combustion occurs from the point of perfect mixing, the amount of unburned methane left in the cylinder at the end of the cycle is reduced by more than half (from 2.29 × 10⁻⁵ kg to 1.06 × 10⁻⁵ kg). This result lowers the methane emissions limit for the improved mixing in the cylinder and shows the potential for the next set of simulations by validating the reduction in methane emissions with enhanced mixing. It indicates that there is an upper limit for how much improved mixing can reduce methane emissions.

3.3. Injection Pressure Variation

The fuel-injection pressures used for this set of simulations varied from low to high pressures. The first runs of fuel-injection simulations were conducted at −115 degrees ATDC and there was a clear trend of improvement in in-cylinder mixing. However, this injection timing was the optimum injection timing for fuel injected at 500 psi and suboptimal for the other injection pressures. An optimal injection-timing investigation was then carried out on the other pressures, 150, 300, 500, 600, 700, and 800 psi. The results of the optimal injection-timing variations are shown in Figure 13.

In Figure 13, the points with the best mixing can be identified for each injection pressure. However, for the 150 psi and 300 psi cases, there were cases with significant fuel slips out of the exhaust ports. Figure 14 shows the fuel slip for 150 psi injection at −145 degrees ATDC. Figure 15 shows the baseline case (500 psi fuel injection at 115 degrees BTDC) with a less than significant fuel slip. Comparison of Figure 14 and Figure 15 shows that after an initial insignificant slip from the pre-chamber in both cases, early fuel injection leads to a significant amount of slip for fuel injected at 150 psi at −145 degrees (Figure 14) as opposed to no slip in the baseline 500 psi (Figure 15). The 150 psi fuel-injection cases had 1.6% fuel slip when the injection timing was at −145 degrees ATDC. At 300 psi there was a 10.3% fuel slip when natural gas was injected at −145 and a 0.6% fuel slip for injection at −135 degrees ATDC. In these cases, the predominant factor that caused this fuel slip was not the injection pressure, but the injection timing. When fuel is injected too early, some of the unburned fuel can escape through the exhaust ports, leading to additional methane emissions.

Figure 16, Figure 17, Figure 18, Figure 19, Figure 20 and Figure 21 show binning plots for the optimal mixing points at each injection pressure. The mean equivalence ratio is represented by the tallest bins and the variation of the equivalence ratio across the mass of the air–fuel mixture in the engine cylinder is illustrated by how close other bins are to the level of the tallest bin. To understand the bin plots, it helps to examine extreme cases. For perfect mixing there would be only one bin that encompasses the average equivalence ratio, with a mass fraction of one. An extremely poor mixing case would show all bins at the same height, spread from across the equivalence ratio range 0–1. Fuel injection at 700 psi gave the best mixing at ignition (−1.5 degrees ATDC). More concisely, Figure 22 shows the summary of the optimal mixing points for each of the injection pressures. The general trend of the optimal points is a downward trend in the direction of improved mixing with increasing pressures. The first spike, from the equivalence ratio standard deviation for 150 psi to that of 300 psi, is because of the limits of injected fuel slip. There is a significant improvement in mixing at 700 and 800 psi compared to the nominal injection pressure of 500 psi. However, the last slight increase in standard deviation at 800 psi requires more investigation in further studies.

3.4. Low Pressures and High Flow Area Injection

The injection of fuel at 150 psi and 300 psi (low pressures) was simulated using the same geometry as the higher pressures of 500 psi, 600 psi, 700 psi, and 800 psi. However, the constant flow area used for these higher-pressure cases made the injection duration for the low pressures relatively very long. To maintain the same injection duration while keeping mass flow constant by increasing the momentum, the flow area for low-pressure fuel injection was increased by applying scale factors to the shroud diameter, poppet diameter, and lift of the fuel valve, as shown in Figure 23. The mass of fuel delivered is shown in Figure 24 for the baseline case (500 psi injected at 115 degrees BTDC), 150 psi injection with the same geometry, and 150 psi injection with an improved geometry–high momentum case. Figure 24 shows that the high-momentum case was matched very closely with the baseline in terms of injection duration and the mass of fuel delivered to the main cylinder of the engine model.

The result in Figure 24 allows for a fair comparison of the two injection pressures. The results in Figure 25 show that the case with an increased flow area has a 39% improvement in mixing at the start of ignition (1.5 degrees BTDC) when compared to the low-pressure (150 psi) case at the same injection timing (115 degrees BTDC) using the initial geometry. The increased flow area case has a standard deviation of equivalence ratio at the ignition spark that is closer to the baseline 500 psi case. This is due to the higher momentum that the low-pressure injection (150 psi) had with the new geometry–increased valve flow area—compared to that of the initial geometry, which impacts the mixing in the cylinder.

4. Conclusions

This study on improved in-cylinder mixing showed that mixing is sensitive to changes in injection pressure. The results from the study were gathered based on the baseline CFD case results, which served as a reference for all other cases. The lowest attainable amount of methane emissions based on the ideal, or perfect, mixing case was identified as being about half of the baseline methane emissions. Beyond the ideal case, varying the injection pressures indicated improved in-cylinder mixing with increasing injection pressures. The key findings from this study include the following:

• Improving the level of in-cylinder mixing in large-bore engines reduces methane emissions, with a maximum attainable reduction of 50%. This emissions reduction effect is due to the improved combustion efficiency that reduces the amount of unburned methane that can escape into the exhaust during scavenging.
• Increasing the injection pressure of fuel in large-bore engines improves the mixing of air and fuel in the main combustion chamber, up to 800 psi, where it slightly declines. The more obvious trend of improved mixing is caused by the high momentum that is associated with the air–fuel mixture when fuel is injected at higher pressures. Further investigation needs to be conducted on the flow field in high-pressure fuel injection to understand the reason for the final decline in mixing (increase in heterogeneity).
• Increasing the flow area in the fuel-injection valve design and geometry for low injection pressures such as 150 psi improves the mixing in the engine cylinder at that same pressure—almost the same mixing as the 500 psi baseline case. The increased flow area matches the injection duration and mass of fuel delivered and allows for increased momentum when fuel is admitted into the cylinder.

This study identified two approaches for improving mixing: increased injection pressure and increased flow area. Both increase the momentum of the fuel jet and improve mixing. There are opportunities to reduce the cost of enhanced mixing and to reduce methane emissions by achieving better mixing than the 500 psi high-pressure fuel-injection (HPFI) baseline. Further studies should investigate medium-pressure high-momentum fuel injection experimentally by using the optimized injection valve proposed for 150 psi injection in this study.