Conversion of a turbocharged gasoline engine to hydrogen operation

Abstract

Fuel cell electric drivetrains, battery electric drivetrains and drivetrains with hydrogen-fueled combustion engines are subject to scientific and political debate. They all have in common that they operate without emitting tank-to-wheel greenhouse gases. The extent to which the use of a specific solution is sensible depends heavily on the application context. In this work, a series production 1.0 L passenger car gasoline engine was converted to hydrogen operation through minor modifications. A lean-burn concept was applied to increase efficiency and mitigate the engine’s nitrogen oxide output. Comparing the combustion process of the gasoline to the hydrogen engine, similar levels of burn velocity, rate of pressure rise, and cyclic variations were observed. However, the series charging system could not meet the air demand of the hydrogen engine. Especially in the low speed range, the exhaust gas enthalpy in lean-burn hydrogen operation is significantly lower than in stoichiometric gasoline operation. As a result, the maximum low speed engine load was limited due to the rapid approach to close-to-stoichiometric operation and the resulting exponentially high NOx emissions. To improve the full load operation in terms of torque output and emissions, the intake and exhaust camshaft phasing was optimized, albeit with slight losses in efficiency.

Keywords

hydrogen internal combustion engine turbocharging port fuel injection NOx emissions variable valvetrain full load optimization

Highlights

Minimum modifications to the engine hardware

Adapting the charging system is crucial to achieve high loads

Camshaft phasing promotes leaner mixture by increasing the air flow rate through the engine

By camshaft phasing and combustion phasing, ISNOx emission can be kept below 1 g/kWh throughout the entire engine map

Introduction

By 2015, the United Nations had already recognized the importance of reducing greenhouse gas emissions in the Paris Agreement.¹ With a share of 25.5% in 2022, the transport sector is still a major source of greenhouse gas emission in the European Union.² In order to reduce this, drive systems and carriers for renewable energy have to be found. Although battery electric vehicles have a higher tank-to-wheel efficiency,³ chemical energy storage (in the form of hydrogen) can offer advantages in terms of driving range, refueling, cost and curb weight.⁴ For the conversion to mechanical energy, either hydrogen fuel cells or hydrogen internal combustion engines (ICE) are a sensible option. Fuel cells have lower process-inherent losses and can therefore achieve higher peak efficiencies.⁵ But since ICEs are a well-developed and cost-efficient⁶ technology, the production infrastructure is available,⁴ they are flex-fuel capable⁷ and they offer the possibility to convert existing engines to hydrogen operation with limited effort,⁸ they are a serious alternative. In 2000, BMW made their H₂ICE powered 750 hL world-famous, by using them as shuttles at ‘EXPO 2000’. However, the torque of the naturally aspirated hydrogen engine was almost 40% lower than that of the gasoline counterpart.⁹ This can only be compensated with large displacements or engine boosting.¹⁰ It has been proven that turbocharged hydrogen engines can reach a similar or higher torque output as gasoline engines with the same displacement.¹¹ With hydrogen being a carbon-free fuel, the only noteworthy emissions are nitrogen oxides (NOx). When it comes to the combustion process, lean-burn operation has proven to be very promising. It leads to suppression of abnormal combustion and an increase in efficiency.¹² Additionally, it comes with the advantage that the endothermic NOx formation process is reliably inhibited due to the low combustion temperatures involved.¹³ Therefore, hydrogen ICEs without the need for an exhaust gas aftertreatment system are conceivable.¹⁴ However, lean operation demands a highly capable charging system, as it combines a high air demand with comparably low exhaust gas temperature, which results in low exhaust gas enthalpy. In addition, the air displacement effect caused by the injection of gaseous, low-density hydrogen into the intake port further aggravates this issue.¹⁵

Prerequisite for wide-spread application of H₂ICE is a stable combustion process, low emissions, high efficiency and high torque output. In this work, the operational characteristics of a turbocharged gasoline direct injection (GDI) engine were compared to its hydrogen port fuel injection (H₂PFI) equivalent. The remainder of this work is organized as follows: Section Test Objects presents the two different engine setups and the conversion process of the GDI engine to H₂PFI operation. Section Test Bench Setup, Instrumentation, Test Procedure outlines the test-bench instrumentation and test procedures. In section Experimental Investigations, results of different experiments made with both engine setups are discussed. Ultimately, the conclusions of the study are presented in section Summary and Conclusion.

Test objects

A series production three-cylinder turbocharged GDI passenger car engine was used for the experimental investigations. Firstly, it was analyzed in series gasoline operation without any hardware modifications. Secondly, the engine was converted to H₂PFI operation. This required both mechanical and engine control modifications. A baseline investigation was conducted. Thirdly, a full load optimization procedure was carried out on the hydrogen engine.

Series gasoline engine

This work utilizes the series engine type ‘Fox 1.0 L Ecoboost GTDI’ from Ford Motor Company as base engine. It is a three-cylinder inline, four-stroke, water-cooled, gasoline engine with turbocharger and charge air cooler. A fully independent double overhead camshaft with variable valve timing (VVT) drives two intake and exhaust valves per cylinder. Fuel injection is designed as high pressure direct injection. The injector and the spark plug are mounted centrally in the combustion chamber. Instead of the series engine control unit (ECU), the Schaeffler ‘PROtronic Topline’, a freely programmable ECU, was used to control the engine in gasoline operation. The ECU calibration was made according to a previous publication.¹⁶ Table 1 shows further technical data.

Table 1.

Technical data of the series production gasoline engine.¹⁷

Ford Fox 1.0 L GTDI ecoboost
Displacement (cm³)	999
Bore × stroke (mm × mm)	71.9 × 82.0
Bore spacing (mm)	78
Compression ratio (–)	10
Max. torque at speed (Nm)/(min⁻¹)	170/1500–4500
Max. power at speed (kW)/(min⁻¹)	92/6000

Converted hydrogen engine

To run on hydrogen, the series gasoline engine was modified as follows:

Removal of the direct injectors and high pressure pump and application of a gaseous port fuel injection with two commercially available compressed natural gas (CNG) injectors per cylinder.

Replacement of the ignition coils with prototype hydrogen ones.

Adaptation of the engine control for quality-controlled instead of quantity-controlled operation.

Quality-control refers to load regulation by adjusting the injected fuel mass, while the air mass results from the air-path setup, as applied in diesel engines. In contrast, quantity-control denotes load regulation by varying the air mass, typically through throttling, as used in gasoline engines.¹⁸

Fuel injection system

Figure 1 shows the designed H₂PFI injection system. It consists of an aluminum port extension which carries two injectors per cylinder to meet the hydrogen flow requirement. It was installed between the cylinder head and the series intake manifold and allows fuel to be injected just before the intake valve. Care was taken to retain the existing runner layout. The hydrogen gas is supplied via an aluminum fuel rail. The fuel injectors used (Bosch, 0280158862) are commercially available PFI injectors, which were originally designed for CNG injection. At the time of the work, no dedicated hydrogen injectors were commercially available to the authors.

Figure 1.

Hydrogen gaseous PFI injection system.

Ignition system

In hydrogen engines, ignition coils can be a major source of backfire due to undesired electric discharge during the intake stroke.¹⁹ Therefore, dedicated prototype hydrogen coils were used to avoid this effect. The series spark plugs were retained, but the spark plug gap was reduced from 0.7 to 0.3 mm to avoid misfiring during highly boosted operation.

Engine control

Hydrogen engines can be considered as a hybrid of gasoline and diesel engines, as their qualitative load control is more similar to diesel engines, while the flame front combustion and the ignition system are similar to gasoline engines. Ultimately, a gasoline motorsport ECU (Bosch MS6.3) was used to control the engine. To achieve quality control, the ECU was operated with load control via injected fuel mass. The electronic throttle was controlled by an external driver device. This makes it possible to operate the engine with independent control of the air and fuel mass.

Test bench setup, instrumentation, test procedure

Test bench setup

The engine was operated in the engine laboratory’s hydrogen test facility at Heilbronn University of Applied Sciences (Heilbronn, Germany). An eddy current brake was used to achieve constant speed operation.

Instrumentation

The test facility featured an indication system with crank angle synchronous data acquisition, as well as a time-based low-frequency data acquisition system.

The indication system used for combustion analysis was the FEVIS by FEV. The crankshaft angular position was measured with a Bosch HA-P2 hall sensor mounted to a 60–2 teeth trigger wheel in the crankshaft pulley. To increase the resolution of the signal, a D2T MIO F01 pulse multiplier was used. Thus, the crank position could be known with a resolution of 0.1° crank angle (dCA). The indication system recorded in-cylinder pressure for all cylinders as well as intake and exhaust port pressure for cylinder 3. In-cylinder pressure was obtained with water-cooled piezoelectric pressure transducers (Kistler, type 6041A), connected to Kistler 5064B charge amplifiers. Port pressures were measured with piezoresistive pressure transducers (intake: Kistler 4045A, exhaust: Kistler 4049B), connected to Kistler 4665B amplifiers.

The time-based acquisition system used type-K thermocouples for temperature measurement and Keller type PAA-33X transducers to obtain absolute pressures. Relative pressures were measured using Keller type PR-33X transducers. The air mass flow rate was measured by an ABB Sensyflow FMT700-P, mounted upstream of the compressor. A Coriolis mass flow meter (Endress + Hauser, Promass A300) measured the hydrogen mass flow. Downstream of the turbine, a gas analyzer (Ecophysics, CLD700) obtained the engine-out NOx emissions. Figure 2 presents a schematic overview of the engine instrumentation. The measurement uncertainties of the main equipment used in the test cell to carry out this work are provided in Table 2.

Figure 2.

Schematic of the test engine instrumentation setup.

Table 2.

Measurement uncertainties of the test cell equipment. ‘FSO’ denotes full scale output.

Measurement device	Measurement uncertainty
Piezoelectric pressure transducer (in-cylinder pressure)	$\pm 0.5 % FSO$
Piezoresistive pressure transducer (intake port pressure)	$0.3 % FSO$
Piezoresistive pressure transducer (exhaust port pressure)	$0.1 % FSO$
Thermocouple type K	$\pm 2^{°} C$
Absolute pressure transducer	$0.1 % FSO$
Relative pressure transducer	$0.1 % FSO$
Thermal mass flow meter	$0.3 %$
Coriolis mass flow meter	$0.35 %$
Exhaust gas analyzer	$\pm 1 % FSO$

Test procedure

Steady state tests were performed with closed-loop temperature control for coolant (90°C, downstream of the engine) and charge air cooler (45°C). The low-frequency signals were sampled at 5 Hz and averaged over the measurement period of 30 s. The indicated data was acquired and averaged along 100 engine cycles for all three cylinders.

During the tests, both the series gasoline and the hydrogen engine were controlled by adjusting the control variables manually at each stationary operating point in the calibration software of the respective used ECU.

Boundaries

Gasoline engine

To control the load of the gasoline engine, a combination of throttle and wastegate (WG) actuation was applied (Figure 3). In the low to medium load range, load was controlled by throttling only, while the WG was kept fully opened. Between 10 and 14 bar indicated mean effective pressure (IMEP), load control is transitioned to WG actuation with wide open throttle (WOT). For loads below 10 bar IMEP, the start of fuel injection (SOI) was kept at 320 dCA before top dead center (BTDC). Beyond 10 bar IMEP, SOI was advanced from 290 to 315 dCA BTDC with rising engine speed to avoid fuel impingement on the piston crown. The fuel injection duration resulted from the required fuel mass to keep the relative air-fuel ratio ( $λ$ ) stoichiometric. The ignition angle was adjusted so that 50% fuel mass fraction burned (50% MFB) was reached at eight dCA after top dead center (ATDC) where applicable (Figure 4). In the high load range, the ignition had to be retarded to suppress knocking. As already mentioned, the ECU calibration in gasoline operation was made according to a previous publication.¹⁶ This also includes the VVT setting.

Figure 3.

Visualization of the gasoline engine’s load control.

Figure 4.

Location of 50% MFB for the hydrogen and gasoline engine at selected engine speeds.

Hydrogen engine

The hydrogen engine load was controlled by adjusting only the fuel injection duration. In other words, a quality-control was applied. The SOI was fixed to 360 dCA BTDC over the whole engine map. Hydrogen rail pressure was kept constant at 8 bar by a mechanical pressure regulator. The throttle was always kept WOT, while the WG was kept close – except for a few specific points at high load and high engine speed. Where applicable, the ignition angle was adjusted in a way that 50% MFB was reached at 10 dCA ATDC (Figure 4). As the air path of the hydrogen engine is expected to be similar to that of a gasoline engine running at full load, the cam phasers for a given speed have been set to the full load setting used for that speed in gasoline operation. Finding the optimal VVT setting was subject of the subsequently performed full load optimization.

Full load limitations for the hydrogen engine have been:

Maximum amplitude of in-cylinder pressure oscillation (MAPO knock indicator) <1 bar

Peak Firing Pressure <110 bar

Relative air-fuel ratio $λ$ > 1.5

The guideline to keep MAPO <1 bar is based on previous experimental tests with hydrogen engines.

Experimental investigations

Baseline hydrogen operation – Comparison to series gasoline operation

Basic operating quantities

Figure 5 compares performance and operational results of both engines. The first row of plots compares the global λ. In hydrogen operation, the air mass resulted from the air-path setup. Consequently, enrichment was observed as the load increased. Comparing different engine speeds, it becomes apparent that the decrease in λ is steepest at low engine speed. This can be attributed to the low exhaust mass flow and the resulting weak performance of the charging system (this will be addressed in the following subsections). As a result, close-to-stoichiometric λ was achieved already at low loads. Since the test methodology prescribed λ to be greater than 1.5, the load could not be increased further. It must be noted that with $λ$ closer to stoichiometry, knocking becomes more frequent, NOx emissions rise and efficiency decreases. Not least for this reason, a large proportion of modern mixture formation concepts are based on lean operation.²⁰

Figure 5.

Comparison of performance and operation results between the converted hydrogen engine (left) and the series gasoline engine (right).

Due to the narrow λ requirements of the three-way catalyst (TWC), the gasoline engine was operated stoichiometrically throughout the whole engine map, except for the high load range where λ up to 0.9 were accepted.

The second row of plots illustrates the indicated thermal efficiency (ITE) of the engines. The hydrogen engine reached a peak ITE of 46%. Particularly impressive is that it has a wide range of ITE greater 44%. It can be postulated that the ITE of the hydrogen engine is enhanced by the WOT²¹ and the lower combustion temperature (Figure 5e), which results in reduced wall heat losses—the latter being a consequence of the lean-burn concept. The gasoline engine reached a peak value of 39% in the medium load range.

Combustion analysis

Figure 6 shows relevant combustion parameters. The first row of plots compares the burn duration of the hydrogen and gasoline combustion, defined as the crank angle interval between 10% and 90% fuel mass fraction burned (10%–90% MFB). Under stoichiometric conditions, hydrogen’s laminar flame velocity is approximately six times higher than that of gasoline. However, at λ = 4, hydrogen’s laminar flame velocity reaches a level four times lower than gasoline at λ = 1.²² Therefore, enleanment can cause hydrogen’s burn duration to exceed that of gasoline. This can be observed when comparing 10%–90% MFB in the low load range, where hydrogen’s burn duration is >32 dCA and therefore greater than that of gasoline. Figure 9a shows that 10%–90% MFB rises rapidly when $λ$ increases at a specific engine speed, emphasizing the dependency of burn duration on $λ$ . However, the figure also indicates that the high turbulence levels and the resulting large flame front present at higher engine speeds can compensate for the increased crank angle interval required to complete combustion. This is why the burn duration above 2500 min⁻¹ is roughly constant, independent of the engine speed. The burn duration of the gasoline engine shows no such dependence on the mixture composition since it is operated stoichiometrically. Below 10 bar of IMEP, it decreases almost linearly from 33 to about 20 dCA as the load increases. Above 10 bar, a dependence on engine speed can be observed. The minimum of 14 dCA is found at 1500 min⁻¹ and 15 bar IMEP. Overall, the burn durations of both engines were comparable (particularly, when $λ$ was kept between 1.6 and 2.5 in hydrogen operation), as they both exhibit a wide range of burn durations between 16 and 32 dCA.

Figure 6.

Comparison of combustion parameters between the converted hydrogen engine (left) and the series gasoline engine (right).

Knocking was quantified using the MAPO knock indicator. From the second row of plots in Figure 6, it can be seen that the MAPO values in hydrogen operation were comparably low throughout the whole engine map. At 13 bar IMEP and 2500 min⁻¹, only 0.4 bar amplitude was detected. At the same operating point, 1.4 bar was measured in gasoline operation. For this reason, the ignition had to be retarded at lower load levels already in gasoline operation, while the hydrogen engine could be operated with maximum brake torque spark timing in wider areas (Figure 4). In gasoline operation, MAPO rises only gradually with increasing load. This is because gasoline is a complex mixture of light and heavy hydrocarbons with varying knock resistance, quantified by the research octane number (RON). In comparison, hydrogen is a single molecule with a defined RON, making the transition between no knock and severe knock very sudden. Therefore, load was not increased further when MAPO approached 1 bar. Knocking was the full load limiting factor at high engine speeds in hydrogen operation.

The third row of plots compares the maximum rate of pressure rise (RPR). In hydrogen operation, the max. RPR was 4.4 bar/dCA. It was highest in the low speed, high load range because of the close-to-stoichiometric λ and therefore high burn velocity. In gasoline operation, the max. RPR rose almost linearly with the load and reached its maximum of 4.1 bar/dCA at medium load. Toward higher loads (>18 bar IMEP) the RPR decreased due to the retarded combustion phasing which was necessary to suppress knocking.

In the fourth row of plots, the coefficient of variation (CoV) of IMEP was used to quantify cyclic variations. It can be seen that the hydrogen engine showed a wide range of CoV of IMEP below 0.9%. The wide flammability limits and higher flame velocity of hydrogen contribute to stable combustion, even during lean operation. However, the richer the mixture, the lower the CoV of IMEP. Accordingly, the combustion becomes unstable in the low load range, where the mixture was leaner (λ > 3). Although other researchers report benefits in terms of combustion stability when throttling in this operating range,²³ this engine was operated with WOT permanently. Also, it has been found that there is negative scavenging during low load operation (Figure 7). It can be assumed that the combustion stability is further impaired by the increased fraction of residual gases in this range. The combustion stability in gasoline operation benefits greatly from the stoichiometric mixture. Even at low load, the CoV of IMEP was low. The best CoV of IMEP values (<0.9%) were observed in the medium load range. Toward higher load conditions, cyclic variability increased, which can be attributed to the spark retardation to limit knocking.

Figure 7.

Scavenging behavior of the hydrogen engine (values > 0 correspond to gas-flow from the intake to the exhaust port).

Figure 7 shows the scavenging gradient of the hydrogen engine, calculated as

Scav . grad . = \frac{1}{Δ α} * \int_{IVO}^{EVC} (p_{int} - p_{exh}) * α d α

The calculation is based on crank angle-synchronous intake (p_int) and exhaust port pressure (p_exh) measurements during the respective valve overlap (from intake valve opening (IVO) to exhaust valve closing (EVC)). For scavenging gradient <0, the gas-exchange process is dominated by gas flow from the exhaust to the intake port, which is referred to as negative scavenging. This occurs mainly during low load operation. The scavenging gradient becomes increasingly positive as the load increases. The largest part of the map is characterized by positive scavenging due to high intake manifold pressures, caused by the turbocharger behavior in combination with the permanently open throttle. Therefore, negative scavenging can only be used to a very limited extent to realize internal exhaust gas recirculation. For the gasoline engine, no data for scavenging analyses is available.

Charging system

Figure 8a shows the full load operating points of the hydrogen and the gasoline engine in the turbine map. Because the WG was kept fully closed, the hydrogen full load line coincides with the turbine curves until the engine speed reaches 3000 min⁻¹. Above 3000 min⁻¹, the WG was slightly opened, meaning that the pressure upstream turbine was reduced, and a part of the exhaust mass flow did not pass the turbine. Accordingly, a more vertical course can be observed because the exhaust mass flow increased, but the expansion ratio was limited by WG opening. Gasoline operation was characterized by a partially open WG even at full load. Consequently, the operating line shows a steep rise in the turbine map.

Figure 8.

(a) Turbine map and (b) compressor map with full load operating points of the hydrogen and gasoline engine.

Comparing both engines, several things can be noticed:

(1) In hydrogen operation, the turbine expansion ratios below 2000 min⁻¹ were low despite the closed WG, resulting in low power being delivered to the compressor. This can be attributed to the low enthalpy flow caused by the low temperature and mass flow of the exhaust gases, resulting from lean-burn operation. Accordingly, in low load operation, where lean mixture was predominantly found, exhaust gas temperature was lower, as shown in Figure 5e. Even at full load, significantly lower temperatures were observed than in gasoline operation. This was reflected in the compressor map (Figure 8b). The compressor pressure ratio in this speed range was significantly lower in hydrogen than in gasoline operation. This led to lower boost pressure, lower cylinder filling and thus a deficit in engine torque.

(2) Between 2500 and 3000 min⁻¹, the expansion ratio was higher in hydrogen than in gasoline operation. Accordingly, the compressor pressure ratio of the hydrogen engine exceeded that of the gasoline engine in this speed range. As can be seen from the engines’ full load lines, the IMEP deficit of the hydrogen engine is lowest in this range (see i.e. Figure 5).

(3) Above 3000 min⁻¹, the expansion ratios of both engines were similar. The compressor pressure ratios were also similar.

In the compressor map, it can be observed that the operating points in hydrogen operation were shifted toward the surge line and away from the line of highest compressor efficiency. The shift can be explained by the air displacement effect that the injected hydrogen has in the intake port. Above 2000 min⁻¹, the hydrogen engine operated at similar or higher pressure ratios than the gasoline engine, even though its full load IMEP is somewhat smaller. The main reason for that was the lean-burn operation, demanding higher boost to keep higher λ levels.

In lean-burn hydrogen operation, the series turbocharger configuration is very unfavorable for passenger cars where the engine must cover a wide speed range. Especially in the low speed range, close-to-stoichiometric mixtures are quickly achieved when the load is increased (Figure 5a), which is not in line with the lean-burn concept. This has been experienced by other researchers as well.^14,24–26

Emission characteristics

Since hydrogen is carbon-free, greenhouse gas and pollutant emissions like hydrocarbons and carbon di-/monoxide are only found in negligible quantities, caused by burned engine oil. The only relevant emissions are nitrogen oxides (NO, NO₂, summarized as NOx), endothermically formed out of the atmospheric nitrogen and oxygen at high temperatures during the combustion. The fourth row of plots in Figure 5 shows the engine-out indicated specific NOx (ISNOx) emissions of the hydrogen and gasoline engine.

Although the peak ISNOx value of the hydrogen engine is higher than that of the gasoline engine, the gasoline engine emits significantly more NOx overall than the hydrogen engine over large parts of the engine map. This happens because the gasoline engine was operated under stoichiometric conditions. Therefore, higher combustion temperatures were present. The ISNOx emissions range from 2 to 10 g/kWh in gasoline operation. The hydrogen engine in comparison shows a wide operating range with quasi-zero engine-out ISNOx emission (<0.1 g/kWh). However, with load increase, which – in this specific engine setup – corresponds to enrichment, ISNOx emissions rise rapidly until they ultimately exceed the peak value in gasoline operation with 23.9 g/kWh. The emission characteristic of the hydrogen engine shows a strong dependency on λ. At low load, where λ was leaner, the combustion temperature was low, which impairs the endothermic NOx formation. This dependency is emphasized by Figure 9b, where ISNOx (in logarithmic scaling) is plotted against λ for all operating points of the hydrogen engine. As long as the engine is operated with $λ$ > 2.4, ISNOx falls below 0.1 g/kWh. When approaching stoichiometric combustion, ISNOx rises rapidly to values above 10 g/kWh. This presents the charging system with the challenge of supplying enough air to maintain lean mixture, even at high load conditions.

Figure 9.

(a) Burn duration (engine speed-resolved) and (b) ISNOx (in logarithmic scaling) of the hydrogen engine over λ.

If operation leaner than $λ$ = 2.4 cannot be ensured for whatever reason, a conventional TWC can be used to reduce harmful nitrogen oxides. It shows high conversion efficiencies in hydrogen engines, but only when operated within its narrow $λ$ requirements (preferably slightly rich to supply hydrogen as reducing agent to the TWC, as reported by^27,28). Therefore, the engine control must switch from lean, quality-controlled to stoichiometric, quantity-controlled operation when λ is about to fall below 2.4. The feasibility of this was proven by BMW in a 7 series vehicle.²⁸

The NOx output of the gasoline engine is less of a concern than that of the hydrogen engine, as the TWC can reliably reduce it at close-to-stoichiometric operation.

Full load optimization

The hydrogen engine described in the previous sections has several shortcomings. Firstly, the high NOx emissions made an after-treatment system indispensable. Secondly, the IMEP was lower than that of its gasoline counterpart.

Without any hardware modifications, the only available degrees of freedom for full load optimization were the VVT setting and the spark timing. Therefore, further experiments were performed to improve the full load curve of the hydrogen engine. Firstly, the VVT setting was optimized through a load cut exemplary at 2500 min⁻¹. Secondly, a spark sweep was performed at 2500 min⁻¹ with different VVT settings to verify its impact on the ISNOx emissions. Lastly, a full load optimization procedure, employing both features, was performed throughout the engine map in order to extend the full load torque. The operating limit was defined as the point at which one of the following threshold values was exceeded:

Peak firing pressure <110 bar

MAPO knock indicator <1 bar

Turbocharger speed < $220 \times 10^{3} \min^{- 1}$

For ISNOx, a target value of 1.0 g/kWh was defined based on WLTC driving cycle simulations from data presented in²⁹. It was further assumed that NOx emissions increase exponentially as λ decreases, while they are negligible in low load operation. This way, some margin is obtained to emit more NOx at full-load conditions and still remaining within Euro 5/6 limits.

Load cut with variation of VVT setting

Firstly, load cuts with different VVT settings were carried out, exemplary at 2500 min⁻¹. The purpose was to identify the optimal VVT setting in terms of NOx output and engine efficiency. The settings shown in Table 3 were applied. They are given in the form ‘(intake phasing|exhaust phasing)’. Both values are given in dCA from the VVT system’s lock position. Negative values indicate an advanced and positive values indicate a retarded valve event.

Table 3.

Overview of the applied VVT settings.

VVT setting	Resulting overlap (dCA)
(−19\|15)	29
(−30\|19)	44
(−45\|19)	59
(−30\|30)	55
(−45\|30)	70
(−30\|45)	70
(−45\|45)	85

50% MFB was kept at 10 dCA ATDC until knocking occurred or the peak firing pressure approached its maximum of 110 bar.

The engine’s VVT system features adjustment angles of 45° advancement for the intake and 45° retardation for the exhaust camshaft. When the intake camshaft is advanced and the exhaust camshaft retarded, the crank angle interval in which both valves are opened simultaneously (so-called valve overlap) increases. It must be noted that the $λ$ values shown below are calculated based on the measured fuel and air mass flow. Accordingly, they do not necessarily equal the in-cylinder $λ$ because a part of the air mass flow may pass through the valves without contributing to the combustion.

Figure 11 shows selected process variables from a load-cut performed at 2500 min⁻¹, plotted over engine load. Each curve represents a VVT setting as described in the legend. Figure 11a underlines the just described theory. The more overlap is applied, the leaner the mixture becomes. The enleanment is very desirable, as premature enrichment was identified as the major full load limit in the baseline investigation.

The temperature upstream turbine behaves inversely proportional to $λ$ because the exhaust gases are continuously cooled due to the air short-circuit through the overlapping valves (Figure 11b). Therefore, the curves with little overlap show the highest combustion and exhaust temperatures due to the comparably rich $λ$ .

VVT also has a significant effect on the charging system. The additional air mass flow resulting from the air short-circuit during high overlap operation shifts the turbine’s operating point toward higher reduced mass flow rates and expansion ratios. Consequently, more expansion work is performed and transferred to the compressor, which converts it into more boost pressure. In the compressor map, operating points with high overlap are shifted to the right toward the map center, where compressor efficiency and the surge margin are higher (Figure 11c). However, even for the greatest overlap, the surge margin falls below 20% at 21 bar of IMEP with a minimum of 0% at 22 bar. This is an indicator that the series compressor is not a good match for the engine in hydrogen operation.

As can be seen in Figure 11d, gas exchange efficiency, defined as

η_{gas - exchange} = \frac{{IMEP}_{net}}{{IMEP}_{gross}}

shows a strong correlation to the applied VVT setting. When comparing the (−45|30) and (−30|45) settings, it can be observed that although the crank angle interval of the valve overlap is equal (compare Table 3), (−45|30), which has the earlier exhaust valve opening (EVO), reaches higher gas exchange efficiencies. Therefore, it can be concluded that the EVO position significantly influences gas-exchange efficiency and ITE (Figure 11e). On the one hand, early EVO values deliver a high-pressure pulse in the beginning of the exhaust stroke, with the drawback of reduced expansion work produced. On the other hand, the later the EVO, the higher the expansion work, with the drawback of reduced pressure to drive the exhaust gases out of the cylinder. This way, piston work is consumed to pump the gases out of the cylinder. This effect can be clearly seen in the p-V-diagram (Figure 10, in logarithmic scaling), where the VVT settings with more exhaust retardation (i.e. the green and the cyan lines) keep a higher pressure during the start of the exhaust stroke, increasing the area of negative work in the pumping loop. Furthermore, with increased valve overlap, a higher amount of air was blown through the cylinder into the exhaust manifold. This short-circuited air aids the turbine expansion work, increasing the amount of boost available from the compressor.

Figure 10.

p-V-diagram at 2500 min⁻¹, 20 bar of IMEP and different VVT settings (in logarithmic scaling).

ITE is illustrated in Figure 11e. Operating points with high overlap were characterized by comparably low ITE, which is a consequence of two aspects. Firstly, high overlap leads to low gas-exchange efficiency, as described in the previous paragraph. Secondly, boost pressure is partly lost through the overlapping valves. It can also be seen that VVT settings which were beneficial in terms of gas-exchange efficiency are not necessarily beneficial in terms of ITE. With rising load, the ITE values of all tested VVT settings decreased. This can be explained by the reduction in $λ$ that came with the load increase and led to higher combustion temperatures, resulting in increased wall heat losses. This effect was further enhanced by the design of the exhaust manifold as an integrated water-cooled component in the cylinder head to reduce the temperature of the exhaust gases entering the turbine.

Figure 11.

Load-cut performed at 2500 min⁻¹ with several VVT settings.

Remarkable was the low level of CoV of IMEP (Figure 11f). Most of the operating points lie below 1.0%, no matter the VVT setting. This shows the distinct advantage of high turbulent tumble flow for premixed combustion processes. This is consistent with the observations made by Salazar and Kaiser in an optical engine running on hydrogen. They report a decrease in CoV of IMEP and in burn duration when introducing tumble flow.³⁰

Figure 11g shows the ISNOx emissions of the engine. It can be observed that the ISNOx emissions start with approximately 0 g/kWh at 10 bar of IMEP. When the load was increased, ISNOx rose rapidly, depending on the VVT setting. The higher the valve overlap, the flatter the ISNOx rise. That is, at 20 bar of IMEP, the ISNOx value of the maximum valve overlap was only 30% of that of the (−30|19) setting. A positive effect of high overlap can therefore be identified on emissions. Nonetheless, NOx emissions were still high in general, no matter the VVT setting. Even with the maximum overlap, the goal of 1 g/kWh could only be maintained until 14 bar of IMEP. Consequently, VVT could not be the only measure to control the emissions and a different approach had to be taken into consideration.

Variation of spark timing

It is known from literature that spark timing has a significant effect on NOx emission.^31,32 The later the spark timing, the lower the NOx emission, as the peak pressure and the combustion temperature drop. Accordingly, spark retardation is seen as a promising approach to keep emissions within low limits. Nonetheless, with the drawback of lower thermal efficiency due to the non-optimized combustion phasing. A test series was conducted at 2500 min⁻¹ with constant fuel injection duration and a variation of spark timing, resulting in 50% MFB ranging from 10 to 30 dCA ATDC. The test series was carried out twice with different VVT settings—once with (−30|30), once with (−45|45). The goal was to study the effect of spark timing on NOx emission and possible undesired side effects.

Figure 12 shows the result of this approach. The main effect of spark retardation was the reduction of the peak firing pressure (Figure 12a), since a larger part of the combustion takes place in the expansion stroke. On the one hand, this reduced the performed piston work, driving a reduction in ITE (Figure 12b), since the injected fuel mass was kept constant. On the other hand, the later combustion delivers a higher exhaust enthalpy, confirmed by the higher exhaust temperatures (Figure 12c). Therefore, the expansion work performed in the turbine was increased, improving the boost levels and slightly increasing the λ values (Figure 12d).

Figure 12.

Spark sweep performed at 2500 min⁻¹ with constant fuel mass at two VVT settings (IMEP ranging from 18 to 15 bar).

The λ curves for the different VVT settings are almost parallel. It can be seen that the λ values for the high overlap setting are consistently greater than those for the low overlap setting. This is a consequence of the higher mass flow of air passing through the engine when high overlap was applied and the better turbocharger behavior at higher air mass flows.

Analogous to Figure 11e, it can be observed that the maximum overlap setting is unfavorable in terms of ITE (Figure 12b)—a consequence of its comparably low gas exchange efficiency and the fact that boost pressure might be partly lost through the overlapping valves.

ISNOx values fall exponentially when the center of combustion is retarded (Figure 12e). The reduced peak firing pressure results in lower combustion temperature and an obstruction of the endothermic NOx formation process. The NOx formation is further reduced by the enleanment, that comes with later center of combustion and also lowers the combustion temperature. It can therefore be observed that the high overlap operation shows lower ISNOx values due to its leaner combustion.

Retarding ignition brings the desired reduction in ISNOx emission but simultaneously, the engine loses efficiency. Fortunately, the search for a trade-off is mitigated by the fact that ISNOx has a strong exponential decay in Figure 12e. Therefore, to achieve the 1 g/kWh limitation, the center of combustion must be retarded to 21.7 dCA ATDC for the high and to 25.9 dCA ATDC for the low overlap operation. Compared to the efficiency-optimized operation at 10 dCA ATDC, this is accompanied by a loss in efficiency of 2.8 and 3.6 percentage points, respectively. It can be concluded that combustion phasing by spark timing adjustment is a sensible way to influence the engine’s ISNOx output.

Figure 12f shows that the combustion stability decreases when the center of combustion is retarded. For the (−45|45) setting, retarding 50% MFB from 10 to 30 dCA ATDC, leads to an increase in CoV of IMEP from 0.75% to 2.2%.

Full load optimization

The previous subsections discussed the influence of VVT and spark timing, specifically at 2500 min⁻¹. For the full load optimization, a combination of both strategies was applied in order to find the best trade-off in terms of ITE and ISNOx. Therefore, engine speeds between 1000 and 4000 min⁻¹ were tested in steps of 500 min⁻¹. The following methodology was established.

At first, a VVT setting was fixed. The tests started at a load point where 50% MFB was at optimal condition (10 dCA ATDC) and ISNOx was still below the 1 g/kWh limit. Next, the injected fuel mass was increased until ISNOx was above the limit and then the spark timing was adjusted to bring the ISNOx value down below 1 g/kWh again. This procedure was repeated until one of the operation limits listed below Table 3 was reached or 50% MFB was retarded to up to 30 dCA ATDC. At this point, a new VVT setting was set and the procedure was repeated. The settings tested first were the ones from the gasoline ECU calibration. After that, only VVT settings with higher overlap were tested.

Figure 13 shows the results of this optimization procedure at 2500 min⁻¹. An IMEP value of 10 bar was selected as the starting point. In the end, IMEP could be increased to up to 19 bar with the described full load optimization procedure.

Figure 13.

NOx-optimized hydrogen engine at 2500 min⁻¹.

From Figure 13a it is apparent that 50% MFB at 10 dCA ATDC could be maintained the longest with high overlaps. In the minimum overlap operation, ignition had to be retarded from 11 bar on to keep ISNOx below the limitation of 1 g/kWh. The immediate NOx response can be explained by the comparably rich $λ$ (Figure 13b) due to the reduced air short circuit and the poor turbocharger behavior. For the medium and the high overlap setting, ignition had to be retarded for loads above 13 and 14.5 bar IMEP, respectively.

Due to the procedure, all operating points show ISNOx emissions below 1 g/kWh, independent of the engine load (Figure 13c). Consequently, all VVT settings are now equal in terms of NOx emission. To identify the optimal VVT setting for each load point, ITE was taken into account.

No matter the VVT setting, the increase in load was followed by a drop in ITE (Figure 13d). This can be explained by the late center of combustion, which was applied to keep ISNOx below the specified limit. It is noticeable that the three curves fall with different gradients. While (−19|15) starts with the highest efficiency, it loses roughly 1.5% points for each bar of increased IMEP. This resulted from the unfavorable combustion phasing which was necessary to contain NOx emissions during operation with lower λ. Compared to the other curves, it can therefore only demonstrate its efficiency advantage in the 10–11 bar IMEP range. (−30|30) had a slightly lower efficiency at 10 bar of IMEP but the curve fell with a significantly lower gradient. Therefore, it can be stated that (−30|30) was the most favorable VVT setting in the IMEP range of 11–14 bar as its efficiency curve lied above the others. Above 14 bar, the best efficiency values were achieved when applying maximum overlap.

At first glance it may seem contradictory that the maximum overlap setting showed the lowest ITE below 14.5 bar of IMEP, although it can maintain the efficiency-optimized combustion phase (50% MFB at 10 dCA ATDC) in this range. But this can be explained by the comparably low gas exchange efficiency at full overlap (as observed from Figure 11d). In the present case, the negative impact of low gas exchange efficiency on the ITE outweighs the positive effect of optimal combustion phasing.

The ITE curve emphasizes the benefit of an engine equipped with VVT. Without the VVT, a compromise in efficiency would have to be made when defining the valve timing. Another important observation is that significantly higher loads could be reached when applying higher overlap settings, underlining the advantages of its use. Ultimately, 50% MFB approaching 30 dCA ATDC was the full load limiting criterion. Possibly, a further retarded combustion center would make higher IMEPs possible while still maintaining 1 g/kWh ISNOx but would also lead to a non-sensible trade-off between efficiency and power output.

From the spark sweep, it is known that the CoV of IMEP reacts sensitively to spark timing (Figure 12f). All in all, it still shows acceptable values below 1.6% (Figure 13e). This may be due to the characteristic of the gasoline intake port to generate turbulence in form of tumble flow.

From this investigation, the most favorable VVT settings have now emerged. Each operating point, defined by engine speed and IMEP, can now be assigned an optimum VVT setting. Only those efficiency-optimized VVT settings are now transferred to the new full load ECU calibration.

Optimized hydrogen engine

The previous optimization was applied for the engine speed range between 1000 and 4000 min⁻¹ in steps of 500 min⁻¹. In the following line plots, the characteristics of the optimized hydrogen engine are shown and superimposed by the characteristic curves of the baseline hydrogen engine and the gasoline engine.

Figure 14 compares the behavior of the gasoline and the baseline hydrogen engine with the optimized hydrogen engine. For the first two engines mentioned, the VVT set values were explained in section Boundaries. For the optimized hydrogen engine, the efficiency-optimized VVT settings, as described in the previous chapter, were applied.

Figure 14.

Optimized hydrogen engine in comparison to baseline hydrogen and gasoline engine at 2500 min⁻¹.

The optimization procedure brought an increase in IMEP in comparison to the baseline hydrogen engine. This can be attributed to the leaner $λ$ (Figure 14a). This was made possible by the high overlap at full load, a result of the high IVO advance (Figure 14b) and high EVC retardation (Figure 14c).

As Figure 14d shows, the optimized hydrogen engine employed later combustion phasing compared to the baseline hydrogen engine. While the baseline hydrogen engine maintained the efficiency-optimized combustion center at 10 dCA ATDC independent of the load, the optimized engine shows heavy retardations from 13 bar of IMEP on. In combination with the comparably low gas exchange efficiency (Figure 14e), this resulted in a significant drop in ITE in this IMEP range (Figure 14f). Nonetheless, the optimized hydrogen engine still shows a considerably higher overall efficiency compared to its gasoline counterpart, promoted by the unthrottled lean-burn operation.

Besides the expansion in IMEP, Figure 14g shows the second major achievement of the optimization procedure. While the baseline hydrogen engine, which was favorable in terms of ITE, showed exponentially rising ISNOx emissions (of up to 17 g/kWh), the optimized hydrogen engine’s ISNOx emissions were constantly below 1 g/kWh, even at full load operation. In this way, it also beats the ISNOx emission behavior of the gasoline engine. However, it must be noted that NOx emissions are less of a concern for a stoichiometrically operated gasoline engine because of the high conversion efficiency of the TWC. The only remaining drawback of the optimized hydrogen engine is that the maximum achievable load is still slightly behind that of the gasoline engine.

Overall, the engine load could significantly be increased while still keeping NOx emissions below the proposed limit of 1 g/kWh. The gasoline engine’s IMEP at 2500 min⁻¹ (21.3 bar) was missed by only 2.2 bar. In comparison to both, the gasoline and the baseline hydrogen engine, the NOx output was heavily reduced.

Figure 15 shows maps of the optimized hydrogen engine in comparison to the baseline hydrogen engine. The following can be observed:

The maximum IMEP was improved from 18.2 to 20.2 bar, achieved at 3000 min⁻¹.

In the range of 1500–2500 min⁻¹, the NOx limitation strategy demanded heavy retardations in spark timing at high load, resulting in comparably poor efficiency for the optimized engine.

At 1500 min⁻¹, the full load of the baseline hydrogen engine could not be surpassed. To maintain the ISNOx limit, the ignition timing had to be retarded to such an extent that 50% MFB rushed against 30 dCA ATDC. This was a consequence of the poor turbocharger behavior at this engine speed, as described in section Charging system. Without adapting a different charging system that responds already to lower exhaust mass flows, the low-speed IMEP cannot be approximated to neither that of the baseline hydrogen engine, nor the series gasoline engine.

The $λ$ -map shows a strong enleanment in comparison to the baseline hydrogen engine which resulted from the higher valve overlap applied.

The ISNOx map shows the successful implementation of the NOx strategy. While the baseline hydrogen engine had tremendous ISNOx emission at full load because of the close-to-stoichiometric mixture, the optimized hydrogen engine maintained ISNOx ≤ 1 g/kWh throughout the whole map.

Figure 15.

Engine map comparison of the baseline (left) and the optimized hydrogen engine (right).

Summary and conclusion

In the present work, a turbocharged 1.0 L gasoline engine was converted for hydrogen combustion. The main modification was the fuel injection system, consisting of H₂ port fuel injectors. A baseline H₂ operation was performed, followed by an optimization procedure, which focused on the manipulation of the VVT system. The main objectives of the optimization were to increase the load and to ensure that ISNOx remains below a level of 1 g/kWh in compliance with Euro 5/6 regulations. The following conclusions can be drawn:

(1) The combustion parameters of the baseline hydrogen engine in terms of burn duration, rate of pressure rise and cyclic variations are similar to the stoichiometric gasoline combustion—especially as long as $λ$ is kept between 1.6 and 2.5 in hydrogen operation.

(2) In hydrogen operation, the ISNOx emission behavior is strongly dependent on $λ$ . Below $λ$ = 2.2, the ISNOx production is close to 0 g/kWh. However, it increases rapidly as $λ$ approaches stoichiometry, making an exhaust gas aftertreatment system indispensable.

(3) The series turbocharger used in this work was not suitable for lean hydrogen operation, as it could not deliver the high boost requirement due to the lower exhaust gas enthalpy—most notably at low engine speeds. When the load was increased, a close-to-stoichiometric mixture was quickly achieved. Ultimately, this was the full load limiting parameter at low engine speed.

(4) During the optimization procedure, it was found that applying high overlap resulted in lower NOx emission and higher achievable loads, but also in a loss of gas-exchange efficiency. Furthermore, spark retardation was identified as a decisive parameter for reducing NOx emissions—even if this is accompanied by a slight loss in ITE. Even with spark retardation, stable combustion in terms of CoV of IMEP could be maintained.

(5) Ultimately, the engine can deliver up to 20.2 bar of IMEP while complying with Euro5/6 regulations across the entire engine map, eliminating the need for an exhaust gas aftertreatment system.

Footnotes

Appendix

ORCID iD

Noah Kastler

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

UNFCCC. Paris agreement. Presented at the Paris climate change conference – November 2015 (December 2015, accessed 02 August 2024). https://unfccc.int/sites/default/files/english_paris_agreement.pdf

European Environment Agency. Greenhouse gas emissions from transport in Europe (2023, accessed 02 August 2024). https://www.eea.europa.eu/en/analysis/indicators/greenhouse-gas-emissions-from-transport

Smith

. Can EV (electric vehicles) address Ireland’s CO2 emissions from transport? Energy 2010; 35(12): 4514–4521.

Kubach

Bucherer

Gerber

, et al. Trends und Herausforderungen bei der Entwicklung von Wasserstoffmotoren. Chem Ing Tech 2024; 96(1-2): 167–181.

Kerstin Mayr Hofer

Ragowsky

, et al. Systemvergleich zwischen Wasserstoffverbrennungsmotor und Brennstoffzelle im schweren Nutzfahrzeug. 2021.

Munshi

Garner

Theissl

, et al. Total Cost of Ownership (TCO) analysis for heavy duty hydrogen fueled powertrains. (2021, accessed 02 August, 2024). (Online). https://wfsinc.com/js/web/viewer.html?file=/file_library/files/wfs-wfsinc/20201225_Westport_AVL_Whitepaper_Hydrogen_HPDI_final.pdf

Verhelst

. Recent progress in the use of hydrogen as a fuel for internal combustion engines. Int J Hydrogen Energy 2014; 39(2): 1071–1085.

Das

. Hydrogen-fueled internal combustion engines. In: Barbir

Basile

Nejat Veziroğlu (eds) Compendium of hydrogen energy. Elsevier, 2016, pp. 177–217.

Schüers

Abel

Fickel

, et al. Der Zwölfzylinder-Wasserstoffmotor im BMW 750hL. MTZ - Motortechnische Z 2002; 63(2): 98–105.

10.

Natkin

Tang

Boyer

, et al. Hydrogen IC engine boosting performance and NOx study. In: SAE 2003 world congress & exhibition, Detroit, MI, 2003, pp. 2003-01–0631. doi: 10.4271/2003-01-0631.

11.

Verhelst

Maesschalck

Rombaut

, et al. Increasing the power output of hydrogen internal combustion engines by means of supercharging and exhaust gas recirculation. Int J Hydrogen Energy 2009; 34(10): 4406–4412.

12.

Sierens

Verhelst

Verstraeten

. EGR and lean combustion strategies for a single cylinder hydrogen fuelled IC engine. In: 10th EAEC European automotive congress, Belgrad, May 2005, p. 10.

13.

White

Steeper

Lutz

. The hydrogen-fueled internal combustion engine: a technical review. Int J Hydrogen Energy 2006; 31(10): 1292–1305.

14.

Schutting

Roiser

Eichlseder

, et al. Hydrogen engine exhaust aftertreatment. In: 43rd International Vienna motor symposium, Vienna, April 2022.

15.

Pauer

Weller

Schünemann

, et al. H2 ICE für zukünftige PKWs und leichte Nutzfahrzeuge. In: 41st International Vienna motor symposium, Vienna, April 2020.

16.

Wittek

Geiger

Andert

, et al. Experimental investigation of a variable compression ratio system applied to a gasoline passenger car engine. Energy Convers Manag 2019; 183: 753–763.

17.

Friedfeldt

Zenner

Ernst

, et al. Three-cylinder gasoline engine with direct injection. MTZ Worldw 2012; 73(5): 4–11.

18.

Pischinger

(ed.). Vorlesungsumdruck Verbrennungskraftmaschinen. 29th ed., Vol. 1, Aachen, RWTH Aachen University, Aachen, Germany, 2013. Vol. 1.

19.

Kondo

Iio

Hiruma

. A study on the mechanism of backfire in external mixture formation hydrogen engines – about backfire occurred by cause of the spark plug. In: International spring fuels & lubricants meeting & exposition, Dearborn, MI, May 1997, p. 971704. doi: 10.4271/971704.

20.

Wallner

. Efficiency and emissions potential of hydrogen internal combustion engine vehicles. In: Symposium on International Automotive Technology 2011, Pune, India, January 2011, pp. 2011-26–0003. doi: 10.4271/2011-26-0003.

21.

Verhelst

Wallner

. Hydrogen-fueled internal combustion engines. Prog Energy Combust Sci 2009; 35(6): 490–527.

22.

Alekseev

Christensen

Konnov

. the effect of temperature on the adiabatic burning velocities of diluted hydrogen flames: a kinetic study using an updated mechanism. Combust Flame 2015; 162(5): 1884–1898.

23.

Swain

Pappas

Adt

, et al. Hydrogen-fueled automotive engine experimental testing to provide an initial design-data base. In: Swain

Pappas

Adt

Jr. , et al. (eds) SAE International congress and exposition, Detroit, MI, February 1981, p. 810350. doi: 10.4271/810350.

24.

Kapus

, et al. High efficiency hydrogen internal combustion engine – carbon free powertrain for passenger car hybrids and commercial vehicles. In: 43rd International Vienna motor symposium, Vienna, April 2022.

25.

Virnich

Durand

Huth

, et al. Exhaust gas aftertreatment system layout for hydrogen internal combustion engines. In: Heintzel

(ed). Proceedings Internationaler Motorenkongress 2022, Springer Fachmedien Wiesbaden, 2024, pp. 241–249. doi: 10.1007/978-3-658-44740-3_17.

26.

Tafel

Martin

. Development of a high-performance hydrogen engine. In: Kulzer

Reuss

Wagner

(eds). Proceedings 2024 Stuttgart International symposium on automotive and engine technology. Springer Fachmedien Wiesbaden, 2024, pp. 130–145. doi: 10.1007/978-3-658-45010-6_8.

27.

Verhelst

Sierens

. Combustion studies for PFI Hydrogen IC engines. In: Asia Pacific automotive engineering conference, August 2007, pp. 2007-01–3610. doi: 10.4271/2007-01-3610.

28.

Enke

Gruber

Hecht

, et al. The bi-fuel V12 engine of the new BMW hydrogen 7. MTZ Worldw 2007; 68(6): 6–9.

29.

López

García

Monsalve-Serrano

, et al. Potential of a two-stage variable compression ratio downsized spark ignition engine for passenger cars under different driving conditions. Energy Convers Manag 2020; 203: 112251.

30.

Salazar

Kaiser

. Influence of the flow field on flame propagation in a hydrogen-fueled internal combustion engine. SAE Int J Engines 2011; 04(2): 2376–2394.

31.

Heywood

(ed.). Internal combustion engine fundamentals. 2nd ed. McGraw-Hill Education, 2018.

32.

Pischinger

. Vorlesungsumdruck Verbrennungskraftmaschinen, 29th ed., Vol. 2, Aachen, RWTH Aachen University, Aachen, Germany, 2013.