Experimental Results on Natural Gas and Liquefied Petroleum Gas Lean Burning in a Diesel Engine Retrofitted for Spark Ignition

Abstract

As part of efforts to support the transition toward a zero-carbon future, this research evaluates how the use of natural gas and liquefied petroleum gas under lean burn conditions affects the energy efficiency and environmental outcomes of a diesel engine that has been retrofitted to operate with spark ignition. The assessment of the ecological potential of these low-carbon gaseous fuels was performed at the engine test bed at optimum spark advance set from the condition of achieving maximum brake thermal efficiency (i.e., lowest carbon dioxide emission, CO₂). The results found with lean mixtures are compared to those obtained under stoichiometric conditions, as well as to those from a commercial gasoline engine of comparable size, equally operated at stoichiometry. With lean burning, a clear improvement is observed for all operating points in terms of brake thermal efficiency with respect to the stoichiometric operation. The results highlight a slightly greater improvement when operating with natural gas lean mixtures: between (1.35 and 2.35) percentage points gained in this case, compared to (1.15–2.10) percentage points gained in the case of liquefied petroleum gas. As for CO₂, a maximum 28% reduction when using natural gas is achieved with lean operation with respect to the commercial gasoline engine. Using lean mixtures also brings an important reduction in the engine-out pollutants (carbon monoxide, nitric oxides and particulate number). However, with respect to stoichiometric operation, cyclic variability of the prototype degrades with lean burning but remains lower than one of the baseline commercial gasoline engines.

1. Introduction and Literature Overview

Electric energy production and road mobility are paramount for human civilization. Considering the impact of these anthropogenic activities upon the environment, it is essential to pursue sustainable development to mitigate negative effects and ensure long-term ecological balance.

Natural gas (NG) and liquefied petroleum gas (LPG) as alternative fuels for transportation were always considered attractive because of (1) the availability of abundant supplies worldwide [1,2] and (2) the environmental benefits compared to conventional fossil fuels [2,3,4,5,6,7,8,9,10,11,12,13,14].

NG is a gaseous mixture of components predominantly composed of methane (CH₄), which is the lightest carbon-based fuel. LPG is a mixture of mainly propane (C₃H₈) and butane (C₄H₁₀) and is classified as a gaseous fuel because, under normal atmospheric conditions, it is in gaseous state. However, it is called “liquefied” because it is stored under moderate pressure (e.g., 15 bar), which causes its condensation. When released from its pressurized tank at room temperature, LPG vaporizes into a gas, making it a practical and versatile energy source. In an LPG fueled spark ignition (SI) engine vehicle, LPG vaporization is facilitated via a heat exchanger using the engine’s cooling liquid. According to [9], LPG is obtained either as a by-product from crude oil refining or extracted from oil or NG production streams. Thus, when LPG is produced from oil, a decrease in gasoline and diesel production leads to a corresponding drop in LPG output, as noted in [12]. That is why it is best to rather rely on NG sources. Report [9] mentions that “the share of LPG coming from gas processing has been growing as output of NG has grown faster than that of oil”. Therefore, one may conclude that NG and LPG are bound one to another.

As discussed in [9], these gaseous fuels are seen as “bridging fuels in the long-term transition to a truly sustainable global energy system”. Indeed, NG and LPG, as fuels with a higher hydrogen-to-carbon (H/C) ratio than gasoline, inherently generate lower amounts of carbon dioxide (CO₂), which is considered the main cause of the global warming phenomenon and hence the main threat to the continued use of carbon-based fuels in internal combustion (IC) engines. To further improve the life-cycle emissions, other NG- and LPG-related possible solutions for meeting the coming climate targets are renewable biogas or bio-methane [15,16,17], bio-LPG [18,19], synthetic methane and Fischer–Tropsch LPG via power-to-gas (PTG) technology [20,21,22]. Moreover, given that the octane numbers (ON) of NG and LPG (fossil or renewable) are higher than gasoline’s, the knock-free operation of the SI engine at a higher compression ratio (CR) is possible, thus further improving the thermodynamic efficiency. Furthermore, considering the extended lean flammability limits of these two gaseous fuels with respect to gasoline [23,24], a further increase in the thermodynamic efficiency with subsequent effects on the CO₂ reduction is achievable thanks to the corresponding increase in the specific heat ratio, γ. Additionally, when comparing to gasoline, the gaseous nature of NG and LPG promotes more homogeneous air–fuel mixtures, enhancing combustion stability and reducing cycle-to-cycle variations. This would allow lean operation with an engine stability still better than the stoichiometric standard gasoline engine, as also pointed out in [25,26].

Despite their advantages, challenges remain and are related (1) to developing an exhaust aftertreatment system (EATS) optimized for lean mixtures in order to ensure an efficient conversion over a wide range of operating conditions despite the relatively low raw engine-out emissions level [11]; (2) to flame propagation speed, which is known to drop quickly as the mixture is leaned from stoichiometric conditions [27]; and (3) to infrastructure requirements. However, these challenges continue to motivate ongoing research and technological development. Especially in the case of NG, the gas slip and unburnt gas at the tailpipe is of great interest, as methane is a greenhouse gas (GHG) with a global warming potential (GWP) 28 times higher than that of CO₂ [11,28,29].

NG or LPG may be used either as a bi-fuel solution, with gasoline and NG or LPG [3,5,12,30,31,32,33,34], or as a single-fuel solution sometimes called “monovalent” [12,29,35]. The former is quite popular in Europe and is used in mass production by many of the European automotive manufacturers (gasoline-LPG more often than gasoline-NG). As shown in [12,29,35], the latter has the advantage that the engine was developed exclusively for NG or LPG, i.e., it is about an engine free from the regular tradeoffs made to allow operating in bi-fuel mode (either with gasoline or with NG/LPG). Such a monovalent engine is designed to achieve the full ecological potential of gaseous fuels like NG or LPG, as also explained in [36,37,38]. Thus, when developing such engines, every decision is made to fully utilize the potential of NG or LPG in terms of energy and environmental performance. For instance, these engines can (1) use a high CR to suitably exploit the high ON of these gaseous fuels, which has the secondary advantage of diminishing the exhaust gas temperature (EGT) via the associated higher expansion, thus permitting the use of a variable geometry or nozzle turbine (VNT) with the benefits that arise from it; (2) apply a higher spark advance (SA), which also results in EGT reduction. On this matter, Warth et al. [35] presented relevant results about a “monovalent NG combustion and downsizing for lowest CO₂ emission”. They showed an up to 31% reduction in CO₂ with respect to the baseline production gasoline engine and 110 kW/L (beating the gasoline baseline by 10 kW/L) from a 1.2 L, three-cylinder, 13.3 CR SI PFI engine used in a compact van class. Regarding their reduction in CO₂ emission, Warth et al. [35] offered a breakdown showing that switching from bivalent to monovalent features is responsible for a further 7% reduction. Concerning the pollutant emissions, paper [35] did not provide any information. In the context of developing an optimized compressed NG downsized demonstrator engine, paper [29] complements paper [35]. However, for the pollutants, paper [29] provides some details only for the particulate mass (PM), which is said to be reduced “from 89% of the EU6 limit for gasoline, to 8% of the same limit when using NG”.

Sometimes, a retrofit (i.e., a transformation or a conversion of an existing engine) has meaning. More specifically, for example, to offer a second life to the diesel engine in a cost-driven market not favoring this technology anymore, as Europe is [39], it might be interesting to convert an existing CI engine to SI operation and use NG or LPG because of their benefits, as already stated above. In this case, the diesel-like architecture is thought to be capable of compensating the lower laminar flame velocities in lean burning via high compression and specific in-cylinder fluid dynamics, as also discussed in [27]. The preceding reasoning is also acceptable for existing diesel-based (1) stationary powerplants used for electricity production and (2) non-road agricultural machines, i.e., offering them a cleaner second life.

Unlike this paper, which investigates a small engine for passenger cars (PCs) and light-duty vehicles (LDVs), previous studies have focused on retrofitting CI engines for NG SI operation in heavy-duty vehicle (HDV) engines [40,41,42,43,44,45,46,47]. It is significant to note that some of these studies were conducted using a single-cylinder engine, which may influence the applicability of their findings to multi-cylinder engines [41,42,45,47,48]. Regarding the CI engine retrofit for NG SI operation, very insightful research on combustion was provided by Liu and Dumitrescu [27,47,49,50,51]. Similarly, Riccardi et al. [45] focused exclusively on combustion modeling in an NG SI heavy-duty (HD) engine, providing the results of using a turbulence model, which takes into consideration the swirl and squish specific to a CI-based architecture, in addition to tumble motion. Nevertheless, they did not reveal details about the effect of NG use on exhaust emissions. Equally, papers [41,52] only present the NG combustion in a single-cylinder research engine specific to an HDV. The results on the emissions (soot, NO_x, THC, CO) of a converted diesel tractor engine to NG SI operation are reported by [42] but only at one operating point: 2300 rpm and full load. More exactly, Aktas F. [42] showed the effect of spark timing on the emissions, but no comparison is provided with the original diesel engine, which was supposed to be replaced with the retrofitted NG application. Sandeep V. and Mittal M. [48] also reported on emissions. However, it was about a pure methane-fueled single-cylinder research application with a hemispherical-shaped bowl-in piston and only for one operating point (1.9 bar BMEP and 1500 rpm with a slightly lean mixture, characterized by an air-excess coefficient of 1.19).

Considering all the above, the experimental research presented in this paper was carried out with a retrofit prototype based on a European automotive manufacturer turbocharged CI engine dedicated to PC and LDV, which was adapted for SI operation with NG and LPG.

In addition to the previously mentioned stakes, this application could also help reduce dependency on traditional fossil fuels. This is especially important for certain regions given the current geopolitical situation.

This paper presents experimental findings on the energy and environmental performance of an inline four-cylinder diesel engine converted to SI operation, operating under lean-burn conditions with two gaseous fuels: natural gas (NG) and liquefied petroleum gas (LPG). The results are also provided for a baseline standard gasoline SI engine for the sake of comparison.

Consequently, after this brief introduction and literature overview outlining this application’s relevance in the current societal context, the paper proceeds with the presentation of the engine prototype and fuel properties, the methodology of research, and the results. Finally, the key findings and conclusions drawn from this research are discussed.

2. Engine Prototype and Fuels Properties

This experimental investigation carried out at the engine test bed (ETB) is part of a project aimed at obtaining a road demonstrator with compressed natural gas (CNG) stored at 200 bar in two cylindrical tanks (Figure 1), as also explained in [38].

As mentioned in the introduction, amongst gaseous fuels with a higher H/C ratio and heating value than gasoline, LPG was also used in the prototype engine to broaden the scope of the study. Thus, the whole enterprise was based on developing a prototype engine via converting an existing new CI engine to SI operation (i.e., retrofit) and burning CNG and LPG. The aim was to have a high CR to take advantage of the free knock operation of CNG (especially) and LPG thanks to their higher octane numbers.

The main technical characteristics of this prototype engine are presented in

Table 1. Details of the prototype engine.

Displacement [cm3]	1461
Bore [mm]	76
Stroke [mm]	80.5
Connecting rod length [mm]	133.5
Number of cylinders [-]	4
Compression ratio, CR [-]	18.25
Number of valves/cylinder [-]	2
Intake and exhaust manifold positioning	On the same side of the cylinder head
Combustion chamber	Bowl-in “omega” type
Injection type	Port-fuel injection (PFI)

.

The engine prototype uses a wiring harness and a control software from a commercial gasoline–LPG SI engine, which is produced by the same manufacturer as the base CI engine and has a displacement analogous to the prototype engine (1598 cm³). Given the particularity of LPG (stored in liquefied form in the tank and vaporized at the engine level via a heat exchanger), the control software taken over from this mass-produced bi-fuel SI engine was designed to always start the engine using gasoline. The engine control unit (ECU) switches to LPG injection only after the coolant reaches the 30 °C temperature threshold; thus, LPG vaporization is possible by extracting heat from the engine coolant in a heat exchanger. Since the modification of the control software was not part of this stage of the project, it means that the engine prototype had to be also equipped with a gasoline supply system. Consequently, the prototype engine will always start in “gasoline” mode. Then, the injection of the gaseous fuels is to be switched over according to the software specification, as explained before. Thus, the engine prototype uses the same gas injection system as the commercial bi-fuel engine and can operate separately either with CNG or with LPG (Figure 2). Obviously, if control software is to be developed specifically for CNG, it could allow the engine to start directly with this fuel because it is always in a gaseous state. This would eliminate the need for a separate vaporization process, as in the case of LPG.

In conclusion, although gasoline mode remains necessary, the prototype engine (created by retrofitting a high-CR diesel engine) serves only to illustrate the full ecological potential of CNG and LPG.

The transformation to SI operation of the original diesel engine was done solely at the cylinder head level. As already explained in papers [36,37], this was difficult because of the large number of parts added. In short, as seen in Figure 2, this involved adapting the spark plugs and their corresponding coils in the place of the diesel injectors, fitting the gasoline and gaseous fuel common rails and their corresponding injectors, and mounting the throttle body, the manifold absolute pressure (MAP) sensor, and the camshaft position sensor.

LPG supply was done from a toroidal tank (15 bar when completely filled-up) placed inside the engine test bed (ETB), while, for reasons of safety, the NG supply was done from outside the ETB, i.e., from the two tanks of a commercial bivalent gasoline-CNG vehicle (200 bar when completely filled-up), such as the blue ones in Figure 1, meaning that a T-type connector was added at the fueling pipe of the vehicle to allow for directing the NG towards the prototype engine being tested at the ETB. For the NG case, a pressure regulator lowers the high pressure from the tank to 5 bar at the level of the injectors, while, for the LPG, as already discussed, there is a vaporizer, which also warrants the pressure decrease to 5 bar. The gas injection is accomplished with a distinct control unit, called a GCU (Gas Control Unit).

Though the CI engine used for the retrofit has a variable nozzle turbocharging (VNT) unit, this investigation was conducted without making use of it, i.e., the engine was exploited as a naturally aspirated (NA) engine.

Regarding the physicochemical properties of the gaseous fuels used in our experiments, first, their components were found based on gas chromatography (GC). An equipment produced by Agilent™ (Santa Clara, CA, USA) was used for the NG analysis, while, for the LPG, the determination was based on Clarus™ (PerkinElmer, Shelton, CT, USA) equipment. The methods used relied on the ISO 6974-6 standard [53] for the NG analysis and on the EN 27941 standard [54] for the LPG analysis [38]. Figure 3 presents the results of chromatography, i.e., the compositions of the gaseous fuels used in the tests.

Based on the data from Figure 3, the number of carbon (C), hydrogen (H), oxygen (O) and nitrogen (N) atoms was calculated: C_1.0102H_3.9793O_0.0098N_0.0264 for NG and C_3.4311H_8.6544O₀N₀ for LPG. This allowed for the determination of the stoichiometric air (i.e., the minimum air mass needed for the ideal complete burning of 1 kg of fuel) for each of these two gaseous fuels used in the experiment. Thus, based on stoichiometric oxidation reactions, the stoichiometric air was found as follows: 16.73 kg of air for 1 kg of NG and 15.41 kg of air for 1 kg of LPG. These values are close to the ones obtained from theoretical calculations for methane–CH₄ (17.23) and propane–C₃H₈ (15.67), as given by Heywood [55].

On the other hand, further to the review of the published scientific results [3,23,24,41,56,57,58,59,60,61,62,63] concerning the physicochemical properties of the fuels used in the experiments, the following conclusions are reached: (1) data are occasionally offered for the alkanes (i.e., paraffinic hydrocarbons) with which NG and LPG are usually linked (methane, propane and/or butane), which seems reasonable since the chemical composition of these two fuels may differ; (2) sometimes, information is offered for NG and LPG without sufficient or even any details about their chemical compositions; (3) for either of the above two cases, one may notice a significant scattering of data; (4) occasionally, the values are given without stating the conditions used to obtain them (e.g., laminar flame velocities and adiabatic flame temperatures provided without mention of the equivalence ratio, the pressure and the temperature used in the corresponding tests, the oxidizer, etc.).

Consequently,

Table 2. Physicochemical properties.

	Fuel	Methane i CH4	NG ii	Propane iii C3H8	Butane iv C4H10	LPG v
Properties
Research octane number, RON [-]		(122–130) [61,63]	120	(110–112) [64,65]	92 [66,67]	105
Autoignition temperature, [°C]		585 [61]	540 [58,60]	450 [62]	490 [68]	470 [56]
Lower heating value, [MJ/kg]		50.0 [61]	47.5 [58,60]	46.4 [62]	45.7 [66,67]	46.1 [60,63]
Laminar flame velocity *, [cm/s]		39.0	41.0	42.0	41.1 [69]	-
Diffusion coefficient in air@25 °C, [cm2/s]		0.16 [61,70]	-	0.10 [70]	-	-
Stoichiometric air [kg air/kg fuel] **		17.17	16.73	15.61	15.39	15.41
Minimal ignition energy in air [mJ]		0.3 [41,61]	-	0.305 [70]	-	-
Air-excess coefficient: flammability limits (rich–lean)		0.70–2.10 [23,24]	-	0.40–1.95 [23,24]	0.34–1.86 [68]	-

ⁱ Methane references [41,59,61,62,63,70]; ⁱⁱ NG references [58,60]; ⁱⁱⁱ propane references [61,62,70]; ^iv butane references [66,67,68,69]; ^v LPG references [56,60,63]. * The data were taken from [59], which presents a comprehensive analysis of the laminar flame velocity providing all the needed data: equivalence ratio = 1, T = 298 K, p = 1 atm. ** Values obtained through calculation, as presented above.

only contains the data judged consistent, specific to the pure substances associated with NG and LPG (methane and propane), which contribute to the comprehension of the results presented in the next section. Regarding the RON specific to the gaseous fuels used in the experiments, the values delivered by the sale unit are 120 for CNG and 105 for LPG.

3. Methodology of Research and Experimental Setup

The resulting prototype was investigated at the engine test bed (ETB) of Pitești university Centre of the National University of Science and Technology “Politehnica” Bucharest, which is equipped with a Horiba™ WT190 Eddy-current brake. The ETB layout and its instrumentation are presented in Figure 4.

The operating points examined at the ETB are presented in Figure 5 in the [effective torque–speed] plane, which also shows the iso-power curves. The four operating points specific to the part loads and low speeds are both in the medium-low load area specific to urban and extra-urban driving. Of these four operating points, point B corresponds to an operation often addressed in ETB research, i.e., a brake mean effective pressure (BMEP) of 2 bar obtained at 2000 rpm. Regarding the other three points (A, C and D), they were taken from a data set specific to a commercial gasoline-powered S.I. engine with a comparable displacement (1598 cm³) produced by the same manufacturer. In other words, these three points were selected to compare different performance metrics obtained by the engine prototype with those of a commercial S.I. engine. Ideally, as also explained in [38], for an even more comprehensive analysis, this comparison should have been made with the original diesel engine, which is supposed to be replaced with the retrofit application. Unfortunately, no data were available regarding the base diesel engine used for the conversion. Efforts were made to find this data through available documents and the literature; however, relevant details may only be accessible through manufacturer records or specialized technical reports that were not accessible during this investigation.

As shown in Figure 4, all tests at the ETB were carried out with the alternator disabled (the battery being charged with a separate power supply) and at 80 °C ± 2 engine coolant temperature via an immersed heat exchanger in cold water whose flow was controlled in a closed loop thanks to an ON–OFF valve commanded by a temperature regulator; 25 °C ± 5 temperature test cell; 55 °C ± 5 intake air temperature (IAT) thanks to a fan–intercooler assembly (i.e., an ON–OFF air fan used to dissipate the intake air heat). Due to the absence of an air conditioning system in the test cell, IAT was controlled through forced ventilation. Under these conditions, a quasi-steady IAT of 55 °C ± 5 was achieved for all tests. Although higher than standard ambient conditions, this temperature level remained quasi-constant throughout the experimental campaign, ensuring good repeatability of the measurements. The reported results should therefore be interpreted as representative of elevated intake temperature operation. Generally speaking, the higher IAT enhances oxidation efficiency, resulting in lower carbon monoxide (CO) and unburnt hydrocarbon (UHC); additionally, the reduction in CO and UHC may also be a consequence of the probability of flame extinction near walls, which lowers with increasing the IAT. Conversely, since the reactants are preheated, the combustion temperature will be higher; this will generate an increased nitric oxides (NOx) emission and will suppress particle formation. At the same time, under these conditions, the likelihood of knock increases. However, as demonstrated in the section presenting the experimental results, if knock does not occur at elevated IAT, it is unlikely to manifest at lower temperatures. On the other hand, the increased IAT reduces air density, which lowers the volumetric efficiency and may locally enrich the mixture and counteract oxidation improvements in some regions. However, the influence of elevated IAT on oxidation kinetics is fuel-dependent. For NG primarily composed of methane (Figure 3a), the increased IAT significantly enhances reaction rates by overcoming the high activation energy required for initial radical formation, leading to faster flame development and the improved oxidation of intermediate species. In contrast, LPG, consisting of propane and butane, as shown in Figure 3b, already exhibits higher reactivity and active low-temperature chemistry pathways; therefore, the relative impact of intake air preheating could be less pronounced. Subsequently, NG benefits more strongly from elevated IAT, showing greater improvements in combustion efficiency and oxidation completeness compared to LPG.

In-cylinder instantaneous pressure acquisition was performed at 600 kHz with AVL’s IndiCom™ dealing, with pressure signals coming from instrumented glow-plugs (i.e., false glow-plugs—Figure 2), while the combustion-related parameters were obtained with Concerto 5 R4™ software based on the first law of thermodynamics. The crank angle encoder triggered the indicating data acquisition every 1 °CA. Considering the inherent cyclic variability, as thoroughly discussed in [71,72], data acquisition was conducted over 500 complete engine cycles for each sampling session, thus resulting in a relevant indicated diagram for each case based on averaging the acquired engine cycles. The number of the recorded cycles was selected considering statistical convergence criteria, as discussed also in papers [72,73]. Therefore, 500 consecutive engine cycles were judged sufficient to ensure statistical representativeness while avoiding excessive acquisition time, i.e., to obtain a statistically robust average and still capture the cycle-to-cycle variability. The latter is estimated based on the coefficient of variation (CoV), which is defined as the ratio between the standard deviation and the average value of the analyzed parameter.

Due to the intrinsic properties of the in-cylinder pressure piezoelectric transducer, it is essential to reference its output against absolute pressure. This means that the measured pressure must be calibrated using a known absolute pressure at a specific point during the engine cycle, a process commonly referred to as pegging. Papers [74,75] provide a comprehensive overview of pegging methods and examine their impact on the calculated parameters indicated. For our study, the cylinder pressure data were pegged by relying on “thermodynamic zero-line detection”, i.e., a fixed polytropic coefficient of 1.32 was applied within the angular interval (100–60) °CA before TDC, corresponding to the end of compression, and an averaging window of 4 °CA was utilized to generate a theoretical pressure curve for referencing the measured pressure curve.

Mass air flow (MAF) was measured every second with AVL Flowsonix™ (Graz, Austria) flowmeter based on an ultrasonic transit time measurement principle, featuring less than ±1% measurement uncertainty for an interval of 0–1400 kg/h. LPG and NG mass flow rates were measured at a frequency of 1 Hz with a Coriolis effect-based flowmeter produced by Emerson™ featuring a ±0.25% rate accuracy, having a range of 0–110 kg/h, working with gas supplied at 5 bar reduced from the high-pressure tanks. These measuring instruments are checked for metrological compliance every 3 months.

Regarding the exhaust gas analysis, as also shown in Figure 6, the usual portable emission measurement system (PEMS) from AVL was used. Consequently, the following raw engine-out gaseous species were recorded at a frequency of 5 Hz: carbon monoxide and dioxide (CO and CO₂) via an NDIR (Non-Dispersive Infrared) analyzer featuring a 1 ppm resolution for a range of 0–49,999 ppm with an absolute error of ±30 ppm between 0 and 1499 ppm and ±2% relative error between 1500 and 49,999 ppm for CO and a 0.01 vol% resolution for a range of 0–20 vol. % with an absolute error of ±0.1 vol.% between 0 and 9.99% and ±2% relative error between 10 and 20 vol.% for CO₂; nitric oxide and dioxide (NO + NO₂ = NO_x) via an NDUV (Non-Dispersive Ultraviolet) analyzer, with a range of 0–5000 ppm for NO and a range of 0–2500 ppm for NO₂, both having a 0.1 ppm resolution and a relative error of ±2%. The gas analyzers were calibrated (zero and span) and linearized before each test. Thanks to the diffusion charging-based particle counter from the AVL’s PEMS featuring a 10 Hz recording frequency, the number of particles larger than 23 nm was also recorded. As seen in Figure 6, the PEMS was placed outside the ETB, meaning that the exhaust pipe of the prototype engine was connected to the PEMS situated on a vehicle. The left picture from Figure 6 also shows the bivalent vehicle used to supply CNG to the ETB, as already explained.

The exhaust gas temperatures (EGTs) were measured for each cylinder at the inlets into the exhaust manifold with 2 mm diameter standard K thermocouples characterized by a sensitivity of 41 µV/°C and an accuracy of ±0.4%.

To reach the goal of the research, i.e., to reach the full ecological potential of these two gaseous fuels within the frame of our retrofit application, an appropriate engine calibration was aimed, that is to say, the spark advance (SA) and the duration of injection (DoI) were subject to change to optimize the engine operation. Changing of the DoI allowed or the exploration of the lean mixtures, meaning that, for each operating point, it was gradually reduced until the brake thermal efficiency (BTE) reached a peak. Obviously, to maintain the desired operating points, this was compensated by further opening the throttle plate. This was possible thanks to the use of ETAS-INCA V7™ software in connection with open ECU and GCU.

The experimental research was carried out as shown in Figure 7. First, a spark ignition advance sweeping, carried out at stoichiometry (λ = 1), was performed for both fuels used and for each of the four operating points approached (Figure 5), as discussed in paper [38]. Through this SA sweeping, the goal was to find the parabolic dependence between engine torque and SA, and hence the optimal values for the SA under the condition of having the maximum brake torque (MBT) and the lowest fuel consumption (therefore, the highest BTE) and operation without knocking.

Mixture leaning was then carried out by gradually reducing the DoI (as already described). Regarding the optimum SA for the lean cases, the iterations carried out at ETB led to the conclusion that the optimal values found when operating with a stoichiometric mixture are also optimal for operation with a lean mixture. This conclusion was not expected because, as is well known, lean mixtures exhibit lower laminar flame velocities, which are to be compensated by higher SA. However, it is believed that, in such applications featuring diesel-like architecture, increasing SA further means producing the spark in a non-favorable environment, thus negatively affecting the BTE.

As for the phases specific to each test, as previously described, the engine always started with gasoline. After reaching the 30 °C temperature threshold, the engine was switched to one of the gaseous fuels and brought to the first operating point (Figure 5). Before starting the first recording session, it was run for 30 min to ensure that any remaining gasoline residues were removed from the fuel system, combustion chamber, and gas analyzers, thereby preventing interference with subsequent measurements. For every engine operation point (Figure 5), before starting the recording session, the engine was run for 5 min to allow the temperature and operating conditions to stabilize.

4. Results and Discussion

Figure 8 shows the impact of mixture leaning on the BTE (i.e., effective efficiency, ɳ_e) for each of the two gaseous fuels, LPG and CNG. The commercial gasoline engine case is also included in this figure (obviously at stoichiometric mixture), from which one may easily see the benefits of the higher compression in the case of the prototype engine (18.25 vs. 9.7): all the BTE points specific to the prototype engine are above the ones corresponding to the commercial gasoline engine—i.e., a gain between 3 and 5 percentage points, which corresponds to a relative improvement of (14.3–20.4)%.

The maximum BTE values (Figure 8) occur, for all cases, in the range λ = (1.24–1.26)—from now on, these λ values are referred to as the lean limits for this application. To be able to extend the lean limits with subsequent benefits for BTE and CO₂, as pointed out in [27], one way would be to improve the spark production, i.e., increasing the discharge energy at the spark electrodes and even relying on a multi-spark event; another method to enhance combustion is by relying on hydrogen addition, as specified in [76].

For all cases, the BTE differences from the stoichiometric mixture operation may also be observed in Figure 8, which also highlights a slightly greater improvement when operating with CNG lean mixtures, i.e., between (1.35 and 2.35) percentage points gained in this case compared to (1.15–2.10) percentage points gained in the case of LPG. Finally, regarding the effective efficiency, it increases with increasing load (B, C, A, D) because of the reduction in pumping losses.

As stated, the optimal values of SA were found also from the condition of having a knock-free operation. To evaluate the knocking intensity, the usual knock pressure peak (kp_pk) was used, as also explained in [38]. This parameter, available in the post-processing software of the indicated equipment (Concerto^TM) can be defined as the peak pressure amplitude deviation from the normal combustion pressure. Figure 9 presents the comparative evolution of the kp_pk parameter, contrasting stoichiometric and lean conditions across 500 recorded cycles. Additionally, statistical outcomes—including minimum, mean, and maximum values, as well as the coefficient of variation (CoV)—are provided. Lean operation demonstrates a consistently knock-free performance despite the high value of CR (18.25), whereas stoichiometric operation exhibits only occasional knocking pressure peaks between 1 and 2 bar in the LPG scenario.

Figure 10 shows the correlation between the indicated thermal efficiency (ITE) and the duration of combustion (DoC), calculated according to the standard convention [55], as the difference between the angles corresponding to a 90% mass burnt fraction (MBF90) and 5% mass burnt fraction (MBF5). In this figure, one may also see the optimum SAs used for both fuels. Therefore, it can be observed that, in every case, the enhancement of ITE is attributable to the decrease in DoC (i.e., when using lean mixtures, the corresponding 500 point clouds move to the upper left direction). The mean values of ITEs and DoCs may be noticed in Figure 10 thanks to the yellow circles. The effect of the compression ratio may also be observed in Figure 10 when comparing the prototype engine (CR = 18.25) with the baseline commercial gasoline engine (CR = 9.7): the higher the CR, the more intense the combustion kinetics; hence, the lower the DoC and consequently the higher the ITE. Figure 10 also outlines slightly higher values of ITE for the CNG case.

The reduction in DoC with the lean mixture (also illustrated in Figure 11d) is explained as follows: as seen in Figure 11a, lean mixtures slightly delay flame kernel inception (the crank angle corresponding to MBF5 increases), as also reported in [27], while the crank angle corresponding to MBF90 (conventionally considered as the end of combustion) either remains quasi-constant or it decreases (Figure 11c). The latter may be attributed to the lower fuel mass burnt when using lean mixtures compared to the stoichiometric case. In other words, it may be about a combustion process which is energetically limited rather than a combustion with a lower kinetics, hence the shorter DoC in the lean case.

MBF50 is usually associated with efficiency. As seen in Figure 11b, for the CNG case, the mean values of MBF50 for the lean limit obtained with maximum BTE were below 15 °CA, which suggests a reliable flame propagation under correct combustion phasing, considering that, according to [47], conventional SI engines typically work with an MBF50 between 10 and 15 °CA after TDC. The same Figure 11b shows slightly higher values for MBF50 in the case of LPG. This is because the MBF5 values in the case of LPG are higher (Figure 11a), meaning that retarded MBF5 values further delay the MBF50 combustion phase (vice versa is valid in the CNG case).

Figure 11 also indicates the extent of the scattering with respect to the mean values of these different metrics used to evaluate combustion. It is clear that leaning the mixture beyond the values corresponding to the maximum BTE (i.e., beyond the lean limits) leads to a further increase in the cyclic variability, which negatively affects engine stability. This is also seen in Figure 12, which presents the cyclic variability applied to the indicated mean effective pressure (IMEP) parameter.

As also reported in [27,47], the probable causes of this increased cyclic variability are spark energy variations and in-cylinder turbulence changing (influencing both the velocity and the direction of flame propagation). However, the positive effect of an increased CR in the case of the prototype engine may also be noticed in the same Figure 12, since the cycle-to-cycle IMEP variation at the lean limits corresponding to the maximum BTE remains inferior to that of the commercial baseline gasoline engine.

Table 3. Impact of lean burn on prototype engine’s emissions.

			A(1600 rpm, 35.6 Nm)			B(2000 rpm, 23.5 Nm)			C(2300 rpm, 30.5 Nm)			D(2300 rpm, 49.8 Nm)
L P G		λ	λ = 1.00	λ = 1.24	Δ [%]	λ = 1.00	λ = 1.24	Δ [%]	λ = 1.00	λ = 1.24	Δ [%]	λ = 1.00	λ = 1.24	Δ [%]
	Emissions
	CO2 [kg/h]		3.61	3.53	−2.23	3.57	3.40	−4.82	5.06	4.86	−3.88	7.43	7.22	−2.83
	CO [kg/h]		0.190	0.040	−76.1	0.170	0.050	−72.5	0.270	0.070	−74.4	0.350	0.080	−76.6
	NOx [kg/h]		0.086	0.067	−21.6	0.069	0.043	−38.1	0.121	0.077	−36.3	0.176	0.119	−32.1
	PN [#1014/h]		4.11	3.54	−13.8	3.31	1.11	−66.4	9.49	4.06	−57.2	3.45	3.50	1.60
C N G		λ	λ = 1.00	λ = 1.25	Δ [%]	λ = 1.00	λ = 1.24	Δ [%]	λ = 1.00	λ = 1.24	Δ [%]	λ = 1.00	λ = 1.26	Δ [%]
	Emissions
	CO2 [kg/h]		3.21	3.07	−4.55	3.15	3.00	−4.79	4.49	4.23	−5.70	6.36	6.14	−3.58
	CO [kg/h]		0.130	0.030	−74.6	0.130	0.030	−73.4	0.210	0.050	−78.3	0.400	0.060	−85.8
	NOx [kg/h]		0.079	0.047	−52.1	0.056	0.029	−48.8	0.108	0.066	−38.4	0.157	0.118	−24.3
	PN [#1014/h]		2.96	2.17	−26.9	3.06	1.86	−39.3	6.51	3.21	−50.8	4.18	0.678	−83.8

presents a comparative analysis between the engine-out emissions specific to the stoichiometric operation and to the lean mixtures corresponding to the maximum BTE (i.e., lean limits), as shown in Figure 8.

As shown in Table 3, the decrease in CO₂ emissions generated by the lean mixtures does not seem to be spectacular. However, for a more comprehensive analysis, two additional aspects are presented:

• CO₂ reduction was achieved through lean operation with respect to the commercial gasoline engine (Figure 13); consequently, thanks to this figure, the perspective is changed—e.g., a 28% reduction when operating with CNG is quite notable (a reminder is considered useful: in the datasheet of the commercial gasoline engine, data are provided only for points A, C and D; point B was added to the testing matrix simply because it corresponds to an operation sometimes addressed in ETB research; hence, the possibility of comparison with other researchers increases).
• It was considered useful to quantify CO2 emissions in [g/km], knowing the limitations related to European regulation no. 83 [77]: starting with 1 January 2025, for the average of the fleet sold, penalties (95 euros for each gram of CO₂ exceeded and for each registered car) began to be applied if the limit value of 94 g CO₂/km, obtained during the homologation on World-wide Harmonized Light-duty Testing cycle (WLTC), is exceeded. Therefore, if point B is assimilated with the steady movement of a mid-class car at a velocity of 54 km/h, then the emission of 3 kg CO₂/h, obtained in the case of using CNG and an excess air coefficient of 1.24, corresponds to a value of 55.5 g CO₂/km, 2.8 g less than the situation of operating with a stoichiometric air-NG mixture and 10.7 g less than operating with a stoichiometric air-LPG mixture.

Regarding the raw engine-out pollutant emissions, as expected, using lean mixtures is quite beneficial, as shown in Table 3. The only exception was recorded for the PN when fueling with LPG for point D and, most likely, was caused by a technical problem with the particle counter. The oxygen excess contributes to the increase in the oxidation degree of the carbon atoms, which in its turn generates the lowering of CO and PN. As for the NO_x decrease, it was caused by the reduction in the in-cylinder maximum temperature while leaning the mixture (Figure 14). This parameter was calculated by employing the state equation under in-cylinder conditions. Where available, Figure 14 also includes the data specific to the commercial gasoline engine, just to have a measure of comparison (i.e., for points A, C and D). Thus, one may clearly see the effect of the prototype engine’s higher compression. However, despite the high CR of the prototype engines, the CNG lean mixtures decreased the NO_x emissions even below the values obtained with the commercial gasoline engine: 0.0478 kg/h vs. 0.0720 kg/h for point A, 0.0665 kg/h vs. 0.0677 kg/h for point C and 0.1189 kg/h vs. 0.1382 kg/h for point D.

Figure 15 highlights the influence of lean mixtures on the heat release rate (RoHR) and on the evolution of the cylinder pressure (pcyl). The effect of the lean mixture on the RoHR (Figure 15a,c,e,g) is as expected, in the sense that lean mixtures generate an obvious reduction in the heat release rate which, beyond a limit value of the excess air coefficient, leads to combustion degradation, with negative effects on the BTE, as shown in Figure 8. Therefore, the internal fluid dynamics of the prototype engine, specific to a CI engine (swirl and squish), are not sufficient to support additional lean mixture, despite the high value of CR of the prototype engine. So, beyond this limit value of λ, the beneficial effects generated by the lean mixture are blurred by the reduction in the RoHR, causing the efficiency to decrease and the CO₂ emission to increase.

Moreover, the evolutions of RoHR (Figure 15a,c,e,g) also highlight how lean mixtures contribute to a decrease in DoC, despite exhibiting lower laminar flame velocities. Following ignition, a noticeable delay in SoC (i.e., MBF5) is observed. However, during the descending phase of RoHR, higher values are present at equal angular positions, indicating that combustion with lean mixtures tends to recover toward its end when compared to stoichiometric mixtures. Consequently, the end of combustion (EoC) remains nearly constant or decreases, as illustrated in Figure 11c.

Regarding the in-cylinder pressure (Figure 15b,d,f,h), higher values are noted during the compression stroke in all cases. This is explained by the additional amount of air inside the cylinder; as already discussed, the operation with lean mixture, while maintaining the operating point, is ensured by an additional opening of the throttle plate, hence the increased manifold absolute pressure (MAP) and consequently the lower pumping work. For example, in the case of the most loaded point (point D), when operating with a lean air-CNG mixture (Figure 15h), this generates pressures that exceed the values measured in the case of the stoichiometric mixture. This occurs for CNG because, as seen in Table 2, the stoichiometric air for CNG is higher than the one specific to LPG. This affects the in-cylinder peak temperature (Figure 14) and, as a result, influences NO_x emissions, which at point D match those produced by burning LPG (see Table 3).

These explanations are also supported by the higher values of volumetric filling efficiency in the case of lean mixture operation, as can be seen in Figure 16. The filling efficiency was calculated by dividing the real fresh fluid consumption (measured MAF and mass fuel flow, MFF) by the ideal or theoretical one. The latter refers to an intake process without pressure and temperature losses, calculated under the standard temperature of 20 °C (NTC—Normal Technical Condition). Obviously, the elevated IAT specific to our tests (mentioned in Section 3) leads to lower values for the calculated filling efficiency. When analyzing at iso-operating point, compared to LPG, the use of CNG, as explained throughout this section, is accompanied by an increase in the air mass per cycle and cylinder, which is obtained through an additional opening of the throttle; in turn, this generates a higher MAP, a reduced pumping, and consequently a higher filling efficiency.

Figure 17 indirectly highlights the mechanical efficiency, i.e., the higher the data points are positioned above the bisector (the dashed line where IMEP equals BMEP specific to idealized zero mechanical losses), the lower the mechanical efficiency.

In addition, when the slope of the data lines decreases and the data points move closer to the bisector line, it means that the mechanical efficiency increases. Thus, Figure 17 shows that the lean mixtures come with lower mechanical efficiencies than the stoichiometric mixtures. Since the data specific to each of the four operating points is represented for the same BMEP, this is because of an increased IMEP obtained with lean mixtures (Figure 18), causing the increased ITE, as shown in Figure 10.

This increase in IMEP is caused by higher MAP values as previously discussed. These higher MAP values raise in-cylinder pressures, which ultimately results in increased friction. For instance, the lowest mechanical efficiency (82.8%) is in point D for the CNG case with lean mixtures. This correlates with the in-cylinder pressure evolution (Figure 15h), which in this case is higher for the lean case than for the stoichiometric case during the whole engine cycle.

In summary, lean mixtures exert a directly positive influence on ITE (Figure 10) while adversely affecting mechanical efficiency. Nonetheless, BTE remains higher under lean conditions (Figure 8), resulting in reduced MFF, as demonstrated in Figure 19, which further highlights the reductions achieved compared to the stoichiometric scenario.

In the case of the lean mixtures, the lower in-cylinder maximum temperatures (Figure 14) corroborated to either quasi-constant or decreased MBF90 (Figure 11c) lead to lower exhaust gas temperatures (EGTs) with respect to the stoichiometric conditions, as displayed in Figure 20.

This has the advantage of making possible the coupling of the engine with a variable nozzle turbine (VNT), with subsequent positive effects on the drivability of the car, as also discussed in [29,35]. However, lower EGTs inherently increase the duration of the EATS light-off, i.e., the duration needed to attain the EATS temperature threshold for efficient treating of the pollutants [11]. This problem, however, may be overcome by different ways, as also pointed out in [78,79].

5. Summary and Conclusions

In an attempt to completely exploit the potential of CNG and LPG in terms of energy and environmental performance, this paper approached lean burning in a diesel engine converted to SI operation. As a comparison measure, the results were also given for a baseline commercial gasoline engine, thus providing a reference required to assess the outcomes of a low-carbon gaseous fuel-based prototype engine.

The relevance of such an application related to the current and future context was substantiated with accurate and complete data attained via experimental research, letting others replicate and develop the obtained results.

The main findings of this experimental investigation regarding the lean burning of LPG and CNG gaseous fuels are given below with respect to the baseline commercial gasoline SI engine and are twofold:

• Energy performance
  • There was a greater gain in BTE when operating with CNG lean mixtures, between 3 and 5 percentage points, which corresponds to a relative improvement of (14.3–20.4)%.
  • There was a slightly lower IMEP variability with LPG than with CNG; however, both were below the values of the baseline stoichiometric commercial gasoline engine (e.g., more than halved in the case of point C for LPG).
• Environmental performance
  • GHG-related CO₂ was clearly decreased with the gaseous fuels, e.g., between a (19.3 and 28.0)% reduction for CNG lean mixtures, the highest drop being recorded at point A.
  • CO emission was also reduced significantly with lean burning: between (70.9 and 79.4)% for LPG and (78.8 and 85.6)% for CNG, the highest cuts being recorded at point D for both fuels.
  • Despite the prototype engine’s higher compression ratio, lower NO_x emissions were observed with gaseous fuel lean burning, e.g., a maximum 33.6% fall for point A in the case of CNG.
  • As for the PN, because there is no data available for the baseline commercial gasoline SI engine, the comparison made between the two gaseous fuels displayed a clear improvement for the CNG case, between a (21.1 and 80.6)% reduction, the highest difference being recorded at point D.
  • The absence of unburnt methane measurement constitutes a limitation of our research, given the GWP associated with methane.

Thus, the results show that the prototype engine was able to provide clear improvements in terms of energy and environmental performance with lean burning. The drawback of the lower laminar flame velocity associated with lean mixtures was compensated by the high compression and the in-cylinder fluid dynamics of the converted diesel engine. The extension of the lean limit with subsequent positive effects on ecological performance may be further explored with an improved ignition system. However, as mentioned, lean mixtures no longer allow the use of the classic and reliable three-way catalytic converter (TWC), meaning that a dedicated EATS must be developed to ensure an efficient conversion over a wide range of operating conditions despite the relatively low raw engine-out emissions level (as also presented in this paper) and finally to comply with future regulations.