Combustion and Emission Analysis of NH₃-Diesel Dual-Fuel Engines Using Multi-Objective Response Surface Optimization

Abstract

As internal combustion engines (ICEs) remain dominant in maritime transport, reducing their greenhouse gas (GHG) emissions is critical to meeting IMO’s decarbonization targets. Ammonia (NH₃) has gained attention as a carbon-free fuel due to its zero CO₂ emissions and high hydrogen density. However, its low flame speed and high ignition temperature pose combustion challenges. This study investigates the combustion and emission performance of NH₃-diesel dual-fuel engines, applying Response Surface Methodology (RSM) for multi-objective optimization of key operating parameters: ammonia fraction (AF: 0–30%), engine speed (1200–1600 rpm), and altitude (0–2000 m). Experimental results reveal that increasing AF led to a reduction in Brake Thermal Efficiency (BTE) from 39.2% to 37.4%, while significantly decreasing NO_x emissions by 82%, Total hydrocarbon emissions (THC) by 61%, and CO₂ emissions by 36%. However, the ignition delay increased from 8.2 to 10.8 crank angle degrees (CAD) and unburned NH₃ exceeded 6500 ppm, indicating higher incomplete combustion risks at high AF. Analysis of variance (ANOVA) confirmed AF as the most influential factor, contributing up to 82.3% of the variability in unburned NH₃ and 53.6% in NO_x. The optimal operating point, identified via desirability analysis, was 20% AF at 1200 rpm and sea level altitude, achieving a BTE of 37.4%, NO_x of 457 ppm, and unburned NH₃ of 6386 ppm with a desirability index of 0.614. These findings suggest that controlled NH₃ addition, combined with proper speed tuning, can significantly reduce emissions while maintaining engine efficiency in dual-fuel configurations.

1. Introduction

At present, more than 80% of all global commerce is carried out by ships, making maritime transportation one of the largest economic industries in the world [1,2]. Since this industry involves high volumes and a great distance, it is mainly powered by internal combustion engines (ICEs) which account for 98% of the merchant fleet [3,4]. This industry accounts for about 10% of total transport emissions and 2.9% of total emissions [5,6]. Therefore, it is essential to identify alternatives which will be economically and ecologically sustainable for maritime transportation. Furthermore, global climate agreements, such as the Paris Agreement [7], which call for increasingly stringent limits on greenhouse gas (GHG) emissions, have directed combustion research toward the development and adoption of low-carbon fuels. This shift is specifically crucial for ICEs [8,9]. The International Maritime Organization (IMO) is actively facilitating this transition by launching extensive capacity-building initiatives and setting a notable goal to cut shipping-related GHG by 50% (in comparison to 2008 levels) by 2050. To achieve this target, the IMO encourages the use of ICEs that run on zero-carbon or low-carbon fuels such as NH₃, H₂, biofuels, methanol, and liquefied natural gas (LNG), along with alternative propulsion systems including fuel cells, electric power, and wind energy [10,11].

The shift to cleaner fuels not only facilitates regulatory compliance but also yields economic advantages by reducing reliance on conventional fuels with volatile prices and by meeting the increasing demand for sustainable mobility [12]. In recent years, NH₃ has emerged as an appealing alternative to ICEs, especially for marine use, because of its environmentally friendly properties and potential for lower carbon emissions [13]. Extensive research into using NH₃ as a fuel source has led to various solutions for combustion challenges [14,15]. Among the possible solutions are increasing spark energy, engine supercharging, and using a second fuel. Most promising, for instance, is a dual-fuel technique called Reactivity Controlled Compression Ignition (RCCI) as proposed by Reitz et al. [16,17]. The RCCI approach seeks to improve thermal efficiency while at the same time lowering NO emissions in diesel engines. This method involves adding two fuels of different reactivity into the engine cylinder: a non-reactive fuel is mixed into a mixture, which is then ignited by a reactive fuel, such as diesel or biodiesel [16]. This dual-fuel strategy aligns well with the low reactivity of NH₃.

Zhang et al. [18] investigated the application of NH₃ in a two-stroke, low-speed marine engine in a laboratory conditions. Their research examined how factors such as NH₃ injection timing, the mass of NH₃ injected, and the timing of diesel injection affected the engine’s performance and emission profiles. Their findings indicated that the timing of diesel injection plays a crucial role in igniting liquid-phase NH₃ spray. Furthermore, optimizing the injection timings for both diesel and NH₃ can lead to a reduction in NO emissions. Additionally, Reiter et al. [19] proposed a scenario in which the majority of the fuel energy (95%) was supplied by NH₃, complemented by a small pilot injection of diesel (5%). In dual-fuel configurations, the combustion of NH₃ often leads to elevated levels of unburned NH₃ and NO emissions. Specifically, when the NH₃ contribution to the total fuel energy is below 40%, NO emissions during dual-fuel operation are lower than those seen when utilizing 100% diesel. However, when the proportion of NH₃ exceeds 60%, emissions of NO rise markedly due to the N₂ present in NH₃. Yousefi et al. [20,21] performed a series of experiments to assess the performance and emission characteristics of diesel engine in an ammonia/diesel dual-fuel (ADDF) setup, which shares similarities with the RCCI engine examined in this study. Their findings revealed that a single diesel injection in the ADDF engine results in an ITE lower than that of a diesel engine operating solely on diesel fuel. Conversely, employing a double diesel injection strategy leads to an enhanced ITE relative to the baseline diesel engine despite a 10% increase in NO emissions. Moreover, the concentration of unburned ammonia in the exhaust consistently exceeds the recommended emission thresholds.

To enhance the performance characteristics and emissions of ICEs, optimization approaches have been employed to fine-tune operational conditions [22,23,24,25,26]. A key technical benefit of optimizing the incorporation of bio-based components into diesel is that it can improve engine performance and emissions while allowing the use of this optimized fuel blend in IC engines without requiring modifications [27,28]. To achieve effective optimization, design of experiments (DoE) methodologies, particularly Response Surface Methodology (RSM), can be employed. RSM is a well-established technique for addressing various industrial challenges. It serves as a practical and cost-effective approach to evaluating both individual and interactive effects of different experimental variables on the desired output responses [29]. The insights gained from RSM analysis facilitate the identification of the most efficient system performance across all optimized parameters [30]. A significant advantage of RSM-based experimental design is that it demands fewer tests and reduces time consumption compared to traditional experimental studies [31]. This methodology has been extensively used in numerous research investigations.

The main objective of the study is to perform MOO of NH₃-diesel dual fuel utilization in ICEs using RSM. The goal is to enhance engine performance and reduce emissions by investigating the effects of various engine parameters, such as ammonia fraction (AF), engine speed, and altitude, on key performance metrics and emissions profiles, including BTE, ignition delay, Center of combustion (CD50), CD90, and various emissions including NO_x, NH₃, THC, and CO₂. The objective of this study is to assess the effectiveness of RSM in enhancing both engine performance and emissions reduction, ultimately determining the optimal ratio of NH₃ in the fuel blend.

2. Methodology and Experimental Setup

The engine used in this research was a 3.0 L, 4-cylinder turbocharged diesel with high pressure common-rail injection. The technical data of the engine is presented in

Table 1. Technical data of the engine.

Parameter
Type of Engine	Four strokes
Cylinder Number	Four
Bore × Stroke	9.5 cm × 10.5 cm
Engine displacement	3.0 L
CR	16:1
Maximum torque at Speed	400 Nm at 1400 r/m
Rated power at Speed	115 kW at 3200 r/m

. There were a couple of key pieces of equipment that were in the experimental design to accurately test and evaluate the performance and emissions of the engine operating with an NH₃–diesel mixture. The main instruments were an AVL power dynamometer for engine performance, an AVL AMA i60 (AVL List GmbH, Graz, Austria) for emission measurement, and an NH₃ laser analyzer for NH₃. The AVL AMA i60 emission analyzer was zeroed and spanned daily using certified calibration gases (NO, NO₂, CO, CO₂, and O₂) traceable to national standards. Characteristics of combustion were measured using an AVL 622 combustion analyzer. The Coriolis NH₃ mass flowmeter and AVL intake air flowmeter were zero-calibrated before each test run using inert gas and ambient air, respectively. Fuel consumption measurements with the AVL 735 were verified using gravimetric checks. Intake air temperature and humidity sensors were cross-checked against reference thermocouples and hygrometers, ensuring accuracies within ±0.5 °C and ±1% RH. During testing, intake pressure, exhaust back pressure, and environmental conditions were continuously monitored and logged, with pressure maintained within ±0.1 kPa of the set value. This rigorous calibration and monitoring procedure ensured measurement accuracy and repeatability across all simulated altitude and NH₃ fraction conditions.

The placement and integration of these instruments in the experimental environment is shown schematically in Figure 1. Such a setup allows for accurate data acquisition, allowing the engine’s behavior and emissions to be measured accurately under a range of operating conditions and NH₃ fractions.

This work explored the impact of varying altitudes—specifically at 0 m, 1000 m, and 2000 m, NH₃ fractions (AF) of 0, 10, 20, and 30%, and engine speeds of 1200, 1400, and 1600 rpm. The tests were conducted at a test bench at 2000 m, with ambient pressure naturally 80 kPa. For lower altitudes, the intake and exhaust lines were pressurized accordingly. For 1000 m, the pressures were further raised by 10 kPa and for 0 m by 20 kPa. Intake pressure was controlled by a dedicated air-conditioning (AC) system with variable-frequency drive (VFD) fan, and exhaust pressure was controlled by a back-pressure valve. The AC system was used to control intake pressure, temperature, and humidity. Intake air temperature and humidity were conditioned by chiller and heater. The AC system is able to maintain a constant intake temperature and humidity within ±2 °C and ±3% RH. In all tests, intake temperature and relative humidity were set at 25 °C and 50% RH, respectively. The AC system therefore provided a constant and reproducible environment for testing.

3. Design of Experiments Analysis

Response Surface Methodology (RSM) is a valuable tool across various engineering disciplines [31]. It employs statistical techniques using quadratic or linear polynomial functions to model the relationship between input variables and responses, aiming to either maximize or minimize the desired response properties. In this experiment, three key parameters were examined: ammonia fractions (AF), along with variations in engine speed and operating altitude. Utilizing RSM, custom-defined experimental designs were implemented. Given the parameters in this study, a discrete function was applied. This discrete approach outlines the specific settings for factors, facilitating easier experimental execution without significantly compromising analytical rigor. To estimate the effects of the three parameters at four levels, 36 experimental runs were conducted.

Table 2. Parameters and their Levels.

Factor	Parameter	Level 1	Level 2	Level 3	Level 4
A	Speed (RPM)	1200	1400	1600
B	NH3 fraction (%)	0	10	20	30
C	Altitude(m)	0	1000	2000

lists the independent variables, their specified levels, and codes. The responses measured included BTE, ignition delay, CD50, D90, and various emissions including NO_x, NH₃, THC, and CO₂. Analysis of variance (ANOVA) was employed to assess the variances between group means and their interactions. In this context, the degree of freedom (DF) indicates the probability distribution in repeated sampling. A high F-value suggests a significant variation between the groups represented by each sample. The significance level was set at a 95% confidence interval (Prob > F ≤ 0.05). The percentage contribution (PC%) serves as an indicator of the relative importance of each model term, calculated as shown in Equation (1):

$$PC={\displaystyle \frac{SSd}{SST}}\times 100\%$$

(1)

where SSd is the sum of squared deviations and SST is the total sum of squared deviations.

The simplest RSM model is a linear function, represented by Equation (2):

$$y=b0+\sum bi xi$$

(2)

If the model indicates curvature, a second-order model is employed, as appears in Equation (3):

$$y=b0+\sum bi xi+\sum bii {xi}^2+\sum bii xi xi$$

(3)

For this study, a quadratic model was suitable to find out the critical points (minimum or maximum), using Equation (4):

$$y=b0+\sum bi xi+\sum bii {xi}^2+\sum bii xi xi+ \epsilon$$

(4)

where k is the number of variables (3 in this case), b0, bi, and bii are the coefficients of the linear, quadratic, and interaction terms, respectively, and ε is the residual error.

Optimal parameter settings were determined using a desirability function approach in Design Expert v.13, aiming to enhance engine performance and minimize emissions. The expected ranges for each response and parameter were set, and multi-response optimization was conducted.

4. Results and Discussion

This section presents an analysis of engine performance and emissions, focusing on the impact of key parameters including AF, speed, and altitude on a CI engine. BTE, ignition delay, center of combustion (CD50), burning duration, and emissions of NO_x, NH₃, THC, and CO₂ were plotted as a function of these inputs. 3D surface plots and corresponding contour plots were generated based on these three independent factors for the CI engine. Employing an empirical model, a correlation was found between the responses, and the variables were developed. After all, a multi-objective optimization is carried out using a desirability function.

Figure 2 and Figure 3 illustrate how BTE varies with different AF under engine speeds of 1200, 1400, and 1600 rpm. When the speed is constant, increasing the AF leads to reduced BTE. This reduction is due to higher NH₃ consumption and lower diesel consumption, along with ammonia’s lower combustion efficiency. Therefore, elevated levels of AF contribute to a decline in BTE. Furthermore, at a fixed AF, BTE decreases significantly with an increase in engine speed.

To examine of impact of an individual parameter on BTE in more detail, Figure 3 and Table S1 illustrate the effects of a single parameter. Table S1 presents the ANOVA data for the BTE various examined parameters. The F-value of 290.28 suggests that the model fits well, with a p-value below 0.05. A high R² value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA Table S1 indicates that AF and speed have significant impacts on BTE, while the altitude does not. The results also confirm that AF has a greater effect on BTE, suggesting a reliable method of BTE estimation. The R² statistic being close to 1 implies a strong correlation between the observed values and measured results. The percentage contribution (PC%) indicates the significance of each model term. The AF contributed the most to BTE variability (74%), followed by speed (17%), with altitude having an insignificant effect, which suggests that ammonia combustion is effective even at high altitudes. This indicates that higher altitudes are suitable for ammonia combustion.

A linear model was used to fit the BTE data. The BTE is expressed via Equation (5).

$$\mathrm{B}\mathrm{T}\mathrm{E}=\mathrm{b}37.28-0.7584\times \mathrm{S}\mathrm{p}\mathrm{e}\mathrm{e}\mathrm{d}-1.7\mathrm{N}\mathrm{H}3\times \mathrm{f}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}-0.898\times \mathrm{A}\mathrm{l}\mathrm{t}\mathrm{i}\mathrm{t}\mathrm{u}\mathrm{d}\mathrm{e}$$

(5)

In the BTE response, the key factors were NH₃ fraction, then speed, as shown in Supplementary file Figure S1. The BTE plotted against AF and speed displayed important variations, while the BTE plotted against altitude showed an insignificant effect. Figure S1A demonstrates a significant decrease in BTE, and Figure S1B shows that the speed–BTE curve was steeper than the AF–BTE curve, indicating a greater impact of AF. The altitude–BTE plot in Figure S1C reveals a slight reduction in BTE with a rise in AF. Overall, the analysis confirms that AF and speed are the primary factors governing BTE, with altitude showing only a minor influence. Increasing AF consistently reduces BTE due to ammonia’s lower combustion efficiency, while higher speeds further decrease efficiency. These results highlight that AF plays the dominant role in shaping CI engine performance, providing a reliable basis for predictive modeling and optimization.

The combustion characteristics of NH₃/diesel blends are influenced by the unique properties of NH₃, which include a low flame speed and high ignition temperature. Understanding how NH₃ affects the ignition delay and combustion duration when blended with diesel is crucial for optimizing engine performance and emissions control. This analysis focuses on the specific impacts of varying AF ranging from 0% to 30% on the combustion behavior in diesel engines, which include ignition delay, center of combustion (CA-50), and combustion duration (CA90).

Figure 4A,B illustrates a 3D plot of ignition delay for diesel and NH₃ at varying AF under different engine speeds and altitudes. As NH₃ replaces more diesel, the ignition delay extends. This delay increases from 8.2 CAD with net diesel to 8.6 CAD at maximum AF, assuming constant engine speed and altitude. This increase in delay is attributed to the entry of gaseous NH₃ into the intake, which results in diminished air mass flow within the cylinder. Additionally, the high minimum ignition energy and temperature required by NH₃ contribute to a prolonged ignition delay. At the highest AF, combustion starts at TDC, where the pressure and temperature make for the best ignition. Figure 5, on the other hand, shows how the ignition delay goes from 8.2 CAD for net diesel to 10.8 CAD at the highest AF and engine speed.

To examine the impact of an individual parameter on ignition delay in more detail, Figure 6 and Table S2 illustrate the effects of a single parameter. Table S2 presents the ANOVA data for the ignition delay various examined parameters. The F-value of 121.23 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is desirable and should reasonably agree with the adjusted R². The ANOVA results in Table S2 indicate that engine speed is the only significant factor, while the AF and altitude are not. The R² statistic close to 1 implies a strong correlation between the observed values and measured result. The engine speed contributed the most to the ignition delay variability (81%), followed by AF (0.1%), with altitude having an insignificant effect. Similar results were reported by Nadimi, et al. [32]; with the increase of AF up to 30%, the start of combustion (SOC) and ignition delay remains almost unchanged at around 8 CAD. When AF rises higher, there would be a prolonged effective ignition delay, from 8.7 CAD (pure diesel) to 15.9 CAD (highest AF). Previous studies have shown that the ignition delay remains relatively insensitive to low ammonia fractions (<20% AF) but increases sharply at higher substitution ratios (>30% AF, significantly affecting the premixed combustion phase (Liu & Liu, 2023 [33]). Junheng Liu, et al. [34]. study also showed that ignition delay remains relatively insensitive to low ammonia fractions, with only minor increases observed at AEF below 30%, while higher ammonia fractions significantly prolong the ignition delay. However, the combustion center (CA50) and combustion duration CA90 reduces significantly with higher AF, from 56 CAD for pure diesel to 24 CAD for the highest AF, due to fast combustion of the premixed NH₃–air mixture and higher heat release through the premixed combustion as shown in Figure 6.

A quadratic model was used to fit the ignition delay data as represented by Equation (6).

$$\begin{array}{cc}\hfill \mathrm{I}\mathrm{g}\mathrm{n}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{d}\mathrm{e}\mathrm{l}\mathrm{a}\mathrm{y}=& 9.619+1.095\times \mathrm{A}+0.046\times \mathrm{B}+-0.008\times \mathrm{C}-0.188\times \mathrm{A}\mathrm{B}+0.091\times \mathrm{A}\mathrm{C}\hfill \\ \hfill & -0.155\times \mathrm{B}\mathrm{C}-0.267\times {\mathrm{A}}^2+0.23\times {\mathrm{B}}^2+0.028\times {\mathrm{C}}^2\hfill \end{array}$$

(6)

The individual parameter effects plot for ignition delay is shown in Supplementary file Figure S2. The ignition delay plotted against AF and speed showed significant variations, while the ignition delay plotted against altitude showed an insignificant effect. Figure S2A demonstrates a significant decrease in ignition delay, and Figure S2B shows that the speed–ignition delay curve was steeper than the AF–ignition delay curve, indicating a greater impact of AF. The altitude–ignition delay plot in Figure S2C implies a slight decrease in ignition delay with an increase in NH₃ percentage.

The start of ignition is described at CA05, the center of combustion at CA50, and the ignition delay time (IDT) is determined from the beginning of injection to CA05. Combustion duration is described as the time from CA05 to CA90, where the combustion duration is the timing elapsed from injection to CA05, at which 90 percent of the total heat release takes place. The effects of AF on CA05, the start of combustion (SOC), CA10, the ignition delay, combustion phasing period, and location of combustion is shown in Figure 6a,b. The CA50 is delayed, and combustion duration further prolonged, with increasing AF, especially at high engine speeds. For example, when the speed is 1400 rpm and at altitude (0), each 10% increase in AF results in a 0.3° CA longer ignition delay period, 0.5° CA later CA50, and 1.5° CA longer combustion duration, respectively. Therefore, the combustion duration was increased from 14.5 CAD for net diesel operation to 19 CAD for the highest AF and engine speed. These results clearly indicate the slower combustion speed of NH₃, given that, the slower the combustion process would be. In summary, the ignition delay depends mostly on engine speed while remaining unaffected by AF and altitude. However, a higher AF shortens CA50 and CA90, reflecting faster premixed NH₃–air combustion and greater heat release.

Figure 7 illustrates a 3D plot of CD50 for diesel and NH₃ at varying AFs under different engine speeds. As NH₃ replaces more diesel, CD50 is increased at high speed. The CD50 increases from 14 CAD in the case of net diesel to 19 CAD for the highest AF at the highest engine speed. Figure 7 and Figure 8 show the impacts on CD90 in more detail on 3D and contour plots. To examine the impact of an individual parameter on CD50 in more detail, Figure S3 illustrates the effects of the single parameter. Table S3 presents the ANOVA data for the CD50 varies examined parameters. The F-value of 1323.99 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA in Table S3 indicates that speed has a greater effect on CD50. The R² statistic close to 1 implies a strong correlation between the observed values and measured result. The engine speed contributed the most to CD50 variability (75.3%), followed by AF (9.3%), with altitude having contributed 1.2%.

Regarding combustion duration (CD90), the impacts of individual parameters on CD90 are illustrated in more detail in Figure 7, Figure 8, Figure 9 and Figure S4 and Table S4. The F-value of 350.28 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA results in Table S4 indicate that AF is the most significant factor. The results also show that engine speed has a slight effect on CD90. The R² statistic being close to 1 implies a strong correlation between the observed values and measured result. The AF contributed the most to CD90 variability (57%), followed by altitude (4.7%), with speed having contributed 1%. Overall, engine speed mainly influences CD50, while AF largely determines CD90, with altitude having minimal effect. This indicates that combustion timing is speed-driven, whereas combustion duration is governed by NH₃ fraction.

A quadratic model was used to fit the 6.CD50 and CD90 data as represented by Equations (8) and (9).

$$\begin{gathered} \mathrm{C}\mathrm{D}50=16.31+1.50\times \mathrm{A}+0.60\times \mathrm{B}+0.22\times \mathrm{C}+0.093\times \mathrm{A}\mathrm{B}+0.041\times \mathrm{A}\mathrm{C}+-0.062\times \mathrm{B}\mathrm{C} \\ +0.254\times {\mathrm{A}}^2+0.105\times {\mathrm{B}}^2+0.058\times {\mathrm{C}}^2 \end{gathered}$$

(7)

$$\begin{gathered} \mathrm{C}\mathrm{D}90=26.9578+0.319131\times \mathrm{A}+2.75481\times \mathrm{B}+1.16322\times \mathrm{C}+0.918334\times \mathrm{A}\mathrm{B} \\ +0.0607792\times \mathrm{A}\mathrm{C}+-0.0679556\times \mathrm{B}\mathrm{C}+1.35463\times {\mathrm{A}}^2+-0.995798\times {\mathrm{B}}^2 \\ +0.536016\times {\mathrm{C}}^2 \end{gathered}$$

(8)

• Engine Emissions

Nitrogen oxides (NO_x) emissions from compression ignition (CI) engines are essential due to their detrimental environmental effects. NH₃, known for being a carbon-neutral fuel, shows promise as a diesel additive. This study examines how blending NH₃ with diesel affects NO_x emissions in CI engines. Figure 10A,B illustrates NO_x emissions at different AF engine speeds and altitudes. The data reveal that increasing the AF results in lower NO_x emissions. The consistent trend across altitudes indicates that NO_x changes are primarily due to differences in fuel NO_x rather than thermal NO_x formation. Additionally, higher engine speeds significantly decrease combustion temperature, leading to a further decrease in thermal NO_x.

Figure S5 and Table S5 illustrate the effects of the individual parameters on NO_x emissions in more detail. Table S5 presents the ANOVA data for the NO_x emissions for various examined parameters. The F-value of 1636.35 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA data in Table S6 indicate that AF and engine speed are significant factors, while altitude is not. The results also show that AF has a greater effect on NO_x emissions, suggesting a reliable NO_x emissions estimation. The R² statistic close to 1 implies a strong correlation between the observed values and measured result. The AF contributed the most to NO_x emissions variability (53.6%), followed by speed (24.9%), with altitude having a negligible influence.

A quadratic model was used to fit the NO_x emissions data as represented by Equation (9).

$$\begin{gathered} {\mathrm{N}\mathrm{O}}_{\mathrm{x}}=450.829-81.360\times \mathrm{A}-135.44\times \mathrm{B}-1.371\times \mathrm{C}+32.554\times \mathrm{A}\mathrm{B}-5.649\times \mathrm{A}\mathrm{C} \\ -0.570\times \mathrm{B}\mathrm{C}-5.97\times {\mathrm{A}}^2+-7.765\times {\mathrm{B}}^2+5.402\times {\mathrm{C}}^2 \end{gathered}$$

(9)

Unburned NH₃ emissions from engines operated with NH₃ blended fuel are a matter of long-term concern for environmental damage and disease risk. Also, NH₃ reacts with water to create corrosive compounds, which makes the substitution rate of NH₃ in combustion critical [35,36]. Ideally, this level must be kept under 10,000 pp. Figure 11 show the unburned NH₃ at various AF and engine speed. If AF rises 10% at 1200 rpm, NH₃ emissions increase by 3197 ppm at sea level, 2814 ppm at 1000 meters and 2387 ppm at 2000 m. Figure S6 and Table S6 provide a detailed illustration of the effects of individual parameters on unburned NH3 emissions.

Figure S6 shows the main factors affecting unburned NH₃ emissions were AF and speed. The unburned NH₃ plotted against AF and speed presented significant variations, while the unburned NH₃ emissions plotted against altitude showed an insignificant effect. Table S6 shows that the ANOVA data for the unburned NH₃ emissions vary according to the examined parameters. The F-value of 303.9 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA data in Table S6 indicate that AF and engine speed are significant factors, while the altitude has slight effects. The results also show that AF has a greater effect on unburned NH₃ emissions, suggesting a reliable method of estimating unburned NH₃ emissions. The R² statistic close to 1 implies a strong correlation between the observed values and measured result. The AF contributed the most to unburned NH₃ emissions variability (82.3%), followed by speed (2.5%), and altitude (0.09%). The unburned NH₃ increases with the increase in NH₃ concentration, which can be explained primarily due to NH₃’s low laminar flame speed, high ignition temperature, and incomplete oxidation under reduced diesel pilot energy fraction, leading to quenching in low-temperature zones. Our findings on the increase of unburned NH₃ at higher ammonia fractions are consistent with Nadimi et al. [32], who also observed elevated unreacted NH₃ due to NH₃’s low flame speed and high ignition temperature. This agreement supports the validity of our experimental results and highlights the fundamental combustion challenges in NH₃–diesel dual-fuel engines. In conclusion, NO_x emissions are primarily reduced by increasing AF and engine speed, with minimal impacts caused by altitude. AF is the dominant factor, confirming its strong influence on emission behavior and providing reliable predictive correlations.

A quadratic model was used to fit the Unburned NH₃ emissions data, as represented by Equation (10).

$$\begin{gathered} \mathrm{U}\mathrm{n}\mathrm{b}\mathrm{u}\mathrm{r}\mathrm{n}\mathrm{e}\mathrm{d} \mathrm{N}\mathrm{H}3=5114.66+785.98\times \mathrm{A}-5100.97\times \mathrm{B}-560.32\times \mathrm{C}+751.15\times \mathrm{A}\mathrm{B}-41.85\times \mathrm{A}\mathrm{C} \\ -715.53\times \mathrm{B}\mathrm{C}-536.31\times {\mathrm{A}}^2+-612.64\times {\mathrm{B}}^2-96.70\times {\mathrm{C}}^2 \end{gathered}$$

(10)

Figure 12 exhibits the emissions of THC at different AF and engine speeds. As one can see from Figure 12, as the AF increased the THC emissions were reduced. Figure S7 shows that the main factor affecting THC emissions was AF. Thus, THC plotted against AF showed significant variations, while the THC emissions plotted against engine speed and altitude showed insignificant effects. Table S7 presents the ANOVA results for the various THC parameters examined. The F-value of 73.4 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.98 value, close to 1, is indicated and should reasonably agree with the adjusted R². The ANOVA results in Table S7 indicate that AF is the significant factor, while the speed and altitude have insignificant effects. The results also show that AF has a greater effect on THC emissions, suggesting reliable THC emissions estimation. The R² value close to 1 implies a good correlation between the observed and measured data. The AF contributed the most to THC emissions variability (61.3%) followed by speed (0.3%), with altitude having an insignificant effect.

A quadratic model was used to fit the THC emissions data as represented by Equation (11).

$$\begin{gathered} \mathrm{T}\mathrm{H}\mathrm{C}=52.14+0.8671\times \mathrm{A}+14.02\times \mathrm{B}-0.2428\times \mathrm{C}+0.3588\times \mathrm{A}\mathrm{B}-0.3780\times \mathrm{A}\mathrm{C}-3.20\times \mathrm{B}\mathrm{C} \\ -10.79\times {\mathrm{A}}^2-2.32\times {\mathrm{B}}^2-1.13\times {\mathrm{C}}^2 \end{gathered}$$

(11)

The most significant advantage of burning NH₃ is its carbon reduction–it is now considered to be a zero-carbon fuel. Figure 13A–C show emissions of CO₂ under various AFs, engine speeds, and altitudes. These data simply show that as the AF increased the CO₂ emissions decreased. That is because diesel is often accompanied by CO₂ that results from burning it, and replacing diesel with NH₃ cuts CO₂ in half. For a 1200 rpm engine speed, an AF increase of 10% decreases CO₂ emissions 5607, 6633, and 7136 ppm at 0, 1000, and 2000 m, respectively. Likewise, at 1600 rpm the same engine runs with a 10% increase in AF that cuts CO₂ emissions by 4185, 4190, and 5987 ppm at the same altitude.

Figure 14A–C show that the main factors affecting CO₂ emissions were AF, speed, and altitude. As shown in Figure 14A, the CO₂ plotted against AF and speed displayed significant reductions in CO₂ at higher speeds and AFs, while the CO₂ emissions plotted against speed and altitude showed an insignificant reduction at high speed and low altitude, as shown in Figure 14B. However, in Figure 14C the highest reduction was found at higher AFs and lower altitudes.

Figure S8 illustrates the effects of individual parameters on CO₂ emissions in more detail. Table S8 presents the ANOVA results for the CO₂ emissions in relation to the examined parameters. The F-value of 288.48 suggests that the model fits well, with a p-value below 0.05. A high R² = 0.99 value, close to 1, is desirable and should reasonably agree with the adjusted R². The ANOVA results in Table S8 indicate that AF and engine speed are significant factors, while altitude is less significant. The results also show that AF has a greater effect on CO₂ emissions, suggesting a reliable method of estimating CO₂ emissions. The R² statistic close to 1 implies a strong correlation between the observed values and measured result. The AF contributed the most to CO₂ emissions variability (36%), followed by speed (30%), and altitude (14%).

A quadratic model was used to fit the THC emissions data as represented by Equation (12).

$$\begin{gathered} \mathrm{T}\mathrm{H}\mathrm{C}=54494.58-6484.50\times \mathrm{A}-8019.69\times \mathrm{B}+5248.12\times \mathrm{C}+1531.24\times \mathrm{A}\mathrm{B}-1487.55\times \mathrm{A}\mathrm{C} \\ -1199.86\times \mathrm{B}\mathrm{C}+3857.44\times {\mathrm{A}}^2-26.10\times {\mathrm{B}}^2+1506.24\times {\mathrm{C}}^2 \end{gathered}$$

(12)

• ⮚ Multi-objective

The parameters for all responses were adjusted via a desirability function to optimize Brake Thermal Efficiency (BTE), ignition delay, center of combustion (CA50), burning duration, NO_x, unburned NH₃, total hydrocarbons (THC), and carbon dioxide (CO₂). The desirability analysis was performed using a “larger-the-better” procedure; the value of desirability that was the largest was the optimal condition [37]. A total of 10 desirable solutions were found, and the solutions with desirability values closest to 1 are the best 10 solutions, as listed in

Table 3. Best 10 Solutions of diesel/ ammonia blended fuel varies examined parameters.

Number	Speed	NH3 Fraction	Altitude	BTE	Ignition Delay	CD50	Burning Duration	NOx	NH3	CO2	THC	Desirability
1.0	1200.0	21.1	0.0	37.4	8.6	15.2	28.0	457.6	6386.7	56,189.7	45.7	0.6	Selected
2.0	1200.0	20.9	0.3	37.4	8.6	15.2	28.0	459.8	6315.8	56,296.9	45.5	0.6
3.0	1200.0	21.2	10.4	37.4	8.6	15.2	28.0	456.7	6407.4	56,182.6	45.8	0.6
4.0	1200.0	21.2	16.1	37.4	8.6	15.2	28.0	457.0	6393.7	56,216.3	45.8	0.6
5.0	1200.0	21.3	17.5	37.4	8.6	15.2	28.1	455.7	6434.2	56,158.6	45.9	0.6
6.0	1200.0	21.0	18.0	37.4	8.6	15.2	28.0	458.8	6334.1	56,311.1	45.6	0.6
7.0	1200.0	20.8	0.1	37.5	8.6	15.2	28.0	461.6	6260.8	56,379.6	45.4	0.6
8.0	1200.0	21.5	0.0	37.4	8.6	15.2	28.1	453.1	6530.8	55,973.6	46.1	0.6
9.0	1200.0	21.7	0.2	37.4	8.6	15.2	28.1	451.4	6587.3	55,889.6	46.3	0.6
10.0	1200.4	21.2	0.3	37.4	8.6	15.2	28.0	456.2	6432.7	56,104.3	45.9	0.6

.

The best solution was an AF of 20% at 1200 speed and 0 altitude. Solution 1, which had a desirability of 0.614, was chosen (Table 3). This solution predicted the best combined BTE, shorter ignition delay, center of combustion (CA50), and burning duration, in additional to lower NO_x, unburned NH₃, THC, and CO₂ emissions.

Figure 15A–I shows contour plots of BTE, ignition delay, center of combustion (CA50), burning duration, and additional NO_x, unburned NH₃, THC, and CO₂ emissions under various NH₃ fractions and speed. Figure 15A shows that the desirability value is very predictive of the parameter values for multi-objective optimization. Figure 15B shows that the BTE decreased with increases in both AF and engine speed. Figure 15C–E show that ignition delays, center of combustion (CA50), and burning duration lengthened with increases in both AF and engine speed. Regarding engine emissions, Figure 15H shows the same trend; as AF and engine speed increased, CO₂ decreased. Figure 15G,I show that both unburned NH₃ and THC increased as the AF and engine speed increased. Overall, statistical analysis shows that AF is a major factor for all of the response outputs. Speed matters less than weight. Figure 16A–I illustrates the effects of an AF parameter on the optimal solution. As you can see in Figure 16A, the desirability value is a good predictor of parameter values in multi-objective optimization. Figure 16B: BTE declines with increasing AF. Figure 16C–E shows ignition delay, CA50s, and burn times that lengthen as the AF increases. Figure 16H shows that CO₂ emissions go down with increasing AF. Figure 16G,I, on the other hand, indicates that unburned NH₃ and THC emissions go up with increasing AFs and engine speeds. In summary, CO₂ emissions decrease markedly with higher AFs, confirming NH₃’s advantage as a carbon-free fuel. AF is the dominant factor, followed by speed and, to a lesser extent, altitude, indicating a reliable method of predicting CO₂ reductions.

The experiments were conducted on a single engine platform and may not fully represent other engine types. The altitude simulation was performed in controlled laboratory conditions, which may not completely reflect real-world dynamic altitude effects. The study focused on low-load operation, limiting generalization to other load conditions.

5. Conclusions

Response Surface Methodology (RSM) was used to optimize the performance, combustion, and emissions characteristics of a four-cylinder turbocharged diesel engine. These characteristics were evaluated against different ammonia fractions (AFs), along with variations in engine speed and operating altitude. The findings can be summarized as follows:

• The results suggest that increasing the AF led to decreases in BTE because NH₃ is less efficient to burn. Higher AF levels result in lower BTEs. At a fixed AF, BTE decreased significantly with increased engine speed. The ANOVA results indicate that AF and speed have significant impacts on BTE, while the altitude does not. AF contributed the most to BTE variability (74%), followed by speed (17%), with altitude having an insignificant effect, which suggests that NH₃ combustion is effective in BTE reduction even at high altitudes. This indicates that higher altitudes are suitable for NH₃ combustion.
• As NH₃ replaces more diesel, the ignition delay extends. This delay increased from 8.2 CAD with net diesel to 8.6 CAD at maximum AF in conditions of steady engine speed and altitude. The engine speed contributed the most to ignition delay variability (81%), followed by AF (17%), with altitude having an insignificant effect. The altitude–ignition delay data imply a slight decrease in ignition delay with an increase in NH₃ percentage.
• The CD50 increased when AF increased, and engine speed contributed the most to CD50 variability (75.3%), followed by AF (9.3%), with altitude contributing 1.2%. However, the results show that engine speed has a slight effect on CD90. The AF contributed the most to CD90 variability (57%), followed by altitude (4.7%), with speed having contributed 1%.
• The results reveal that increasing the AF results in lower NO_x emissions. The consistent trend across altitudes indicates that NO_x changes are primarily due to differences in fuel NO_x rather than thermal NO_x formation. AF contributed the most to NO_x emissions variability (53.6%), followed by speed (24.9%), with altitude having a negligible influence. However, one must take care of the unburned NH₃ emissions in order to comply with regulations.
• On the other hand, the results reveal that AF and engine speed are significant factors leading to increased unburned NH₃, while the altitude has a slight effect. The AF contributed the most to unburned NH₃ emissions variability (82.3%), followed by speed (2.5%), and altitude (0.09%).
• As the AF increased, THC emissions reduced. The results show that AF has a greater effect on THC emissions, suggesting a reliable method of THC emissions estimation. The ANOVA results reveal that the AF contributed the most to THC emissions variability (61.3%) followed by speed (0.3%), with altitude having an insignificant effect. The data simply show that as the AF increased the CO₂ emissions decreased. The AF contributed the most to CO₂ emissions variability (36%), followed by speed (30%), and altitude (14%).

These studies illustrate how NH₃-diesel dual-fuel systems have a promising future in meeting international maritime emission reduction goals, laying a foundation for sustainable fuels that would support global environmental goals. The future research must be on refining ignition and understanding the long-term effects of marine engines using NH₃ to take full advantage of the low-carbon options.