Introduction

Globally, the demand for essential fuel commodities has been steadily increasing due to their critical role in transportation, food production, agriculture, and other sectors1,2. While electric vehicles are gaining traction, the demand for internal combustion engine vehicles remains strong due to their durability and reliability3,4. However, rising fuel prices and environmental impacts, including harmful emissions, present significant challenges to both economies and ecosystems. As a result, recent initiatives have prioritized the development of energy solutions rooted in renewable sources5,6.

Initially, edible oils were considered substitutes for diesel. However, concerns about food scarcity led researchers to shift toward second-generation renewable fuels using non-edible oils including mahua seed, cottonseed, neem seed, and others. Many of these seed oils are highly viscous due to their fatty acid content7,8. To reduce this viscosity, esterification is employed, although this process increases production costs. Some researchers have investigated low-viscosity alternatives like leaf and resin-based oils, but issues such as low yield and high cost remain challenges9. Biofuel production not only offers a pathway to reduce tailpipe emissions and fossil fuel dependency but also supports agricultural development and waste reduction through biomass-to-energy strategies. Recently, biomass-based renewable fuels have attracted significant attention for addressing environmental concerns like landfill waste and vehicular emissions10,11.

In India, poultry meat consumption reached approximately 5000 metric tons in 2024, representing a 60% increase compared to 2010. It is estimated that 25% of chicken meat is wasted during production, contributing to growing environmental concerns12. To address this issue, some researchers have explored the use of waste chicken meat for bio-oil production. Transesterification is employed to extract usable chicken oil from waste materials, which can then serve as an alternative fuel for IC engines13,14. Converting domestic chicken biowaste into biofuel represents a promising option for alternative fuel generation and landfill waste reduction15. Transesterification remains the preferred method for bio-oil extraction, with the yield depending on parameters such as reaction time, mixing speed, catalyst type, and alcohol-to-feedstock ratio16,17. Researchers report that optimal biodiesel yield is achieved using a 1% catalyst, a 6:1 methanol-to-feedstock ratio, and a 60-min reaction time18. This duration has been found more effective than other reaction times in maximizing output19,20.

Recent advancements have increasingly prioritized the environmental sustainability of biofuel production via transesterification. Green chemistry has emerged as a sustainable strategy to minimize environmental impact, particularly by reducing waste generation during the production process21,22. The Environmental Factor (E-factor) serves as an important metric in green chemistry, indicating the amount of waste generated relative to the desired product. A lower E-factor reflects an eco-friendlier process. Achieving higher yields while reducing waste is a major objective in biodiesel production23. Biodiesel has been shown to support smoother engine operation with lower emissions, due to its elevated oxygen concentration, which facilitates more efficient combustion24. However, pure biodiesel (100% blend) often results in reduced efficiency and elevated CO and smoke emissions due to lower energy content, higher ignition delay, and incomplete combustion25.

To analyze the effects of biodiesel, various blend ratios were tested in IC engines. Among them, a 20% biodiesel blend exhibited substantial improvements26. A detailed study using different biodiesel fractions in IC engine found that all blends produced lower BTE than diesel, likely due to different combustion behaviors. Nevertheless, biodiesel blends significantly reduced hydrocarbon and smoke emissions. Enhanced combustion arises from biodiesel’s elevated oxygen levels, promoting oxidation and aiding the decomposition of soot particles27. While biodiesel blends effectively reduce HC, CO, and smoke emissions, they are less effective in controlling NOx and CO2 emissions. Elevated NOₓ emissions arise from increased peak combustion temperatures driven by the high oxygen concentration in biodiesel28,29. To mitigate this, various fuel modification techniques have been explored, including water emulsions, fuel additives, and nanofluids. Nanofluids, in particular, show promise due to their elevated surface area relative to volume (HSAVR), which enhances combustion efficiency and reduces multiple emission components27.

Water-emulsified fuels, another modification technique, can effectively lower NOx and smoke emissions. Water helps reduce peak combustion temperatures due to its low heating value, and micro-explosions during combustion promote secondary atomization, enhancing fuel-air mixing. However, challenges include potential engine corrosion, reduced fuel energy content, fuel stability issues, and the need for modified injection systems30. A prior investigation assessed the influence of a cobalt chromite-formulated nanofluid on the performance of a diesel engine operating under three nanoparticle dosage levels: 50 ppm, 100 ppm, and 150 ppm. The 100-ppm blend achieved approximately 10% NOx reduction compared to the biodiesel-only blend. In contrast, the 150-ppm blend was less effective, likely due to nanoparticle aggregation31. Nanofluid blends have consistently shown enhanced combustion performance and reduced emission levels in contrast to diesel and biodiesel formulations. Contributing factors include higher energy content, shorter ignition delay, improved cetane index, and catalytic effects that enhance oxidation reactions during combustion. The HSAVR also improves heat and mass transfer, supporting better combustion conditions32,33.

Although biodiesel-based nanofluids share similar thermo-physical properties with diesel, they are not identical due to the biological nature of biodiesel. Evaluating the engine impact of these nanofluids is complex and costly, involving extensive titration and testing. To address this, researchers have introduced ANN tools to optimize fuel blends and engine performance more efficiently34,35. ANN models, inspired by human brain function, have become extensively adopted owing to their accuracy and adaptability across various scientific and engineering domains. These models simulate brain-like information processing and have become valuable tools for prediction and optimization in biodiesel research36. An ANN framework was constructed and trained to predict engine output behavior. Training was based on a comprehensive dataset obtained from experimental trials. Key performance indicators included BTE, BSEC, and emissions such as NOx, CO, HC, CO2, and smoke. The model’s predictions closely matched experimental results, indicating high reliability and accuracy. These findings demonstrate the utility of ANN models in guiding the development of optimized fuel blends and improving engine performance. This predictive capability is particularly valuable amid growing regulatory demands and the global push for sustainable, low-emission energy solutions37,38.

Despite significant progress in biofuel research, most existing studies tend to evaluate biodiesel blending, nanofluid use, and engine modeling separately, with little focus on their combined impact on engine performance and emissions using locally sourced biomass. Additionally, there is a lack of comprehensive studies applying ANN to optimize and predict the behavior of biodiesel-nanofluid blends made from domestic biowaste such as chicken waste. This investigation aims to bridge that gap by examining the combined effect of ANN-based optimization, nanotechnology, and biowaste-derived biodiesel on the performance and emission profile of IC engines. The hypothesis guiding this work is that biodiesel-nanofluid blends derived from chicken waste, when optimized using ANN, can enhance combustion efficiency and reduce emissions more effectively than conventional biodiesel or diesel. By integrating waste utilization, emission reduction techniques, and predictive modeling, this study proposes a practical and sustainable approach to developing cleaner fuels for IC engines, contributing to ongoing efforts in renewable energy and environmental protection. In the investigation. The entire procedural framework was outlined in the Fig. 1.

Fig. 1
figure 1

Schematic diagram of overall work flow.

Full size image

Materials and methodology

In this work, chicken oil biodiesel was synthesized through acid separation subsequently chemical esterification. In the first step, acid treatment was employed to reduce the acid value of chicken oil derived from poultry waste. This was followed by a transesterification step to produce chicken oil biodiesel.

Acid treatment

The feedstock, consisting of waste chicken collected from a local market, underwent acid treatment. Initially, feedstock was heated at 100 °C for up to 1 h to convert the solid material into a liquid. The acid treatment was conducted using a three-neck flask, as illustrated in Fig. 2. The extracted chicken oil was poured into the flask, which was placed on a heating element equipped with a thermometer. A mechanical stirrer was attached to the setup to ensure continuous agitation. The heating temperature was maintained at 100 °C, and the stirrer operated consistently at 600 rpm. For every gram of chicken oil, 50 mg of H2SO4 and 2 g of methanol were added. The reaction continued until the acid value was minimized to the threshold level12. Following the acid treatment phase, the reaction contents were poured into a separation vessel and allowed to settle. After a period of phase separation, three distinct layers were observed: The uppermost phase comprised residual methanol, followed by a middle stratum of chicken-based fatty acid methyl esters, while the aqueous component settled at the bottom. The middle phase was collected and used for subsequent transesterification processes.

Fig. 2
figure 2

Schematic setup for acid treatment.

Full size image

Transesterification process

The alcohol and catalyst used in this process were procured from Subra Scientific Company, Puducherry. The transesterification was carried out in intermittently operated reactor system integrated with a mechanical agitation unit, as shown in Fig. 3. Methanol was as the alcohol and KOH as the catalyst for the reaction. A measured volume of methanol and KOH was initially blended in a conical vessel and subsequently introduced into the processing unit, which had been preloaded with the treated chicken oil. The yield of chicken oil biodiesel depended on four primary control variables: the catalyst-to-chicken-oil mass ratio, the methanol-to-chicken-oil molar ratio, mixing intensity, and reaction time. After the transesterification process, the biodiesel was filtered from the by-products using a separating vessel. In the funnel, the top layer consisted of raw chicken oil biodiesel, while the bottom layer contained waste glycerol, which gradually settled. The raw chicken oil biodiesel was then washed three times with purified water to remove any unreacted oil, alcohol, and residual catalyst. Finally, it was heated in an open pan to eliminate any remaining moisture. After heating and purification, the final chicken oil biodiesel was obtained39. The biodiesel blend was enriched with zinc sulfide (ZnS) nanoparticles at varying mass concentrations of 50, 100, and 150 ppm using ultrasonication, and its thermochemical properties were analyzed (Table 1). ZnS was selected for its distinctive properties, including high thermal conductivity, visible-light-driven catalytic activity, and superior surface reactivity. Unlike conventional metal oxide nanoparticles, ZnS exhibits efficient photocatalytic behavior under visible light, enhancing its practicality for combustion applications. Additionally, ZnS demonstrates lower agglomeration tendencies and improved dispersion stability in liquid fuels, critical for ensuring uniform combustion and optimizing engine performance40.

The selected concentration range (50–150 ppm) was determined based on a balance between enhancing combustion characteristics and maintaining nanoparticle stability. The lower limit of 50 ppm was chosen to ensure a measurable impact on engine performance without affecting fuel stability or increasing preparation cost. The upper limit of 150 ppm was based on preliminary studies and literature indicating that concentrations beyond this threshold often result in nanoparticle agglomeration, reduced dispersion uniformity, and marginal improvement in engine parameters36,38. This range allows for evaluating both the onset of nanoparticle influence and the saturation point beyond which additional concentration may lead to diminishing returns or adverse effects on combustion efficiency.

Fig. 3
figure 3

Schematic setup for transesterification process.

Full size image
Table 1 Comparison of thermo-chemical property.
Full size table

E-factor (environmental factor)

This study also focused on evaluating the Environmental Factor (E-factor) associated with chicken oil extraction. Assessing the E-factor is a green chemistry initiative aimed at identifying and minimizing waste generation during biodiesel production. The E-factor serves as a green metric used to quantify the sustainability during production23. In this work, the E-factor for biodiesel synthesis was calculated using Sheldon’s expression.

$${\text{E -factor}}\,=\,{\text{Total Waste }}\left( {{\text{kg}}} \right){\text{ }}/{\text{ Product }}\left( {{\text{kg}}} \right)$$

In this context, waste refers to by-products generated during oil production, such as glycerol and unreacted alcohol, excluding water. The quantity of water is not considered in the E-factor assessment. An ideal Environmental Factor (E-factor) is zero, indicating a completely waste-free process. Higher E-factor values signify that the oil production method is less environmentally friendly and not suitable for green practices.

Characterisation of nano powder

In this investigation, zinc sulfide (ZnS) nanopowder was used to prepare the nanofluids. The nanopowder was procured from Subra Scientific Company, Puducherry. Detailed characterization of the ZnS nanopowder was carried out using SEM, XRD, and FT-IR to assess its suitability for application.

XRD for ZnS

An X-ray diffractometer was employed to perform the XRD analysis, which is commonly employed to govern the crystalline structure and phase of a sample. In this investigation, ZnS nanopowder was analyzed using XRD. Copper as the target material to examine the ZnS sample. To confirm and validate the sample’s structure, the resulting diffraction pattern was obtained and is presented in Fig. 4. The crystallite size of the ZnS nano-additive was calculated to be 59.4 nm using the Scherrer equation, based on the XRD analysis. For more detailed information, please refer to our previous work40.

Fig. 4
figure 4

XRD analysis for ZnS.

Full size image

SEM for ZnS

Scanning Electron Microscopy was employed to examine the morphology of the ZnS nanopowder. The analysis was conducted at an accelerating voltage of 8000 kV, with magnifications ranging from 20,000× to 150,000×. The resulting image is presented in Fig. 5. For more detailed information, please refer to our previous work40.

Fig. 5
figure 5

SEM analysis for ZnS40.

Full size image

FT-IR for ZnS

The FTIR absorption spectrum of ZnS was recorded post-synthesis at 80 °C, as shown in Fig. 6. Notably, the broad peaks detected in the 3399–3454 cm−1 range are attributed to the presence of hydroxyl (OH) groups, which are commonly found in ZnS nanoparticles. These peaks arise due to the strong binding energy of the OH groups within the ZnS matrix, resulting in prominent transmission bands in the 3400–3465 cm−1 region.

Fig. 6
figure 6

FT-IR analysis for ZnS40.

Full size image

The vibrations of zinc and sulfide bonds are closely associated with the peaks detected at 609 and 659 cm−1, which are characteristic of the cubic ZnS structure. Additionally, the occurrence of a symmetric carboxyl group related to sodium is indicated by the peak at 1548 cm−1. For further details, please refer to our previous work40.

Test fuel preparation

The synthesized chicken oil biodiesel was utilized in engine, both as a pure fuel and in blends with diesel and a nano-additive. The nano-additive was mixed with the biodiesel-diesel blend to prepare a nanofluid. This nanofluid was formulated using an ultrasonic generator, as shown in Fig. 7. In this study, a total of five test fuels were prepared. The comprehensive composition of the test fuel is presented in Table 2.

Table 2 Test fuel composition.
Full size table
Fig. 7
figure 7

Work flow of ultra sonication method.

Full size image

Experimental test ring

To rigorously evaluate the performance of the synthesized fuel variants, a water-cooled, single-cylinder engine operating steadily at 1500 rpm was employed throughout the testing process. The engine featured a bore of 86.6 mm, a length of 112 mm, and a CR of 17:1 with delivered an output of 5.2 kW. An eddy-current dynamometer was used to apply the load, while fuel flow was measured using a burette over a specified time interval. For emission analysis, a Krypton 290 was employed to quantity CO, CO2, HC, and NOx levels, and an AVL smoke meter was employed to determine smoke. Prior to testing, the engine oil and lubrication levels were checked to ensure proper functioning. Initial trials were conducted using standard diesel to establish a steady-state condition. After stabilization, the fuel was drained, and the engine was operated using the prepared test fuels41.

AI learning model

The ANN was designed to emulate intricate nonlinear interactions between fuel characteristics and engine output parameters. The input was selected based on their impact on engine performance and emissions, including biodiesel-nanofluid blend ratio, engine load, brake mean effective pressure, and injection timing. Prior to training, all input and output data were normalized using min-max scaling to a range between 0 and 1 to improve model convergence and ensure uniform feature contribution. The dataset was arbitrarily partitioned into training (80%), and testing (20%) subsets42,43.

Multiple learning algorithms, such as SCG, LM, RP, and BFGS, were evaluated for their training efficiency and prediction accuracy. These optimizers were chosen for their proven ability to minimize RMSE and adapt to varying network complexities. In contrast, algorithms such as Support Vector Machines and Decision Trees were found less suitable due to their higher computational cost and limited performance on nonlinear datasets44,45.

Manual hyperparameter tuning was performed to identify an optimal trade-off between predictive accuracy and computational efficiency. Specifically, hidden neuron counts of 5, 10, and 15 across epoch settings of 250, 500, and 750 were used. Results consistently demonstrated superior performance metrics higher regression coefficients, and lower RMSE and MAPE values for the configuration with 10 neurons and 500 epochs. Hence, the 10 number of hidden neurons and 500 training epochs were chosen to develop the learning models. Early stopping criteria were applied based on regression coefficient (R2) thresholds (> 0.99), RMSE (< 0.001), and Mean Absolute Percentage Error (MAPE < 1%). Otherwise, the training was limited to 500 epochs46,47.

Model performance was comprehensively assessed using standard evaluation metrics, includes R2, RMSE, and MAPE. These metrics provided robust insight into model accuracy and reliability across both performance parameters and emission levels as illustrated in Fig. 8. This structured ANN approach ensured high-fidelity prediction and validation of complex engine behavior under varying fuel compositions.

Fig. 8
figure 8

ANN configuration.

Full size image

Cross validation (K-fold)

In conventional regression analysis, model validation typically relies on an 80:20 holdout method, where 80% of the data is used for training and the residual 20% for testing. Model accuracy is evaluated using statistical metrics such as R² and error values, with higher R2 and lower errors reflecting better predictive performance. However, this single partitioning approach may introduce bias, particularly when the data distribution is uneven.

To improve reliability, five-fold cross-validation was adopted. This technique partitions the dataset into five equally sized folds and performs five successive training-testing cycles, with each fold serving once as the testing set while the residual folds are used for model training. Such an iterative approach enhances generalizability, reduces susceptibility to overfitting, and ensures more balanced utilization of the dataset especially crucial when working with limited data samples. The averaged R² across the five folds provides a more stable and unbiased assessment of model performance. A schematic representation of this methodology is illustrated in Fig. 9.

Fig. 9
figure 9

The basic concept of cross validation.

Full size image

Ambiguity examination

Uncertainty analysis played a significant role in ensuring the accuracy of the experimental results. The focus was specifically on evaluating instrumental errors, which are a key component of overall uncertainty. These errors primarily stem from factors such as instrument condition, environmental influences, methods of usage, test procedures, calibration practices, and data acquisition errors48. Conducting a thorough uncertainty analysis helps minimize fluctuations and enhance the reliability of experimental outcomes. The uncertainty values associated with various instruments and measured parameters are presented in Table 3. The total ambiguity for each performance and emission parameter was quantified through the following equation:

Table 3 Uncertainty of different measuring parameters.
Full size table

In this study, fuel blend composition and brake power were chosen as the primary input features for the ANN model due to their direct and dominant influence on engine performance and emission behavior. These parameters are consistently controllable and experimentally repeatable, making them ideal predictors for modeling real-time engine responses. Other potentially relevant variables, including injection timing, ambient temperature, and EGR, were deliberately excluded to maintain a focused scope and reduce model complexity. These parameters were held constant throughout the experiments under controlled laboratory conditions, thus having negligible impact on the variability of the output response within the tested range. Moreover, including additional inputs without significant influence could lead to overfitting and reduce the generalizability of the ANN model23,34.

Result and discussion

E-factor assessment

Figure 10 presents the Environmental Factor (E-factor) results, calculated using Sheldon’s widely accepted expression for waste quantification in chemical processes. Among the 81 experimental trials conducted, a focused subset of 27 trials with a fixed reaction time of 60 min was selected for detailed analysis, in alignment with prior literature validating this duration for effective transesterification23. The minimum E-factor value observed was 0.5, which occurred when chicken oil biodiesel was synthesized using a 9:1 MOR, 1% catalyst, a mixing speed of 500 RPM, and a 60-min period. This condition produced the least amount of waste and is considered the most environmentally sustainable setup within our experimental framework. Approximately 80% of the tested conditions exhibited E-factor values between 0.4 and 0.8, while 40% of these clustered around the 0.6 mark. Conversely, 20% of the trials exceeded an E-factor of 1.0, with the highest value reaching 1.7, recorded at a lower alcohol molar ratio (6:1) and a higher mixing speed (700 RPM). These results confirm that excessive reactants or improper agitation can significantly increase waste generation, thus compromising environmental sustainability.

Fig. 10
figure 10

E-factor assessment for biodiesel production.

Full size image

To contextualize these findings, traditional biodiesel production methods from vegetable oils typically report E-factors ranging from 1.1 to 3.0 depending on feedstock quality and reaction efficiency49. In contrast, petroleum-based chemical processes often exhibit even higher E-factors, with many industrial syntheses exceeding 25–100, especially in pharmaceutical production. Therefore, the E-factor values obtained in this study (0.5–1.7) compare favourably with both biofuel and conventional chemical benchmarks, indicating a significantly lower environmental burden. Furthermore, nearly 50% of the evaluated experimental conditions fall below an E-factor of 1.0, aligning with green chemistry principles that emphasize waste minimization and efficient resource utilization. This supports the environmental merit of using chicken oil, a waste-derived, non-edible feedstock, as a renewable source for biodiesel production. The low E-factor values demonstrate that the process is not only technically feasible but also environmentally responsible, particularly when optimized for reactant ratios and mechanical agitation.

Experimental results

Performance response

BTE vs. BSEC

Figure 11 illustrates the relationship between BTE and BSEC under varying load conditions for five test fuels: Diesel, C100, C20D80 + 50 ppm, C20D80 + 100 ppm, and C20D80 + 150 ppm. BTE represents the ability of a fuel to convert its chemical energy into mechanical work, while BSEC reflects the fuel’s energy efficiency. Generally, the trend observed in BSEC closely mirrors that of BTE, fuels demonstrating higher mechanical energy conversion also exhibit lower BSEC due to their greater energy content. At low and medium load conditions, all test fuels displayed reduced BTE and increased BSEC compared to full-load. This can be ascribed to less favorable combustion dynamics at lower loads. At higher loads, combustion became more effective for all fuels, driven by factors such as shorter ignition delay, elevated peak temperatures, and longer combustion durations.

Among the fuels tested, diesel consistently unveiled the uppermost BTE and lowest BSEC across all loads, primarily due to its high CV and favorable combustion characteristics. The C100 blend, composed solely of chicken oil biodiesel, showed the poorest performance among all test fuels. It exhibited approximately 20% lower BTE and 25% higher BSEC than other test blends. This decline in performance is chiefly due to the lower CV of C100 and the absence of diesel, which typically enhances ignition quality and atomization. The lower cetane number and higher viscosity of C100 also contributed to poor combustion efficiency, increased fuel consumption, and reduced overall performance.

Fig. 11
figure 11

BTE and BSEC vs. BP.

Full size image

In contrast, the C20D80 + 100 ppm blend demonstrated performance characteristics closest to diesel, particularly under medium and high load conditions. This blend recorded a 3–21% higher BTE and 11–26% lesser BSEC than other chicken oil biodiesel blends. The superior performance is a result of the synergistic effect imparted by the oxygenated biodiesel, the high energy content of diesel, and the enhanced HSAVR provided by the ZnS nanoparticles, which improved combustion efficiency.

However, the C20D80 + 150 ppm blend exhibited a 5.3% lower BTE compared to the C20D80 + 100 ppm blend. This decline in efficiency is likely attributed to the excessive concentration of nano-additives beyond the optimal threshold (100 ppm), which may promote nanoparticle aggregation. Such aggregation adversely affects the fuel’s atomization and air-fuel mixing properties, resulting in incomplete combustion and reduced thermal efficiency. Their research demonstrated that increasing the nano-additive concentration beyond an optimal point led to a reverse trend, reducing efficiency by nearly 6% due to similar aggregation effects. Overall, the C20D80 + 100 ppm blend demonstrated superior performance among the test fuels, achieving results comparable to diesel. Specifically, when compared to diesel, this blend showed only a 1.7% reduction in BTE and a 10.73% increase in BSEC, confirming its potential as a viable alternative fuel with minimal compromise in thermal efficiency.

Emission response

NOx and HC vs. BP

Figure 12a, b illustrate the NOx and HC emissions, respectively, from a CI engine operated with five different test fuels. In general, higher NOx emissions are indicative of more complete combustion, as they are primarily formed due to elevated peak combustion temperatures and the availability of excess oxygen during combustion. In contrast, hydrocarbon (HC) emissions follow an opposite trend, typically increasing with incomplete combustion due to the occurrence of unburned fuel residues. Interestingly, the incorporation of ZnS as a nano-additive contributed to a drop in NOx emissions. This effect is ascribed to the catalytic properties of ZnS, which promote more controlled combustion and potentially lower peak temperatures, thereby mitigating NOx formation. Moreover, the catalytic action of ZnS facilitates enhanced oxidation of HC, further improving the overall emission profile of the fuel blends.

Fig. 12
figure 12

NOx and HC vs. BP.

Full size image

The emission results revealed that the C20D80 + 100ppm blend delivered superior performance in both NOx and HC reductions than other test fuels. Specifically, this blend achieved a 5.7% reduction in NOx emissions relative to conventional diesel. This improvement can be ascribed to the oxygen in the biodiesel and the enhanced combustion characteristics imparted by the ZnS nanopowder. The presence of nano-additives promotes complete combustion by improving fuel atomization and accelerating oxidation reactions. Despite these benefits, the C20D80 + 100ppm blend also exhibited a slightly higher in NOx formation due to intensified oxidation. However, the overall impact remains minimal, as NOx compounds (including NO and NO₂) are generally more harmful than NO alone, and the total NOx emissions were still reduced.

The hydrocarbon (HC) emission analysis further confirmed the effectiveness of the C20D80 + 100ppm blend. Diesel exhibited the highest HC levels among all tested fuels, likely due to its lower oxygen level. In contrast, the C20D80 + 100ppm blend produced approximately 29.23% lower HC emissions than diesel. This reduction is chiefly ascribed to the oxygen-rich and the HSAVR of the ZnS nanopowder, which enhances combustion completeness. Other blends, such as C100 and C20D80 + 50ppm, also showed reduced HC emissions (by approximately 7–12%) compared to diesel. These findings align with prior studies, such as that by Hansen et al., who reported a 7% HC reduction using B100 soya methyl ester, highlighting the benefits of biodiesel’s inherent oxygen for complete oxidation. Additionally, the improved thermal conductivity offered by the nanopowder enhances heat transfer during combustion, leading to more efficient energy conversion. Overall, the C20D80 + 100ppm blend demonstrated the best emission performance, effectively reducing both NOx and HC while maintaining high combustion efficiency, making it a promising candidate for sustainable CI engine operation.

CO and CO2 vs. BP

The formation of CO and CO2 is strongly influenced by the availability of oxygen in the combustion process. Fuels with inherent oxygen content, such as biodiesel and its blends, are therefore considered more favorable for achieving cleaner combustion and ensuring future sustainability41. The Fig. 13a, b present the CO and CO2 characteristics of all test fuels, as outlined in Table 2. As engine load increased, CO emissions decreased at low to medium loads across all fuel types, indicating enhanced combustion efficiency due to higher in-cylinder temperatures and better air-fuel mixing. In contrast, CO2 emissions increased progressively with load, reflecting more complete combustion at higher load conditions. These trends reaffirm that oxygenated fuels promote better oxidation of carbon species, thereby reducing incomplete combustion products like CO and enhancing the generation of CO2 through more efficient fuel oxidation. This underscores the environmental advantage of using biodiesel blends enriched with nano-additives, which facilitate improved combustion and emission characteristics in CI engines.

Fig. 13
figure 13

CO2 and CO vs. BP.

Full size image

Except for the C100 blend, all other oxygen-enriched fuel blends exhibited higher CO₂, which can be ascribed to more complete combustion and increased oxidation rates. Interestingly, the C100 blend produced approximately 2% lower CO2 emissions than diesel, primarily due to incomplete oxidation and longer ignition delay. However, this benefit was offset by its elevated CO emissions, a direct consequence of incomplete combustion. While CO2 is a greenhouse gas, its environmental impact is partially mitigated by photosynthetic carbon fixation in plants, which utilizes CO2 and releases oxygen. In contrast, CO poses more immediate environmental and health risks, contributing to ground-level ozone formation and exhibiting toxic effects on the human cardiovascular and respiratory systems. This underscores the importance of minimizing CO emissions, even in the pursuit of reducing CO2 output.

The higher CO emissions from the C100 blend highlight the limitations of using pure biodiesel, as observed in the verdicts of Krishnamoorthy et al.34, who reported an 11% increase in CO emissions for B100 related to diesel. This was ascribed to longer ignition delay, higher viscosity, and diffusion-dominated combustion, all of which impair the completeness of fuel oxidation. In contrast, the C20D80 + 100ppm blend demonstrated superior emission performance, achieving a 37% reduction in CO and a 6.3% increase in CO2 related to diesel. This improvement is ascribed to the enhanced combustion characteristics imparted by the nano-additives, which promote better atomization, oxygen availability, and thermal conductivity, thereby facilitating more complete combustion. The balanced composition of diesel and biodiesel in the blend, coupled with the catalytic effects of nano-materials, contributes to both emission reductions and maintained engine performance. Overall, the C20D80 + 100ppm blend emerged as the most promising alternative, combining the environmental advantages of biodiesel with the performance reliability of diesel, making it a feasible and sustainable fuel.

Smoke vs. BP

Figure 14 illustrates the soot-based smoke emission results for all the test fuels outlined in Table 2. As the engine load increases, all fuel blends demonstrated rising smoke emissions, consistent with increased fuel injection and combustion under higher thermal loads. The formation of smoke is closely related to the availability of oxygen, reinforcing the suitability of oxygenated fuels for future sustainability due to their potential to reduce particulate matter (PM) emissions. Among the fuels tested, diesel exhibited higher smoke emissions, approximately 8.5–13% greater than the other oxygenated fuel blends, with the exception of C100. This increase is primarily ascribed to partial combustion arising from poor air/fuel entrainment, improper atomization, which limits oxidation reactions during combustion.

Fig. 14
figure 14

Smoke vs. BP.

Full size image

Interestingly, the C100 blend (pure biodiesel) showed the highest smoke, from 3.5 to 14.5% more than other blends. This counterintuitive result is attributed to the high viscosity of C100, which leads to larger fuel droplet formation, impairing atomization and delaying evaporation and ignition. Furthermore, longer ignition delay and a dominant diffusion combustion phase in C100 hinder complete combustion, resulting in increased soot and smoke generation. In contrast, the C20D80 + 100ppm blend emerged as the most effective in minimizing smoke emissions, exhibiting reductions of approximately 3–14% compared to other test fuels. Most notably, it achieved a 14% drop in smoke relative to diesel. This improvement is ascribed to the combined effects of better oxidation, enabled by the biodiesel fraction, and heightened combustion efficiency facilitated by ZnS nano additives. These additives contribute to finer fuel droplet formation, improved air-fuel mixing, and faster oxidation of soot precursors. These discoveries are in line with those of Arulprakasajothi et al.50, who reported 7–15% reductions in smoke emissions using nano additive–enriched biodiesel blends. The study emphasized that complete oxidation, enhanced by both the oxygen-rich biodiesel and catalytic effects of nano powders, played a critical role in reducing soot formation. Beyond emission reduction, the C20D80 + 100ppm blend also offers engine performance benefits. The improved atomization and combustion characteristics reduce particulate matter, contributing to cleaner engine operation, better fuel economy, and potentially extended engine life. The more complete combustion also ensures that a higher proportion of the fuel’s energy content is harnessed efficiently, thereby minimizing energy losses and lowering maintenance needs over the engine’s operational lifespan.

ANN evaluation

ANN holdout validation

In this computational study, ANN models are employed to predict complex and nonlinear behaviours. The primary goal is to assess engine performance and emission characteristics using various fuels: Diesel, C100, C20D80 + 50 ppm, C20D80 + 100 ppm, and C20D80 + 150 ppm. The ANN model, created in google Colab using Python, uses two input variables: the type of fuel blend and brake power. It focuses on analysing seven key output parameters: BTE, BSEC, NOx, CO, HC, CO2, and smoke utilizing data collected from experimental runs.

To validate the engine output response, it was crucial to identify the most suitable learning algorithms. This predictive process involved extensively to identify the most effective learning algorithms, guided by RMSE, MAPE outcomes and R2-value assessments. Generally, configurations with higher R2-values and lower RMSE & MAPE values were considered for effective learning algorithm selection. The computational study determined that the LM learning algorithm, provided the most precise predictions compare with experimental data. This decision was based on achieving the highest R2-value and the lowest RMSE and MAPE, as detailed in Tables 4 and 5.

Table 4 Comparative statistical summary of key output parameters across diverse neural network training algorithms.
Full size table
Table 5 Comparative statistical summary of key output parameters using suitable neural network training algorithm.
Full size table

Figure 15 presents a comparison between the ANN model’s predictions and the experimental results. For parameters such as HC, CO, CO2, NOx, smoke opacity, BTE, and BSEC, the R-values are 0.9329, 0.9197, 0.9989, 0.9961, 0.9523, 0.9961, and 0.9431, respectively. The corresponding RMSE values are 0.0159, 0.1008, 4.858, 0.0629, 3.9601, 0.3414 and 1.1016, reflecting the learning error. Additionally, the MAPE values for HC, CO, CO2, NOx, smoke opacity, BTE, and BSEC are 8.36%, 9.38%, 1.03%, 1.37%, 9.12%, 1.55% and 5.15% respectively. This low error reflects a high degree of agreement between the model’s predictions and the observed experimental data. This study underscores the effectiveness of ANN in streamlining resource expenditure, minimizing analysis duration, and simplifying traditionally complex experimental procedures. Employed as a forecasting framework, the model delivers notable benefits in capturing engine response characteristics and optimizing control mechanisms, thereby yielding insightful outcomes51,52,53.

Fig. 15
figure 15

Comparison between ANN-based predictions and experimental observations.

Full size image

ANN K-fold cross validation

A comprehensive K-fold cross validation analysis was conducted to evaluate the performance of the ANN prediction models. A total of 140 R2 values were generated for each output parameter, with each learning algorithm contributing five top-performing R2 scores across multiple folds. These values were subsequently averaged to determine the mean R2 score for each parameter and learning algorithm, as presented in Table 6. The summarized results indicate that the LM algorithm consistently achieved the highest average R2 values across all output parameters when compared to other learning algorithms. This affirms the robustness and predictive accuracy of the LM-based model, further validated by the cross-validation procedure. The ability of LM to maintain superior performance across fold iterations reinforces its suitability as a reliable learning approach in modeling engine output responses under varying training conditions.

Table 6 K-fold cross validation on average R2 results.
Full size table

Economic feasibility analysis

The economic feasibility of integrating nano-additives, specifically ZnS nanopowder, into biofuel blends was analyzed to assess cost implications and potential benefits. This analysis evaluates the cost of nano-additive procurement, nanofluid preparation (C20D80 + 50 ppm, C20D80 + 100 ppm, and C20D80 + 150 ppm), and subsequent integration into the fuel blends. While the initial cost of ZnS nano-additives is higher compared to conventional diesel production, the long-term benefits offset these expenses. The C20D80 + 100 ppm blend, which achieved a 29.23% reduction in HC emissions, a 37% reduction in CO, and a 5.7% reduction in NOx, offers significant environmental and regulatory advantages. These reductions can potentially lead to cost savings through lower emission taxes, reduced environmental penalties, and eligibility for incentives promoting cleaner fuels. Furthermore, the use of nano-additives is expected to minimize engine wear and maintenance costs due to improved combustion efficiency and reduced particulate formation. These factors contribute to the economic viability of adopting nano-additives in commercial applications. Among the tested blends, the C20D80 + 100 ppm formulation emerged as the most cost-effective, balancing a moderate upfront investment with substantial emission reductions and consistent performance.

Conclusion and SWOT

This study highlights the potential of C20D80 + 100 ppm as an efficient and environmentally sustainable alternative to conventional diesel fuels.

  • The experimental results demonstrated that this blend achieved 3–21% higher BTE and 11–26% lower BSEC compared to other alternative fuels. Furthermore, it recorded 5.7% lower NOx emissions and 29.23% lower HC emissions than diesel, attributed to enhanced combustion facilitated by ZnS. Additionally, the blend reduced smoke emissions by 3–14% and improved the emission profiles of CO2 and CO, promoting environmental sustainability.

  • The ANN based LM model proved highly effective, delivering predictive accuracy with R2-values ranging from 0.9197 to 0.9961 and lowest RMSE & MAPE values. These findings reflects a high degree of agreement between the model’s predictions and the observed experimental data, underscoring the ANN model’s reliability in predicting engine performance and emission metrics.

  • Overall, this study establishes the C20D80 + 100 ppm blend as a viable eco-friendly fuel option, offering improved fuel efficiency, reduced emissions, and a pathway toward sustainable energy solutions. The ANN based LM model further enhances its value by reducing the time, cost, and complexity associated with experimental analysis, demonstrating its utility as a predictive tool for optimizing engine performance and emissions.

Numerical summary

The numerical summary of the present work is detailed below:

Parameter

Diesel

C100

C20D80 + 50ppm

C20D80 + 100ppm

C20D80 + 150ppm

BTE

Reference

− 21%

− 18%

+ 3% to + 21%−

5.3%

BSEC

Reference

− 23%

− 20%

− 11% to − 26%

+ 5%

NOx

Reference

− 23%

− 21%

− 5.7%

+ 2%

HC

Reference

− 12%

− 10%

− 29.23%

− 5%

CO2

Reference

+ 2%

+ 1%

+ 2.5%

+ 1.5%

CO

Reference

+ 2%

+ 1%

− 3%

+ 1.5%

Smoke

Reference

− 14.5%

− 12%

− 3% to − 14%

− 5.5%

This conclusion underscores the blend’s superior performance, significant emission reductions, and high predictive accuracy, affirming its potential as a sustainable alternative.

SWOT analysis of the C20D80 + 100ppm blend

Strengths:

  1. 1.

    High Efficiency: Achieves 3–21% higher BTE compared to other alternative fuels.

  2. 2.

    Lower Energy Consumption: Records 11–26% lower BSEC, indicating improved fuel economy.

  3. 3.

    Emission Reduction: Significantly reduces NOx emissions by 5.7% and HC emissions by 29.23% compared to conventional diesel.

  4. 4.

    Cleaner Combustion: Demonstrates 3–14% lower smoke emissions and improved CO2 and CO emission profiles, supporting environmental sustainability.

  5. 5.

    ANN Predictive Accuracy: Offers high predictive accuracy with R-values ranging from 0.9197 to 0.9961 and low RMSE & MAPE values closely aligning with experimental data.

Weaknesses:

  1. 1.

    Higher Production Costs: Involves increased initial investment due to the use of nano-additives and the need for specialized fuel processing techniques.

  2. 2.

    Aggregation Challenges: Excessive nano-additive concentrations may lead to particle aggregation, adversely affecting fuel atomization and air-fuel mixing efficiency.

  3. 3.

    Limited Availability: Restricted accessibility of specific components, such as ZnS nanopowder, may pose challenges for large-scale production and widespread adoption.

Opportunities:

  1. 1.

    R&D Advancements: Continued research can optimize fuel blend formulations and investigate alternative nano-additives to further improve engine performance and emission control.

  2. 2.

    Policy Support: Engagement with policymakers to establish supportive regulations and incentives can accelerate the adoption of sustainable fuel technologies.

  3. 3.

    Market Expansion: The blend shows promise for broad applicability across sectors such as transportation, agriculture, and power generation.

  4. 4.

    Environmental Benefits: Supports global sustainability efforts by significantly reducing carbon emissions and other harmful pollutants.

Threats:

  1. 1.

    Competition from Other Fuels: Emerging renewable energy sources and advanced biofuels may challenge the market penetration of this blend.

  2. 2.

    Regulatory Changes: Shifts in environmental policies and fuel quality standards could influence approval and commercial viability.

  3. 3.

    Economic Factors: Variability in raw material prices and production costs may impact the economic feasibility and scalability of the blend.

  4. 4.

    Technological Barriers: Integration into existing systems may demand engine modifications and infrastructure upgrades, posing adoption challenges.

This SWOT analysis offers a comprehensive evaluation of the C20D80 + 100ppm blend, emphasizing its potential as a sustainable and efficient alternative fuel. It also acknowledges the associated challenges and identifies opportunities that can facilitate its broader adoption and future development.