Introduction

The use of the previous global energy drive summary indicated that the primary source of energy utilised stems from the internal combustion (IC) engines’ igniting of petroleum-based fuels. According to current estimates, the world’s recoverable stocks of petrol and petroleum are expected to reach 7.289 trillion cubic feet and 1.44 trillion barrels, respectively, based on the global energy scenario1. A total of 261 billion cubic feet of petrol and 86.5 million barrels of oil are consumed daily by people worldwide. Global oil and natural gas reserves were depleted in 40 and 60 years, respectively, due to recent consumption patterns2. Because of their inherent physiochemical and oxidation components, hydrocarbonated and liquidity fuels consume a significant portion of the market. Unfortunately, the capitals of non-renewable energy fuels have declined, thus it may remain empty for a few decades.

Fig. 1
figure 1

Fuel consumption rate by various continents (Data sourced from2).

Full size image

Figure 1 shows that, as of 20192, Asia is the continent that consumes the most oil, hydropower, lignite, and coke on a global scale. In the field of electricity generation, Europe and Eurasia are the biggest users of nuclear power, petrol fuel, and renewable resources. In terms of worldwide coal consumption rate, accounting for three-quarters of global energy consumption, the Asian continent holds the top spot among all users. In every sector, the energy consumption ratios in the Middle East and Africa were negligible, whereas North America’s energy consumption rate is moderate. In the years that followed, the universal oil consumption rate climbed gradually (Fig. 2).

Fig. 2
figure 2

Global energy consumption by various sectors (Data from3, available online: https://ourworldindata.org/energy-productionconsumption).

Full size image

The alarming impacts of global warming are attributed to a number of sources, including hazardous waste produced by companies, toxic goods derived from petroleum products, and some chemical items. This will be the main cause of rising food prices, changing climatic patterns, and threats to the survival of humanity. In the next 25 years, India hopes to achieve the sustainable evolution rate of 8.5–10% in the economic sector, in order to maintain the people’s living standards and further personal growth events. In the future, it will be possible to increase the major energy band supply by 4–5%. Regaining the nation’s future energy demands may depend on striking a balance between the energy crisis and the upcoming investigation of alternative fuel energy sources with low emission levels.

Due to the lack of conservative fossil fuels, the increase in fuel prices, and the multiplied emissions of toxins produced during the oxidation process, biodiesel and biofuel sources have proven to be highly amazing in the current decades. The deposits of petroleum oil and its byproducts are concentrated in a few isolated areas around the country, and they regularly run out. The likelihood of producing biomass and biodiesel from locally grown sources and using them to replace different petroleum-based commodities is the most appealing way to overcome the fuel energy deficit. Biodiesel and biomass sources are advantageous renewable fuels that are generated through natural processes such as aero-biotic or agricultural digestion, as opposed to the geological delivery method used to manufacture raw crude oils like lignite, coke, and crude oil4.

In their study, Kasiraman et al.5 examined the effects of using cashew nut methyl ester, raw cashew nut oil, and various combinations of vegetable oil, oxygenating additives, and alcohols on the performance of the CI engine. The authors concluded from their observations that, of the different oxygenate additives, Die-ethyl ether 30% (DEE) evolved to have a more excellent Brake Thermal Efficiency (BTE) of approximately 29.68% at peak load conditions. Additionally, the authors found that tailpipe emissions, including nitrogen oxide (NOx) and smoke opacity, also showed better results in comparison to other additives. When compared to other alcohol additive mixes, butanol (butanol 30%) has superior fuel economy, BTE, and Heat Release Rate (HRR). Then, on a volume basis, the analysis was expanded to include several vegetable oil blends and cashew nut methyl ester. Among these, blends of camphor oil (CMPRO30) showed the highest levels of thermal efficiency, HRR rate, NOx, and smoke opacity. Using three different Pistacia lentiscus biodiesel ratios with diesel—5%, 30%, and 50% by volume basis—Khiari et al.6 studied the naturally occurring CI engine. According to the author’s observations, Pistacia lentiscus biodiesel has a roughly 3% increase in thermal efficiency when compared to hydrocarbon fuel. It also significantly reduces tailpipe emissions of unburned hydrocarbon fuel, carbon, and particulate matter at 100% load condition—roughly 45%, 25%, and 17%, respectively. Higher levels of BSFC (Brake Specific fuel consumption, which is a measurement of how much fuel an engine consumes in relation to the power it produces) and NOx are also found—about 10% and 4%, respectively.

A diesel engine running on petroleum-diesel mixes and fuelled by Ceiba pentandra and Jatropha curcas biodiesel had its performance and exhaust emissions examined by Dharma et al.7. More engine torque can be achieved using B10’s CI engine performance, which is generally comparable to that of petroleum-Diesel fuel; different mixes were shown to correspond with varied levels of brake power (BP) and brake torque (BTE). When 50% Jatropha curcas and 50% Ceiba pentandra biodiesel are blended with hydrocarbon fuel, the amount of CO and smoke opacity emissions are reduced; however, nitrogen oxide and CO emissions are increased simultaneously. The test engine exhaust emissions and performance metrics were also predicted using ANN models, which produced outstanding results with a coefficient of determination of greater than 98% and a mean absolute percentage error of less than 5% for all parameters. An explicit mathematical representation is typically not necessary for ANN models to produce accurate and trustworthy results. In order to predict the engine exhaust emission and performance parameters of the CI engine operating in dual fuel mode using hydrogen gas and Jatropha biodiesel, Javed et al.8 investigate the application of the ANN model. The ANN tool was developed to estimate engine performance and emissions, such as NO, CO, O2, HC, CO2, and EGT, by adjusting load, hydrogen flow rate, and biodiesel blends. Regarding this, the suggested ANN tool’s gross regression coefficient, mean square error, and mean absolute percentage error are, respectively, 0.9936, 0.0011, and 4.86%.

The integration of ANN with Ant Colony Optimisation (ACO) algorithm was handled by Mohammadhassani et al.9. The relationship between input parameters was illustrated using the feed-forward multilayer perceptron network and the multilayer perceptron (MLP) network. With pertinent correlation factors of 0.96 and 0.98, ANN suggests the appropriate soot and NOx model. The ACO algorithm was developed to determine the ideal temperature for engine air consumption and injection fuel at varying power and speed levels, taking into account engines with significantly reduced emissions of NOx and soot. In comparison to the combination of Genetic Algorithm (GA) and ANN, the convergence speed of the ACO and ANN tools was comparatively more significant. The ANN modelling technique in this investigation primarily estimated tailpipe emissions levels at the lowest possible time and produced a higher anticipated rate with greater accuracy. The author proposes using this modelling technique to lower soot and NOx emissions in the electronic control unit industry.

The ANN tool was used to forecast the CI engine’s output when Ude et al.10 studied the engine’s performance using cottonseed biodiesel. Petroleum-Diesel, biodiesel, and their blends are used in experimental testing in a two-cylinder DI Diesel engine to achieve varying engine speeds. Raw cottonseed oil and mineral diesel test fuel mixes were used to experiment with CI engines, and the results showed better performance levels with low emissions. When comparing raw petroleum-diesel fuel outcomes, such as BP, BTE, BSFC, minimum CO, and HC emissions, B20 blend has produced results that are comparable. The ANN tool, which finds the R-correlation coefficient by 0.9743, 0.9914, and 0.9788, optimised and predicted the experimental data, such as BP, BTE, and BSFC, in the interim. Furthermore, the model illustrates the range of 0.0001 for the mean square error between the simulated and measured output. In conclusion, the engine efficiency of the biodiesel mix derived from cottonseed oil is improved, and the engine parameter is accurately predicted by the ANN tool. The quality of biofuel is expected to significantly improve with the addition of hydrogen flow. Hydrogen gas is needed in many industries, such as the transportation, fertilizer, and petroleum refining sectors. Hydrogen is an environmentally friendly fuel that may be produced using nuclear power, natural gas, biomass, and sustainable energy sources including solar and wind power. Hydrogen is an energy carrier that improves the performance of the primary fuel.

Nano additives play a major role in lowering the exhaust emissions from diesel engine. Several researches were done on nano additives which are metal oxide based additives which can liberated oxygen during combustion and thereby oxidize the unburnt hydrocarbons and carbon monoxide to carbon dioxide and water vapour. Table 1 shows the literatures used with different nano particles and their effect in base fuel in enhancing the characteristics.

Table 1 Overview of a variety of nano additives and their influence on characteristics of engine.
Full size table

According to the literature review mentioned above, diesel engines are thought to produce more harmful pollutants since they use fossil fuels, and there will be a greater need for this kind of energy in the near future. The world’s fuel and food crises are lessened, and pollution of the land, water, and atmosphere is reduced thanks to the innovative lipase immobilisednanocatalystutilised to produce Garciniabiodiesel from inedible Garcinia seeds11. Due to the low percentage of biodiesel produced using standard methods, a lipase immobilisednanocatalyst was employed to increase the total yield of Garcinia biodiesel, which met the requirements to be utilised as a substitute fuel for diesel engines. Next, different test fuel blends made from the Garcinia biodiesel were created, and the fuel’s characteristics regarding pollutant emissions and engine performance were examined using a diesel engine. Next, using the ANN modelling tool, the diesel engine properties of Garciniabiodiesel and its fuel mixtures were anticipated and optimised. Research on the expected and optimised properties of the innovative Garcinia seed biodiesel and its engine characteristics is carried out in the research community using the ANN modelling tool. This study was completed using waste Garcinia seed biodiesel that was trans-esterified using a lipase immobilisednanocatalyst. The biodiesel was then synthesised and characterised using TEM, XRD, and SEM instruments, yielding a yield of over 92% biodiesel. The fuel blends were then prepared using a magnetic stirrer method and analysed using FT-IR and GC-MS instruments. To reduce the need for and demand for fossil fuels, the Garciniabiodiesel and its fuel mixes were tested in a single-cylinder diesel engine. The ANN modelling tool was then used to forecast and optimise the engine’s characteristics.

Materials and methods

Novel nanocatalyst preparation, Garcinia oil and biodiesel production

Novel nanocatalyst preparation and its technique

Nowadays, there are a lot of active applications using nanotechnology and related nanomaterials. As a result, nanostructured additives are used to immobilise the well-known biocatalyst. In this study, the chemical binding of Fe3O4 magnetic nanoparticles, which acted as a core and an enzyme as a shell to create core-shell nanoparticles, successfully completed the immobilisation process. Thermomyceslanuginosus lipase (TL) was positively immobilised in conjunction with the Fe3O4 magnetic nanoparticle for this reaction by means of a co-precipitated approach. This collection of reactions is described as an innovative technique using lipase-immobilized Fe3O4 magnetic nanoparticles.

Fig. 3
figure 3

(a) Immobilized Fe3O4 nanoadditives separated by the external magnet. (b) Synthesized Fe3O4 nanoadditives (c) dried Fe3O4 nanoadditives.

Full size image

The co-precipitation method was first used to create iron oxide nanoparticles. Ammonia solution was added after ferric and ferrous chlorides were mixed at a molar ratio of 2:1, reserved in 0.2 mol, and dissolved in 200 millilitres of distilled water for this synthesis. The pH of 9 was achieved by including this solution, and a mechanical stirrer was used to mix it often. Following a 12-h ageing period, the necessary Iron Oxide Magnetic Nano Particle (IOMNP) material was obtained. To remove any contaminants from the obtained solution, the produced IOMNPs were then separated using the centrifugation technique. Three or four times, it was cleaned using ethanol and acetone. After removing the moisture content, the produced nanoparticles were vacuum-sealed at 60 °C for 11 to 12 h, and the IOMNP were stored in airtight bottles.

Previous research by Subramani et al.50 was cited as the source of the overall process for immobilising lipase enzyme within IOMNP. The carbodiimide activation procedure achieved the covalent connection of Thermomyceslanuginosus (TL) lipase immobilisation with IOMNP, as was previously discussed. In relation to this, IOMNP were permitted to suspend in 0.2 mol of NaCl and 0.005 mol of phosphate buffer at eight pH. After 0.6 ml of 0.04 g/ml carbodiimide solution mixed with phosphate buffer at room temperature, it is combined for 20 min of sonification process. The carbodiimide agent and TL were mixed repeatedly after being left in the stirrer for 40 min at 5 degrees Celsius. With IOMNP attributable to its magnetic nature, the TL was effectively immobilised now. The external magnetic field separated it, and it was then recycled for the subsequent batch operation, which is illustrated in Fig. 3. Finally, immobilised IOMNP was placed in storage to be examined later.

After using the supernatant approach, the Bradford process was used to obtain the lipase enzyme. Removing the excessive immobilised lipase from the net assessed amount of lipase for the immobilisation process allowed for the regular determination of the immobilised lipase quantity. Equation (1) below was used to determine how well immobilised lipase reacted to form magnetic nanoparticles.

$${\text{I}}\,{\text{=}}\,\,\frac{{{\text{(C}}{{\text{O}}_{\text{i}}}\,{\text{-}}\,{\text{C}}{{\text{O}}_{\text{f}}}{\text{)}}{{\text{L}}_{{\text{1}}\,\,\,\,\,}}}}{{{\text{C}}{{\text{O}}_{\text{i}}}\,{{\text{L}}_{\text{2}}}\,}}$$
(1)

Herein, I is the Immobilization efficiency (%), L1 and L2 is the volume of solution in cm3, COi and COf is the initial and final lipase concentration rate in the supernatant.

Garcinia seed oil extraction and its preparation approaches

Garcinia seeds are mostly obtained from the National Bureau of Plant Genetic Resource (NBPGR) located in the Kerala, India, Thrissur region. After the seeds were sorted according to size and shape, they were dried for two to five days to remove any remaining moisture, and then they were roasted for twelve hours at 101ºC. The grinder is then used to grind the available seeds into a fine powder, and the traditional mechanical expeller process is used to extract the raw Garcinia seed oil from the powdered seeds. With 1 kg of Garcinia seed supply, this approach yields a greater level of 40%. The resultant consequence ought to surpass the results of the solvent extraction technique. Yet, due to its higher inherent oil content than other conventional seed sources, Garcinia seed oil can provide higher output levels at the mechanical extraction stage itself (Yatish et al., 2018). Using European and ASTM standards as a basis, the desirable fuel qualities were measured and examined. Figure 4 shows the flow of the garcinia seed oil process. The following Eq. (2) was used to predict the raw Garcinia oil yield:

$$\begin{gathered} {\text{Total}}\,{\text{Weight}}\,{\text{of}}\,{\text{Oil}}\,{\text{Yield}}\,{\text{(\% )}}\,{\text{=}}\,\frac{{{\text{Weight}}\,{\text{of}}\,{\text{Extracted}}\,{\text{Oil}}\,{\text{(}}\,{\text{g}}\,{\text{)}}}}{{{\text{Weight}}\,{\text{of}}\,\,Garcinia\,\,{\text{Seeds}}\,{\text{(}}\,{\text{g}}\,{\text{)}}}}\,{\text{x}}\,{\text{100\% }} \hfill \\ \hfill \\ \end{gathered}$$
(2)

Lipase immobilized with Fe3O4 catalyst transesterification process of Garcinia gummi-gutta seed oil

As was previously mentioned, this process involved the careful selection and use of a unique lipase immobilised transesterification activity to achieve an exceptional yield and improved biodiesel quality. Initial steps of this technique included TL immobilised Fe3O4 nanoparticles (1–6%), raw Garcinia fuel (1 L), and methanol (2–14%). The enhanced reaction temperature (20–70 ºC) and time (12–74 min) were used for this study.

Fig. 4
figure 4

Oil extraction process of raw Garcinia oil from Garcinia seed.

Full size image

Previous study findings12,13,14,15 were used to determine these ideal settings for this lipase-based transesterification procedure. From the Nanoscience Laboratory at the AC Tech Campus of Anna University in Chennai, the lipase-based transesterification reaction was carried out. The laboratory also handled the raw ingredients, glassware, and other equipment. The initial transesterification activity using a 500 ml airtight empty bottle and a lipase immobilised catalyst (Fe3O4 + Thermomyceslanuginosus) was conducted. 250 ml of raw Garcinia oil and 1% of immobilised Fe3O4 were then added to the bottle, which was closed and divided. The mixture was then manually swirled to combine the ingredients. With pellets in the arrangement, it was placed in a magnetic stirrer, and 800–1000 rpm stirring was used to achieve the desired operating condition. A magnetic stirrer is often used to initiate the reaction at 20 ºC for 12 min. Lipase immobilised transesterification starts under ideal working circumstances when the filled airtight bottle is stirred magnetically.

Fatty Acid Methyl Ester of Garcinia Oil (FAME GO) and glycerol are produced when the immobilised lipase enzyme (TL) breaks down all of the triglycerides in the raw Garcinia oil at the conclusion of the stirring process. Additionally, to overcome the excess biodiesel yield, the immobilised lipase enzyme drives the reaction and aids in converting the raw Garcinia oil’s highest ratio of triglycerides.” After stopping the lipase-based reaction, the natural reaction mixtures were added to the separating funnel arrangement, which allowed them to remain in a cool, balanced environment. With a higher yield of production rate, the FAME GO was extracted straight from raw Garcinia seed oil following the completion of the separation process, stimulated by lipase immobilise catalyst. The extractable FAME GO was gathered in the beaker that was left empty, and the glycerol content was separated. Figure 5 illustrates the lipase immobilised transesterification development. Following this, the FAME GO (reaction mixture) was heated to a maximum of 40 ºC in order to remove the biodiesel’s accessible contaminants as well as its greater water and methanol content. Several sets of the batch method of producing garcinia biodiesel were used to continue this same process.

The best conditions for FAME GO conversion were found to be reaction temperature (40 ºC), catalyst immobilisation (5% w/w), reaction duration (62–74 min), and methanol (6%) with the highest average biodiesel percentage of 92.12 yielded as a result of these factors. This evaluation was based on the results of all batch processes. Due to a higher reaction rate and greater biodiesel output, methanol content was added to the reaction mixture three times in this case. FAME GO (biodiesel) was found to be highly miscible with petroleum diesel when the mixture was centrifuged for 12 to 15 min.When biodiesel was combined with raw diesel, various miscibility ratios were tested; no changes were seen after 15–20 days.

Fig. 5
figure 5

Integrated process flow of GarciniaBiodiesel production with lipase immobilized nanocatalyst.

Full size image

With the assistance of the external magnetic field, immobilised lipase was finally removed from the mixture and cleaned with distilled water and alcohol to get rid of any undesired contaminants. For the following batch process, it was then frozen and kept intact. Utilising the biodiesel reactor’s higher capacity for this reaction under optimal conditions allowed for the bulk synthesis of FAME GO (biodiesel). Instruments from FT-IR and GC-MS were used to diagnose the acquired FAME GO.

Assessment of raw Garcinia biodiesel properties compared with petroleum-diesel fuel

Since Garcinia gummi-gutta seed oil has unfavourable fuel qualities and satisfies both ASTM D6751 and EN 14214 standards, it was selected as the innovative alternative feedstock for this experimental study project. Raw Garcinia oil also satisfies ordinary petroleum diesel requirements. The lipase immobilised catalyst achieves the best biodiesel output while also contributing to the reduction of environmental contaminants. This generates an environmentally benign natural energy fuel for human use. The attributes of unrefined GGME blends with petroleum-based diesel are explained in the description that follows. The characteristics of raw Garcinia and Jatropha biodiesel were then contrasted. The lab was used to measure and estimate all pertinent parameters. Table 2 compares the fuel characteristics of Garcinia biodiesel and its blend.

Table 2 Test fuel properties of Garcinia Biodiesel and its blends compared with petroleum-diesel52.
Full size table

ANN modelling and its approach

  1. (a)

    The most advanced ANN technique is part of an experimental data-processing toolkit developed by biological neural networks, and it shares some performance and technical characteristics with them. The main use for this ANN model is to analyse non-linear problems using an experimental and dependable method in order to test, train, and target the forecast of various parameters, such as engine performance and tailpipe emissions of GGME and its blends with petroleum-Diesel. The R-regression coefficient and relativity errors of the input data were obtained by running this ANN model. The GGME Biodiesel engine’s efficiency and tailpipe emissions were then predicted and optimised by the model.

  2. (b)

    ANN modeling and its organization.

The current study used the MATLAB software platform to build an artificial neural network (ANN) model and program to forecast and analyse the BTE, BSEC, HC, CO, CO2, NOx, and smoke opacity of GGME blends mineral diesel of experimental engine output. The learning method known as LM, or Levenberg Marquardt, was also utilised to run this ANN model. ANN network architectures are created using this algorithm between16,19. Parallel to this, it was demonstrated that the LM method was a learning algorithm that performed expertly, built the network fairly, and was referred to as having an excellent network structure in ANN modelling with rapid data processing. To guide simulation and forecast of GGME and its mixtures’ performance and emission, this learning method was put into practice. A target data set was constructed at the time of model execution using 30% of the engine experimental data that was framed for this modelling, with the remaining 70% of the data being assigned for testing and training purposes17,18,20,21. With values of 0 s and 1 s, the task was carried out by the famed logistic sigmoid transfer function, which was coupled in this ANN model. Primarily, input and output dataset normalisation was completed previously.

From 10 to 19, neurones were initiated in the system, which was run with two sets of inputs (Blend and Brake power). ANN’s input, hidden, and output layers—all produced by a group of nodes—are its primary processing steps. Projected to the hidden layer are artificial neurones that process input data obtained from experimental engine tests. When processing took place in the hidden layer, data from ANN neurones that were accessible in the input layer helped to record the pertinent output. The desired output was sent to the output layer once the hidden layer satisfied its goals of screening and prediction. The system calculates the weighted sum values and can be used to instigate the system summation function to fix the net input transformed into the cell.

Fuel mixes were the input parameter, and BP may be identified as input data to the cells. With this information, the generated network can train on the inputs. In addition, one can estimate the input data and processing method with its components by accessing weight functions. Each value of input into this system must be multiplied by its weight value. The processing parts are then connected using a summation function grouping scheme. The net information of the network was ascertained following these assessments. This curvilinear approximation of the relationship between the given input and output data was obtained by computing the system activation function. Furthermore, by using the net-input value to the designated cell, the estimated output of the cell was determined. Various functions that have been allowed may have an impact on the activation function’s effectiveness. In order to overcome these challenges, the logistic sigmoid transfer function was chosen as an initiating function. This function has several advantages, including the ability to handle non-linear functions and continuous and diverse processes.

The backpropagation method was designated as a general learning algorithm, and a variety of learning algorithms were worked on to improve the weight function in the ANN model. The weight function is updated only by this backpropagation technique in tandem with the variations in the network’s output and its data, which was processed to produce the best outcomes40. In the artificial neural network model, these machine learning parameters were often constrained and active. Adaptive rule, conjugate gradient, Bayesian regularises, and Levenberg-Marquardt method were the frequently used training functions for the artificial neural network (ANN). The LM method was applied to a hidden layer consisting of numerous neurones that were set up in a network structure for a powerful operation that produced an excellent result and a perfect match between the output values and the experimental test values.

  1. (c)

    ANN architecture and its network activity.

  2. (d)

    Neurones in the hidden layers with a transfer function are followed by neurones trained with layers in the general architecture of artificial neural networks19. Typically, the output layers of the ANN model were configured with the hidden layers, and the input layer was configured with hidden layers. The suggested study employed a feed-forward neural network criteria, wherein the training set and network underwent continuous modification through successive model iterations contingent on data inputs. Incorporating production and a hidden layer of the ANN model, the output of this network was finally evaluated. The ANN model’s output layer and hidden layer were previously assessed, as was the feed-forward network and transfer function. A crucial requirement for networks and network performance simulation, the transfer function selection for this ANN model is required to be included in the learning rule. To evaluate the descriptive response, this model’s transfer function was put into practice.

Fig. 6
figure 6

Pictorial representation of artificial neural network (ANN) architecture.

Full size image

Gradient descent activity gradually dictated the errors of the ANN network through the backpropagation learning technique, which was formerly known as feed-forward architecture. The Levenberg–Marquardt training technique and the back propagation neural network—the artificial neural network architecture shown in Fig. 6—were both selected for the accurate models. Biodiesel is a valuable substitute for traditional fossil fuel, according to Karthickeyan et al.25. The ANN model tool utilised in this work estimates and forecasts performance and emissions using seven different training techniques. Levenberg–Marquardt (trail), which yields log and tan sigmoidal transfer function outputs, is the finest of the seven methods. Using various mixtures of biodiesel, the recommended ANN modelling method predicts engine performance and tailpipe emissions with high accuracy. With petroleum-Diesel at all compression ratios, the fuel mix B20 demonstrated the BTE and decreased the BSFC. Except for NOx emissions, this fuel blend has substantially lower CO and HC emissions than petroleum-diesel22. Kannan et al. investigates the effects of using Waste Cooking Palm Oil (WCPO) biodiesel in a single-cylinder, 1500 rpm CI engine. Using an ANN model with injection timing and pressure, the researcher evaluates and predicts engine performance, combustion, and tailpipe emissions. At 280 bars of injection pressure and 25.5degbTDC timing, the WCPO methyl ester offers better combustion and performance parameters. To carry out the ANN training, the model was equipped with the Backpropagation learning algorithm. The input and output criteria of the MLP network were employed in the ANN model for non-linear mapping. The ANN network with two hidden layers and eleven neurones produces the exceptional correlation coefficient23.

This ANN prediction was carried out using MATLAB software and a single hidden layer for the performance and emissions parameters of the diesel engine, based on the literature review. Two input layers, ten to nineteen neurones in the hidden levels, two neurones in the output layer, and (2-5-1-1) for performance were used to develop an architecture set. In order to examine the correlation metrics, this model was put into use. To identify implementation flaws, model performance metrics and other mistakes were compared. Up till the mean square error was less than 0.01, several neurones were changed. In order to remove values that are difficult to read, one must remove the most common measurement errors, which include zeros and ambiguous values with necessary scale independence.

  1. (e)

    Developed ANN model to forecast the performance and emission measurement of GGME and its mixtures.

Based on the various literature reviews, this ANN model was used for this research work. Then the statistical analysis and error measurement was estimated and performed with RSME, R2, MSRE, NSE, and MAPE for the evaluation of the performance of the proposed ANN model, which was displayed below:

$$\,\,\,\,\,{\text{MAPE }}\left( {{\text{Mean Absolute Percentage Error}}} \right)=\frac{1}{n}\sum\limits_{{i=1}}^{n} {\left( {\left| {\frac{{{Z_e}\,-\,{Z_p}}}{{{Z_e}}}} \right|} \right)\,} \,x\,100$$
(3)
$${{\text{R}}^{\text{2}}}{\text{ (Determination}}\,{\text{coefficient) = }}1\, - \,\,\left( {\sum\limits_{{i=1}}^{n} {\left[ {\frac{{{{({Z_e}-\,{Z_p})}^2}}}{{{{(\,{Z_p})}^2}}}} \right]} \,\,} \right)$$
(4)

In these equations, n is the Experimental data, Ze is the Experimental yield, Zp is the predicted yield, Za is the Average experimental yield.

Typically, the minimum and superior value of MAPE and R2 must be lesser than 80% to accomplish the accurate model.

$${\text{RMSE}}\,\,{\text{=}}\,\,\sqrt {\,\frac{1}{n}\,\sum\limits_{{i\,=1}}^{n} {\,{{({Z_{{\text{ }}i}}-\,{X_i})}^2}} }$$
(5)
$${\text{MSRE}}\,\,{\text{=}}\,\,{\left[ {\frac{1}{n}\,\sum\limits_{{i\,=1}}^{n} {\,{{\left( {\frac{{{Z_i}-{X_i}}}{{{Z_i}}}} \right)}^2}} } \right]_q}$$
(6)
$${\text{NSE}}\,\,{\text{=}}\,\,{\left[ {1 - \,\left( {\sum\limits_{{i\,=1}}^{n} {{{\left( {\frac{{{Z_i} - {X_i}}}{{{X_i} - \overline {X} }}} \right)}^2}} } \right)} \right]_q}$$
(7)

The performance and tailpipe emission of raw GGME and its fuel blends with petroleum-Diesel blends were examined as BTE, BSEC and HC, CO, NOx, CO2, and smoke opacity with the experimental input of fuel blends and BP. Both the Figures and Tables of the ANN model exhibits accurate, descriptive, and predicted nature of model assessed values for the experimental data of performance and emission of GGME models. The model resulted in errors also in line with the output results.

Diesel engine experimental setup and its measurements

The basic empirical study was conducted in an air-cooled, four-stroke, 5.2 kW Kirloskar brand DI Diesel engine operating at 1500 rpm. In order to conduct an experiment, the principal instrument was mounted in the engine and tested under various load conditions while maintaining a steady speed. Regular petroleum-Diesel fuel was first used to evaluate the test engine in order to excite it and obtain baseline measurements. At first, an eddy current dynamometer with a dynamometer and electrical resistance was coupled to the air-cooled test engine. It was tested with 5.2 kW of power and a steady speed of 1500 rpm. In order to monitor the pressure rise versus crank angle on a flywheel, a Kistler-made piezoelectric pressure transducer, model number 7063-A, was mounted to the test engine and coupled to a crank angle encoder. The engine was installed alongside the SES combustion analyser. The encoder, analyser, and charge amplifier made by Kistler Instrument AG in Switzerland were extremely important since they allowed the input signal from the diesel engine to be utilised by a large number of people. In order to precisely quantify the buildup of tailpipe emissions, the QROTECH Co Ltd. five gas analyser and the QRO-402 models of Korean equipment were used to measure the emissions of HC, CO, CO2, and NOx. The AVL 437 C smoke opacimeter device was used to measure the emissions of smoke opacity. Figure 7 describes the Kirloskar made DI Diesel engine arrangement. During this engine test, SAE 40 lubricating oil was used to increase engine effectiveness and prevent wear and tear issues. Table 3 provided a complete specification for the diesel engine.

Fig. 7
figure 7

Layout illustrating the diesel engine experimental setup53.

Full size image
Table 3 The detailed specification of the diesel engine.
Full size table

Uncertainty analysis of the experimental work

Numerous factors, including observations, testing standards, instrument calibration for purchased instruments, and equipment selection, would contribute to the uncertainty and inaccuracies in the instruments. In general, random mistakes and fixed errors can be used to combine uncertainty. For this uncertainty analysis, the analytical quantities are manageable. Next, the Gaussian distribution function shown in Eq. (8), along with the confidence bounds, was used to estimate the uncertainty of the evaluated parameters. The figure is known as the mean limit. About 96% of the deals in this were measured.

$${\text{\varvec{\Delta}}}{{\text{Z}}_{\text{i}}}\,{\text{=}}\,\frac{{{\text{2}}{{\text{\varvec{\upsigma}}}_{\text{i}}}}}{{\overline {{{{\text{Z}}_{\text{i}}}}} }}\,{\text{x}}\,{\text{100}}$$
(8)

In the above equation, the parameter \(\overline {{{{\text{Z}}_{\text{i}}}}}\) is the experimental interpretations, \({{\text{Z}}_{\text{i}}}\) is a number of readings, \({{\text{\varvec{\upsigma}}}_{\text{i}}}\) is the value of standard deviation. The estimated uncertainty of various constraints was explored as follows in Eq. (9):

$${\text{P}}\,{\text{=}}\,{\text{f(}}{{\text{Z}}_{\text{1}}}{\text{,}}{{\text{Z}}_{\text{2}}}{\text{,}}{{\text{Z}}_{\text{3}}}{\text{,}}{{\text{Z}}_{\text{4}}}{\text{,}}{{\text{Z}}_{\text{5}}}{\text{,}} \ldots {{\text{Z}}_{\text{n}}}{\text{)}}$$
(9)
$${\text{\varvec{\Delta}P=}}\,\sqrt {\,\,\left( {{{\left[ {\frac{{\partial {\text{P}}}}{{\partial {{\text{Z}}_{\text{1}}}}}{\text{\varvec{\Delta}}}{{\text{Z}}_{\text{1}}}} \right]}^{\text{2}}}{\text{+}}\,{{\left[ {\frac{{\partial {\text{P}}}}{{\partial {{\text{Z}}_{\text{2}}}}}{\text{\varvec{\Delta}}}{{\text{Z}}_{\text{2}}}} \right]}^{\text{2}}}{\text{+}}{{\left[ {\frac{{\partial {\text{P}}}}{{\partial {{\text{Z}}_{\text{3}}}}}{\text{\varvec{\Delta}}}{{\text{Z}}_{\text{3}}}} \right]}^{\text{2}}}{\text{+}}\, \cdots {{\left[ {\frac{{\partial {\text{P}}}}{{\partial {{\text{Z}}_{\text{n}}}}}{\text{\varvec{\Delta}}}{{\text{Z}}_{\text{n}}}} \right]}^{\text{2}}}} \right)}$$
(10)

In the above equation, the functions\({{\text{Z}}_{\text{1}}}\), \({{\text{Z}}_{\text{2}}}\), \({{\text{Z}}_{\text{3}}}\),…\({{\text{Z}}_{\text{n}}}\)delivers the number of readings taken from the experiment. Thence, “\({\text{\varvec{\Delta}P}}\)”was calculated by the root mean square and the errors related to the measured limits. Uncertainty analysis of several measuring instruments was exhibited in Table 4. The above Eq. (10) was assessed for measuring constrains and it was explored in Table 5.

Table 4 Uncertainty range of several instruments used in the experiments.
Full size table
Table 5 The uncertainty of measured constraints of the diesel engine.
Full size table

Results and discussion

Nanoparticles characterization and its techniques

XRD analysis of the Fe3O4 nanoparticles

Fig. 8
figure 8

X-ray diffraction pattern of Fe3O4.

Full size image

An equipped powered sample of Fe3O4 nanoparticles was exposed to XRD on a diffractometer to perceive the crystalline nature, phase status, and average grain size of Fe3O4 nanoparticles were exemplified in Fig. 8. The XRD patterns were noted from 2Ө = 39.4° active with D2 phaser XRD by Bruker operating organization by resources of CuKα (1.5406Å) radiation with a scanning rate of 5º/min. The provided voltage was established at 20 KeV and 150 mA, correspondingly. The crystallized structure of Fe3O4 nanoparticles was recognized by increased planes, viz. (5 1 1), (4 0 0), (4 2 2), (2 2 0), (3 1 1) and (4 0 0). In connector with these high and penetrative peaks (3 1 1) confirms the survival of Fe3O4 nanoparticles. The deflection peaks of Fe3O4 with planes (2 2 0), (3 1 1), (4 0 0), (4 2 2), (5 1 1), and (4 0 0) support the crystalline system of synthesized nanoparticles. Powerful peaks were noted in the planer of (3 1 1), and by this resource, the grain size was premeditated by using the Debye–Scherer technique, and its common grain size was ambitious to 29.22 nm. The lattice structures were discovered from the JCPDS line numbers (89-3854). By using this, it was addictive that the system is Face Centred cubic (FCC) systemwith mean axial ratio of 0.14 and the theoretical density of Fe3O4 nanoparticles, also known as magnetite were found to be 5.17 g/cm3.

Fourier transform-infrared spectroscopy (FT-IR) analysis for Fe3O4 and lipase immobilized Fe3O4

FT-IR characterization was victimized to attach the capping molecules state on the layer of nanoparticles by physiochemical technique and FT-IR absorption peaks for elegant Fe3O4 and lipase immobilized Fe3O4, as verified in Fig. 9. An identified characteristics spot at 3412 cm− 1 approves the existence of OH bond attached to iron oxide tangible.

Fig. 9
figure 9

(A,B) exhibit the FT-IR analysis of Fe3O4 nanoadditives and lipase immobilized Fe3O4 nanocatalyst.

Full size image

This way, the NH2 stretch was missing in the nanoparticles, and therefore, it is inferred that these compounds have to undergo reduction method while C=O bond was known at 1626 cm− 1 the distinguishing peak at 571 cm− 1 of Fe–O character confirms the survival of Fe3O4 nanoparticles in the gained sample. Lipase immobilized Fe3O4 peak disclosed the distribution peaks at 3326 cm− 1 regions, and one more characteristic peak was identified at 1427 cm− 1. These peaks show the proximity of lipase bounded in the Fe3O4 nanoparticles.

Fig. 10
figure 10

(A) SEM characterization of Fe3O4 and immobilized Fe3O4, (B) surface morphology of Fe3O4 nanoparticle and lipase immobilized Fe3O4 nanocatalyst.

Full size image

Figure 10A,B shows the external morphology of Fe3O4 nanoparticle and immobilized Fe3O4 analyzed by SEM make: VEGA-3-TESCAN. Commonly, electrons reacted with atoms in the sample, and it results in different signals. These signals are treated, and an SEM image was produced. In the specific SEM images, the morphological nature of Fe3O4 magnetic nanoparticles exhibits roughly in nature due to the agglomeration of fine particles to magnetic coupling and the attracting nature of individual nanoparticles Fig. 10A, agglomeration and inelegance are abstracted in Fig. 10B. This result, better molecules, and binding occurred during the reaction. Figure 10A confirm the clear surface of Fe3O4 with 60 nanometres (nm) range, and Fig. 10B shows the more grain growth surface of immobilized Fe3O4 nanoparticle with 60 nm scale.

Characterization of Garcinia Biodiesel

FT-IR analysis of FAMEGO

To examine the FAMEGOs compound profile, the current generation high-resolution compound analyzer, FT-IR, was handled in this experimental study. The FT-IR analysis duration was 1–2 s only. The FT-IR instrument was presently preferred to examine the compound’s presence in the extracted FAME GO and raw Garcinia seed oil from the version of the latest omnic software shown in Fig. 11A,B. The whole analysis was operated with the Nichrome source configuration, 7.3 KBr splitter, Deuteratedtriglycinesulfate-based detector installed with Magna 750 FT-IR spectrometer arrangement. The overall range of 500 to 4000 cm−1 with a resolution (5 cm−1) was fixed in the FT-IR instrument for this analysis. The FT-IR spectrum ranges create a chemical inequity for both FAME GO, and raw Garcinia seed oil ranges from 3000 to 2800 cm−1, named a fingerprint region.

Fig. 11
figure 11

(A,B) Illustrate the FT-IR spectrum of raw Garcinia seed oil and fames of Garcinia.

Full size image

FT-IR for extracted FAME GO produces the sharp peaks of 1721 cm− 1 and 1843 cm−1, determine the functionality of esters, acids in –C =O stretch, carbohydrate, and saturated fatty acid in the acquired sample. Another successive peak, 1570 cm− 1 in the spectrum band, belongs to an amide group of compounds. Shaped peak (1321 cm− 1) deserves O–CH3, which a glycerol-based group is attributed to FAME GO. For raw Garcinia seed oil samples, the FT-IR results in the existence of fatty acid groups and their compounds. Figure 11B shows sharp peaks from 2850 to 3003 cm− 1 raised regions to agree to sp2–CH, sp2–CH, and aliphatic CH (hybridized alkene group). Meanwhile, 1737 cm− 1 stretched peak related to the keto group of esters, and the 1627 cm− 1 region is for non-keto group with unsaturated double bond availability. Numerous categorized firm peaks raised in-between 3000 and 2800 cm− 1 area that is 3003 and 2850 cm− 1 peaks related to the methyl ester group. Another peek at 1469 cm− 1 in the region gives evidence of methylene groups. The other peaks, like, 1627 cm− 1, fit into (olefin group C = C), 3003 cm− 1 for (Alkene C–H) that are an Aliphatic group with sp2-CH, 1731 cm− 1 stretch (Carbonyl single bond ester O–C = O) and last peak in the spectrum band 2850 cm− 1 got Alkane (C–H sp3–CH) group.

Gas chromatography-mass spectroscopy characterization of extracted FAMEGO

Fig. 12
figure 12

GC analysis of extracted Garcinia FAME peaks.

Full size image

GC–MS is a well-organized instrument used to scrutinize attracted FAME GO physiochemical properties from raw Garcinia, including finding the fatty acid profile and the compositions in a well-sophisticated manner. Also, prone to search the various carbon double chains and the fuel properties, namely, oxidation rate, combustion rate of the oils, fuel viscosity, cetane number, carbon-chain length, cold fuel property, and heating value of the biodiesel24.

In general, the fatty acid configuration of the crude oil was greatly influenced by the physicochemical properties of biodiesel and evolved the knowledge of the fatty acid profile of the raw crude oil that is used to produce biodiesel. Meanwhile, biodiesel physiochemical properties are the foremost ingredient for all the fuel performance and economy. Presently, GC analysis was engaged for the extracted FAMEGO, and it discovered the available fatty acid methyl esters, fatty acids, fatty acid ratios, and other significant information. The chemical composition of Garcinia seed oil and comparison of fatty acid methyl ester of Garcinia seed oil with various fatty acid methyl esters were displayed in the Tables 6 and 7 and the GC analysis result was represented in Fig. 12.

Table 6 Chemical composition of raw Garcinia seed oil52.
Full size table

Figure 12 shows that the GC analyzer generates various peaks related to relative abundance (%) versus Time (min) deals with extracted FAMEGO. As pertaining to further peaks, 39.09% and 35.69% stretch was sharper and broad peaks in the graph. The peak 35.69% shows (E)—methyl octadic—13 enoate related to oleic acid and additional peak 39.69% stretch equivalent for methyl 16—methyl heptadecanoate it known as margaric acid and other peaks results, myristic acid, linoleic and arachidic acid consistently. Commonly, unsaturated fatty acids presented in the raw oils results in a lower level of biodiesel production.

Table 7 Fatty acid methyl ester of raw Garcinia seed oil and other type of oils52.
Full size table

Mass Spectroscopy (MS) is an instrument specialized to determine the structures and compounds exposed in the extracted oil samples. It works in the mechanism of NIST, which generates ion peaks correlated to the molecules. THE extracted FAME GO sample was located in the MS, then the β-carbon atoms were developed by the instrument, and the β-carbon atoms were permitted to pass over the FAME GO molecules.

Fig. 13
figure 13

Mass spectroscopy structure of extracted Garcinia FAME molecules52.

Full size image

Hence, it possesses an MC Lafferty molecular rearrangement spectrum. This spectrum had a gathering of all mass fragmentation linked to the sample compounds. These collections of fragment patterns were recorded by mass spectroscopy, and based on the design, the structure and compounds were estimated from the FAME GO molecules, and it is exemplified in Fig. 13. As defined in Fig. 13, the structure A—10, 13-trimethyl myristate, B—Methyl linoleate, C—Methyl oleate, D—16-dimethyl margarate, and E-Methyl arachidonate compound. These are the IUPAC names of the essential components that are presented in the extracted FAME GO.

Synthesis and characterization of TiO2 nanoparticles

The process commences with the treatment of titanium tetraisopropoxide, the precursor material, using ethanol, hydrochloric acid, and deionized water. The resulting mixture underwent stirring for 30 min until it attained a pH of 1.5. Following that, 10 mL of deionized water was introduced, and the stirring continued at room temperature for 2 h, eventually reaching a pH of 6. Further titration with deionized water led to the formation of cladigel, with a pH of 8. Subsequently, the cladigel was dried and subjected to calcination at 150 °C for 12 h. Lastly, the TiO2 nanoparticles were obtained by subjecting the dried samples to a temperature of 300 °C for 2 h. The physical properties of the synthesized TiO2 nanoparticles are elaborated in Table 8. HR-TEM and TEM morphology (Fig. 14a,b) carried out using the JEM-3010 ultra-high-resolution analytical electron microscope revealed distinct lattice fringes, indicating the presence of nanocrystalline TiO2 particles. The SEM image (Fig. 14c) obtained using the VEGA3-TESCAN preparation showed an irregular spherical distribution of nanoparticles with slight aggregation.

Fig. 14
figure 14

(a) HR-TEM, (b) TEM, (c) SEM of TiO251.

Full size image
Table 8 Properties of synthesized TiO2 nanoparticles51.
Full size table

Engine characteristics

This four-stroke, single-cylinder, water-cooled compression ignition engine was used for all of the testing. To ensure unambiguous findings, the CI engine was driven for 10 min without fuel after the nanocatalyst mixed biofuel was obtained. This was done in preparation for conducting experimental tests. The various gasoline mixes, including Diesel, B10, B20, Diesel-TiO2, B10-H2-TiO2, and B20-H2-TiO2, were tested for each one. The dynamometer has the engine attached to it. The data procured through PC was linked to an AVL gas analyzer, which was utilized in that configuration to obtain the emission patterns21. To obtain performance data, a PC with the performance measuring system installed was used. To regulate H2 gas supply, a filter was utilized. Through the conveyance of the data to PC, the data logger was used to calculate the data. Table 1 enumerates the Physico-chemical properties of the test fuels. According to Tables 2 and 3 every test was carried out at a different engine speed ranging from 1800 rpm to 2200 rpm.

Performance characteristics

Torque

Each test fuel (Diesel, B10, B20, Diesel-TiO2, B10-H2-TiO2, and B20-H2-TiO2) provided the same amount of engine torque. As the engine speed increases, the torque produced decreases, as seen in Fig. 15. This occurs as a result of both incomplete combustion and disrupted air intake at faster speeds. The torque at high speeds is therefore decreased by the increasing pressure. Yet, the B10-H2-TIO2 fuel blend produced more torque (21.3 N-m) than other fuel blends when a 5 lit/min hydrogen supply was applied at a slower speed of 1800 rpm. Diesel-TiO2, B10-H2-TiO2, B20-H2-TIO2, DIESEL, B10 and B20 blend possesses torque of about 21.8, 21.3, 20.5, 19.5, 20.1, and 19.5 Nm at low engine speed. Engine torque was improved by 6% and 5% for B10-H2-TiO2 and B20-H2-TiO2, respectively, when hydrogen and TiO2 were added to the normal blend. Figure 16 also shows that the engine’s torque was increased by the addition of hydrogen and nanoparticles. Torque decreased dramatically as engine speed rises.

Fig. 15
figure 15

Variation of torque with respect to engine speed.

Full size image

For example, between lower engine speeds and 2800 rpm, the torque loss for the blends Diesel, B10, B20, Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 was 24%, 27%, 28%, 28%, 26%, and 27%. Due to the significant decrease in thermal efficiency, torque has drastically decreased.

Brake thermal efficiency (BTE)

Fig. 16
figure 16

Variation of brake thermal efficiency (BTE) with respect to engine speed.

Full size image

The BTE values for all gasoline blends—Diesel, B10, B20, Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 at various engine speeds from 1800 rpm till 2700 rpm were displayed in Fig. 16. The diagram illustrates the basic fact that an increase in engine speed lowers the BTE in a manner akin to the torque pattern of the test engine. The Diesel fuel produced a higher BTE at 1800 rpm than the B10 and B20 fuel without the influence of hydrogen or nanoparticles. B10 and B20 blends had BTEs of 21.5% and 21%, respectively, but pure DIESEL had a BTE of 24%. By adding nanoparticles to fuel mixes with hydrogen supply as Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2, the corresponding percentages at 1800 rpm were 24.2%, 22%, and 21.6%, respectively. The highest BTE value was obtained by the biofuel blend B10 due to combined influence of nanoparticles with hydrogen resulted in a reduced viscosity. The increased BTE is caused by the hydrogen’s stronger diffusivity and faster flame41., With nanoparticle addition in biofuel, the BTE is found to increase in several literatures (Attia et al.54; Arockiasamy& Anand55; Aalam & Saravanan56; Shaafi & Velraj57; Anbarasu et al.58). However, some contradictory results with higher nanoparticles giving rise to lowered BTE were reported in few literatures (Prabu & Anand59; Sharma et al.60).Additionally, hydrogen has a reduced quenching distance. In addition, the BTE is increased by the nanoparticles, which are vital. One more factor contributing to a full combustion and the BTE rise is the increasing oxygen content and decreased viscosity. The fuel Diesel, B10, B20, Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 exhibited a progressive decrease in BTE values at 2400 rpm and 2700 rpm. However, the maximal hydrogen supply with nanoparticles at each speed produced the highest values (Fig. 17).

Brake specific fuel consumption (BSFC)

Fig. 17
figure 17

Variation of brake specific fuel consumption (BSFC) with respect to engine speed.

Full size image

Regarding the BSFC, the result is the reverse of the BTE. A higher BSFC results from faster speeds. In addition, at lower engine speeds (especially 1800 rpm), the BSFC range of Diesel, B10, and B20 were about 421, 382 and 358 g/kW-h. The minimum value for B20’s specific fuel usage was this. In addition, the BSFC decreased at all speeds as the percentage of hydrogen and TiO2 particles increased. The significantly reduced value of BSFC, 349 g/kWh, was obtained, particularly for the fuel blend B20-H2-TiO2.Similarreductions in BSFC with nano additives were also quoted by several researchers (Prabu and Anand59; Shaafi and Velraj57; Patel and Kumar61).

With increasing nanoparticle concentration, the BSFC is found to decrease.

in the literatures of Attia et al.54; Arockiasamy and Anand55; Aalam and Saravanan56; Shaafi and Velraj57 and Anbarasu et al.58. However, some contradictory results with higher nanoparticles giving rise to higher BSFC were reported such as Prabu and Anand59; Sharma et al.60 which could be attributed to prolonged ignition delay along with shorter spray penetration causing poor combustion efficiency.While the engine speed rose, the BSFC values showed a considerable increase in fuel consumption42,43. At lower speeds, the BSFC data was lesser for Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 (421, 382, 358, 403, 368, and 344 g/kW-h). The full combustion and increased heating value brought about by the blending of nanoparticles and hydrogen are responsible for the BSFC decrease. It is comparatively, cost-effective in comparison with Diesel with surplus H2 supply or even pure Diesel (excluding additives) since B20-H2-TIO2 has a lower brake-specific fuel consumption44.

Emission characteristics

Upon experimentation for the blends of Diesel, B10, B20, Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 the following figures illustrate the varying emission rates of carbon monoxide, Carbon dioxide, and nitrous oxides at various engine operating speeds from 1800 rpm till 2700 rpm. Emission rates from fuel must be measured in order to prevent the greenhouse effect from getting worse.

Carbon monoxide (CO) emission

Figure 18 displays the value of CO emission for several gasoline mixes (Diesel, B10, B20, Diesel-TiO2, B10-H2-TiO2, and B20-H2-TiO2) at various engine speeds. It is clear from the figure that the addition of hydrogen and nanoparticles reduces the amount of CO. This is due to the fact that an increase in biofuel will result in an increase in oxygen content45. Similar studies with Al2O3 nanoparticles addition lowering the CO emissions were shown by several researchers (Shaafi and Velraj57; Sivakumaret al.62; Hosseini et al.63). With increasing nanoparticle concentration, the CO is found to decrease in the literatures of Attia et al.54, Gumus et al.64, Arockiasamy and Anand55 and Chandrasekaran et al.65. However, some contradictory results with higher nanoparticles giving rise to higher CO were reported such as Kannan et al.66, and Aalam and Saravanan56 which is possibly due to the hindering of oxidation process with more nanoparticles stopping CO oxidation to CO2 molecule and stays as CO itself.The biofuel’s higher oxygen content, the addition of nanoparticles that reduce viscosity, and the addition of hydrogen that increases diffusivity and flame speed all contribute to a clearer combustion rate that leaves no fuel unburned. The CO emissions are also caused by the fuel’s unburned components. Furthermore, because hydrogen lacks the carbon structure, it emits less CO. Full combustion is also made possible by the TiO2’s large energy surface area. Conversely, when the speed is increased, so did CO emissions. Fuel mixes Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 had CO levels of 0.079, 0.043, 0.06, 0.081, 0.044, and 0.387% vol. at 1800 rpm, respectively. Compared to standard blends of B10 and B20, the emission of CO values was greater with the hydrogen supply.

Fig. 18
figure 18

Variation of carbon monoxide (CO) emission with respect to engine speed.

Full size image

The blends containing hydrogen B10-H2-TiO2 exhibited a 2% increase in hydrogen generation as compared to B10. However, the B20-H2-TiO2 reported 25% lower CO emissions than the B20, which is only attributable to the fuel burning through to the end. Higher concentrations of biodiesel resulted in a significant decrease in CO emissions. Hydrogen added to the biofuel blends and pure Diesel reduced CO by an average of 26%.

Carbon dioxide (CO2) emission

Fig. 19
figure 19

Variation of carbon dioxide (CO2) emission with respect to engine speed.

Full size image

The CO2 emission as a function of engine speed is shown in Fig. 19. While CO2 emissions rise with speed, at a certain point in time, clean Diesel and biofuel blends containing hydrogen and nanoparticles as well as fuels without additives—such as DIESEL, B10, and B20 have showed higher CO2 emissions46. For instance, at lesser engine speeds, the CO2 emissions of Diesel, B10, and B20 was found to be 4.2%, 3.8%, and 3.5%, respectively, whereas the CO2 emissions of Diesel-TiO2, B10-H2-TiO2, and B20-H2-TiO2 were only 4.2%, 3.7%, and 3.5%, respectively, significantly lower than the plain fuel mixes. According to the research, running an engine at a fast speed causes a higher percentage of full combustion, which increases the release of CO2 and other gases that are comparable to CO32. Because of the carbon concentration of the biodiesel mixes, there has been an increase in emissions. Since hydrogen gas doesn’t contain carbon molecules, adding hydrogen to biodiesel decreased the amount of CO that was released.

Nitrogen oxides (NOX) emission

Fig. 20
figure 20

Variation of oxides of nitrogen (NOx) emission with respect to engine speed.

Full size image

Figure 20 displays the variation of nitrogen oxides emissions of Diesel, B10, B20, Diesel-TIO2, B10-H2-TiO2, and B20-H2-TiO2 at various. NOX emissions, in fact, rises in tandem with increased velocity.It results from the higher temperature of combustion. The emission of NOX is enhanced not only by increasing speed but also by including hydrogen47. The reason for the elevated cylinder temperature can be attributed to the elevated calorific value of hydrogen. With increasing nanoparticle concentration, the NOx emissions are found to decrease in the literatures of Attia et al.54, Gumus et al.64, Arockiasamy and Anand55, and Chandrasekaran et al.65. However, some contradictory results with higher nanoparticles giving rise to higher NOx emissions were reported such as Shaafi and Velraj57, Mohan et al.67 and Prabu and Anand59. This is due to the fact that, the oxygen buffering tendency of nanoparticles getting shattered with higher viscous blend which along with higher in-cylinder temperatures promoting the O2 and N2 mixing followed by higher NOx emission.The NOX emission increases in conjunction with the cylinder temperature peak48,49. Diesel, B10, and B20 plain fuels had NOX emissions of 220 ppm, 350 ppm, and 560 ppm, respectively; in contrast, fuel blends with hydrogen and nanoparticles added, B10-H2-TiO2 and B20-H2-TiO2, reported NOx reductions of 20 ppm and 30 ppm compared to B10 and B20. Figure 20 shows that the temperature rise for NOX emission is the major factor to take into account in this situation.

Proposed ANN model validation and its analysis of the GGME fuelled engine output

The Diesel engine experimental work was performed based on the design matrix output, which was designed and developed by experiment software design its details were displayed in Table 9. Then the ANN model was used to optimize and predict the experimental data’s (Diesel engine output) and for this ANN model, Diesel engine brake power, biodiesel, and its fuel blends, Diesel fuel was assigned as an input parameter, and the BTE, BSFC, CO, CO2 and NOx, emissions were assigned as an output parameter37.

Table 9 Design matrix output for various performance and emission parameters.
Full size table
Fig. 21
figure 21

Experimental and predicted results of (a) BTE (b) BSFC (c) CO (d) CO2 and (e) NOx.

Full size image

Then the Diesel engine experimental work results, Garcinia B20 fuel blend generate 8.08% drop in BTE compared to petroleum-Diesel, and mineral Diesel holds lesser energy consumption rate but B20 blend yield higher energy consumption range by its properties like calorific value and viscosity. In the case of emissions correlated with other fuel blends, B10 and B20 blends punch lowered rate of 8.39% and 10.35% with petroleum-Diesel at topmost load condition. Both B20 and Diesel-TiO2 blends hold significantly reduced CO emissions of 18.42% and 19.20% with petroleum-Diesel, and this could be owing to the higher carbon chain length of GGME fuel, lessened viscosity, and density. Compared with mineral Diesel fuel, the B10 and B20 fuel blends result in a higher range of CO2 emissions at peak load conditions, and B10 and B20 blends have 700 ppm and 744 ppm of higher NOx emissions related to mineral Diesel fuel at higher load condition.

The ANN model error and performance was tabulated in Table 10. The Fig. 21a,b reveals the experimental, prediction and optimization of the Diesel engine performance (BTE and BSEC) and its R2, RMSE and R value for BSEC (0.99373, 0.250301 and 0.9996) and BTE (0.992448, 0.419935 and 0.993). From these results, the R value had better accuracy rate for the Diesel engine performance model with an RMSE lesser value. Next, Fig. 22c–e display the predicted, optimized and experimental result of CO, CO2 and NOx emissions with the R2, RMSE and R values are HC (0.995783, 0.635321 and 0.99587), CO (0.964267785, 0.002488626 and 0.9983), CO2 (0.997672579, 0.031630703 and 0.99725), NOx (0.984559489, 44.89075405 and 0.9847) and smoke emission (0.986814591, 1.00294331 and 0.9990). Moreover, the RSME had lowered rate of percentage error which was related with R value of the ANN model result38,39.

Table 10 Error and performance of the ANN model.
Full size table

The MSRE, NSE and MAPE error standard measurements values for ANN tool optimized Diesel engine performance and pollutant emissions model was noticed in the range of (3.02 to 4.07% and 0.009057 to 0.008773). Then the overall correlation coefficient of the developed ANN network for the performance and pollutant emissions model were displayed from the Figs. 22, 23, 24, 25 and 26. In the Figs. 22 and 23, it was perceived that the ANN performance model (BTE and BSEC) results the R value in terms of training, validation, testing and all simulated value as (BTE) 0.99916, 0.99993, 0.99955 and 0.9993 and (BSEC) 0.99963, 0.99988, 0.99991 and 0.99969 correspondingly. Then Figs. 24, 25 and 26, it was noticed that the ANN emission model (CO, CO2 and NOx emission) yields the R value in terms of training, validation, testing and all simulated value as CO (0.993606, 0.99945, 0.9998 and 0.99838), CO2 (0.99795, 0.99997, 0.99456 and 0.99725) and NOx (0.9735, 1, 0.99847 and 0.98478).

Fig. 22
figure 22

Overall correlation coefficient and optimization of BTE.

Full size image
Fig. 23
figure 23

Overall correlation coefficient and optimization of BSFC.

Full size image
Fig. 24
figure 24

Overall correlation coefficient and optimization of CO emission.

Full size image
Fig. 25
figure 25

Overall correlation coefficient and optimization of CO2 emission.

Full size image
Fig. 26
figure 26

Overall correlation coefficient and optimization of NOx emission.

Full size image

Conclusion

This contemporary experimental investigation deals with biodiesel prepared with a novel immobilized lipase catalyst, and the present transesterification process was achieved. The collected GGME was analyzed with both physical and chemical properties, and it was correlated with ASTM and EURO fuel standards. The prepared GGME (biodiesel) achieved above 92% of biodiesel yield by eco-friendly lipase immobilized nanocatalyst which was tested with Kirloskar Diesel engine in terms of performance and tailpipe emissions and it results as follows: Diesel, B10, B20, Diesel-TiO2, B10-H2-TiO2, and B20-H2-TiO2 were used for execution of all tests at varying engine speeds from 1800 rpm till 2700 rpm. The outcomes are provided below.

  • The inclusion of nanoparticles and hydrogen improved the performance attributes, such as torque and brake thermal efficiency (BTE). In comparison with B10 (15.2%), the B10-H2-TiO2 has higher BTE (17%) which has an increase in BTE by about 11.84%. The primary cause of improved combustion resulting from inclusion of H2 gas (with higher flame velocity) and TiO2 (improved diffusivity), followed by the lowered viscosity and higher oxygen content of the nanoparticle addition.

  • When hydrogen and nanoparticles were added to fuel blends, the BSFC was likewise decreased. The enhanced fuel blends were not as high as fuel consumption of DIESEL, B10, and B20. Low specific fuel consumption can be attributed to improvement in combustion efficiency with addition of nano additives and hydrogen fuel which therefore contributes to higher fuel efficiency.

  • It is evident from a comparison of the emission characteristics that enriched biofuel lowers the emissions of hazardous gases like CO and CO2. The decrease in CO and CO2 rate is caused by complete combustion, which is achieved by the addition of hydrogen and nanoparticles, as well as charge dilution due to hydrogen results in overall lean combustion and carbon neutral characteristics. However, due of the elevated cylinder temperature, there was an increase in NOX emissions. It is reasonable to say that adding hydrogen and nanoparticles improves the quality of the biodiesel overall.Higher NOx emissions are environmental hazards but it can be mitigated by retarding the injection timing, water-in-emulsion, urea injection, etc.

  • The proposed ANN model achieved accurate performance and pollutant model with optimal architecture, and the Diesel engine experimental data were precisely predicted and optimized. This ANN model was approved with the statistical error measurement values, namely relative mean square error, mean square relative error, Nash and Sutcliffe efficiency, an absolute fraction of variance, and mean fundamental percentage error. Moreover, the RSME had lowered rate of percentage error which was related with R value of the ANN model result.