Fuel 269 (2020) 117472

Contents lists available at ScienceDirect

Fuel
journal homepage: www.elsevier.com/locate/fuel

Full Length Article

Performance and emission prediction of a compression ignition engine T

fueled with biodiesel-diesel blends: A combined application of ANN and
RSM based optimization
⁎
Mustafa Aydına, Samet Uslub, , M. Bahattin Çelikb
a
Pamukkale University, Technology Faculty, Automotive Engineering Department, Denizli, Turkey
b
Karabuk University, Engineering Faculty, Mechanical Engineering Department, Karabuk, Turkey

A R T I C LE I N FO A B S T R A C T

Keywords: In the present study, the performance and emission parameters of a single cylinder diesel engine powered by
Artificial neural network biodiesel-diesel fuel blends were predicted by Artificial Neural Network (ANN) and optimized by Response
Response surface methodology Surface Methodology (RSM). The data to be used for ANN and RSM applications were obtained by using bio-
Biodiesel diesel/diesel fuel blends at different engine loads and various injection pressures. ANN model has been devel-
Optimization
oped to predict the outputs such as brake thermal efficiency (BTE), brake specific fuel consumption (BSFC),
Prediction
exhaust gas temperature (EGT), nitrogen oxides (NOx), hydrocarbons (HC), carbon monoxide (CO) and smoke
Diesel engine
regarding engine load, biodiesel ratio and injection pressure. A feed-forward multi-layer perceptron network is
used to show the correlation among the input factors and the output factors. The RSM is applied to find the
optimum engine operating parameters with the purpose of simultaneous reduction of emissions, EGT, BSFC and
increase BTE. The obtained results reveal that the ANN can correctly model the exhaust emission and perfor-
mance parameters with the regression coefficients (R2) between 0.8663 and 0.9858. It is seen that the maximum
mean relative error (MRE) is less than 10%, compared with the experimental results. The RSM study demon-
strated that, biodiesel ratio of 32% with 816-W engine load and 470 bar injection pressure are the optimum
engine operating parameters. It is found that the ANN with RSM support is a good tool for predict and optimize
of diesel engine parameters powered with diesel/biodiesel mixtures.

1. Introduction vegetable or animal fats species and they have similar heating value as
compared to the diesel fuel [9–11]. Moreover, they do not include
Today, diesel engines are preferred in several industries because of harmful substances such as sulfur, minerals and releases lowest level of
their higher fuel economy, better efficiency and low fuel costs [1,2]. emissions because of high-level oxygen content [12]. In addition, the
However, the exhaust emissions of these engines have been adversely similar physico-chemical characteristic of biodiesel to diesel fuel are
affecting humanity and habitat for many years. Currently, investigators allowed for its adoption without any engine modification [6,13]. Since
emphasis on alternative fuels because of diminution of fossil fuel source biodiesel is recognized as a fuel alternative to conventional diesel fuel,
and concerns regarding the influence of fossil fuels use on ecological some researcher has been working on diesel engines powered with
problems such as air pollution and global warming [3–5]. To defeat this biodiesel which is produced from numerous sources to investigate their
problems, diesel engine investigators are aiming to discover an alter- effects on engine parameters [14–18].
native fuel that is able to increase the performance and reduce emis- Because of the traditional methods are very time-consuming and
sions of the engine [6]. Biofuels are one of the most remarkable alter- costly, researchers have turned to methods that can achieve the same
native fuels to conventional fuels owing to their parallel fuel performance faster and cheaper [19]. Nowadays, with the development
characteristic to fossil fuels and their ability to produce less greenhouse in computer technology, ANN and RSM has been used for numerous
gas and smoke emissions [7,35]. Between many possible alternative automotive engineering troubles. There are many studies in the litera-
fuels, biodiesel has recently gained rapid progress because their prop- ture using only RSM or ANN in diesel engines. Krishnamoorthi et al.
erties are very close with diesel fuel [8]. They can be obtained from performed the tests with different injection pressure and injection

⁎
Corresponding author.
E-mail address: sametuslu@karabuk.edu.tr (S. Uslu).

https://doi.org/10.1016/j.fuel.2020.117472
Received 3 December 2019; Received in revised form 4 February 2020; Accepted 23 February 2020
Available online 28 February 2020
0016-2361/ © 2020 Elsevier Ltd. All rights reserved.
M. Aydın, et al. Fuel 269 (2020) 117472

Nomenclature EGT exhaust gas temperature

HC hydrocarbon
AI artificial intelligence MRE mean relative error
ANN artificial neural network MSE mean square error
BSFC brake specific fuel consumption NOx nitrogen oxides
BTE brake thermal efficiency R2 correlation coefficient
CO carbon monoxides RMSE root mean square error
ECU electronic control unit RSM response surface methodology

timing at various compression ratio on a changeable compression ratio emission parameters of a diesel engine powered by diesel/biodiesel/
engine. After experiments, an optimization was made with RSM by diethyl ether combination. They created an ANN and RSM model to
taking the input parameters injection pressure, compression ratio and make minimum engine emission responses and to make maximum the
injection timing. According to the optimization results, compression engine performance responses. They used speed, load and compression
ratio of 18, injection timing of 21° before upper dead point and injec- ratio being input parameters in ANN and RSM models. They found that
tion pressure of 250 bar were found as optimum operating parameters CO, NOx and HC were decreased about 12.8%, 4.19% and 9.61%
[20]. Hosseini et al. created a model on ANN for estimating the power, compared to diesel, respectively [23].
torque and emissions on a single cylinder, direct injection diesel engine The studies cited above considered the impacts of biodiesel/diesel
powered with diesel/biodiesel/alumina fuel mixtures by different per- fuel combinations to engine performances and exhaust emissions of a
centages. Engine operating parameters such as fuel mixture, fuel den- diesel engine. However, the ability of the combined usage of artificial
sity, fuel viscosity and engine speed were selected as input parameters. intelligence (AI) techniques such as RSM and ANN to predict and op-
The R2 values for training, validation and testing in ANN were higher timize of a diesel engine input-output parameter have not been ade-
than 0.99 which are indicating high accuracy and performance. They quately investigated in the literature. Henceforward, this research was
argue that the results obtained from this study, efficiency and correct- carried out to fill this gap in the literature. Considering the above si-
ness of the ANN model was totally acceptable [21]. Baranitharan et al. tuation, an ANN is created as an accurate step to estimate the perfor-
investigated to estimate and optimize the diesel engine parameters mance parameters and exhaust emissions of diesel engine. Furthermore,
fueled with pyrolysis oil/diesel/Tert-butyl hydroxyl quinone anti- RSM model is formulated with ANN predicted values to discover the
oxidant blend with ANN and RSM. A single cylinder diesel engine suitable working conditions of diesel engine.
powered by 80% diesel + 20% Aegle marmelos bio-oil + 1000 ppm
Tert-butyl hydroxyl quinone antioxidant was used for getting the data 2. Materials and methods
for the RSM and ANN models. According the responses, the optimum
values of performance parameters such as BSFC and BTE were obtained Fig. 1 schematically illustrates the installation of the test device
as 0.33 kg/kWh and 22.01%, respectively. In addition, for CO, HC, CO2 including a diesel engine, electronic scale and fuel reservoir, an Elec-
and NOx, optimum values were found as 0.67%, 244 ppm, 8.33% tronic Control Unit (ECU), an electrical dynamometer, dyno controller,
and 351 ppm, respectively by RSM optimization. The MRE values were an emission analyzer and an opacimeter. In this study, experiments
0.552% and 1.231% for optimized RSM and predicted ANN model. The were conducted on a single cylinder diesel engine with 200 bar injec-
outcomes of this research show that A20D80T fuel mixture gives im- tion pressure. The injection pressure of the engine was adjusted by the
proved performance and emission values against with diesel fuel [22]. ECU with the fuel injection system assembly as 200–400–600 bar and
Krishnamoorthi et al. researched the performance, combustion and tested with four different test fuels at 3000 rpm engine speed, at five

Fig. 1. Schematic illustration of installation of the test device.

2
M. Aydın, et al. Fuel 269 (2020) 117472

Table 1 procedure, test data were compared with the estimated information.
Properties of test fuels [18,24,25]. The training phase of network was stopped once the test error achieves
Properties Diesel Biodiesel the desired level [24].
Regression coefficients, MRE and RMSE which were created with
Chemical Formula C13H28 C18H34 targets and outputs of the ANN model were used to measure estimation
Cetane number > 50 60.4
success of the models. The following equations were used for evaluation
Density (kg/m3, 15 °C) 0.839 0.87
Ignition temperature (℃) 240 316
[24]:
Boiling temperature (℃) 160–370 182–338 n 2
Kinematic viscosity (mm2/s – at 40 °C) 3.09 5 ⎛ ∑ (ti − oi ) ⎞
R2 = 1 − ⎜ i = 1n 2 ⎟
Lower calorific value (kJ/kg) 43 39.576
Latent heat of evaporation (kJ/kg) 260 325 ⎝ ∑i = 1 (oi ) ⎠ (1)

n
1 ti − oi
MRE(%) =
n
∑ 100
ti
various engine loads (500, 750, 1000, 1250, 1500 W). The test fuel i=1 (2)
obtained by mixing diesel and biodiesel at different ratios; 100% diesel,
80% diesel + 20% biodiesel, 60% diesel + 40% biodiesel and 40% 1
n

diesel + 60% biodiesel. The properties of the test fuels are presented in RMSE =
n
∑ (ti − oi)2
Table 1. i=1 (3)
In the experiments, firstly, engine was operated with standard diesel where ‘n’ is number of points in the information group, ‘o’ is predicted
fuel and reference data were generated. Subsequently, fuel blends in output data and ‘t’ is actual output.
different ratios were tested under the same test conditions. Before In this study, injection pressure, biodiesel ratio and engine load
starting the experimental measurements, the engine was operated with were selected as input parameter for input layer and EGT, BSFC, BTE,
test fuels to 90 °C operating temperature. Measurements were taken HC, NOx, CO and smoke were chosen as output parameter for output
after the engine started to run stable by loading the load unit. During layer. Schematic view of the developed ANN is shown in Fig. 2. The
the experimental studies, the measurements were repeated 3 times at feed-forward backpropagation network type was chosen which is
each measurement point. commonly used in order to explain complicated troubles in system
modeling and identification [28]. Levenberg-Marquardt (Trainlm)
2.1. Application of ANN training function was applied which is generally has been applied for
precise estimates wherein mean square error (MSE) determines the
ANNs are information processing techniques that are encouraged by failure function of the network [29]. (3-10-7) topology has been used
the biological neural system and are applied to explain a large diversity for predicting the input-output parameters, wherein three neurons for
of troubles, especially in engineering where traditional modeling input layer, ten neurons for hidden layer and seven neurons for output
techniques are inadequate [26]. ANN is popular in many fields of sci- layer. Most of the researchers [30–32] found that logistic sigmoid
ence and engineering and has also become popular in the automotive (logsig) activation function produces better success than other functions
industry. ANN is a mathematical and computational modeling tech- in formulating the appropriate model because it is a differentiable,
nique that can estimate the performance and emission parameters of continuous and nonlinear function. Also, the gradient descent with
engine [19]. ANN can have any number of layers, but generally consists momentum weight and bias learning function (LEARNGDM) which
of three layers: the input layer, the output layer, and the hidden layer computes the mass alteration for a dedicated neuron from the neuron's
[24]. The input layer contains of all the input parameters, data from the input and error, the mass, learning percentage, and momentum con-
input layer is then computed in the hidden layer and following output stant, with respect to gradient descent with momentum back-
vector is calculated in the output layer. There are three consecutive propagation was used to minimize the errors [33]. Particulars of the
steps in the estimation using ANN technique: Modeling; learning or neural network are given in Table 2.
training phase; and testing phase [27]. An ANN model was created ANN has 3 parts. In part 1, injection pressure, biodiesel ratio and
according to the selected input and output factors of the engine on engine load were selected as input data for ANN. Nearly 75% of the
modeling stage. During the training phase, the network was prepared to overall test data were selected for training and 25% for testing phase. In
create a target prediction based on input information. On the test Part 2, the ANN has been trained with this information many times

Fig. 2. ANN structure.

3
M. Aydın, et al. Fuel 269 (2020) 117472

Table 2
Particulars of the neural network.
Network 3 inputs, 7 outputs and one hidden layer
Data Training: 75% randomly selected data from experimental data
Test: 25% randomly selected data from experimental data
Network Type Feed-forward back propagation
Training function Trainlm
Adaptation learning function Learngdm
Transfer function Logsig
Performance function Mean square error
Stopping criteria Break the training of network when the confirmation error begins increasing

Fig. 3. Regression graphs of training, test, validation and all results of ANN.

because of obtaining the best trained ANN by validation, training and close to 1, thus ANN shows great correctness in modeling the outputs.
analysis of test schedules. Finally, in Part 3, certainty control of ANN The evaluation of the estimated and measured results for BTE, BSFC
was applied. and EGT are shown in Fig. 4. The R2 were found as 0.9144, 0.9626 and
0.9811, respectively. The results obtained from the ANN model show
that the use of ANN is enough to predict BTE, BSFC and EGT. RMSE
3. Results and discussion values are 0.55%, 25.65 g/kWh and 4.18 °C and MRE values are 5.92%,
5.44% and 3.49% for BTE, BSFC and EGT, respectively.
In this research, an ANN was designed utilizing the data collected in The evaluation of the estimated and measured outcomes for exhaust
the tests to estimate the diesel engine parameters. Engine load, bio- emissions such as CO, HC, NOx and smoke are demonstrated in Fig. 5.
diesel ratio and Injection pressure were selected as the input parameters R2 values were found as 0.9771, 0.8663, 0.8917 and 0.9858, respec-
while BSFC, EGT, BTE, NOx, HC, CO and smoke were chosen being tively. The results obtained from the ANN model show that the use of
output parameter. Overall regression graph from ANN in this study is ANN is enough to estimate the exhaust emissions for instance NOx, CO,
shown in Fig. 3. Correlation coefficients were found to be 0.9475, 0.997 HC and smoke. RMSE values are 6.87 ppm, 0.0064%, 9.02 ppm and
and 0.94875 for training, validation and testing, respectively. Correla- 0.70% and MRE values are 6.93%, 9.51%, 9.84% and 6.49% for NOx,
tion coefficient for all that includes training, validation and test is CO, HC and smoke, respectively.
0.95763. The fact that the correlation coefficient is close to 1 is an
indication of its high accuracy. It is observed from these values are very

4
M. Aydın, et al. Fuel 269 (2020) 117472

Fig. 4. Experimental results versus ANN results in terms of engine performance.

4. RSM based optimization applied in the area of engineering especially where it is difficult to
accurately represent the performance variables of multiple input
The main purpose in optimization is to achieve the best output re- parameters. In this study, RSM is used because there are more than one
sults by maximizing or minimizing some factors within the selected input and output parameters. The optimization is mainly aimed at
limits. There are many applications used for this purpose, which can be maximizing BTE while minimizing BSFC, EGT, NOx, CO, HC and smoke.
optimized by making common solution of inputs and outputs. RSM is a At the same time, it is aimed to create functional relations between the
mathematical statistics-based optimization application that can si- ANN estimated target parameters (BTE, EGT, BSFC, NOx, HC, CO and
multaneously model the input parameters and output parameters af- smoke) and design parameters. Design factors used to find optimum
fected by input parameters. In recent years, RSM has been extensively performance and emission parameters of engine are engine load (L),

5
M. Aydın, et al. Fuel 269 (2020) 117472

Fig. 5. Experimental results versus ANN results in terms of exhaust emissions.

6
M. Aydın, et al. Fuel 269 (2020) 117472

Fig. 6. Optimization setup.

factors for estimating the responses of BSFC, EGT, BTE, NOx, CO, HC
and smoke are presented by Eqs. (4)–(10), respectively:

BSFC = 2067 − 1, 470IP + 4, 80BR − 1, 536L + 0, 000576IP

∗ IP + 0, 0173BR ∗ BR + 0, 000527L ∗ L − 0, 00001IP
∗ BR + 0, 000213IP ∗ L − 0, 00275BR ∗ L (4)

EGT = 52, 8 + 0, 2741IP − 0, 157BR + 0, 1098L − 0, 000084IP

∗ IP − 0, 00430BR ∗ BR − 0, 000003L ∗ L − 0, 000186IP
∗ BR − 0, 000136IP ∗ L + 0, 000117BR ∗ L (5)

BTE = −2, 877 + 0, 02479IP − 0, 0067BR + 0, 01376L − 0, 000026IP

∗ IP − 0, 000045BR ∗ BR − 0, 000006L ∗ L − 0, 000019IP
∗ BR + 0, 000011IP ∗ L + 0, 000023BR ∗ L (6)

NOx = 98, 2 − 0, 1325IP − 1, 442BR + 0, 0006L + 0, 000204IP

∗ IP + 0, 01742BR ∗ BR + 0, 000047L ∗ L + 0, 000200IP
∗ BR + 0, 000001IP ∗ L + 0, 001205BR ∗ L (7)

CO = 0, 2574 − 0, 000472IP − 0, 001399BR − 0, 000096L + 0, 000000

IP ∗ IP + 0, 000012BR ∗ BR + 0, 000000L ∗ L + 0, 000000IP
∗ BR + 0, 000000IP ∗ L − 0, 000000BR ∗ L (8)

HC = 331, 1 − 0, 1444IP − 2, 899BR − 0, 2142L − 0, 000029IP

∗ IP + 0, 01394BR ∗ BR + 0, 000097L ∗ L + 0, 002363IP
∗ BR + 0, 000014IP ∗ L + 0, 000633BR ∗ L (9)

Smoke = 26, 09 − 0, 0827IP + 0, 328BR − 0, 0085L + 0, 000075IP

∗ IP − 0, 00093BR ∗ BR + 0, 000030L ∗ L − 0, 000466IP
∗ BR − 0, 000005IP ∗ L − 0, 000265BR ∗ L (10)

The RSM model was employed to find the best engine operation
Fig. 7. Optimization results. factors which are the engine load, biodiesel ratio and injection pressure
considering to the ANN model estimated values of BTE, EGT, BSFC,
biodiesel ratio (BR), and injection pressure (IP). Because the combus- NOx, HC, CO and smoke. Fig. 6 shows optimization setup that mini-
tion in an internal combustion engine is an extremely complicated case mizes or maximizes selected parameters according to the upper and
including non-linear relationship between the variables, a second order lower limits used in the RSM model. The optimized parameters ob-
model was created to demonstrate the relation among the input vari- tained from the optimization setup are shown in Fig. 7. Three vertical
ables and the responses [34]. Second-degree equations related to design red lines shows the best amount of engine load, biodiesel ratio and
injection pressure according to output parameters, respectively. The

7
M. Aydın, et al. Fuel 269 (2020) 117472

Table 3
Validation tests for optimized and experimental values.
Biodiesel Ratio (%) Load (W) I. Pressure (bar) Value BTE (%) EGT (℃) BSFC (g/kWh)

32 816 470 Optimized 14.54 184 783.95

Experimental 14.36 188 771.93
Error (%) 1.25 2.13 1.56

CO (%) HC (ppm) NOx (ppm) Smoke (%)

Optimized 0.079 120.30 120.05 9.95

Experimental 0.090 129.40 120 10.45
Error (%) 7.99 7.03 0.04 4.83

Optimized Experimental Optimized Experimental

Optimized Experimental
Fig. 8. Comparison of RSM optimized results and experimental results in terms of engine performance parameters.

optimum operation factors were obtained as 816-W engine load, 32% highest error proportion of 7.99% and lowest error proportion of 0.04%
biodiesel ratio and 470 bar injection pressure. Corresponding to op- were achieved with formula (11). The overall engine performance and
timum operating factors, best responses are 14.54% of BTE, 184.14 °C exhaust emission error rate is less than 8%. Clearly, the values opti-
of EGT, 783.95 g/kWh of BSFC, 120.05 ppm of NOx, 120.30 ppm of HC, mized by RSM are meaningful.
0.079% of CO and 9.95% of smoke. These RSM results represent that
Experimental result − Optimized result
performance parameters and exhaust emissions were significantly af- Percentage of error = × 100
fected by engine load, biodiesel ratio and injection pressure. Experimental result
A validation is required to assess the usability of the applied (11)
method. Hence, an experimental study was carried out with values
Comparison of RSM optimized results and actual outcomes re-
close to the optimized engine operating parameters obtained from RSM.
garding the performance and emission parameters of engine are pre-
The confirmation of the RSM was performed by comparing the opti-
sented in Figs. 8 and 9, respectively. The experimental outcomes and
mized results with actual results which is displayed in Table 3. The
the RSM results show close values in both engine performance and

8
M. Aydın, et al. Fuel 269 (2020) 117472

Optimized Experimental Optimized Experimental

Optimized Experimental Optimized Experimental

Fig. 9. Comparison of RSM optimized results and experimental results in terms of engine exhaust emissions.

Table 4 desirability was found to be 0.8539, which is much nearer to 1. This

Evaluation of performance parameters of two techniques for engine perfor- result shows that RSM models can be used effectively to discover the
mance. best engine operating factors with best engine performance and exhaust
BTE (%) EGT (℃) BSFC (g/kWh) emission parameters.
Evaluation of performance parameters of two techniques for engine
ANN RSM ANN RSM ANN RSM performance and emissions are shown in Tables 4 and 5. By comparing
MRE, RMSE and R2 values, the performance of the ANN model was
MRE 5.92% 2.37% 3.49% 2.41% 5.44% 3.93%
RMSE 0.55% 0.19% 4.18 °C 2.80 °C 25.65 g/kWh 18.615 g/kWh generally higher than that of the RSM model. While RSM model for BTE
R2 0.914 0.985 0.981 0.925 0.963 0.9398 and NOx was more successful than ANN, for other performance and
emission parameters ANN was more successful.

exhaust emissions. The combined desirability (D) indicates that how the 5. Conclusion
input factors optimize a set of overall targets. In an optimization, the
proximity of combined desirability to 1 is an indication that the per- The first aim of this study was to reduce the number of experiments
formance of the optimization is good. In this study, the combined to determine the engine parameters by creating an ANN model. The

Table 5
Evaluation of performance parameters of two techniques for emissions.
CO (%) HC (ppm) NOx (ppm) Smoke (%)

ANN RSM ANN RSM ANN RSM ANN RSM

MRE 9.51% 10.12% 9.84% 9.98% 6.93% 6.82% 6.47% 7.80%

RMSE 0.0064% 0.0095% 9.02 ppm 8.98 ppm 6.87 ppm 5.88 ppm 0.70% 1.18%
R2 0.866 0.782 0.892 0.837 0.977 0.984 0.986 0.932

9
M. Aydın, et al. Fuel 269 (2020) 117472

other aim of this research was to find the optimum operating para- biodiesel and diesel fuels for a diesel engine. J Environ Manage 2019;243:218–26.
meters regarding performance and exhaust emissions of a diesel engine [10] Chandran D. Compatibility of diesel engine materials with biodiesel fuel. Renew
Energy 2020;147:89–99.
based on various engine load, biodiesel ratio and injection pressure [11] Simsek S. Effects of biodiesel obtained from Canola, sefflower oils and waste oils on
with the aim of minimize BSFC, EGT, NOx, HC, CO and smoke as well as the engine performance and exhaust emissions. Fuel 2020;265.
to maximize BTE. [12] Elkelawy M, Bastawissi H, Esmaeil KK, Radwan AM, Panchal H, Sadasivuni KK,
et al. Experimental studies on the biodiesel production parameters optimization of
An ANN model was performed to estimate the performance and sunflower and soybean oil mixture and DI engine combustion, performance, and
emission parameters of the diesel engine fueled by biodiesel/diesel fuel emission analysis fueled with diesel/biodiesel blends. Fuel 2019;255.
mixtures with different proportion. Approximately 75% of the overall [13] Chauhan BS, Kumar N, Cho HM. A study on the performance and emission of a
diesel engine fueled with Jatropha biodiesel oil and its blends. Energy
test data was selected for the training part and 25% for the test part. 2012;37(1):616–22.
Biodiesel ratio, injection pressure and engine load were selected being [14] Ogunkunle Oyetola, Ahmed NA. Performance evaluation of a diesel engine using
input parameters while BSFC, EGT, BTE, NOx, HC, CO and smoke were blends of optimized yields of sand apple (Parinari polyandra) oil biodiesel. Renew
Energy 2019;134:1320–31.
selected as output parameters on ANN model. The R2 values were found
[15] Gharehghani A, Pourrahmani H. Performance evaluation of diesel engines (PEDE)
as 0.9144, 0.9626, 0.9811, 0.9771, 0.8663, 0.8917 and 0.9858, RMSE for a diesel-biodiesel fueled CI engine using nano-particles additive. Energy Convers
values were found as 0.55%, 25.65 g/kWh, 4.18 °C, 6.87 ppm, Manage 2019;198.
0.0064%, 9.02 ppm and 0.70%, MRE values were found as 5.92%, [16] Doğan O, Çelik MB, Özdalyan B. The effect of tire derived fuel/diesel fuel blends
utilization on diesel engine performance and emissions. Fuel 2012;95:340–6.
5.44%, 3.49%, 6.93%, 9.51%, 9.84% and 6.49% for BTE, BSFC, EGT, [17] Emiroğlu AO, Şen M. Combustion, performance and exhaust emission character-
NOx, CO, HC and smoke, respectively. When ANN values and experi- izations of a diesel engine operating with a ternary blend (alcohol-biodiesel-diesel
mental results are evaluated, it is understood that performance para- fuel). Appl Therm Eng 2018;133:371–80.
[18] Nirmala N, Dawn SS, Harindra C. Analysis of performance and emission char-
meters and exhaust emissions of diesel engines powered with biodiesel/ acteristics of Waste cooking oil and Chlorella variabilis MK039712.1 biodiesel
diesel mixtures can be modeled with ANN. ANN showed that it is a blends in a single cylinder, four strokes diesel engine. Renew Energy
suitable method for estimating engine parameters since R2 values are so 2020;147:284–92.
[19] Mehra RK, Duan H, Luo S, Rao A, Ma F. Experimental and artificial neural network
close to 1 and RMSE and MRE values are quite low. (ANN) study of hydrogen enriched compressed natural gas (HCNG) engine under
In addition, RSM is applied to obtain the best engine performance various ignition timings and excess air ratios. Appl Energy 2018;228:736–54.
values and exhaust emissions. The RSM outcomes show that, the 32% [20] Krishnamoorthi M, Malayalamurthi R, Shameer PM. RSM based optimization of
performance and emission characteristics of DI compression ignition engine fuelled
biodiesel ratio, 816-W engine load and 470 bar injection pressure are with diesel/aegle marmelos oil/diethyl ether blends at varying compression ratio,
the optimum engine operating factors. The optimized values with op- injection pressure and injection timing. Fuel 2018;221:283–97.
timum operating parameters were determined 14.54%, 184.14 °C, [21] Hosseini SH, Alisaraei AT, Ghobadian B, Mayvan AA. Artificial neural network
modeling of performance, emission, and vibration of a CI engine using alumina
783.95 g/kWh, 120.05 ppm, 120.30 ppm, 0.079% and 9.95% for BTE,
nano-catalyst added to diesel-biodiesel blends. Renew Energy 2019.
EGT, BSFC, NOx, HC, CO and smoke, respectively. Moreover, these [22] Baranitharan P, Ramesh K, Sakthivel R. Measurement of performance and emission
optimum results of engine load, injection pressure and biodiesel ratio distinctiveness of Aegle marmelos seed cake pyrolysis oil/diesel/TBHQ opus pow-
combinations were tested in the same engine. According to tests, results ered in a DI diesel engine using ANN and RSM. Measurement 2019;144:366–80.
[23] Krishnamoorthi M, Malayalamurthi R, Sakthivel R. Optimization of compression
were found as 14.36%, 188 °C, 771.93 g/kWh, 120 ppm, 129 ppm, ignition engine fueled with diesel – chaulmoogra oil – diethyl ether blend with
0.090% and 10.45% for BTE, EGT, BSFC, NOx, HC, CO and smoke, engine parameters and exhaust gas recirculation. Renew Energy
respectively. These test results and optimized results are relatively close 2019;134:579–602.
[24] Uslu S, Celik MB. Prediction of engine emissions and performance with artificial
to each other. Therefore, this study can support the engine producer neural networks in a single cylinder diesel engine using diethyl ether. Eng Sci
and researchers to predict and optimize the optimum operating para- Technol 2018;21(6).
meters for advanced performance and emission values with fewer ex- [25] Bencheikh K, Atabani AE, Shobana S, Mohammed MN, Uğuz G, Arpa O, et al. Fuels
properties, characterizations and engine and emission performance analyses of
periments. ternary waste cooking oil biodiesel–diesel–propanol blends. Sustain Energy Technol
Assessments 2019;35:321–34.
CRediT authorship contribution statement [26] Niu X, Yang C, Wang H, Wang Y. Investigation of ANN and SVM based on limited
samples for performance and emissions prediction of a CRDI-assisted marine diesel
engine. Appl Therm Eng 2016.
Mustafa Aydın: Conceptualization, Investigation. Samet Uslu: [27] Najafi G, Ghobadian B, Tavakoli T, Buttsworth DR, Yusaf TF, Faizollahnejad M.
Conceptualization, Methodology, Software, Writing - original draft, Performance and exhaust emissions of a gasoline engine with ethanol blended ga-
soline fuels using artificial neural network. Appl Energy 2009;86(5):630–9.
Writing - review & editing. M. Bahattin Çelik: Supervision.
[28] Yap WK, Ho T, Karri V. Exhaust emissions control and engine parameters optimi-
zation using artificial neural network virtual sensors for a hydrogen-powered ve-
References hicle. Int J Hydrogen Energy 2012;37(10):8704–15.
[29] Deb M, Majumder P, Majumder A, Roy S, Banerjee R. Application of artificial in-
telligence (AI) in characterization of the performance–emission profile of a single
[1] Goga C, Chauhan BS, Mahla SK, Cho HM. Performance and emission characteristics cylinder CI engine operating with hydrogen in dual fuel mode: an ANN approach
of diesel engine fueled with rice bran biodiesel and n-butanol. Energy Rep with fuzzy-logic based topology optimization. Int J Hydrogen Energy
2019;5:78–83. 2016;41(32):14330–50.
[2] Fayyazbakhsh A, Pirouzfar V. Investigating the influence of additives-fuel on diesel [30] Javed S, Murthy YVVS, Baig RU, Rao DP. Development of ANN model for prediction
engine performance and emissions: Analytical modeling and experimental valida- of performance and emission characteristics of hydrogen dual fueled diesel engine
tion. Fuel 2016;171:167–77. with Jatropha Methyl Ester biodiesel blends. J Nat Gas Sci Eng 2015;26:549–57.
[3] Şanlı BG, Uludamar E, Özcanlı M. Evaluation of energetic-exergetic and sustain- [31] Chakraborty A, Roy S, Banerjee R. An experimental based ANN approach in map-
ability parameters of biodiesel fuels produced from palm oil and opium poppy oil as ping performance-emission characteristics of a diesel engine operating in dual-fuel
alternative fuels in diesel engines. Fuel 2019;258. mode with LPG. J Nat Gas Sci Eng 2016;28:15–30.
[4] Uslu S, Celik MB. Combustion and emission characteristics of isoamyl alcohol-ga- [32] Bhowmik S, Panua R, Debroy D, Paul A. Artificial neural network prediction of
soline blends in spark ignition engine. Fuel 2020;262. diesel engine performance and emission fueled with diesel–kerosene–ethanol
[5] Simsek S, Ozdalyan B. Improvements to the composition of fusel oil and analysis of blends: a fuzzy-based optimization. J Energy Resour Technol 2017;139(4).
the effects of fusel oil-gasoline blends on a spark-ignited (SI) engine’s performance [33] Baruah D, Baruah DC, Hazarika MK. Artificial neural network based modeling of
and emissions. Energies 2018;11(3):625. biomass gasification in fixed bed downdraft gasifiers. Biomass Bioenergy
[6] Karikalan L, Baskar JS, Venugopal S. Analyzing the influence of varied fuel injection 2017;98:264–71.
pressure on diesel engine fueled with Karanja biodiesel. Mater Today: Proc 2019. [34] Uslu S, Celik MB. Performance and exhaust emission prediction of a SI engine fueled
[7] Scaldaferri CA, Pasa VMD. Production of jet fuel and green diesel range biohy- with I-amyl alcohol-gasoline blends: an ANN coupled RSM based optimization. Fuel
drocarbons by hydroprocessing of soybean oil over niobium phosphate catalyst. 2020;265:116922.
Fuel 2019;245:458–66. [35] Simsek S, Ozdalyan B, Saygın H. Improvement of the Properties of Sugar Factory
[8] Venu H, Subramani L, Raju VD. Emission reduction in a DI diesel engine using Fusel Oil Waste and Investigation of its Effect on the Performance and Emissions of
exhaust gas recirculation (EGR) of palm biodiesel blended with TiO2 nano ad- Spark Ignition Engine. Bioresources 2019;14(1):440–52. https://doi.org/10.15376/
ditives. Renew Energy 2019;140:245–63. biores.14.1.440-452. (in press).
[9] Yildiz İ, Açıkkalp E, Caliskan H, Mori K. Environmental pollution cost analyses of

10