SETC2015
JSAE 20159710 / SAE 2015-32-9710
Engine Performance, Emission and Combustion in Common Rail
Turbocharged Diesel Engine from Jatropha Curcas Using Artificial
Neural Network
A.S.Silitongaa,b
*, H.H.Masjukia, Hwai Chyuan Ong
a**,
F. Kusumoa,c
, H.G. Howa, Y.H. Teoh
a,d, T.M.I. Mahlia
e
aDepartment of Mechanical Engineering, Faculty of Engineering, University of Malaya,50603 Kuala Lumpur, Malaysia
bDepartment of Mechanical Engineering, Medan State Polytechnic,20155 Medan, Indonesia
cCenter for Artificial Intelligence Technology (CAIT), Faculty of Information Science and Technology Universiti Kebangsaan
Malaysia, Bangi, Selangor, 43600, Malaysia dSchool of Mechanical Engineering, Universiti Sains Malaysia, Engineering Campus, 14300 Nibong Tebal, Penang, Malaysia
eDepartment of Mechanical Engineering, Universiti Tenaga Nasional, 43000 Kajang, Selangor, Malaysia
Copyright © 2015 SAE Japan and Copyright © 2015 SAE International
ABSTRACT
This paper investigates the performance, emission and
combustion of a four cylinder common-rail turbocharged
diesel engine using jatropha curcas biodiesel blends (JCB).
The test was performed with various ratios of jatropha
curcas methyl ester (JCME) in the blends (JCB10, JCB20,
JCB30, and JCB50). An artificial neural networks (ANN)
model based on standard back-propagation algorithm was
used to predict combustion, performance and emissions
characteristics of the engine using MATLAB. To acquire
data for training and testing of the proposed ANN, the
different engine speeds (15003500 rpm) was selected as
the input parameter, whereas combustion, performance and
emissions were chosen as the output parameters for ANN
modeling of a common-rail turbocharged diesel engine. The
performance, emissions and combustion of the ANN were
validated by comparing the prediction dataset with the
experimental results. The results show that the correlation
coefficient was successfully controlled within the range
0.9798–0.9999 for the ANN model and test data. The value
of MAPE (Mean Absolute Percentage Error) was within the
range 1.2373–6.4217 and the Root Mean Square (RSME)
value was below 0.05 by the model, which is acceptable.
This study shows that modeling techniques as an approach
in alternative energy can give improvement advantage of
reliability in the prediction of performance and emission of
internal combustion engines.
INTRODUCTION
Researchers nowadays are focusing on the efforts seeking
renewable energy resources due to limited and increasing
prices of fossil fuels, as well as the harm impacts on the
environment including toxic emissions, GHG and global
warming. Alternative fuel is gaining its popularity for its
ability to solve the issues since it discharges low levels of
pollutants in both gaseous and particulates. As one of
alternative fuels, biodiesel is a renewable and eco-friendly
biofuel. It is regarded as biodegradable with higher cetane
number, no aromatics, very low sulfur, and more than 10%
oxygen compared to diesel fuel [1, 2]. Moreover, the
biodiesel emission characteristics such as carbon monoxide
(CO), hydrocarbons (HC) and particulate matter (PM) are
lesser in the exhaust gas, although NOx emission of
biodiesel is higher due to combustion characteristic [3, 4].
The used of blending biodiesel with diesel fuels have been
widely implemented for internal combustion engine. The
blends are preferred to be used because their availability,
competitive prices and comply with ASTM D7467 standard
[5].
In the practice of biodiesel production, the exploitation of
jatropha curcas as the feedstock for biodiesel is widely
practiced due to its beneficial effects in reducing the fossil
diesel consumption and exhaust emissions [6]. Sahoo et al.
[7] tested jatropha biodiesel-diesel blends (JB0, JB20, JB50
and JB100) in a water-cooled three cylinder tractor engine.
Their results showed that the use of 20% of biodiesel in the
blend improves the brake power for 0.09–2.64%, and
improvement of 0.05–3.8% for 50% biodiesel blend
compared to diesel fuel alone. Besides, the authors also
claimed that the use of the blends also reduces hydrocarbon
(HC) and particulate matter (PM). In another study, El-
Kasaby et al. [8] found that 10% Jatropha biodiesel (JB10)
has the highest brake thermal efficiency compared to the rest
of the biodiesel blended fuels (JB0, JB20, JB30 and JB50),
while JB30 and JB50 produced the lowest CO
concentration. Similarly, Chauhan et al. [9] studied that
preheated Jatropha biodiesel-diesel blend fuel decreases
CO, HC and smoke opacity. Ganapathy et al. [10] tested the
combustion, performance and emission on a single-cylinder,
Greaves Cotton model GL 400 II A, diesel engine. In their
study, Jatropha biodiesel reduced CO, HC and smoke with
heat release rate at 340 oCA, and NOx emission was lower at
350 oCA. Other researchers also had performed experiment
engine performance and emissions on diesel engine. It was
also found that by blending biodiesel with regular diesel fuel
can help in reducing the exhaust emissions without
significantly sacrificing engine’s power and economy [11,
12]. These are the reasons why the use of biodiesel instead
of diesel attracted numerous researches. Another advantage
SETC2015
are including such as lower fuel costs and no engine
modification required.
ANN is computational elements (neurons) working to solve
specific problems and functions. ANN is formed by
interconnecting a group of relatively simple processing
elements called neurons to predict multiple output variables
[13]. The application of ANN for modeling performance,
emission and combustion parameters is the ability to model
a multivariate problem without making complex
dependency assumptions among prediction and
experimental data. Moreover, it is trained to predict
performance, emission and combustion characteristics of
internal combustion engine [14, 15]. ANN can predict those
variables by simulating the results from the training based
on experimental data and validation by the independent data
[15, 16]. Besides, the aim of ANN is to determine the
characterization of engine performance for different
operating conditions and fuel to reduce time and cost
consumed. Some researchers had reported the proficiency of
ANN to predict performance and emission in internal
combustion engine characteristics in their researches.
Ghobadian et al. [17] used ANN to predict brake power,
torque, specific fuel consumption and emissions exhaust of
diesel engine using waste cooking oil. The experimental
results revealed that waste cooking methyl ester-diesel
blends provide better engine performance and improved
emission characteristics. Shivakumar et al. [13] studied that
the performance and emissions can be affected by injection
timing on of a single cylinder with variable compression
ratio. Their result showed the mean relative error (MRE)
between ANN predicted values and experiment data is 8%
which is reliable and acceptable. Maheswari et al. [18]
investigated the performance of an CI engine fueled with a
karanja biodiesel blends. They found that the karanja
biodiesel can improve efficiency engine and reduce HC
emissions. The prediction using ANN was given a low root
mean square error (RSME) that is about 0.950.99 for
performance and emission characteristics.
Since the use of biodiesel in diesel engine has different
physical and chemical properties compared to the fossil
diesel fuel, this is will affect the performance and emissions
of the engine. However the experimental studies and
measurements found in the literatures conclude that the use
of biodiesel-diesel blends are capable as the alternative fuel
leads to the improvement of engine power. This paper is
aimed to compare combustion, engine performance and
emissions between the experimental data and the prediction
using ANN modeling.
METHODOLOGY
BIODIESEL PRODUCTION
The biodiesel used in this research was produced from crude
Jatropha curcas oil (CJCO) by using a laboratory-scale
reactor. A double jacketed reactor equipped with a reflux
condenser with the refrigerator cooling bath and water bath
was used to produce Jatropha curcas biodiesel. The
production used two-stage esterification transesterification
process. The water bath was heated and maintained at 55oC.
A constant stirring speed of 1000 rpm was maintained
throughout the reactions process. The molar ratio of
methanol to oil ratio was 9:1 and the catalyst concentration
of H2SO4 and KOH were 1.0 vol. % and 1.0 wt. % of oil
respectively. All of the processes were carried out for
duration of 2.5 h. At the end of transesterification process,
glycerol and methyl ester were transferred to a separation
funnel and allowed to settle down for 4 h, then methyl ester
was washed with distilled water for three times and dried
using anhydrous sodium sulfate (Na2SO4). Finally, the
methyl ester was filtered and evaporated using rotary
evaporator to obtain the pure biodiesel.
FUEL CHARACTERIZATION
Biodiesel quality was tested according to following the
biodiesel standards ASTM D6751 and EN 14214. ASTM
D7467 standard was used to characterize the biodiesel-diesel
blends. The fuel properties of CJCO, JCME and JCB10,
JCB20, JCB30, and JCB50 were examined for their
viscosity, density, acid value, calorific value, flash point,
cloud and pour point, oxidation stability, water content, etc.
Table 1 shows the summary of the equipment used to
analyze the properties of biodiesel in this study. Each
measurement was repeated for three times and the mean
value was calculated for each sample. After all, the JCME
properties were obtained and the results are compared to
other researchers [6, 9, 11]. The properties and composition
CJCO, the properties JCME and JCB are summarized in
Tables 25. It is observed from Table 2 that properties of
CJCO are quite similar to the results from other researchers
and Table 3 shows fatty acids content of CJCO. The CJCO
has a greater part of unsaturated acid which are 44.5% and
35.4% for oleic acid and linoleic acid respectively. The
unsaturation of oil depends upon the origin of the vegetable
oil and is described by the iodine value [15]. The iodine
value of CJCO obtained in this study was (Table 2) 84.6, as
the value is within the accepted range of 80–130 suggested
by ASTM D6751 [19]. The unsaturation fatty acid chains of
triglycerides is the main parameter of oxidation and
polymerization when vegetable oils are used to produce
biodiesel [6]. The properties of crude oil depend upon the
feedstock, specifically on its fatty acid content [20]. Finally,
the study of biodiesel quality in terms of raw material was
performed based on the reaction conditions and the crude oil
was then tested to achieve better quality of biodiesel.
Table 1: Summary of the equipment
Property Equipment
Kinematic
viscosity StabingerTM viscometer SVM 3000 (Anton Paar)
Density DM40 LiquiPhysics™ density meter (Mettler
Toledo)
Flash point NPM 440 Pensky-martens flash point tester
(Normalab)
Cloud and pour point
NTE 450 Cloud and pour point tester (Normalab)
Calorific value 6100EF Semi auto bomb calorimeter (Perr)
Acid number and Automation titration rondo 20 (Mettler Toledo)
SETC2015
iodine value
Conradsons carbon residue
(100 sample)
NMC 440 micro-Carbone conradson residue
tester (Normalab)
Copper strip
corrosion (3 h at
50◦C)
Seta copper corrosion bath 11300-0 (Stanhope-Seta)
Sulphur content (S 15 grade and S500
grade)
Multi EA 5000 (Analytical jena)
Oxidation stability
at 110oC 873 Rancimat (Metrohm)
Methanol content Agilent 7890 gas chromatograph (Agilent)
Water content 831 KF coulometer (Metrohm)
Carbon CE 440 CHN Elemental Analyzer (EIA)
Table 2: Properties of CJCO
Properties CJCO CJCO [7]
CJCO [13]
Kinematic viscosity at
40oC (mm
2/s)
32.83 52 38
Density at 15oC (kg/m
3) 912.3 921 921.0
Calorific value (MJ/kg) 38.295 39.584
Flash point (oC) 172.5 220 235
Acid value (mg KOH/g) 25.41 30 2.32
Iodine value (I2/100g) 84.6
Table 3: The fatty acid composition of CJCO
Fatty acid Chemical
structure (wt.%)
Lauric (dodecanoic) C12:0 0.1
Myristic (tetradecanoic) C14:0 0.1
Palmitic (hexadecanoic) C16:0 13.0
Palmitoleic (hexadecenoic) C16:1 0.7
Stearic (octadecanoic) C18:0 5.8
Oleic (octadecenoic) C18:1 44.5
Linoleic (octadecadienoic) C18:2 35.4
Linolenic (octadecatrienoic) C18:3 0.3
Arachidic (eicosanoic) C20:0 0.1
Table 4: The properties of JCME
Properties Unit JCME JCME
[12] JCME
[13]
Kinematic
viscosity mm
2/s
4.57 4.34 4.12
Density kg/m3
872.5 881.0
Flash point °C 158.5 130 162
Cloud point °C 1.0 4
Pour point °C 2.0 8
Calorific
value MJ/kg
40.204 39.594
Acid value mg
KOH/g
0.391 0.09
Iodine value g
I2/100g
109.06
CCR m/m 0.03
Sulphur
content ppm
8.016 10 8
Cooper
corrosion -
1a
Oxidation
stability h
10.12
Water
content %vol.
0.0021
Cetane
number -
58.2 52.7 48.13
Methanol
content % m/m
0.061
Carbon %wt. 72.5
Hydrogen %wt. 12.8
Oxygen %wt. 11.8
Table 5: The properties of JCB
Properties Unit JCB10 JCB20 JCB30 JCB50
Kinematic viscosity
mm2/s 3.56 3.67 3.81 4.41
Density kg/m3 832.9 836.6 838.6 852.4
Flash point ºC 80.5 81.5 81.5 85.5
Cloud point ºC 0.0 1.0 1.0 1.2
Pour point ºC 1.0 1.0 2.0 2.0
Calorific
value MJ/kg 44.581 43.481 42.517 40.432
Acid value mg KOH/g 0.115 0.118 0.121 0.285
Iodine
value I2/100 g 30.26 42.71 49.62 53.58
CCR 0.0741 0.0599 0.0571 0.0482
Sulphur
content mg/kg 355.624 280.476 190.573 186.374
Cooper
corrosion _ 1a 1a 1a 1a
Oxidation
stability hours 20.33 18.56 15.48 12.82
Water
content v/v 0.0011 0.0013 0.0015 0.0018
Cetane
number - 52.8 53.9 54.5 55.7
Methanol
content 0.011 0.015 0.018 0.020
Carbon 80.2 78.9 77.3 76.3
Hydrogen 13.3 13.1 13.0 12.9
Oxygen 6.8 7.2 9.5 10.3
SETC2015
Kinematic viscosity is defined as the resistance of liquid to
flow and that is one of the most important fuel features. It
affects the operation of fuel injection, blending formation
and combustion process [20]. The viscosity of jatropha
curcas biodiesel and blends is presented in Table 4. From
this study, the viscosity of JCME was 4.57 mm2/s which is
the highest compared to JCME viscosity from other studies,
which are 4.35 mm2/s [11] and 4.12 mm
2/s [12]. The
viscosity of JCME and JCB were in agreement with ASTM
D6751 (3.55.0 mm2/s) and ASTM D7467 (1.94.5 mm
2/s).
JCB10, JCB20, JCB30 and JCB50 did not vary significantly
with regard the viscosity and the results were as agreed with
the previous studies [6, 21], which also reported that the
increase in length of carbon chain and number of double
bonds can increase the kinematic viscosities of biodiesel.
Density is an important property of biodiesel that affects the
efficiency of atomization of fuel in the combustion system
in which the increase of the fuel density tends to increase the
amount of fuel injected to the combustion chamber [22].
The density values for the fuels from this study, JCME,
JCB10, JCB20, JCB30 and JCB50 are described in Table 5.
The results revealed that the densities of JCB10, JCB20,
JCB30 and JCB50 (832.9 kg/m3, 836.6 kg/m
3, 838.6 kg/m
3
and 852.4 kg/m3
respectively) were higher than diesel (822.2
kg/m3). This density value was matched with Alptekin and
Canakci [23] which reported that the blending up to 20% of
biodiesel with diesel was very close to diesel. Additionally,
they found that there is no significant difference of the
density of the blends with up to 20% of methyl esters. The
density JCB50 was within the recommended limits for
biodiesel diesel blends by ASTM D7467 standard of max.
858 kg/m3 [24].
Flash point is a measure of flammability of fuels and it is an
important safety criterion in transport and storage [25].
Observed from this study, flash point of JCME was not
lower than 120oC and is higher than diesel which 79.5
oC
(Table 4). This is consistent with the studies reported by
Abedin et al. [26] and Gendure et al. [27] who reported that
biodiesel has higher flash point than diesel fuel. This implies
that the main advantage of biodiesel-diesel blends is better
in storing and safer to handle for transportation.
Acid number provides an indication of the level of
corrosiveness and long-term stability of the fuel [28]. As
shown in Table 4, the acid values for JCME and JCB10,
JCB20, JCB30 and JCB50 from this study are fulfilled the
ASTM D6751 and EN 14214 standards, which is below 0.5
mg KOH/g after the transesterification and blending process.
The results from this study are also approved by Tan et al.
[11] and Chauhan et al. [12]. Moreover, the acid values of
JCME and JCB10, JCB20, JCB30 and JCB50 from this
study are within specification of ASTM D7467 standard
which specifies a maximum of 0.3 mg KOH/g.
In this study, calorific value of biodiesel was measured
based on EN 14214 (35 MJ/kg), and the result was in
compliance with the standard. Calorific values of JCME and
JCB10JCB50 were found to be very stable and lower than
diesel fuel (45.448 MJ/kg) due to the oxygen content in the
fuel. It was observed that the calorific value of JCME
(40.204 MJ/kg) was higher than the one presented by
Chauhan et al. (39.594 MJ/kg) [12]. In Table 5, the calorific
value of JCB10, JCB20, JCB30 and JCB50 are ranges from
40.432 to 44.581 MJ/kg, and it is noted that biodiesel-diesel
blends have good calorific value and closed to diesel fuel.
The higher oxygen content in biodiesel could improve the
combustion process and decrease its oxidation potential.
Oxidative stability is one of the major technical issues
affecting the quality of fuel. The content of polyunsaturated
fatty acid chains in biodiesel is a major component that
affects oxidation stability [29]. It was observed that the
oxidation stability values were varied between 12.05 to
20.33 h, while the value for JCME was 10.12, as displayed
in Table 4. These results are also in compliance with the
ASTM D6751 and EN 14214 standards, but lower than
diesel fuel (22.85 h).
The cloud point is the temperature at which wax is first
precipitated and becomes visible when the fuel is cooled
[30]. Meanwhile, the pour point is the temperature at which
the wax is out of the solution and is sufficient to gel the fuel
[25]. As seen in Tables 4-5, both cloud and pour points of
JCME and JCB resulted from this study are quite close to
diesel fuel. This is in regards of the fatty acid composition
of the various individual fatty esters depends on chain
length, hydroxyl group, degree of unsaturation, double
bonds and type of ester, hence the occurrence is because
CJCO has double bond which influences the low
temperature [31]. The cloud and pour point values were
observed higher than the values by Chauhan et al. [12], of
which 4.0oC for cloud point and 8.0
oC for pour point of
JCME. These points, the cloud and pour points are, highly
affected by the proportion of saturated fatty acid esters in
biodiesel [20].
Cetane number is one fuel property which affects the engine
parameters such as combustion, oxidation stability and
exhaust emissions [25]. The biodiesel and biodiesel-diesel
blend cetane number are shown in Table 5. The cetane
number of the fuels in this study was higher than diesel fuel
(49.7), value by Tan et al. [11] (52.7) and by Chauhan et. al.
[12] (48.13). Moreover, the resulted cetane value was in
agreement with the specification standard of biodiesel
(ASTM D6751), which min. 47 and biodiesel-diesel blends
(ASTM D7467) , which min. 40 [24, 25].
EXPERIMENTAL SET UP
The experimental engine to test the performance, emission
and combustion was a four-cylinder, common-rail
turbocharged diesel engine. The engine was coupled with a
150 kW eddy current engine dynamometer and equipped
with a KoBOLD fuel flow meter. Besides, K-type
thermocouple was installed to monitor the fuel flow to the
engine and ambient air temperature, emission exhaust and
lubricant oil. The detail specifications and flow diagram of
the engine test bed are given in Table 6, Fig. 1 and Fig. 2. A
Kistler 6058A piezoelectric sensor was used to measure in-
cylinder pressure was mounted in the head of the first
cylinder by means of a glow plug adapter. The signal of
cylinder pressure was acquired at 0.125 °CA (crank angle)
and the resolution was recorded for averaged a 100
consecutive cycles. The engine was run at 1500 rpm to 4000
rpm with interval of 500 rpm for full load (100% throttle
open). The tests were repeated three times keeping the same
SETC2015
operating conditions. During each experiment, the engine
was warmed up and reached stable condition for about 57
min. The emission exhaust and smoke opacity were
measured by an AVL DICOM 4000 5-gas analyzer, AVL Di
Smoke 4000 portable and Bacharach Analyzer 300 NSX gas
analyzer. The uncertainty measurement of the results are
given in Table 7.
Fig. 1 Schematic diagram of the experiment setup
1. Test engine
2. Eddy current dynamometer
3. Air mass sensor
4. Pressure sensor
5. Data acquisition system
6. Dynamometer controller
7. Bosch gas analyzer
8. AVL gas analyzer9. Smoke opacity meter
Fig. 2 Photo of engine tested
Table 6: Specifications of tested diesel engine
Engine Type Diesel, four strokes, turbocharged,
direct injection
Fuel system High-pressure common-rail
Number of
cylinders
4
Number of
valves per
cylinder
2
Bore 76.0 mm
Stroke 80.5 mm
Displacement 1461 cm3
Compression
Ratio
18.25 : 1
Maximum
power
48 kW @ 4000 rpm
Maximum
torque
160 Nm @ 2000 rpm
Catalytic
converter
No
Table 7: List of instruments and the range accuracy and
percentage uncertainties
Measurement Measurement
range Accuracy
%
Uncertainty
Load ±600 Nm ±0.1 Nm ±0.25
Speed 010,000
rpm ±1 rpm ±0.1
Time ±0.1 s ±0.2
Fuel flow
measurement 0.536 L/hr ±0.1 L/hr ±0.2
Air flow
measurement
0.25-7.83
kg/min
±0.11
kg/min ±2.3
CO 010% by
vol. ±0.001% ±1
NOx 05,000 ppm ±1 ppm ±0.6
Smoke 0100% ±0.1% ±1
EGT sensor 0900°C ±0.3°C ±0.1
Pressure
sensor 025,000
kPa ±10 kPa ±0.5
oCA encoder
012,000
rpm ±0.125° ±0.03
Brake power ±0.04 kW ±0.1
BSFC ±2.7
g/kWh ±1.1
ARTIFICIAL NEURAL NETWORKS (ANN)
Artificial neural network (ANN) is a mathematic nonlinear
computer algorithm that is inspired by the structure or
functional aspects of biological neural systems. ANN
systems are computational function composed of neurons
and are an alternative way to solve specific functions [16].
Moreover, ANN is a massively parallel distributed processor
made up of simple processing units that have a natural
tendency for storing experiential knowledge and making it
available for use [13]. The ANN application in engineering
field has the capability to investigate the characteristics of
performance, emissions and combustion such as fuel
consumption and length time of consumption which from
many different fuel blends and different operating
conditions [32, 33]. Another advantage of using ANN is its
ability to handle large and complex systems with many
interrelated parameters. Furthermore, ANN has ability to
ignore the excess data in large and complex systems with
multidimensional information domain [34]. In this study, an
experimental investigation of the performance, emissions
and combustion of the turbocharged diesel engine was
conducted for different blending ratios (JCB10, JCB20,
JCB30 and JCB50) at full load (100%) without any
modification to the diesel engine. Using experimental data
in this study, ANN model was developed to predict BSFC,
BTE, and EGT for the performance model. The emission
model consisted of the prediction of CO, CO2, NOx and
SETC2015
smoke opacity, while in the combustion model the
predictions were Pmax, CADPmax, RHORmax and
CADROHRmax. The input parameters were JCB percentage
(JCB10, JCB20, JCB30 and JCB50) and engine speed
(15004000 rpm) for the entire model. The application of
ANN for modeling the emission characteristics will develop
strong to predict accuracy of experimental data.
STATISTICAL EVALUATION OF OUTPUT
PARAMETERS
The correlation coefficient (i.e. R) and coefficient of
determination (i.e. R2) are to evaluate the prediction
performance of the proposed ANN [14, 34, 35]. Applying
the statistical evaluation of output parameters, the details of
the ANN setting are presented in Table 7, and the absolute
fraction of variance (R2) is defined as below [14]:
n
i ibk
n
i ibia
YY
YYR
1
2
1
2
2
)(
)(1 (1)
MSE, RMSE and MAPE are commonly used to evaluate the
statistical error in ANN performance [32, 36]. The error
described by RMSE and MAPE are defined, respectively as
follows [32]:
21
1
2)(1
n
iibia YY
nRMSE (2)
1001
1
nY
YYMAPE
n
iia
ibia (3)
Where ‘n’ is the number of pattern in the data set,‘ iaY ’ is
the actual output, ‘ ibY ’ is the predicted output value and
’ kY ’ is the average of the actual value [37, 38]. The
developed ANN model used for the performance of a MAPE
of 5% was taken as a limit for the output parameters.
MODELING AND DESIGN
ANN model contains of three layers which are: input layer,
some hidden layers and an output layer. The information
that set in the input will be sent to the neurons with a certain
weight to the hidden layers. The information from the input
data is processed with the activation functions of the hidden
layer to approximate the non-linear behavior of the input
data set. The summation function that is used to calculate
the net input of the cell is given as below [13] :
bxwNETn
j jji 1 (4)
Where n is the number of input, wj is the connecting line
weight, xj is the input signal and b is an external bias.
The weights modify the connection between layers in
training network continuously until the expected responses
provide. Target value is the network output compared,
calculated and modified where the connecting line weight
and external bias has been minimized. The target is to
minimize the average sum of square errors which is called
as MSE of the output, as follow [16] :
n
j ij ndndn
MSE1
)()(1
(5)
Where: jd represents the experiment data
id is the ANN output or predicted
n is the neuron number of the output data
When the MSE falls below a determined value, or the
maximum number of epochs is exceeded, the training
process is terminated. Then, this trained network can be
used for simulating the system outputs for the inputs that
have not been introduced before [37, 39]. The selection of
an appropriate MATLAB equation significantly affects
network performance as below [37] :
nensig
1
1)(log (6)
nnPurelin )( (7)
The input and output data points in this study were
first normalized given by expression [13] where Zi is the
input or output variable and 21 : is the ranges of the
transfer function [33].
minmax
min
121
ii
iii
zz
zzX (8)
The 30 experimental data were prepared for the training and
testing for ANN modeling. The learning algorithm called
back-propagation was applied for the single hidden layer.
Levenberg–Marquardt (LM) was used for the variants of the
algorithm, while the logsig and purelin transfer function
were normalized between (0, 1) for logsig and (1, 1) for
purelin. The architecture for all the models is shown in Fig.
3. The independent model was initially developed for
performance parameters, emission and combustion
characteristics of diesel engine. The ratio for ANN training
(83% 25 data) and testing (17% 5 data) from the
experimental data were randomly selected for the testing
data and training data, respectively.
Each presented output of the neural network was compared
to the desired output and an error was computed. The
statistical values for the performance, emissions and
combustion model are presented in Tables 810. These
errors were then fed back (back propagated) to the neural
network and used to adjust the weights such that the error
decreases with each iteration. The network was initially
SETC2015
trained with different number of neurons in the hidden layer.
By increasing or decreasing the number of neurons, then
MSE can reach the minimum value. The number of neurons
is the minimum MSE which selected as the optimum
number of neurons in the hidden layer.
Fig. 3 ANN architecture for (a) performance model, (b)
emission model and (c) combustion model
RESULT AND DISCUSSION
ENGINE PERFORMANCE PARAMETER
BRAKE SPECIFIC FUEL CONSUMPTION (BSFC)
The BSFC with speed variation for diesel fuel and different
volume percentages of JCB10, JCB20, JCB30 and JCB50
are presented in Fig. 4. The BSFC trend decreased at
medium speeds (2000 rpm) and increased to high speeds
(4000 rpm) for turbocharged engine. BSFC for JCB10
(208.825 g/kWh) and diesel fuel (215.130 g/kWh) were
lower compared to JCB20, JCB30 and JCB50 (225.925
g/kWh and 228.322 g/kWh and 237.944 g/kWh
respectively) at 2000 rpm. It is believed that the lower BSFC
for JCB10 and diesel fuel was due to the better viscosity and
calorific value they possessed than JCB20JCB50 at the
operation conditions performance at 2000 rpm. Besides,
density of biodiesel was found higher compared to diesel
fuel, and although fuel consumption are the same on volume
basis it resulted in higher BSFC in the case of
JCB20JCB50. Similar trends of inconsistent BSFC values
at high speed for various biodiesel blends are also reported
by other researchers [6, 40]. Oi et al. [41] reported that
increasing volume percentage of biodiesel will decrease
calorific value and it causes higher BSFC to produce the
same amount of energy in the engine compared to diesel
fuel. For instance, JCB10 tends to have lesser BSFC than
JCB20, JCB30 and JCB50.
[Fig. 4]
Fig. 4 The BSFC variation with diesel fuel and
JCB10JCB50 at full throttle
BRAKE THERMAL EFFICIENCY (BTE)
The brake thermal efficiency for diesel fuel and blends of
JCB10, JCB20, JCB30 and JCB50 are shown in Fig. 5. The
results show that BTE have similar trends for all fuels. It
was observed that the BTE of diesel fuel (38.28%) was
slightly higher than the obtained BTE for JCB10 (37.18%)
at 2000 rpm. BTE values of JCB20, JCB30 and JCB50 were
found as 36.49%, 36.12% and 36.34% respectively at 100%
load, which are also lower than BTE of diesel fuel. This is
due to higher kinematic viscosity and density of JCB20,
JCB30 and JCB50, which cause lower volatility of the
blends and hence results in a poor mixture formation and
lower brake torque output. However, this delay the fuel
atomization that heads to reduce fuel air mixing. This study
also observed that JCB10 gave better fuel combustion and
improved the thermal efficiency compared to JCB20, JCB30
and JCB50. This occurrence is agreed with Muralidharan et
al. [42] which reported that the high oxygen content and less
C/H ratio of the blends cause complete combustion which
extends the premixed combustion phase. With higher
viscosity and density in JCB20, JCB30 and JCB50, lower
BTE was resulted due to larger droplet size in the fuel spray.
Other researches have studied cylinder pressure traces to
determine the heat release and any other parameters that
may affect the quality of the fuel-air mixture [9, 26].
(a)
(b)
(c)
SETC2015
Fig. 5 The BTE variation with diesel fuel and
JCB10JCB50 at full throttle
EXHAUST GAS TEMPERATURE (EGT)
The variation of EGT with engine speed for diesel fuel and
different volume percentage of JCB10, JCB20, JCB30 and
JCB50 are shown in Fig. 6. As seen from this figure, the
EGT increases as the speed increases for all test fuels in
common-rail turbocharged engines. This is explained as the
fuel is combusted in the combustion chamber, the unit time
as well as the heat energy are increased to generate faster
engine speed. The highest EGT for diesel fuel and JCB10,
JCB20, JCB30 and JCB50 were observed 515.00oC,
518.74oC, 530.00
oC, 539.13
oC and 550.00
oC respectively at
4000 rpm. According to these values, the EGT was
increased accordingly at higher engine speeds. The EGT for
diesel fuel showed a fairly lower trend compared with JCB
due to the low volatility during ignition delay and the fuel
spray atomizes into small droplets and would lead excessive
smoke (incomplete combustion). Kumar et al. [43] reviewed
that biodiesel-diesel blends have higher EGT and can cause
increase of BSFC of the blends compared to diesel fuel.
Additionally, Usta et al. [44] agreed that biodiesel has longer
ignition delay and slower burning rate that causes the
combustion duration affects the EGT. This is due to the
biodiesel is made from vegetable oil and its lower heating
value are not adequately evaporated during the main
combustion phase and continue to burn in the late
combustion phase [45]. Moreover, Rizwanul Fattah et al.
[22] revealed that higher CN and poor atomization due to
higher viscosity causes the presence of unburnt fuels in the
premixed combustion phase. These unburnt portions
continue can lead to the higher exhaust temperature. In this
study, all JCBs were tested and showed that JCB10 has low
EGT compared to JCB20, JCB30 and JCB50 which is an
indication of good burning of fuel inside cylinder.
Fig. 6 The EGT variation with diesel fuel and
JCB10JCB50 at full throttle
EXHAUST EMISSIONS PARAMETER
CARBON MONOXIDE (CO)
Fig. 7 illustrates the variation of CO emissions with engine
speed for diesel fuel and JCB10, JCB20, JCB30 and JCB50.
The highest CO emission was by diesel fuel (0.360%), then
followed JCB10 (0.163%), JCB20 (0.139%), JCB30
(0.135%) and JCB50 (0.091%) at 1500 rpm. The results
show that CO emission for JCB10, JCB20, JCB30 and
JCB50 were reduced for an average of 54.80% compared to
diesel fuel. It is clear that CO emissions significantly
decrease when the JCME ratio is increased since JCB10,
JCB20, JCB30 and JCB50 have higher cetane number than
diesel fuel. Knothe [46] observed that higher cetane number
causes longer ignition delay time for the biodiesel due to
less volatility leads late premixed combustion and more fuel
early fuel injection. It is acceptable for biodiesel because
higher cetane number is related to the fuel-rich zones
formation, which reduces CO emissions. Moreover, CO
concentration rate increases at low speed due to poor
atomization and uneven distribution of fuel across to the
combustion chamber and can lead to oxygen deficiency and
incomplete combustion [4].
Fig. 7 The comparison of CO emissions for diesel and
JCB10JCB50 at full throttle
CARBON DIOXIDE (CO2)
In general, diesel fuel and biodiesel-diesel blends have
similar trends of which CO2 emissions decrease with an
increase in the biodiesel blends ratio and engine speed. Fig.
8 illustrates the highest CO2 emission which by diesel fuel
(13.70%), follow by JCB10 (13.41%), JCB20 (13.19%),
JCB30 (13.17%) and JCB50 (13.01%) at 1500 rpm.
However, there is a slight increase CO2 for JB10 (11.81%)
compared to diesel fuel (11.79%) at 2000 rpm, which the
occurrence was caused high cetane number and excess
oxygen content in JCB10 developed combustion quality and
decreased fuel delay period compared to diesel fuel. It is
also showed that the average CO2 emission for JCB20,
JCB30 and JCB50 were reduced of 3.94% than diesel fuel.
From the results, JCB10, JCB20, JCB30 and JCB50 can
replace diesel oil in combustion turbocharged engine
without major adjustments and it reduces CO2 emissions.
The reduction CO2 emission for JCB10, JCB20, JCB30 and
JCB50 can be explained by the increasing injection pressure
SETC2015
with constant displacement volume and biodiesel density is
higher than diesel fuel so the amount of fuel mass injected is
larger when the engine is operated with JCB10, JCB20,
JCB30 and JCB50. This is in agreement with Huang et al.
study [47]. They observed that high viscosity and density
cause difficulty for atomization with biodiesel at low engine
speed and it can lead to incomplete combustion which
affects as decreasing in CO2 emissions during the
combustion phase. Canakci [48] and Ozener et al. [3]
investigated engine performance and emission exhaust using
biodiesel-diesel blends. They stated that the increasing
percentage volume of biodiesel-diesel blends produces low
CO2 emissions, as it caused by the availability of oxygen
molecules in biodiesel-diesel. Moreover, in an
environmental view, the use of biodiesel-diesel blends
reduces CO2 emission and pollutant from transportation
sector [43, 49, 50].
Fig. 8 The comparison of CO2 emissions for diesel and
JCB10JCB50 at full throttle
NITROGEN OXIDES (NOX)
NOx emission for diesel fuel and the blends (JCB10, JCB20,
JCB30 and JCB50) are shown in Fig. 9. From the results,
the highest NOx emission was found at JCB50 (1877 ppm)
at 2500 rpm and the lowest was obtained from diesel fuel
(1200 ppm) at 1500 rpm. In average, the increase of NOx
emission of JCB10, JCB20, JCB30 and JCB50 is 12.20%
compared to diesel fuel. This is because JCB10, JCB20,
JCB30 and JCB50 have higher viscosity that reduces
injection timings in the combustion process and leads to
higher NOx emission. Higher density, viscosity, speed of
sound, and bulk modulus leads to early fuel injection can
cause more fuel early fuel injection which leads to early
combustion timing [51]. Therefore, advanced combustion
timing gives higher temperature and increased NOx. In
addition, the double bonded molecule presents in biodiesel
leads higher production of NOx from the biodiesel-diesel
blends compared to diesel fuel [8]. It is caused that biodiesel
have a slightly higher adiabatic flame temperature, which
tend to the increase in NOx production for biodiesel.
Kivevele et al. [50] reported that increasing in NOx emission
is varied by fuel injection parameters and air fuel ratio.
Jindal et al. [52] and Banapurmath et al. [53] reported that
higher injection pressure and longer ignition delay period
may facilitates more NOx emission since high injection
pressure tends to increase nitrogen.
Fig. 9 The comparison of NOx emissions for diesel and
JCB10JCB50 at full throttle
SMOKE OPACITY
The main problem faced by the diesel engine is smoke
opacity, and hence this study examined the smoke emission
level from diesel fuel, JCB10, JCB20, JCB30 and JCB50.
The smoke opacity values in this study were recorded as
7.39% (JCB10), 6.21% (JCB20), 5.85% (JCB30) and 4.37%
(JCB50), which were lower compared to diesel (8.81%) at
2000 rpm (Fig. 10). The average smoke opacity of JCB10,
JCB20, JCB30 and JCB50 was decreased of 37.21%
compared with diesel fuel. This is due to the sufficient
pressure and temperature that can improve complete
combustion and reduce smoke opacity from internal
combustion engines [54, 55]. In addition, the lower emission
of CO is related to the decrease of smoke opacity which
indicates excesses of the oxygen content in the JCB10,
JCB20, JCB30 and JCB50, so that they perform easier to be
burnt at higher temperature when the engine speed increase.
Karabektas [56] and Hwang et al. [49] reported that smoke
opacity and CO emission of biodiesel-diesel blends is lower
than diesel fuel, as it is caused by the higher oxygen content
and cetane number in the blends than diesel fuel only. This
emission is concerned since more smoke opacity can cause
longer ignition delay period and auto ignition capability
compared to diesel fuel. Ushakov et al. [57] observed that
engine performance using biodiesel in turbocharged engine
can reduce smoke opacity and CO emission. The reduction
of smoke opacity from turbocharged engine is as the result
of the delivery of oxygen into chamber to complete the
burning of the hydrocarbon and carbon particles in the fuel.
It is also proven that the increase air to the turbocharged
engine provides easiness in mixing of fuel-air in the
combustion chamber and complete combustion [48, 50].
SETC2015
Fig. 10 The comparison of smoke opacity for diesel and
JCB10JCB50 at full throttle
COMBUSTION PARAMETER AT FULL
THROTTLE
PEAK CYLINDER PRESSURE
The variations of the cylinder pressure and heat release rate
with crank angle (oCA) for diesel fuel and JCB10, JCB20,
JCB30 and JCB50 at different engine speed are plotted on
Fig. 11. The peak cylinder pressure of diesel fuel was higher
compared to JCB10, JCB20, JCB30 and JCB50 at 1500 rpm
and 2000 rpm. However, at 2500-3500 rpm, peak cylinder
pressure values for JCB10, JCB20, JCB30 and JCB50 were
close to diesel fuel; and peak cylinder pressure values for
JCB10, JCB20 and JCB30 at 4000 rpm were higher than
diesel fuel with JCB50 at the lowest since JCB50 has lower
calorific value and higher viscosity compared to the rest of
the tested fuels. At 1500 rpm, the peak cylinder pressure
values were recorded as 114.059 bar at 11oCA (diesel fuel),
103.318 bar at 12oCA (JCB10), 100.964 bar at 13
oCA
(JCB20), 103.140 bar at 13oCA (JCB30) and 97.907 bar at
13oCA (JCB50). From 2500 rpm to 3500 rpm it can be seen
the peak cylinder pressures for JCB10, JCB20, and JCB30
shared the same rising pattern with diesel fuel. A small
increase in peak gas pressure was observed at 4000 rpm as
130.786 bar at 8oCA (JCB10), 131.516 bar at 7
oCA (JCB20)
and 130.898 bar at 8oCA (JCB30). In contrary, decrease as
126.875 bar at 7oCA (JCB50) was occurred due to higher
viscosity content than other fuels which cause longer
ignition delay of blends and slower pressure rise than
JCB10, JCB20 and JCB30. Also, the reduction in ignition
delay is as the regard of the higher oxygen content in
JCB50. Since JCB50 has lower calorific value hence more
fuel needs to be burned in premixed combustion phase and
this may result in slightly lower peak pressure. The high
viscosity and low volatility of the blends result to poor
combustion and cause low peak pressure, as reported by
Buyukkaya [58]. Likewise, Ushakov et al. [57] mentioned
that physical fuel properties such as high density and
viscosity can affect spray characteristics of fuel atomization
and it will reduce combustion stage since it will longer
injection period for each operating cycle. The graph for
JCB10, JCB20 and JCB30 at 4000 rpm has advantage to the
blends in spray atomization and evaporation. These results
are in accordance with Siwale et al. [59] whereas B5 to B20
can improve the combustion process of diesel engines due to
the oxygen content of the blends. In addition, Devan and
Mahalakshmi [45] tested the poon oil based biodiesel and
reported that peak gas pressure of B20 was higher than
B100 due to the lesser viscosity and better spray formation.
Ozsezen et al. [60] confirmed that the oxygen content of the
blends increases the fuel mixing ratio in the cylinder
compared to the only diesel fuel and this condition might
lead to the high cylinder pressure in the combustion
chamber.
RATE HEAT RELEASE (ROHR)
The variations of rate heat release (ROHR) with crank angle
for diesel fuel and JCB10, JCB20, JCB30 and JCB50 at full
throttle (100% load) are shown in Fig. 11. The ignition
delay for diesel was shorter than JCB10, JCB20, JCB30 and
JCB50 at 1500 rpm, and ROHR was increased for JCB10 at
2500 rpm 4000 rpm due to the cetane number and excess
oxygen content as a consequence of the good ignition delay
and complete combustion stage (Table 4). Also, increasing
accumulation of fuel during delay period gave higher ROHR
when running with JCB10. ROHR of JCB10 was slightly
higher than diesel fuel. The obtain values of ROHR were
48.314 J/oCA (JCB10), 47.278 J/
oCA (JCB20) and 46.817
J/oCA (JCB30), which were higher compared to diesel fuel
(45.003 J/oCA) at 13
oCA for 2500 rpm. The premixed
combustion heat release was higher for JCB10, JCB20 and
JCB30 due to higher volatility and good blending with
diesel fuel and air. Long ignition delay of JCB10, JCB20
and JCB30 leads slower ignition delay due to greater fuel
accumulated in the combustion chamber at the time of
premixed burning phase. Besides, the availability of higher
oxygen content is to improve the combustion stage and to
burn fuel complete as well. The lowest ROHR occurred to
JCB50 of 44.602 J/oCA at 13
oCA for 2500 rpm due to the
lower calorific value of JCB50 during the combustion phase
as it reaches unsufficient stoichiometric ratio (excess of
fuel) at higher compression ratio. The low ROHR in JCB50
affects premixed combustion stage hence ignition delay was
shorter compared to JCB10, JCB20, JCB30 and diesel fuel
due to the chemical reactions and composition of biodiesel
at high temperature. Pullen et al. [61] reported that carbon
chain on biodiesel affects the degree of oxidation of the
biodiesel and it causes lesser ignition delay due to higher
cetane number. Sayin et al. [62] investigated that biodiesel-
diesel blends have lower ROHR than diesel fuel due to high
fuel viscosity, which causes deterioration in preparation of
airfuel mixture hence shorter ignition times for the
oxidized fuel.
SETC2015
(a) 1500 rpm
(b) 2000 rpm
(c) 2500 rpm
(d) 3000 rpm
(e) 3500 rpm
(f) 4000 rpm
Fig. 11 Cylinder pressure and ROHR curves for diesel fuel
and JCB at various engine speed at 100% load
MODEL PREDICTION USING ANN
MODEL PREDICTION OF PERFORMANCE
PARAMETERS
The statistical analysis for performance model revealed that
the developed ANN model had low RMSE content of
BSFC, BTE and EGT for the training and testing data (Table
8). For the training data, the resulted values of RMSE were
2.9178%, 2.4316%, and 0.658 for the BSFC, BTE and EGT
respectively, while for the test data were 3.942%, 2.717%,
and 0.977% respectively. Table 8 shows the values of
MAPE, which are 3.5259%, 2.4289% and 1.2373% for
BSFC, BTE and EGT of training data respectively, and
4.7929%, 3.0999%, 1.5311% for test data respectively. The
test data (MAPE values) was found that below the limit
model. The correlation coefficient (R) value for training and
data testing for all the data was close to 1, indicates that
there is a strong correlation in modeling BSFC, BTE and
EGT as depicted in Figs. 1214.
SETC2015
Table 8: Statistical value for the performance model
Variable
Training set Testing set
RMSE R2 MAPE
(%) RMSE R2
MAPE
(%)
BSFC
BTE
EGT
0.029178
0.024316
0.00658
0.9968
0.9922
0.9999
3.5259
2.4289
1.2373
0.03942
0.02717
0.00977
0.9954
0.9941
0.9986
4.7929
3.0999
1.5311
Table 9: Statistical value for the emission model
Variable
Training set Testing set
RMSE R2 MAPE
(%) RMSE R2
MAPE
(%)
NOx
CO
CO2
S.Opc
0.0251
0.0413
0.0463
0.0328
0.9936
0.9832
0.9798
0.9923
3.2789
5.3135
5.7324
4.1137
0.04942
0.02651
0.03235
0.04113
0.9871
0.9967
0.9956
0.9808
6.2131
3.3999
4.5338
6.1724
Table 10: Statistical value for the combustion model
Variable
Training set Testing set
RMSE R2
MAPE
(%) RMSE R
2
MAPE
(%)
Pmax
CADPmax
ROHRmax
CADROHRmax
0.0292
0.0302
0.0473
0.0364
0.9912
0.9937
0.9899
0.9917
5.1027
5.5207
6.1022
4.5723
0.0392
0.0342
0.0436
0.0311
0.9895
0.9905
0.9826
0.9932
6.0765
5.8752
6.4217
5.0233
Fig. 12 Comparison of ANN predicted BSFC with measured
data
Fig. 13 Comparison of ANN predicted BTE with measured
data
Fig. 14 Comparison of ANN predicted EGT with measured
data
SETC2015
MODEL PREDICTION OF EMISSION
PARAMETERS
Table 9 shows the values of RMSE for the training data,
which are 2.5102%, 4.1316%, 4,6371 % and 3,2842 % for
NOx, CO, CO2 and smoke opacity respectively, and 4.942%,
2.6511%, 3.235 and 4.113% respectively for the test data.
The observed MAPE values for the training data were
3.2789 %, 5.3135 %, 5,7324 % and 4,1137% for NOx, CO,
CO2 and smoke opacity respectively, and 6.2131%,
3.3999%, 4,5338% and 6.1724% respectively for the test
data. The increase and decrease percentage accuracy of
MAPE can be specified as error by the measurement of
emission parameters, and MAPE has the range error limits
(±5%) for the emissions. The accuracies of the prediction
for the variable performances are shown in Figs. 1518. The
correlation coefficient (R) for training and data testing for
all the data were found to be close to 1, which indicates a
good reliability in predicting the engine emission.
Fig. 15 Comparison of ANN predicted CO2 with measured
data
Fig. 16 Comparison of ANN predicted CO with measured
data
Fig. 17 Comparison of ANN predicted NOx with measured
data
SETC2015
Fig. 18 Comparison of ANN predicted smoke opacity with
measured data
MODEL PREDICTION OF COMBUSTION
PARAMETERS
The values of RMSE for the training and testing data are
shown in Table 10. Values of 2.9226%, 3.0217%, 4.7351%
and 3,6436% for Pmax, CADPmax, ROHRmax and
CADROHRmax respectively were observed for the training
data, while 3.921%, 3.417%, 4.361 and 3.105% respectively
were observed for the test data. The MAPE values for the
training data were 5.1027%, 5.5207%, 6.1022% and
4.5723% for Pmax, CADPmax, ROHRmax and CADROHRmax
respectively, while for test data were 6.0765%, 5.8752%,
6.4217% and 5.0233% respectively. The predictions for the
combustion parameters are shown in Figs. 1922. The
correlation coefficient (R) for training and data testing for
all the data were close to 1, which indicates a reliable
prediction.
Fig. 19 Comparison of ANN predicted Pmax with measured
data
Fig. 20 Comparison of ANN predicted CADPmax with
measured data
SETC2015
Fig. 21 Comparison of ANN predicted ROHRmax with
measured data
Fig. 22 Comparison of ANN predicted CADROHRmax with
measured data
CONCLUSION
The experimental study was conducted on a four cylinder
common-rail turbocharged diesel engine using Jatropha
curcas biodiesel and its blends with diesel fuel. Resulted
from this study, engine performance values when operated
with JCB10 were higher compared to diesel fuel and others
blends (JCB20, JCB30 and JCB50) at 2000 rpm. The BSFC
of JCB10 was found to be lesser as 208.825 g/kWh
compared to JCB20, JCB30, JCB50 and diesel fuel.
Moreover, the rate heat release of JCB10 was higher of
48.314 J/oCA compared to JCB20, JCB30, JCB50 and diesel
fuel, which concludes the suitability of Jatropha curcas
biodiesel as an alternative fuel in diesel engine.
The application of ANN in this study was developed to
predict engine performance, emission exhaust and
combustion in turbocharged engine based on a model of a
feed-forward back propagation network combine with
logsig/purelin. The best network architecture of ANN
developed the engine performance, emission and the
combustion were 2-8-3, 2-7-4 and 2-6-4 respectively. The
results show that the training algorithm of feed-forward
back-propagation was sufficient enough in predicting BSFC
and EGT outputs for various engine speeds and fuel blends
ratios. Additionally, R values were close to 1 for the entire
model. The RMSE for all the models were smaller than 0.05
and the values of MAPE were within the range 1.2373–
6.4217. The values of correlation coefficient showed a very
strong and positive correlation in both the predicted data and
experimental data. Therefore, the ANN can be applied to
estimate desired output engine parameters. Finally, the
utilization of ANN can be considered as a practical
technique to predict the engine parameters instead of
generalizing in a wide range of experimental studies. For the
further study, ANN can be utterly useful as a tool to
optimize engines using different feedstock in the future.
REFERENCES
[1] Um S, Park SW. Numerical Study on Combustion and
Emission Characteristics of Homogeneous Charge
Compression Ignition Engines Fueled with Biodiesel.
Energy & Fuels. 2010;24(2):916-27.
[2] Sadeghinezhad E, Kazi SN, Badarudin A, Oon CS,
Zubir MNM, Mehrali M. A comprehensive review of
bio-diesel as alternative fuel for compression ignition
engines. Renewable and Sustainable Energy Reviews.
2013;28:410-24.
[3] Özener O, Yüksek L, Ergenç AT, Özkan M. Effects of
soybean biodiesel on a DI diesel engine performance,
emission and combustion characteristics. Fuel.
2014;115:875-83.
[4] Silitonga AS, Masjuki HH, Mahlia TMI, Ong HC,
Chong WT. Experimental study on performance and
exhaust emissions of a diesel engine fuelled with Ceiba
pentandra biodiesel blends. Energ Convers Manage.
2013;76:828-36.
[5] Atadashi IM, Aroua MK, Aziz AA. High quality
biodiesel and its diesel engine application: A review.
Renewable and Sustainable Energy Reviews.
2010;14(7):1999-2008.
SETC2015
[6] Mofijur M, Masjuki HH, Kalam MA, Atabani AE.
Evaluation of biodiesel blending, engine performance
and emissions characteristics of Jatropha curcas methyl
ester: Malaysian perspective. Energ. 2013;55:879-87.
[7] Sahoo PK, Das LM, Babu MKG, Arora P, Singh VP,
Kumar NR, Varyani TS. Comparative evaluation of
performance and emission characteristics of jatropha,
karanja and polanga based biodiesel as fuel in a tractor
engine. Fuel. 2009;88(9):1698-707.
[8] El-Kasaby M, Nemit-allah MA. Experimental
investigations of ignition delay period and performance
of a diesel engine operated with Jatropha oil biodiesel.
Alex Eng J. 2013;52(2):141-9.
[9] Chauhan BS, Kumar N, Du Jun Y, Lee KB.
Performance and emission study of preheated Jatropha
oil on medium capacity diesel engine. Energ.
2010;35(6):2484-92.
[10] Ganapathy T, Gakkhar RP, Murugesan K. Influence of
injection timing on performance, combustion and
emission characteristics of Jatropha biodiesel engine.
Appl Energ. 2011;88(12):4376-86.
[11] Tan P-q, Hu Z-y, Lou D-m, Li Z-j. Exhaust emissions
from a light-duty diesel engine with Jatropha biodiesel
fuel. Energ. 2012;39(1):356-62.
[12] 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. Energ.
2012;37(1):616-22.
[13] Shivakumar, Srinivasa Pai P, Shrinivasa Rao BR.
Artificial Neural Network based prediction of
performance and emission characteristics of a variable
compression ratio CI engine using WCO as a biodiesel
at different injection timings. Appl Energ.
2011;88(7):2344-54.
[14] Yusaf TF, Buttsworth DR, Saleh KH, Yousif BF. CNG-
diesel engine performance and exhaust emission
analysis with the aid of artificial neural network. Appl
Energ. 2010;87(5):1661-9.
[15] Atadashi IM, Aroua MK, Abdul Aziz AR, Sulaiman
NMN. Production of biodiesel using high free fatty
acid feedstocks. Renewable and Sustainable Energy
Reviews. 2012;16(5):3275-85.
[16] Canakci M, Ozsezen AN, Arcaklioglu E, Erdil A.
Prediction of performance and exhaust emissions of a
diesel engine fueled with biodiesel produced from
waste frying palm oil. Expert Syst Appl.
2009;36(5):9268-80.
[17] Ghobadian B, Rahimi H, Nikbakht AM, Najafi G,
Yusaf TF. Diesel engine performance and exhaust
emission analysis using waste cooking biodiesel fuel
with an artificial neural network. Renew Energ.
2009;34(4):976-82.
[18] Maheshwari N, Balaji C, Ramesh A. A nonlinear
regression based multi-objective optimization of
parameters based on experimental data from an IC
engine fueled with biodiesel blends. Biomass
Bioenerg. 2011;35(5):2171-83.
[19] Sahoo PK, Das LM. Process optimization for biodiesel
production from Jatropha, Karanja and Polanga oils.
Fuel. 2009;88(9):1588-94.
[20] Canakci M, Sanli H. Biodiesel production from various
feedstocks and their effects on the fuel properties. J Ind
Microbiol Biot. 2008;35(5):431-41.
[21] Ong HC, Masjuki HH, Mahlia TMI, Silitonga AS,
Chong WT, Yusaf T. Engine performance and
emissions using Jatropha curcas, Ceiba pentandra and
Calophyllum inophyllum biodiesel in a CI diesel
engine. Energ. 2014;69:427-45.
[22] Rizwanul Fattah IM, Masjuki HH, Liaquat AM, Ramli
R, Kalam MA, Riazuddin VN. Impact of various
biodiesel fuels obtained from edible and non-edible
oils on engine exhaust gas and noise emissions. Renew
Sust Energ Rev. 2013;18:552-67.
[23] Alptekin E, Canakci M. Characterization of the key fuel
properties of methyl ester–diesel fuel blends. Fuel.
2009;88(1):75-80.
[24] National Renewable Energy Laboratory. Biodiesel
Handling and Use Guide. Fourth ed. Oak Ridge: U.S.
Department of Energy; 2009.
[25] Atabani AE, Silitonga AS, Ong HC, Mahlia TMI,
Masjuki HH, Badruddin Irfan Anjum., Fayaz H. Non-
edible vegetable oils: A critical evaluation of oil
extraction, fatty acid compositions, biodiesel
production, characteristics, engine performance and
emissions production. Renew Sust Energ Rev.
2013;18:211-45.
[26] Abedin MJ, Masjuki HH, Kalam MA, Sanjid A,
Rahman SMA, Fattah IMR. Performance, emissions,
and heat losses of palm and jatropha biodiesel blends
in a diesel engine. Ind Crop Prod. 2014;59:96-104.
[27] Gandure J, Ketlogetswe C, Temu A. Fuel properties of
biodiesel produced from selected plant kernel oils
indigenous to Botswana: A comparative analysis.
Renew Energ. 2014;68:414-20.
[28] Fernando S, Karra P, Hernandez R, Jha SK. Effect of
incompletely converted soybean oil on biodiesel
quality. Energy. 2007;32(5):844-51.
[29] Pullen J, Saeed K. An overview of biodiesel oxidation
stability. Renew Sust Energ Rev. 2012;16(8):5924-50.
[30] Singh SP, Singh D. Biodiesel production through the
use of different sources and characterization of oils and
their esters as the substitute of diesel: A review. Renew
Sust Energ Rev. 2010;14(1):200-16.
[31] Refaat AA. Correlation between the chemical structure
of biodiesel and its physical properties. International
Journal of Environmental Science & Technology.
2009;6(4):677-94.
[32] Çay Y, Çiçek A, Kara F, Sağiroğlu S. Prediction of
engine performance for an alternative fuel using
artificial neural network. Appl Thermal Eng.
2012;37:217-25.
[33] Basheer IA, Hajmeer M. Artificial neural networks:
fundamentals, computing, design, and application. J
MicrobioI Meth. 2000;43(1):3-31.
[34] Kara Togun N, Baysec S. Prediction of torque and
specific fuel consumption of a gasoline engine by
using artificial neural networks. Appl Energ.
2010;87(1):349-55.
[35] Najafi G, Ghobadian B, Tavakoli T, Buttsworth DR,
Yusaf TF, Faizollahnejad M. Performance and exhaust
emissions of a gasoline engine with ethanol blended
gasoline fuels using artificial neural network. Appl
Energ. 2009;86(5):630-9.
[36] Mohamed Ismail H, Ng HK, Queck CW, Gan S.
Artificial neural networks modelling of engine-out
responses for a light-duty diesel engine fuelled with
biodiesel blends. Appl Energ. 2012;92:769-77.
SETC2015
[37] Sayin C, Ertunc HM, Hosoz M, Kilicaslan I, Canakci
M. Performance and exhaust emissions of a gasoline
engine using artificial neural network. Appl Thermal
Eng. 2007;27(1):46-54.
[38] Sözen A, Arcaklioǧlu E. Solar potential in Turkey.
Appl Energ. 2005;80(1):35-45.
[39] Traver M, Atkinson R, Atkinson C. Neural Network-
Based Diesel Engine Emissions Prediction Using In-
Cylinder Combustion Pressure. SAE International.
1999;1532:1–15.
[40] How HG, Masjuki HH, Kalam MA, Teoh YH. An
investigation of the engine performance, emissions and
combustion characteristics of coconut biodiesel in a
high-pressure common-rail diesel engine. Energ.
2014;69:749-59.
[41] Qi DH, Chen H, Geng LM, Bian YZ. Experimental
studies on the combustion characteristics and
performance of a direct injection engine fueled with
biodiesel/diesel blends. Energ Convers Manage.
2010;51(12):2985-92.
[42] Muralidharan K, Vasudevan D, Sheeba KN.
Performance, emission and combustion characteristics
of biodiesel fuelled variable compression ratio engine.
Energ. 2011;36(8):5385-93.
[43] Kumar N, Varun, Chauhan SR. Performance and
emission characteristics of biodiesel from different
origins: A review. Renew Sust Energ Rev.
2013;21:633-58.
[44] Usta N, Öztürk E, Can Ö, Conkur ES, Nas S, Çon AH,
Can AÇ, Topcu M. Combustion of biodiesel fuel
produced from hazelnut soapstock/waste sunflower oil
mixture in a Diesel engine. Energ Convers Manage.
2005;46(5):741-55.
[45] Devan PK, Mahalakshmi NV. Study of the
performance, emission and combustion characteristics
of a diesel engine using poon oil-based fuels. Fuel
Process Technol. 2009;90(4):513-9.
[46] Knothe G. A comprehensive evaluation of the cetane
numbers of fatty acid methyl esters. Fuel.
2014;119(0):6-13.
[47] Huang J, Wang Y, Qin J-b, Roskilly AP. Comparative
study of performance and emissions of a diesel engine
using Chinese pistache and jatropha biodiesel. Fuel
Process Technol. 2010;91(11):1761-7.
[48] Canakci M. Combustion characteristics of a
turbocharged DI compression ignition engine fueled
with petroleum diesel fuels and biodiesel. Bioresource
Technol. 2007;98(6):1167-75.
[49] Hwang J, Qi D, Jung Y, Bae C. Effect of injection
parameters on the combustion and emission
characteristics in a common-rail direct injection diesel
engine fueled with waste cooking oil biodiesel. Renew
Energ. 2014;63:9-17.
[50] Kivevele TT, Kristóf L, Bereczky Á, Mbarawa MM.
Engine performance, exhaust emissions and
combustion characteristics of a CI engine fuelled with
croton megalocarpus methyl ester with antioxidant.
Fuel. 2011;90(8):2782-9.
[51] Agarwal AK, Dhar A. Experimental investigations of
performance, emission and combustion characteristics
of Karanja oil blends fuelled DICI engine. Renew
Energ. 2013;52:283-91.
[52] Jindal S, Nandwana BP, Rathore NS, Vashistha V.
Experimental investigation of the effect of
compression ratio and injection pressure in a direct
injection diesel engine running on Jatropha methyl
ester. Appl Thermal Eng. 2010;30(5):442-8.
[53] Banapurmath NR, Tewari PG, Gaitonde VN.
Experimental investigations on performance and
emission characteristics of Honge oil biodiesel
(HOME) operated compression ignition engine. Renew
Energ. 2012;48:193-201.
[54] Hazar H, Aydin H. Performance and emission
evaluation of a CI engine fueled with preheated raw
rapeseed oil (RRO)–diesel blends. Appl Energ.
2010;87(3):786-90.
[55] Yuste AJ, Dorado MP. A Neural Network Approach to
Simulate Biodiesel Production from Waste Olive Oil.
Energy & Fuels. 2006;20(1):399-402.
[56] Karabektas M. The effects of turbocharger on the
performance and exhaust emissions of a diesel engine
fuelled with biodiesel. Renew Energ. 2009;34(4):989-
93.
[57] Ushakov S, Valland H, Æsøy V. Combustion and
emissions characteristics of fish oil fuel in a heavy-
duty diesel engine. Energ Convers Manage.
2013;65:228-38.
[58] Buyukkaya E. Effects of biodiesel on a DI diesel engine
performance, emission and combustion characteristics.
Fuel. 2010;89(10):3099-105.
[59] Siwale L, Kristóf L, Adam T, Bereczky A, Mbarawa
M, Penninger A, Kolesnikov A. Combustion and
emission characteristics of n-butanol/diesel fuel blend
in a turbo-charged compression ignition engine. Fuel.
2013;107(0):409-18.
[60] Ozsezen AN, Canakci M. Determination of
performance and combustion characteristics of a diesel
engine fueled with canola and waste palm oil methyl
esters. Energ Convers Manage. 2011;52(1):108-16.
[61] Pullen J, Saeed K. Factors affecting biodiesel engine
performance and exhaust emissions – Part II:
Experimental study. Energy. 2014;72:17-34.
[62] Sayin C, Gumus M, Canakci M. Effect of fuel injection
pressure on the injection, combustion and performance
characteristics of a DI diesel engine fueled with canola
oil methyl esters-diesel fuel blends. Biomass Bioenerg.
2012;46:435-46.
ACKNOWLEDGMENTS
The authors would like to acknowledge for the Ministry of
Education of Malaysia And The University Of Malaya,
Kuala Lumpur, Malaysia for the financial support under
High Impact Research Grant (HIRG) entitled clean diesel
technology for military and civilian transport vehicles
M.C/HIR/MOE/ENG/07 (D000007-16001), UMRG:
RP022A-13AET and FRGS: FP009-2014A.
DEFINITIONS/ABBREVIATIONS
Symbol Description
ANN Artificial neural network
BTE Brake thermal efficiency
BSFC Brake specific fuel consumption oCAD Crank angle distance
CJCO Crude Jatropha curcas oil
CO Carbon monoxide
SETC2015
CO2 Carbon dioxide
EGT Exhaust gas temperature
JCME Jatropha curcas methyl ester
JCB Jatropha curcas biodiesel blends
JCB10 Jatropha curcas biodiesel blends 10%
with diesel fuel
JCB20 Jatropha curcas biodiesel blends 20%
with diesel fuel
JCB30 Jatropha curcas biodiesel blends 30%
with diesel fuel
JCB50 Jatropha curcas biodiesel blends 50%
with diesel fuel
MSR Mean square error
NOx Nitrogen oxide
ROHR Rate heat release
TDC Top death center
RMSE Root mean square error
MRE Mean relative error
MSE Mean square error
R Correlation coefficient
MAPE Mean absolute percentage error