Performancel Engine
Performancel Engine
Performancel Engine
ENERGY
www.elsevier.com/locate/apenergy
a,*
Abstract
In this study, the applicabilities of Articial Neural Networks (ANNs) have been investigated
for the performance and exhaust-emission values of a diesel engine fueled with biodiesels from different feedstocks and petroleum diesel fuels. The engine performance and emissions characteristics
of two dierent petroleum diesel-fuels (No. 1 and No. 2), biodiesels (from soybean oil and yellow
grease), and their 20% blends with No. 2 diesel fuel were used as experimental results. The fuels were
tested at full load (100%) at 1400-rpm engine speed, where the engine torque was 257.6 Nm. To
train the network, the average molecular weight, net heat of combustion, specic gravity, kinematic
viscosity, C/H ratio and cetane number of each fuel are used as the input layer, while outputs are the
brake specic fuel-consumption, exhaust temperature, and exhaust emissions. The back-propagation learning algorithm with three dierent variants, single layer, and logistic sigmoid transfer function were used in the network. By using weights in the network, formulations have been given for
each output. The network has yielded R2 values of 0.99 and the mean % errors are smaller than 4.2
for the training data, while the R2 values are about 0.99 and the mean % errors are smaller than 5.5
for the test data. The performance and exhaust emissions from a diesel engine, using biodiesel
blends with No. 2 diesel fuel up to 20%, have been predicted using the ANN model.
2005 Elsevier Ltd. All rights reserved.
Keywords: Articial neural-network; Biodiesel; Engine performance; Exhaust emissions
Corresponding author. Tel.: +90 262 339 4031; fax: +90 262 305 8010.
E-mail addresses: mustafacanakci@hotmail.com, canakci@kou.edu.tr (M. Canakci).
0306-2619/$ - see front matter 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.apenergy.2005.05.003
595
Nomenclature
ANN
BSFC
BSO
BYG
CN
KV
LM
NH
R
RMS
R2
SCG
SG
SN
Texh
W
articial neural-network
brake specic fuel-consumption (g/kWh)
biodiesel soybean-oil
biodiesel yellow-grease
cetane number
kinematic viscosity (mm2/s)
LevenbergMarquardt
net heat of combustion (kJ/kg)
C/H ratio
root-mean-squared
fraction of variance
scaled conjugate gradient
specic gravity
smoke number
exhaust-manifold temperature (C)
average molecular-weight (kg/kmol)
1. Introduction
Global air-pollution is a serious problem. Much of this pollution is caused by the
use of fossil fuels for transportation. Therefore, engine manufacturers have designed
alternatively fueled engines and fuel systems, which provide sucient power while
staying within regulatory emission-limits. At the same time, a great deal of research
and development on internal-combustion engines has taken place not only in the design area but also in nding an appropriate fuel. Many researchers have concluded
that biodiesel holds promise as an alternative fuel for diesel engines, since its properties are very close to those of No. 2 diesel fuel [13]. Therefore, biodiesel can be
used in diesel engines with few or no modications. Diesel-fuel blends with biodiesel
have superior lubricity, which reduces wear and tear on the diesel engine and makes
the engine components last longer. Biodiesel mixes well with diesel fuel and stays
blended.
Biodiesel has a higher cetane number than petroleum diesel fuel, no aromatics,
and contains 1011% oxygen by weight. These characteristics of biodiesel
reduce the emissions of carbon monoxide (CO), hydrocarbons (HC), and particulate matter (PM) in the exhaust gas compared with diesel fuel [4]. However,
NOx emissions of biodiesel increase because of combustion and some fuel characteristics [1]. When a higher percentage of biodiesel is used in the diesel engine
during cold weather, it thickens more than diesel fuel and special systems may
be required. Equipment made before 1993 may have rubber seals in the fuel
596
2. Experimental work
In this study, experiments were performed on a John Deere 4276T, four-cylinder,
four-stroke, turbocharged direct-injection (DI) diesel-engine. The basic specications
of the engine are shown in Table 1. A 112 kW General Electric (Schenectady, NY)
model TLC 2544 direct-current dynamometer, assembled on the engine, was used.
An electronic scale and a stopwatch were used to measure the fuel ow rate, and
the intake airow rate was measured using a laminar-ow element. The fuels were
597
Table 1
Specications of test engine
Make and model
Motor type
Number of cylinders
Compression ratio
Bore and stroke
Connecting-rod length
Maximum engine power
Maximum torque
Type of injection pump
Fuel-injector holes
Fuel-injector opening pressure
Table 2
List of emission analyzers used in the tests
Analyzer
Smoke meter
NO/NOx analyzer
HC analyzer
CO2 analyzer
CO analyzer
O2 monitor
tested at full load (100%) at 1400-rpm engine speed, where the engine torque was
257.7 Nm. The fueling rate of the engine was adjusted to maintain this torque level
for all fuels.
The tests were performed under steady-state conditions. The fuels were tested in
random order and each test was repeated 3 times. The results of the three repetitions
were averaged to decrease the uncertainty. No. 1 diesel fuel, No. 2 diesel fuel, BSO,
BYG, and 20% biodiesel blends with No. 2 diesel fuel were used. The instruments
used to measure the engines exhaust-emissions are given in Table 2. Calibration
of each analyzer was done before each test. Using the appropriate calibration curve,
the measurement error for each analyzer was reduced to less than 2%, as recommended in the exhaust analyzer bench-manual. The test results obtained in the experimental study have been used to train and test the ANN.
598
R 1
j t j
oj 2
2
j oj
!
;
Mean% Error
599
1 X t j oj
100
;
j
p
tj
where t is the target value, o is the output value, and p is the pattern [6].
Experimental results for dierent fuels and biodiesel blends are used as the training and test data for the ANN. The experimentally-tested fuels are No. 2 Diesel,
BSO, BYG, No. 1 Diesel, 20% BSO and 20% BYG with No. 2 diesel-fuel. Dierent
networks were used in the ANN study. The dierence was only in the selection of the
test fuel. In other words, No. 2 Diesel, No. 1 Diesel, 20% BSO and 20% BYG fuels
were used as test data in dierent networks. However, ve fuels were used in the
learning layer and one fuel was also used in the test for all the networks. The
RMS, R2 and the mean error percentage values were used for comparison.
Table 3
Statistical values of predictions
Outputs
RMS
(training)
R2
(training)
Mean % error
(training)
RMS
(test)
R2
(test)
Mean % error
(test)
BSFC
SN
CO
CO2
HC
Texh
O2
NOx
0.004264
0.029216
0.008778
0.006779
0.003551
0.003558
0.005497
0.010851
0.999954
0.996702
0.999635
0.999913
0.9999
0.999976
0.999935
0.999659
0.573866
4.132617
1.505732
0.780031
0.808099
0.320963
0.662413
1.589002
0.000674
0.032582
0.016838
0.004808
0.005782
0.000993
0.006556
0.000657
0.999999
0.997051
0.998927
0.999964
0.999859
0.999998
0.999925
0.999999
0.100623
5.430333
3.27612
0.59898
1.185666
0.123447
0.866655
0.103291
600
Predicted
Actual
SN
BSFC (g/kWh)
1.2
250
225
0.8
0.4
200
0
1
3
Actual
Predicted
3
Actual
Predicted
850
CO2 (g/kWh)
0.6
CO (g/kWh)
Predicted
1.6
275
0.5
0.4
800
750
700
3
Actual
Predicted
3
Actual
Predicted
810
T ( oC )
HC (g/kWh)
0.5
0.35
0.2
795
780
3
Actual
Predicted
3
Actual
Predicted
570
NOx (g/kWh)
O2 (g/kWh)
21
560
550
18
15
1
;
1 e2.9038 F14.6293 F20.7157 F32.3663 F40.3416 F51.5021 F61.42 F70.2832
4
601
SN
1
;
1 e3.6081F10.2425F21.9101F311.1771F48.7307F53.9425F64.9152F75.0621
CO
1
;
1 e1.6786F11.1652F23.352F32.3968F41.9249F53.2227F61.1016F71.0821
CO2
HC
T
1
1
e3.3975F10.1471F21.4536F30.8451F44.8292F50.022F63.5789F72.9965
1
e2.4194F14.6647F22.7628F38.2077F40.8052F51.039F65.2902F71.8493
7
8
1
;
1 e2.7774F14.0336F21.5327F30.7189F44.0896F52.4595F61.5176F72.4029
1
;
1 e3.1106F12.1033F23.6928F30.1042F43.184F51.4274F61.116F74.2499
10
O2
NOx
1
1
e1.6005F11.1653F27.3322F31.042F40.1532F50.5913F61.6203F70.5812
11
1
;
1 eEi
12
C2
C3
C4
C5
C6
C7
1
2
3
4
5
6
7
0.6693
2.8581
2.6895
2.8897
5.9609
8.2284
2.7972
0.5495
3.5098
2.1628
2.6564
2.9266
1.9820
5.2028
0.7362
1.6790
2.5692
1.0565
5.7404
6.7795
3.7080
2.2348
0.1170
7.0595
1.9906
0.6876
3.3239
5.4424
3.3178
5.8240
0.7040
2.3271
6.0617
2.4814
2.2330
5.8244
2.9026
0.0845
1.8872
1.9154
2.5003
3.1103
9.4929
3.0292
2.1046
0.1816
3.4363
1.9834
5.8612
602
SN, CO, CO2, HC, Texh, O2 and NOx values need to be multiplied by 350, 1.5, 1,
1000, 1, 1000, 750 and 30, respectively.
The causes of the air pollution from diesel engines have been identied as smoke
and NOx in the exhaust emissions. Most products occur as a consequence of combustion processes inside the engine, and these unwanted emissions are exhausted into
the atmosphere. The amounts of emission products vary for dierent engines and depend on the operating conditions and fuel properties. As stated above, dierent fuels
are used in the test data. The network using 20% BSO test data results in the best
results. The network results have shown that the model can be used for predicting
the engine performance and emissions. Therefore, all output values are predicted
by using weight values for 1%, 2%, 5%, 10%, 15% BSO and BYG. Fig. 2 shows
the predicted values of BSFC, SN, CO, CO2, unburnt HC, Texh, O2 and NOx for
these compositions.
Fuel properties, such as specic gravity, viscosity and heating value inuence the
BSFC of a diesel engine. Therefore, the biodiesel blends eects on the BSFC were
calculated for both biodiesels. As seen in the gure, the predicted values for the
BSFC increase with the increasing amount of biodiesel in the blend. The increase
in BSFC for biodiesel is understandable, since both the BSO and BYG have lower
heating values than for No. 2 diesel fuel. The BSFC for the BYG is slightly higher
than that for the BSO due to the lower heating-value of the BYG. This result can
be clearly seen at higher ratios. The lower heating value requires that a larger
amount of fuel is injected into the combustion chamber to produce the same power.
In the literature, higher values of the BSFC were noted for biodiesel fuels [1,12,13].
The smoke level of the biodiesel blends, as seen in the gure, starts to decrease
even at 1% blends for both biodiesels. For all the predicted results, almost no dierence has been observed in the SNs between the two biodiesel blends. The researchers
have observed that the smoke levels of the biodiesels and their blends were signicantly lower than that for the diesel fuel.
Carbon monoxide (CO) in diesel engines is formed during the intermediate combustion stages. For fuel-rich mixtures in the spark-ignition engines, CO concentrations in the exhaust increase steadily. Diesel engines, however, always operate well
on the lean side of stoichiometric. Therefore CO emissions from the diesel engines
are usually low and most engine manufacturers meet CO regulations easily [14].
The CO emissions have decreased with increasing amounts of biodiesel in the blend.
The reduction of CO emissions is signicant when the blend percentage is 15 or higher. The gure shows that both biodiesels have similar CO reductions with increasing
amounts of biodiesel in the blends. In the literature, researchers [12,15] have concluded that biodiesel and blend lowered the CO emissions have benets. As seen
in the Figure, the CO2 emissions for the biodiesels were increasing with the increasing ratio of biodiesels in the blend. When compared with each other, the CO2 for the
BYG is slightly higher than that for the BSO due to the better combustion achieved.
This result can be clearly obtained at higher blend-ratios since the BYG has a higher
cetane number than that for the BSO. Higher cetane numbers indicate better combustion. However, the increases of the CO2 emissions are found not to be so
signicant.
603
1.10
234
1.08
1.06
1.04
230
1.02
SN
BSFC (g/kWh)
232
228
1.00
0.98
226
0.96
0.94
224
0.92
0.90
222
1% 2% 5% 10% 15% 1% 2%
BSO
564.5
807
564.0
563.5
O 2 (g/kWh)
Texh ( o C)
806
805
804
803
563.0
562.5
562.0
802
801
561.5
1%
2% 5% 10% 15% 1%
BSO
2% 5% 10% 15%
BYG
0.544
812
810
0.542
808
806
CO 2 (g/kWh)
0.540
CO (g/kWh)
5% 10% 15%
BYG
0.538
0.536
0.534
804
802
800
798
796
794
0.532
792
0.530
790
19.4
0.510
19.2
0.505
NO x (g/kWh)
HC (g/kWh)
19.0
0.500
0.495
0.490
18.8
18.6
18.4
0.485
18.2
18.0
0.480
1% 2% 5% 10% 15% 1% 2% 5% 10% 15%
BSO
BYG
604
5. Conclusions
The applicability of ANNs has been investigated for the performance and exhaust
emission values of a diesel engine fueled with biodiesels and petroleum diesel fuels.
To train the network, the fuel properties of each fuel are used as the input layer,
while the outputs are BSFC, Texh, and exhaust emissions. By using the back-propagation learning algorithm with three dierent variants, single layer, and logistic sigmoid transfer function, the weights of the network, formulations have been given for
each output. The network has yielded R2 values of 0.99 and the mean % errors are
smaller than 4.2 for the training data, while the R2 values are about 0.99 and the
mean % errors are smaller than 5.5 for the test data.
The results may easily be considered to be within the acceptable limits. The relationship between fuel properties and engine performance-emissions can be determined for dierent biodiesel blends by using the network. Therefore, the usage of
ANNs may be highly recommended to predict the engines performance and emissions instead of having to undertake complex and time-consuming experimental
studies.
605
References
[1] Canakci M, Van Gerpen JH. Comparison of engine performance and emissions for petroleum dieselfuel, yellow-grease biodiesel and soybean-oil biodiesel. Trans ASAE 2003;46:93744.
[2] Chang DYZ, Van Gerpen JH, Lee I, Johnson LA, Hammond EG, Marley SJ. Fuel properties and
emissions of soybean oil esters as diesel fuel. JAOCS 1996;73:154955.
[3] Freedman B, Pryde EH. Fatty esters from vegetable oils for use as a diesel fuel. In: Vegetable-oil fuels:
Proceedings of the int. conf. on plant and vegetable oils as fuels. Fargo, ND: ASAE; 1982. p.
11722.
[4] Graboski MS, McCormick RL. Combustion of fat and vegetable-oil derived fuels in diesel engines.
Prog Energy Combus Sci 1998;24:12564.
[5] Yuanwang D, Meilin Z, Dong X, Xiaobei C. An analysis for eect of cetane number on exhaust
emissions from an engine with the neural network. Fuel 2002;81:196370.
[6] Sozen A, Arcakhoglu E. Prediction of solar potential in Turkey. Appl Energ 2005;80:3545.
[7] Arcakhoglu E, C
avusoglu A, Erisen A. Thermodynamic analyses of refrigerant mixtures using
articial neural-networks. Appl Energ 2004;78:21930.
_
[8] Massie DD. Neural-network fundamentals for scientists and engineers. ECOS01, 46 July, Istanbul,
Turkey; 2001. p. 123.
[9] Golcu M, Sekmen Y, Erduranli P, Salman MS. Articial neural-network based modeling of variable
valve-timing in a spark-ignition engine. Appl Energy 2005;81:18797.
_ Performance and exhaust emissions in a diesel engine. Appl Energ
[10] Arcakhoglu E, C
elikten I.
2005;80:1122.
[11] More A, Deo MC. Forecasting wind with neural networks. Marine Struct 2003;16:3549.
[12] Yahya A, Marley SJ. Performance and exhaust emissions of a compression-ignition engine operating
on ester fuels at increased injection-pressure and advanced timing. Biomass Bioenerg 1994;6:297319.
[13] McDonald JF, Purcell DL, McClure BT, Kittelson DB. Emission characteristics of soy methyl ester
fuels in a compression-ignition engine. SAE Paper 950400, 1995.
[14] Heywood JB. Internal-combustion engine fundamentals. New York: McGraw Hill; 1988.
[15] Monyem A, Van Gerpen JH, Canakci M. The eect of timing and oxidation on emissions from
biodiesel-fueled engines. Trans ASAE 2001;44:3542.
[16] Chang DYZ, Van Gerpen JH. Fuel properties and engine performance for biodiesel prepared from
modied feedstocks. SAE Paper No. 971684, 1997.
[17] Ali Y. Beef tallow as a biodiesel fuel. Ph.D. Dissertation. University of Nebraska, Lincoln, NE, 1995.
[18] Borman G, Ragland KW. Combustion engineering. New York: McGraw Hill; 1998.
[19] Mittelbach M, Tritthart P. Diesel fuels derived from vegetable oils, III. Emission tests using methyl
esters of used frying oil. JAOCS 1988;65:11857.
[20] Rickeard DJ. Thompson ND. A review of the potential for bio-fuels as transportation fuels. SAE
Paper No. 932778, 1993.