i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1

Available online at www.sciencedirect.com

ScienceDirect

journal homepage: www.elsevier.com/locate/he

Mathematical modeling for the performance and

emission parameters of dual fuel diesel engine
using hydrogen as secondary fuel

A.E. Dhole a,*, R.B. Yarasu a, D.B. Lata b, S.S. Baraskar c

a
Department of Mechanical Engineering, Government College of Engineering, Amravati, Maharashtra 444603, India
b
Department of Mechanical Engineering, G.H. Raisoni College of Engineering and Management, Amravati 444705,
India
c
Department of Mechanical Engineering, Government Polytechnic, Amravati, Maharashtra 444603, India

article info abstract

Article history: In this work, mathematical models were developed to correlate the brake thermal effi-
Received 23 April 2014 ciency, un-burnt hydrocarbons, carbon monoxides and oxides of nitrogen by varying en-
Received in revised form gine parameters like Load and Gaseous (H2) fuel substitution. The developed models can be
12 June 2014 used to predict the important performance and emission parameters for diesel-hydrogen
Accepted 15 June 2014 operation in various combinations at different loads within the experimental domain.
Available online 11 July 2014 Response surface methodology (RSM) has been applied for developing the models using the
techniques of design of experiments and multi linear regression analysis. General factorial
Keywords: design was used to plan the experiments. Second order response surface models were
Dual fuel diesel engine found to be the most suitable in the present work. Analysis of variance (ANOVA) of the
Hydrogen experimental results at 95% confidence level revealed that the developed models are sig-
Design of experiments nificant. Comparison of experimental output with those predicted by the developed models
General factorial design showed close proximity having high correlation coefficients R2 for the various response
Mathematical model variables.
Response surface methodology Copyright © 2014, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights
reserved.

hydrogen and producer gas, which have been studied exten-

Introduction
Concern over current crude oil supplies in addition to varying
oil prices has resulted in the wide evaluation of substitution of
alternative sources of fuel. With the increasing need to
conserve fossil fuel and minimize toxic emissions much effort
is being focused on the advancement of present combustion
technology. This has encouraged exploration and testing of
several alternative fuels such as alcohol, gas viz. CNG, LPG,
sively [1e5]. The main exhaust emissions from diesel engine
are smoke, NOx. The only option to reduce these pollutants is
to use alternative fuels which do not have sulfur dioxide, al-
dehydes and ketones [6]. Amongst these alternative fuels,
hydrogen shows great potential. The advantage of using
hydrogen as fuel for internal combustion engine in dual fuel
mode is less polluting fuel, non-toxic, odorless, and has wide
range of flammability [6,7].

* Corresponding author. Tel.: þ91 9404242407.

E-mail addresses: ananta172002@yahoo.co.in, annt64@rediffmail.com (A.E. Dhole).
http://dx.doi.org/10.1016/j.ijhydene.2014.06.084
0360-3199/Copyright © 2014, Hydrogen Energy Publications, LLC. Published by Elsevier Ltd. All rights reserved.

Downloaded from http://www.elearnica.ir

12992 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1
Hydrogen dual fuel engines have many attractive features,
but they tend to suffer from premature ignition called
knocking, particularly under high load conditions because of
hydrogen's lower ignition energy, wider flammability range
and shorter quenching distance [6]. Hydrogen has auto-
ignition temperatures of 858 K and requires an ignition
source to be used in an I.C engine [8]. The diesel fuel which has
an auto-ignition temperature of 525 K can be used as a pilot
fuel to ignite hydrogen. Considerable research works have
been done on hydrogen as an alternative fuel in the I.C. engine
as it undergoes complete combustion [9]. Dual fuel engines
have been a subject of high interest due to their potential to
reduce emission with improved performances [10]. They
exhibit good thermal efficiency and low smoke level at high
power output [11]. The majority dual fuel engines use gaseous
fuel mostly inducted with the air during induction process as
secondary fuel. Therefore, hydrogen becomes a natural choice
as secondary fuel since it exhibits wide flammability limits,
high flame velocity and reduced pollution.
Several researchers have carried out works on hydrogen
[6e17]. Saravanan et al. [6] worked with hydrogen as an air-
enrichment medium while diesel as an ignition source and
found increase in brake thermal efficiency (hBTH ) with reduc-
tion in emission of NOx. Saravanan et al. [12] observed 15% rise
in hBTH at 75% load as compared to pure diesel operation by
injecting hydrogen into the intake port of a single cylinder
diesel engine while NOx emission was raised by 1e2% at full
load condition. Boretti [13] investigated that the dual fuel
hydrogen-diesel engine had higher brake thermal efficiency
nearer to 40% as compared to original diesel engine by intro-
ducing double injector one for hydrogen and other for diesel in
an injector of a common rail diesel engine. Gomes Antunes
et al. [14] obtained approximately 43% higher fuel efficiency in
hydrogen-fueled engine in comparison with 28% for the con-
ventional diesel engine in addition to 20% reduction in NOx
formation than diesel engine due to direct injection of
hydrogen in a diesel engine.
Lata et al. [15] observed appreciable and eco-friendly per-
formance of the engine using hydrogen, LPG and mixture
combinations of hydrogen and LPG in different proportions as
secondary fuels. Santoso et al. [16] identified slight increase in
indicated efficiency with hydrogen enrichment at 15 Nm load
and at lower load the efficiency decreases. Liu et al. [17]
observed that an addition of a small amount of hydrogen
significantly increased the emissions NO2/NOx ratio particu-
larly at low load condition. Poor combustion results in high
carbon monoxide and un-burnt emission [18]. Recently, au-
thors have presented the effect on performance and emis-
sions of a dual fuel diesel engine in which hydrogen, producer
gas and mixture of hydrogen and producer gas were used as
secondary fuels and the performances in these three distinct
cases were compared [19]. Lata et al. [20] have reported theo-
retical and experimental studies on dual fuel diesel engine
using hydrogen; LPG and a mixture of both were used as
secondary fuels. Theoretical models were presented to predict
brake thermal efficiency, pressure, and net heat release rate
and compared with experimental results.
Baraskar et al. [21] applied response surface methodology
for developing the models using the techniques of design of
experiments and multi-linear regression analysis. Validation
of the model was confirmed by conducting the confirmation
experiments. Mu'azu et al. [22] developed a mathematical
model for the esterification of J. curcas seed oil and studied
effect of methanol to oil ratio, catalyst concentration and re-
action time on free fatty acid (FFA).
Mu'azu et al. [23] developed mathematical model for the
transestrification of jatropha curcas seed oil by factorial
analysis of design of experiment. The factors which studied
were methanol to oil molar ratio (6e10), catalyst concentra-
tion (4e8 wt %), reaction time (1e2 h) and stirrer speed
(100e700 rpm). Analysis of variance (ANOVA) of the experi-
mental results at 95% confidence level exposed that the
developed model equation was significant and stirrer speed
was the major parameter in the transestrification reaction
followed by reaction time and methanol to oil ratio having
effects of 24.59, 21.16 and 8.07% respectively. Comparison of
experimentally obtained biodiesel yields with the predicted
biodiesel yields by the developed model showed close prox-
imity having high correlation coefficient (R2 ¼ 0.995). The
developed mathematical model could be employed in simu-
lation of biodiesel production under the reaction conditions
studied.
Dhar et al. [24] presented work which evaluates the effect
of current (c), pulse-on time (p) and air gap voltage (v) on metal
removal rate (MRR), tool wear rate (TWR), radial over cut (ROC)
on electric discharge machining of Ale4Cue6Si alloye10 wt %
SiCP composites. Three factors, three level full factorial de-
signs were adopted for analyzing the results. A second order,
non-linear mathematical model had been developed for
establishing the relationship among machining parameters.
Analysis of variance (ANOVA) had been performed to verify
the fit and adequacy of the developed mathematical models.
Karthikeyan et al. [25] developed mathematical models for
optimizing electric discharge machining (EDM) characteristics
such as the metal removal rate (MRR), the tool wear rate (TWR)
and the surface roughness (CLA value). The process parame-
ters had been taken in to consideration were the current (I),
the pulse duration (T) and the percent volume fraction of SiC
(25 m size) (V) present in LM25 aluminium matrix. A three
level full factorial design was chosen for experimentation and
mathematical models with linear, quadratic and interactive
effects of the parameters chosen were developed. Finally the
significance of the models was verified by using the analysis of
variance technique (ANOVA).
Saidur et al. [26] investigated the effect of parameters like
engine speed, throttle position, and operating time on engine
emissions at different load conditions. The significance and
adequacy of these parameters were evaluated using analysis
of variance (ANOVA) and fisher's statistical test (F-test). From
the analysis, it was concluded that emissions were more
affected by engine speed and throttle position. Muhammad
et al. [27] described the use of response surface methodology
for the optimized biodiesel production using chemical and
enzymatic transesterification of rice bran and sunflower oils.
Based upon (ANOVA) and response Surface plots, significant
impact of reaction parameters under study was ascertained.
The monitoring of exhaust emission of synthesized biodiesels
and their blends revealed a marked reduction in carbon
monoxide (CO) and particulate matter (PM) levels, whereas an
irregular trend was observed for NOx emissions.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1 12993

delay, combustion duration, heat release rate and combustion

Table 1 e Engine specification.
phasing.
Sr.No. Parameter Engine specification Hence, this paper presents thorough experimental results
1. Make and Model Ashok Leyland ALU WO4CT at different load conditions on the performance and emission
Turbocharged, inter-cooler, parameters of a dual fuel turbocharged multi-cylinder diesel
Gen- Set engine with hydrogen as secondary fuels. In addition, this
2. General Details Four Stroke, Compression
study was evaluated the suitability of response surface
Ignition, Constant Speed,
methodology for the development of a mathematical model
vertical, water-cooled,
direct injection, turbo describing the relationships and subsequent effects of the
charger, Intercooler, Gen-Set primary process variables.
3. No. of Cylinder 4
4. Bore mm 104
5. Stroke mm 113
6. Rated Speed (rpm) 1500 Experimentation
7. Swept volume (cc) 3839.67
8. Clearance volume (cc) 84.90 A diesel engine test setup was developed to carry out the
9. Compression ratio 17.5:1 experimentation on dual fuel engine. A 4-stroke CI engine,
10. Injection Pressure (bar) 260
model Ashok Leyland ALUWO4CT, turbocharged with inter-
11. Injection Timing BTDC 160
12. Rated Power kW at 62.5
cooler and gen-set was used for the experimental investiga-
1500 rpm tion. Detailed specification for the present work is given in
13. Inlet Pressure (bar) 1.06 Table 1. The diesel engine was modified by attaching
14. Inlet Temperature K 313 hydrogen gas cylinder with the intake manifold to work on
15. Nozzle Diameter mm 0.285 dual fuel mode through flame traps, mass flow meters. It was
16. Number of hole 5
followed by a one-way non-return valve and common flame
arrestor by keeping turbocharger and its bypass active. The
engine was coupled to a 62.5 kW D.C. generator. The load on
In all the referred work, research on the addition of the engine was varied by introducing five water pumps and
hydrogen to heavy-duty diesel engines has been reported to twelve 3 kW industrial water heaters in a set of four each. The
very little extent. Existing experimental data indicated to engine was run at a constant speed of 1500 RPM. A schematic
some extent the reliance of the improvement to the brake layout and view of the diesel engine test setup used during the
thermal efficiency on the engine load and the amount of experimentation is shown in Figs. 1 and 2 respectively.
hydrogen added [28]. A well designed experiment will help to Exhaust gas emissions explicitly, CO, NOx and un-burnt
explain the effects of the addition of H2 on the thermal effi- hydrocarbons (UBHC) were measured by an AVL 5000 DI Gas
ciency and emission parameters like un-burnt hydrocarbon analyzer. The CO was measured in volume percentage basis
(UBHC), carbon monoxide (CO) and oxides of nitrogen (NOx) at while NOx and UBHC were measured in ppm units. The cyl-
different engine loads. For better understanding of the po- inder pressure was measured with the help of piezoelectric
tential of hydrogen in affecting the brake thermal efficiency pressure transducer (pressure range 0e250 bars) and a charge
and exhaust emissions of heavy duty diesel engines, there is a amplifier. The pressure data were transferred to data acqui-
need to explore the detailed effect of hydrogen addition and sition system for further analysis. A Kistler make crank angle
engine load on the combustion process including the ignition encoder with an accuracy of 1 was used for angle

Fig. 1 e Schematic of the experimental set up.1-Engine, 2-Gen-set, 3-Diesel tank and measurement system, 4-Air tank and
measurement system, 5-Gas supply, 6-Gas analyzer, 7-PC based data acquisition system, 8-Charge amplifier,9-Cylinder
pressure sensor, 10-Crank angle encoder, 11-Hydrogen gas flame trap, 12-Gas flow meter, 13-Gas cylinder control valve, 14-
Pressure regulator, 15- Solenoid switch valve,16-Temperature and Pressure measurement locations, 17- Hydrogen gas cylinder.
12994 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1

Table 3 e Design factors and their corresponding

variation levels.

Symbols Design Units levels

Factor
A GFS (H2) % 0 5 10 15 20 25
B Load % 13 40 60 80
H2- Hydrogen

BP
hBTH ¼
 (1)
md qd þ mH2 qH2  1000
Fig. 2 e View of experimental engine Set up.
where, md and mH2 are the mass flow rate of diesel and
hydrogen respectively, qd and qH2 are the lower heat of com-
bustion of diesel and hydrogen respectively.
measurement. After 15 min of engine operation on stabilized
Exhaust gas emissions explicitly, CO, NOx and UBHC are
conditions, pressure data were obtained for an average of 100
measured by an AVL 5000 DI Gas analyzer. The CO is
cycles. The mass flow rate of hydrogen was measured by mass
measured in volume percentage basis while NOx and UBHC
flow meters in liters per minute. The experiments were con-
are measured in ppm units which then converted to g/kWh.
ducted for five times to ensure repeatability. The working
conditions used during experimentation are given in Table 2.
Factorial design employed

Experiments were designed on the basis of design of experi-

Design of experiments ments. The design finally chosen was a General factorial
design having two factors i.e. GFS and Load. Levels selected for
The design factors, response variable as well as the method- GFS and Load were six and four respectively. The total
ology employed for the experimentation is described below. numbers of experiments performed were 6  4 ¼ 24 (Table 4).
Based on General factorial design, experiments were con-
ducted to develop mathematical models for hBTH , UBHC, CO
Design factors and NOx in terms of the two input variables as GFS and Load.

The design factors considered in the present work were

Gaseous fuel substitution (GFS) in % and Load in %. The se-
lection of these two factors has been made to study the effect
of various hydrogen-diesel combinations on the response
variables at different load conditions in order to analyze the
prospect of an alternative fuel such as hydrogen. Table 3
shows the relationship between the design factors and their
corresponding selected variation levels. H2-0 shows that the
engine operated on pure diesel only.
Response variables
Response surface methodology
Response surface methodology is a collection of mathematical
and statistical technique that is useful for modeling and
analysis of problems in which a response of interest is influ-
ence by several variables and the objective is to optimize the
response [29,30]. In order to study the effect of dual fuel engine
process parameters on the hBTH , UBHC, CO and NOx, a second
order polynomial response is fitted into the following
equation-

X
k X
k X
k X
k
The selected response variables are Brake thermal efficiency Y ¼ bo þ bi Xi þ bii X2i þ bij Xi Xj (2)
(hBTH ) in %, Un-burnt Hydrocarbon (UBHC) in g/kWh, Carbon i¼1 i¼1 i1 > j j

monoxide (CO) in g/kWh and Oxides of nitrogen (NOx) in g/

where, Y is the response, i and j are the linear and quadratic
kWh. hBTH was calculated by equation given below.
coefficients respectively, bo is the regression coefficient and k
is the number of factors studied in the experiment.
Table 2 e Working conditions used during
experimentation.
Primary Secondary Load (%) Secondary fuel Experimental results
fuel fuel substitution as %
of diesel at each load %
Table 4 shows the design matrix developed for the proposed
Diesel GFS (Hydrogen) 13,40,60,80 H2-0, H2-5, H2-10, model as well as the performance characteristics value ob-
H215, H2-20, H2-25
tained in the experiments for hBTH , UBHC, CO and NOx.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1 12995

Model adequacy test for hBTH

Table 4 e Design of experiment matrix and performance
characteristics.
The ANOVA [30] and Fisher‘s statistical test (F-test) were
Expt. run Expt. GFS Load hBTH HC CO NOx
performed to check the adequacy of the model as well as the
No. A % B % % g/kWh g/kWh g/kWh
significance of individual parameters. Table 5 shows the pre-
16 1 0.0 13.0 20.8 1.50 0.22 1.80 ANOVA model summary statistics for hBTH . It can be seen
17 2 5.0 13.0 18.8 1.07 0.23 1.36
that standard deviation (SD) and predicted error sum of
8 3 10.0 13.0 19.3 1.19 0.21 1.31
4 4 15.0 13.0 20.5 1.28 0.28 1.03
squares (PRESS) of the suggested cubic model is 0.69 and 25.91
13 5 20.0 13.0 20.7 3.65 0.37 0.76 respectively, which is much better as compared to the lower
9 6 25.0 13.0 20.2 3.89 0.43 0.66 order model for R-squared. Hence the cubic model suggested
15 7 0.0 40.0 27.7 1.65 0.24 6.15 is most appropriate.
7 8 5.0 40.0 26.4 1.48 0.23 4.18 Table 6 shows the variance analysis results of the proposed
14 9 10.0 40.0 26.9 1.64 0.26 3.29
model of hBTH . The ANOVA table includes Sum of Squares (SS),
5 10 15.0 40.0 27.1 2.12 0.29 2.61
Degrees of Freedom (DF), Mean Square (MS), F-value and P-
18 11 20.0 40.0 27.4 3.89 0.39 2.39
11 12 25.0 40.0 26.8 4.42 0.41 2.15 value. The MS was obtained by dividing the SS of each of the
6 13 0.0 60.0 29.6 1.71 0.29 6.59 sources of variation by the respective DF. The P-value is the
2 14 5.0 60.0 30.3 1.81 0.24 5.27 smallest level of significance at which the data are significant.
3 15 10.0 60.0 31.4 2.18 0.27 5.19 The F-value is the ratio of MS of the model terms to the MS of
12 16 15.0 60.0 31.7 2.49 0.31 4.89 the residual.
23 17 20.0 60.0 32.5 4.09 0.34 4.52
As seen from Table 6, the P-values for developed model of
10 18 25.0 60.0 33.1 4.72 0.39 4.18
hBTH are less than 0.05, which indicates that model is significant
24 19 0.0 80.0 31.5 1.75 0.30 9.21
19 20 5.0 80.0 32.1 1.96 0.26 5.48 at 95% confidence level. It was noted that MS of the model (64.7)
21 21 10.0 80.0 32.4 2.31 0.28 5.78 is many times larger than MS of the residual (0.47), thus the
1 22 15.0 80.0 33.2 2.59 0.33 6.03 computed F-value of the model (F ¼ 64.7/0.47) of 137.0 implies
20 23 20.0 80.0 33.7 4.17 0.36 5.97 that the model is significant. The R-Squared is defined as the
22 24 25.0 80.0 31.7 4.92 0.40 6.24 ratio of variability explained by the model to the total variability
in the actual data and is used as a measure of the goodness of fit.
The more R2 approaches unity, the better the model fits the
Modeling response variables
experimental data. For instance, the obtained value of 0.9888 for
R2 in the case of hBTH implies that the model explains variations
Design expert 8.0 (Statease, Mananpolis) software was used
in hBTH to the extent of 98.88% in the current experiment and
for developing the mathematical models for hBTH , HC, CO and
thus the model is adequate to represent the process. The “Pre-
NOx and presented in equations (3)e(6) respectively.
dicted R2” of 0.9560 is in sound agreement with the “Adjusted R2”
of 0.9816 because the difference between the adjusted and
hBTH ¼ þ19.17733 predicted R2 is within 0.2 as recommended for model to be
 0.26336*GFS þ 0.064335*Load þ 0.010035*GFS2 þ 5.13584E  adequate. The value of Predicted R2“ of 0.9560 indicates the
003*Load2 þ 3.88450E  003*GFS*Load  1.06250E  prediction capability of the regression model. It means that the
004*GFS3  5.03900E  005*Load3  4.73677E  005*GFS2*Load  model explains about 95.60% of the variability in predicting new
1.17919E  005*GFS*Load2 (3) observations as compared to the 98.88% of the variability in the
original data explained by the least square fit. “Adequate Preci-
sion” measures the signal to noise ratio. A ratio greater than 4 is
UBHC ¼ þ0.97594  0.028099*GFS þ 0.020031*Load þ desirable. The ratio of 31.74 indicates an adequate signal. Thus,
1.67321E  003*GFS2  1.09590E  004*Load2 þ 1.26026E  the overall prediction capability of the model based on these
004*GFS*Load (4) criteria seems very satisfactory.
Fig. 3 presents a plot of experimental versus the predicted
values of hBTH . Since all the predicted values are close to the
CO ¼ þ0.18237  5.28698E  003*GFS þ 4.10849E  003*Load þ
3.57071E  004*GFS2  7.51870E  005*Load2  1.31674E 
005*GFS*Load  3.15972E  006*GFS3 þ 5.25948E 
007*Load3  5.87576E  007*GFS2*Load þ Table 5 e Pre-ANOVA model summary statistics for hBTH .
9.56534E  008*GFS*Load2 (5) Underline signifies the suitability conditions to suggest
the model i.e. minimum SD and PRESS Value and R2 value
close to unity.

NOx ¼ þ0.11742  0.076132*GFS þ 0.15574*Load þ 4.96166E  Source SD R2 Adj. R2 Pred. R2 PRESS

003*GFS2  2.91238E  004*Load2  4.71224E  003*GFS*Load  Linear 1.49 0.9205 0.9129 0.8923 63.38
6.96528E  005*GFS3  4.59201E  006*Load3 þ 3.72461E  2FI 1.52 0.9218 0.9101 0.8738 74.28
005*GFS2*Load þ 2.96870E  005*GFS*Load2 (6) Quadratic 0.90 0.9751 0.9682 0.9430 33.56
a
Cubic 0.69 0.9888 0.9816 0.9560 25.91

where, the values of the variables have been specified ac- SD ¼ Standard deviation.
a
cording to their original units. Suggested.
12996 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1

Table 6 e ANOVA for cubic model of hBTH . Table 7 e - Pre-ANOVA model summary statistics for HC.
Source SS DF MS F-Value P-Value Source SD R2 Adj. R2 Pred. R2 PRESS
Model 581.8 9 64.7 137 <0.0001 Linear 0.56 0.8154 0.7978 0.7553 8.74
GFS 1.6 1 1.60 3.39 0.0869 2FI 0.57 0.8173 0.7899 0.7253 9.81
a
Load 42.3 1 42.3 89.7 <0.0001 Quadratic 0.36 0.9364 0.9188 0.8848 4.12
GFS2 2.94 1 2.94 6.22 0.0257 Cubic 0.36 0.9495 0.9170 0.8527 5.26
Load2 12.1 1 12.1 25.6 0.0002
Underline signifies the suitability conditions to suggest the model
GFS*Load 1.3 1 1.28 2.72 0.1213
i.e. minimum SD and PRESS Value and R2 value close to unity.
GFS3 2.9 1 2.93 6.20 0.0259 a
Suggested.
Load3 2.8 1 2.79 5.92 0.0290
GFS2*Load 2.1 1 2.06 4.36 0.0555
GFS*Load2 0.25 1 0.25 0.53 0.4785 capability of the model. The results of the statistical analysis
Residual 6.61 14 0.47 and Fig. 5 shows that model can satisfactorily be used in
Total 588.4 23 predicting the response of CO.
Adequate precision 31.74

Model adequacy test for NOx

experimental values, confirming that the model could predict
the responses accurately. The statistical analysis of the model of NOx is presented in
Tables 11 and 12 respectively. Since, cubic model is having the
SD of 0.53 as well as the PRESS 21.66 which is much better as
Model adequacy test for UBHC
compared to other lower order model for R squared, hence
suggested. Similarly, the plot in Fig. 6 shows the prediction
Similarly, a pre-ANOVA model statistics, the ANOVA results
capability of the model. The results of the statistical analysis
and the post-ANOVA model adequacy for the developed
and plot shows that model can satisfactorily be used in pre-
model of UBHC are shown in Tables 7 and 8 respectively.
dicting the response of NOx.
Least SD (0.36) and PRESS (4.12) of quadratic model confirm
that quadratic model is most suitable. Also R2 Statistics
shown in Table 7 further confirms the suitability of the
suggested model. Similarly, P-value and F-value of the
Results and discussion
developed model as depicted in Table 8 showed the signifi-
The influences of parameters like GFS and Load on the
cance of the model. Furthermore, the plots from Fig. 4
selected response variables (hBTH , UBHC, CO and NOx) were
confirm that the developed model can be used to predict
assessed on dual fuel diesel engine. In order to predict the
the UBHC efficiently.
responses accurately, second order models were postulated in
obtaining a relationship between design factors and response
Model adequacy test for CO
variables. The analysis of variance (ANOVA) was used to check
the adequacy of the model. Values of various regression sta-
The statistical analysis of the model of CO is presented in
tistics were compared to identify the best fit model. The
Tables 9 and 10 respectively. Since, cubic model is having least
developed response surface methodology (RSM) based math-
SD (0.016) and PRESS (0.010) among the other models, hence
ematical models of hBTH , UBHC, CO and NOx are discussed
suggested. Similarly, the plot in Fig. 5 shows the prediction
below.

36 Analysis of hBTH
Predicted Brake Thermal Efficiency (%)

33 ANOVA for response surface reduced cubic model ofhBTH is

depicted in Table 6. P-value of the load is less than 0.05 in-
30 dicates that load has a significant effect on the hBTH . Fig. 7(a)
shows the estimated response surface for hBTH . It is observed
27

24 Table 8 e ANOVA for cubic model of HC.

Source SS DF MS F-Value P-Value
21 Model 33.45 5 6.69 53.04 <0.0001
GFS 11.58 1 11.58 91.82 <0.0001
18 Load 1.22 1 1.22 9.68 0.0060
GFS2 4.18 1 4.18 33.15 <0.0001
18 21 24 27 30 33 36 Load2 0.074 1 0.074 0.59 0.4534
GFS*Load 0.068 1 0.068 0.54 0.4713
Actual Brake Thermal Efficiency (%)
Residual 2.27 18 0.13
Fig. 3 e Experimental vs. Predicted value of Brake thermal Total 35.72 23
Adequate precision 22.75
efficiency (hBTH ).
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1 12997

6 0.5
Predicted Un-burnt Hydrocarbons (g/kWh)

Predicted Carbon Monoxide (g/kWh)

5
0.4
4

3 0.3

2
0.2
1

0.2 0.3 0.4 0.5

0 Actual Carbon Monoxide (g/kWh)
0 1 2 3 4 5 6
Actual Un-burnt Hydrocarbons (g/kWh) Fig. 5 e Experimental vs. Predicted value of Carbon
monoxides (CO).
Fig. 4 e Experimental vs. Predicted value of Un-burnt
hydrocarbons (UBHC).
distance and more thermal conductivity [13]. Further, even at
low concentration of hydrogen in the intake air, the com-
bustion spreads throughout the gas-air mixture; this causes
Table 9 e Pre-ANOVA model summary statistics for CO. high heat transfer losses to the neighboring walls. While, in
case of diesel engines under light load condition, the distri-
Source SD R2 Adj. R2 Pred. R2 PRESS
bution of the diesel spray is such that it does not reach the
Linear 0.034 0.7628 0.7402 0.6736 0.033 cylinder walls and the combustion is confined to piston bowl.
2FI 0.031 0.8108 0.7824 0.6980 0.031
Also, the coatings of surrounding air acts as insulation in be-
Quadratic. 0.019 0.9341 0.9158 0.8913 0.011
a
Cubic 0.016 0.9657 0.9437 0.9009 0.010
tween burnt gases and the walls, which reduces heat losses
thereby giving better hBTH with pure diesel.
Underline signifies the suitability conditions to suggest the model
At 60% and 80% load conditions using 25% and 20% GFS,
i.e. minimum SD and PRESS Value and R2 value close to unity.
a
Suggested.
figure shows that the maximum hBTH is 33.1% and 33.7%
respectively which is more as compared to 29.6% and 31.5%
for pure diesel operation. This may be due to the fact that an
that the hBTH using 20% GFS is less than pure diesel operation increase in combustion rate with larger pilot diesel quantities
at lower load conditions. This may be because of the fact that leads to stronger ignition sources and hence, more compre-
the pilot diesel fuel is low in quantity at lower load conditions. hensive and better combustion of gaseous fuel. The flamma-
Thus, fewer ignition centers are formed due to small quantity bility limits of premixed hydrogen flame are much wider than
of injected pilot diesel. Addition of gaseous fuel further lowers the diesel fuel. Therefore, cylinder chamber gets fumigated
pilot diesel fuel hence, poor ignition of gaseous fuel results in with hydrogen which enhances burning of the diesel fuel.
less hBTH .The maximum hBTH at 13% load and 20% GFS condi- With the increase in hydrogen concentration, the pressure
tion is found to be 20.7% as compared to 20.8% of pure diesel increases due to high flammability limits and rate of com-
operation. The brake thermal efficiency is less due to more bustion of hydrogen. It was observed by Boretti [13] that the
cooling losses which may be due to shorter quenching hBTH increases up to 40% as compared to original diesel engine.
Fig. 7(b) is the two-dimensional contour plot obtained by
connecting points of GFS and Load. If a particular value ofhBTH
is desired, for example 26.5%, according to Fig. 7(b) there are
Table 10 e ANOVA for cubic model of CO.
many combinations of GFS and load, on the contour line of
Source SS DF MS F-Value P-Value hBTH ¼ 26.5%.
Model 0.098 9 0.011 43.85 <0.0001
GFS 1.38 1 1.38 5.55 0.0336
Load 1.01 1 1.01 4.06 0.0636
Table 11 e Pre-ANOVA model summary statistics for NOx.
GFS2 9.29 1 9.29 3.73 0.0741
Load2 1.72 1 1.72 0.07 0.7967 Source SD R2 Adj. R2 Pred. R2 PRESS
GFS*Load 1.34 1 1.34 5.38 0.0360
Linear 0.75 0.8997 0.8902 0.8600 16.65
GFS3 2.59 1 2.59 10.4 0.0062
2FI 0.77 0.9002 0.8852 0.8328 19.89
Load3 3.04 1 3.04 1.22 0.2883
Quadratic 0.67 0.9318 0.9128 0.8479 18.09
GFS2*Load 3.17 1 3.17 1.27 0.2789 a
Cubic 0.53 0.9666 0.9451 0.8179 21.66
GFS*Load2 1.65 1 1.65 0.066 0.8010
Residual 3.49 14 2.49 Underline signifies the suitability conditions to suggest the model
Total 0.10 23 i.e. minimum SD and PRESS Value and R2 value close to unity.
a
Adequate precision 22.48 Suggested.
12998 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1

spreading of the pilot fuel prior to ignition. This will lead to

Table 12 e ANOVA for cubic model of NOx.
poor combustion of the gaseous fuel-air mixture. Larger
Source SS DF MS F-value P-value gaseous fuel substitutions which also displace some part of the
Model 115 9 12.7 44.98 <0.0001 air in the intake process may lead to more utilization of oxygen
GFS 0.76 1 0.76 2.68 0.1237 and cause reduction in the available oxygen for further com-
Load 6.77 1 6.77 23.84 0.0002 bustion of pilot diesel fuel, thus increased UBHC levels. Fig. 8(b)
GFS2 0.63 1 0.63 2.22 0.1586
shows the effect of GFS and load on the estimated response of
Load2 0.60 1 0.60 2.10 0.1690
UBHC. This graph is very useful in predicting the parameters
GFS*Load 1.06 1 1.06 3.74 0.0735
GFS3 1.26 1 1.26 4.43 0.0539 GFS and load for given value of UBHC.
Load3 0.02 1 0.02 0.082 0.7793
GFS2*Load 1.27 1 1.27 4.48 0.0527 Analysis of CO
GFS*Load2 1.59 1 1.59 5.58 0.0331
Residual 3.98 14 0.28 The significant effect of design factors on CO is depicted in
Total 119 23
Table 10. It is seen that GFS is the most significant factor. The
Adequate precision 23.2
progress of CO depends on post-oxidation reaction. In com-
mon diesel engine, the carbon oxidation reaction is almost
Analysis of UBHC
completed due to the presence of more excess air. Fig. 9(a)
shows the estimated response surface of CO, varying the
The effect various design factors on HC is represented in Table
factors of GFS and load. It is observed that at 13% load con-
8. The GFS is the most significant factor among the design
dition, rise in CO emission is 95% using 25% GFS as compared
factors. Fig. 8(a) shows the estimated response surface of
to pure diesel operation. Higher concentration of CO in the
UBHC by varying GFS and load. The extra leaning and
exhaust is a clear indication of incomplete combustion of the
improper mixing are responsible for UBHC emission in diesel
premixed mixture. At this load condition gaseous fuel-air
engine. As seen from this figure, at 13% load condition, 25%
mixture near the pilot is burned due to less turbulence. Thus
GFS exhibits maximum UBHC emissions of 3.89 g/kWh as
some partial oxidation product like carbon monoxide may
compared to 1.5 g/kWh of pure diesel operation. Reduction in
come out in the exhaust. At higher concentration of gaseous
pilot fuel quantity causes poor ignition of gaseous fuel and
fuel, the concentration of the partial oxidation product could
inducted mixture is too lean to burn could be the reason for
increase [5,11].
this. Also, the gas temperature is low at low loads, while at
At 80.0% load condition using 25% GFS, maximum rise in
higher load this rises due to quicker burning of hydrogen. This
CO emission is 33% as compared to pure diesel. This may be
leaves diesel fuel injected toward the end of injection period,
due to richness of mixture and major temperature at this
inadequate in oxygen. Decrease in oxygen concentration
condition. Hydrogen shows different behavior in dual fuel
would cause an increase in total unburned UBHC [2]. There-
engine due to presence of liquid hydrocarbon. As soon as the
fore, rate of UBHC emissions is more.
ignition starts, the spontaneous combustion occurs due to the
Figure also shows that the maximum UBHC emission of
presence of more percentage of hydrogen. Consequently, the
4.92 g/kWh occurs at 80% load condition using 25% GFS against
diesel fuel is more subjected to higher combustion tempera-
1.75 g/kWh of pure diesel operation. At higher load condition
ture in an atmosphere due to lack of oxygen. Further, CO
due to a low pilot quantity, UBHC emission is high at gaseous
emission is increased at all load conditions due to delayed
fuel substitution. High load condition results in increased
ignition period. Fig. 9(b) shows the effect of GFS and load on
ignition delay and cylinder gas temperature which may lead to
the estimated response of CO. This graph is very useful in
predicting the parameters GFS and load for given value of CO.
10
Analysis of NOx
Predicted Oxides of Nitrogen (g/kWh)

9
8 Nitric oxide (NO) and a small amount of nitrogen dioxide (NO2)
7 are the main constituents of nitric oxides, mostly formed by
the oxidation of atmospheric nitrogen in the combustion
6
chamber. NO formation is mainly controlled by combustion
5 temperatures and the availability of oxygen. Whereas, for-
4 mation of NOx in the dual fuel engine mostly depends on
diesel pilot spray region. The formation of NOx increases with
3
the increase in the size and amount of pilot diesel fuel. Also,
2 the nitrogen oxide emission rises with the increase in cylinder
1 temperature, oxygen concentration and combustion duration.
ANOVA for response surface reduced cubic model of NOx is
0
0 1 2 3 4 5 6 7 8 9 10 depicted in Table 12. Fig. 10(a) shows the estimated response
surface of NOx, varying the factors of GFS and load exhibiting
Actual Oxides of Nitrogen (g/kWh)
the variation of NOx. It is observed that dual fuel operation
Fig. 6 e Experimental vs. Predicted value of Oxides of produces less NOx at all load conditions than pure diesel
nitrogen (NOx). operation. At 13.0% and 80.0% load conditions using 25% GFS,
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1 12999

Fig. 7 e (a) Response surface of hBTH vs. (b) Contours of hBTH vs. Load and GFS Load and GFS.

drop in NOx emission is observed 63% and 32.3% as compared

to pure diesel. This may be due to increase in hydrogen sub-
stitution simultaneously increases the mole fraction of H2O
i.e. the moisture increases which finally brings down the peak
temperature. Hence, NOx decreases with the increase in
hydrogen substitution [2]. Fig. 10 (b) shows the effect of GFS
and load on the estimated response of NOx. This graph is very
useful in predicting the parameters GFS and load for given
value of NOx.
Conclusions
On the basis of the results and discussions presented above,
the following conclusions could be drawn.
1. The performance study of CI engine operated on diesel and
hydrogen in dual fuel mode indicates that no major
modification is required in an existing diesel engine.

80.0
UBHC

5.16
63.2
4.15

3.14 1.79 4.48

B: LOAD

2.12
2.46 3.81
UBHC

46.5
1.11 3.14

29.8

80.0
25.0
63.2
18.8
46.5 13.0
12.5
0.0 6.3 12.5 18.8 25.0
B: LOAD 29.8 6.3
A: GFS
13.0 0.0 A: GFS

Fig. 8 e (a) Response surface of UBHC vs. (b)Contours of UBHC vs. Load and GFS Load and GFS.
13000 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1

80.0
CO
0.28

0.43

0.37 63.2

0.32
0.24
0.26

B: LOAD
0.35
0.28
CO

0.32
0.20 46.5

0.39

29.8

80.0
25.0
63.2
18.8
46.5 13.0
12.5
0.0 6.3 12.5 18.8 25.0
B: LOAD 29.8 6.3
A: GFS
13.0 0.0 A: GFS

Fig. 9 e (a) Response surface of CO vs. (b) Contours of CO vs. Load and GFS Load and GFS.

2. Use of hydrogen as secondary fuel enhances the hBTH at
high load conditions while, it produces reasonably good
effect at low load conditions.
3. UBHC/CO emissions for various combinations of diesel-
hydrogen in dual fuel mode are found to be more than
pure diesel fuel engine operation.
4. NOx emissions for various combinations of diesel-
hydrogen in dual fuel mode are found to be lower than
pure diesel fuel engines.
5. The best performances of the engine employed for inves-
tigation are obtained by the substitution of 20% of
hydrogen at higher loads.

80.0
NOx
5.70

8.36
7.03
6.36 63.2 5.70

4.37
4.37
2.37
B: LOAD
NOx

0.37 46.5

3.04

29.8

80.0
1.70
25.0
63.2
18.8
46.5 13.0
12.5
0.0 6.3 12.5 18.8 25.0
B: LOAD 29.8 6.3
A: GFS
13.0 0.0 A: GFS

Fig. 10 e (a) Response surface of NOx vs. (b)Contours of NOx vs. Load and GFS Load and GFS.
 i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 9 ( 2 0 1 4 ) 1 2 9 9 1 e1 3 0 0 1 13001

6. Experimental values of hBTH , UBHC, CO and NOx can satis- [13] Boretti A. Advantages of the direct injection of both diesel
factorily be predicted from experimental diagrams of and hydrogen in dual fuel H2ICE. Int J Hydrogen Energy
response surfaces and contour graphs. Results showed 2011;36:9312e7.
[14] Gomes Antunes JM, Mikalsen R, Roskilly AP. An experimental
that general factorial design is a powerful tool for providing
study of a direct injection compression ignition hydrogen
experimental diagrams and statistical mathematical engine. Int J Hydrogen Energy 2009;34:6516e22.
models, to perform the experiments efficiently and [15] Lata DB, Misra A, Medhekar S. Effect of hydrogen and LPG
economically. addition on the efficiency and emissions on a dual fuel diesel
7. Analysis of variance (ANOVA) of the experimental results engine. Int J Hydrogen Energy 2012;30:1e13.
at 95% confidence level revealed that the developed model [16] Santoso WB, Nur A, Ariyono S, Bakar RA. Combustion
is significant. Comparison of experimental output with characteristics of diesel-hydrogen dual fuel engine.
ISBN:978-967-0120-04-1.
the predicted by the developed model showed close
[17] Liu S, Li H, Liew C, Gatts T, Wayne S, Shade B. An
proximity having high correlation coefficients R2 for the experimental investigation of NO2 emission characteristics
various response variables. of a heavy-duty H2-diesel dual fuel engine. Int J Hydrogen
Energy 2011;36:12015e24.
[18] Lambe SM, Watson HC. Low polluting, energy efficient C.I.
hydrogen engine. Int J Hydrogen Energy 1992;17:513e25.
references [19] Dhole AE, Yarasu RB, Lata DB, Priyam A. Effect on
performance and emissions of a dual fuel diesel engine using
hydrogen and producer gas as secondary fuels. Int J
Hydrogen Energy 2014;39:8087e97.
[1] Namasivayam AM, Korakianitis T, Crookes RJ, Bob-
[20] Lata DB, Misra A. Theoretical and experimental
Manuel KDH, Olsen J. Biodiesel, emulsified biodiesel and
investigations on the performance of dual fuel diesel engine
dimethyl ether as pilot fuels for natural gas fuelled engines.
with hydrogen and LPG as secondary fuels. Int J Hydrogen
Appl Energy 2010;87:769e78.
Energy 2010;35:11918e31.
[2] Karthikeyan B, Srithar K. Performance characteristics of a
[21] Baraskar SS, Banwait SS, Laroiya SC. Mathematical modeling
glow plug assisted low heat rejection diesel engine using
of electrical discharge machining process through response
ethanol. Appl Energy 2011;88:323e9.
surfaces methodology. Ijser November-2011;2(11). ISSN 2229-
[3] Chandra R, Vijay VK, Subbarao PM, Khura TK. Performance
5518.
evaluation of a constant speed IC engine on CNG, methane
[22] Mu’azu K, Mohammed-Dabo A, Waziri SM, Ahmed AS,
enriched biogas and biogas. Appl Energy 2011;88:3969e77.
Bugaje IM. Development of mathematical model for the
[4] Agarwal AK, Rajamanoharan K. Experimental investigations
esterification of Jatropha curcas seed oil. JPTAF March
of performance and emissions of Karanja oil and its blends in
2013;4(3):44e52.
a single cylinder agricultural diesel engine. Appl Energy
[23] Dhar S, Purohit R, Saini N, Sharma A, Hemath Kumar G.
2009;86:106e22.
Mathematical modeling of electric discharge machining of
[5] Mansour C, Abdelhamid B, Abdelkader A, Francoise G. Gas-
cast Ale4Cue6Si alloye10 wt.% SiCP composites. J Mater
diesel (dual-fuel) modeling in diesel engine environment. Int
Process Technol 2007;194:24e9.
J Therm Sci 2001;40:409e24.
[24] Muazu K, Mohammed-Dabo IA, Waziri SM, Ahmed AS,
[6] Saravanan N, Nagarajan G. An experimental investigation of
Bugaje IM, Ahmad MS. Development of mathematical model
hydrogen-enriched air induction in a diesel engine system.
for the transesterification of Jatropha curcas seed oil in
Int J Hydrogen Energy 2008;33:1769e75.
heterogeneous catalyst. Int J Adv Renew Energy Res
[7] Choi GH, Chung JY, Han SB. Performance and emissions
2012;11(1):658e65.
characteristics of a hydrogen enriched LPG internal
[25] Karthikeyan R, Lakshmi Narayanan PR, Naagarazan RS.
combustion engine at 1400 rpm. Int J Hydrogen Energy
Mathematical modeling for electric discharge machining of
2005;30:77e82.
aluminiumesilicon carbide particulate composites. J Mater
[8] Sahoo BB, Sahoo N, Saha UK. Effect of parameters and type of
Process Technol 1999;87:59e63.
gaseous fuel on the performance of dual-fuel gas diesel
[26] Saidur R, Jahirul MI, Hasanuzzaman M, Masjuki HH. Analysis
engines- A critical review. Renew Sustain Energy Rev
of exhaust emissions of natural gas engine by using response
2009;13:1151e84.
surface methodology. J Appl Sci 2008;19:3328e39.
[9] Porpatham E, Ramesh A, Nagalingam B. Effect of hydrogen
[27] Muhammad WM, Ahmad A, Farooq A, Hamid M,
addition on the performance of a biogas fuelled spark
Muhammad AR, Umer R. Response surface methodology: an
ignition engine. Int J Hydrogen Energy 2006;32:2057e65.
emphatic tool for optimized biodiesel production using rice
[10] Karim GA. Combustion in gas fueled compression: ignition
bran and sunflower oils. Energies 2012;5:3307e28.
engines of the dual fuel type. J Eng Gas Turbines Power
[28] Liew C, Li H, Nuszkowski J, Liu S, Gatts T, Atkinson R. An
2003;125:827e36.
experimental investigation of the combustion process of a
[11] Senthil Kumar M, Ramesh A, Nagalingam B. Use of hydrogen
heavy-duty diesel engine enriched with H2. Int J Hydrogen
to enhance the performance of a vegetable oil fuelled
Energy 2010;35:11357e65.
compression ignition engine. Int J Hydrogen Energy
[29] Montgomery DC. Design and analysis of experiments. India:
2003;28:1143e54.
Wiley; 2007.
[12] Saravanan N, Nagarajan G. Performance and emission
[30] Myers, Montgomery DC, Anderson-Cook. Response surface
studies on port injection of hydrogen with varied flow rates
methodology. New York: John Wiley; 2009.
with diesel as an ignition source. Appl Energy
2010;87:2218e29.