Optimization of application of 2ethylhexylnitrate on partial substitution of ethanol in CI engine for fuel economy and emission control using MADM method
 166 Downloads
Abstract
The present study is performed to identify the blending proportion of cetane improver for partial substitution of ethanol with diesel to achieve fuel economy and emission control. This is case of multiple attribute decision making problem which is solved using Taguchi GRA. Taguchi method is used to decide the proportion of blend in fuel sample to achieve maximum improvement with minimum number of fuel sample and grey relation analysis is done to identify the significant factor along with optimize fuel proportion. Using Taguchi method, nine different fuel samples containing mixture of ethanol, 2 ethyl hexylnitrates and diesel were prepared. Each sample was tested on computerized single cylinder CI engine test rig at constant speed and variable compression ratio at 16, 17 and 18 and variable load conditions. After Grey relation analysis of performance and emission results of experimentation, it is observed that D80E20EH3 by ranking method and D80E10EH3 by mean method are optimum fuel sample sequence. Confirmatory experimentation also shows that experimental GRG is close to predicted GRG of 0.8387 hence optimum sequences is acceptable. The optimum sequence shows improvement in fuel economy and lesser NOx emission with little increases in smoke.
Keywords
Ethanol Cetane improver MADM Taguchi GRA CI engineAbbreviations
 CR
Compression ration
 BSFC
Brake specific fuel consumption
 NOx
Nitrogen oxide
 PM
Particulate matter
 CR
Compression ration
 kW
Kilo watt
 CO
Carbon monoxide
 HC
Hydro carbon
 ANNOVA
Analysis of variance
 2EHN
2 Ethyl hexyl nitrate
 DF
Degree of freedom
1 Introduction
Ethanol is getting widespread acknowledgement as renewable alternative fuel for diesel and as oxygenate providing potential to reduce particulate emissions [1, 2]. With ethanol blend of up to 5–10% in diesel, smoke density greatly reduces with higher percentage while NOx show little or no reduction compared to neat diesel [3]. The better atomization and vaporization of fuel is achieved with use of 15% ethanol due to its lower density and NO emission of the engine run on the is also found to be lower than that of diesel due to the higher value of latent heat of vaporization [4]. The work on ethanolmethyl soyate blend with diesel (5:20:75) for heavy duty engine shows significant reduction in PM emission while 2–14% rise in NOx emission without changes in CO emission as it is more dependent on operating conditions [5]. This work on blending of nbutanol with diesel shows that, smoke formation is lower, CO formation is lower at all load condition except at higher load while HC formation is high at all load condition except at low load condition [6, 7].
With the increased in ethanol results in Uses of ethanol lower the cetane value of Diesel which lend result in increase in ignition delay and decrease in total combustion duration [8]. Study on use of bioethanol at 5%, 10% and 15% with diesel shows longer ignition delay by 4.4% at full load condition than that for pure diesel. 5% Ethanol emulsion gives better performance and lower emission than 10% and 15%. NOx and smoke were reduced by 4% and 20% [9].
Uses of ethanol lower the cetane value of Diesel fuel which has undesirable effect on combustion properties. Low cetane number value generally has a tendency to exhibit longer ignition delay due to their ignition quality. This can be improved by adding small amount of ignition improvers or cetane number improvers. Examples of cetane improvers are organic peroxides, nitrates, nitrites and various sulphur compounds. Earlier, alkyl nitrates isopropyl nitrate, primary amyl nitrates, primary hexyl nitrates, octyl nitrate were commercially used [10, 11]. THC and CO slightly increases for ethanol–diesel blend but addition of cetane improver increases Total Hydrocarbon (THC) conversion efficiency up to 40–60% [12, 13, 14]. An ignition improver of up to 7% by volume would normally be required to enable the ignition of alcohol fuels in CI engines [15]. Due to high their cost, cetane improver application is limited and it also makes the fuel expensive in the present study possibility of use of DoE and its application in identify the optimize percentage of 2EHN and Ethanol is investigated with load conditions.
2 Design of experiment
Design factors and their levels
Design factors  Level 1  Level 2  Level 3 

C.R  16  17  18 
Load (kg)  0  4.5  9 
Ethanol (%)  10  15  20 
2EHN (%)  3  5  7 
2.1 Selection of orthogonal array
Before selecting the orthogonal array the minimum no. of experiment to be conducted shall be fixed based on the total no. of degree of freedom. In counting the total degree of freedom the investigator commits 1 degree of freedom to the overall mean of the response under study. So in our experiment, we have selected L9 array, because we have 3 levels and 4 factors i.e. compression ratio (C.R.), load, ethanol, 2EHN. That’s why we selected L9 array and hence selected orthogonal array shall have 9 run.
Array selector table [19]
Allocation of variable factor in L9 array
Run no.  A  B  C  D  Code 

1  1  1  1  1  D80E10EH3 
2  1  2  2  2  D80E15EH5 
3  1  3  3  3  D80E20EH7 
4  2  1  2  3  D80E15EH7 
5  2  2  3  1  D80E20EH3 
6  2  3  1  2  D80E10EH5 
7  3  1  3  2  D80E20EH5 
8  3  2  1  3  D80E10EH7 
9  3  3  2  1  D80E15EH3 
L9 orthogonal design matrix for experimental data
Run no.  Factor  Response  

CR  Load (Kg)  Ethanol (ml)  2EHN (ml)  NO_{x} (ppm)  Smoke (%)  BSFC (kg/kWh)  
1  16  0  80  24  –  –  – 
2  16  4.5  120  40  –  –  – 
3  16  9  160  56  –  –  – 
4  17  0  120  56  –  –  – 
5  17  4.5  160  24  –  –  – 
6  17  9  80  40  –  –  – 
7  18  0  160  40  –  –  – 
8  18  4.5  80  56  –  –  – 
9  18  9  120  24  –  –  – 
2.2 Sample preparation
Details composition of fuel sample (ml)
Run no.  Samples  Diesel  Ethanol  2EHN 

1  D80E10EH3  800  80  24 
2  D80E15EH5  800  1200  40 
3  D80E20EH7  800  160  56 
4  D80E15EH7  800  120  56 
5  D80E20EH3  800  160  24 
6  D80E10EH5  800  80  40 
7  D80E20EH5  800  160  40 
8  D80E10EH7  800  80  56 
9  D80E15EH3  800  120  24 
3 Experimentation
3.1 Fuel sample preparation
There are multiple additives available for diesel fuel. Oxygenate additives is used to improve the oxygen content of diesel fuel during combustion while cetane improver improves cetane number hence burning capacity of fuel. During literature survey it is found that ethanol is getting popularity as alternate fuel/additive for diesel while 2EHN is most commonly used cetane improver [22, 23, 24].
3.2 Experimental setup
Physiochemical key properties of additives
Property  Diesel  Ethanol  2EHN 

Density at 150 °C (Kg/m^{3})  832  778.87  963 
Kinematic viscosity @ 40 °C (cst)  4.70  1.08  1.78 
Calorific value (MJ/kg)  42.49  29.02  29.86 
Flash point (°C)  53  12.78  27 
Fire point (°C)  58  16.60  76 
Cloud point (°C)  − 2  5  4.44 
Pour point (°C)  − 5  5  4.44 
Engine test rig specification
Parameter  Specification 

Engine type  Single cylinder DI, water cooled 
No. of cylinder  01 
Rated power  5.20 kW, 1500 rpm 
Displacement  661.45 cc 
Bore*stroke length  87.50 mm*110 mm 
Compression ratio  18:1 
Orifice diameter  20 mm 
Orifice coefficient of discharge  0.60 
No. of cycle  10 
Dynamometer arm length (mm), hydraulic  185 mm 
Results of Experimentation
Run no.  BSFC (kg/kWhr)  NOx (ppm)  Smoke (%) 

1  1.71  360  0.1 
2  0.58  263  0.5 
3  0.39  288  2.4 
4  1.3  660  0.2 
5  0.53  184  0.5 
6  0.39  256  3.1 
7  1.1  570  0.6 
8  0.54  173  1.6 
9  0.38  295  4.7 
4 Multiple attribute decision making (MADM)
There are many multiple attribute decision making (MADM] problems we face in our day to day life. MADM aims to select the best from the existing ‘‘alternatives’’ by considering multiple ‘‘attributes’’ which are frequently in conflict with each other. There are several common techniques for solving MADM problems, such as simple additive weighting (SAW), Technique for order preference by similarity to ideal solution (TOPSIS), analytical hierarchy process (AHP), data envelopment analysis (DEA) and Grey Relation Analysis.
Grey relational analysis (GRA) is part of grey system theory, which is useful for solving problems with complicated interrelationships between multiple factors and variables. GRA has been successfully used in solving a variety of MADM problems in past.
GRA solves MADM problems by combining the entire range of performance attribute values while being considered for every alternative into one, single value. This reduces it to a single attribute decision making problem. Therefore, alternatives with multiple attributes can be compared easily after the GRA process [25].
4.1 Normalizing/grey relation generation
In GRA method, raw data of response parameter were normalized between 0 and 1 with smaller the better characteristic equation no 2. This is called grey relational generation. Here ‘1’ being most deserved parameter [22, 26, 27, 28, 29].
Grey relation generation of each response variance
Run no.  BSFC  NOx  Smoke 

1  0.000  0.616  1.000 
2  0.850  0.815  0.913 
3  0.992  0.764  0.500 
4  0.308  0.000  0.978 
5  0.887  0.977  0.913 
6  0.992  0.830  0.348 
7  0.459  0.185  0.891 
8  0.880  1.000  0.674 
9  1.000  0.749  0.000 
4.2 Deviation sequence
Deviation sequence of each response variance
Run no.  BSFC Δ_{1}  NOx Δ_{2}  Smoke Δ_{3} 

1  1.000  0.384  0.000 
2  0.150  0.185  0.087 
3  0.008  0.236  0.500 
4  0.692  1.000  0.022 
5  0.113  0.023  0.087 
6  0.008  0.170  0.652 
7  0.541  0.815  0.109 
8  0.120  0.000  0.326 
9  0.000  0.251  1.000 
4.3 Grey relational coefficient

ξ = Distinguishing coefficient (0 < ξ < 1)
(0.5 is value used in most situations)

Δ _{min} = Smallest value of y_{0}(k) − yi(k)

Δ _{max} = Largest value of y_{0}(k) − yi(k)
Grey relation coefficient of each output parameter
Run no.  NO_{X}  Smoke  BSFC 

1  0.333  0.566  1.000 
2  0.769  0.730  0.852 
3  0.985  0.679  0.500 
4  0.420  0.333  0.958 
5  0.816  0.957  0.852 
6  0.985  0.746  0.434 
7  0.480  0.380  0.821 
8  0.806  1.000  0.605 
9  1.000  0.666  0.333 
4.4 Grey relational grade
Grey relation grade is calculated using Grey Relation Coefficient. The higher the value of GRG is the greater is the desirability.
W_{p} = weighting value for each grey relation coefficient ranging from 0 to 1 and sum of W_{p} is always 1. In present study since we want to control each response equally important so are assign weighting factor as 0.33 [30, 31].
Grey Relation grade calculated by above formula is tabulated as below. The higher grade correspond to a better S/N ration respectively as it’s is closer to computed ideal S/N Ration.
4.5 Estimation of mean grey relational grade
Grey relational grades with rank
Run no.  Grey relational coefficient  Grey relation grade G = 1/3(a + b + c)  Rank  

NO_{x} (a)  Smoke (b)  BSFC (c)  
1  0.333  0.566  1.000  0.633  7 
2  0.769  0.730  0.852  0.784  3 
3  0.985  0.679  0.500  0.721  5 
4  0.420  0.333  0.958  0.570  8 
5  0.816  0.957  0.852  0.875  1 
6  0.985  0.746  0.434  0.722  4 
7  0.480  0.380  0.821  0.561  9 
8  0.806  1.000  0.605  0.804  2 
9  1.000  0.666  0.333  0.667  6 
Response table for grey relational grade
Factor/level  CR  Load  Ethanol  2EHN 

1  0.7127  0.5880  0.7195  0.7248 
2  0.7223  0.8207  0.6735  0.6886 
3  0.6770  0.7032  0.7190  0.6986 
Max effect (Max–Min]  0.0453  0.2328  0.0460  0.0362 
Rank  3  1  2  4 
Total mean GRG = γ _{mean} =  0.7040 
4.6 Confirmatory experiment for grey relational analysis
Results of confirmatory experiment
Initial factor setting  Optimal condition  

Predicted  Experimental  
Level  CR1 L1 Eth1 EH1  CR2 L2 Eth1 EH1  CR2 L2 Eth1 EH1 
CR  16  17  
Load  0  4.5  
Ethanol  10  10  
2EHN  3  3  
BSFC  1.71  1.67  
NOx  360.00  350.00  
Smoke  0.10  0.11  
GRG  0.6330  0.8387  0.8155 
5 ANNOVA for GRG of responses
The general linear model procedure is used to conduct an ANOVA tests the hypothesis that the means of several populations are equal. During ANNOVA for GRG of response, it is observed that, F factor and p value results are not possible due to limitation of degree of freedom as denominator of F is zero, Hence we need to skip one of factor during ANNOVA. It is observed that factor ‘Load’ is continuous variable and can’t be kept constant for optimum operation of engine. Hence it was skipped during ANNOVA analysis [32].
Results of ANNOVA
Source  DF  Seq SS  Adj SS  Adj MS  F  p 

CR  2  0.5343  0.5343  0.2671  0.04  0.959 
Ethanol  2  0.6257  0.6257  0.3128  0.05  0.953 
2EHN  2  0.3100  0.3100  0.1550  0.02  0.976 
Residual error  2  12.6364  12.6364  6.3182  
Total  8  14.1064 
6 Interaction analysis for factor
6.1 Effect of factor interaction on GRG
7 Normality analysis of response
We generate a normal probability plot and perform a hypothesis test to examine whether or not the observations follow a normal distribution.

H0: data follow a normal distribution vs.

H1: data do not follow a normal distribution.
The vertical scale on the graph resembles the vertical scale found on normal probability paper. The horizontal axis is a linear scale. The line forms an estimate of the cumulative distribution function for the population from which data are drawn. A onesample hypothesis test to determine whether the population from which you draw your sample is nonnormal. Many statistical procedures rely on population normality, and using a normality test to determine whether to reject this assumption can be an important step in your analysis. The null hypothesis for a normality test states that the population in normal. The alternative hypothesis states that the population is nonnormal. To determine whether your sample data come from a nonnormal population, you can choose from four tests. We have performed Normality test using Anderson–Darling test. This test compares the empirical cumulative distribution function of your sample data with the distribution expected if the data were normal. If this observed difference is sufficiently large, the test will reject the null hypothesis of population normality. We can test normality by two way, one if the p value of these test is less than your chosen alevel, you can reject your null hypothesis and conclude that the population is nonnormal. And secondly using graph. If the observation follows straight line it indicate it follow normal distribution and population is nonnormal.
7.1 Normality test for BSFC
7.2 Normality test for NOx
The reason for this may be uneven temperature distribution during the combustion process and also consideration of combustion at various load conditions.
7.3 Normality test for smoke
7.4 Normality test for GRG
8 Regression analysis
Regression analysis is used to investigate and model the relationship between a response variable and one or more predictors (factors).There are different methods of regression analysis and selection is dependent on type of response. Both generalized linear models and least squares regression investigate the relationship between a response variable and one or more predictors. A practical difference between them is that generalized linear model techniques are used with categorical response variables, and linear regression techniques are used with continuous response variables. Both generalized linear model techniques and least squares regression techniques estimate parameters in the model so that the fit of the model is optimized. Least squares minimizes the sum of squared errors to obtain maximum likelihood estimates of the parameters, whereas generalized linear models obtain maximum likelihood estimates of the parameters using an iterativereweighted least squares algorithm [33]. In many cases, the differences between the LS and MLE results are minor, and the methods can be used interchangeably [31].
As observed in normality test, response not following exact normal distribution, hence BoxCox transformation is used to make our data approximately normal so that we can complete analysis. After regression analysis, we get following equations using optimal λ Box–Cox transformation.
9 Conclusion
The higher the GRG is the closer will be output value to the ideal value. Thus, higher GRG is desired for optimum performance. Therefore, the optimal parameters setting for better BSFC and lesser NOx and Smoke are (A2B2C3D1) for run no 5 as presented in Table 11 highest grey relational grade.
GRA can also be used to separate the effect of each factor and identify the significance of individual factor and its optimum level using mean GRG. As seen in Table 12, we get optimum response when CR is 17, Load at 4.5 kg, Ethanol with 10%.
As per confirmatory experimentation as shown in Table 13, It is observed that our predicated optimum sequence GRG is 0.8155 which quite close to predicated GRG i.e. 0.8387. Compared with initial setting of sequence at level 1, our optimum sequence gives better Fuel economy and lesser NOx emission with little increases in smoke.
Notes
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
References
 1.Manimaran R, Gopalakrishnan V, Senthil Kumar R (2013) Experimental investigation on D. I. diesel engine fuelled by ethanol diesel blend with varying inlet air temperature. Adv Appl Sci Res 4:319–329Google Scholar
 2.Wang Y, Li S, Roskilly AP, Yu H, Li H, Huang J (2009) Experimental investigation on the performance and emissions of a diesel engine fuelled with ethanol–diesel blends. Appl Therm Eng 29:2484–2490CrossRefGoogle Scholar
 3.Canakci M, Turkcan A (2011) Combustion characteristics of an indirect injection (IDI) diesel engine fueled with ethanol/diesel and methanol/diesel blends at different injection timings. Sustain Transp (ST) 57:3565–3572Google Scholar
 4.Zhen H, Lu X, Zhang Wg, Yang Jg, Li Dg (2005) Physicochemical properties of ethanol–diesel blend fuel and its effect on performance and emissions of diesel engines. Renew Energy 30:967–976CrossRefGoogle Scholar
 5.Taji SG, Patil AR (2015) Enhancement of CI engine performance and emission by use of composite additive. Int J Eng Manag Res (IJEMR) 17–23 (special issue)Google Scholar
 6.Murugan S, Das LM, Hansdah D (2013) Experimental studies on a DI diesel engine fueled with bioethanol–diesel emulsions. Alex Eng J 52:267–276CrossRefGoogle Scholar
 7.Mahapatra SS, Hansdah D, Sivalingam M (2015) Validation of some engine combustion and emission parameters of a bioethanol fuelled DI diesel engine using theoretical modelling. Alex Eng J 54:993–1002CrossRefGoogle Scholar
 8.Dandakouta H, Yahuza I (2015) A performance review of ethanol–diesel blended fuel samples in compressionignition engine. J Chem Eng Process Technol 6:1–6Google Scholar
 9.Rakopoulos CD, Kakaras EC, Giakoumis EG, Rakopoulos DC (2008) Effects of ethanol–diesel fuel blends on the performance and exhaust emissions of heavy duty DI diesel engine. Energy Convers Manag 49:3155–3162CrossRefGoogle Scholar
 10.Choi B, Kima H (2008) Effect of ethanol–diesel blend fuels on emission and particle size distribution in a commonrail direct injection diesel engine with warmup catalytic converter. Renew Energy 33:2222–2228CrossRefGoogle Scholar
 11.Panga X, Mua Y, Hea H, Shuaib S, Wangb J, Chenb H, Li R, Shia X (2006) Emission reduction potential of using ethanol–biodiesel–diesel fuel blend on a heavyduty diesel engine. Atmos Environ 40:2567–2574CrossRefGoogle Scholar
 12.Vijay P, Venkat Ram Puli R, Dadi Y, Pavan Nihanth Chaluvadi CHVVN (2013) Diesel engine performance improvement by using cetane improver using cetane improver. Int J Eng Innov Technol (IJEIT) 2(10):179–182Google Scholar
 13.Parlak A, Bağci E, Aydin Z, Kökkülünk G (2014) Application of Taguchi Methods for the optimization of factors affecting engine performance and emission of exhaust gas recirculation in steaminjected diesel engines. Acta Polytech Hung 11:95–107Google Scholar
 14.Won Kim J, Sik Lee C, Hyung Lee J, Yoon Kim M (2006) Effect of compression ratio and spray injection angle on HCCI combustion in a small DI diesel engine. Energy Fuels 20(1):69–76CrossRefGoogle Scholar
 15.Nagarajan G, Sampath S, Saravanan S (2012) Application of Taguchi’s orthogonal array in multi response optimization of NOx emission of crude rice bran oil methyl ester blend as a CI engine fuel. Open J Optim 1:25–33CrossRefGoogle Scholar
 16.Jindal TK, Valli DM (2014) Application of Taguchi method for optimization of physical parameters affecting the performance of pulse detonation engine. J Basic Appl Eng Res 1:18–23Google Scholar
 17.Ross PJ (2014) Taguchi techniques for quality improvement. Tata MacGraw Hill, New York CityGoogle Scholar
 18.Yang T, Huang GW, Kuo Y (2008) The use of grey relational analysis in solving multiple attribute decisionmaking problems. Comput Ind Eng 55:80–93CrossRefGoogle Scholar
 19.Tabatabaei SA, Amiri G, Hasani H (2012) Grey relational analysis to determine the optimum process parameters for openend spinning yarns. J Eng Fibers Fabr 7:81–86Google Scholar
 20.Pohit G, Misra D (2013) Optimization of performance and emission characteristics of diesel engine with biodiesel using greyTaguchi method. J Eng. https://doi.org/10.1155/2013/915357 CrossRefGoogle Scholar
 21.Tripathy DK, Tripathy S (2016) Multiattribute optimization of machining process parameters in powder mixed electrodischarge machining using TOPSIS and grey relational analysis. Eng Sci Technol Int J 19:62–70CrossRefGoogle Scholar
 22.Taji SG, Patil AR (2014) Investigation on effect of oxygenated additive on multicylinder diesel engine performance and emission. Int J Innov Res Sci Eng Technol 3(6):14113–14120Google Scholar
 23.Taji SG, Patil AR (2013) Need of composite additives for diesel fuel : a review. Int J Eng Res Technol (IJERT) 2(11):2652–2656Google Scholar
 24.Kuszewski H (2018) Effect of adding 2ethylhexyl nitrate cetane improver on the autoignition properties of ethanol–diesel fuel blend—investigation at various ambient gas temperatures. Fuel 224:57–67CrossRefGoogle Scholar
 25.Sivaramakrishnan K (2017) Investigation on performance and emission characteristics of a variable compression multi fuel engine fuelled with Karanja biodiesel–diesel blend. Egypt J Petrol 27(2):177–186CrossRefGoogle Scholar
 26.Arunachalam VP, Dhandapani N, Nataraj M (2005) Optimizing diesel engine parameter for low emission useing taguchi method: varition risk analysis approch—Part I. Indian J Eng Mater Sci 12:169–181Google Scholar
 27.DULARI HANSDAH, “Phd Thesis _Experimental Studies On Partial Substitution Of Diesel With Bioethanol (Derived From Madhuca Indica Flowers) Using Different Techniques,”,ROURKELA769008 (INDIA), 2015Google Scholar
 28.Das AK, Banerjee R, Roy S (2014) Application of GreyeTaguchi based multiobjective optimization strategy to calibrate the PMeNHCeBSFC tradeoff characteristics of a CRDI assisted CNG dualfuel engine. J Nat Gas Sci Eng 21:524–531CrossRefGoogle Scholar
 29.Tripathy DK, Tripathy S (2015) Multiattribute optimization of machining process parameters in powder mixed electrodischarge machining using TOPSIS and grey relational analysis. Eng Sci Technol Int J 19:62–70CrossRefGoogle Scholar
 30.Pandey PM, Venkateswara Rao P, Patel KM (2010) Optimisation of process parameters for multiperformance characteristics in EDM of Al_{2}O_{3} ceramic composite. Int J Adv Manuf Technol 47(9–12):1137–1147Google Scholar
 31.Tandale Sangram MS, Jadhav D (2016) Multiobjective performance optimization of compression ignition engine operated on mangifera indica biodiesel by applying Taguchi Grey relational analysis. Waste Biomass Valor 7(5):1309–1325CrossRefGoogle Scholar
 32.Tandale MS, Jadhav SD (2015) Biodiesel (mangifera oil methyl ester) derived from triglycerides of mangifera kernel seed and leaves oil by using heterogeneous catalyst. SAE Tech Pap 1:1682Google Scholar
 33.McCullagh P, Nelder JA (1992) Generalized linear model. Chapman & Hall, Boca RatonzbMATHGoogle Scholar