Optimization of CI Engine Performance and Emissions Using Alcohol–Biodiesel Blends: A Regression Analysis Approach

Abstract

This research paper investigates the optimum engine operating parameters, namely engine load, palm biodiesel, and ethanol percentage, by using a regression analysis approach. The study was conducted on a single-cylinder, four-stroke diesel engine at varying engine loads and constant speed. A general full factorial design was established using Minitab software (Version 17) for three different input factors with their varying levels. The test results based on the regression model are used to optimize the engine load and percentages of palm biodiesel and ethanol in diesel–biodiesel–ethanol ternary blends. The analysis of variance (ANOVA) revealed a significant effect on performance and emission parameters for all three factors at a 95% confidence level. From the regression study, optimum brake thermal efficiency (BTE), nitrogen oxide (NO_x), carbon monoxide (CO), and unburnt hydrocarbon (UHC) emissions were found to be 12.57%, 436.2 ppm, 0.03 vol.%, and 79.2 ppm, respectively, at 43.43% engine load, 11.06% palm biodiesel, and 5% ethanol share. The findings of this study can be used to optimize engine performance and emission characteristics. The regression analysis approach presented in this study can be used as a tool for future research on optimizing engine performance and emission parameters.

1. Introduction

The rapid decline in fossil fuel stocks, hikes in crude oil import bills, and strict policies on emission regulations are the major cruces that have threatened the future development of the modern world. Transportation and power generation are the two major areas where the application of diesel engines is predominant. Diesel engines are widely used worldwide for many transport applications like buses, trucks, diesel locomotives, marine engines, etc., which make global transportation much easier and more convenient. The diesel engine is supreme, and it provides high fuel economy but at the cost of very high NO_x and particulate matter (PM) emissions [1]. In recent years, automotive industries have been facing a twin crisis of fossil fuel depletion and engine emissions, which threatened engine manufacturers to a great extent. Hence, renewable fuels’ importance is growing as a promising sustainable energy resource [2]. Global warming and other carbon footprint issues due to diesel combustion have encouraged the use of bio-based alternative fuels in diesel engines without any engine modification. Based on the high turnover per year, rapeseed [3], soybean [4,5], palm [6], jatropha [7] oil, etc. have become some of the major biodiesel feedstocks. However, higher kinematic viscosity, density, pour, and flashpoint of biodiesel deteriorate the engine combustion. To overcome these difficulties, many researchers have introduced methanol, ethanol, octanol, pentanol, and diethyl ether (DEE) as an additive [8,9,10]. Yasin et al. [11] experimentally investigated the performance of a palm biodiesel blend in a diesel engine and observed a 4.7% increase in NO_x and a 3.5% decrease in CO emissions compared to diesel. However, they also mentioned that the use of exhaust gas recirculation (EGR) can reduce NO_x emission by 22% during engine operation with a palm biodiesel blend. Appavu et al. [12] observed better engine performance and lower engine emissions while operating palm biodiesel (PB100) in an unmodified direct-injection diesel engine. They found 23, 24, 39, and 5% lower CO, HC, smoke, and NO_x emissions, respectively, compared to diesel but at the cost of higher fuel consumption. Ma et al. [13] performed a study where they experimentally compared diesel and biodiesel operations with ethanol and pentanol blended ternary fuels at varying engine speeds. They observed higher indicated thermal efficiency for diesel–biodiesel–ethanol ternary blends compared to baseline diesel and biodiesel operation, irrespective of engine speed. Using non-edible biodiesel and ethanol, Sathish et al. [14] experimentally found that biodiesel–ethanol, diesel–ethanol, and diesel–biodiesel–ethanol blends resulted in higher BTE and lower engine emissions compared to diesel. Devarajan et al. [15] experimentally investigated the combustion performance of octanol, palm biodiesel, and diesel ternary blends and observed earlier and smooth combustion compared to palm biodiesel blend operation. They observed higher BTE and low fuel consumption during operation with ternary blends. The better performance could be accounted for by the reduction in viscosity due to octanol addition, which helps in better atomization of fuel that leads to faster combustion. Alcohol, like n-butanol, is effective in increasing the BTE and can decrease NO_x emissions by 20–60% when used in a ternary blend [16]. Uslu and Aydin [17] investigated DEE addition in palm biodiesel–diesel blends and observed that lower DEE and palm biodiesel fractions help in improving BTE and fuel economy. They also reported that engine input variables like advance injection, engine load, and DEE percentage in the blends need to be optimized to obtain better performance and emissions.

Conventional methods of engine experiments are considered cost-ineffective and take much time. These drawbacks can be overcome by using new computational techniques by optimizing the working parameters. To obtain a trade-off between engine performance and emission characteristics, optimization of engine parameters is one of the prime choices [18]. In addition to conventional engine experiments, the prediction and optimization of experimental design by using soft computational tools may encourage the making of optimal decisions on engine operating parameters. Many techniques, like response surface methodology (RSM) [19,20], artificial neural network (ANN) [21], genetic algorithm [22], etc., are used for the prediction of input variables like load, injection pressure, start of injection, biodiesel proportion, etc. From the viewpoint of multi-objective problems [23], engine manufacturers are concerned about predicting responses that are required for selecting the optimum design parameters. RSM results in a better combination in terms of improving performance and reducing emissions with the lowest prediction errors. Using the RSM technique, Singh et al. [24] optimized BTE, UHC, and NO_x emissions with error values of 2.4, 4.95, and 0.93%, respectively.

From the above studies, it was noted that compared to biodiesel blends, better engine performance and emissions have been observed using ternary blends. In the previous literature, studies on the trade-off between engine performance and emission parameters in the optimal engine operating range using optimization techniques are rare. None of the studies mentioned in the literature refer to the application of the regression approach on palm biodiesel. Also, no previous study has investigated the optimization using the regression technique for diesel–palm biodiesel–ethanol ternary blends. This motivates us to explore a general full-factorial-design-based regression model to quantify the optimum engine input parameters for a better performance–emission balance. This present work aims to optimize engine operating parameters for optimum performance and emission characteristics using a regression model. For this, three factors, namely engine load, palm biodiesel, and ethanol percentage at different levels, are considered as the input factors for the investigation. A non-linear regression model was developed for BTE, NO_x, CO, and UHC emissions. Significant effects of linear, square, and interaction terms of all three factors are investigated by ANOVA analysis. Finally, using a regression model, the optimization of performance–emission characteristics was evaluated.

2. Materials and Methods

2.1. Engine Setup

The engine that was used in the experimental investigation is a single-cylinder, four-stroke, water-cooled, DI (direct injection) computerized diesel engine. The schematic diagram and full specifications of the engine are shown in Figure 1 and

Table 1. Engine specifications.

Parameters	Specifications
Engine	1-C, 4-S, VCR (variable compression ratio) diesel engine
Product code	234
Dynamometer	Eddy-current-type
Cooling type	Water-cooled
Data acquisition system	NI USB-6210, 16-bit, 250 kS/s
Crank angle sensor	1° resolution, speed 5500 rpm with TDC pulse
Temperature sensor	RTD, PT100, and K-type thermocouple
Load sensor	Strain-gauge-type (range 0–50 Kg)
Rotameter	Engine cooling (40–400 LPH); calorimeter (25–250 LPH)
Stroke	110 mm
Bore	87.5 mm
Displacement	661 cc
Compression Ratio	17.5:1
Output power	3.5 kW
Speed	constant 1500 rpm
Fuel Injection pressure	220 bar

. An eddy current dynamometer at a constant speed of 1500 rpm connected to a speed sensor was coupled to the engine crankshaft to measure the outputs from the engine. The engine speed for every 1° crank angle was measured using a crank angle sensor (make: Kubler) fixed with the crankshaft. The engine was connected to a data acquisition (DAQ) system comprising a computer with a crank angle encoder and graphical user interface (GUI)-based Engine Soft post-processing software (Version 9.0) [25]. The DAQ system that was installed was designed to measure cylinder pressure and temperature at every 1° crank angle interval. A piezoelectric transducer was installed at the top of the engine cylinder head to measure in-cylinder gas pressure. All the measurements of exhaust gas temperature (EGT), cooling water outlet and inlet temperature, and performance data are reported over a period of time. The whole computerized system was then connected with the engine setup using the NI lab view centralized data acquisition system (NI USB-6210 Bus Powered M Series) interfaced with “Engine soft” software. Exhaust gases were measured using a 5-gas analyzer (Make: AVL India; model: 444) fitted with a Digas sampler to measure the NO_x, UHC, CO, CO₂, and O₂.

2.2. Experimental Methodology and Fuel Preparation

The present experimental investigation was performed at varying engine loads from 20 to 100%. For the present study, diesel and ethanol were purchased from a local vendor, while palm biodiesel was prepared in the lab, as shown in Figure 2. Different proportions of diesel, palm biodiesel, and ethanol were mixed for the preparation of ternary blends. For the preparation of different ternary blends, the mix proportions of diesel, palm biodiesel, and ethanol were varied between 70 and 90%, 5 and 20%, and 5 and 10%, respectively. For the total volume of 100%, the proportions of diesel, palm biodiesel, and ethanol were accurately measured using a measuring cylinder. Finally, the entire mixture was stirred to make a homogeneous mix of the blends before running the engine. The blends are denoted by D90B5E5 where D, B, and E stand for diesel, palm biodiesel, and ethanol, respectively, and the subsequent number shows their respective volumes in percentages. The physicochemical properties of different blends are listed in

Table 2. Properties of the different diesel, palm biodiesel, and ethanol blends.

Samples	Density at 20 °C (kg/m3)	Cetane Number	Kinematic Viscosity at 40 °C (cSt)	Calorific Value (kJ/kg)	Flashpoint (°C)
	ASTM D-1298	ASTM D-613	ASTM D-445	ASTM D-240	ASTM D-93
Diesel	836	49	2.45	42,800	100
Palm Biodiesel	925	62	4.56	39,849	167
Ethanol	789	8	1.09	29,700	16.60
D90B5E5	838.1	47.6	2.49	41,998	99.2
D85B10E5	842.6	48.3	2.59	41,850	102.8
D80B15E5	847	48.9	2.7	41,702	105.9
D75B20E5	851.5	49.6	2.8	41,555	109.2
D85B5E10	835.8	45.6	2.42	41,343	95
D80B10E10	840.2	46.2	2.53	41,195	98.4
D75B15E10	844.7	46.9	2.63	41,047	101.7
D70B20E10	849.1	47.5	2.7	40,900	105.1
% measurement uncertainty	±0.3	±0.15	±0.22	±0.75	±0.2

. The fuel consumption was measured using a fuel burette (12.4 mm diameter) for an interval of 60 s. Ambient temperature and relative humidity during the tests were recorded as 27 °C and 60%, respectively. Before taking the reading, the engine was allowed to run for 10–15 min to come to a steady condition. For each ternary blend, engine operation was performed at 20, 40, 60, 80, and 100% load. For each individual blend, the load was varied from 20 to 100% using the engine control panel. After being set to a particular load, the engine was allowed to run a minimum of five minutes to take the reading of engine performance for 60 s of fuel consumption, and the same procedure was repeated for the other blends. For each blend at different load conditions, NO_x, CO, and HC emissions were recorded five times and their average was taken.

2.3. Uncertainty of Measurement

The purpose of the uncertainty measurement is to evaluate the quality of the experimental readings obtained from any measurements. Providing an exact count of the errors in the measurements, uncertainty analysis is very important in meeting the standard quality of explanation. Uncertainty analysis gives a proper explanation of the repeatability of the investigations. By using the root mean square (RMS) method, the total uncertainty of the engine performance parameters is calculated. The total percentage uncertainty of the computed performance parameters is listed in

Table 3. Total uncertainty analysis of performance parameters [21].

Performance Parameter	Measured Variables	Instrument Involved in the Measurement	% Uncertainty of the Instrument	Calculation	Total % Uncertainty of Parameters
BP	Load, RPM	Load sensor, load indicator, speed measuring unit	0.2, 0.1, 0.9	$\sqrt{{0.2}^2{+0.1}^2+{0.9}^2}$	0.9
BSFC	SFC (Liquid Fuel), BP	Fuel measuring unit, fuel flow transmitter, as For BP measurement	0.05, 1.5, 0.92	$\sqrt{{0.05}^2{+1.5}^2{+0.92}^2}$	1.8
BSEC	SFC (Liquid Fuel), BP	As for SFC measurement, as for BP measurement	1.84, 0.92	$\sqrt{{1.84}^2+{0.92}^2}$	2

. Total percentage uncertainty was calculated using Equation (1), where $∆U$ is total uncertainty, and $∆x_1, ∆x_2, ∆x_3\dots \dots ∆x_n$ are the errors of$x_1,{ x}_2,{ x}_3{\dots \dots x}_n$. The accuracy of the emission measuring instrument is shown in

Table 4. Accuracy of the emission measuring instrument (AVL DIGAS 444-5 gas analyzer) [26].

Measured Parameter	Measurement Principle	Measuring Range	Resolution	Accuracy	% Uncertainty in Sampling
CO	NDIR	0–10% vol.	0.01% vol.	<0.6% vol.: ±0.03% vol.; ≥0.6% vol.: ± 5% of value	±0.2 ± 0.3
CO2	NDIR	0–20% vol.	0.1% vol.	<10% vol.: ±0.5% vol.; ≥10% vol: ±5% of value	±0.15 ±0.2
HC	NDIR	0–20,000 ppm vol. (n-hexane equivalent)	≤2000:1 ppm vol. >2000:10 ppm vol.	<200 ppm vol.: ±10 ppm; ≥200 ppm vol.: ±5% of value	±0.1 ±0.2
O2	Electro chemical sensor	0–22%vol.	0.01% vol.	<2% vol.: ±0.1% vol.; ≥2% vol.: ±5% of value.	±0.2 ±0.3
NO	Electro chemical sensor	0–5000 ppm vol.	1 ppm vol.	<500 ppm vol: ±50 ppm vol ≥500 ppm vol: ±10% of value	±0.2 ±0.9

.

$$∆U=\sqrt{{\left(\frac{\partial U}{{\partial x}_1}∆x_1\right)}^2+{\left(\frac{\partial U}{{\partial x}_2}∆x_2\right)}^2+{\left(\frac{\partial U}{{\partial x}_3}∆x_3\right)}^2+\dots {+\left(\frac{\partial U}{{\partial x}_n}∆x_n\right)}^2}$$

(1)

3. Results and Discussion

3.1. Effect of Control Factors on Performance and Emission Characteristics

BTE is the measure of the conversion of heat energy by an engine from fuel to mechanical power. The variations in BTE with load, biodiesel percentage, and load ethanol percentage are shown in Figure 3a,b. The contour plots in Figure 3a,b express the effect of individual variations in palm biodiesel and ethanol on BTE. It was observed that BTE increases with an increase in load and is found to be highest at 100% load. A BTE maximum of 18.96% was found for 5% biodiesel and 5% ethanol addition at 100% load. It was observed that a minimal substitution of biodiesel and ethanol results in higher BTE. The addition of more ethanol, from 5 to 10%, leads to a decrease in the overall calorific value of the diesel–biodiesel–ethanol blend that retards the combustion performance of the engine. Engine emissions like NO_x, CO, and UHC are shown in Figure 4, Figure 5 and Figure 6. NO_x and UHC emissions were found to have an increasing trend with load because of biodiesel and ethanol in the ternary blend. The rapid rise in temperature generation during combustion at full load is the main reason for high NO_x emission. Similar trends were reported by Sathish et al. [14], who observed 4.4 to 6.3% higher NO_x emissions from different ternary blends compared to baseline diesel. Most of the CO emissions were found in the range from 0.038 to 0.054 vol.%. A high amount of oxygen content in both biodiesel and ethanol accelerates the combustion, which results in lower CO emissions. The more complete combustion can indicate a drop in CO emission due to the ethanol addition. The oxygenated property of ethanol and palm biodiesel enhances the rate of combustion and the blend burns faster, which leads to low CO emissions [27,28].

3.2. Non-Linear Regression Analysis

Due to the different complexities and difficulties of running experiments in a conventional way, the design of experiments (DOE) is one efficient statistical technique to reduce the number of experiments [29]. The developed regression model is a tool for the prediction of engine performance–emission characteristics for the optimization of multivariable problems. The BTE is the main performance index, whereas NO_x, CO, and UHC are the important pollutants that are used for model optimization. In this paper, before performing the regression analysis, a general full factorial design matrix was developed for conducting the experiments. In this design, three factors were selected, namely load, biodiesel, and ethanol percentage at five, four, and two levels each, respectively (in

Table 5. Experimental factors and their levels.

Factors	Symbolic Representation	Levels
Load (%)	A	20	40	60	80	100
Palm Biodiesel (vol.%)	B	5	10	15	20	-
Ethanol (vol.%)	C	5	10	-	-	-

). An ordinary second-order non-linear regression model [30] was developed for the prediction of performance emissions of the diesel engine. The experiments were carried out at different loads (20, 40, 60, 80, and 100%), and varying palm biodiesel (5, 10, 15, and 20% by vol.) and ethanol (5 and 10% by vol.) fractions. The model was analyzed after generating a full factorial design of the experiment among three design variables for 20 different experimental runs. A relationship was developed between the outputs and the input design variables to evaluate statistical terms like F-value, p-value, and R² values [31]. The regression equations for BTE, NO_x, CO, and UHC were calculated by using load (A), palm biodiesel (B), and ethanol (C), as shown below.

BTE = 7.065 + 0.13600 A + 0.0114 B + 0.1142 C − 0.000096 A × A + 0.00152 B ×
B − 0.000253 A × B − 0.001841 A × C − 0.01048 B × C

(2)

CO = 0.0678 − 0.000741 A − 0.00284 B + 0.00115 C + 0.000007 A × A + 0.000170
B × B + 0.000012 A × C − 0.000144 B × C

(3)

NO_x = −344.3 + 18.658 A + 10.89 B + 17.65 C − 0.07905 A × A − 0.093 B × B −
0.0396 A × B + 0.0211 A × C − 1.152 B × C

(4)

UHC = 54.6 + 0.599 A − 1.65 B + 1.68 C − 0.00106 A × A + 0.0245 B × B +
0.00023 A × B + 0.0179 A × C + 0.0636 B × C

(5)

3.2.1. Analysis of Variance (ANOVA)

The main purpose of the analysis of variance (ANOVA) was to observe the significant influence of input variables on the engine output responses since ANOVA reveals the percentage significance of input design factors on a response. ANOVA analyses for the BTE, NO_x, CO, and UHC are shown in

Table 6. Analysis of variance for BTE, NO_x, CO, and UHC.

Source	BTE		NOx		CO		UHC
	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value	F-Value	p-Value
Regression	746.21	0.000	957.45	0.000	4.86	0.001	39.30	0.000
A	229.39	0.000	795.38	0.000	6.57	0.015	5.72	0.023
B	0.06	0.814	9.42	0.004	3.36	0.077	1.50	0.229
C	4.90	0.034	21.57	0.000	0.48	0.494	1.36	0.252
A × A	2.60	0.117	325.54	0.000	12.24	0.001	0.41	0.527
B × B	0.91	0.348	0.63	0.434	11.00	0.002	0.30	0.585
A × B	1.01	0.322	4.55	0.041	0.00	0.951	0.00	0.974
A × C	10.70	0.003	0.26	0.614	0.48	0.495	1.30	0.263
B × C	13.55	0.001	30.14	0.000	2.47	0.126	0.64	0.429
R-sq.	99.48		99.49		55.66		91.02
R-sq. (adj.)	99.35		99.23		44.21		88.71

to observe the influence of linear, square, and interaction terms of the input factors in the model [32]. To meet the 95% level of significance, p-values less than 0.05 for the load were found for all targets [33]. P-values less than 0.05 in the interaction terms of different targets were found for different interactions of factors. A significant effect of biodiesel was observed at a 95% confidence level for both its linear and interaction effect in the NO_x emission. However, no such significant effect of ethanol on NO_x, CO, and UHC emissions was observed. Except for CO emission, an R² value of more than 90% was observed for BTE, NO_x, and UHC, resulting in a high accuracy of the model with the experimental values.

3.2.2. Response Optimization

The optimization of different engine operating parameters using a regression model is shown in Figure 7. For the test conditions, a maximum value of BTE and minimum values of NO_x, CO, and UHC emissions were set as the optimum model target. In Figure 7, the optimization of BTE-NO_x-CO-UHC is shown where D and d signify composite and individual desirability of the response, respectively. Composite and individual desirability (d) evaluate the optimization of a set of responses and a single response, respectively. Desirability ranges between 0 and 1, and the higher value represents a favorable result overall. Desirability analysis was performed on the response values. With the condition that the larger the value, the better the desirability function, a desirability value of 0.6053 was obtained as the optimal condition for BTE, NO_x, CO, and UHC. A similar kind of desirability analysis has been performed by Awad et al. [34], who obtained a 0.7 desirability value, which is very similar to the result obtained in this study. Optimization using a regression model reveals 43.43% engine load, 11.06% palm biodiesel, and 5% ethanol as the optimal input variables, which can optimize BTE, NO_x, CO, and UHC emissions at 12.57%, 436.21 ppm, 0.037 vol.%, and 79.24 ppm, respectively. The optimization of the BTE-NO_x-CO-UHC parameters describes the contributions of palm biodiesel and ethanol that were effectively optimized by the regression model for the performance–emissions synergy. Corresponding to the optimized engine load, palm biodiesel, and ethanol percentage from the regression analysis, the experimental results at 40% engine load, 10% palm biodiesel, and 5% ethanol share were compared for validation. The detailed comparison study is shown in

Table 7. Comparison of performance–emission parameters between experimental and optimized input variables.

Engine Output Parameters	Experimental (Input Variables: 40% Engine Load, 10% Palm Biodiesel, 5% Ethanol)	Optimized (Input Variables: 43% Engine Load, 11% Palm Biodiesel, 5% Ethanol)
BTE (%)	12.52	12.57
NOx (ppm)	401	436.2
UHC (ppm)	87	79.24
CO (vol.%)	0.05	0.037

. From Table 7, it is clear that the optimized performance–emission parameter values are almost similar to the experimental values. Hence, the regression model can be used as an optimization tool to predict and optimize the engine output variables within the range of the tested targets, which can reduce experimental runs by saving time and money.

4. Conclusions

This paper discusses the effect of biodiesel and ethanol at different proportions in ternary blends at different engine loads. Using a non-linear regression model, a four-stroke diesel engine’s prediction of optimum performance and emission characteristics was investigated. A significant effect of various input factors was found for the optimization of engine parameters. Optimum performance–emission parameters were observed as 12.57%, 436.2 ppm, 0.03 vol.%, and 79.2 ppm for BTE, NO_x, CO, and UHC, respectively, at optimum input parameters of 43.43% engine load, 11.1% palm biodiesel, and 5% ethanol share. Based on the aforementioned results, BTE-NO_x-CO-UHC optimization reveals an active contribution of palm biodiesel and ethanol that can be used for diesel engine combustion. Hence, this paper contributes to accurate decision-making by optimizing engine performance and emission parameters using a statistical regression model. This approach provides practical ideas to the decision-maker for the development of IC engine research. Further, for future work, other operating parameters like varying the compression ratio (CR), main injection pressure, timing, and duration of main injection could be taken into consideration for obtaining an optimized share of palm biodiesel and ethanol fractions in ternary blend operation.