Parametric Optimization of Biodiesel Fuelled Engine Noise using the Taguchi Method

Biodiesel is a renewable, biodegradable, and efficient fuel that can be blended with petro-diesel in any proportion. The noise in the engine resulting from the combustion has a direct effect on the engine’s performance. Many studies have examined the engines’ vibration and noise when using diesel and biodiesel blends. This study examines the optimization of diesel blends, load, and compression ratio in the aspect of reducing noise on a Kirloskar single-cylinder diesel engine. Noise was measured at the engine and its exhaust on a computerized setup and for different loads. The experimental results showed that a blend with 15% biodiesel, at 7kg load, and 18 compression ratio produced the lowest noise. Moreover, the Taguchi method was utilized, and experimental results were validated by an ANN. Keywords-transesterification; biodiesel; noise; optimization

INTRODUCTION Any alternative to diesel fuel should be replicable, economical, and technically acceptable [1]. Biodiesel is produced by the transesterification of renewable vegetable oils and animal fats with the use of alcohol. Biodiesel is highly degradable and nontoxic. Meanwhile, it has low emissions of carbon monoxide, particulate matter, and unburned hydrocarbons. Due to these properties, biodiesel has attracted wide attention as a replacement to diesel fuel [1,2]. Biodiesel can be used without modifications in conventional compression ignition engines. Noise and vibrations are major issues of diesel engines [3,4]. Engine body vibrations and noise are rich in information about the engine's operating parameters and physical condition [4,5]. Excess noise and vibrations wear out components such as bearings, grouting, and couplings, increasing maintenance cost due to more component failures and unplanned operations. Due to noise and vibrations' importance, there is a need to study the effect of biodiesel and its blends on engine's life and performance [6][7][8]. Noise level depends on the load and the blending ratio of biodiesel [5]. As a result, it is necessary to extend an engine's life by using optimal blends, after analyzing their impact in noise [5][6][7][8][9].
II. EXPERIMENTAL PROCESS A Kirloskar TV1 VCR single cylinder, four stroke, constant speed, water-cooled diesel engine, having 3.5HP at 1500rpm, was used on a computerized test bed equipped with measuring instruments such as thermocouples, dynamometer, tachometer and flow meters. The engine's specifications are shown in Table I. Noise levels were measured by a noise meter for four different fuel blends on variable load conditions and compression ratios as per the Taguchi array. The study focused on the input parameters of biodiesel blends for examining the diesel engine's operating conditions. Noise was measured at the engine and its exhaust. A noise meter was placed at 0.5m distance from the engine for measuring its noise, and another was placed outside the room near the exhaust pipe end to measure the noise at the exhaust [3][4][5]. The noise meter and its specifications are shown in Figure 1 and Table II respectively. Four fuel types were tested, namely: B0 consisting of 100% diesel, B15 consisting of 15% biodiesel and 85% diesel, B20 consisting of 20% biodiesel and 80% diesel, and B25 consisting of 25% biodiesel and 75% diesel [2,3,9]. Biodiesel blend, load on the engine, and compression ratio were the parameters whose effects on the engine's noise were studied. The parameters' levels are listed in Table III. A. Noise Analysis The orthogonal array of the input parameters indicates the number of combinations for the experiments. This selection of orthogonal array is based on three parameters and four levels for each parameter [2,5]. The array was obtained by Minitab using the following operating parameters:

C. Validation of Experimental Results by Artificial Neural Network (ANN)
The results of noise at the engine and the exhaust were validated by an ANN. An ANN script, shown in Table VII, was used for obtaining the output from the input parameters.

A. Noise at the Engine
The experimental results for noise at the engine, the values calculated by the ANN, and the error between them are shown in Table VIII and a comparative graph of these values is shown in Figure 2. Apparently, there is a small difference, less than 1.2%, between the experimental and the ANN calculated values. The regression plot obtained by the Taguchi model for the experimental results was compared with the ANN regression plot. The regressions' R-square value was around 80%. The straight line in these plots shows that the data fit a normal probability distribution. There are very low residual values, as

B. Noise at the Exhaust
The experimental results of noise at the exhaust, the values calculated by the ANN, and the error between them are given in Table IX. Moreover, a comparative graph of these values is shown in Figure 4. Apparently, there is a tiny difference between experimental and ANN results, less than 0.3%, for noise at the exhaust.

IV. CONCLUSION
This study examined the optimization of noise reduction at the engine and its exhaust with biodiesel blend, load, and compression ratio of the engine as input parameters. Analysis was carried out utilizing the Taguchi method, and optimization of the input parameters was performed by using SNR [10,11]. The experimental results obtained by Minitab were validated by an ANN. The main conclusions of this study are: • Optimal input parameters were: a blend with 15% biodiesel, applied load of 7kg, and compression ratio 18, resulting to 95dB noise at the engine and 110.45dB at its exhaust.
• R-square values obtained by regression analysis were around 80% and more, indicating that the obtained model fits to the actual data.
• There are small to tiny differences between the experimental and the ANN's noise values.
• All regression residuals of both Minitab and ANN were very low and almost along the line in both methods. The similarities in both plots validated the results.