Predictive modeling and optimization of noise emissions in a palm oil methyl ester-fueled diesel engine using response surface methodology and artificial neural network integrated with genetic algorithm

Abstract

This research examines the predictive performance of two modeling techniques—Response Surface Methodology (RSM) and an Artificial Neural Network enhanced by a Genetic Algorithm (ANN-GA)—in relation to noise emission levels from a single-cylinder diesel engine running on palm oil methyl ester (POME). By employing different engine speeds and loads within the low to high range, noise emissions were recorded from multiple engine components to evaluate the performance of each predictive model. The outcomes of the experiments were contrasted with the forecasts produced by the RSM and ANN-GA models, emphasizing the goal of reducing error percentages. The analysis indicates that the ANN-GA model consistently yields predictions that align more closely with the experimental noise values compared to the RSM model. The average error for the ANN-GA model was 1.03%, significantly less than the 1.82% average error found with the RSM model. This demonstrates a significant enhancement in predictive accuracy using ANN-GA, highlighting its potential as a dependable tool for forecasting noise emissions in biodiesel-powered engines. Specific components, such as the radiator, crankshaft, and crankcase, exhibited minimal prediction errors under ANN-GA, suggesting that this model is particularly adept at capturing the complex noise emission patterns associated with POME-fueled engines. In summary, the results illustrate that the ANN-GA model outperforms the RSM model in predicting noise emissions under the conditions tested, providing a more accurate and effective method. These findings endorse the feasibility of applying ANN-GA in scenarios where precise noise prediction is crucial, particularly in relation to alternative fuels like POME.