An Assessment of Six-Cylinder Diesel Engine Performance and Vibration Features by Diesohol Fuel Employing Neural Networks
Abstract
This study's main goal was to thoroughly assess the performance and vibration of a six-cylinder engine utilizing diesel-bioethanol fuel mixtures. These blends incorporated bioethanol in ratios below 15% and were created by combining anhydrous ethanol (C2H5OH) with standard diesel fuel in varying proportions. The study employed seven distinct blends to assess engine performance and vibration levels thoroughly. Multi-layered perception (MLP) neural networks were applied namely a feedforward back-propagation neural system. The chosen training algorithm was the Levenberg-Marquardt, while activation functions employed were logsig, tansig, and purelin transferring functions. The study findings revealed that the optimal neural network model consisted of two hidden layers comprising 15 neurons. The recommended transfer functions for the first and second hidden layers were logsig and logsig, respectively. Overall, this study demonstrated the neural system model's remarkable efficacy in accurately predicting the performance and vibration levels of engines operating on blends of diesel and bioethanol fuels, commonly known as diesohol fuel blends.
