Optimization of biodiesel production, engine exhaust emissions, and vibration diagnosis using a combined approach of definitive screening design (DSD) and artificial neural network (ANN)

Abstract

In this study, definitive screening design (DSD) optimization and artificial neural network (ANN) modelling techniques are applied for the production of palm oil biodiesel (POBD). These techniques are implemented to examine the vital contributing factors in achieving maximum POBD yield. For this purpose, seventeen experiments are conducted randomly by varying the four contributing factors. The results of DSD optimization reveal that a biodiesel yield of 96.06% is achieved. Also, the experimental results are trained in ANN for predicting the biodiesel yield. The results proved that the prediction capability of ANN is superior, with a high correlation coefficient (R²) and low mean square error (MSE). Furthermore, the obtained POBD is characterized by significant fuel properties and fatty acid compositions and observed within the standards (ASTM-D675). Finally, the neat POBD is examined for exhaust emissions and engine cylinder vibration analysis. The emissions results confirm a significant drop in NO_x (32.46%), HC (40.57%), CO (44.44%), and exhaust smoke (39.65%) compared to diesel fuel at 100% load. Likewise, the engine cylinder vibration measured on top of the cylinder head reveals a low spectral density with low amplitude vibrations observed for POBD at measured loads.

Keywords: Biofuel production; Correlation coefficient; Engine performance; Engine vibration; Feed forward back propagation algorithm; Machine learning; Palm oil; Research operation.