Optimal Performance and Emission Analysis of Diesel Engine Fuelled with Palm Oil Methyl Ester with an Artificial Neural Network

Abstract

Biofuels or biodiesels are fuels that are in biodegradable and non-toxic. They are manufactured from waste cooking oils, vegetable oils and animal fats or tall oil (a by-product of the pulp and paper industry). These oils undergo a procedure called transesterification whereby they are subjected to a reaction with an alcohol usually ethanol or methanol by means of a catalyst such as sodium hydroxide. This results formation of ethyl or methyl ester called biodiesel and a by-product called glycerin. Pure biodiesel fuel is considerably not as much of flammable than petroleum diesel which burns at 50 degrees Celsius. Biodiesels are frequently used in blend with petroleum diesel and are named as biodiesel blends. These blends will contain a flashpoint and a gel point wherever between the two pure fuels depending on the mixture. Artificial neural networks (ANNs) are recently developed techniques which are in variably used in obtaining exact correlations which involves non-linear data. An ANN can be considered to be consisting of interconnected group of relatively simple processing elements or nodes, called neurons, where the global performance is resolute by the relations between the processing nodes and the network parameters. Neural networks, when trained properly are good at providing very fast, extremely close approximations of the correct output for nonlinear data. This study deals with artificial neural network modeling a diesel engine using palm oil methyl ester to predict brake power, brake thermal efficiency, specific fuel consumption and exhaust emission of engine. This property of biodiesel produced from palm oil was measured and the experimental results reveal that blends of palm oil with diesel fuel provide improved engine performance and improved emission characteristics. Using some of the experimental data for training, an ANN model program for the engine was developed. Different activation functions and several rules were used to assess the percentage error between the desired and predicted values. It was observed that the ANN model can predict the engine performance and exhaust emission quite well.