A comparative analysis of two neural network predictions for performance and emissions in a biodiesel fuelled diesel Engine

In this research, back-propagation (BP) and generalized regression (GR) GR neural networks are developed for predicting the performance and emissions of direct injection diesel engine fuelled with the mixtures of diesel and castor oil fuels. The neural network models for the engine were trained by using some of the experimental data. Experimental test are carried out on a semi-heavy duty Motorsazan MT4.244 direct injection diesel engine fuelled with blends of diesel fuel with 0%, 5%,10%,15%,20%, 30% of Castor oil%(by volume) at various speeds and loads. Then, the performance of these neural networks predictions are compared by comparing predictions with the experimental results which were not used in the training process. The comparison of the predicted values shows that the computational accuracy of both GR and BP neural networks are appropriate, however the GR presents slightly better performance with very faster training compared with the BP. therefore, it can be concluded that GR can be used to predict performance and emissions with high accuracy and faster training.