Wear element analysis using neural networks of a DI diesel engine using biodiesel with exhaust gas recirculation

Abstract

Wear is an important characteristic because of its great value in connection with the engine parts. The main focus of this work is to analyze the effect of neural network models for predicting engine performance such as wear of the DI diesel engine using B20 blend of Methyl Ester of Mahua (MEOM) and diesel. Experimental results revealed that 20% biodiesel blend is the optimum blend in terms of performance, emission and combustion characteristics. For B20 blend, it was also found experimentally that 15% hot EGR and 20% cold EGR were the optimum EGR ratios. Under the optimum EGR ratios identified, a series of experimental work was done with B20 blend and diesel at various loads to obtain the concentration of wear metals from the lubricating oil. Experimentally, it was found that wear metals were found to be lower for B20 biodiesel compared to diesel. Artificial neural networks (ANN) have become the premier candidate as the modeling tool. Using the experimental data, ANN models based on probabilistic neural networks (PNN) and radial basis function neural networks (RBFN) for predicting the engine wear were developed. The results show that ANN is sufficient enough in predicting the engine wear in terms of mean square error (MSE) and regression coefficient (R). Also among the ANN models tested, RBFN performs significantly better than PNN.