Abstract:
The endeavour of this study is to analyze the emission and performance characteristics of a single cylinder; four stroke diesel engine using a Polanga biodiesel blends as fuel. The assortment of Polanga biodiesel is based on a widespread review of literature which indicated that this is relatively unexplored fuel for a diesel engine. The experimental investigation is followed by a computational study consisting of modelling using artificial neural network and optimization for forecasting the best possible emission and performance parameters constituents of the diesel engine. The experimental investigation is conducted on a Kirloskar make single cylinder, four stroke diesel engine using Polanga biodiesel blends with diesel as a test fuel. The exhaust emissions and performance characteristics are assessed by operating the diesel engine at altered predetermined fuel injection timings of 15, 19, 23, 27 and 31obTDC and changed fuel injection pressures of 160 bar, 180, 200, 220 and 240 bar at varying load from zero to full load in steps of 20 percent increments. The exhaust emission constituents measured are unburnt hydrocarbons (UHC), carbon monoxide (CO), carbon dioxide (CO2), smoke opacity and oxides of nitrogen (NOx). The performance parameters are brake specific fuel consumption (BSFC), brake thermal efficiency (BTE) and exhaust gas temperature (EGT). The experimental study accomplished provides a very large number of results. But, it is impossible to select the input parameters such as injection timing, injection pressure and blend for obtaining best exhaust emissions and performance and from the diesel engine. Hence, an appropriate computational study is required to be carried out to meet the goal of finding the best grouping of the input parameters under different specific priorities. ANN is used to obtain the output parameters using different input parameters. ANN modelling is a very complex technique which is capable of modelling different functions and processes. In this study, ANN modelling is developed using the neural network feature of MATLAB (R2010a).
