Optimization and ANN Modelling for Performance and Emissions Prediction of a Biodiesel-Diesel Blend Engine with EGR using Response Surface Methodology

Abstract

This study evaluates the effects of Jatropha and Karanja biodiesels and their blends on a diesel engine's fuel economy, performance, and exhaust emissions. Response Surface Methodology was applied to model engine responses using four key input parameters: blend percentage, load, injection timing, and Exhaust Gas Recirculation. RSM projected multiple Pareto-optimal solutions through multi-objective optimization and contour plots. The study investigated important engine responses like Brake Thermal Efficiency, Brake Specific Fuel Consumption, HC emissions, smoke, NOx, and EGR. RSM models for Jatropha biodiesel and blends displayed high R² values, ranging from 0.93 to 0.99. Similarly, Karanja biodiesel and blends exhibited R² values ranging from 0.94 to 0.98. The results indicate that all tested fuels provided accurate approximations for the engine responses. Furthermore, an Artificial Neural Network model was developed to predict input parameters based on desired performance and emission constraints. The ANN approach proved effective in predicting engine responses based on operating conditions, injection system parameters, and exhaust gas recirculation.