AI-Based ANN Modeling of Performance–Emission Profiles of CRDI Engine under Diesel-Karanja Strategies

Abstract

The current investigation highlights the impact of Diesel–biodiesel blends on performance and exhaust emission profiles of a single-cylinder, common rail direct injection (CRDI) engine. Experiments were performed at constant engine speed (1500 rpm) and three engine loads (50, 75 and 100%) under high fuel injection pressure (900 bar) with volume proportions (10, 20 and 30%) of Karanja with Diesel. Utilizing CRDI engine experimental data, an artificial intelligence (AI)-affiliated artificial neural network (ANN) model has been created with the intention of forecasting brake thermal efficiency, oxides of nitrogen, unburned hydrocarbon and carbon monoxide emissions. From various tested ANN models, one hidden layer with three neurons along with logsig transfer function has been noticed to be optimum network for Diesel-Karanja paradigms under high fuel injection pressure. While developing the optimum model, standard Levenberg–Marquardt training algorithm has been employed. The optimum ANN model is capable to estimate the CRDI engine performance–emission profiles with an overall correlation coefficient value of 0.99742, wherein 0.99783, 0.99951 and 0.99969 for training, validation and testing datasets, respectively. Results made clear that the formulated AI-based ANN model is viable for predicting the existing CRDI engine performance and emission profiles of Diesel-Karanja blends operating under high fuel injection pressure.