Optimizing diesel engine heterogeneous combustion performance and NOx emissions: A next energy perspective with AI

Abstract

This study investigates experimental and artificial intelligence-based predictions of heterogeneous combustion performance in a diesel engine fueled with neat biodiesel. The combustion aspects, including cylinder pressures, heat energy developed and released, mass burnt fractions (MBF), mean gas temperatures (MGT), and the influence of combustion temperatures on NOx formation, are examined experimentally. The combustion results are trained in a feed-forward artificial neural network (ANN) algorithm for the predictions, and an error histogram with 20 bins helps identify the accuracy of the trained model. The prediction results of combustion parameters are recorded quite accurately for most instances, as the errors are centered around 0. The overall accuracy of the trained model is achieved with a high correlation coefficient (R = 0.99) and a low mean square error (MSE). In addition, the influence of combustion temperature on NO_x emissions is highlighted, and a correlation is developed with errors of 2.22% and 1.96% at 75% and 100% loads, respectively. Finally, biodiesel exhibits controlled diffusion combustion, achieving more sustained combustion, with 6.19% and 6.18% lower NO_x formation compared to diesel fuel at 75% and 100% loads.