Artificial Neural Network-Based Prediction of Performances-Exhaust Emissions of Diesohol Piloted Dual Fuel Diesel Engine Under Varying Compressed Natural Gas Flowrates

The present study surveys the effects on performance and emission parameters of a partially modified single cylinder direct injection (DI) diesel engine fueled with diesohol blends under varying compressed natural gas (CNG) flowrates in dual fuel mode. Based on experimental data, an artificial intelligence (AI) specialized artificial neural network (ANN) model have been developed for predicting the output parameters, viz. brake thermal efficiency (B_th), brake-specific energy consumption (BSEC) along with emission characteristics such as oxides of nitrogen (NO_x), unburned hydrocarbon (UBHC), carbon dioxide (CO₂), and carbon monoxide (CO) emissions. Engine load, Ethanol share, and CNG strategies have been used as input parameters for the model. Among the tested models, the Levenberg–Marquardt feed-forward back propagation with three input neurons or nodes, two hidden layers with ten neurons in each layer and six output neurons, and tansig-purelin activation function have been found to the optimal model topology for the diesohol–CNG platforms. The statistical results acquired from the optimal network topology such as correlation coefficient (0.992–0.999), mean square error (MSE) (0.0001–0.0009), and mean absolute percentage error (MAPE) (0.09–2.41%) along with Nash–Sutcliffe coefficient of efficiency (NSE), Kling–Gupta efficiency (KGE), mean square relative error, and model uncertainty established itself as a real-time robust type machine learning tool under diesohol–CNG paradigms. The study also incorporated a special type of measure, namely Pearson's Chi-square test or goodness of fit, which brings up the model validation to a higher level.