Appraisal of artificial neural networks to the emission analysis and prediction of CO2, soot, and NOx of n-heptane fueled engine

The present investigation was carried out to assess the potential of applying Artificial Neural Network (ANN) technique to the prediction of n-heptane fueled Direct Injection (DI) Diesel engine emissions of CO2, soot, and NOx. A code was developed to simulate the combustion process using computational fluid dynamics (CFD) approach employing n-heptane fuel under the effect of crank angle, temperature, pressure, liquid mass evaporated, equivalence ratio, and O2 concentration at two engine speeds of 2000 and 3000 rpm. In the next step, a supervised ANN model coupled with CFD approach was trained. Therefore, a feed-forward with back-propagation (BP) learning algorithm was applied and the network training approach of Levenberg-Marquardt was evaluated within varying number of neurons. While a decreasing trend was observed regarding O2 concentration, equivalence ratio, liquid mass evaporated, and temperature were increased by the increment of the crank angle (CA). Additionally, the exhaust emissions of CO2, soot, and NOx were increased by the increment of CA, and the engine speed. Finally, it was discovered that the lowest (Mean square error) MSE value of 0.0001086 is yielded at 18 neurons in the hidden layer. The remarkable R2 amounts of 0.9976, 0.9995, and 0.9951 were obtained for CO2, soot and NOx emissions, respectively.