Modeling the impact of in-cylinder combustion parameters of di engines on soot and NOx emissions at rated EGR levels using ANN approach

This study examines the effect of in-cylinder combustion parameters on soot and NOx emissions at rated EGR levels by using the data obtained from the CFD implemented code. The obtained data were subsequently used to construct an artificial neural network (ANN) model to predict the soot and NOx productions. To this aim, at three different engine speeds of 2000, 3000 and 4000 rpm, heat release rate, equivalence ratio, turbulence kinetic energy and temperature varied to obtain the relevant soot and NOx data at three EGR levels of 0.2, 0.3 and 0.4. It was discovered that wherein the application of higher EGR rates reduced the NOx as a result of mixture dilution, equivalence ratio increment makes soot production to be increased as well as NOx emission. It was also found that the application of higher EGR from 20% to 40% decreased soot mass fraction in the combustion chamber. Increment of EGR reduced the emissions where the equivalence ratio had contradictory effect on the produced emissions. Various ANN topological configurations and training algorithms were incorporated to yield the optimal solution to the modeling problem applying statistical criteria. Among the four adopted training algorithms of trainlm, trainscg, trainrp, and traingdx, the training function of Levenberg–Marquardt (trainlm) with topological structure of 5-19-17-2 denoted MSE equal to 0.0004627.