Investigating the effect of combustion properties on the accumulated heat release of DI engines at rated EGR levels using the ANN approach

• An accumulated heat release analysis for DI engines using computational fluid dynamics (CFD) was covered.
• Artificial neural network (ANN) prognostication of accumulated heat release was carried out.
• A 5-17-1 topology with Levenberg–Marquardt training algorithm (trainlm) was selected as the supervised ANN.
• Mean squared error (MSE) of 0.1057 and R2 of 0.99 were obtained for the outperforming model, respectively.

This study is dedicated to explore the effect of in-cylinder combustion parameters on the accumulated heat release at rated EGR levels using the CFD implemented code data which were coupled to the artificial neural network (ANN) approach to construct a model to predict the accumulated heat release of DI diesel engines. To this end, at two different engine speeds of 3000 and 4000 rpm, crank angle, equivalence ratio, turbulence kinetic energy and temperature varied to obtain the corresponding accumulated heat release data at three EGR levels of 0.2, 0.3 and 0.4. It was discovered that that application of higher EGR is conducive to temperature reduction while it leads to the decreased equivalence ratios. It was also concluded that the accumulated heat increases with equivalence ratio and temperature but decreases with increment of Exhaust Gas Recirculation (EGR) levels. Numerous ANN modeling implementations were carried out using different training algorithms of trainlm, trainscg, traingdx, and trainrp at diversified number of neurons in the single hidden layer. At 17 neuron numbers in the hidden layer, the trainlm method denoted MSE equal to 0.1057 which was the best performance among the various implemented models. The coefficient of determination (R2) values equal to 0.99 and 0.99 were obtained for training and testing phases. The obtained results confirm the promising ability of ANN for the prognostication of accumulated heat release of DI engines.