A neural network approach on forecasting spark duration effect on in-cylinder performance of a large bore compression ignition engine fueled with propane direct injection

In the current study, we examined the impact of spark duration strategy on a large bore compression ignition engine fueled with propane direct injection. An artificial neural network also was used to forecast engine in-cylinder performance characteristics. A rapid compression and expansion machine (RCEM) with a spark plug was tested with a high-pressure direct injection propane of 200 bar. While the timing of the injection was set to 20 °CA bTDC, the spark duration can range from 0.7 to 5.0 milliseconds. Crank angle degree, pressure, ignition coil number and spark duration were used as input parameters in the ANN model to predict in-cylinder performance, while engine performance parameters such as heat release rate (HRR), turbulent kinetic energy (TKE), tumble ratio, indicated power, and combustion efficiency (ηc) were used as output parameters. The ANN model was created using the neural network toolbox and standard backpropagation with the Levenberg-Marquardt training algorithm was used with the learning rate and training epochs of the ANN model set to 0.001 and 1000, respectively. The accuracy of the model was validated by comparing the predicted datasets with the experimental data. The five projected parameters of heat release rate (HRR), turbulent kinetic energy (TKE), tumble ratio, indicated power, and combustion efficiency (ηc) showed R2 values of 0.9833, 0.9860, 0.9728, 0.9807, 0.9052, and 0.9999, respectively, and MSE values of 0.1419, 0.0023, 0.6428, 0.0106, 0.0050, and 0.0134. The R2 of the validation dataset was nearly 0.98, which is close to that of the training dataset. The coefficients of determination (R2) were greater than 0.9 in the projected results, and the MSE was reasonably low, indicating that a predictive model based on ANN model could predict in-cylinder performance of a large bore compression ignition engine.