Estimation of the COVIMEP Variation in a HCCI Engine

Abstract

In this study, variation of the COVIMEP was tried to be predicted by using the artificial neural network method for 4-stroke, 4-cylinder, direct injection and supercharged HCCI engine experimental data obtained by using n-heptane fuel at 60 degrees C intake air temperature, 1000 rpm engine speed at different inlet air intake pressure. Intake air inlet pressure and lambda were used as input data in artificial neural network model. The COVIMEP value was used as the target. Three layers and five neurons were used to construct the network using the Levenberg-Marquardt algorithm. Correlation between targets and outputs for teaching, accuracy and testing were obtained as 0.97989, 0.9504 and 0.91644, respectively. Total correlation factor was found as 0.96983. As a result of the study, it was seen that the stored data and the estimated COVIMEP data were compatible.