SI Engine Simulation Using Residual Gas and Neural Network Modeling to Virtually Estimate the Fuel Composition

Research in electronic controlled internal combustion engines mainly focuses on improving performance and lowering the emissions. Combustion performance depends on the geometry of cylinders and on the design of all mechanical parts, which are based on the laboratory experimental research. Due to the limitations of the materials used in the engine and the continuous high operating temperature, engines function in either spark ignition or charge ignition processes. Recent research on computer controlled engines uses sensors and electronic actuators which allows switching the engine operational mode between spark ignition and charge ignition. Thus, this makes possible to mix intake fuel compositions in order to give more choices to consumers. This study employs a neural network which is capable of estimating fuel composition using the parameters of residual gas. The simulation is based on a thermodynamic engine model implemented in Matlab Simulink. The main advantages are the capabilities of the model to 1) calculate the gas exchange as a function of time in transient mode, and 2) to generate data for the design control algorithms without the need of the engine bed test environment to test various fuel compositions.