Development of Method to Measure the State of Fuel Supply During Cold Start and Application of the Data to Modeling of Emission Using Machine Learning

Abstract

To investigate the behavior of injected fuel during cold start of gasoline engines, a non-combustive method using exhaust gas without combustion was constructed. The THC concentration of exhaust gas is relevant to the quantity of fuel introduced into the cylinder at the current and next cycle. The sensitivities of the concentration to the amount of injected fuel, fuel property and intake port temperature were proved by the experimental data. The proved data were used to model THC behavior applying machine learning. The model was demonstrated high accuracy and applied to derive the injection strategy to achieve low THC emission.