Adaptive neuro-fuzzy method to estimate virtual SI engine fuel composition using residual gas parameters

This paper proposes a neuro-fuzzy model to estimate an engine fuel composition from the residual gas information. While fuels available from fuel producers are categorized by names, i.e. gasoline or diesel, the exact chemical composition in terms of number of hydrocarbon atoms remains unknown to the user. Meanwhile, the engine combustion performance depends on the fuel composition. This study researches advanced engine control methodologies by identifying fuel composition from the exhaust gases to improve the engine performance while reducing the emissions. Fuel composition can be estimated by the inverse of engine combustion process model; In order to achieve this, the development of the suitable artificial intelligence system identification was required. Here, a computer based engine model which uses Adaptive Neuro-Fuzzy Interface System (ANFIS) to identify the fuel composition is used. The residual composition contains the levels of Carbon Dioxide (CO₂) Oxygen (O) Carbon Monoxide (CO) and Nitric Oxide (NO). The model is developed to estimate the Hydrocarbon level of the original fuel. Results show that ANFIS control can reasonably distinguish the mixtures of three given fuel compositions, namely, pure Isooctane, Isooctane-Methanol and Isooctane-Ethanol.