The comprehension of combustion mechanisms enables supervision of reaction rates. By adjusting factors such as heat transfer rates, combustion duration, self-ignition propensity, ignition delay and laminar flame speeds, it is possible to minimize emissions and enhance fuel conversion efficiency in internal combustion engines (ICE). The present study aims to develop and explore a methodology employing an Artificial Neural Network that uses Mass Burned Fraction data as a function of crankshaft angular position to determine combustion kinetics in ICE. The Artificial Neural Network was programmed in this work as a home-made code and produced accurate results. The kinetic triplet consisting of Activation Energy (Ea), Frequency Factor (A) and Reaction Model throughout the combustion process was determined to explore the combustion characteristics of different gasoline formulations and ICE operation conditions. The experimental data were obtained in a Single Cylinder Research Engine (SCRE) operating with gasoline formulations commercialized in Brazil. The methodology determines the kinetics of combustion along the process and recovers the values of Ea and A without resorting to mechanisms that describe each reaction individually, describing, instead, the global contribution of physical models. Because the kinetic models activate the neurons in the hidden layer, they accurately reproduce the experimental Mass Burned Fraction data and bring physical information to the network about the combustion process. The kinetic study showed that the samples with higher values of Ea also had higher ignition delay. The rate constant was also related to the consumption and combustion efficiency during the combustion process, i.e., the fuel with a higher rate constant presents greater combustion efficiency and smaller consumption.
Read full abstract