Supervised Machine Learning Dual-Fuel Ignition with a Glass Box - High Dimensional Model Representation (HDMR) and a Black Box - Convolutional Neural Network (CNN)

To meet stringent pollutant regulations, many IC engines have been developed; one of them is a dual-fuel engine that is a compression ignition engine and uses gaseous fuel, e.g., methane, as the main fuel and diesel/reactive fuel as pilot fuel. Ignition delay time (IDT) is one of the most important parameters to characterize and model the dual-fuel combustion process. Accurate calculation of the IDTs over a wide range of fuel blends, pressures and strain rates is a complicated procedure. In this work, we use two supervised machine learning methods, a glass box – high dimensional model representation (HDMR) and a black box - Convolutional Neural Network (CNN), to seek models for accurate and efficient prediction of the IDT. HDMR, also known as the generalized functional ANOVA expansion, provides useful insight into the model output dependence on important input variables, which helps to study low-temperature chemistry effects on the IDT prediction. To generate datasets, computational parametric studies are conducted in a transient counterflow mixing layer with detailed chemistry and transport. Different sample datasets (referred to training data) are used to train the HDMR. The remaining datasets (referred to testing data) are used to evaluate the IDT model performance. Detailed comparisons and analysis will be shown.