Combustion Chemistry of Chemical Kinetics and Machine Learning

Understanding low-temperature combustion mechanisms of hydrocarbons is aided by isomer-resolved experiments and chemical kinetics modeling of complex species. Alkyl-substituted cyclic ethers are intermediates formed during low-temperature oxidation and are derived from unimolecular reactions of hydroperoxyalkyl radicals (Q̇OOH). To understand the combustion of these cyclic ethers and differences in stereochemistry, comparison of chemical kinetics modeling with species profiles produced from the competing network reactions are required. For this present work, N-propyloxirane and cis-/trans-2-ethyl-3-methyloxirane oxidation experiments were conducted in a jet-stirred reactor at 1.1 atm, from 600 – 950 K, at 2 s residence time and stoichiometric conditions using an initial concentration of 2500 ppm. Isomer-resolved species profiles were quantified using tandem vacuum-ultraviolet absorption spectroscopy and electron-impact mass spectrometry. Species such as ketones, aldehydes, cyclic ethers, and alkenes were observed from each set of oxidation experiments. To complement the measurements, a mechanism was produced for both n-propyloxirane and 2-ethyl-3-methyloxirane using Reaction Mechanism Generator (RMG) which yielded 560 species and 1849 reactions and 852 species and 3753 reactions, respectively. Additionally, machine learning models were developed as a tool for predicting contributions of the carbonyl group and consisted of various pre-processing steps including Principal Component Analysis, energy range prediction importance, Synthetic Minority Oversampling Technique, and Tomek Links.