Experimental measurement and application of artificial intelligence (AI) for modeling of threshold sooting index (TSI)

Soot formed during combustion of fossil fuels is a serious health and environmental hazard. The propensity of a fuel to form soot is dependent on its chemical composition and the conditions under which combustion takes place. Theoretical models for soot formation have largely been unsuccessful to predict soot due to the coupling of complex chemical kinetics and transport flow. Therefore, empirical correlations like the threshold-sooting index (TSI) have been used to measure the sooting propensity of hydrocarbon fuels and mixtures. The aim of this project is to experimentally measure the TSI of various fuel surrogates (gasoline, diesel and jet fuels) using a non-premixed diffusion flame in a smoke point lamp as specified by ASTM D 1322 standard. The experimental data measured will help understand the relationship between the fuels functional groups and TSI. This data along with other experimental data from the literature will be used to develop a machine learning based artificial intelligence (Artificial Neural Network) model to predict the TSI of fuel mixtures and practical fuels. The chemical composition of the fuels will be represented in terms of the fuels constituent functional groups. Transportation fuels like gasoline, diesel, jet fuel etc. are made up of several hundred to thousands of individual molecules, which are in turn made up of a small number of molecular fragments or functional groups that determine the physical, thermo-chemical and combustion properties of the fuel. These functional groups along with molecular weight and a new parameter called as branching index will be used as input features to develop the TSI machine learning models.