Machine Learning and Stochastic Chemical Modelling: A Novel Synthesis for Emissions Mitigation

Combustion-generated particles called ‘soot’ are the second worst contributor to climate change and are carcinogenic to humans. The formation of soot in combustion devices, however, remains poorly understood. It is therefore essential to gain a better fundamental understanding of soot formation to help reduce soot emissions and mitigate its negative effects on humans and the planet. To facilitate this understanding, the proposed research will develop an Artificial Intelligence (AI) algorithm to predict the rate at which soot is nucleated in flames. Chemical modeling software developed at the University of Michigan will provide highly detailed information about thousands of chemical species in flames. An AI algorithm will learn to predict soot formation from this data pairing, and can then be applied to predict soot formation in other systems. The AI algorithm will improve our understanding of soot formation; how much forms and under what conditions. Improving our scientific understanding and creating a robust soot prediction algorithm will inform engine and other combustion device designers and enable them to select designs with minimal soot production.

Faculty Supervisor:

Seth Dworkin

Student:

Partner:

University of Michigan

Discipline:

Engineering

Sector:

Education

University:

Toronto Metropolitan University

Program: