Ph.D. Proposal

Anant Girdhar

(Advisor: Prof. Jechiel Jagoda)

"Characterization of Chemical Kinetic Model Performance in Emissions
Prediction using Data-Driven Techniques"

Tuesday, Apr. 1 

1:30 p.m.
Savant 308
 

Abstract
Understanding emissions is crucial to the design, implementation, and analysis of everyday engineering
systems. If left unchecked, these emissions have a wide range of negative impacts on the environment and
our health, including global warming, acid rain, etc. Therefore, we need to gain an understanding of the
thermochemical processes occurring in these combustors. Most modern combustors are operated at fuel
lean conditions. While this has many advantages (higher combustion efficiency, reduced emissions, and
lower operating costs, it also creates various challenges (blow-off, flashback, autoignition). Bluff bodies
are commonly employed as a flame-holding mechanism to increase flame stability. Characterizing bluff
body stabilized flows is, therefore, of practical importance in understanding the engineering trade-offs of
balancing performance and emissions.

These devices work by creating low-speed recirculation zones in the wake of the bluff body. This
provides a region where the hot combustion products and cold reactants can interact and sustain continuous
combustion. Maintaining a flame in the recirculation zone is controlled by a competition between
convection, mixing, and chemistry. Each of these processes has their own timescales and the flow and thermochemical
quantities (temperature, pressure, composition, velocity) are non-linearly coupled with each
other. Many studies have been conducted to understand these relationships, providing insight into the fluid
mechanics, flame dynamics, NOx and CO emissions, etc. Modeling such flows, with their extremely stiff
chemical kinetics, is still expensive even with modern computing capabilities. This has led to the popularity
of reduced chemical kinetic models, which in turn, need to be characterized. However, even for simple
fuels, these models can include tens of species and hundreds of reactions making it difficult to analyze the
intimate relationships that these quantities have with each other.

Data-driven methods have been shown to be incredibly useful in uncovering pattens in highly correlated
data. Additionally, such techniques are able to decompose these relationships and quantify which of them
are the most important. However, with the vast range of such techniques available in the literature and
the high number of variables in combustion problems, additional work needs to be conducted not only to
understand combustion systems, but also the applicability of such methods to this data. Specifically, we
propose using the Principal Component Analysis (PCA) family of methods for this research effort. Through
this research effort we hope to comment on the chemical mechanisms at play in bluff body emissions. That
said, while this study is performed in the context of a bluff body combustor, we also hope to explore PCA and
its variants and extend their application to combustion and chemical kinetics datasets, and comment on their
use, challenges, and possible solutions for gaining physically meaningful insights from such data-driven
physics-agnostic methods.

Committee

• Prof. Jechiel Jagoda - School of Aerospace Engineering (advisor)
• Prof. Lakshmi Sankar - School of Aerospace Engineering
• Prof. Andrew Medford - School of Chemical and Biomolecular Engineering