Author ORCID Identifier
Date Available
1-26-2027
Year of Publication
2026
Document Type
Doctoral Dissertation
Degree Name
Doctor of Philosophy (PhD)
College
Engineering
Department/School/Program
Mechanical Engineering
Faculty
Savio J. Poovathingal
Faculty
Alexandre Martin
Faculty
Qiang Cheng
Abstract
Thermal Protection System (TPS) materials are critical to the survivability of hypersonic vehicles during atmospheric entry and sustained high-speed flight. These systems are exposed to extremely high temperatures, pressures, and velocity gradients, leading to the formation of strong shock waves and chemically reacting boundary layers. The interaction between the hypersonic flow and the TPS material gives rise to a range of complex thermo-chemical non-equilibrium phenomena, including gas dissociation, ionization, vibrational and rotational excitation, and gas-material interactions such as volumetric diffusion and ablation. Accurately modeling these coupled multi scale processes is essential for evaluating the performance of TPS materials. However, the state-of-the-art simulation approaches often rely on empirical correlations formulated based on physical experiments. These empirical models ignore the effects of microscale and mesoscale phenomena on macroscale closure parameters like recession rate and material properties of the underlying TPS material.
This thesis focuses on the development of data-driven closure models to enhance the simulation of gas-material thermo-chemical processes relevant to TPS materials in hypersonic conditions. These models aim to bridge the gap between high-fidelity mesoscale physics and computationally affordable engineering simulations by incorporating detailed data into reduced-order supervised learning models suitable for use in macroscale solvers. The closure models are designed to enhance the modeling of TPS materials, where accurate prediction of material properties and recession rate is most critical.
The approach begins with the collection of high-fidelity data from direct simulation Monte Carlo (DSMC), and an in-house simulation framework which uses DSMC simulations to simulate the ablation of TPS materials. These datasets are then used to train supervised learning models that approximate closure terms traditionally used in macroscale solvers, such as material properties and recession rates of TPS materials. Sophisticated data sampling techniques are adopted to ensure that the closure models are trained on flow conditions encompassing the entire spectrum of values typically encountered in hypersonic flight. Sampled data are trained using support vector regression techniques to ensure a closed form relationship between the predicted variable and the user-defined input parameters. A closed form solution is of paramount requirement to couple the trained supervised learning model with macroscale simulations. A closed form solution ensures that the closure terms are continuous and differentiable, a necessity to find the Jacobian matrix in macroscale solvers. The robustness of the closure models is corroborated by verifying its performance on both trained and test data. The closure model developed for predicting the material properties is further validated with experimental results and successfully coupled to material response solvers. The closure models, once trained on high fidelity data are capable of instantaneous predictions which provided a cost effective alternative and exhibited indifference to interpolation errors unlike their empirical counterparts.
In conclusion, this work demonstrates the potential of data-driven closure models to fundamentally enhance the simulation of complex thermo-chemical phenomena encountered by TPS materials during hypersonic flight. By combining machine learning with physical constraints and high-fidelity data, the resulting models offer a path toward predictive, efficient, and robust tools for TPS design and analysis. The frameworks and methodology developed in this work can be easily leveraged to re-train new closure models for any type of TPS material developed now and in the future.
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2026.08
Funding Information
This work was supported by a Space Technology Research Institutes grant from NASA’s Space Technology Research Grants Program under grant number 80NSSC21K1117.
Research was sponsored by U.S. Army Combat Capabilities Development Command (DEVCOM) Army Research Laboratory (ARL) and was accomplished under Cooperative Agreement Number W911NF-21-2-0075, and supported by the Office of the Under Secretary of Defense for Research and Engineering under award number FA9550-22-1-0342.
Recommended Citation
Mohan Ramu, Vijay B., "Data driven closure framework for hypersonics" (2026). Theses and Dissertations--Mechanical Engineering. 254.
https://uknowledge.uky.edu/me_etds/254
