ORCID

0009-0002-0318-3263

Keywords

Machine Learning, Ceramic Matrix Composites, Physics Informed Neural Networks, Hydrogen Torch Test, Gas Turbine Engines, Thermal Characteristics Prediction

Subject Categories

Aerospace Engineering | Materials Science and Engineering | Mechanical Engineering

Abstract

Through the integration of experimental characterization, machine learning, deep learning, numerical modeling, and physics-informed artificial intelligence, this dissertation explores the thermal performance of polymer-derived ceramic matrix composites (CMCs) for hydrogen combustion environments. The polymer infiltration and pyrolysis (PIP) process was used to create Yttria-stabilized zirconia (YSZ)-fiber-reinforced ceramic matrix composites, which were then experimentally assessed under hydrogen torch and hydrogen combustion conditions typical of next-generation gas turbine and aerospace propulsion systems. Front- and back-surface temperature measurements were used to continually monitor thermal reactions, and post-test microstructural characterization and numerical simulations were carried out to evaluate material integrity and heat-transfer behavior. To forecast the ablation performance of ceramic matrix composites, the dissertation first develops six supervised machine-learning regression models: Decision Tree, Random Forest, Support Vector Machine, Gradient Boosting, Extreme Gradient Boosting, and AdaBoost. The models based on boosting had the best prediction accuracy among these methods. Building on these findings, back-surface temperature prediction was further enhanced by an optimized deep artificial neural network (DANN) using two hidden layers with ReLU activation, which produced a R² of 0.9671, RMSE of 16.45 °C, MAE of 14.07 °C, and MAPE of 3.92%.Finally, a physics-informed neural network (PINN) framework was developed to estimate the time-dependent effective thermal conductivity of YSZ/Si(B)CN ceramic matrix composites under repeated exposure to hydrogen torches and to address the inverse transient heat-conduction problem. PINN effectively recreated physically significant conductivity evolution by incorporating the governing heat-transfer equation into the learning process. The developed methodology offers a dependable and data-efficient way to speed up the design, validation, and optimization of high-temperature materials for advanced thermal protection systems, gas turbine engines, and hydrogen-powered aerospace propulsion.

Completion Date

2026

Semester

Summer

Committee Chair

Gou, Jihua

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Mechanical Engineering

Format

PDF

Document Type

Dissertation

Language

English

Share

COinS
 

Accessibility Statement

This item was created or digitized prior to April 24, 2027, or is a reproduction of legacy media created before that date. It is preserved in its original, unmodified state specifically for research, reference, or historical recordkeeping. In accordance with the ADA Title II Final Rule, the University Libraries provides accessible versions of archival materials upon request. To request an accommodation for this item, please submit an accessibility request form.