Prediction of Aerodynamic Heating on High-Speed Missiles Using Gaussian Based Surrogate Models

Date

2025-1

Author

Sivri, Sezer

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

731
views

1581
downloads

Cite This

This thesis presents a detailed study on the use of advanced surrogate modeling techniques for predicting aerodynamic heating in high-speed missiles. Aerodynamic heating, a critical factor in the design of high-speed aerospace vehicles, requires precise and efficient prediction methods to ensure thermal protection and structural integrity. Traditional Computational Fluid Dynamics (CFD) simulations, while accurate, are often impractical for iterative design processes and real-time applications. Previous attempts to reduce computational costs and accelerate the process have utilized loosely coupled CFD methods and low-fidelity models to efficiently estimate surface heat transfer rates under high-speed conditions. With recent machine learning advancements, this research applies surrogate models—including Gaussian Processes, neural networks, and hybrid approaches—to model the complex, nonlinear relationships of aerodynamic heating. Specifically, this study evaluates models such as Exact Gaussian Processes (EGP), Deep Gaussian Processes (DGP), Deep Sigma Point Processes (DSPP), Deep Neural Networks (DNN), and Deep Kernel Learning (DKL). A multi-fidelity model combining high- and low-fidelity data is also explored to improve predictive accuracy and reduce computational cost. The methodology involves applying these surrogate models with data preprocessing steps. The impact of varying training data sizes on the models' predictive capabilities is also investigated. The results demonstrate that these surrogate models offer high accuracy in predicting non-linear heat flux values, highlighting their potential for use in aerospace engineering and other engineering applications. This research establishes a framework for using surrogate models in the design and optimization of high-speed vehicles, enabling more efficient design cycles.

Subject Keywords

Aerodynamic Heating, Machine Learning, Surrogate Modelling, Gaussian Process, Multi-Fidelity Models

URI

https://hdl.handle.net/11511/113385

Collections

Graduate School of Natural and Applied Sciences, Thesis