Abstract
Simulation-led design is becoming an important part of thermal systems design. Simulations of thermal systems continue to improve in their fidelity, but this comes with an increased computational cost. Additionally, uncertainty in simulation inputs, such as thermophysical properties, leads to uncertainty in simulation outputs. For simulations to be used rigorously for simulation-led design, this uncertainty must be quantified. Performing uncertainty quantification on thermal simulations is made difficult by their many uncertain parameters and high computational cost. Multifidelity uncertainty quantification is a set of methods that reduce the cost of uncertainty quantification by leveraging high-fidelity, high-cost simulations with low-fidelity, low-cost simulations. Reduced order models are used as the low-fidelity models in this research. Data driven reduced order models accelerate modeling by using past data from a system to identify a small set of basis functions that can be used to describe the state of the system. In Chapter 2, it is shown that the proper orthogonal decomposition can be used to form an effective basis for reduced order methods. This research demonstrates that multifidelity methods using reduced order models can be used to accelerate uncertainty quantification for computationally expensive thermal simulations with many uncertain input parameters. This is demonstrated with thermal simulations of two scenarios: a model of a notional munition in a fire environment used by Sandia National Labs for developing uncertainty quantification workflows and a focused ultrasound thermal ablation treatment of a breast tumor. Multifidelity uncertainty quantification decreased error in mean estimates on average by approximately 30% relative to the traditional Monte Carlo uncertainty quantification method for both of these scenarios.
Degree
MS
College and Department
Ira A. Fulton College of Engineering; Mechanical Engineering
Rights
https://lib.byu.edu/about/copyright/
BYU ScholarsArchive Citation
Bates, Jakob G., "Leveraging Reduced Order Models and High-Fidelity Simulations for Efficient Multifidelity Uncertainty Quantification of Thermal Systems" (2025). Theses and Dissertations. 11284.
https://scholarsarchive.byu.edu/etd/11284
Date Submitted
2025-05-21
Document Type
Thesis
Permanent Link
https://arks.lib.byu.edu/ark:/34234/q2e1ae4c1e
Keywords
Uncertainty quantification, reduced order modeling, heat transfer, focused ultrasound
Language
english