Leveraging Gaussian Process Sampling for Sensitivity Analysis and Optimization in Engineering Design

High-fidelity simulations and physical experiments are fundamental in engineering analysis and design. However, their high computational cost often prohibits their application in global sensitivity analysis (GSA), optimization, and automated structural health monitoring (SHM). Gaussian processes (GPs) are proposed as a promising solution to this challenge. GPs inherently facilitate efficient sampling strategies, enabling informed decision-making under uncertainty by extracting information from a subset of potential functions for the model of interest. Despite their widespread use in machine learning and scientific computing, and the potential they hold for realizing intelligent infrastructural systems via Digital Twin, GP sampling strategies have received little attention in engineering applications. This thesis thus presents the mathematical foundations of GPs and provides a detailed implementation of two sampling methods—Fourier decomposition-based and pathwise conditioning—for generating approximate stochastic functions from GPs. It then discusses the application of these sampled stochastic functions in engineering tasks such as GSA, single-objective optimization, and multi-objective optimization. Towards realizing intelligent engineering systems, this thesis finally proposes a DT framework that leverages GP for efficient model updating and optimal decision-making.

Name thesis:degree_name

Master of Science in Civil Engineering

Discipline thesis:degree_discipline

Civil Engineering

Grantor

University of Houston

Year dc:date.issued

2025

Author dc:creator

• Ajenifuja, Nafeezat Adetoro 1999-

Advisor dc:contributor.advisor

• Zhang, Ruda

Committee members dc:contributor.committeemember

• Beck, Abigail
• Cao, Jian
• Nakshatrala, Kalyana Babu

• Civil engineering

Language dc:language.iso

English

Handle dc:identifier.uri

https://hdl.handle.net/10657/20699

OAI identifier oai:identifier

oai:uh-ir.tdl.org:10657/20699