Graduate Thesis Or Dissertation

 

Hybrid Statistical and Engineering Optimization Architectures in Early Multidisciplinary Designs of Resilience and Expensive Black-box Complex Systems Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/cc08hp06m

Descriptions

Attribute NameValues
Creator
Abstract
  • Practical engineering design problems are generally multi-disciplinary with limited budget and high risk in terms of life loss, economic resources, etc. In the early phase of such problems, selection of true efficient designs is desired while minimizing overall design cost by avoiding expensive search processes. However, the task is difficult for a simple optimization framework due to the formulation complexity, high function evaluation cost, uncertain design parameters etc. Thus, the overall research goal is to develop complex, hybrid optimization architectures for solving early design problems considering the trade-off among model complexity, performance and cost. We start by comparing multiple architectures, and investigated a nested bi-level architecture for early resilience design with discrete design space and with a trade-off among multiple objectives at different risk level scenarios. The work then focused on increased problem complexity with black-box functions in a mechanical design classification problem with discontinuous design space using a sequential Bayesian Optimization (BO) architecture to locate an unknown creep-fatigue failure constraint boundary. The work then extends a weighted Tchebycheff black-box multi-objective BO (MO-BO) architecture for mechanical design with a trade-off between design risk and cost, with model calibration through regression analysis of unknown parameters. Finally, we investigate an iterative regression model selection procedure, nested into the proposed MO-BO, to enhance design flexibility, estimation and overall performance. This work can be applicable to any domains of complex or/and expensive black-box system design problems.
Contributor
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Funding Statement (additional comments about funding)
  • This research was funded in part by NASA Grant 80NSSC18M0106. The opinions, findings, conclusions, and recommendations expressed are those of the authors and do not necessarily reflect the views of the sponsor.
  • This research was funded in part by DOE NEUP DE-NE0008533. The opinions, findings, conclusions, and recommendations expressed are those of the authors and do not necessarily reflect the views of the sponsor.
Publisher
Peer Reviewed
Language

Relationships

Parents:

This work has no parents.

In Collection:

Items