Ph.D. Thesis Defense
Brandon J. Johnson
(Advisor: Prof. Dimitri N. Mavris)
12:00PM, Monday, March 27, 2017
Weber Space Science and Technology Building (SST-II)
Collaborative Visualization Environment (CoVE)
An Uncertainty Quantification and Management Methodology to Support Rework Decisions in Multifidelity Aeroelastic Load Cycles
Cost overruns and schedule delays have plagued almost all major aerospace development programs and have resulted in billions of dollars lost. Design rework has contributed to these problems and one approach to mitigating this risk is reducing uncertainty. Failure during flight test results is one of the most significant and costly rework efforts. The main purpose of this thesis is to reduce the risk of this rework by improving the loads analysis process. The main objective of loads analysis is to determine the worst-case loading conditions to design the structure. Observing the current approach has revealed some shortcomings related to uncertainty and the allocation of load and structural margins. Uncertainty quantification and management were chosen to address these limitations and a framework is proposed to support decisions for rework in loads analysis.
Key aspects of the framework include utilizing a Bayesian network for modeling the loads process as well as propagating various uncertainty sources. Bayesian-based resource allocation optimization is used to reduce and manage uncertainty. Finally, the goal of the framework is to determine the optimal tradeoffs between aerodynamic fidelity and margin allocation to minimize the risk of rework while considering their respective costs within a finite budget. Assigning costs related to fidelity and margins are intended to reflect the users’ prioritization of uncertainty, computational cost and performance degradation through weight penalties.
Five experiments were conducted related to epistemic uncertainty quantification, sensitivity analysis, developing the uncertainty management system and finally experiments to improve and evaluate the framework against the current approach. The contributions of this thesis include; an integrated modeling and simulation environment for the load analysis process, uniquely applying a Bayesian network for efficient uncertainty modeling and propagation, and a viable cost-based uncertainty management system for loads analysis among others.
Prof. Dimitri Mavris
Prof. Stephen Ruffin
Prof. Daniel Schrage
Dr. Frode Engelsen
Dr. Neil Weston