Ph.D. Proposal Presentation by Jason Aughenbaugh
Wednesday, August 31, 2005

(Dr. Chris Paredis, Chair)

"Managing Uncertainty in Engineering Design Using Imprecise Probabilities and Principles of Information Economics"

Abstract

The engineering design community recognizes that an essential part of the design process is decision-making. Each decision consists of two main phases—problem formulation and problem solution. Existing literature focuses on problem solution using precisely known probabilities. Problem formulation has received considerably less attention. The objective of this thesis is to investigate methods for managing uncertainty during the formulation phase of engineering design decisions, focusing on situations in which probabilities are not precisely known. The existence of such situations has been recognized in the decision theory community but has not been addressed substantially in engineering design problems. The thesis seeks to identify the fundamental characteristics of uncertainty in the context of engineering design, and hypothesizes that subjective, imprecise probabilities are more general and appropriate than currently used representations. Many methods of uncertainty representation are rigorous and internally consistent, but the applicability of the starting axioms must be evaluated. Further, the thesis will develop a method for comparing the practical value of alternative problem formulations and uncertainty representations, taking into account not only the mathematical expressiveness of the formalism, but also the cost of the computations. This is an information management problem to which principles of information economics will be applied for determining an appropriate cost-benefit tradeoff. Finally, the thesis will evaluate decision-policies for problems with imprecisely known probabilities.