Date of Award

12-2010

Document Type

Thesis

Degree Name

Master of Science (MS)

Legacy Department

Mechanical Engineering

Advisor

Fadel, Georges M

Committee Member

Kurz , Mary E

Committee Member

Thompson , Lonny L

Abstract

In the contemporary world of engineering, engineers strive towards designing reliable and robust artifacts while considering and attempting to control manufacturing costs. In due course they have to deal with some sort of uncertainty. Many aspects of the design are the result of properties that are defined within some tolerances, of measurements that are appropriate, and of circumstances and environmental conditions that are out of their control. This uncertainty was typically handled by using factors of safety, and resulted in designs that may have been overly conservative. Therefore, understanding and handling the uncertainties is critical in improving the design, controlling costs and optimizing the product. Since the engineers are typically trained to approach problems systematically, a stepwise procedure which handles uncertainties efficiently should be of significant benefit.
This thesis revises the literature, defines some terms, then describes such a stepwise procedure, starting from identifying the sources of uncertainty, to classifying them, handling these uncertainties, and finally to decision making under uncertainties and risk. The document elucidates the methodology introduced by Departments of Mathematical Science and Mechanical Engineering, which considers the after effects of violation of a constraint as a criterion along with the reliability percentage of a design. The approach distinguishes between aleatory and epistemic uncertainties, those that can be assumed to have a certain distribution and those that can only be assumed to be within some bounds. It also attempts to deal with the computational cost issue by approximating the risk surface as a function of the epistemic uncertain variables.
The validity of this hypothesis, for this particular problem, is tested by approximating risk surfaces using various numbers of scenarios.

Share

COinS