Ph.D. Proposal Presentation by Hae-Jin Choi
Monday, April 26, 2004
( Dr. Farrokh Mistree, Chair)
"A Robust Design Method under Model and Propagated Uncertainty"
Abstract
Uncertainty in a model estimating performance of a system is one of the important factors to be considered in an engineering system design because engineers usually employ limited number of sampled data or a simplified model rather than a complex model in order to reduce analysis lead-time or due to lack of knowledge. The uncertainty is propagated and accumulated in a series of design tasks so that it could make a final estimated value much deviated from the true one. Many studies have been done to design a system to be robust and reliable from the unexpected effect of the uncertainty. However, the studies have not considered embedded uncertainty which is difficult to be parameterized in a system, and it is common in a complex system design, such as materials design.
In this study, a new approach for robust multi-scale materials design will
be developed that will facilitate search for robust materials design solution
against external uncertainty, model uncertainty, and those propagation along
multiple length-scaled models. Several accomplishments will make this approach
possible, including (1) a robust design exploration method incorporated with
the ‘Error Margin Indices’ concept, (2) a method for formulation
of uncertainty boundaries in a model with fewer experimental data extending
a model calibration technique, (3) a graphical grammar for representing uncertainty
propagation within robust design network, and (4) a method for mitigating propagated
uncertainty in hierarchical and sequential design tasks. Potential application
could be extended to general multiple robust designs with external and model
uncertainty.