Hayes Stripling

As a result of advances in massive computational techniques in the last decade, the scientific community has taken more stock in predictive mathematics and physics to substantiate high-consequence decisions (weapons testing, for example). Essential to this growing dependence is the accurate reporting of uncertainty and confidence intervals. Therefore, recent work in predictive methods has focused on capturing the numerous sources of error involved with any given scheme.

My current research involves rigorous verification and validation of mainstream uncertainty quantification (UQ) methods. Most of these methods are statistical by design and therefore are not informed by the physics of the problem of interest. Therefore, the use of these methods must account for the strengths, weaknesses, and assumptions in the given technique. A framework under development, called the Method of Manufactured Universes (MMU), allows a modeler to control the uncertainties in a given problem and thereby assess the effect of model assumptions on the predictive capability of the UQ method.

After some very simple iterations of the MMU framework, we have illustrated many of the dangers of `black box' application of UQ methods. We have also shown that the MMU framework is a valuable tool in the development of a UQ strategy for a given problem. We hope to expand this framework and further illustrate the need and benefits of a fully informed UQ study.

Submitted Manuscript: Stripling, HF, Adams, ML, McClarren, RG, and Mallick, BK. The Method of Manufactured Universes for Testing Uncertainty Quantification Methods. Submitted for Review to the Journal of Reliability Engineering and Safety Systems special QMU Edition.

Podium Presentation: Stripling, HF, Adams, ML, McClarren, RG, and Mallick, BK. The Method of Manufactured Universes for Testing Uncertainty Quantification Methods. ANS Student Conference, April 2010. Ann Arbor Michigan.

Poster Presentation: Stripling, Hayes, Marvin Adams, and Ryan McClarren. The Method of Manufactured Universes for Testing Uncertainty Quantification Schemes. Presented at the 2009 Winter Meeting of the American Nuclear Society in Washington DC.