Woodruff School of Mechanical Engineering

Seminar

Title:

The Default Uncertainty is Always ZERO

Speaker:

Dr. William Rider

Affiliation:

Sandia National Laboratories

When:

Thursday, April 26, 2018 at 2:00:00 PM

Where:

MRDC Building, Room 4211

Host:

Yan Wang
yan.wang@me.gatech.edu
4048944714

Abstract

We tend to accept a phenomenally common and unfortunate practice where an uncertainty is defined implicitly as identically equal to ZERO. In other words, if we do nothing at all, no work, no judgment, the modeling or experimental uncertainty is ZERO. In reality, this is a situation where the lack of knowledge is dramatic, and the estimated uncertainty is large to account for this lack of knowledge. Uncertainty quantification is a major component of well executed verification and validation (V&V). The conduct of V&V can be thought of as following the classical scientific method as applied to modern modeling and simulation. Today, we have access to numerous tools such as Dakota that automate key aspects of uncertainty assessment. Often the use of these tools seemingly provides a single turnkey uncertainty estimate. Frequently this is accepted as complete without the subtle realization that it is a part of a far broader scientific investigation. As such, the automation of uncertainty quantification is elusive and requires a complex work flow utilizing multiple tools used in concert with detail analysis of computational and experimental work each having its uncertainties. One of the key things to utilize is the application of expert knowledge and judgment where existing tools either do not or cannot exist. When used in a coordinated and mindful manner, the combination of tools, workflows and expertise can successfully avoid unjustifiable ZERO uncertainties. To demonstrate the execution of V&V properly we will highlight current work in validating hypersonic fluid flow with turbulence modeling. This work will demonstrate how we are concretely examining the uncertainty in a validation case to elucidate the uncertainties in the modeling.


Biography

Dr. William Rider is a distinguished member of technical staff who joined Sandia in 2007 after working 18 years at LANL. Bill serves as the Chief Scientist for Sandia ASC's ATDM Project, and has worked in the ASC V&V Program since its inception. He is an associate editor of the ASME Journal of Verification Validation and UQ, and a voting member of the ASME V&V 20 Standards Committee. He is an international expert on V&V, computational physics, turbulence and computational fluid dynamics. He is the author of two books and over 200 journal articles and technical reports. He is a member of ASME and SIAM, and writes a blog that touches upon a broad spectrum of topics including V&V, UQ, CFD and Computational Physics, 'The Regularized Singularity' http://wjrider.wordpress.com.

Notes

Meet the speaker