Uncertainty Quantification at the Confluence of Engineering, Science, Mathematics and Decisions
Thursday, September 13, 3:30PM – 5PM
Dr. Roger Ghanem
The pace of scientific discovery has been accelerated in recent decades due largely to technological innovations in sensing and computing. Our ability to observe the physical world at scales ranging from quantum to stellar is indeed matched by our ability to interpret these observations, with some sophistication, using current computational resources. The simultaneous maturity of these experimental and interpretive capabilities has accelerated the creation of knowledge, and has even redefined its significance.
Uncertainty can be constructively attributed to lack of knowledge, either experimental or fundamental. It thus stands to reason that as the landscape of knowledge evolves, so should our understanding of associated uncertainties. Scientific discourse about Uncertainty Quantification, Verification and Validation has been on a high note for the past decade, driven to a large extent by the demand from various constituencies that computational resources deliver on their promise of reproducing if not predicting physical reality. Clearly, the value of such an achievement will be enormous, ranging from superior product design to disaster mitigation and the management of complex systems such as financial markets, SmartGrids, and society itself. A number of challenges are quickly identified on the path to delivering this computational twin. First, reality itself is elusive and is not always described with commensurate topologies. Second, mathematical models by necessity are incomplete, introducing further uncertainties as to the proximity of their solution to reality. Third, even when these models are accurate and well-resolved numerically, uncertainties are introduced at the physical realization stage, for instance when a device is actually built, reflecting additional tolerances and imperfections. Thus, while probabilistic modeling itself has been a mainstay of mechanics for many decades, one of the recent challenges has been to adapt probabilistic representations and computations to the needs of highly-resolved physical models.
This talk will provide an overview of the work of my group at tackling many of technical issues relevant to the above challenges. Innovations in physical modeling that are uniquely adapted to large computing and large data will be described. I will describe applications from across science and engineering.
Hosted by Robert Moser