MEMS Seminar: Predicting Rare Events in Complex Dynamical Systems

Nov 7

Wednesday, November 7, 2018

12:00 pm - 1:00 pm
Fitzpatrick Center Schiciano Auditorium Side A, room 1464

Add to calendar

Presenter

Dr. Tuhin Sahai | United Technologies Research Center

Modern day engineering systems can display a wide range of complex behavior that can make the design and analysis of robust systems particularly challenging. In particular, predicting the dynamic probability (along with the associated mechanism) of high consequence catastrophic events is especially germane to this setting. Accurate and efficient algorithms for the computation of rare events are critical for safe operation and design of such systems. 

In this work, we develop novel dynamic algorithms for computing the probability and “most likely paths” for system failure. These methods rely on optimal path computations from states to failure regions. It is well known that these methods require the solution of a Hamilton-Jacobi equation which is computationally intractable for realistic systems. Subsolutions of the underlying PDE give rise to asymptotically efficient estimators for rare events but their performance degrades in the presence of attractors. We outline a novel approach for constructing subsolutions in high-dimensional dynamical systems with non-self-adjoint dynamics which correspond to efficient methods for escaping the attractor dynamics. We show how this approach can be used for dynamic importance sampling schemes and splitting based approaches for rare event simulation.

We apply our approach to a problem of rotorcraft stall and demonstrate orders of magnitude improvement over state-of-the-art methods for rare event prediction. This work is supported by the DARPA EQUiPS program.

Tuhin Sahai is an Associate Director at United Technologies Research Center (UTRC) where he leads multiple projects in the area of Applied Mathematics. He currently serves as a PI on the DARPA EQUiPS program where his project focuses on rare event estimation in complex systems. He also serves as a PI for the DARPA Lagrange program, where his team is building efficient algorithms for nonlinear mixed-integer programs. He previously has served as a PI for AFOSR’s CCMI, DARPA GUARD-DOG and DARPA N7 programs. His research expertise lies in the areas of uncertainty analysis, dynamical systems, and scalable graph and data analytics. Sahai received the 2012 Technical Excellence Award, the highest individual award at UTRC. In 2013, he was invited for the Frontiers of Engineering (FoE) Symposium and was awarded the subsequent Grainger Foundation award, both by the National Academy of Engineering (NAE). Sahai earned his PhD on January 2008 from Cornell University where he was the McMullen Fellow and a recipient of the H.D. Block teaching award. Prior to that, Sahai earned his BTech and MTech degrees in Aerospace Engineering from the Indian Institute of Technology, Bombay.

Lunch will be served at 11:30 am.

Hosted by Dr. Wilkins Aquino

Contact

Laura Dzwonczyk
laura.dzwonczyk@duke.edu