Predicting Rare Events in Complex Systems

Tuesday, March 6, 2018 - 10:30am - 11:30am
Lind 305
Tuhin Sahai (United Technologies Corporation)
The accurate prediction of the probability and mechanism of catastrophic rare events in complex systems is of paramount importance in a wide variety of applications. In this work, we will explore the construction of novel dynamic importance sampling (DIS) schemes for rare event prediction that are scalable and computationally feasible. In particular, we will explore the use of transport maps coupled with new ansatz that relax restrictive assumptions that prevent the application of these methods in real applications. We will apply these methods to rotorcraft stall and compare them to Monte Carlo based schemes.