H23B-0884:
Noise propagation in hybrid models of nonlinear diffusion systems
Tuesday, 16 December 2014
Soren Taverniers and Daniel M Tartakovsky, University of California San Diego, La Jolla, CA, United States
Abstract:
In order to increase the fidelity of current state-of-the-art multiphysics simulations, it is necessary to take into account the effects of noise generated either in the bulk or at the boundary of the domain of interest. In highly nonlinear systems, where fluctuations can significantly alter the mean macroscale behavior of the system of interest, the accuracy of the numerical time integration scheme depends critically on its ability to correctly propagate noise both in each physics component and across the intercomponent interface. Using two multiscale nonlinear parabolic problems excited by a white noise at one of the boundaries as test cases, we propose a partitioned, tightly-coupled time integration scheme and analyze its ability to capture noise propagation across a coupling interface and evaluate the computational efficiency and accuracy of the coupling.