Mitigating nonlinearity in PDEs to improve nonlinear convergence
April 19, 2023
3:00 pm
Abstract
I will argue that nonlinearity in PDE systems can be mitigated through
lifting the systems to a higher-dimensional space. After reviewing basic
properties of Newton's method for solving nonlinear equations, I will
detail the proposed lift-transform-linearize (LTL) approach and show that
the resulting Newton-type algorithms may yield favorable convergence
properties. I will illustrate the ideas first on simple examples, and show
connections to primal-dual interior point methods and mixed finite element
methods. The resulting solvers will be studied for the solution of
large-scale flow problems with severely nonlinear constitutive laws arising
in sea ice modeling.
Webmaster University of Houston
---
Last modified: April 11 2016 - 18:14:43