Degree

Doctor of Philosophy (PhD)

Department

Mathematics

Document Type

Dissertation

Abstract

We present an approach to shape optimization problems that uses an unfitted finite element method (FEM). The domain geometry is represented, and optimized, using a (dis- crete) level set function and we consider objective functionals that are defined over bulk domains. For a discrete objective functional, defined in the unfitted FEM framework, we show that the exact discrete shape derivative essentially matches the shape derivative at the continuous level. In other words, our approach has the benefits of both optimize-then- discretize and discretize-then-optimize approaches.

Specifically, we establish the shape Fréchet differentiability of discrete (unfitted) bulk shape functionals using both the perturbation of the identity approach and direct perturbation of the level set representation. The latter approach is especially convenient for optimizing with respect to level set functions. Moreover, our Fréchet differentiability results hold for any polynomial degree used for the discrete level set representation of the domain. We illustrate our results with some numerical accuracy tests, a simple model (ge- ometric) problem with known exact solution, as well as shape optimization of structural designs.

We also present some analysis of the Landau–de Gennes model for liquid crystals in an unfitted framework and derive a consistency estimate for a scalar-valued version of this PDE. These results will (eventually) form the foundation of an unfitted method for the Landau–de Gennes model.

Date

4-3-2025

Committee Chair

Walker, Shawn

Download

Share

COinS