Time-domain electromagnetic solvers on heterogeneous systems: Theory and implementation

Feng, Junda

Permalink

https://hdl.handle.net/2142/116130 Copy

Description

Title

Time-domain electromagnetic solvers on heterogeneous systems: Theory and implementation

Author(s)

Feng, Junda

Issue Date

2022-07-21

Director of Research (if dissertation) or Advisor (if thesis)

Peng, Zhen

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• computational electromagnetics
• time-domain Maxwell equations
• high-performance computing

Abstract

Nowadays, the time-domain electromagnetic solver plays an important role in the field of computational electromagnetics due to its ability to simulate time-domain wave physics and efficiency in solving wide-band structures. In this thesis, we focus on the derivation, implementation and optimization of time-domain electromagnetic solvers based on partial differential equations. To be specific, we first investigate and validate the Yee’s FDTD (finite difference time-domain), nodal FETD (finite element time-domain) and nodal DGTD (discontinuous Galerkin time-domain) in two dimensions. Secondly, we formulate the 3D DGTD with absorbing boundary condition based on curl-conforming vector elements and multiple numerical fluxes from an interior penalty approach. Due to the great computing flexibility the discontinuous Galerkin method offers, we explore the parallelization of the DGTD by implementing a CPU (central processing unit) version and several GPU (graphic processing unit) counterparts on a typical heterogeneous system where electromagnetic simulation software is usually installed. Especially, in addition to a basic GPU implementation, three parallelization strategies that make good use of available computing resources are proposed for the case where GPU memory is not sufficient. We conclude the thesis with a few numerical examples and show the capability of our proposed method by a long time simulation for more than 10^5 time steps of a model meshed into approximately 1.7 million tetrahedral elements.

Graduation Semester

2022-08

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Junda Feng

Owning Collections