Doctoral Defense - Drew Murray

About the Event

The Department of Computer Science and Engineering
Michigan State University
Ph.D. Dissertation Defense

July 11th, 2025 at 1:00pm EST
3105 Engineering Building and https://msu.zoom.us/j/93033959140
Passcode: Upon Request from Vincent Mattison or Advisor

ABSTRACT

SCALABLE NUMERICAL SIMULATIONS OF NONLINEAR AND ATTENUATING ACOUSTIC WAVES

By: Drew Murray
Advisor: Robert McGough

Computer simulations provide important assistance in the development, design, and optimization of ultrasound transducers and systems.  Most ultrasound simulations model linear wave propagation, but better models of shock waves and power law attenuation are still needed.  Furthermore, existing numerical models of acoustic wave propagation either rely on simplifying assumptions of underlying physics and/or cannot be efficiently parallelized for distributed memory computer hardware.

To address these deficiencies, two new numerical implementations are presented that broaden the possible phenomena that can be efficiently modeled with parallel architectures.  First, a new method of injecting power-law attenuation into spatial impulse response calculations is introduced.  This new method is embarrassingly parallel and thus well suited for 3D simulations.  Second, a scheme for modeling nonlinear, attenuating, 3D wave propagation in both homogenous and heterogenous media using discontinuous Galerkin (DG) is derived and implemented.

In addition to full descriptions of these methods, simulation results obtained from the corresponding codes are presented.  3D simulations and comparisons to existing theory are shown for both schemes.  The discontinuous Galerkin (DG) method requires frequent communication between processors but only between nearest-neighbors, which enables excellent parallel performance of DG simulations on large computer clusters.  This is demonstrated on up to 1024 cores through scaling studies.  For the lossy spatial impulse response, simulation studies assess the convergence as the attenuation parameter and spatial resolution are varied.  These computer simulations enable evaluations of fundamental properties that are either presently unavailable or difficult to describe with existing ultrasound software models.