Towards Deep Learning-enabled Engineering for Expedited Understanding and Mitigation of Physical Phenomena

Abstract

The dissertation is dedicated to developing deep learning-enabled engineering methods for achieving expedited understanding and mitigation of physical phenomena. While physical phenomena are crucial in comprehending our world, from the behavior of matter to the functioning of complex systems, challenges still exist in facilitating them for various scientific and engineering fields. The dissertation addresses the challenges and limitations associated with physical phenomena via data-driven approaches. Notably, the dissertation introduces a novel field of study called a deep learning-enabled engineering (DLE), which integrates physical knowledge into data-driven deep learning frameworks to enhance our understanding of physical phenomena and improve the ability to mitigate their effects. The dissertation presents a comprehensive study on the development of DLE methods that are tailored for several scientific and engineering domains concerned with physical phenomena: 1) heat transfer behaviors in the steel-manufacturing process, 2) wave propagation and scattering characteristics, 3) resonance-based acoustic metamaterials, and 4) localization of sound-emitting sources. The findings of the proposed DLE methods indicate their feasibility in achieving a precise and accelerated understanding of physical phenomena and overcoming the limitations that arise in physical systems.