Theses and Dissertations

Advisor

Mun, Sungkwang

Committee Member

Sescu, Adrian

Committee Member

Sharma, Gehendra

Committee Member

Kim, Han-Gyu

Date of Degree

8-7-2025

Original embargo terms

Immediate Worldwide Access

Document Type

Graduate Thesis - Open Access

Major

Computational Engineering

Degree Name

Master of Science (M.S.)

College

James Worth Bagley College of Engineering

Department

Computational Engineering Program

Abstract

Over the past decade, additive manufacturing (AM) has gained substantial use in industry and research. It is a promising alternative to traditional subtractive and casting manufacturing. AM not only saves time and materials but can also construct intricate structures within components for specific purposes. Wire-Arc Directed Energy Deposition (WA-DED) is one of the additive manufacturing technologies that offers a high extrusion rate of metal wire materials to fabricate near-net-shape metal components. Accurate temperature control is crucial for controlling the microstructure of metal components fabricated through the AM process. Machine learning has the ability to analyze real-time sensor data, such as temperature, into Data-driven models to predict defects. Data-driven models can deliver accurate predictions when large and clean datasets are available. However, obtaining quantity and quality data is expensive. An effective solution for monitoring temperature is Physics-Informed Neural Networks (PINNs). PINNs integrate physics equations with conventional neural networks as a framework. This framework addresses problems governed by physics equations while reducing the reliance on datasets and ensuring physical consistency.

Sponsorship (Optional)

Engineer Research and Development Center (ERDC), Center for Advanced Vehicular Systems (CAVS)

Share

COinS