Image-Based Methods for Materials Characterization and Process Control in Polymer Additive Manufacturing
Date Issued
May 2026
Author(s)
O'Brien, Christopher
The University of Tennessee
Advisor(s)
Chad Duty
Kris Villez
Additional Advisor(s)
Andrew Sarles
Seokpum Kim
Seokpum Kim
Abstract
Modern manufacturing has been transformed by the development of large-format
additive manufacturing (LFAM). Layerwise deposition of three-dimensional structures
enables rapid prototyping and the fabrication of parts with complex geometries that
are impossible with traditional methods. This dissertation focuses on polymerbased
LFAM systems, which face critical challenges including part distortion and
interlayer bonding. Realizing the full potential of LFAM requires overcoming
limitations in materials characterization and process control. The presented work
develops two image-driven solutions through interdisciplinary approaches. First,
deep learning-based segmentation approaches were developed for automated materials
characterization, including the measurement of residual fiber length and the
quantification of porosity. Statistical comparisons demonstrate generalizable, reliable
measurement across varying fiber orientations, morphologies, and microstructures,
reducing measurement time from hours to minutes. Second, thermal imaging is
used for in-situ layer deposition time control. This addresses the limitation that
the deposition process is inherently manual and cannot adapt to varying thermal
conditions. The developed closed-loop feedback control system used real-time
substrate temperature to maintain a setpoint temperature of the previous layer
dynamically. Together, both branches of the dissertation demonstrate how imagebased
methods can enable quantitative process understanding and control for LFAM.
Degree
Doctor of Philosophy
Major
Data Science and Engineering
Embargo Date
May 1, 2027