Doctoral Dissertations
Date of Award
5-2021
Degree Type
Dissertation
Degree Name
Doctor of Philosophy
Major
Data Science and Engineering
Major Professor
Jamie B. Coble
Committee Members
Vincent C. Paquit, Sudarsanam S. Babu, Anahita Khojandi
Abstract
In recent years, metal additive manufacturing processes have become popular choices for part production especially for low volume, high complexity parts. To enable widespread adoption of these methods, it is essential to understand the link between process parameters and part properties. This is particularly because additive manufacturing processes cause inherently complex thermo-mechanical cycles and drastically different local process conditions within a part, compared to conventional manufacturing processes like casting and forging. Additionally, properties of feedstock material like metal powder impact final part properties. The focus of this dissertation is on development of data-driven methods using in situ monitoring, as a step towards part qualification.
The primary contributions are:
(a) In situ monitoring of powder feedstock: Monitoring in situ powder spreading in powder bed AM is important to avoid part defects. Data-driven methods are developed to study new and recycled powder feedstock in the Electron beam Powder Bed Fusion (E-PBF) process, using log files generated in the Arcam ™ system. Results indicate that powder raking behavior reflects powder spreadability on the powder bed and is closely related to powder morphology.
(b) In situ monitoring of thermal behavior in the Laser Powder Bed Fusion (L-PBF) process: Thermal signatures in metal AM processes are closely linked to part microstructure and properties. Thermal signatures vary as a complex function of scan strategy, part geometry and process parameters. A similarity analysis methodology is developed to identify regions of self-similar thermal signatures in the L-PBF process using infrared imaging. Thermal signatures reflect localized solidification conditions and their similarity across regions points to similar regional microstructures.
(c) Demonstration of generalizability of the similarity analysis methodology to lattice structures: To demonstrate validity of similarity analysis methodology to different geometries, the algorithm is applied to lattice structures to evaluate the effect of geometry and process parameters on thermal signatures. Results indicate dependence of thermal signatures on nodal connectivity in the lattice and on inert gas used. Builds in Helium are observed to cool faster than in Argon. The observed thermal signature difference may lead to different part microstructures and associated properties.
Recommended Citation
Chandrasekar, Sujana, "Development of Data Science Tools for Part Qualification in Additive Manufacturing. " PhD diss., University of Tennessee, 2021.
https://trace.tennessee.edu/utk_graddiss/6690