Date of Award

2025-12-01

Degree Name

Doctor of Philosophy

Department

Mechanical Engineering

Advisor(s)

Ryan Wicker

Second Advisor

Francisco Medina

Abstract

This dissertation addresses the persistent challenges of repeatability, reproducibility, and qualification in metal additive manufacturing, with a specific focus on laser-based powder bed fusion of metals (PBF-LB/M). Although this process has matured into a technologically attractive option for various applications, industrial deployment remains constrained by the difficulty of qualifying machines and processes across sites, the limited accessibility of current standards, and the presence of original equipment manufacturer-controlled “black box” parameters that influence part quality but are not visible to or controllable by the user. Existing qualification approaches typically emphasize tensile and fatigue properties derived from lengthy, expensive builds and often overlook the behavior of individual subsystems and time-resolved laser-scanning parameters that directly affect the thermal history of parts. As a result, users may operate machines that appear nominally qualified yet still exhibit variability, hidden processing delays, and machine-to-machine differences that are difficult to diagnose or compare. Guided by these gaps, this work seeks to improve the reliability and reproducibility of laser-based powder bed fusion of metals through three complementary contributions: a user-enabled installation qualification method for the scanner subsystem, an investigation of the microstructural impact of a typical scanner delay parameter, and the development of an additive manufacturing-specific univariate statistical equivalency framework The first part of the dissertation introduces an installation qualification method that enables machine users to assess the health of the scanner subsystem using a single-layer plate melt test. A 25-millimeter-square stainless-steel plate is scanned with a pattern of predesigned vectors that probe laser-on and off delays, scanner jump and polygon delays, scanner acceleration, dimensional accuracy, and slicer file interpretation. Each region of the plate is evaluated through a series of “pass” or “attention required” criteria that can be applied with simple imaging tools. This user-enabled approach eliminates the need for lengthy builds, specialized metrology, or original equipment manufacturer intervention, while retaining sensitivity to scanner and laser timing behavior under actual build conditions. Demonstration across four commercial machines showed that all systems exhibited the attention required condition for multiple tests, including those addressing laser and scanner delays, dimensional accuracy, and slicer handling of thin walls and unmelted regions. These results underscore the prevalence of scanner subsystem variability in nominally qualified machines and illustrate how a low-cost, cross-platform plate melt test can provide actionable diagnostic information and a practical pathway for installation, and potentially operational, qualification by end users. Additionally, an updated version of the user-enabled installation qualification method is provided as a foundation for an upcoming publication detailing its implementation and results. Building on the insight that two-dimensional scan behavior can translate into three-dimensional part inconsistencies, the second part of the dissertation examines the effect of a typical scanner delay parameter, the polygon delay, on local microstructure. Polygon delay influences how the laser beam dwells as the scanner changes direction between two straight vectors, thereby affecting local energy input along contour vectors. Using an open-architecture system that allows individual delay settings, single-track and low-height three-dimensional geometries were fabricated over a range of polygon delay values. Electron backscatter diffraction was employed to quantify grain morphology and local misorientation along the scan paths. The results show that increasing the polygon delay leads to measurable changes in grain aspect ratios and retained plastic strain at single-track corners. In three-dimensional builds, the relationship between polygon delay and local microstructural metrics is more complex due to the thermal history introduced by an added hatch strategy. Yet, the study demonstrates that scanner delay parameters, which are rarely documented or user-configurable, can have a microstructural impact. This finding highlights both the importance of improved transparency in scanner control and the potential of microstructural analysis as a qualification and diagnostic tool for scanner-related process behavior. The third part of the dissertation introduces a novel univariate equivalency framework created for additive manufacturing, implemented as an RStudio-based package designed to determine whether two processes produce statistically equivalent part populations for a selected metric. The method begins with the identification of a reference process that is stable and characterized, followed by the collection of a candidate population from another process or machine operated with nominally identical user-controllable parameters. Rather than relying on traditional hypothesis tests, which are oriented toward detecting if samples are different and do not fully confirm equivalence, the proposed method evaluates whether the candidate distribution falls within an acceptable margin of the reference distribution. This is achieved by partitioning the reference data into discrete bins and using chi-square-based distribution comparisons and effect size measures, such as Cramér’s V, to determine whether the selected populations are equivalent relative to a predefined threshold. A case study involving three commercial PBF-LB/M systems demonstrated that the univariate equivalency method can distinguish machines whose corner deviation distributions are statistically equivalent to a reference from those that are not, even when all are operated with the same nominal processing parameters. In doing so, the method revealed hidden sources of variability associated with scanner behavior and timing parameters that would potentially not be apparent from mean comparisons or a limited set of mechanical tests alone. Taken together, the three contributions form a coherent pathway toward more reliable and reproducible additive manufacturing. The user-enabled plate melt installation qualification method provides a practical tool for routine assessment of scanner performance and early identification of issues that can propagate into part defects. The microstructural investigation of polygon delay establishes that microsecond-level scanner timing parameters, often not available to users, can significantly influence local grain morphology and defect content and therefore require greater transparency and control. The univariate equivalency framework offers a formal statistical mechanism to test if two populations of parts are statistically equivalent, enabling machine-to-machine, parameter-set-to-parameter-set, or site-to-site comparisons. Collectively, these methods address critical gaps in current qualification practices by emphasizing subsystem-level behavior, time-resolved scanning effects on two- and three-dimensional parts, and more sophisticated yet practical statistical comparisons. The work reinforces the central hypothesis that improving additive manufacturing reliability requires increased qualification, greater transparency into process parameters that have historically remained within the manufacturer’s control, and tools to establish statistical equivalence for robust part-property comparisons.

Language

en

Provenance

Received from ProQuest

File Size

193 p.

File Format

application/pdf

Rights Holder

Brenda Leticia Valadez Mesta

Download

Available for download on Sunday, December 31, 2028

Share

COinS