Back to search

McMaster University

Geometry-Aware Sensor Fusion for Defect Detection in Laser Powder Bed Fusion

Abstract

dc:description.abstract

Laser powder bed fusion (LPBF) enables fabrication of complex metal components with high geometric freedom, but internal defects such as porosity and cracking continue to limit qualification for safety-critical applications. These defects are difficult to observe directly during fabrication, particularly when complex features alter heat flow, acoustic propagation, and sensor response. This thesis develops and evaluates a geometry-aware, multi-sensor monitoring framework for LPBF defect screening using synchronized structure-borne acoustic emission and coaxial two-colour pyrometry. The thesis is presented in integrated-article format. Chapter 2 reviews machine-learning applications in LPBF process monitoring and control and establishes the literature foundation. Chapter 3 develops the sensing and data-generation foundation, including an acoustic-emission-only pilot, optical integration of coaxial two-colour pyrometry, synchronization strategy evaluation, high-rate data acquisition, geometry-aware design of experiments, Archimedes density measurements, micro-computed tomography validation, and signal repeatability/alignment analysis. Chapter 4 presents the main monitoring contribution: feature-level fusion of synchronized acoustic and thermal representations for crack-like and porosity-like defect screening across geometry-diverse LPBF coupons. The monitoring study uses synchronized AE and analogue pyrometer-output channels sampled on a common 1 MHz DAQ time base, with pyrometer interpretation constrained by the instrument update rate. Additionally, it uses independent Archimedes and micro-computed tomography ground truth, modality-specific latent representations, feature-level fusion, and grouped held-out evaluation. The primary fused model achieved F1 = 0.974 for crack-like detection and F1 = 0.987 for porosity-like detection, with recall = 1.00 for both heads and AUROC values of approximately 0.998 and 0.993, respectively. The results support synchronized acoustic–thermal fusion as a promising high-sensitivity in situ screening approach for LPBF. By emphasizing synchronization, geometry-aware experimental design, independent ground truth, and grouped evaluation, the thesis establishes a monitoring foundation for future STL-aware spatial mapping, geometry-stratified validation, cross-machine and multi-material transfer, and supervisory process-control integration.

Degree

thesis:*
Department dc:contributor.department
Mechanical Engineering
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Magolon, Marcin Stanisław
Advisor dc:contributor.advisor
  • Elbestawi, Mohamed

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
McMaster University
Base URL
macsphere.mcmaster.ca/server/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Magolon, Marcin Stanisław. Geometry-Aware Sensor Fusion for Defect Detection in Laser Powder Bed Fusion. 2026. https://hdl.handle.net/11375/33364