Automated metrological analysis of Voronoi lattice structures using X-ray computed tomography

Abstract: The advent of metal laser powder bed fusion (LPBF) has revolutionized the production of functional components, enabling the fabrication of complex geometries previously unattainable by subtractive methods. Among these, stochastic Voronoi lattice structures have attracted significant attention due to their distinct mechanical and functional properties. These structures are increasingly adopted in diverse engineering fields: from aerospace components, where they maximize stiffness while minimizing weight, to automotive energy-absorption systems, and biomedical implants, where their porosity mimics trabecular bone to facilitate osseointegration. The geometric complexity that defines these structures renders conventional inspection methods, such as tactile coordinate measuring machines (CMMs) and optical measurement systems, ineffective, as they cannot access internal features. X-ray Computed Tomography (CT) represents a viable non-destructive inspection solution for the characterization of internal features. However, volumetric data acquired using CT can be affected by noise and image artefacts, requiring dedicated post-processing methodologies. Moreover, the stochastic nature architecture and high geometrical variability of Voronoi lattice structures make reliable inspection challenging even when using CT. While LPBF lattice structures are commonly evaluated by comparison with nominal computer-aided design (CAD) models, this global approach often leads to large discrepancies due to pronounced distortions and local geometric deviations introduced during fabrication. Consequently, localized measurement strategies based on single-strut analysis are required. However, Voronoi lattices exhibit a stochastic distribution of struts, with orientation, length and diameter varying throughout the structure. This variability introduces significant measurement challenges, particularly the isolation and alignment of struts from CT data. The intricate wireframe topology and substantial deviations from the nominal CAD model further complicate this task. This thesis addresses these limitations by developing an automated methodology for the localized metrological characterization of Voronoi lattice structures from CT data. The primary objective is the design and validation of a robust computational framework capable of extracting statistically significant geometric and dimensional information directly from raw CT voxel data. The proposed approach is based on a tailored algorithmic pipeline. CT noise and image artefacts are mitigated through volumetric denoising and adaptive binarization, followed by morphological skeletonization to extract the lattice topology. A key contribution of this work lies in the development of an automated strut isolation solution. By leveraging principal component analysis (PCA), each strut is spatially aligned to a normalized coordinate system, enabling accurate, orthogonal cross-sectional analysis along its entire length. The results demonstrate the ability of the proposed pipeline to isolate and characterize thousands of struts within a stochastic Voronoi lattice structure with high repeatability. The localized analysis provided critical insights into the manufacturing quality, enabling the detection of dimensional and geometric deviations along the strut length, as well as the comprehensive characterization of sintered and partially sintered powder clusters. By correlating such features strut spatial orientation, the developed framework represents a scalable and automated tool for the quality assurance of complex additively manufactured components, effectively bridging the gap between qualitative CT imaging and quantitative metrology.