Factory-Level Analysis of Additively Manufactured Lattice Structures Using Automated Metrology

Modern equipment for additive manufacturing (AM) enables scalable production of parts having complex geometry, for which each part can have thousands of features. Defects in these parts affect the parts’ functionality; however, it is technically challenging to measure these complex geometries and track the variance of printed parts across multiple AM machines in a factory. This presentation describes tools for automated metrology and their application for tracking part defects at the factory level.

In this study, we fabricated 90 regular hexagonal lattice parts using continuous liquid interface production (CLIP). Each part had 237 individual thin-wall features, with designed thickness of 0.5 mm and length of 2.5 mm. Fabrication of the parts mimicked a production schedule, utilizing multiple printers, interchangeable hardware, and locations across the build area of each printer. Detailed geometric measurements of each lattice wall were collected using an automated scanning metrology, and we used statistical methods to investigate the role of factory-level manufacturing parameters on the geometric quality of the lattice parts.

The mechanical performance of many lattice structures depends on the geometry of the individual features. We developed an automated metrology capable of efficiently extracting hundreds of geometric features from part image data. We begin by collecting an image of the part using a high-resolution document scanner. The metrology algorithm ingests this image data and identifies wall edges using Canny edge detection. A corner detection algorithm identifies lattice nodes, and the image is segmented into snapshots of individual walls. A Hough transform resolves two linear equations that best describe the identified wall edges with respect to a reference coordinate system. Finally, geometric measurements such as length, thickness, and orientation, are calculated from the extracted equations. Length is measured as the distance between intersecting walls, thickness is measured as the separation between wall edges, and orientation is measured as the slope of the Hough lines. The described metrology rapidly and accurately captures hundreds of geometric features per part, adapts to changing lattice geometries within a part, and functions on different lattice structures entirely. This metrology is successfully implemented in analysing two-dimensional hexagonal, as well as three-dimensional octet-truss lattice parts.

The speed and flexibility of this automated metrology allowed us to study defects across a batch of parts manufactured in a factory. We investigated three M2 printers from Carbon, and three cassettes that can be interchangeably paired with the printers. The cassette is a stationary reservoir that mounts inside the printer, and holds liquid resin during printing. A window at the base of the cassette permits the transmission of light and oxygen, which are necessary for controlled photo-polymerization. To mimic a real production order, parts were fabricated in 9 batches, one batch for each printer-cassette combination, and were spread across the build area in 10 unique locations.

Utilizing the automated metrology, we investigated geometric deviations as a function of printer, cassette, and location in a build arrangement. Statistical methods such as Analysis of Variance (ANOVA) indicate that different printers have statistically significant differences in wall thicknesses and lengths. Thickness also varies as a function of location in the build, but is not affected by the cassette used. In contrast, length is significantly affected by the cassette, but does not depend on location.  

Based on this efficient, image-based metrology approach and associated statistical analysis, we demonstrate that choices in factory-level manufacturing parameters can have a statistically significant impact on the geometry of fabricated parts. While some variation appears to be random, we hypothesize that strategic manufacturing plans can improve the overall quality of a batch of parts.