Detecção de Defeitos em Impressões 3D com Visão Computacional e Aprendizado Não Supervisionado

Abstract

The quality monitoring of 3D-printed components is essential to ensure product reliability and optimize manufacturing processes. Traditional inspection methods often rely on expensive equipment, making them unfeasible for many industries. This study explores the use of accessible and low-cost technologies for defect detection in 3D-printed parts. Images of the manufactured components were captured and analyzed using analytical algorithms, with the assistance of OpenCV for edge detection and image segmentation. The analysis identified defects such as bubbles, under-extrusion, and over-extrusion. Unsupervised learning, utilizing Kmeans clustering, enabled the classification and characterization of defects based on their visual patterns and spatial distribution. The results demonstrate the potential of integrating low-cost computer vision techniques with unsupervised learning to enhance inspection in additive manufacturing.