Classificação de defeitos a partir de imagens de superfícies fabricadas por manufatura aditiva robotizada

Abstract

With the rapid evolution of industries in recent years, new, more flexible, and cost-effective production methods are required to meet growing demands. Among the fastest-growing methods, additive manufacturing, also known as 3D printing, stands out. This process offers significant advantages for the industry, such as a more flexible design, the ability to produce parts with complex geometries at low costs, and strong consistency with the original model. Within additive manufacturing, there is a specific process where an extruder is mounted on a robotic manipulator, allowing for even greater flexibility during production. However, this flexibility, while advantageous, also increases the number of process variables and, consequently, the potential for errors. In this context, the present work aims to develop an algorithm capable of classifying surface defects that occur on a part during its production using robotic additive manufacturing, as a means of enabling effective monitoring. This classification was performed by a machine learning algorithm through the analysis of images captured by a camera connected to a Raspberry Pi mini-computer. The goal was to achieve an algorithm capable of accurately classifying defects that arise during the robotic additive manufacturing process. Through these analyses, preventive measures can be taken to reduce the frequency of such errors or even eliminate them, as well as to implement a correction system during the manufacturing process.