Medição de desgaste de inserto de usinagem in loco utilizando um sistema de visão computacional

• Ano de defesa: 2024
• Autor(a) principal: David Lelis Filho
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA MECÂNICA; Programa de Pós-Graduação em Engenharia Mecanica; UFMG
• Palavras-chave em Português: Medição de desgaste; Visão computacional; Usinagem; Medição in loco; Baixo custo; Sistema embarcado; Engenharia mecânica; Engenharia econômica; Custo-benefício; Medição; Desgaste mecânico; Visão por computador; Sistemas embutidos de computador
• Link de acesso: http://hdl.handle.net/1843/75988

Machining plays an important role in manufacturing, involving the removal of material to give the desired shape, dimensions and finish to products. An obstacle to the increase of productivity in machining is the downtime generated by planned and unplanned interruptions, particularly the replacement of worn or prematurely failed cutting tools, which can represent a considerable portion of the total production time. To reduce these effects, computer vision techniques have been used to monitor the evolution of tool wear. However, traditionally, the application of these techniques are carried out with the machine stopped, which affects production time. The present work proposes the construction of a low-cost computer vision system to measure the flank wear of milling inserts in loco and compare two different lighting methods, one that uses only one image of the insert and the other that uses a combination of three images. All procedures involving from image acquisition to wear determination are performed using a dedicated low-cost system (Raspberry Pi and camera module v2). The firmware was developed using open source programming language and libraries (Python). The system was validated through bench tests and in a machining center. Tool wear measurement were carried out on different types of inserts and compared the results of the proposed system with measurements carried out using an optical microscope, through the normalized error test. The results showed that the method that uses the combination of three images of the insert is the most advantageous, with a success rate of 100% for average flank wear and 92% for maximum flank wear in the tests conducted in the machining center.