Monitoramento do desgaste de ferramentas no fresamento de topo através dos sinais de potência e emissão acústica e redes neurais

• Autor(a) principal: Silva, Rodrigo Henriques Lopes
• Data de Publicação: 2010
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/14840
• topic: Fresamento de topo; Desgaste; Monitoramento do desgaste; Emissão acústica; Potência; Redes neurais artificiais; Metal duro; Fresagem (Trabalhos em metal); Ferramentas para cortar metais; End miling; Wear; Tool condition monitoring; Acoustic emission; Cutting power; Neural networks; Carbide tools; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA

This study aims to monitor the cutting tools wear, during an end milling of VP80 steel and moreover use artificial neural networks (ANN) as a tool to predict the cutting condition. Therefore, the acoustic emission (AE) and the cutting power, were chosen as the techniques. Statistical parameters were extracted from the signals and compared with the maximum flank wear (VBBmáx). During the study, Sensis equipments were used to purchase the raw signal AE, and a sensor based on the Hall effect to the acquisition of the current, which will be posteriorly converted in effective cutting power. An end mill with R390-032A32-11M (Sandvik Coromant) specification and 32 mm in diameter was used. The inserts were coated with TiN and had R390-11 T3 10M-PH 1025 (Sandvik Coromant). During the life tests (40 in total) were acquired signals and carried out the wear measurement. The intention was try to relate them after tests conclusion and statistical parameters extraction of the signals. These parameters extracted was analyzed and those with had correlation with flank wear values were used in the training and validation of an ANN. The results show that skewness of the AE signal (frequency band 50-500 kHz), the RMS level extracted from power spectra of AE (frequency band 100-230 kHz) and the effective cutting power are correlated with wear. Furthermore, the use of these parameters as input values in an ANN, provides excellent feedback to the network while trying to predict have or not, working condition.