Modelagem utilizando redes neurais artificiais para predição da percentagem de ferrita e parâmetros geométricos de cordões de solda de aços inoxidáveis austeníticos

• Ano de defesa: 2013
• Autor(a) principal: Marina Spyer Las-casas
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Simulação computacional; Redes neurais; Aços inoxidáveis austeníticos; Soldagem; Ferrita; Processos de fabricação Modelos matemáticos; Aço inoxidável austenítico Soldagem; Engenharia mecânica
• Link de acesso: http://hdl.handle.net/1843/BUOS-9ABGGH

A model was developed based on experimental data obtained under laboratory conditions. To acquire these data we used an industrial robot that made welds with GMAW (Gas Metal Arc Welding) process. Welds were made with different values of voltage, current and filler material while all others parameters were kept constant. The following austenitic stainless steel wires were used: ER 308LSi ER, ER 309LSi and ER 312. All weld beads were performed on AISI 304 plates. The input parameters of the network are Vweld (input parameter that determines the robot welding voltage), Aweld (input parameter that determines the robot welding current) and values of Nickel and Chromium equivalent of wires calculated using the Schaeffler formula. The quantity of ferrite was analyzed by magnetic methods calibrated according to the AWS standard procedure and therefore will be adopted the term "Ferrite Number" (FN) in place of percent ferrite to identify this variable. In addition to FN, the model predicts the width, reinforcement and penetration of the weld beads.