Modelagem de curvas de fluxo plástico de um aço bifásico utilizando inteligência artificial

• Ano de defesa: 2022
• Autor(a) principal: Contini Junior, Leones
• Orientador(a): Balancin, Oscar
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de São Carlos; Câmpus São Carlos
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Mecânica - PPGEMec
• Palavras-chave em Português: Conformação; Rede neural; Modelo
• Palavras-chave em Inglês: Forming; Neural network; Model; Fuzzy
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; ENGENHARIAS::ENGENHARIA DE MATERIAIS E METALURGICA::METALURGIA DE TRANSFORMACAO; ENGENHARIAS::ENGENHARIA MECANICA::PROCESSOS DE FABRICACAO; ENGENHARIAS::ENGENHARIA MECANICA
• Link de acesso: https://repositorio.ufscar.br/handle/20.500.14289/15653

The data used for the development of this work were obtained from previous research, in which samples of a super duplex stainless steel with ferritic matrix and dispersed austenite particles were deformed with torsion tests at temperatures from 900°C to 1200°C with strain rates ranging from 0.01 s-1 to 10 s-1. The results of these experiments were presented in the form of plastic flow curves with constant temperature and strain rate. The shapes of the curves depend on temperature and strain rate and vary with the volumetric fraction of austenite, since in two-phase materials the plastic flow is more complex than in single-phase materials. Data from this trial were used to build a spreadsheet with four columns. The first three columns contain the input attributes (temperature, strain rate and strain) and the fourth the resistance imposed by the material when deformed (stress). These data were submitted to two machine learning algorithms, one consisting of a shallow neural network (ANN) and the other of a neural network with an expert system (ANFIS). After the machine learning process, the plastic flow curves were reconstructed and compared with those obtained experimentally. Curves were predicted under conditions not measured experimentally and the results obtained are discussed. The ability of both algorithms to reconstruct the plastic flow curves of super duplex stainless steel was associated with changes in the shape of the flow curves and microstructure evolution.