Aprimoramento das técnicas de simulação dos fenômenos associados ao escoamento do aço em distribuidores para lingotamento contínuo

• Ano de defesa: 2019
• Autor(a) principal: Bernardo Martins Braga
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ENG - DEPARTAMENTO DE ENGENHARIA METALÚRGICA; Programa de Pós-Graduação em Engenharia Metalúrgica, Materiais e de Minas; UFMG
• Palavras-chave em Português: Modelagem matemática; Modelagem física; Remoção de inclusões; Placa de mistura; Otimização baseada em metamodelo; Engenharia metalúrgica; Metalurgia extrativa; Fundição contínua; Aço - Metalurgia
• Link de acesso: http://hdl.handle.net/1843/30413

In the continuous casting process, the tundish performance is key for the production of high- quality steels. For this reason, techniques of physical and mathematical modeling are widely used in improvement projects of tundishes. However, the modeling techniques currently employed present significant limitations and uncertainties. The similarity relations between physical models and industrial tundishes are not rigorously deduced. The mathematical models, which predicts the removal of non-metallic inclusions, fail to treat the variability introduced by turbulence. Parameters calculated from Residence Time Distribution (RTD) curves are often used to evaluate the fluid-dynamic performance of tundishes. Nevertheless, these parameters do not provide an estimate of the fraction of inclusions removed in the tundish or of the length of intermix slab generated during ladle changes in which the steel flow rate and the bath level in the tundish are variable. In addition, it is not clear what parameter of the RTD curve should be primarily considered during a comparison of different tundish configurations. A few efforts were also performed in the use of approximated optimization techniques capable of estimating the optimum point of the system with a feasible calculation time. The present work faced these shortcomings through the development of seven tasks: 1) rigorous deduction of the similarity relations between physical models and industrial tundishes; 2) development of a mathematical model capable of predicting the mean and the standard deviation of the mass fraction of non-metallic inclusions within the tundish; 3) elaboration of new compartment models to describe results of RTD curves; 4) deduction of an analytical model for estimating the fraction of non-metallic inclusions removed in the tundish that is based in the RTD curve; 5) application of an analytical model, which relies on RTD data, for predicting the amount of intermix slab generated in the tundish; 6) recommendation of procedures for evaluating the fluid-dynamic performance of tundishes by means of RTD curves; 7) assessment of a tundish optimization approach based on metamodels of a CFD model. The software programs ANSYS CFX and IChrome Nexus were employed in the calculations. The results obtained were compared with physical modeling data available in the literature. For each of these seven tasks, it was concluded that: 1) factors, such as whether the fluid flow in the tundish is steady or unsteady, the temperature of the working fluid, and the relative importance between buoyancy effects and fluid acceleration effects on particles, affect significantly the similarity criteria that should be used in physical models; 2) turbulence generates regions with high variability of inclusion mass fraction in the tundish, which should be avoided during a sample collection; 3) the proposed compartment models enable both a detailed analysis of RTD curves with one or two peaks and a simplified analysis of the global performance of the tundish; 4) in the tested conditions, the RTD-based method for estimating mean fraction of inclusions removed in the tundish presented an excellent agreement with the CFD model while , where is a dimensionless measure of the particle size; 5) for the tested condition, wherein the flow rate and the bath level were variable, the accuracy of the Niemi-Zenger method was similar to that of multiphase CFD models for ladle changes; 6) when several configurations of one-strand tundishes (or of symmetrical two-strand tundishes) are studied in a general manner, it is recommended the use of the hold-back parameter; 7) due to the highly complex shape of the response surface of the tundish CFD model, metamodels often fail to predict accurately it.