Caracterização do escoamento bifásico em golfadas utilizando redes neurais artificiais

• Ano de defesa: 2016
• Autor(a) principal: Cozin, Cristiane
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Curitiba; Brasil; Programa de Pós-Graduação em Engenharia Elétrica e Informática Industrial; UTFPR
• Palavras-chave em Português: Escoamento bifásico; Gás - Escoamento; Métodos de simulação; Redes neurais (Computação); Engenharia elétrica; Two-phase flow; Simulation methods; Neural networks (Computer science); Electric engineering; CNPQ::ENGENHARIAS::ENGENHARIA MECANICA::FENOMENOS DE TRANSPORTE::MECANICA DOS FLUIDOS; Engenharia Elétrica
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/2654

Gas-liquid two-phase flows are present in nature and in different industrial activities. In this type of flow, the liquid and gas phases assume different spatial configurations inside the pipe, called flow patterns. Slug flow is one of the most frequent flow patterns in industrial applications, occurring over a wide range of phase velocities according to studies presented by several authors. The mathematical modelling of slug flow comprises from simple steady state models to more complex models for transient regimes. Those models require closure relationships, e.g. empirical correlations and statistical distributions of characteristic flow parameters. In this work, several models based on artificial neural networks are presented as a support to the characterization of the two-phase slug flow parameters that depend on experimentally obtained void fraction time series. The void fraction time series are measured with a pair of wiremesh sensors installed in a test section of an experimental rig in the premises of the NUEM/UTFPR labs and described in Castillo (2015). From the time series of void fraction measurements relevant parameters to the flow under consideration are computed: the length of the elongated gas bubble, the liquid slug length, the translational velocity of the elongated bubble and the standard deviations for those variables. Those measured and calculated variables are used to obtain a set of artificial neural network-based models. After obtaining such models, a simulation study in which those models are used to estimate the parameters that characterize the two-phase slug flows is carried out. Detailed analysis of the results showed that the variables related to the gas phase are estimated with greater accuracy than the ones related to the liquid phase. As an immediate application of the obtained model, its use as a tool to calculate the initial conditions for a phenomenological mathematical model of twophase slug flow with a slight change of inclination based on a slug tracking method is presented. The differential of this study is to predict the intermittent features of the twophase slug flow by means of a neural model, as well as the estimation of average parameters for the variables of interest with uncertainly rates ranging between 10% and 16%.