Implementação dos efeitos da não linearidade do tensor de Reynolds em modelos de turbulência baseados na hipótese de Boussinesq / Implementation of non linear effects of Reynolds tensor on turbulence models based on Boussinesq approximation

AUTOR(ES)
DATA DE PUBLICAÇÃO

2009

RESUMO

The main goal of this work is to implement and evaluate the correction proposed by Spalart [53] on turbulence models based on Boussinesq approximation. The objective of this correction is to capture non linear effects of Reynolds tensor, which are not correctly predicted and modeled by Boussinesq approximation. In his work, Spalart [53] proposed the correction by introducing a non linear to Reynolds tensor obtained through the velocity gradients correlation. The implementation has been done in commercial package Fluent, from ANSYS Inc., through a User-Defined Function (UDF), which is executed with the main program. The modification has been tested for three turbulence models: Spalart-Allmaras, k−" and k − ! SST, which had their results compared to experimental data and Reynolds Stress Model (RSM) results. This last model does not use Boussinesq approximation and has turbulent tensor components modeled by an evolutive equation. Three test cases have been selected in order to explore experimental results and compare them to those obtained with turbulence models in default and modified formulations, proposed by Spalart [53]. The first test case, proposed by Melling [41], represented by the flow inside a square duct, has the development of boundary layer and secondary flows at duct corner as main characteristics. At the second test case, the flow inside a rectangular curved duct, proposed by Kim and Patel [29], has been simulated. At this case, the secondary flow is also presented and, moreover, specific characteristics of turbulence kinetic energy production and dissipation at inner (convex) and outer (concave) wall have been evaluated. At the third test case, the flow around Ahmed body [1] has been simulated and the results obtained by Becker et at [17] have been used. On this case, the boundary layer detachment and the wake behind the body are the characteristics evaluated. For the first test case, longitudinal velocity component profiles have been obtained, besides transversal turbulent kinetic energy and the secondary flow represented by transversal velocity vector components. These results presented good correlation between numerical and experimental data. The modified turbulence models based on Spalart [53] modification have been able to represent the secondary flow, which has not been possible with turbulence models in their default formulation. For the second test case, pressure coefficient, transversal velocity and turbulent kinetic energy results have been obtained. As described by Kim and Patel [29], the secondary flow inside the bend has been captured by all of turbulence models evaluated on this work. For the third test case, the results of pressure coefficient, transversal velocity and turbulent kinetic energy have been obtained and presented good agreement with experimental data. Moreover, the secondary flow has been represented by transversal component of velocity vectors. Drag coefficient has also been evaluated and presented differences of 2.1% and 8.1% varying between turbulence models simulated. Finally, for all test cases, the computational time has been evaluated, comparing turbulence models in their default formulation with the implementation done on this work.

ASSUNTO(S)

dinâmica dos fluidos computacional engenharia mecanica turbulência não-linearidades métodos numéricos

ACESSO AO ARTIGO

Documentos Relacionados