Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
Stereo-PIV Velocity Measurements of Turbulent Flows in a Square Duct
Submission Author:
Leonardo Soares Fernandes , RJ , Brazil
Co-Authors:
Leonardo Soares Fernandes, Luis Fernando Alzuguir Azevedo
Presenter: Leonardo Soares Fernandes
doi://10.26678/ABCM.COBEM2021.COB2021-0257
Abstract
Turbulent flows in ducts with a square cross section are found in many engineering applications and environmental applications, such as ventilation systems and combustion engines. Moreover, its relatively simple geometry is useful to study the effect of the less intense secondary flows on the overall flow structure, and other quantities of interest, such as pressure drop and wall shear stresses. This work details an experiment designed to simultaneously measure pressure drop and three-component velocity vector fields in a water channel with a square cross-section of 40x40 mm2, using the Stereoscopic Particle Image Velocimetry (SPIV) technique. A total of 10 different cases were measured, with the Reynolds number based on the hydraulic diameter of the test section ranging from 7000 to 45000. For each case, the number of independent samples was increased until the averaged velocity fields and the Reynolds stress tensor quantities were well converged. The results obtained clearly show the secondary motions at the duct corners, which influence the mean streamwise velocity field. This behavior was already reported in the literature, but detailed data was based on direct numerical simulations (DNS). The mean velocity fields and the Reynolds stress tensor results obtained can be used both to serve as a base case for studies with drag-reducing agents as well as to improve the performance of turbulence models. Recent studies have shown that Machine Learning techniques can be used to improve the prediction capabilities of Reynolds Averaged Navier-Stokes (RANS) models. These studies usually make use of DNS databases both during the training and the validation of the algorithms, what limit the maximum Reynolds number available. The use of well-converged experimental databases such as the one obtained in this work is a valuable information to such machine learning initiatives. Both studies with drag reducing agencies and the use of such data to improve RANS models with machine learning techniques are the next phase of this work.
Keywords
Turbulence, Square Duct, Stereo-PIV
