Heitor Gomes de Souza Batista

Abstract

The project described in this report aims to use Regression Analysis to produce adequate and consistent predictions in regards to the dimensions of a turbulent flow over a backwards step, based on its Reynolds Number. The analysis is based on computational simulations data, obtained from the Finite Element Method with spatial and temporal discretization. The Regression is then validated in its precision, such that it can be determined the viability of this approach. Python libraries were utilized for all steps of this project. Particularly, Gmsh and FeniCS were the base for the construction of the Navier-Stokes equations and proper meshing of the problem’s domain, while Skleanr was chosen for the Regression Analysis. Throughout this study, multiple hypotheses were tested as to the relationship between the magnitude of the fluid dynamic phenomena and the Reynolds Number. For each hypothesis, a new regression model is developed, which is then analyzed in its accuracy. From the tested models, it is concluded that the vortices’ magnitude inside the pipe follows a polynomial function of order 3, based on the Reynolds Number. The final accuracy for the regression was approximately 82.96%, based on the Coe!cient of Determination R2

Advising

Advisors: Anjos, G. R.