Ricardo Serfaty

Mathematical and Numerical Modeling of Turbulent Flows

The present work is devoted to the development and implementation of a computational framework to perform numerical simulations of low Mach number turbulent flows over complex geometries. The algorithm under consideration is based on a classical predictor-corrector time integration scheme that employs a projection method for the momentum equations. The domain decomposition strategy is adopted for distributed computing, displaying very satisfactory levels of speed-up and efficiency. The Immersed Boundary Methodology is used to characterize the presence of a complex geometry. Such method demands two separate grids: An Eulerian, where the transport equations are solved with a Finite Volume, second order discretization and a Lagrangian domain, represented by a non-structured shell grid representing the immersed geometry. The in-house code developed was fully verified by the Method of Manufactured Solutions, in both Eulerian and Lagrangian domains. The capabilities of the resulting computational framework are illustrated on four distinct cases: a turbulent jet, the Poiseuille flow, as a matter of validation of the implemented Immersed Boundary methodology, the flow over a sphere covering a wide range of Reynolds numbers, and finally, with the intention of demonstrating the applicability of Large Eddy Simulations - LES - in an industrial problem, the turbulent flow inside an industrial fan.

A novel immersed boundary/Fourier pseudospectral method for flows with thermal effects

ABSTRACT A novel immersed boundary method (IBM) for flows with thermal effects is proposed, combining high accuracy and low computational cost, provided by the Fourier pseudospectral method (FPSM), for the possibility of handling complex and nonperiodical geometries using the IBM. With focus on incompressible flow problems modeled by Navier-Stokes, mass, and energy equations, the method of manufactured solutions is used for the numerical verification of Dirichlet boundary conditions imposed via the IBM. Then, the proposed method is applied on two different 2-D cases: (1) energy transfer due to natural convection in a square cavity, and (2) an annulus between horizontal concentric cylinders nonuniformly heated. Good agreement with available data in the literature has been achieved.

Fourier pseudospectral method for nonperiodical problems: A general immersed boundary method for three types of thermal boundary conditions

ABSTRACT In the present paper a general scheme for the three types of thermal boundary conditions is proposed and applied to natural convection and diffusion problems. The numerical algorithm, denominated thermal IMERSPEC, consists of the application of the Fourier pseudospectral method, where Dirichlet, Neumann, or Robin boundary conditions are modeled through immersed boundary method (IBM). The methodology is to impose the boundary conditions on the interface and transmit through distribution functions. Source terms are added to the two-dimensional Navier-Stokes and energy equations on the Cartesian mesh. Manufactured solutions are used for the numerical verification of Dirichlet, Neumann, and Robin boundary conditions, imposed through IBM. The proposed method is applied for solving problems involving thermal energy transfer for natural convection in an annulus between horizontal concentric cylinders. Results for these applications using the thermal IMERSPEC and the traditional finite volume method are compared and a good agreement is obtained for both methodologies.

Numerical Simulation of Natural Gas Combustion Process Using a One-Step and a Two-Step Reaction

The work evaluates the combustion of natural gas in a cylindrical furnace. The Generalized Finite Rate Reaction Model was selected for predicting the reactions. Two situations were considered. In the first case the combustion of the fuel was predicted by a single global reaction, and in the second case a two-step reaction was considered for predicting the combustion process. The conservation equations of mass, momentum, energy and chemical species were solved by the finite volume procedure, with the commercial software FLUENT. The turbulent flow was modeled by employing the two differential equation κ–e model. The solutions obtained with the two reaction models, for the temperature and species concentration fields, were compared among them and against experimental data available in the literature. It was observed that the two-step reaction model represents better the physical phenomena, showing a better agreement with the experimental data.Copyright