Bulletin of the American Physical Society
65th Annual Meeting of the APS Division of Fluid Dynamics
Volume 57, Number 17
Sunday–Tuesday, November 18–20, 2012; San Diego, California
Session D5: Computational Fluid Dynamics II |
Hide Abstracts |
Chair: Elias Balaras, George Washington University Room: 24A |
Sunday, November 18, 2012 2:15PM - 2:28PM |
D5.00001: Tweed Relaxation: a new multigrid smoother for stretched structured grids Thomas Bewley, Alireza Mashayekhi In DNS/LES of the NSE using a fractional step method, one must accurately solve a Poisson equation for the pressure update at each timestep. This step often represents a significant fraction of the overall computational burden and, when Fourier methods are unavailable, geometric multigrid methods are a preferred choice. When working on an unstretched Cartesian grid, the red-black Gauss-Seidel method is the most efficient multigrid smoother available. When working on a Cartesian grid that is stretched in 1 coordinate direction to provide grid clustering near a wall, zebra relaxation, on sets of lines perpendicular to the wall, is most efficient. When working on a structured grid that is stretched in 2 or 3 coordinate directions, however, one is forced to alternate the directions that the zebra relaxation is applied in order to pass information quickly across all regions of grid clustering. A new relaxation method is introduced which is shown to significantly outperform such alternating direction line smoothers. This new method is implicit along sets of lines that branch and form 90${}^\circ$ corners, like the stripes at the shoulder of a tweed shirt, to stay everywhere perpendicular to the nearest wall, thus passing information quickly across all regions of grid clustering. [Preview Abstract] |
Sunday, November 18, 2012 2:28PM - 2:41PM |
D5.00002: A new expicit projection method for incompressible flows Sangro Park, Changhoon Lee When solving unsteady incompressible flows, the divergence-free condition should be satisfied. For this, the non-linear terms in the Navier-Stokes equation should be projected onto divergence-free space by an operator which arises from taking divergence of the Navier-Stokes equation. The calculation of projecting non-linear fields requires a lot of computational cost because the projection typically relies on an iterative solution of pressure. In this study, we propose an explicit projection method based on the spectral solution of the Poisson equation in the infinite domain and local truncation in the physical space, which does not require iterations. For validations of our methods, we applied the proposed method to the 2-dimensional Taylor-Green vortex simulation and forced isotropic turbulence simulation. The test results show that our method saved computational cost enormously while maintaining reasonable accuracy of flow field. More details about the suggested method and the performance of the method will be discussed in the meeting. [Preview Abstract] |
Sunday, November 18, 2012 2:41PM - 2:54PM |
D5.00003: Tetrahedralization of Isosurfaces with Guaranteed-Quality by Edge Rearrangement (TIGER) Shawn Walker We present a method for generating tetrahedral meshes of solids whose boundary is a smooth surface. The method uses a background grid (body-centered-cubic (BCC) lattice) from which to build the final conforming 3-D mesh. The algorithm is fast, robust, and provides useful guaranteed dihedral angle bounds for the output tetrahedra. The dihedral angles are bounded between 8.5 and 164.2 degrees. If the lattice spacing is smaller than the ``local feature size,'' then the dihedral angles are between 11.4 and 157.6 degrees (c.f. Labelle, Shewchuk 2007). The method is simple to implement and performs \emph{no} extra refinement of the background grid. The most complicated mesh transformations are 4-4 edge flips. Moreover, the only parameter in the method is the BCC lattice spacing. Applications of the method range from free boundary flows, to modeling deformations, shape optimization, and to anything that requires dynamic meshing such as virtual surgery. A MATLAB demonstration will be given to show case the method. [Preview Abstract] |
Sunday, November 18, 2012 2:54PM - 3:07PM |
D5.00004: On a robust ALE method with the discrete primary and secondary conservation properties Seongwon Kang, Nahmkeon Hur The objective of the present study is to construct a robust, implicit discretization method for the arbitrary Lagrangian-Eulerian (ALE) method for deforming grids. In order to minimize the effect of an artificial diffusion, we present a novel implicit method derived using the secondary conservation property enforced in both spatial and temporal discretization. When applied to the Navier-Stokes equation, the proposed method satisfies conservation of the discrete mass, momentum and kinetic energy in both incompressible and compressible flows. We compared the different choices for discretization in the ALE method by an analysis of the truncation errors. With the numerical tests using the cases with high Reynolds numbers, an improved stability was observed using the revised discretization method compared to the existing methods. [Preview Abstract] |
Sunday, November 18, 2012 3:07PM - 3:20PM |
D5.00005: Wavelet-based adaptive numerical simulation of unsteady 3D flow around a bluff body Giuliano De Stefano, Oleg Vasilyev The unsteady three-dimensional flow past a two-dimensional bluff body is numerically simulated using a wavelet-based method. The body is modeled by exploiting the Brinkman volume-penalization method, which results in modifying the governing equations with the addition of an appropriate forcing term inside the spatial region occupied by the obstacle. The volume-penalized incompressible Navier-Stokes equations are numerically solved by means of the adaptive wavelet collocation method, where the non-uniform spatial grid is dynamically adapted to the flow evolution. The combined approach is successfully applied to the simulation of vortex shedding flow behind a stationary prism with square cross-section. The computation is conducted at transitional Reynolds numbers, where fundamental unstable three-dimensional vortical structures exist, by well-predicting the unsteady forces arising from fluid-structure interaction. [Preview Abstract] |
Sunday, November 18, 2012 3:20PM - 3:33PM |
D5.00006: Adaptive Wavelet Collocation Method in Shallow Water Model: Validation Study Shanon Reckinger, Oleg Vasilyev, Baylor Fox-Kemper The adaptive wavelet collocation and Brinkman penalization methods are applied to the shallow water model and validated. The wavelet method solves the equations on temporally and spatially varying meshes, which allows a higher effective resolution to be obtained with less computational cost. The grid adaptation is achieved by using the ability of wavelet multiresolution analysis to identify and isolate localized dynamically dominant flow structures, e.g., vortices, and to track these structures on adaptive computational meshes. In addition to studying how the shallow water model behaves on non-uniform, time varying grids, this work also sets out to improve the representation of continental topology through an extension of the Brinkman penalization method. This numerical technique works by altering the governing equations in such a way that no slip boundary conditions are enforced. When coupled with the adaptive wavelet collocation method, the flow near a complex boundary can be well defined. In previous work, the methods were presented, fully derived, and convergence was demonstrated. In this work, a variety of benchmark studies will be presented to validate the model and insight will be given on possible directions for wavelets in ocean modeling. [Preview Abstract] |
Sunday, November 18, 2012 3:33PM - 3:46PM |
D5.00007: On the POD based reduced order modeling of high Reynolds flows Fariduddin Behzad, Brian Helenbrook, Goodarz Ahmadi Reduced-order modeling (ROM) of a high Reynolds fluid flow using the proper orthogonal decomposition (POD) was studied. Particular attention was given to incompressible, unsteady flow over a two-dimensional NACA0015 airfoil. The Reynolds number is $10^5$ and the angle of attacked of the airfoil is $12^\circ$. For DNS solution, hp-finite element method is employed to drive flow samples from which the POD modes are extracted. Particular attention is paid on two issues. First, the stability of POD-ROM resimulation of the turbulent flow is studied. High Reynolds flow contains a lot of fluctuating modes. So, to reach a certain amount of error, more POD modes are needed and the effect of truncation of POD modes is more important. Second, the role of convergence rate on the results of POD. Due to complexity of the flow, convergence of the governing equations is more difficult and the influences of weak convergence appear in the results of POD-ROM. For each issue, the capability of the POD-ROM is assessed in terms of predictions quality of times upon which the POD model was derived. The results are compared with DNS solution and the accuracy and efficiency of different cases are evaluated. [Preview Abstract] |
Sunday, November 18, 2012 3:46PM - 3:59PM |
D5.00008: Projection of Discontinuous Galerkin Variable Distributions During Adaptive Mesh Refinement Carlos Ballesteros, Marcus Herrmann Adaptive mesh refinement (AMR) methods decrease the computational expense of CFD simulations by increasing the density of solution cells only in areas of the computational domain that are of interest in that particular simulation. In particular, unstructured Cartesian AMR has several advantages over other AMR approaches, as it does not require the creation of numerous guard-cell blocks, neighboring cell lookups become straightforward, and the hexahedral nature of the mesh cells greatly simplifies the refinement and coarsening operations. The \emph{h}-refinement from this AMR approach can be leveraged by making use of highly-accurate, but computationally costly methods, such as the Discontinuous Galerkin (DG) numerical method. DG methods are capable of high orders of accuracy while retaining stencil locality---a property critical to AMR using unstructured meshes. However, the use of DG methods with AMR requires the use of special flux and projection operators during refinement and coarsening operations in order to retain the high order of accuracy. The flux and projection operators needed for refinement and coarsening of unstructured Cartesian adaptive meshes using Legendre polynomial test functions will be discussed, and their performance will be shown using standard test cases. [Preview Abstract] |
Sunday, November 18, 2012 3:59PM - 4:12PM |
D5.00009: A Spectral Adaptive Mesh Refinement Method for Homogenous Isotropic Turbulence Leila Nasr Azadani, Anne Staples We present an algorithm for accelerating simulations of homogenous isotropic turbulence. The method is akin to an adaptive mesh refinement (AMR) technique, applied in Fourier space. In direct numerical simulations of turbulence (DNS) the mesh size or the number of Fourier modes is defined based on the ratio of the sizes of the largest to smallest eddies that can be formed during the computation. The range of spatial scales in evolving turbulent flows changes with time. Early in a computation there may exists only large eddies and a coarse mesh will be enough to capture all the details of the flow, while at another time smaller eddies may form and a finer mesh will be required to resolve all scales. Therefore, instead of performing DNS with a constant fine mesh, AMR techniques can be applied and the mesh size can be varied during the computation in order to have optimum mesh sizes and save computational time. The spectral AMR method we present here is applied to 2D and 3D homogenous isotropic turbulence and results are compared with the DNS performed using a fine mesh. [Preview Abstract] |
Sunday, November 18, 2012 4:12PM - 4:25PM |
D5.00010: ABSTRACT WITHDRAWN |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700