Bulletin of the American Physical Society
63rd Annual Meeting of the APS Division of Fluid Dynamics
Volume 55, Number 16
Sunday–Tuesday, November 21–23, 2010; Long Beach, California
Session CD: CFD: Algorithms II |
Hide Abstracts |
Chair: Jean-Christophe Nave, Massachusetts Institute of Technology Room: Long Beach Convention Center 102B |
Sunday, November 21, 2010 1:00PM - 1:13PM |
CD.00001: An Adaptive Mesh Refined Gradient-Augmented Level Set Method Jean-Christophe Nave, Benjamin Seibold, Ruben Rosales The Gradient-Augmented Level Set method (GA-LS) was introduced at the 62$^{nd}$ annual APS-DFD meeting by Nave et al. (arXiv:0905.3409). Leveraging the optimal locality and unconditional stability of the method, we present a natural extension to adaptive quad-tree meshes. The new method possesses many desirable features such as improved mass conservation, reduced computational effort, and is, due to the optimal locality property of the underlying GA-LS, very easy to implement. Several key benchmark tests will be presented to help demonstrate the benefits of the approach, and the overall simplicity of the algorithm. [Preview Abstract] |
Sunday, November 21, 2010 1:13PM - 1:26PM |
CD.00002: Adaptive mesh refinement for large-eddy simulation using the dynamic reconstruction model Lauren Goodfriend, Fotini Chow, Marcos Vanella, Elias Balaras Combining large-eddy simulation (LES) with adaptive mesh refinement (AMR) to reduce computation costs yields a powerful technique for modeling flows with complex geometries. Increased errors associated with the variable mesh sizes used in AMR have limited computations combining these methods. Using explicit filtering to separate turbulent stresses into resolvable sub-filter scale (RSFS) and sub-grid scale (SGS) terms may help control these errors. In this study, the dynamic reconstruction model (DRM) is used to approximate the RSFS stress using a series expansion to invert the explicit filter operation, with a dynamic eddy viscosity model for the SGS stress. DRM has previously been shown to reduce numerical truncation errors, leading to a more accurate turbulence closure. Here, the effect of using discontinuous versus continuously varying filter width across a grid refinement interface is investigated in the context of DRM. Decaying isotropic turbulence is advected past a refinement interface in which the grid is coarsened. Results are compared using explicit filtering with different levels of reconstruction and without using explicit filtering. [Preview Abstract] |
Sunday, November 21, 2010 1:26PM - 1:39PM |
CD.00003: Characteristic Boundary Conditions with Transverse Effects: a Comparative Study Guido Lodato, Heinz Pitsch The inclusion of transverse effects in designing characteristic boundary conditions for the Euler and compressible Navier-Stokes equations was discussed by many authors and proved to give significant improvement in reducing numerical perturbations from open boundaries. Starting from the characteristic formulation of the Euler equations using generalized coordinates, an analysis of the different transverse terms is carried out. Based on the identification of different types of transverse effects, it is suggested that, in order to achieve the best performance for the numerical behavior of the boundary conditions, different transverse terms should be treated differently. The analytical check on well-posedness and reflection coefficients gives evidence that stability might not be an issue regardless of the way these terms are treated, but that numerical reflection from the boundary might be negatively affected by the occurrence of a direct coupling between outgoing vorticity modes and incoming acoustic modes. This analysis is supported by numerical tests on the inviscid convected vortex problem at different Mach numbers. [Preview Abstract] |
Sunday, November 21, 2010 1:39PM - 1:52PM |
CD.00004: Multi-core/GPU accelerated multi-resolution simulations of compressible flows Babak Hejazialhosseini, Diego Rossinelli, Petros Koumoutsakos We develop a multi-resolution solver for single and multi-phase compressible flow simulations by coupling average interpolating wavelets and local time stepping schemes with high order finite volume schemes. Wavelets allow for high compression rates and explicit control over the error in adaptive representation of the flow field, but their efficient parallel implementation is hindered by the use of traditional data parallel models. In this work we demonstrate that this methodology can be implemented so that it can benefit from the processing power of emerging hybrid multicore and multi-GPU architectures. This is achieved by exploiting task-based parallelism paradigm and the concept of wavelet blocks combined with OpenCL and Intel Threading Building Blocks. The solver is able to handle high resolution jumps and benefits from adaptive time integration using local time stepping schemes as implemented on heterogeneous multi-core/GPU architectures. We demonstrate the accuracy of our method and the performance of our solver on different architectures for 2D simulations of shock-bubble interaction and Richtmeyer-Meshkov instability. [Preview Abstract] |
Sunday, November 21, 2010 1:52PM - 2:05PM |
CD.00005: Stability and Accuracy of Coupling Strategies in Hybrid LES/PDF Algorithms for Turbulent Reactive Flows Pavel Popov, Haifeng Wang, Stephen Pope Hybrid Large Eddy Simulation/Probability Density Function (LES/PDF) algorithms for turbulent reactive flow consist of two main components: a grid-based LES solver, and a Monte Carlo solver which approximates the composition PDF via an ensemble of Lagrangian particles. As a part of the interplay between these two codes, the PDF solver communicates to the LES solver the information which it needs in order to evaluate the filtered density field. Here, we assess two alternative strategies for implementing this coupling: either by directly passing density information from the PDF to the LES solver, or by evaluating density from an enthalpy field which is solved for by the LES code, and relaxed towards a consistent (up to numerical errors) PDF-based enthalpy field. We compare the stability of these two approaches, and their accuracy, defined as the level of consistency between a standalone LES and a coupled LES/PDF solution for a laboratory-scale jet flame, with flamelet chemistry modeling. The benefits of second-order time accuracy, relative to a first-order-accurate in time implementation, are examined, and different approaches for performing filtering of the PDF fields, which contain significant noise due to their stochastic nature, are evaluated. [Preview Abstract] |
Sunday, November 21, 2010 2:05PM - 2:18PM |
CD.00006: Theoretically Simple Thermal Lattice Boltzmann Method and Its Application to a 2-D Shock Tube Simulation Jae Wan Shim We propose a new approach to derive the thermal lattice Boltzmann method (LBM), and show its accuracy and stability by a 2-D shock tube simulation. The derivation is as simple as that of the isothermal LBM obtained by using Taylor expansion. We do not use Gauss-type quadratures because the results are inapplicable in regular lattices for thermal flows. The derivation enables us to simulate thermal flows with accuracy, stability, and use of regular lattices, simultaneously. [Preview Abstract] |
Sunday, November 21, 2010 2:18PM - 2:31PM |
CD.00007: Shape Optimization for Drag Reduction in Linked Bodies using Evolution Strategies and the Hybrid Wavelet Collocation - Brinkman Penalization Method Oleg V. Vasilyev, Mattia Gazzola, Petros Koumoutsakos In this talk we discuss preliminary results for the use of hybrid wavelet collocation - Brinkman penalization approach for shape optimization for drag reduction in flows past linked bodies. This optimization relies on Adaptive Wavelet Collocation Method along with the Brinkman penalization technique and the Covariance Matrix Adaptation Evolution Strategy (CMA-ES). Adaptive wavelet collocation method tackles the problem of efficiently resolving a fluid flow on a dynamically adaptive computational grid, while a level set approach is used to describe the body shape and the Brinkman volume penalization allows for an easy variation of flow geometry without requiring body-fitted meshes. We perform 2D simulations of linked bodies in order to investigate whether flat geometries are optimal for drag reduction. In order to accelerate the costly cost function evaluations we exploit the inherent parallelism of ES and we extend the CMA-ES implementation to a multi-host framework. This framework allows for an easy distribution of the cost function evaluations across several parallel architectures and it is not limited to only one computing facility. The resulting optimal shapes are geometrically consistent with the shapes that have been obtained in the pioneering wind tunnel experiments for drag reduction using Evolution Strategies by Ingo Rechenberg. [Preview Abstract] |
Sunday, November 21, 2010 2:31PM - 2:44PM |
CD.00008: Xwing: A 3D Viscous Design Tool for MAVs Don Gipe, Kamran Mohseni Characterizing the 3D effects of low Reynolds number flow over low aspect ratio wings is computationally expensive using the full Navier-Stokes Equations. Recently, the use of Micro Aerial Vehicles as a tool for both scientific and military applications has grown leading to a diverse range of applications as well as designs. This development has created the need for a less computationally expensive approach for design of Micro Aerial Vehicles (MAVs). Xwing is a design tool for rapid calculation of aerodynamic forces of MAVs. Xwing uses an invisid solver to compensate for the effect of tip vortices over a low aspect ratio wing. A potential flow -- boundary layer matching technique was used at each 2D cross section using this effective angle of attack. The calculated displacement thickness at each cross section was used to obtain a new effective airfoil at each section. This iterative process could continue until convergence is achieved. The validity of the technique was tested in several cases including tapered, twisted, and swept wings at both low and high Reynolds number flows. [Preview Abstract] |
Sunday, November 21, 2010 2:44PM - 2:57PM |
CD.00009: A Poisson-Boltzmann solver on Non-Graded Adaptive Grid with Robin boundary conditions on Irregular Domains Asdis Helgadottir, Frederic Gibou We introduce a second-order solver for the Poisson-Boltzmann equation in arbitrary geometry in two and three spatial dimensions. The Poisson-Boltzmann equation can be used to represent the electric potential of a solution and is, therefore, of great interest in micro fluidics. The method introduced differs from existing methods solving the Poisson-Boltzmann equation in the two following ways: First, non-graded Quadtree (in two spatial dimensions) and Octree (in three spatial dimensions) grid structures are used; Quadtree/Octree grid structures save a significant amount of computational power at no sacrifice in accuracy. Second, Robin boundary conditions are enforced at the irregular domain's boundary. The irregular domain is described implicitly and the grid does not need to conform to the domain's boundary, which makes grid generation straightforward and robust. The resulting matrix is an M-matrix, thus the linear system is invertible, leading to a simple and robust second- order accurate solver. [Preview Abstract] |
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