National Institute of Aerospace
Computational Fluid Dynamics Seminar

A place to share ideas and problems for barrier-breaking developments

NIA CFD Seminar, Season 3 (2013-2014)
#50:   08-04-2014, Seongim Choi
High-/Multi-Fidelity Aerodynamic Analysis and MDO with Uncertainty Quantification
#49:   07-17-2014, Nicolas R. Gauger
Efficient Techniques for Optimal Active Flow Control
#48:   06-03-2014, Debojyoti Ghosh
Compact-Reconstruction WENO Schemes - Theory, Implementation and Applications
#47:   05-13-2014, Alex Povitsky
Adjusted Formulation of Vorticity Confinement with Applications to Tip Vortices of Stationary and Rotaing Wings
#46:   05-06-2014, Andrey Chernikov
Parallel Tetrahedral Mesh Generation
#45:   04-22-2014, Peter Gnoffo
Walsh Functions in Numerical Simulation: A New Framework for Solving Nonlinear Systems of PDEs
#44:   03-04-2014, Atsushi Hashimoto
Development of Fast and Efficient CFD Tools, HexaGrid and FaSTAR
#43:   02-18-2014, Boris Diskin
Evaluation of Multigrid Solutions for Turbulent Flows
#42:   12-03-2013, Elbert Jeyapaul
Higher-Order Moments and Their Modeling Approximations in a Turbulent Channel Flow Subjected to Mean Strain
#41:   09-24-2013, Steven Diot
The MOOD Method --- Multidimensional Optimal Order Detection -- a first a posteriori approach to Very-High-Order Finite Volumes methods
#40:   09-10-2013, Robert L. Ash
Non-equilibrium Pressure Considerations in Modeling Viscous Rotating Flows
#39:   08-27-2013, Qiqi Wang
Chaotic adjoint: from massive linear algebra to a light-weight wrapper of existing solvers

08-04-2014   2:00pm - 3:00pm (EDT)   NIA Room 138      video

High-/Multi-Fidelity Aerodynamic Analysis and MDO with Uncertainty Quantification

Simulation-based design optimization method depends heavily on the specifics of aerospace vehicle system and flow characteristics at design points. This talk focuses on how different design optimization methods with aerodynamic analysis of various fidelities and sensitivity analysis are utilized for rotary and fixed wing vehicles: a helicopter rotor in forward flight and a low-boom supersonic jet. If time permits, a design application of low-noise co-axial contra-rotating rotor, i.e., open rotor, in consideration of aleatory uncertainties will be introduced as a design extension for rotor flows. First, an efficient CFD method to optimize helicopter rotor at unsteady forward flight, a time-spectral and adjoint-based optimization method, is developed. A direct application of the optimization techniques common to the steady flow problems to the unsteady flows is difficult, as the flows are complicated and design cost is very high in resolving transient, time-varying aerodynamic performances. The time-spectral method is based on the frequency-domain analysis and reduces the unsteady form of the Navier-Stokes governing equations into a periodic steady state. A powerful adjoint solution method can be effectively integrated into the unsteady design framework for helicopter rotors through the time-spectral formulation of the equations. The design framework of time-spectral and adjoint-based method can be applied to many areas of the engineering design problems including co-axial contra-rotating open rotor, wind-turbine blade and flapping wing. Next, a multi-fidelity and multidisciplinary design optimization method for supersonic business jets (SBJ) is introduced. Optimal shape and mission profile of SBJ are sought to reduce sonic boom on the ground during the cruise condition. Inviscid Euler computation with solution-adaptive mesh refinement is utilized, and gradient-free optimization is used in combination with multi-fidelity response surface model.

[ presentation file (pdf) ] Seongim Choi

Speaker Bio: Dr. Seongim Choi is currently an Assistant Professor in the dept. of Aerospace and Ocean Engineering, Virigina Tech. Prior to working in Virginia Tech, she worked as an Assistant Professor in Korea Advanced Institute of Science and Technology (KAIST) from 2010~2013. She received her Ph.D from Stanford in 2006, and B.S. and M.S. from Seoul National University. She received an AIAA MDO best paper award for her Ph.D work on the MDO of low-boom supersonic jets. She worked as a post doctorate fellow in the U.S. Aeroflightdynamics Directorate (AFDD) working on helicopter rotor flow analysis and design. She received a young investigated award from the 1st Asian Rotorcraft Forum in 2012 for her work on the time-spectral and adjoint-based design method for helicopter rotor flows.

07-17-2014   11:00am - noon   NIA Room 101      video

Efficient Techniques for Optimal Active Flow Control

For efficient optimal active control of unsteady flows, the use of adjoint approaches is a first essential ingredient. We compare continuous and discrete adjoint approaches in terms of accuracy, efficiency and robustness. For the generation of discrete adjoint solvers, we discuss the use of Automatic Differentiation (AD) and its combination with checkpointing techniques. Furthermore, we discuss so-called one-shot methods. Here, one achieves simultaneously convergence of the primal state equation, the adjoint state equation as well as the design equation. The direction and size of the one-shot optimization steps are determined by a carefully selected design space preconditioner. The one-shot method has proven to be very efficient in optimization with steady partial differential equations (PDEs). Applications of the one-shot method in the field of aerodynamic shape optimization with steady Navier-Stokes equations have shown, that the computational cost for an optimization, measured in runtime as well as iteration counts, is only 2 to 8 times the cost of a single simulation of the governing PDE. We present a framework for applying the one-shot approach also to optimal control problems with unsteady Navier-Stokes equations. Straight forward applications of the one-shot method to unsteady problems have shown, that its efficiency depends on the resolution of the physical time domain. In order to dissolve this dependency, we consider unsteady model problems and investigate an adaptive time scaling approach.

[ presentation file (pdf) ] Nicolas R. Gauger

Speaker Bio: Prof. Dr. Nicolas R. Gauger received his Master in Mathematics (Dipl.-Math.) from University of Hannover in 1998, and his Ph.D. in Applied Mathematics (Dr.rer.nat.) from TU Braunschweig in 2003. From 1998 to 2010 he was Research Scientist in the field of Numerical Methods for Aerospace Science at German Aerospace Center (DLR) in Braunschweig. Furthermore, he was appointed as Assistant Professor (Jun.-Prof.) for Applied Mathematics at the Department of Mathematics of the Humboldt University Berlin from 2005 to 2010. In Addition, he was Member of the DFG Research Center MATHEON (Mathematics for Key Technologies) in Berlin from 2006 to 2010. Since October 2010 he is now Professor for Computational Mathematics at the Department of Mathematics and the Center for Computational Engineering Science (CCES) at RWTH Aachen University. Furthermore, he is Principal Investigator at the Aachen Institute for Advanced Study in Computational Engineering Science (AICES), which is a Graduate School funded by the German Excellence Initiative.

06-03-2014   11:00am - noon   NIA Room 138      video

Compact-Reconstruction WENO Schemes - Theory, Implementation and Applications

The Compact-Reconstruction WENO (CRWENO) schemes are a new class of weighted, non-linear compact finite-difference schemes that have a high spectral resolution and yield non-oscillatory solutions across discontinuities. These schemes are thus well-suited for applications characterized by a large range of length scales as well as discontinuities and/or sharp gradients. Due to its non-linear nature, the CRWENO scheme requires the solution to a tridiagonal system of equations at each time-integration step/stage. This talk will describe the CRWENO schemes and their numerical properties. A scalable implementation of these schemes will be introduced based on an iterative sub-structuring of the tridiagonal system of equations. Our approach avoids the drawbacks of current approaches to implementing compact schemes on parallel platforms and scales well till very small sub-domain sizes per processor. The numerical properties as well as the efficiency and scalability of the CRWENO schemes will be demonstrated through several benchmark problems. Some of the results that will be presented are the largest parallel simulations with compact schemes till date, to the best of our knowledge.

[ presentation file (pdf) ] Debojyoti Ghosh

Speaker Bio: Dr. Debojyoti Ghosh is a post-doctoral researcher in the Mathematics & Computer Science Division at the Argonne National Laboratory. He received his doctorate in Applied Mathematics & Scientific Computation from the University of Maryland College Park. His research interests include the high-resolution finite-difference schemes, their scalable implementation and applications.

05-13-2014   11:00am - noon   NIA Room 141      video

Adjusted Formulation of Vorticity Confinement with Applications to Tip Vortices of Stationary and Rotaing Wings

Computational Fluid Dynamics solvers employ upwind discretization for convective terms and, while this improves numerical stability, it comes at the cost of numerical dissipation causing smearing of the solution at a much greater rate than physical diffusion. Modeling of the generation of vortex and its impingement remain a very computationally intensive task. To improve accuracy of CFD simulations by combating numerical viscosity using moderate-size grids, the vorticity confinement (VC) approach is able to capture tip vortices as they convect downstream through the addition of a centrifugal body force term in the momentum equations. We developed a fully adaptive VC based on local evaluation of the amount of numerical dissipation and on coupling of VC and Total variation Diminishing (TVD) approaches. Applications to helicopter blade tip vortices and to induced drag generated by fixed wing will be discussed.

[ presentation file (pdf) ] Alex Povitsky

Speaker Bio: Dr. Alex Povitsky is an Associate Professor of Mechanical Engineering at the University of Akron, Akron OH from 2003. Prior to it he was a lecturer at the Technion-Israel Institute of Technology (1994-97), Senior Scientist with the Institute of Computer Applications (ICASE) at NASA Langley Research Center (1997-2001), and Associate Professor at Concordia University (Montreal) (2001-2003). He published 50 journal papers in vorticity confinement approach, CFD modeling of micro- and nano- fluidics, development of parallel algorithms for implicit and high-order numerical schemes, modeling of laser ablation and hypersonic ablation, unsteady aerodynamics of micro air vehicles and aero-acoustics of vortex-body interaction. His research has been supported by Canadian NSERC, British Royal Society, Israel Academy of Sciences, NSF, NASA, AFOSR, ARO, US Navy, AFRL, ASEE, industry and the State of Ohio. He is a Senior Member of AIAA. He has been an Air Force Summer Faculty Fellow in 2005-2009, 2011-2012 and 2014.

Relevant Publication: T. Snyder and A.Povitsky, Far-field Induced Drag Prediction Using Vorticity Confinement Technique, AIAA J. of Aircraft, April 2014 [online ]

K. Pierson and A. Povitsky, Vorticity Confinement for Turbulent Wing Tip Vortices for Wake-integral Aircraft Drag Prediction, 21st AIAA Computational Fluid Dynamics conference, AIAA Paper 2013-2574, June 2013, San Diego, CA

K. Pierson and A. Povitsky, Preservation of Tip Vortices of Helicopter Blades by Vorticity Confinement, AIAA Paper 2013-2420, the 31st AIAA Applied Aerodynamics conference, San Diego, CA, June 2013.

05-06-2014   11:00am - noon   NIA Room 141      video

Parallel Tetrahedral Mesh Generation

With the deployment of extreme-scale HPC systems, generating finite element meshes in parallel is now necessary to achieve scalability of the overall simulation pipeline. The parallelization of the mesh generation algorithms requires that we not only preserve the quality of the meshes generated by the sequential algorithms, but also address the requirements of robustness and scalability. Robustness of a mesh generation algorithm is its ability to correctly process any geometric domain from the class of domains it is designed for. Scalability is the ability of an algorithm to achieve performance proportional to the number of hardware cores. In this talk I will focus on tetrahedral meshing algorithms due to their ability to adapt to complex geometric domains, and, furthermore, on algorithms that are accompanied by theoretical proofs of their correctness. I will review both the theoretical foundations and the practical aspects related to the implementation of parallel meshing methods on current and emerging architectures. I will organize the methods in terms of two basic attributes: the underlying sequential techniques and the degree of coupling between the subproblems.

[ presentation file (pdf) ] Andrey Chernikov

Speaker Bio: Dr. Andrey Chernikov is Assistant Professor of Computer Science at Old Dominion University. He obtained his Ph.D. from the College of William and Mary in 2007. His interests include parallel computational geometry with a focus on guaranteed quality mesh generation, and image analysis in medical and bio-material modeling and simulation.

04-22-2014   11:00am - noon   NIA Room 141      video (no audio due to a technical problem)

Walsh Functions in Numerical Simulation: A New Framework for Solving Nonlinear Systems of PDEs

A segmented, orthonormal basis function set composed of Walsh Functions is used for deriving global solutions (valid over the entire domain) to nonlinear differential equations that include discontinuities. A powerful, self-mapping characteristic of this set is closure under multiplication -- the product of any two elements of the set is also an element of the set: gn(x) gm(x) = gk(x) (xb - xa)^{-1/2}. In the same way that Fourier series are used to generate global solutions to linear problems, this self-mapping property under multiplication allows similar approaches to non-linear problems. A new derivation of the basis functions applies a fractal-like algorithm (infinitely self-similar) focused on the distribution of segment lengths. Only two segment lengths are allowed in a group p. A recursive-folding algorithm that propagates fundamental symmetries to successive functions in the series determines the distribution of segment lengths. Functions, including those with discontinuities, may be represented on the domain as a series in gn(x) with no occurrence of a Gibbs phenomenon (ringing) across the discontinuity. Integrals and derivatives are computed using simple matrix multiplication of series representations of functions without the need for divided differences. A FORTRAN module for supporting Walsh function simulations is discussed. Examples are discussed for solution of the time dependent problems: an advection equation, a Burgers equation, and a Riemann problem.

[ presentation file (pdf) | Movie (Shock Tube)] Peter Gnoffo

Speaker Bio: Dr. Peter Gnoffo is Senior Computational Aerothermodynamicist at NASA Langley Research Center. He earned his Ph.D in Mechanical and Aerospace Engineering at Princeton University in 1983. He joined NASA in 1974 after receiving a B.S. in Aerospace Engineering from Polytechnic Institute of Brooklyn. The subject lecture follows recent work published in the Journal of Computational Physics, Vol 258, pp 650-688, Feb 2014, titled 'Global Series Solutions of Nonlinear Differential Equations with Shocks Using Walsh Functions.'

Relevant Publication: "Global Series Solutions of Nonlinear Differential Equations with Shocks using Walsh Functions'", Journal of Computational Physics, Volume 258, p. 650-688. [online ]

03-04-2014   11:00am - noon   NIA Room 137      video

Development of Fast and Efficient CFD Tools, HexaGrid and FaSTAR

We have developed efficient CFD tools: an automatic hexahedra grid generator "HexaGrid" and a fast flow solver "FaSTAR". HexaGrid can generate Cartesian-based unstructured grids with body-fitted layer grids just by inputting CAD geometry data and several control parameters such as surface grid sizes and domain sizes. Since the layer grids are generated by a projection method, the time-consuming work that repairs gaps and overlaps of surface data is minimized. FaSTAR is a compressible flow solver of unstructured grids. It employs a simple face-based data structure and this helps performance optimization of the codes. An agglomerated multigrid method is also utilized to accelerate convergence. The octree data structure of the Cartesian grid is used for the agglomeration to keep the coarse-grid quality. Steady and unsteady simulations with HexaGrid and FaSTAR are presented at the seminar. We finally show a future outlook of CFD code development.

[ presentation file (pdf) ] Atsushi Hashimoto

Speaker Bio: Dr. Atsushi Hashimoto is a researcher at Japan Aerospace Exploration Agency (JAXA). He earned Ph.D. in Aerospace Engineering at Nagoya University, Japan in 2007. He joined JAXA in 2007. His area of expertise is aerodynamic prediction with unstructured-grid CFD, aeroacoustics, and aeroelasticity.

Relevant Publication: "Toward the Fastest Unstructured CFD Code 'FaSTAR'", AIAA2012-1075 [online]

02-18-2014   11:00am - noon   NIA Room 101      video

Evaluation of Multigrid Solutions for Turbulent Flows

A multigrid methodology has been recently developed in a NASA solver, FUN3D, and successfully applied for a wide range of turbulent flows, from simple two-dimensional geometries to realistic three-dimensional configurations. The methodology is applicable to structured- and unstructured-grid solutions and includes both regular and agglomerated coarse meshes. Significant speed-ups over single-grid computations have been demonstrated. In the current work, we report on a detailed evaluation of the solver performance in computing benchmark turbulent flows. For those benchmark computations, multigrid solutions are compared with the corresponding single-grid solutions in terms of time-to-solution characteristics measured in the same computing environment. Multigrid efficiency enables a grid-refinement study of a turbulent flow around an airfoil that is discussed in detail. This talk is an extended version of the talk presented at AIAA SciTech-2014 conference.

[ presentation file (pptx) ] Boris Diskin

Speaker Bio: Dr. Boris Diskin is Research Fellow at NIA. He earned Ph.D. in Applied Mathematics at The Weizmann Institute of Science, Israel in 1998. He was a Senior Research Scientist at ICASE and joined NIA from its inception in 2003. His area of expertise is adjoint-based optimization and grid adaptation methods, finite-volume discretizations, and multigrid methods.

12-03-2013   11:00am - noon   NIA Room 137      video

Higher-Order Moments and Their Modeling Approximations in a Turbulent Channel Flow Subjected to Mean Strain

Third- and fourth-order moments and the budget terms in their transport equations are evaluated using results from Direct Numerical Simulations (DNS) of a turbulent channel flow at Re_tau = 395, to aid in the development of higher-order Reynolds-Averaged Navier Stokes (RANS) closure models. These models have been proposed as a means of obtaining more accurate predictions of complex flows. Strain has been applied to this channel to mimic a spatially-evolving adverse pressure gradient flow leading to separation. Existing modeling hypothesis that relate higher-order moments to their lower-order moments, such as gradient diffusion and assumed probability distribution functions have been tested. Insights into where to focus modeling efforts will be presented.

[ presentation file (pdf) ] Elbert Jeyapaul

Speaker Bio: Dr. Elbert Jeyapaul is a post-doctoral fellow at NIA, working on RANS turbulence modeling. He received his doctorate in Aerospace engineering in 2011 from Iowa state university.

09-24-2013   11:00am - noon   NIA Room 101      video

The MOOD Method --- Multidimensional Optimal Order Detection -- a first a posteriori approach to Very-High-Order Finite Volumes methods

This talk will be dedicated to the new type of very high-order Finite Volume methods for hyperbolic systems of conservation laws that I introduced and developed during my doctoral studies. This method, named MOOD for Multidimensional Optimal Order Detection, provides very accurate simulations for two- and three-dimensional unstructured meshes. The design of such a method is made delicate by the emergence of solution singularities (shocks, contact discontinuities) for which spurious phenomena (oscillations, nonphysical values creation, etc.) are generated by the high-order approximation. The originality of this work lies in a new treatment for theses problems. Contrary to classical methods which try to control such undesirable phenomena through an a priori limitation, we propose an a posteriori treatment approach based on a local scheme order decrementing. In particular, we show that this concept easily provides properties that are usually difficult to prove in a multidimensional unstructured framework (positivity-preserving for instance). The robustness and quality of the MOOD method have been numerically proved through numerous test cases in 2D and 3D for single-material compressible flows, and a significant reduction of computational resources (CPU and memory storage) needed to get state-of-the-art results has been shown. I will moreover present some preliminary results on my current work at LANL for the case of multi-material compressible flows.

[ presentation file (pdf) ] Steven Diot

Speaker Bio: Dr. Steven Diot is Postdoctoral Research Associate at the Los Alamos National Laboratory. He received his Ph.D. degree in Applied Mathematics from the University of Toulouse (France) in 2012. During his Ph.D. research studies, he developed a new approach to Very-High-Order Finite Volume methods for single-material compressible flows called the Multidimensional Optimal Order Detection (MOOD) method. An important part of his postdoctoral studies is the extension of the MOOD method to multi-material compressible flows and to coupled physics.

[ Home Page | PhD Defense Slides | Award-winning movie ]

09-10-2013   11:00am - noon   NIA Room 141      video

Non-equilibrium Pressure Considerations in Modeling Viscous Rotating Flows

Hamilton's Principal of Least Action has been employed to incorporate non-equilibrium fluid behavior in the study of simple viscous flows. The original goal was to clarify the role of bulk viscosity in modeling high-speed viscous and reacting flows.[1,2] The theory demonstrated that bulk viscous transport effects and non-equilibrium density effects coexisted, but were intertwined. Employing acoustically-based estimates for this overall bulk viscous effect, the theory predicted correctly the density-jump and shock thickness for air flows at various Mach numbers.[2] The theory also exhibited a separate non-equilibrium pressure effect, and the associated pressure relaxation coefficient could be estimated using acoustic data. Recently-published work [3] has identified a possible new low-speed sound source, while predicting that pressure deficits within the central cores of tornadoes and dust devils (and similarly aircraft trailing line vortices) can be substantially larger than equilibrium-based estimates. Modifying the Navier-Stokes equations to incorporate non-equilibrium pressure gradient forces in simple vortical flows resulted in an exact solution for an axial vortex [3] that bypasses the shear stress discontinuity found in Rankine vortex models by way of a a sheath-like non-equilibrium pressure zone that effectively isolates a rigidly-rotating central fluid region from an outer potential vortex region. Subsequently, the near-isolation of such a rotating central fluid column has been used to show that non-equilibrium pressure gradient forces can control the height of the central column. [4] The rotating column solution was employed successfully to predict the heights of terrestrial dust devil columns.

Non-equilibrium pressure gradient forces offer the potential for modeling Mars dust devil behavior more accurately, possibly enabling the estimation of maximum pressure deficits and swirl velocities via satellite observation. Of potentially greater importance, the theory may explain how streaks are formed and sustained in the wall region of turbulent boundary layers and thus offer the potential for improved turbulence models.

[ presentation file (pdf) ] Robert L. Ash

Speaker Bio: Dr. Robert L. (Bob) Ash joined the engineering faculty at Old Dominion University at the time when LaRC was heavily involved in the design and development of their spectacularly successful Viking missions to Mars. His research has been sponsored primarily by NASA, and accomplishments have ranged from co-inventing riblets for turbulent skin friction drag reduction through determination that the most cost-effective and practical way to accomplish sample return and human missions to Mars was to manufacture the rocket propellant required for the return trip in situ using local Mars water and carbon dioxide (while he was working as a National Research Council Senior Resident Researcher at JPL). He and now-retired LaRC scientist, Dr. Allan J. Zuckerwar, have been collaborating for more than 30-years, working to develop a better understanding of the role of bulk viscosity in modeling fluid behavior, and that effort has led to the work that will be discussed in this seminar. Professor Ash is designated as an Eminent Scholar in the Mechanical and Aerospace Engineering Departmetn at Old Dominion University.
[ Home Page ]
Relevant Publications: 1. Zuckerwar, A. J. and R.L. Ash, "Variational approach to the volume viscosity of fluids", Physics of Fluids, Vol. 18, 047101 (10 pages), April 2006.
[ journal ]

2. Zuckerwar, A.J., and R.L. Ash, "Volume viscosity in fluids with multiple dissipative processes", Physics of Fluids, Vol. 21, 033105 (12 pages), March 2009.
[ journal ]

3. Ash, R.L., I. Zardadkhan, and A.J. Zuckerwar, "The influence of pressure relaxation on the structure of an axial vortex", Physics of Fluids, Vol. 23, 073101, (12 pages), July 2011.
[ journal ]

4. Ash, R.L. and I.R. Zardadkhan, "Non-equilibrium pressure control of the height of a large-scale, ground-coupled, rotating fluid column", Physics of Fluids, Vol. 25, 053101, (20 pages), doi: 10.1063/1.4807068, May 2013.
[ journal ]

08-27-2013   11:00am - noon   NIA Room 141      video

Chaotic adjoint: from massive linear algebra to a light-weight wrapper of existing solvers

Many high fidelity unsteady flow simulations at high Reynolds number exhibit chaotic dynamics. Failure of conventional tangent and adjoint solvers for these simulations has been demonstrated on flow solvers including FUN3D. The Least Squares Shadowing approach recently developed by the speaker overcomes the ill-conditioning of chaotic dynamical system, and lead to stable tangent and adjoint sensitivity analysis. This talk will start with the problem faced in chaotic adjoint. We then introduce the Least Squares Shadowing approach, and the massive space-time 4D linear system it requires to solve. Finally, we show how solving the linear system can be reduced to a light-weight wrapper of an existing PDE and adjoint solver.

[ presentation file (pdf) ] Qiqi Wang

Speaker Bio: Dr. Qiqi Wang is Assistant Professor of Aeronautics and Astronautics in Massachusetts Institute of Technology. His areas of expertise include design, optimization and uncertainty quantification, unsteady flows, aeroelastics and aeromechanics, computational sensitivity analysis. [ Home Page ]

Copyright 2011- . National Institute of Aerospace. All rights reserved.