Unless otherwise noted, all seminars take place in ACE 6.304 from 3:30 – 5:00 PM; Everyone is welcome to attend. Refreshments are served at 3:15pm.
For information on seminars sponsored by the Mathematics Department, go to Math Seminars.
Jump to seminars for:
December |
November |
October |
September |
August |
July |
June |
May |
April |
March |
February |
January |
Thursday, May 22
Dr. Yuri Bazilevs
ICES Postdoctoral Fellow
ICES Seminar: “Isogeometric Analysis of Fluids and Fluid-Structure Interaction”
Abstract:
Recently, Isogeometric Analysis has emerged as a new computational technology and as an alternative to the standard finite element method. Isogeometric analysis improves upon finite elements in the areas of geometric modeling and solution representation. The first instantiation of isogeometric analysis was based upon non-Uniform Rational B-Splines NURBS), although other alternatives, such as subdivision and T-Splines, are possible and are currently under investigation. Despite its recent emergence, NURBS-based isogeometric analysis has already been applied to many areas of contemporary interest in computational mechanics. These include: fluids and turbulence, solids and thin structures, fluid-structure interaction and, recently, phase-field modeling. Improved geometry and solution approximation properties of NURBS functions has led to superior performance of the isogeometric approach in comparison to standard finite elements in these applications. This presentation focuses on application of isogeometric analysis to wall-bounded turbulent flows, flows about rotating components, and vascular fluid-structure interaction. Basic ideas on how to develop discrete formulations that yield accurate and stable solutions for these applications are presented. Implementation of these ideas within a NURBS-based, largescale computational framework is discussed and computations that demonstrate good performance of the proposed methodology are shown.
Tuesday, May 20
Dr. Bernd Flemisch
IWS, Universitaet Stuttgart
ICES Seminar: “Multi–scale Multi–physics Modeling of Flow and Transport Processes in Porous Media”
Abstract:
Flow and transport processes in porous media occur on different spatial and temporal scales and may also exhibit different physical behavior in different parts of the model domain. Additionally, the structure of the porous medium itself generally shows a high dependence on the spatial scale. Employing a complex fine scale model throughout the whole domain is in general very expensive and often limited by the available computational resources. Moreover, in many cases, it is not needed to use the full model everywhere, and one can choose a simpler one and/or consider one on a coarser scale in large parts of the domain.
Our goal is to develop a multi-scale multi-physics simulator DUMUX for nonisothermal compositional multiphase flow and transport processes in porous media which makes use of the locally different behavior. In particular, we aim to use in each part of the domain the model which is the less expensive but still accurate enough to describe the physics correctly. Furthermore, we want to be able to choose the best discretization in space and time for each model. We use modern programming techniques within the framework of DUNE, the Distributed and Unified Numerics Environment which has been developed recently. It allows to implement the necessary numerical algorithms independent of the ultimately used data structures. This talk presents our modeling concept and the current possibilities of our simulator.
Thursday, May 15
Dr. Elizabeth Olson
Assistant Professor of Biomedical Engineering and Auditory Biophysics, Columbia University
Joint ICES/Acoustics Seminar: “The Size and Shape of Intracochlear Pressure”
Abstract:
Upon sound stimulation, the eardrum and ossicles are set in vibration. The stapes
vibration pressurizes the cochlear fluids. Due to the cochlea's geometry and its
approximately symmetric division by a flexible cochlear partition (which includes
the sensory tissue), the intracochlear pressure can be decomposed into two dominant
pressure modes. One of these, the compression mode, fills the cochlea nearly uniformly.
The other, traveling-wave mode, is anti-symmetric across the partition.
The traveling-wave mode creates the cross-partition pressure difference that leads to sensory hair-cell motion and excitation (and damage, with loud sounds).
As far as we know, the compression mode serves no physiological function and is
simply a by-product of the piston-like drive of the stapes upon the cochlea. Small intracochlear pressure sensors developed a decade ago have been useful in parsing the two intracochlear pressure modes and for probing the traveling-wave mode where it is largest, close to the cochlear partition. The traveling-wave mode shows many of the same interesting features as cochlear partition motion, in particular frequency tuning and nonlinearity that arises from active hair-cell - based mechanical nonlinearity. In this talk these fundamental characteristics of intracochlear pressure will be reviewed.
Wednesday, May 14 (Location: ACES 2.402) (Time: 1:30 — 3:00 PM)
Dr. Nawaf Bou–Rabee
DFG Research Center MATHEON and Free University of Berlin, Germany
ICES Seminar: “Structure–Preserving Langevin Integrators for Computational Femtochemistry”
Abstract:
This talk is on discrete Markov chains for simulation of mechanical systems at uniform temperature. These Markov chains are based on a probablistic/geometric discretization of Ornstein-Uhlenbeck processes in a potential force field. I will show that these Markov chains possess attractive structure-preserving properties; in particular, I will show that they exponentially converge to their equilibrium
distribution, and I will quantify the appealing forward error they make in preserving the Boltzmann-Gibbs measure. I will also describe how one can control this sampling error. These integrators can readily incorporate holonomic constraints (which often arise in molecular simulations), and are well-suited for rare event simulations. They arose from a nascent continuous and discrete Lagrangian theory for stochastic Hamiltonian systems due to myself and Houman Owhadi (ACM/CDS Caltech). This theory is built on a stochastic version of mechanics, originally defined by Bismut, and along the lines of Ornstein-Uhlenbeck theory. The talk will include a discussion of applications, latest developments, and future directions.
Short Bio:
Nawaf Bou-Rabee completed his PhD in Applied and Computational Mathematics in 2007 at the California Institute of Technology. He is currently a visiting guest scientist at MATHEON in Berlin, Germany and plans to begin a NSF mathematical sciences postdoctoral research fellowship at Courant Institute in Fall 2008. His research interests include stochastic variational integration theory; structure-preserving integrators; stochastic numerics; multiscale integrators for biomolecular systems and atomic clusters; and Markov-Chain Monte-Carlo Methods.
Thursday, May 8
Dr. Martin Vohralik
Laboratoire Jacques–Louis Lions Université Pierre et Marie Curie (Paris 6)
ICES Seminar: “Guaranteed (and Robust) a–Posteriori Error Estimates in Continuous and Discontinuous Galerkin Finite”
Abstract:
We establish a general framework which allows for optimal a posteriori error estimation in approximation of linear second-order elliptic partial differential equations by different numerical methods. Fully computable upper bounds are derived, so that the estimators allow for the overall energy error control, whereas adaptive mesh refinement is supported by the local efficiency of the estimators. For continuous Galerkin finite elements or vertex-centered finite volumes, the estimates are also fully robust with respect to diffusion inhomogeneities (without any "monotonicity" assumption) and reaction dominance. For convection-dominated problems, the estimates are semi-robust, with the efficiency depending on the local mesh Peclet number only. The case of inexact solution of the associated linear systems is also addressed. Numerical results illustrate the theoretical developments.
Wednesday, May 7 (Location: ACES 2.402)
Benjamin S. Kirk, Adam J. Amar
NASA Lyndon B. Johnson Space Center, Aerosciences & Flight Mechanics Division
ICES Seminar: “Orion Reentry: Modeling the Aerothermodynamic Environment and Thermal Protection System Response”
Abstract:
Atmospheric entry at superorbital speeds involves a number of complex physical phenomena including (i) thermochemical nonequilibrium fluid mechanics, (ii) radiative heat transfer, and (iii) ablative thermal protection system material response. These processes, which are intrinsically coupled, are often treated in a decoupled fashion in engineering analysis and design. This presentation will review the physical processes and governing equations for the aforementioned phenomena, and will discuss the particular challenges which complicate performing predictive simulations. The current approach being used at NASA for the design of the Orion Crew Module will be discussed, with particular emphasis on known limitations of our current modeling schemes. Relevant Orion and historical Apollo design challenges will be reviewed so as to provide motivation for a more rigorous, fully coupled simulation capability which could be used in the future to more accurately model the coupled multiphysics lunar & planetary reentry environment.
Thursday, April 10 (Location: ACES 2.302/Avaya Auditorium)
Horst D. Simon
Associate Laboratory Director for Computing Sciences, Lawrence Berkeley National Laboratory, Adjunct Professor of Computer Science, University of California, Berkeley
ICES/TACC Distinguished Lecture Series in Petascale Simulation: “The Greening of HPC — Will Power Consumption Become the Limiting Factor for Future Growth in HPC?”
Abstract:
In a recent survey by IDC, facilities managers by an overwhelming majority named power and cooling to be the most pressing issues to them. A study of Exaflops computing concluded that by projecting today’s technology, an Exaflops computer might require 120 megawatts of power. A different study commissioned by the EPA estimates that server power consumption doubled in the period from 2000 to 2005 worldwide, and that the total amount of electricity consumed by servers worldwide now costs about $7.2B. This is the same order of magnitude as the investment in today’s HPC technology ($9.2B). Thus, we have reached a critical threshold that gives us cause to consider the question of power consumption as a potentially limiting factor to the future growth in HPC.
What are the power limitations of current technology, and how can we change the equation to assure the future rapid growth of HPC performance without contributing even more to carbon emissions and global warming? Dr. Simon will discuss several research projects that have started in Berkeley to address the issue of reducing power consumption in HPC, both at the systems and at the building level.
Speaker Biography: Horst Simon is Associate Laboratory Director at Lawrence Berkeley National Laboratory for Computing Sciences, Division Director for the Computational Research Division, and Adjunct Professor of Computer Science at the University of California, Berkeley. His research interests are in the development of sparse matrix algorithms, algorithms for large–scale eigen value problems, and domain decomposition algorithms. His recursive spectral bisection algorithm is a breakthrough in parallel algorithms honored with the 1988 Gordon Bell Prize. He has served in senior management positions at Silicon Graphics, the Computer Sciences Corporation, Boeing Computer Services, and was a faculty member at the State University of New York.
Currently, he serves on advisory boards at research organizations located throughout the world, is a member of many journal editorial boards, and is one of four editors of the twice–yearly “TOP500” list of the world’s most powerful computing systems.
Reception: 3:00 – 3:30 pm, ACES Connector Lobby
Live Webcast: http://www.tacc.utexas.edu/petascale
Tuesday, April 8
Prof. Benqi Guo
Department of Mathematics, University of Manitoba, Canada
ICES Seminar: “Some Important Issues in Approximation Theory for High–order Finite Element Method in Three Dimensions”
Abstract:
he approximation theory for high–order finite element method in three dimensions is much less developed than in one and two dimensions because of several fundamental issues which have not been answered in the past two decades. The talk will address these fundamental issues and update the recent progresses and challenging problems on these issues.
Wednesday, April 2 (Time: 2:30 — 3:30 PM)
Krzysztof Kuczera
Departments of Chemistry and Molecular Biosciences, University of Kansas
ICES/CLSB Seminar: “Simulations of Peptide and Protein Dynamics in Solution”
Abstract:
I will present results of molecular dynamics simulations of peptide and protein systems of increasing degree of complexity that have been studied in my group recently. The starting point will be the analysis of reorientational motions of protein sidechain models in solution, which will also serve to introduce the tools used to analyze rotational motions, and present a baseline for expected accuracy. At the next level, I will describe reorientations in aromatic residue dipeptides — Phe, Tyr and Trp, and in an enkephalin analog, the YGGWL pentapeptide. The final topic will be the analysis of domain motions in a large tetrameric protein, S–adenosyl–L–homocyteine hydrolase. In all cases the simulations will be compared to available experimental data and new microscopic insights obtained from the simulations discussed.
Wednesday, March 26 (Time: 2:30 — 3:30 PM)
Jarek Meller
Cincinnati Children’s Hospital Medical CenterCincinnati Children’s Hospital Medical Center
ICES/CLSB Seminar: “Model Assessment for Soluble and Membrane Proteins Using Solvent Accessibility-based Scoring Functions”
Abstract:
Folding simulation and protein structure prediction methods are capable of generating models with near native structures. However, the problem of selecting such near native structures from a population of alternative (and often very different) models remains challenging, owing to inherent limitations of empirical force fields and folding potentials. Knowledge-based scoring functions for model assessment, often optimized using statistical and machine learning approaches, have therefore become an important component of structure prediction efforts. We present a novel scoring function for model assessment, which integrates state-of-the-art predictions of solvent accessibility with contact models in order to yield improved accuracies. The new model is assessed and compared with other approaches from the literature on several different types of models of soluble proteins. Extensions to membrane proteins are also discussed.