(September 06) Prof. Jian-Guo Liu:
Novel Finite Element Navier-Stokes Solvers without Inf-Sup Conditions -
We will show that in bounded domains with no-slip boundary conditions,
the Navier-Stokes pressure can be determined in a such way that it is
strictly dominated by viscosity in an equivalent unconstrained
formulation. As a consequence, in a general domain we can treat the
Navier-Stokes equations as a perturbed vector diffusion equation,
instead of as a perturbed Stokes system. We provide a simple proof of
unconditional stability and convergence for discretization schemes
that are implicit only in viscosity and explicit in both pressure and
convection terms, requiring no solution of stationary Stokes
systems or inf-sup conditions, particularly for corresponding
fully discrete finite-element methods with $C^1$ elements for velocity
and $C^0$ elements for pressure. It is important to note that we
impose {\em no} inf-sup compatibility condition between the
finite-element spaces for velocity and pressure. The inf-sup
condition (also known as the Ladyzhenskaya-Babuska-Brezzi
condition) has long been a central foundation for finite-element
methods for all saddle-point problems including the stationary Stokes
equation. Its beautiful theory is a masterpiece documented in many
finite-element books. In the usual approach, the inf-sup condition
serves to force the approximate solution to stay close to the
divergence-free space where the Stokes operator is
dissipative. However, due to the fully dissipative nature of the new
unconstrained formulation, the finite-element spaces for
velocity and pressure can be completely unrelated.
This is a joint work with Bob Pego and Jie Liu.
(September 08) Prof. Christoph Schwab:
Sparse Adaptive FEM for Multiple Scale Problems -
Elliptic homogenization problems in a domain $\Omega \subset \R^d$
with $n+1$ separated scales are reduced to elliptic one-scale
problems in dimension $(n+1)d$. These one-scale problems are
discretized by a sparse tensor product finite element method (FEM).
We prove that this sparse FEM has accuracy, work and memory requirement
comparable to standard FEM for single scale problems in $\Omega$ while
it gives numerical approximations of the correct homogenized limit
as well as of all first order correctors, throughout the physical
domain with performance independent of the physical problem's scale
parameters. Anisotropic Besov regularity in the scales and adaptive wavelet FEM
will be addressed. Numerical examples for model diffusion problems
with two and three scales will be given. The complexity of the method
will be compared to that of the recently proposed hierarchical
multiscale methods. (September 13) Prof. Ohannes Karakashian:
A Posteriori Error Estimates and Convergence of Adaptive Methods
for a Discontinuous Galerkin Method for Elliptic Problems
-
A posteriori error estimates are derived for an interior penalty
formulation for discontinuous Galerkin methods for elliptic problems.
Residual type estimators as well as those based on the solution of
local problems are formulated. The estimators are used to devise
adaptive refinement strategies. It is shown that under certain
conditions the algorithms are convergent with a guaranteed error
reduction rate.
(September 20) Prof. Eberhard Baensch:
Uniaxial, Extensional Flows in Liquid Bridges
-
Linear flow fields are commonly used for rheological studies,
e.g. to measure the fluid viscosity or the deformation behavior
of the whole sample or components in it. In this presentation we
consider the possibility of generating homogeneous
flows with a nearly constant strain rate. This is achieved by stretching an almost cylindrical liquid bridge
under microgravity. One key issue is the adjustability of the disk
diameters, necessary for maintaining ideal boundary conditions.
This method gives rise to much weaker end-effects than the commonly
used method with unchangeable disks. For the numerical simulation of the problem we use a finite
element method with the following key ingredients: a variational
formulation for the curvature of the free boundary, yielding an
accurate, dimensionally-independent and simple-to-implement
approximation for the curvature; a stable time discretization,
semi-implicit with respect to the treatment of the curvature terms.
This firstly allows one to choose the time step independently
of the mesh size in contrast to common ``explicit'' treatments
of the curvature terms, and secondly decouples the computation of the
geometry and the flow field. By systematic numerical simulations we could identify different
regimes depending on $\Ca$, $\We$ and $\Re =\We / \Ca$. Regions of
{\it capillary-dominated flow}, \hbox{$\Ca \ll 1$}, $\We \ll 1$,
{\it viscous-dominated flow}, $\Ca > {\cal O}(0.1)$, $\Re < {\cal O}(0.1)$,
and {\it inertia-dominated flow}, $\We > {\cal O}(0.1)$, $\Re >{\cal O}(0.1)$
have been detected, which exhibit mutually different bridge deformations
during stretching. (September 27) Chensong Zhang:
A Posteriori Error Analysis for Parabolic Variational Inequalities
in Finance -
Motivated by the pricing American options for baskets we consider
a parabolic variational inequality in a bounded polyhedral domain
$\Omega\subset\mathbb{R}^d$ with an obstacle $\chi$. We formulate
a fully discrete method by using piecewise linear finite elements
in space and the backward Euler method in time. We define an
a-posteriori error estimator and show that it gives an upper
bound for the error in $L^2([0,T],H^1(\Omega)$. The error
estimator is localized in the sense that the size of the elliptic
residual is only relevant in the approximate non-contact region,
and the approximability of the obstacle is only relevant in the
approximate contact region. We also lower bound results for the
local space error estimators in the non-contact region, and for
the time error estimator. We present numerical results for
$d=1,2$ which show that error estimator decays with the same rate
$O(\tau + h)$ as the actual error when the space mesh size $h$
and the time step $\tau$ tend to zero. Also, the local error
estimators capture the correct behavior of the errors in both the
contact and the non-contact regions.
This is joint work with K-S. Moon, R.H. Nochetto and T. von
Petersdorff. (October 05) Prof. Chi-Wang Shu:
Anti-diffusive High Order Weighted Essentially Non-Oscillatory Schemes
for Sharpening Contact Discontinuities -
In this talk we will first describe the general framework of high
order weighted essentially non-oscillatory (WENO) finite
difference schemes for solving hyperbolic conservation laws and
in general convection dominated partial differential equations.
We will then discuss our recent effort in designing
anti-diffusive flux corrections for these high order WENO
schemes. The objective is to obtain sharp resolution for contact
discontinuities, close to the quality of discrete traveling waves
which do not smear progressively for longer time, while
maintaining high order accuracy in smooth regions and
non-oscillatory property for discontinuities. Numerical examples
for one and two space dimensional scalar problems and systems
demonstrate the good quality of this flux correction. High order
accuracy is maintained and contact discontinuities are sharpened
significantly compared with the original WENO schemes on the same
meshes. We will also report the extension of this technique to
solve Hamilton-Jacobi equations to obtain sharp resolution for
kinks, which are derivative discontinuities in the viscosity
solutions of Hamilton-Jacobi equations. This is joint work with
Zhengfu Xu.
(October 11) Prof. Alfred Schmidt:
Adaptive Finite Element Methods for Allen-Cahn and Phase Field
Problems -
Phase field models, combining an Allen-Cahn equation for
interface motion with a degenerate heat equation, are widely used
to model phase transitions on a mesoscopic scale. Based on a
posteriori error estimates, adaptive finite element methods for
phase field problems are presented. For a double obstacle
potential, the error indicators combine estimates for parabolic
variational inequalities and degenerate diffusion problems. For a
smooth potential, spectral estimates can be used to reduce the
strong dependence of the estimator's constants on the inherent
relaxation parameter. The talk reports on (ongoing) joint works with Ricardo H.
Nochetto, Zhiming Chen, and Daniel Kessler. (October 18) Prof. Homer Walker:
Globalization techniques for Newton-Krylov methods -
A Newton-Krylov method is an implementation of Newton's
method in which a Krylov subspace method is used to solve
approximately the linear subproblems that determine Newton
steps. To enhance robustness when good initial approximate
solutions are not available, these methods are often
"globalized", i.e., augmented with auxiliary procedures
("globalizations") that improve the likelihood of convergence
from a poor starting point. In recent years, globalized Newton--Krylov
methods have been used increasingly for the fully-coupled solution of
large-scale CFD problems. In this talk, I will review several
representative globizations, discuss their properties, and
report on a numerical study aimed at evaluating their relative
merits on large-scale 2D and 3D problems involving the
steady-state Navier--Stokes equations. This is joint work
with John Shadid and Roger Pawlowski at Sandia National
Laboratories and Joseph Simonis at WPI. (October 25) Cornelia Fermuller:
Encounters of Vision with Scientific Computing -
I will present two of our encounters of Scientific Computing with
the vision system during the past few years. The first encounter
gave rise to a theory that predicts many geometric optical
illusions. The theory is based on estimation from limited amounts
of data under noisy conditions. The second encounter was in the
study of superresolution image reconstruction from sequences of
low resolution images. Using the theory of wavelet filter banks
we investigated the theoretical limits on the reconstruction and
developed an efficient algorithm which is optimal when the noise
model is unknown. (November 1) Prof. Zdenek Strakos:
Estimating the A-norm of the error in (Preconditioned) CG. -
Estimation of the A-norm of the error in the conjugate gradient
method is studied in many papers, reports and sections of books.
In most of them the estimates are based on the fundamental relationship
of the conjugate gradient and Lanczos methods to Gauss quadrature
of the Riemann-Stieltjes integral defined by the spectral decomposition
of the system matrix and components of the right hand side in
the individual invariant subspaces.
After summarizing this fundamental relationship and recalling
several possible ways of convergence evaluation,
we present a simple estimate for the A-norm of the error in the
preconditioned conjugate gradient method. We then describe results
of numerical stability analysis and illustrate the effectivity and
possible drawbacks of the proposed estimate on numerical experiments. This represents joint work with Petr Tichy.
(November 8) Prof. Jerry Prince:
Cortical Surface Alignment Using Optical Flow in an Eulerian Framework -
How does one match locations on the human brain cortex of one person
to that of another? A standard approach is deformable 3D-3D
registration. This talk describes an alternative approach that focuses
directly on the cortical surfaces and their geometric properties.
Gyri and sulci are aligned in a multiresolution fashion using multiple
geometric features computed on both cortices and computed in an Eulerian
fashion. The result is an alignment of the cortices, which can then be
used to apply a coordinate system to a given subject or to perform
population analyses of function in a standardized coordinate system
on the cortical surface.
(November 15) Prof. Marco Verani:
A Finite Element Formulation for a Shape Optimization Problem -
Several problems arising in Engineering can be formulated
as shape optimization problems, i.e. the optimal design of an
aorto-choronaric bypass, of a drug eluting stent or the optimal
design of a rowing boat. In this talk I will show how it is
possible to write a suitable FEM formulation for a simple Shape
Optimization problem by coupling Shape Calculus tools and
Finite Elements Methods for geometric flows. (this is a joint project with G. Dogan, P. Morin and R. Nochetto)
(November 22) Prof. L. Ridgway Scott:
The FEniCS Project to Automate Computational Mathematical Modeling -
The objective of computational mathematical modeling is to make
quantitative predictions of natural phenomena based on some type
of mathematical description. Sophisticated models using partial
differential equations can be found today in disciplines as
diverse as economics, financial modeling, protein design,
materials engineering, as well as more traditional areas.
However, the development of software to implement these models
remains costly and error-prone. Numerous attempts have been made to develop problem-solving
environments for partial differential equations (p.d.e's). Some
of these have allowed (1) the specification of arbitrary p.d.e's,
(2) the use of families of finite elements of arbitrary degree,
(3) the incorporation of industrial-strength geometry engines,
mesh generators, error estimators and adaptive mesh techniques,
(4) the integration with parallel support tools, and (5) the use
of multi-level and domain-decomposition solvers in addition to
efficient sparse direct methods. However, no system has made
use of all of these to the fullest extent possible, and there
are important research issues related to doing so. The FEniCS project is an international effort to study the
fundamental issues related to automating the generation of
software that can be used to approximate the solutions of p.d.e's
and other similar systems. In addition to providing multiple
end-user systems, FEniCS is developing what is called middle-ware
to support such efforts. The concept of middle-ware goes beyond
that of simple libraries, but the individual components of the
middle-ware can be used independently and are being adopted by
developers of other systems. We will discuss the general outlines of the FEniCS project and
then describe some specific efforts to generate efficient code
for matrix assembly using a system called FErari. FErari can
be thought of as part of a code generator which can be used for
general finite elements and p.d.e's. We show that it is possible
to have very efficient assembly for low-order methods, comparable
to what is known for high-order methods using spectral-element
technology.
(November 29) Prof. Alex Kiselev:
Diffusion and Mixing in Fluid Flow -
Enhacement of diffusion by advection is a classical subject that has
been extensively studied by both physists and mathematicians. In this
work, we consider enhancement of diffusive mixing on a compact
Riemannian manifold by a fast incompressible flow. Our main result
is a sharp description of the class of flows that make the deviation
of the solution from its average arbitrarily small in an arbitrarily
short time, provided that the flow amplitude is large enough. The
necessary and sufficient condition on such flows is expressed
naturally in terms of the spectral properties of the dynamical system
associated with the flow. In particular, we find that weakly mixing
flows always enhance the relaxation speed in this sense. The proofs
are based on a new general criterion for the decay of the semigroup
generated by a dissipative operator of certain form. They employ
ideas from quantum dynamics, in particular the RAGE theorem
describing evolution of a quantum state belonging to the continuous
spectral subspace of the hamiltonian (and related to a theorem of
Wiener on Fourier transforms of measures). (December 06) Prof. Chun Liu:
Macro-Micro Models in Viscoelastic Materials: An Energetic
Variational Approach
-
I will describe an unified energetic variational framework
for elastic complex fluids. It highlights the competition of the kinetic
energy and the elastic energies, through the transport of the internal
elastic variables. As applications, I will discuss
some macro-micro scale coupling effects in these materials.
Corresponding analytical and numerical issues will be
addressed, in particular, I want to present some recent well-posedness
results and a new moment closure algorithm that preserves the energy law
of the original system.
(December 08) Prof. Willi Jaeger:
Reactive Flow Through Porous Media -
Consider flow, diffusion, transport and reactions through porous
media consisting of a fluid and a solid phase. Assuming periodic
media consisting of a fluid and a solid phase. Assuming periodic
structures and a scale parameter \epsilon (measuring e.g. the pore size)
and properly scaled model equations (micro-system) the scale
limit \epsilon\to0 is studied under different assumptions. The aim is to
derive macroscopic laws (macro-system) reducing the complexity of
the model. The main mathematical problem consists in deriving the
proper scale dependent estimates for the solutions to the given
micro-system needed to apply the functional analytic tools of
multi-scale convergence. Here, as usual nonlinear terms are
posing obstacles in passing to the limit. Using localized
coordinate systems and introducing macro- and microscopic
variables, the arising compactness problems can be reduced
essentially. In this seminar, the underlying concepts are
introduced and applied to selected examples. Open problems
arising in important applications will be discussed.
(December 09) Prof. Willi Jaeger:
Multi-scale Modelling in Biosciences- Ion Transport Through Membranes
-
Modelling and simulation of processes in biosciences lead in
general to complex model systems This complexity is caused by the
system size, the arising nonlinearities, the range of scales
involved, the stochastic nature of the processes and of the
underlying geometry. It is a main challenge for Mathematics to
reduce the complexity e.g. by analytic or numerical multi-scale
methods. Already in setting up models these techniques are
necessary to include information about the real processes on
different scales and to link the corresponding models. The effect
of microscale processes on the macro-scale behaviour has to be
analysed. As an important example we consider the transport of
ions through membranes. Mathematical modelling and simulation of
ion concentrations inside and outside living cells, in their
cyto-plasma and their nucleus, separated by membranes, are
decisive for a better understanding of the bio-system "cell". Due
to that fact that there is more and more information available
about the processes on the micro-scale it is necessary to link
model equations on micro-scale to a macroscopic description. It
is important that model data can be computed using micro-scale
information. Here two domains are considered separated by a membrane
perforated by channels placed in periodically distributed cells.
The thickness of the membrane and the diameter of the cells are
of order ?. The transport of ions is modelled by the
Nernst-Planck equations, properly scaled in the channels. Charges
fixed to the channels are included modelling the influence of the
channels and the changes of its conformation. Effective laws for
the ion transport trough membranes are derived performing an
asymptotic analysis with respect to the scale parameter ?. The
effective model consists in the Nernst-Planck equations on both
sides of the membrane together with appropriate transmission
conditions for the ion concentrations and the electric potential
across the cell membrane. These conditions are determined solving
micro-problems for cell problems in the membrane. New methods of homogenization have to be developed and applied in
order to deal with the nonlinear model equations and the
reduction of the membrane to a two dimensional interface.
(December 13) Prof. Alexei Novikov:
A homogenization approach to large-eddy simulation of incompressible fluids -
In the development of large-eddy simulation one makes two primary
assumptions. The first is that a turbulent flow can be categorized by a hierarchy of lengthscales. The second assumption states that the small scales have universal properties, characterized by, e.g. a spectral power law. This motivated a number of physical models that attempt to account for the presence of small scales by suitably modifying the corresponding partial differential equations (PDE), the Navier-Stokes equations. Homogenization theory addresses rigorously the issue of modification of PDE in the presence of small scales. The goal of this talk is to apply homogenization methods to LES modeling of fluid flows. We will start with the phenomenon of eddy viscosity in two-dimensions: in the presence of small-scale eddies the transport of large-scale vector quantities can be accompanied with depleted, and even ``negative" diffusion, when the Reynolds number is sufficiently large.
|