Some Specific Projects

1
Some Specific Projects
US France Young Engineering Scientists symposium

• Modeling and HPC

2
The Life Sciences Session
US France Young Engineering Scientists symposium
3
Life Sciences Session
Common features complex problems (multiphase,
multiphysics, multidomain, multi scale), often
"early stage" in modelization.
4
Life Sciences Session
Common interest Four subjects Soft
tissue/fluid (medical image driven
problems) Contact model/moving
boundaries Concentrate suspension Complex
geometry and (medical...) imaging.
5
Medical image driven problems
Georges Biros, Didier Auroux, Marcela Szopos,
Benjamin Mauroy, Mourad Ismail, Boyce
Griffith. Right now integration of existing
codes in an open source framework
development of benchmarks for arterial flows
sharing image-driven model datasets (when
possible) development of novel parameter
estimation algorithms.
6
Contact model /moving boundaries
Judith Hill, Vincent Martin, Marcela Szopos,
Arnaud Ducrot, Olivier Saut, Benjamin Mauroy,
Boyce Griffith. Right now a discussion of
methods for moving boundaries (we all use
different ones) a beginning of a discussion on
the theory behind the contact problem (what's
the right thing to do).
7
Concentrate suspension
Mourad Ismail, Judith Hill, Olivier Saut,
Benjamin Mauroy. Right now compare different
numerical methods for complex fluids
simulation include some US researchers in our
ANR project "MOSICOB" on numerical
simulation of complex fluids. share
experimental data for validation of numerical
methods.
8
Complex geometry and imaging
Olivier Saut, Boyce Griffith, Arnaud
Ducrot. Right now discuss methods for mesh
generation discuss compare methods for
interpolation of medical data on the mesh (and
image reconstruction).
9
The Algorithm Session
US France Young Engineering Scientists symposium
10
Error estimate-based adaptivity for fluid
structure problems

• Martin Vohralik, Virginie Bonnaillie-Noel, Mourad
  Ismail, Martin Campos-Pinto, Boyce Griffith,
  Sreekanth Pannala

11
Develop and implement a general parallel
adaptive scheme based on local error estimate
for problems involving fluid structure
interactions. Some of the properties desirable
in this scheme area) Optimal geometric and
hierarchical adaptivity based on local errorb)
Load balanced to ensure scalabilityc) Amenable
to implicit fluid structure coupling
12
Numerical methods for singular reaction diffusion
equations arising in population dynamics

• Arnaud DucrotMayya Tokman

13
We want to construct numerical schemes using the
weak formulation of pray-predator models with
Holling-Tanner like interaction. The scheme
will capture the singular behaviour of
thesolutions
14
Inverse problems, parameter estimation and data
assimilation

• Didier Auroux
• George Biros

15
In most life and physical sciences, a crucial
issue in the modelisation process is the
estimation and identification of the model
parameters (or some boundary conditions, or some
unknown terms in the model equations). Inverse
problem and data assimilation techniques (e.g.
optimal control theory, Kalman filters, dual
variational algorithm) allow us to calibrate the
model parameters from real data sets and identify
more precisely the system state. We will combine
Kalman filters with dual variational methods to
explore novel methodologies for large scale data
assimilation. We will conduct numerical
experiments to compare the new methods with the
existing state of the art.
16
Development of Parallel Solvers for Highly
Anisotropic Parabolic Linear Systems Arising in
Resistive MHD and Radiation Transport

• Frederic Magoules
• Daniel Reynolds
• Bronson Messer

17
We consider linear systems arising from highly
anisotropic, parabolic differential equations
relevant to fusion plasmas and astrophysical
radiation transport. We will investigate
parallel Domain Decomposition algorithms on these
problems. Such approaches may promise increased
robustness over multigrid methods for highly
anisotropic and spatially adaptive systems on
such problems.
18
Non-life sciences and Software and
librariesSessions
US France Young Engineering Scientists symposium
19
Parallel in Time and Space Algorithms for Fluid
Mechanics

• F. Magoules
• K. Evans
• G. Staffelbach
• R. Mills

20
Parallel in Time and Space Algorithms for Fluid
Mechanics

• We will investigate computational efficiency
  improvements for computational fluid dynamics
  through an adaptation of the parallel method both
  in time and space. First, development of an
  implicit solver will allow larger time steps on
  relatively coarse grids to create 'seed' values
  along a time dependent run. The seed values allow
  a subsequent refinement of decomposed time
  domains to occur in parallel. An investigation of
  the treatment will be performed to determine the
  feasibility of scalability to 500K processors
  using space and time decomposition.

21
Combined Finite Element and Finite Volume Schemes
for Subsurface Flows

• Martin Vohralik
• Richard Mills
• Sreekanta Parmala

22
Combined Finite Element and Finite Volume Schemes
for Subsurface Flows

• Develop and implement a scalable scheme
  based on combined Finite Element (FE) and Finite
  Volume (FV) method for subsurface flow and
  transport with full anisotropic heterogeneous
  tensor and the following properties
• One unknown/element
• Symmetric Positive Definite matrix
• With proven existence and uniqueness
• General mesh (non-convex, non-matching)
• Local conservation
• Linear
• Discrete maximum principle

23
Efficient Preconditioning Strategies for Neutral
Particle Transport

• Dinesh Kaushik
• Broson Messer
• Laura Grigori
• Julien Salomon

24
Efficient Preconditioning Strategies for Neutral
Particle Transport

• The neutron transport equation is seven
  dimensional (three in space, two in angle, one in
  energy, and the last in time). The discretized
  form of this equation gives rise to massive
  linear systems that need to be solved on
  large-scale parallel machines. In order to do
  this in reasonable amount of time, efficient
  preconditioners are essential. In this
  collaborative effort, we will work on custom
  precondtioners that take advantage of the matrix
  structure. These preconditioners will be applied
  to the astrophysics (neutrino transport) and
  nuclear reactor applications (neutron transport).
  We will also explore the opportunities for
  preconditioning using techniques from
  parallelization in the time dimension.

25
Theoretical Analysis of the Eigenspectrum of the
Dirac Equation

• James Brannick
• Virginie Bonnaillie-Noel

26
Theoretical Analysis of the Eigenspectrum of the
Dirac Equation

• The aim of the project is to analyze the
  properties of the eigenspectrum of the Dirac
  equation of
• QCD. Initially, we propose to study the
  simplified Schwinger model of Quantum
  Electrodynamics
• with a U(1) potential.
• The goals will be as follows
• Conduct theoretical analysis to determine the
  behavior and localization of the eigenfuntions
• Develop a gauge invariant discretization using
  Finite Elements --
• current discretizations are essentially limited
  to finite difference schemes.
• Explore the theoretical results using this
  numerical model.
• Generalize the results obtained for this model to
  the QCD equation with SU(3) gauge.

27
Integrating Adaptive Grids with Nonlinear Solvers
for Problems in Plasma Physics

• Martin Compos-Pinto
• Mayya Tokman
• Daniel Reynolds

28
Integrating Adaptive Grids with Nonlinear Solvers
for Problems in Plasma Physics

• The presence of complex nonlinear
  interactions of multiple spacial and tem- poral
  scales make numerical solutions of equations such
  as Vlasov or MHD a challenging task. To address
  this problem, it is highly desirable to construct
  numerical schemes which integrate efficient
  adaptive approaches to discreti- zations in space
  and time. By combining expertise of French
  researchers in time evolution of adaptive space
  discretizations and American counterparts in
  efficient time integrators for nonlinear systems,
  we plan to investigate pos- sibilities for
  designing innovative numerical methods for
  problems in plasma physics.

29
Exploring Coupling Strategies Using PALM for
Multiphysics Nuclear Reactor Simulations

• Dinesh Kaushik
• Gabriel Staffelbach
• Laura Grigori

30
Exploring Coupling Strategies Using PALM for
Multiphysics Nuclear Reactor Simulations

• Nuclear reactor core simulations require coupling
  among different physics areas such as neutronics,
  thermal hydraulics, and structural mechanics.
  This coupling needs to be accurate (not to
  compromise accuracy from each physics component)
  and parallel (to support large-scale
  simulations). We will explore using PALM software
  for coupling mutiphysics codes from Argonne. PALM
  is developed by the PALM Team at CERFACS
  (http//www.cerfacs.fr/palm/). Various coupling
  approaches will be tested with scalability and
  ease of use in mind. We will also attempt to
  construct accurate interpolation schemes and
  preconditioning techniques designed for the
  coupled systems.