2004 Newsnotes
Sandia Software to Help
Design New Commercial Jet
The Boeing Company
plans to use the Emu code, developed in SNL Center 9200, to help model
structural damage tolerance in the new 7E7 commercial aircraft. This code
is based on
the peridynamic model of solid mechanics, a Sandia-developed mathematical
technique that predicts crack growth with greater ease and generality than
previously
possible. Under Sandia’s CRADA with Boeing, SNL staff member Stewart
Silling (9232) spent the summer of 2004 as a Visiting Scholar at Boeing
Phantom Works in the Seattle area. Resulting progress included adaptation
of Emu to
modeling crack growth in strongly anisotropic composite structures such
as the proposed 7E7 wings and fuselage. During a recent review meeting
at Boeing,
this modeling technology was immediately recognized by 7E7 team management
as providing a new capability for design and interpretation of structural
tests on the new aircraft, which is making extensive use of composite materials.
Applications of the code to the 7E7 program include prediction of the residual
strength of composite structures following damage due to impact by runway
debris,
hail, bird strike, and ground vehicles. The FAA requires aircraft manufacturers
to demonstrate the resistance of new aircraft to threats of this type.
(Contact: Stewart Silling)
December
6, 2004 LLT meeting
Emu prediction of crack
growth in a strongly anisotropic composite panel.
Colors indicate material damage level.
ALEGRA
A multi-disciplinary team of code developers, material modelers, and analysts
is working to deploy advanced earth penetration algorithms in the ALEGRA
finite element shock physics code. Their efforts are focused on developing
predictive modeling capability for penetration through complex, highly
variable geologic targets. Three ASC Level III Milestones were completed
by the ALEGRA development team at the end of FY04 that result in a much
tighter, more consistent coupling between the penetrator and target, leading
to more accurate solutions and better performance. These efforts are the
basis of application of ALEGRA to simulations defined in an ASC Level II
Milestone for FY05.
An ALEGRA development team led by Kevin Brown (formerly 9231) and Mike Wong
(9231) implemented an advanced isotropic geologic material model being developed
by Arlo Fossum and Rebecca Brannon (both 6117) that accounts for shear strength,
damage, jointing, micro-cracking, pore collapse, and rate dependence. Dave
Hensinger (9231) has developed and implemented a multi-block structured Eulerian
mesh capability in ALEGRA that provides an efficient capability to model large
target regions. As these new capabilities have come on-line, Tom Voth (9231)
has led an ASC V&V Advanced Deployment project with analysts Luba Kmetyk
and Joe Bishop (both 9127) to compare the new penetration capabilities to a
variety of experiments, ranging from small laboratory tests of steel penetrators
on aluminum targets to large-scale tests on concrete and limestone targets.
Simulation of the EQ test on Sandia’s sled track is planned for FY05.
For the FY05ASC Level
II Milestone, ALEGRA developers in 9231 and ALEGRA analysts in 9127 will
collaborate on simulations designed to demonstrate the capability of ALEGRA's
advanced penetration algorithms to simulate the impact and penetration
of a large, 5000-lb penetrator into geologic materials, with varying angles
of obliquity (normal impact to 20 degrees). The objectives are to demonstrate
an ability to (1) predict deceleration histories and mechanical loading
experienced by a penetrator body for both normal and oblique impacts; (2)
treat surface effects, including fracture, on crater formation and trajectory;
and (3) demonstrate the effects of a major rock fault on the penetrator
trajectory and loads.
The ALEGRA penetration
capabilities demonstrated to date and maturing in FY05 and beyond will
give Sandia an increasingly accurate, robust tool for high-fidelity, predictive
analyses of the penetration of projectiles into geologic targets of interest
for Hard, Deeply Buried Target defeat and other DOE/DoD applications. (Contacts: Kevin
Brown and Mike
Wong)
November
1, 2004 LLT meeting
Computer & Information
Sciences (CIS) Technical Review
The Technical Review
of the Computer & Information Sciences (CIS) S&T Council was
held on 18-19 August, on site in the RMSEL Auditorium with the out-briefing
by the external panel in the IPB on 20 August. In response to Pace
Vandevender’s request, the Center 9200 (CCIM) review was expanded
this year to include Center 8900’s R&D arm, which with CCIM
constitutes the CIS Council.
The review panel was
again chaired by Prof. Michael Levine, Director of Pittsburgh Supercomputing
Center and Professor of Physics, Carnegie Mellon University. Also on the
panel were:
Dr. Evi Dube, Division Leader, DNT & PAT Computing Applications Division,
Computing Applications and Research Department, Lawrence Livermore National
Laboratory;
Prof. A. B. (Barney) Maccabe, Professor of Computer Science, University of
New Mexico; Interim Director, UNM Center for High Performance Computing;
Prof. Marc Snir, Professor and Chair of Computer Science, University of Illinois
at Urbana-Champagne;
Prof. Danny Sorensen, Professor of Computational and Applied Mathematics, Rice
University;
Prof. Jacob Fish, Professor of Civil, Mechanical, and Aerospace Engineering,
Rennselaer Polytechnic Institute.
Dr. Marshall Peterson, J. Craig Venter Science Foundation, was unable to attend
due to hurricane Charley.
Agenda Highlights:
- 3 Sessions – Applications
Development; Algorithms and Enabling Technology; (Computer) Systems.
- Two of 13 scheduled
CIS presentations by 8900 staff members.
- Customer Perspective
presentation by Tom Mehlhorn (1674), “High Energy Density Physics”
Outcome Highlights (adapted
from Panel’s out-briefing*):
- Overall the panel
was impressed and very positive about the work that was presented, including
its mission relevance, technical direction, and awareness and involvement
with the R&D community external to Sandia.
- Red Storm stands to
be highly successful and is seen as a critical contribution shaping the
future of high performance computing (HPC) in the U.S. In Red Storm,
Sandia is providing leadership in HPC. It is to be congratulated for,
again, succeeding in creating a production computing system in the face
of the vagaries of the supercomputing market and for establishing a close,
long-term collaboration with one of the few extant supercomputer vendors.
It is critical to preserve and enhance this ability.
- The panel expressed
concern about the potential for pressures towards short term results
to adversely impact the ability of CIS, and Sandia as a whole, to attract
and retain new, talented technical staff. They urged that attention be
given to several pairs of opposing forces: Short-term vs. long-term projects,
high-risk vs. low-risk; technology push vs. customer pull. They note
that it is vital to CIS and Sandia’s ability to respond properly
to unforeseen needs that we give due attention to this issue of ‘refreshing
our stock of seed corn.
- The collaboration
between 9200 and 1600 (Pulsed Power Sciences Group), from managers down
to individual contributors, impressed the Panel. They view this as another
example of the value in vertical integration of effort, from system and
software designers to the ultimate customers - a Sandia’s strength.
The CIS’s R&D strategic portfolio in applications and algorithm
development has strengthened High Energy Density Physics.
- It is clear that the
work of the algorithms and enabling technologies has been incorporated
into the applications efforts both within and outside of CIS. This "buy
in" to the technology produced by this group is unusual and is a
clear indication of success. Based on Mehlhorn’s customer perspective
there is need for increased investment in solver technology and adaptivity.
To do this and to address the needs of other organizations, there is
a need for adding skilled staff (internal or external hires).
- The panel encouraged
CIS leadership in promoting collaborations with Engineering Sciences.
Examples this year show that this can be very productive and has potential
to influence national practice. (Contact: John
B. Aidun)
November
1, 2004 LLT meeting
Article
on Quantum DFT Calculations Accepted for Publication by "Modeling
and Simulation in Materials Science and Engineering"
A multi-Center team of authors, led by Ann Mattsson (9235), has had a
significant review article on quantum density functional theory (DFT)
calculations accepted
for publication in the peer-reviewed journal “Modeling and Simulation
in Materials Science and Engineering.” The charter for this invited article
was to provide practical guidance in the use of, and cautionary tales in the
misuse of DFT calculations applied to materials science problems. The invitation
recognizes Sandia’s growing expertise and leadership in this emerging
technology, and brings greater visibility to our growing capabilities in
DFT applications and theory at Sandia.
The article by Ann, Peter
Schultz (9235), Michael Desjarlais (1674), Thomas Mattsson (1674), and
Kevin Leung (1116) is entitled “Designing meaningful density functional
theory calculations in materials science – A primer.” Performing
DFT calculations that provide useful simulations of physical phenomena
and chemistry in materials (as opposed to molecules or clusters of atoms)
requires careful assessment of the problem and carefully chosen manipulations
of intricate computer codes. Well-constructed DFT calculations can have
significant predictive capability. However, blind applications of DFT codes
can give meaningless and misleading results. This article addresses issues
that confront a user of DFT methods, providing practical guidance to overcoming
many of the common challenges in calculation construction, and presents
multiple cautionary examples of how poorly designed DFT calculations fail
to give accurate predictions of material properties.
At Sandia, as in the
wider materials research community, quantifying the chemistry of individual
atoms is becoming more critical to predicting the macroscopic response
of materials. Consequently, DFT methods are playing an increasingly important
role. This extends to their application to stockpile issues such as radiation
effects in electronics (e.g. QASPR, ELDRS), equations of state of metals
and hydrogen, including the conductivity and opacity of plasmas in the
Z machine, and both neutron tube manufacture and aging. DFT is also an
enabling capability for nanotechnology. Sandia has developed considerable
expertise in DFT theory aimed at improving its accuracy (http://dft.sandia.gov/functionals/),
in two complementary, large-scale computational codes, SeqQuest (http://dft.sandia.gov/Quest/)
and Socorro (http://dft.sandia.gov/Socorro/), and
in applications in several Centers including 9200, 8700, 6100, 1800, 1600,
and 1100. (Contact:
Ann
Mattsson)
November 1, 2004 LLT meeting
Using ASCI Red
to Address DP Issues
Sandia is using ASCI Red
to collect scalability data on several key applications that address a number
of high priority Defense Programs issues: MEMS design and fabrication analysis,
abnormal facilities environments, neutron generator tube design, simulation
capability to enable SPR Retirement, and high energy density physics analysis
and design of Z-Pinch experiments. This data and application performance
modeling analysis will be used to establish a technical foundation for the
1Q-FY05 ASC Level 1 Milestone to authorize the ASC program to pursue Petaflop
scale computing and beyond. ASCI Red appeared in the number 1 position on
the Top500 supercomputer list from December 1996 to November 2000, when it
was surpassed by ASCI White. But even today, 4 years after giving up the
title of "World's Fastest," ASCI Red is still the "World's
Largest," not in physical size, but in logical size, based on total
number of compute processors. As such, among the Tri-Lab ASC capability systems,
ASCI Red remains a unique resource for its ability to collect scalability
data, and as a platform for testing and improving the performance of applications
on large parallel systems, such as Red Storm, White and Q.
(Contact: Jim Ang)
September
27, 2004 LLT meeting
Petaflops,
NASA, Supercomputing, and Red Storm
1.
Our proposal to DOE Office of Science for runtime systems software for petaflops
computing is reportedly approved for 3-year funding starting FY05. Our collaborators
are Caltech and the University of New Mexico. This, along with the continuing
funding for scalable systems software, increases DOE Science funding to over
$1M per year for systems software.
2. We have been selected
by NASA to submit a full proposal for a ground based supercomputing system
for design and analysis of ambitious NASA missions.
3. We are hosting a Workshop
on the future of supercomputing in October 2004 in Santa Fe. Our continued
research into future supercomputing suggests that there will be increasing
needs. These include enhancing confidence in the results of our simulations
of the nuclear stockpile stewardship, earthquake prediction, and climate
modeling. The answers with high confidence may require zettaflops (10**21
floating point operations per second) capability.
4. With the upcoming
Red Storm, Sandia has re-joined the Cray User Group, and is hosting its
annual conference in May 2005 in Albuquerque. (Contact: Neil
Pundit)
September 13, 2004 LLT meeting
New Collaboration with
the Stanford Linear Accelerator Center
The Zoltan project is beginning a new collaboration with the Stanford Linear
Accelerator Center, funded by the SciDAC program of DOE's Mathematical, Information
and Computational Sciences (MICS) office. The collaboration will examine parallel
data partitioning methods for specialized geometries and communication-dominated
applications. Zoltan is a toolkit of parallel dynamic load balancing and data
management services for dynamic, adaptive, and unstructured applications. Zoltan
is used in many of Sandia's simulation codes, including SIERRA, ACME, ALEGRA/NEVADA,
ChemCell, CHISELS, Xyce, Trilinos, and CTH-AMR. Because of its open-source
distribution and web-based release, Zoltan is also used widely in other DOE
facilities, universities, and industry. (Contact: Karen
Devine)
August 2, 2004 LLT meeting
DAKOTA
DAKOTA Optimization and Uncertainty Quantification Toolkit version 3.2 is being
released June 2004 by Department 9211. DAKOTA is an open-source software
framework used throughout the Tri-Labs to answer fundamental engineering
questions such as "What is the best design?" and "How
reliable is it?" DAKOTA closes the design and analysis loop by automatically
driving Sandia's high performance computing simulation codes (e.g. Alegra,
MPSalsa and Xyce) through ranges of design and uncertain parameters for
complex system studies. DAKOTA v3.2 deploys new optimization, sampling,
and statistical analysis algorithms we've researched over the last year
to optimize and assess the reliability of weapons components, such as
the recent W80 NEP pad parameter sensitivity study. DAKOTA has over 1600
registered installations in the broader community, and can also be used
for applications such as finding minimum energy states for protein folding
problems and least-squares fits for calibrating simulation codes' internal
parameters. (Contact: Scott A. Mitchell <http://endo.sandia.gov/DAKOTA/>)
July
12, 2004 LLT meeting
Trilinos Wins R&D 100 Award
Researchers at Sandia National Laboratories received a 2004 R&D 100
award for the development of the Trilinos framework to facilitate the
design, development, integration and ongoing support of mathematical
software libraries. Every year since 1963, R&D Magazine has showcased
the 100 best ideas in industrial and technical innovation through the
annual awards program known informally as "the Oscars of Invention." Trilinos
is an object-oriented software framework that provides numerical solver
functionality for the solution of large-scale, complex multiphysics engineering
and scientific applications. Figure 2 (next page) shows the structure
of Trilinos, including the major packages and capabilities that are currently
available.
Trilinos
has been under development at Sandia for the past 4 years, and
has had a major impact on Sandia’s modeling and simulation
capabilities during the past several years by providing uniform
access to accurate, robust and efficient solvers and tools. It
also facilitates more rapid development of new libraries by providing
important core functionality and software engineering processes
for developers. Trilinos unifies a diverse collection of libraries
that have been developed at Sandia, as well as tools developed
by other researchers. Trilinos has also been the development framework
for fundamental algorithmic advances in nonlinear solvers and continuation
methods, time integration methods, eigenvalue solvers and multi-level
preconditioners. Trilinos was released externally for the first
time in September 2003 under an open-source license and has been
eagerly adopted by academia, industry and other laboratories. Trilinos
4.0 was released in June 2004. (Contact: Michael
A, Heroux)
July
12, 2004 LLT meeting
|
Figure 1: The 2004 R&D 100 Award identifies Trilinos as one of the
100
most technologically significant
products introduced this year.
|
|
Figure 2. The Trilinos framework provides access to the wide range of solver
capabilities shown above.
|
Security of Nation's
Water Distribution Networks
A cross-disciplinary team of Sandia's computational scientists, geohydrologists,
and facilities experts are teaming with the Environmental Protection Agency
(EPA) to apply science and technology to improve the security of the Nation's
water distribution networks. The EPA recently signed an Inter-Agency Agreement
with Sandia for $1.5M. The goal is to develop algorithms to determine the optimal
placement of sensors within a municipal water network, and, once these sensors
detect a chem/bio attack, algorithms to determine the location and concentration
of the attack, and optimal damage mitigation and clean-up strategies. It is
also necessary to validate these methods to determine our level of confidence
in the results, and determine the overall level of risk any given municipality
is subject to. Timothy Oppelt, Director of the EPA National Homeland Security
Research Center, wrote "This is an excellent opportunity to leverage the
analytical and computational skills at Sandia with EPA's long history in protecting
water resources." The Computational Algorithms for Water Homeland Security
team effort is part of a larger effort including air security, with additional
funding from the Department of Homeland Security (DHS) and other federal agencies. (Contacts: Scott
A. Mitchell, Ray E. Finley)
June
21, 2004 LLT meeting
Spectral
Element Atmospheric Model (SEAM)
Sandia
is working in collaboration with the National Center for Atmospheric Research
on the development of a Spectral Element Atmospheric Model (SEAM). We have
recently demonstrated that SEAM has excellent scalability on ASCI Red, running
on 8938 CPUs with over 80% parallel efficiency. This is the largest (in terms
of processor count) and one of the highest resolution atmospheric model runs
to date. We have used the performance of SEAM on ASCI Red to estimate its
performance on the 10,000 processors of Red Storm. On Red Storm, we will
be able to perform global simulations of the Earth's atmosphere with an average
grid spacing of 10km, running at close to 40 times faster than reality. This
will be the first time it is possible to obtain 10km global resolution on
an MPP supercomputer. (Contact: Jennifer
Nelson)
June
7, 2004 LLT meeting
Return
to recent newsnotes
|
