DAKOTA software soon to be available free on web
Weapons designers and analysts often ask themselves questions such as: “What is the best design?” “How safe is my design?” “How much confidence do I have in my answer?” A Sandia-developed software toolkit can help answer these questions and assist engineers in designing anything from components to sophisticated systems.
The toolkit, DAKOTA version 3.0, will soon be on the Web and available for free.
DOE recently granted DAKOTA 3.0 (Design Analysis Kit for Optimization and Terascale Applications) an open-source release under a GNU General Public License. This means any company engineer or university researcher will be able to download DAKOTA and use it to improve their product design or impact their research.
Written in the C++ computer language, DAKOTA provides a flexible interface between the designer’s simulation software and the latest algorithms for optimization, uncertainty quantification, parameter estimation, design of experiments, and sensitivity analysis. Interfaces between DAKOTA and user simulation codes can be developed rapidly. To date more than 20 simulator programs have been interfaced with the software.
One of DAKOTA’s key features is its ability to use parallel computing resources. For example, DAKOTA was recently interfaced with the SALINAS structural dynamics computer code as part of DOE’S Accelerated Strategic Computing Initiative (ASCI) milepost.
“DAKOTA and SALINAS performed a large weapon component design study on 2,560 ASCI Red processors that accomplished in four days what would have taken more than 10 years to complete on a single workstation,” says Mike Eldred (9211), principal investigator.
With its powerful algorithms and ability to manage complex simulations, DAKOTA allows designers to develop virtual prototypes of products that can be modified within the computer to minimize weight, cost, or defects; limit critical temperature, stress, vibration, or other responses; or maximize performance, reliability, agility, and robustness. The result: better designs and reduced dependence on prototypes and testing, which shortens design cycles and lowers development costs.
Sandia researchers have been developing DAKOTA for about eight years. Starting out as a Laboratory Directed Research and Development (LDRD) program, the initial work focused on optimization methods, but has since branched out into uncertainty quantification and other areas. It has been used internally by analysts from Centers 9100, 9200, 15200, 8700, 2100, 2300, and 2500 and externally by researchers at Los Alamos and Lawrence Livermore national laboratories in conjunction with DOE’s ASCI program.
In addition, 15 industrial companies and universities have been granted DAKOTA licenses under the old licensing system.
Now this sophisticated and flexible engineering software will be made available free to virtually anyone who wants it without the hassles of custom licenses.
“Some commercial products exist that allow users to do optimization, but there are a variety of features that make DAKOTA unique — like having the ability to use thousands of processors,” Mike says.
Other unique features include support for surrogate-based optimization, optimization under uncertainty, mixed integer nonlinear programming, and simultaneous analysis and design, all of which are useful tools for real-world engineering applications.
There are several reasons for making DAKOTA readily available. It will encourage collaborations between Sandia, universities, and other research organizations. This will help infuse the latest research in optimization and related areas into DAKOTA.
Also, the public release will give the software more exposure and use. That is beneficial because as more people use the software, they will identify problems and contribute enhancements that can be shared with the user community. This expanded use could extend to commercial software companies as well — several software vendors have expressed interest in using DAKOTA services along with their proprietary software systems.
“The only restriction is that people cannot take the DAKOTA software, change it, and then sell it,” Mike says. “They can, however, design products with DAKOTA and sell their products.”
The expected positive result will be an improved software toolkit that will be even more efficient and capable for supporting Sandia’s mission.
DAKOTA is a flexible software that can be used both on massively parallel computing platforms that have thousands to tens of thousands of processors and on a single workstation or a network of workstations. It has been successfully ported to most common UNIX-based workstations including Sun, SGI, DEC, IBM, and LINUX-based PCs.
Technical manuals have been prepared to help walk DAKOTA users through the software. Also, the DAKOTA team has begun giving tutorial workshops and expects to do more as the software is publicly released.
Many engineers and researchers around the country are already awaiting the open source release. These “friendly customers” will be given the first access to the software and will serve as testing sites.
“We’ll let them download it and help us work out the kinks in the new process,” Mike says. “Then when we feel completely comfortable with it, we’ll open it up for total public access.”
Download information is available here , and the DAKOTA team can be contacted via e-mail at firstname.lastname@example.org.