High-Risk, Big Payoffs from DOE’s ASCR Leadership Computing Research Program
The Department of Energy's Advanced Scientific Computing Research (ASCR) program has a reputation for supporting high-risk, high-payoff computer simulations.
Through its ASCR Leadership Computing Challenge (ALCC), ASCR offers up to 30 percent of its computational resources at NERSC (National Energy Research Scientific Computing Center) and the Leadership Computing Facilities at Argonne and Oak Ridge national labs to scientists and engineers who are investigating cutting edge technologies that support DOE's energy mission.
Specifically ALCC supports researchers who are:
Advancing DOE's clean energy agenda.
Understanding the environmental impacts of global energy systems.
Reacting to natural and man-made disasters including hurricanes, earthquakes and pandemics.
Explore new frontiers in physical and biological sciences.
In addition to government and academic researchers, awards are made to scientists and engineers from industry, including leading manufacturers.
All three facilities have some very impressive hardware to offer. Included is Oak Ridge's Jaguar, which, until the Chinese made their move, headed up the TOP500 supercomputer list. The Cray XT5 was the first petaflop machine available for open scientific research.
Dave Goodwin, who became the ALCC program manager earlier this year, is the perfect choice to head up an effort that welcomes high-risk ventures. Not only does Goodwin have the scientific and engineering chops that the job requires — he's trained as a particle physicist — but he also has a daredevil streak. One of his hobbies was flying acrobatics in a propeller driven Czech Zlin, a nimble little aircraft specially designed to cut intricate patterns in the sky.
Pushing the Envelope
"Our intent with ALCC is to provide high performance computing (HPC) time for ambitious proposals that require millions of processor hours," Goodwin says. "We're specializing in areas where the technical risk is high — these are projects that might not be funded in other, more conservative applications programs. And we are particularly interest in industrial involvement, more so than many other programs granting time on their supercomputers."
Goodwin says that his program differs in another important respect. Most agencies allocate two months between the time the grant availability is announced and the submission deadline. ALCC allows researchers five and a half months to hone their proposals. "Competition for these large blocks of time is quite intense (the average award is 19 million hours)," he explains. "Because these projects are usually quite complex, this gives the researchers more time to prepare a really good proposal."
In addition to a variety of academic projects, this year's awards include four manufacturers: Boeing, Ramgen Power Systems, the United Technology Research Center and Pratt & Whitney, and General Electric. All of them have been awarded time on Jaguar at Oak Ridge.
Comments Suzy Tichenor, director of the Industrial Partnership Program for the Computing and Computational Sciences Directorate at Oak Ridge: "ALCC is another terrific pathway for companies to apply for time on Jaguar and participate in our Industrial HPC Partnership Program. We're very excited that this year four manufacturers were selected through the ALCC process — and that they represent a mix of small and large corporations. Each of these companies is pursuing challenging science research that will help them develop new products to enhance their competitiveness in the global market."
The Favored Four
Here's a brief summary of their investigations in cooperation with ALCC and Oak Ridge:
Boeing: Designing and Analyzing High Lift Take-Off and Landing Configurations
In the submission to ALCC, Boeing's principal investigator on this project, John Bussoletti, notes that current methods for designing and analyzing commercial aircraft in high lift take-off and landing configurations rely heavily on wind tunnel tests. "These costly tests can take months of flow time to perform, and include the building of very complicated wind tunnel models," he says.
There is plenty of incentive to replace much of the wind tunnel testing with full-scale simulations using computational fluid dynamics (CFD).
Boeing already has a track record conducting these kinds of investigations. In 2010, access to Jaguar under Oak Ridge's INCITE (Innovative and Novel Computational Impact on Theory and Experiment) program, allowed Bussoletti's team to test and verify the accuracy, scalability, sensitivity to grid density, and robustness of a CFD code to assess aircraft performance in take-off and landing configurations. In particular, Jaguar with its 16GB of memory and 12 cores per compute node, enabled them to carry out CFD simulations using up to 50 million grid points, an effort that required only a small fraction of the total system resources.
"We have found that our code scales extremely well, scaling nearly linearly," Bussoletti reports. "We have recently demonstrated the ability run our code analyzing a takeoff configuration in as little as two hours. When fully validated, such a capability could allow us to make radical changes to our wing design process."
So far most of the team's runs at Oak Ridge have been with fixed grids. While this has confirmed that their solver technology holds up when tackling large problems — e.g., up to 50 million vertices or 300 million degrees of freedom — the team is using this year's ALCC award to develop an adaptive grid capability in three dimensions in order to move them toward their goal of fully analyzing high-lift configurations during takeoff and landing under a wide variety of conditions.
Ramgen Power Systems: Developing CO2 Compressing Technology
Sequestering CO2 is a way to help curb global warming and one of the primary — and highly expensive — ways to achieve this is by compressing large volumes of the gas at high temperatures. DOE is looking for novel concepts to accomplish this task more efficiently and at a far lower cost.
Ramgen's shock compression technology uses a supersonic aerodynamic design techniques that is expected to dramatically decrease the size, cost and power required for compressing CO2. In the proposal, principal investigator Allan Grosvenor notes: "This promising breakthrough in turbomachinery efficiency and cost is achieved by combining aerodynamics and mechanical engineering in new ways. To optimize the new technology, very high grid resolution is required to capture the 3D shock distributions and accurately model the resulting compressible viscous flow behavior." Obviously a job for CFD and Jaguar.
And there is a bonus: advancing the Ramgen shock wave technology also promises a major breakthrough in small engine performance and cost. The engine will help reduce greenhouse gases as support for renewable technologies and by using dilute methane as fuel at industrial sites.
United Technologies Research Center (UTC) and Pratt & Whitney: Large Eddy Simulation for Turbomachinery
UTC's Gorazd Medic, principal investigator, writes "This project will explore the application of large-eddy simulation to realistic three-dimensional turbomachinery configurations. Case studies will focus on the centrifugal compressor and low-pressure-ratio axial fan."
Centrifugal compressors are a key technology in many UTC products, including commercial chillers, aircraft environmental control systems, aircraft auxiliary power units, fuel pumps, and in rotorcraft, and business jet engines. Developing an efficient low-pressure-ratio axial fan is a key component for a geared turbofan architecture that Pratt & Whitney has developed for its next generation of commercial jet engines. These engines will have a substantially lower fuel burn than traditional designs.
The processing power of Jaguar is allowing UTC and Pratt & Whitney researchers to use large-eddy simulation to model turbulent flows in the turbomachinery and move beyond the challenges that have hampered standard turbulence models widely used in industry.
Says Medic, "Our ability to better understand the flow physics in these specific product components using wall-resolved large-eddy simulation on grids of sufficient resolution will help lead future design improvements and will be critical to improved energy efficiency of our product going forward."
General Electric: Non-icing Surfaces for Cold Climate Wind Turbines
It's cold and windy high up in the mountains of Northern and Central Europe, North America and Asia. A mixed blessing: the wind makes these areas perfect for electricity-generating wind turbines, but the cold makes it very difficult to operate these machines effectively.
GE principal investigator on this study, Masako Yamada, notes, "Ice buildup on wind turbine blades in cold climates (temperatures of <20C) drastically reduces the efficiency of power generation, often requiring turbine shutdown."
In fact, even a small layer of ice on the wind turbine sensors can result in 30 percent underestimation of the wind speed, resulting in energy losses. And, of course, wind turbines are not the only applications effected by ice buildup. Aircraft, oil and gas rigs, heat exchangers, transmission lines, buildings and boats are just a few of the other applications that would welcome an effective, low cost solution.
Many ice mitigation strategies are based on active de-icing processes requiring additional energy to generate heat. It follows that passive strategies such as ice resistant coatings, are a more desirable solution.
GE Global Research facilities have been investigating passive coatings at the molecular level. However, the problem is so complex that its in-house HPC systems can only handle limited simulations comprising about 1000 molecules with a total dimension of a few nanometers.
GE is using molecular dynamics (MD) on Jaguar to gain a far better understanding of icephobicity on a molecular level. Yamada's team is probing the early stages of ice formation on a variety of surfaces with varying degrees of hydrophobicity to understand how they hinder the formation of ice.
Short and Long Term
For 2012, Goodwin says they have made no major changes in the program, but have expanded the emphasis on emergency preparedness and are looking for new directions in the physical and biological sciences.
In the long run, the drive to exascale computing will make time on these DOE supercomputers even more desirable as researches explore scientific avenues that we cannot even conceive of today. But in the meantime, the constantly upgraded petascale supercomputers at NERSC, Argonne and Oak Ridge will continue to be available through ALCC for those willing to take high risks in hopes of a high reward.
For more about the program and the awards, click here.