picture: ORNL’s Gremlin software program, proven above in use by CARLA, an open-source simulator for autonomous driving research, is designed to enhance weaknesses in machine studying.
view extra
Credit: CARLA
Researchers on the Department of Energy’s Oak Ridge National Laboratory and their technologies have acquired seven 2022 R&D 100 Awards, plus particular recognition for a battery-related inexperienced know-how product.
R&D World journal has introduced the winners from their collection of finalists who shall be acknowledged on the group’s sixtieth awards ceremony on Nov. 17 in Coronado, California.
Established in 1963, the R&D 100 Awards, dubbed “the Oscars of Innovation,” yearly acknowledge 100 accomplishments in research resulting in new business merchandise, technologies and supplies from world wide notable for his or her technological significance. This 12 months’s wins convey ORNL’s complete R&D 100 Awards to 239 for the reason that award’s inception.
“Each day, ORNL strives to ship scientific breakthroughs for the advantage of society,” ORNL Director Thomas Zacharia stated. “The R&D 100 Awards are an amazing recognition of the onerous work and dedication required by our researchers to develop these impactful technologies.”
Among 9 ORNL finalists, the profitable ORNL researchers and technologies embody:
DuAlumin-3D: An Additively Manufactured Dual-Strengthened Aluminum Alloy Designed for Extreme Creep and Fatigue Resistance, developed by ORNL, General Motors and Beehive3D.
In response to a necessity for extra resilient, light-weight aluminum alloys, ORNL researchers designed DuAlumin-3D, an aluminum alloy with a mixture of tensile, creep, fatigue and corrosion properties superior to all recognized forged, wrought and printable aluminum alloys.
DuAlumin-3D is designed to benefit from the distinctive thermal situations that happen through the laser additive manufacturing course of. The alloy takes its identify from twin strengthening mechanisms: a nanoscale microstructure that varieties throughout printing and precipitates that type upon warmth therapy. Because of those microstructural options, the alloy retains greater than half its power at excessive temperatures of 300 to 315 levels and is steady as much as 400 levels C.
Funding for this mission was offered by the Office of Energy Efficiency and Renewable Energy’s Vehicle Technologies Office and Advanced Manufacturing Office.
Principal investigators for this research embody ORNL’s Alex Plotkowski, GM’s Qigui Wang and Beehive3D’s Jonaaron Jones; GM’s Andy Wang, Devin Hess, Dan Wilson and Dale Gerard; Beehive3D’s Devon Burkle, Rachel Jones and Charles Stansberry; and ORNL’s Amit Shyam, Ryan Dehoff, Allen Haynes, Richard Michi, Sumit Bahl, Ying Yang, Larry Allard, Jon Poplawsky, Bill Peter, Derek Splitter and Jiheon Jun. The University of Tennessee’s Kevin Sisco additionally contributed to the event.
Gremlin: Adversarial Discovery of Weaknesses in Machine Learners, developed by ORNL.
Weaknesses in machine studying know-how can have critical penalties, similar to improperly skilled facial recognition synthetic intelligence yielding inaccurate identification. To enhance machine studying, ORNL researchers developed Gremlin, a studying system designed to establish and tackle the worst-performing neural community characteristic units.
Gremlin identifies issues inside a machine studying system, typically by way of inverting a mannequin’s coaching metrics. For instance, a mannequin could also be skilled to drive a digital autonomous automobile, so a easy coaching metric for that mannequin is perhaps maximizing the size of time earlier than crashing; Gremlin would invert that metric to find situations the place the mannequin crashes the soonest.
The system can then be used to replace the mannequin coaching knowledge with extra examples of these poor performing situations, and the mannequin is retrained utilizing that up to date knowledge.
Gremlin decreases time wanted to deal with machine studying mannequin weaknesses and may be scaled for software from laptop computer computer systems to machines like ORNL’s Summit supercomputer.
A versatile framework bettering upon comparable techniques, the know-how can be utilized on machine studying fashions designed for many any software.
Funding for this mission was offered by the Office of Energy Efficiency and Renewable Energy’s Vehicle Technologies Office and the DOE Office of Science’s Advanced Scientific Computing Research.
ORNL’s Mark Coletti led the event. ORNL’s Robert Patton and Quentin Haas additionally contributed to the event.
RapidCure: High-Speed Electron Beam Processing of Battery Electrodes, developed by ORNL. This know-how additionally acquired the Silver Award within the Special Recognition: Green Tech class.
In typical lithium-ion battery electrode manufacturing, supplies are blended in N-Methyl-2-pyrrolidone, an natural solvent, to type a slurry throughout manufacturing. There are a number of drawbacks to this technique: the solvent is poisonous and explosive, the method requires long-drying ovens and solvents have to be recovered after manufacturing.
To tackle the disadvantages of this manufacturing course of, ORNL researchers developed a cleaner and extra environment friendly technique to fabricate electrodes. A high-speed electron beam basically replaces the long-drying ovens to evaporate the solvent, serving because the vitality supply to chemically polymerize and crosslink small molecules into excessive molecular weight polymers.
Additionally, this know-how produces electrodes quicker — seconds to minutes in contrast with the solvent technique — and reduces the vitality and gear crucial for manufacturing. When the method is full, no recycling unit is required, in contrast to the necessary restoration when utilizing the solvent.
Funding for this mission was offered by DOE’s Office of Energy Efficiency and Renewable Energy.
ORNL’s Zhijia Du led the event. ORNL’s Chris Janke, David Wood and Jianlin Li and Carrier Global’s Claus Daniel additionally contributed.
SolidPAC: A Comprehensive Solid-State Battery Design Tool, developed by ORNL.
Solid-state batteries, or SSBs, are composed of stable electrolytes, versus the liquid electrolytes in lithium-ion batteries. With excessive vitality and energy density ranges, SSBs have the potential to be an efficient method to electrify the transportation sector.
However, the shortage of an current framework for developing SSBs poses a barrier to their financial feasibility.
To overcome this downside, ORNL researchers developed SolidPAC, a standard spreadsheet and graphical consumer interface-based device for analyzing and creating SSB properties. The open supply toolkit contains common design pointers to foretell cell-, module- and pack-level vitality densities based mostly on user-defined parameters for the battery system.
SolidPAC gives particular design rationales for constructing extremely energy-dense SSBs and can assist decide battery metrics wanted for SSBs to turn out to be akin to lithium-ion batteries.
Funding for this mission was offered by DOE’s Laboratory Directed Research and Development program.
ORNL’s Ilias Belharouak led the event. ORNL’s Marm Dixit, Nitin Muralidharan, Ruhul Amin, Rachid Essehli and Mahalingam Balasubramanian additionally contributed to SolidPAC.
Ultraclean Condensing Gas Furnace, developed by ORNL.
Commercial and residential condensing pure fuel furnaces contribute to local weather change by releasing acidic water and dangerous fuel emissions, all able to inflicting long-term hurt to soil, water and air.
To mitigate injury from these pollution, ORNL researchers developed the Ultraclean Condensing Gas Furnace, which makes use of monolithic acidic fuel discount, or AGR, because the catalyst to take away greater than 99.9% of acidic gases and different emissions, similar to carbon monoxide, hydrocarbons and methane, from furnaces.
This results in not solely impartial condensate that’s extremely environmentally pleasant, but additionally ultraclean flue fuel that meets future emissions laws. Neutral condensate allows a less complicated and cheaper furnace design, which yields the next effectivity ultrahigh furnace and a discount in set up prices.
AGR capabilities like a catalytic converter in a automobile, passing the exhaust over metals to scale back acidic gases and pollutant emissions that contribute to international local weather change.
Ultraclean may be built-in into present furnace designs with out altering manufacturing processes and utilized to different gas-driven units like fuel boilers, business pure fuel gear, industrial furnaces and pure fuel water heaters.
Funding for this mission was offered by the DOE Office of Energy Efficiency and Renewable Energy’s Building Technologies Office.
ORNL’s Zhiming Gao led the event. Research contributors included ORNL’s Kyle Gluesenkamp, Kashif Nawaz, Anthony Gehl, Josh Pihl, Dino Sulejmanovic, Tim LaClair, Mingkan Zhang, Lingshi Wang, Van Baxter, Bo Shen, Xiaobing Liu, Jeff Munk and Jim Parks.
Flash-X, a Multiphysics Simulation Software, developed by Argonne National Laboratory, ORNL, Michigan State University, University of Chicago, University of Illinois at Urbana-Champaign and RIKEN Center for Computational Science, Japan.
Flash-X is a extremely versatile software program instrument that makes use of a mixture of partial and strange differential and algebraic equations to simulate several types of bodily phenomena, together with astrophysics, computational fluid dynamics and cosmology.
The know-how is extremely accessible; Flash-X has a efficiency portability layer that’s language agnostic, making it appropriate with quite a lot of pc techniques. The open-source software program options elements in an simply customizable plug-and-play mode for many scientific functions. The configuration of particular functions is split into smaller parts so that every particular person configuration device stays comparatively easy. Flash-X additionally publishes its auditing and high quality management processes and options.
A earlier model of the software program, FLASH, was employed for quite a lot of scientific discovery functions over the previous decade however is now not totally appropriate with state-of-the-art computing techniques and supercomputers, particularly hybrid CPU-GPU techniques just like the Frontier and upcoming Aurora supercomputers at ORNL and ANL, respectively. FLASH was used as a device to show astrophysical ideas, and Flash-X could possibly be employed for instructing functions, as properly.
Funding for this mission was offered by the DOE Office of Science’s Advanced Scientific Computing Research program as a part of the Exascale Computing Project, a joint effort of two DOE organizations, the Office of Science and the National Nuclear Security Administration.
Argonne’s Anshu Dubey led the event. Research contributors included ORNL’s Bronson Messer, J. Austin Harris, Thomas Papatheodore, Eirik Endeve and (*100*) Raphael Hix; Argonne’s Klaus Weide, Jared O’Neal, Akash Dhruv, Johann Rudi, Tom Klosterman, Rajeev Jain, Paul M. Rich and Katherine M. Riley; Michigan State University’s Sean M. Couch; RIKEN Center for Computational Science’s Mohammed Wahib; the University of Illinois’ Paul Ricker; the University of California Santa Cruz’s Dongwook Lee; Google’s Muralikrishnan Ganapathy; California Institute of Technology’s Michael Pajkos; the University of Tennessee’s Ran Chu; Lawrence Berkeley National Laboratory’s Christopher Steven Daley and Katie Antypas; Amazon’s Shravan Kumar Gopal; Nvidia’s John Bachan; and the University of Alabama’s Dean M. Townsley.
GridEye: A Wide-Area Power Grid Real-Time Situational Awareness System, developed by the University of Tennessee and ORNL.
As local weather change causes frequent main climate occasions and energy grids rely more and more on renewable vitality sources, the necessity for larger situational consciousness and occasion monitoring continues to develop.
To shortly detect and supply details about main occasions throughout the North American energy grid, researchers at UT and ORNL developed a monitoring system, GridEye.
GridEye makes use of greater than 300 frequency disturbance recorders — screens that may be put in wherever with an 110V outlet, Ethernet and GPS entry — to gather knowledge on frequency variation throughout the grid. Sudden modifications in frequency point out an uncommon occasion, similar to an electrical generator shutdown.
Within seconds, GridEye can detect anomalies inside the system to pinpoint their areas and the dimensions of the facility loss. The know-how then sends out alerts with electronic mail occasion evaluation reviews that includes incident particulars and placement data to energy firms, grid operators and different stakeholders, so affected events can take the right actions, similar to ramp up further energy era.
GridEye is the primary and solely monitoring resolution for electrical energy grids throughout North America and permits energy firms to see exterior their very own service areas.
Funding for this mission was offered by DOE, the National Science Foundation, Dominion Energy, Tennessee Valley Authority, North American Electric Reliability Corporation and Electric Power Research Institute.
UT-ORNL Governor’s Chair Yilu Liu and UT’s He Yin and Wenpeng Yu led the research. Contributors to GridEye’s improvement embody ORNL’s Thomas J. King Jr. and Lingwei Zhan; UT’s Shutang You, Yi Zhao, Jiaojiao Dong, Yuru Wu, Zhihao Jiang, Xinlan Jia, Wei Qiu, Chengwen Zhang, Chang Chen, Chujie Zeng and Hongyu Li; Dominion Energy’s Matt Gardner; and Electric Power Research Institute’s Lin Zhu.
UT-Battelle manages ORNL for the Department of Energy’s Office of Science, the only largest supporter of fundamental research within the bodily sciences within the United States. The Office of Science is working to deal with a few of the most urgent challenges of our time. For extra data, please go to vitality.gov/science. – Alexandra DeMarco
https://www.eurekalert.org/news-releases/962947
