Nov_Dec_AMP_Digital

A D V A N C E D M A T E R I A L S & P R O C E S S E S | N O V E M B E R / D E C E M B E R 2 0 2 0 7 MACHINE LEARNING | AI A team led by Stanford University professors recently developed a machine learning method that slashes battery testing times by 98%. Although the group developed their method by testing battery charge speed, they say it can be applied to many other parts of the battery development pipeline and even to non-energy technologies. stanford.edu. BRIEF CREATING MATERIALS WITH AI Researchers at Delft University of Technology (TU Delft), the Nether- lands, developed a highly compressible and strong material by using only arti- ficial intelligence (AI). Miguel Bessa, as- sistant professor of materials science and engineering, got the inspiration for this project during his time at the Cal- ifornia Institute of Technology. In the Space Structures Lab, he noticed a sat- ellite structure that could open long so- lar sails from a tiny package. He won- dered if it would be possible to design a compressible yet strong material that could be squeezed into a small fraction of its volume. “If this was possible, ev- eryday objects such as bicycles, dinner tables, and umbrellas could be folded into your pocket,” he says. “Metamate- rial design has relied on extensive ex- perimentation and a trial-and-error ap- proach. We argue in favor of inverting the process by using machine learning for exploring new design possibilities, while reducing experimentation to an absolute minimum.” Guided by machine learning, Bes- sa fabricated two designs at different length scales that trans- form brittle polymers into lightweight, recov- erable and super-com- pressible metamaterials. The macro-scale design is tuned for maximum compressibility, while the micro-scale is de- signed for high strength and stiffness. “We follow a computational data-driven approach for exploring a new metamaterial con- cept and adapting it to different target properties, choice of base materials, length scales, and manufacturing pro- cesses,” he says. “The important thing is that machine learning creates an op- portunity to invert the design process by shifting from experimentally guided investigations to computationally da- ta-driven ones, even if the computer models are missing some information.” Bessa believes that the most important aspect of the work is not the particular material that was created, but the abili- ty to reach untapped regions of the de- sign space via machine learn- ing. www.tudelft.nl . SEARCHABLE DATABASE FOR LEGACY PARTS Imaginestics LLC, a soft- ware company headquar- tered at Purdue Research Park, West Lafayette, Ind., re- ceived a one-year contract from the U.S. Army’s Red- stone Arsenal, Huntsville, Ala., to make its engineering data in- stantly accessible for part identifica- tion, maintenance, and repairs. The Army base currently has tens of thou- sands of paper blueprints for legacy parts, which were created before the digital design era. Working in partner- ship with a team from the University of Alabama in Huntsville, Redstone Arse- nal is now scanning these physical doc- uments and saving them as PDF files. The data management team will then use the VizSeek visual search en- gine from Imaginestics to scan each PDF and extract data from the engi- neering drawing, creating a database that can be searched by text, attribute, geometry, or a combination of these. Army personnel will be able to instant- ly locate the data by searching the da- tabase. “When there is a critical need to identify a replacement part, this can now be done in seconds instead of hours,” says Jamie Tan, CEO of Imagin- estics. “The benefit to military person- nel at the base and in the field is signifi- cant.” purdue.edu , vizseek.com. Metamaterial created with AI transforms a brittle material into a sponge-like substance. Courtesy of TU Delft. Blueprint from a 1968 commercial aircraft. Courtesy of Imaginestics.