February AMP_Digital

1 0 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 | F E B R U A R Y / M A R C H 2 0 1 9 “Magnetic signals provide a wide range of possible national security ap- plications,” says R&D engineer engi- neer David Mascarenas. “It’s a promis- ing phenomenon that we hope to lever- age to uniquely identify different pieces of artillery.” The researchers applied Barkhau- sen noise, a magnetic phenomenon, to two types of steel—conventional steel and an abrasive-resistant version used in mining equipment. A sensor measured electromagnetic signals by repeatedly scanning the different steels over a period of time. They then com- pared the signals from those two sets of scanned images and found signa- tures that were intrinsic to each type. The variations that occur from the pro- duction of various kinds of steel are reflected as distinct fingerprints and appear to be repeatable. Those intrinsic signatures could help identify counter- feit or low-grade steel parts in construc- tion by looking for differences in the electromagnetic signatures. lanl.gov. TESTING | CHARACTERIZATION ANALYZING ELECTRON MICROSCOPY WITH DEEP LEARNING A significantly faster way to find defects in electron microscopy images is now possible. The Multinode Evo- lutionary Neural Networks for Deep Learning (MENNDL) creates artificial neural networks that tease defects out of dynamic data. It runs on all available nodes of the Summit supercomputer, performing 152 thousand million mil- lion calculations per second. In only a few hours, scientists us- ing MENNDL created a neural network that performed as well as a human expert. It reduces the time to analyze electronmicroscopy images by months. MENNDL is the first known approach to automatically identify atomic-level struc- tural information in scanning transmis- sion electron microscopy data. MENNDL creates and evaluates millions of networks using a scalable, parallel, asynchronous genetic algo- rithm augmented with a support vec- tor machine to automatically find a superior deep learning network topol- ogy and hyper-parameter set. For the application of electron microscopy, the system furthers the goal of better understanding the electron-beam-mat- ter interactions and real-time image- based feedback, which enables a huge step beyond human capacity toward nanofabricating materials automati- cally. energy.gov. STEEL FINGERPRINTS HELP IDENTIFY TAMPERED PARTS Researchers at Los Alamos Nation- al Laboratory (LANL), N.M., are using magnetic signals to find distinguish- able “fingerprints” on steel. These innate signatures could help to verify weapons treaties and reduce the use of counter- feit bolts in the construction industry. David Mascarenas of LANL uses Barkhausen noise to find unique “fingerprints” in steel. Courtesy of Furhana Afrid/LANL. Teledyne Technologies Inc., Thousand Oaks, Calif., and Roper Technol- ogies Inc., Sarasota, Fla., announced that Teledyne will acquire Roper’s scientific imaging businesses for $225 million. These businesses include Princeton Instruments, Photometrics, Lumenera, and others, and provide a range of imaging solutions for life sciences, academic research, and cus- tomized OEM industrial imaging solutions. teledyne.com , ropertech.com . BRIEFS Laboratory Testing Inc. (LTI), Hatfield, Pa., expanded its scope of Nadcap accreditation with the addition of Z2 low stress grinding and polishing for fatigue speci- men preparation. LTI’s fracture mechanics lab performs high- cycle fatigue, low-cycle fatigue, fatigue crack growth, and fracture toughness testing services. labtesting.com . The same image shown using different analysis methods: a) raw electron microscopy image; b) defects (white) as labeled by a human expert; c) defects (white) as labeled by a Fourier transformmethod; and d) defects (white) as labeled by the optimal neural network. Defects that do not exist are shown in purple, and defects that were not identified are orange. Courtesy of ORNL.