May/June_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 | M A Y / J U N E 2 0 1 8 9 Future engines made with these new alloys may also reduce carbon dioxide emissions without sacrificing vehicle performance, say researchers. ornl.gov. With this approach, the team was able to determine the microstructures of low-carbon steel at a level of accu- racy not previously possible—93% ac- curate compared with 50% for conven- tional methods. “The newdeep learning methods will not only help us assess the quality of steel more objectively and more accurately, we also anticipate that our results will be transferable to many other production processes and materials,” says Professor Frank Mücklich, who supervised the study. www.uni-saarland.de . Scientists from FCA US, ORNL, and Nemak are using neutrons to compare residual stresses in engine cylinder heads made of different cast aluminum alloys. Courtesy of Genevieve Martin/ORNL. NEW COLLABORATION OPTIMIZES STEEL CLASSIFICATION Using machine learning techni- ques, computer scientists and mate- rials scientists at Saarland University, Germany, developed a classification method that is reportedly much more accurate and objective than conven- tional quality control procedures. First, the materials researchers needed to help the computer scientists under- stand how the internal structures of a material are related to its proper- ties. Likewise, the computer scientists showed the materials researchers how machine learning methods can produce more accurate results than any of the image analyses conducted manually by materials scientists. During the study, which focused on classifying steel microstructures, the team needed to understand how every stage of the steel production process influences the metal’s internal struc- ture. Traditionally, during the material development and quality control stag- es, samples are evaluated using optical and electron microscopy. In the new study, the materials scientists were in- terested in finding an objective pro- cedure that was far less prone to user error and that could be applied irre- spective of the user’s level of expertise. Saarland University’s atomprobe tomography lab.