Feb_March_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 | F E B R U A R Y / M A R C H 2 0 2 0 8 MACHINE LEARNING | AI HYBRID METHOD SPEEDS MATERIALS DEVELOPMENT Scientists often use high through- put experimentation to map relation- ships between the compositions, struc- tures, and properties of materials made of varying quantities of the same ele- ments. Although this approach helps accelerate materials development, it usually requires expensive equipment. In contrast, high throughput calcula- tion uses computational models to de- termine a material’s properties based on its electron density. This is faster and cheaper than physical experiments, but much less accurate. Yuma Iwasaki of the Central Research Laboratories of NEC Corp., together with his colleagues in Japan, combined these methods and paired them with machine learning to streamline the process. They tested their approach using a 100-nm thin film made of iron, co- balt, and nickel spread on a sapphire substrate. Various combinations of the three elements were distributed along the film, enabling testing of many sim- ilar materials in a single sample. The re- sulting x-ray diffraction curves provide detailed information about the crys- tallographic structure, chemical com- position, and physical properties. The team then used machine learning to break the data into individual diffrac- tion curves for every combination of the elements. High throughput calculations helped define the magnetic properties of each combination. Further calcula- tions were performed to reduce the dif- ference between the experimental and calculation data. The hybrid approach allowed them to map the Kerr rotation of the elements’ composition spread— without requiring costly high through- put experiments. Knowledge of this property is important for a variety of applications in photonics and semicon- ductor devices. www.nims.go.jp/eng. 3D PRINTER USES MACHINE VISION AND AI A startup out of MIT called Ink- bit is aiming to bring the benefits of 3D printing to products that have never been printed before. Further, the com- pany is planning to do so at volumes that would radically disrupt production processes across many industries. Ink- bit is doing this by pairing its multima- terial inkjet printer with machine vision and machine learning systems. The vi- sion system scans each layer of the ob- ject as it is printed to correct errors in real-time, while the machine learning system uses that information to predict the warping behavior of materials and make more accurate final products. The company says it can print more flexible materials much more ac- curately than other printers. Inkbit cur- rently has one production-grade print- er being used in a pilot project for John- son and Johnson, and plans to start selling printers later this year. The ex- isting printer features 16 print heads, which are used to create multimateri- al parts, and a print block large enough to produce hundreds of thousands of fist-sized products each year. The ma- chine’s contactless inkjet design means increasing the size of later iterations will be as simple as expanding the print block. mit.edu . Kerr rotation mapping of an iron, cobalt, and nickel composite spread using (a) high throughput experimentation, (b) using only high throughput calculation, and (c) using a combined approach. Courtesy of National Institute for Materials Science. Inkbit’s multimaterial 3D printer can output extremely flexible materials. Welcome to Machine Learning/AI, a new department focusing on the explosive growth of these disciplines within the materials science and engineering community. Turn here to learn about the latest research on artificial intelligence and how it impacts computer- aided materials development. BRIEF Noodle.ai, San Francisco, and SMS Group, Pittsburgh, launched an AI-driven application for the steel industry called MPV (Mechanical Properties Variability). MPV uses AI and machine learning to address challenges associated with the variability of mechanical properties. The application senses patterns within mill data and predicts when increased variability will occur. It then recommends the input parameters required to optimize target properties such as yield strength, tensile strength, and elongation. www.noodle.ai .