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 | A P R I L 2 0 2 2 6 MACHINE LEARNING | AI NEW METHOD INSPECTS 3D-PRINTED METAL PARTS Researchers from Nanyang Technological University, Singapore developed a fast and economical imaging technique that can examine the structure of 3D-printed metal parts. Most of these alloys comprise numerous microscopic crystals, which vary in size, shape, and atomic lattice orientation. By extracting this information, the alloy’s properties can be deduced. Traditionally, examining the microstructure of 3D-printed metal alloys involves time-consuming and costly measurements with scanning electron microscopes. The technique created by assistant professor Matteo Seita and his team offers the same quality of information within minutes by using a system comprising a flashlight, an optical camera, and a laptop computer that runs machine learning software designed by the team. First, the metal surface is treated with chemicals to expose the microstructure. Next, the sample is positioned to face the camera and numerous optical images are taken as the flashlight irradiates the metal from many directions. The software then examines the patterns generated by light reflected from the surface of diverse metal crystals and infers their orientation. The whole process takes about 15 minutes. The researchers then applied machine learning to program the software by inputting hundreds of these optical images. Ultimately, the software learned how to forecast the arrangement of crystals in the metal from the images, based on variances in how light disperses off the metal surface. It was then tested to develop a full crystal orientation map. Seita believes the new imaging technique could streamline the certification and quality assessment of 3D-printed metal parts. The technology could be especially useful in applications such as aerospace manufacturing and repair. www.ntu.edu.sg. PREDICTING DIRECTIONDEPENDENT PROPERTIES A machine learning algorithm developedat SandiaNational Laboratories could lead to a faster and more cost-efficient way to test bulk materials for use in automotive manufacturing, aerospace, and other industries. To screen materials such as sheet metal for formability, companies often use commercial simulation software calibrated to the results of various mechanical tests. However, these tests can take months to complete. Cer- tain high-fidelity computer simulations can assess formability in a few weeks, but companies need access to a supercomputer to run them. The Sandia team has shown machine learning can dramatically cut the time and resources required to calibrate commercial software because the algorithm does not need information from mechanical tests—or a supercomputer. The new algorithm is called MAD3, short for Material Data Driven Design. Researchers say the model has been trained to understand the relationship between crystallographic texture and anisotropic mechanical response. Working with The Ohio State University, Sandia trained the algorithm on the results of 54,000 simulatedmaterials tests using a feed-forward neural network. The Sandia team then presented the algorithm with 20,000 new microstructures to test its accuracy, comparing the algorithm’s calculations with data gathered from experiments and supercomputer-based simulations. “The algorithm is about 1000 times faster compared to high-fidelity simulations. We are actively working on improving the model by incorporating advanced features to capture the evolution of the anisotropy since that is necessary to accurately predict the fracture limits of the material,” says Sandia scientist Hojun Lim. sandia.gov. A low-cost imaging system that uses an optical camera and machine learning can analyze the properties of a 3D-printed metal alloy in 15 minutes. Researchers examine data generated by a newmachine learning algorithm.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==