October_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 | O C T O B E R 2 0 2 0 7 MACHINE LEARNING | AI Researchers at Japan’s National Institute for Materials Science and Toyota Motor Corp. developed a technique that feeds information from aluminum alloy databases into a machine learning model. This trains the model to under- stand the relationships between an alloy’s mechanical properties and its elements, as well as the type of heat treat- ment applied during manufacturing. The model can then predict what is required to manufacture a new alloy with specific mechanical properties. www.nims.go.jp/eng. BRIEF SOFTWARE ASSESSES 3D PRINT QUALITY Scientists at the DOE’s Oak Ridge National Laboratory (ORNL), Tenn., re- cently developed artificial intelligence software called Peregrine for powder bed 3D printers. The program assess- es parts quality in real time, without the need for characterization equip- ment. To devise a control method for surface-visible defects that would work on multiple printer models, the team created a novel convolutional neural network—a computer vision technique that mimics the human brain in quickly analyzing images captured from camer- as installed on the printers. The software uses a custom algo- rithm that processes pixel values of im- ages, taking into account the compo- sition of edges, lines, corners, and tex- tures. If Peregrine detects an anomaly that may affect part quality, it automat- ically alerts operators so adjustments can be made. The software is well suit- ed to powder bed printers, which are popular for producing metal parts. However, problems during printing such as uneven distribution of the pow- der or binding agent, spatters, insuffi- cient heat, and porosities can result in defects at the surface of each layer. “One of the fundamental chal- lenges for additive manufacturing (AM) is that you’re caring about things that occur on length-scales of tens of mi- crons and happening in microseconds, and caring about that for days or even weeks of build time,” says principal in- vestigator Luke Scime. Peregrine is being tested on mul- tiple printers at ORNL. The researchers stress that by making the software ma- chine-agnostic, printer manufacturers can save development time while of- fering an improved product to indus- try. Peregrine produces a common im- age database that can be transferred to each new machine to train new neural networks quickly, and it runs on a single high-powered lap- top or desktop. ornl.gov. MACHINE LEARNING ENABLES REVERSE ENGINEERING Key to the strength and versatility of glass and carbon fiber reinforced composites is the orientation of fibers in each layer. Recent innovations in 3D printing have made it possible to finetune this factor, due to the abil- ity to include printer-head orientation instructions within the CAD file for each layer of the part being printed. How- ever, a research team from NYU Tan- don School of Engineering showed that these toolpaths are also easy to repro- duce—and therefore steal—with ma- chine learning (ML) tools applied to the microstructures of the part obtained by a CT scan. The team showed that the print- ing direction can be captured from the printed part’s fiber orientation via mi- cro-CT scan image. However, since the fiber direction is difficult to discern with the naked eye, researchers used ML al- gorithms trained over thousands of mi- cro-CT scan images to predict the fi- ber orientation on any fiber-reinforced 3D-printed model. The study raises con- cerns for the security of intellectual property in 3D-printed composite parts, where significant effort is invested in development but modern ML methods can make it easy to quickly replicate them at low cost. engineering.nyu.edu. Peregrine software detects an anomaly in a component being made on a powder bed printer. Courtesy of Luke Scime/ORNL. Reconstructed CT scan model of a 3D-printed com- posite part shows overall dimensions and geometry .