Photonics

New imaging method to assess quality of 3D-printed metal parts

Low-cost system uses an optical camera and machine learning

21.03.2022 - Scientists from NTU Singapore have developed a fast imaging system to analyze the inner structure of 3D-printed metal parts.

Most 3D-printed metal alloys consist of a myriad of microscopic crystals, which differ in shape, size, and atomic lattice orienta­tion. By mapping out this information, scientists and engineers can infer the alloy’s properties, such as strength and toughness. This is similar to looking at wood grain, where wood is strongest when the grain is continuous in the same direction. This new technology, made by researchers of the Nanyang Techno­logical University, Singapore, could benefit the aerospace sector, where low-cost, rapid assessment of mission critical parts – turbine, fan blades and other components – could be a game­changer for the maintenance, repair and overhaul industry.

Until now, analyzing this micro­structure in 3D printed metal alloys has been achieved through laborious and time-consuming measure­ments using scanning electron micro­scopes. The method designed by Matteo Seita and his team, provides the same quality of information in a matter of a minutes by using a system consisting of an optical camera, a flashlight, and a notebook computer that runs a pro­prietary machine-learning software developed by the team – with the hardware costing about S$ 25,000.

The team’s new method first requires treating the metal surface with chemicals to reveal the micro­structure, then placing the sample facing the camera, and taking multiple optical images as the flashlight illuminates the metal from different directions. The software then analyses the patterns produced by light that is reflected off the surface of different metal crystals and deduces their orienta­tion. The entire process takes about 15 minutes to complete.

“Using our inex­pensive and fast-imaging method, we can easily tell good 3D-printed metal parts from the faulty ones. Currently, it is impossible to tell the difference unless we assess the material’s micro­structure in detail,” explains Seita. “No two 3D-printed metal parts are created equal, even though they may have been produced using the same technique and have the same geometry. Conceptually, this is akin to how two otherwise identical wooden artefacts may each possess a different grain structure.”

Seita believes that their inno­vative imaging method has the potential to simplify the certi­fication and quality assessment of metal alloy parts produced by 3D printing. One of the most commonly used techniques to 3D print metal parts uses a high-powered laser to melt metal powders and fuse them together, layer by layer, until the full product is printed. However, the micro­structure and thus the quality of the resulting printed metal depends on several factors, including how fast or intense the laser is, how much time is given for the metal to cool before the next layer is melted, and even the type and brand of metal powders used.

This is why the same design printed by two different machines or production house may result in parts of different quality. Instead of using a complicated computer programme to measure the crystal orienta­tion from the optical signals acquired, the smart software uses a neutral network. The team then used machine learning to programme the software, by feeding it hundreds of optical images. Eventually, their software learned how to predict the orientation of crystals in the metal from the images, based on differences in how light scatters off the metal surface. It was then tested to be able to create a complete crystal orientation map, which provides compre­hensive information about the crystal shape, size, and atomic lattice orientation.

To commercia­lize their method, the team is now in discussion with NTUitive, NTU’s inno­vation and enter­prise company, to explore the possi­bility of starting a spin-off company or to license their patent to interested industry players. (Source: NTU)

Reference: M. Wittwer & M. Seita: A machine learning approach to map crystal orientation by optical microscopy, npj Comput. Mat. 8, 8 (2022); DOI: 10.1038/s41524-021-00688-1

Link: Seita research group, School of Mechanical and Aerospace Engineering, Nanyang Technological University, Singapore, Singapore

Top Feature

Vote now at the inspect award 2022 and choose the best machine vision products.

Vote Now!

Digital tools or software can ease your life as a photonics professional by either helping you with your system design or during the manufacturing process or when purchasing components. Check out our compilation:

Proceed to our dossier

Top Feature

Vote now at the inspect award 2022 and choose the best machine vision products.

Vote Now!

Digital tools or software can ease your life as a photonics professional by either helping you with your system design or during the manufacturing process or when purchasing components. Check out our compilation:

Proceed to our dossier