Machine learning helps detect two ways the keyhole oscillates

The new breakthroughs were made by a research group headed by Tao Sun, associate professor of materials science and engineering at the University of Virginia. It has the potential to expand additive manufacturing in the aerospace industry and other industries that rely on robust metal parts.

Tao Sun, associate professor at UVA, in his lab at the University of Virginia. Image credit: Tom Cogill for UVA Engineering.

Their peer-reviewed paper was reported on January 6y2023, at Sciences magazine.

It satisfies the problem of detecting keyhole pore formation, which is one of the significant drawbacks of the common additive manufacturing method known as laser powder bed fusion or LPBF.

Introduced in the 1990s, LPBF uses lasers and powder metallurgy in 3D printed metal parts. However, porosity defects remain a challenge in stress-sensitive applications such as aircraft wings. A few pores were attached to deep, narrow vapor depressions, known as keyholes.

Keyhole configuration and size are defined as a function of scanning velocity and laser power, as well as the ability of materials to absorb laser energy. If the keyhole walls tend to be stable, it improves the laser absorption of the encapsulated material and also helps improve the laser machining efficiency.

But if the walls appear to wobble or collapse, the material hardens next to the keyhole, thus trapping the existing air pocket within the newly developed material layer. This makes the material very brittle and more likely to break under environmental stress.

Sun and his group, such as Carnegie Mellon University materials science and engineering professor Anthony Roulette and mechanical engineering professor Lianyi Chen of the University of Wisconsin-Madison, have come up with a way to detect the exact moment a keyhole is developed during the printing process.

By integrating operando synchrotron X-ray imaging, near-infrared imaging, and machine learning, our approach can capture the unique heat signature associated with keyhole generation with sub-millisecond temporal accuracy and a prediction rate of 100%..

Tao Sun, Associate Professor, Materials Science and Engineering, University of Virginia

In expanding the technology for real-time keyhole detection, the scientists also developed the way an advanced instrument — called the Obrando Synchrotron X-ray Imaging — can be used. Using machine learning, they also discovered two modes of keyhole oscillation.

Our findings not only advance additive manufacturing research, but can also practically expand the commercial use of LPBF for metal parts manufacturing.Roulette said. Roulette is also the co-director of the Next Manufacturing Center at Carnegie Mellon University.

Porosity in metal parts remains a major obstacle to the widespread adoption of LPBF technology in some industries. Keyhole porosity is the most challenging type of defect when it comes to real-time detection using lab-scale sensors because it occurs randomly below the surface.

Tao Sun, Associate Professor, Materials Science and Engineering, University of Virginia

Sun added,Our approach provides a viable solution for high-resolution, high-accuracy keyhole creation detection that can be easily applied in many additive manufacturing scenarios.. “

The group’s research was supported financially by the Department of Energy’s Kansas City National Security Campus operated by Honeywell FM&T.

Journal reference:

rang, z, et al. (2022) Machine learning-assisted real-time discovery of keyhole generation in laser powder layer fusion. Sciences.


Leave a Comment