Researchers at University of Virginia have published a paper highlighting their work in the use of machine learning (ML) for the real-time detection of keyhole defects during the LPBF printing of titanium.
Keyhole defects are common in laser welding and LPBF printing processes, and can result in the formation of random pores through the printed material, resulting in a weak and brittle part.
Keyholes
A keyhole defect in metal 3D printing refers to a cavity that forms in the melt pool of the material being printed. This cavity is shaped like a keyhole (hence the name), with a narrow opening at the top and a wider area at the bottom.
Keyhole defects can occur during laser-based 3D printing processes when the laser is operating at a slow speed and high power. Keyhole defects are caused by a variety of factors, including the properties of the material being printed, the laser parameters being used, and the design of the 3D printing setup. You can read our previous article on keyhole formation here.
The team, consisting of researchers from University of Virginia, Carnegie Mellon University, and the University of Wisconsin-Madison used simultaneous high-speed operando synchrotron x-ray imaging and thermal imaging, along with multiphysics simulations, to discover two types of keyhole oscillation in laser powder bed fusion of Ti-6Al-4V.
They then utilized machine learning to further amplify their understanding of this phenomenon and developed an approach for detecting the stochastic (random) keyhole porosity generation events with sub-millisecond temporal resolution and a near-perfect prediction rate.
You can see one of the ML-aided parts in the image below.
The operando x-ray imaging provided highly accurate data labeling, enabling the researchers to demonstrate a practical and straightforward method for implementing their approach in commercial systems.
Accelerating Widespread Adoption
This approach was developed to detect the exact moment when a keyhole pore forms during the printing process, which they succeeded in achieving.
“By integrating operando synchrotron X-ray imaging, near-infrared imaging, and machine learning, our approach can capture the unique thermal signature associated with keyhole pore generation with sub-millisecond temporal resolution and 100% prediction rate,” said Tao Sun, associate professor of materials science and engineering at the University of Virginia.
The porosity resulting from keyhole formation has been seen as a hurdle in the widespread adoption of large scale metal additive manufacturing, especially in industries requiring a high level of quality and reliability in their parts. It is difficult to detect using typical sensors as the defect formation occurs randomly beneath the surface.
“Our findings not only advance additive manufacturing research, but they can also practically serve to expand the commercial use of LPBF for metal parts manufacturing,” said Anthony Rollett, co-director of the NextManufacturing Center at CMU.
You can read more in the paper titled “Machine learning–aided real-time detection of keyhole pore generation in laser powder bed fusion”, published in Science, which is available at this link.
