A technology for enhanced inspection
X-ray computed tomography – a CT scan such as those used in the medical field – is used to check the interior quality of 3D-printed objects without damaging them. A series of these X-ray images are combined and reconstructed to reveal the internal structure, identifying weaknesses or printing errors.
However, scanning the same part from many angles can be time-consuming and expensive. ORNL’s technology, named Simurgh for a mythological winged beast, offers a solution. Simurgh uses realistic training data to teach a neural network, leveraging physics-based simulations with computer-aided design to reconstruct more accurate images with fewer CT scans than the conventional method. Scan times for very dense materials are now 12 times faster with a fourfold greater ability to detect defects.
ORNL researcher Amir Ziabari and his colleagues developed the technology in 2022 under DOE’s Advanced Materials & Manufacturing Technologies Office, or AMMTO, for use with 3D-printed metal parts.
The applications and performance of the technology has since been expanded under the Advanced Materials and Manufacturing Technologies program, or AMMT, in DOE’s Office of Nuclear Energy. Through this cross-cutting program, researchers at both labs translated the benefits of Simurgh to the stringent demands of the nuclear field. This has opened a host of new applications, from rapidly characterizing hundreds of nuclear parts and materials to safely inspecting irradiated parts for improving the printing process.
“Nuclear is a high-cost environment with extremely high standards for precision, materials and safety,” said Ryan Dehoff, director of DOE’s Manufacturing Demonstration Facility, or MDF, at ORNL. “The fact we’re using this tool suite in the nuclear sphere speaks to the quality and reliability of the technology.”
The MDF, supported by AMMTO, is a nationwide consortium of collaborators working with ORNL to innovate, inspire and catalyze the transformation of U.S. manufacturing.
The recent research partnership began after INL encountered a logistical challenge when trying to link defects to specific printing parameters: Researchers needed to scan more than 30 samples to recognize patterns, but each scan took 30 hours. Before paring down the effort, Chuirazzi consulted ORNL’s Ziabari. Licensing his colleague’s algorithm enabled collection of all the data in a timely way.
“Including prep, it now takes about 15% of the time it did to scan something with our setup,” Chuirazzi said. “We can do three scans in the amount of time it took us to complete one.”
Chuirazzi realized that these benefits could be applied to a different nuclear challenge under another federal program that focuses on nuclear fuels.
This Oak Ridge National Laboratory news article "National lab collaboration enables faster, safer inspection of nuclear reactor components, materials" was originally found on https://www.ornl.gov/news
