Algorithm boosts inspection speed and accuracy for additive manufacturing

Researchers at the DOE’s Oak Ridge National Laboratory developed an inspection algorithm to enhance X-ray computed tomography (X-ray CT) reconstructions of 3D-printed parts. The approach produces more accurate images while reducing the number of scans needed to verify the accuracy of a 3D-printed part. This increases inspection speed and reduces costs for time, labor, maintenance, and energy. The technology enables more rapid verification of components, including parts for nuclear reactors, large-scale castings and additive manufactured dense metallic parts.

Why it matters:

X-ray CT imaging is a common non-destructive method for inspecting the inside of 3D-printed parts for quality control. However, conventional methods are time-intensive, costly, and imprecise to be widely used. Increasing detection accuracy and speed for a high volume of specialized 3D-printed parts would reduce US supply chain bottlenecks and speed up rollout of new nuclear power sources. These advances boost the manufacturing industry and expand the energy supply in the US while reducing the cost of both electricity and domestically manufactured products.

The innovation:

• The technology, called Simurgh, involves a trained neural network that leverages physics-based simulations and computer-aided design to reconstruct accurate images with fewer CT scans.
• The algorithm was developed in 2022 and first incorporated in software by commercial partner ZEISS in machines at DOE’s Manufacturing Demonstration Facility at ORNL. The platform has since been adapted for applications like scanning large metal casted parts and evaluating nuclear components made of new metals and alloys.

Real-world impact:

ORNL partnered with Idaho National Laboratory to apply Simurgh to a challenging project, reducing X-ray CT inspection time from nearly 40 days to less than a week. This laid the groundwork for using the technology to scan radioactive materials and irradiated fuels in the future. The reduced scan time could also reduce radiation exposure for lab technicians, as well as wear on the detector, which lengthens its operating life and maintaining scan accuracy. Faster results translate to faster feedback to performance models, speeding the timeline for bringing new types of nuclear reactors to the power grid.

The benefits:

• The software platform can reduce sample scan time by more than 90% for some applications.
• Repeated analysis shows Simurgh’s reconstructed images are superior to those produced by either conventional software or commercially-available deep learning models, so fewer scans are needed for flaw detection.
• The most recent adaptations to Simurgh allow it to function with dedicated hardware to analyze images of complex metallic parts more than 100 times faster than standard software-only techniques, reducing scan durations from hours to minutes. This improvement enables high-throughput inspection at industrial scale, even in challenging industrial environments.
• Using the technology to inspect very large printed parts can provide a new domestic pipeline for key replacement parts used in nuclear or industrial processes. Many of these parts are currently produced abroad. Expanding this supply chain with custom-printed parts supports the nation’s energy and economic security.

Backed by science:

The technology was developed in the MDF with support from DOE’s Advanced Materials and Manufacturing Technologies Office (AMMTO). The applications and performance of Simurgh was subsequently expanded under both AMMTO and the DOE Office of Nuclear Energy’s Advanced Materials and Manufacturing Technologies program.

Deep dive: 

• Learn more about how the Simurgh technology works and how it could propel additive manufacturing toward industrial scale.
• Read about how Simurgh enhances nuclear reactor research.

The big picture:

The ORNL-developed Simurgh technology is poised to enable much broader use of 3D printing for rapid industry deployment of new alloys, novel nuclear energy technologies, and large metal parts that present supply chain challenges. The result will be a more robust, nimble US industrial sector and secure, reliable energy.

Read more stories about ORNL's science with impact.