Scientists at the Argonne National Laboratory, part of the U.S. Department of Energy (DOE), have devised a novel method for identifying flaws in additively manufactured metal components. The novel technology has the potential to revolutionise the additive manufacturing industry by allowing for the early detection and prediction of defects in 3D printed materials.
Many different sectors used additive manufacturing with metals because it allows for the rapid production of complex structures. Additive manufacturing, in which parts are built layer by layer using a 3D printer, is used to create everything from rocket engine nozzles and pistons for high-speed cars to custom orthopaedic implants. Although additive manufacturing facilitates the rapid construction of complicated components, its widespread adoption has been hampered by the emergence of structural faults during the building process.
Researchers could monitor the growth of pores in 3D-printed metals in near-real time by combining imaging and machine-learning approaches. The novel technology has the potential to revolutionise the additive manufacturing industry by allowing for the early detection and prediction of flaws in 3D printed materials.
Laser powder bed fusion was utilised to construct the metal samples for the investigation; this technique involves melting metal powder into the desired shape using heat from a laser. However, this method often results in the creation of pores, which might reduce the effectiveness of the part.
Several AM machines contain thermal imaging sensors to keep an eye on the building process, but these sensors only photograph the parts’ outside surfaces; therefore, they overlook the pore creation. The Advanced Photon Source (APS) in Argonne, a user facility of the Department of Energy’s Office of Science, provides the only technique to identify pores inside solid, metal objects directly.
“Our X-ray beams are so strong that we can image more than a million frames per second,” explained Samuel Clark, an assistant physicist at Argonne. Researchers could watch pores form in real-time according to these photos. Scientists revealed that pores produced within a sample cause unique heat fingerprints at the surface, detectable by thermal cameras and comparing X-ray and thermal images.
Researchers then used thermal images alone to train a machine-learning model to forecast the development of pores within 3D metals. To ensure their model was correct, they compared it to data from X-ray pictures, which they knew to be an accurate reflection of pore creation. The model was then tested in an unlabeled sample setting, where it was asked to identify heat signals and forecast pore development.
“The APS gave the 100% correct ground truth that allowed us to achieve perfect prediction of pore production with our model,” explained Tao Sun, an associate professor at UVA.
Sensors are included on many commercially available additive manufacturing machines, but they are less precise than the technology the scientists developed. “Our approach may quickly be implemented in commercial systems,” added Kamel Fezzaa, a physicist at Argonne. Machines should be able to monitor the printing process, identify when and where pores are created, and alter their settings automatically using simply a thermal camera.
If a machine detects a severe flaw in a component early in the production process, for instance, it can halt further assembly of that component. The new method can provide information on where pore faults might be within the part even if the construction process isn’t paused, which saves users time during inspection.
“If you have a log file that tells you these four spots potentially have faults, then you’re going to check out these four locations instead of looking at the entire part,” Sun noted.
The team plans to investigate more sensors that can identify additive manufacturing errors in the future. So they can create a system that can spot and fix problems as they arise in production. Finally, Sun revealed that they want to construct a comprehensive system to inform users and show exactly where the defect is and how it might be remedied.
