Oak Ridge National Laboratory has developed a machine-vision software called Peregrine that monitors powder bed 3D printers in real time and flags surface defects as they form, eliminating the need for separate post-build characterization equipment.
The software uses a convolutional neural network trained on images captured at each layer of the build. A custom algorithm processes pixel values, analyzing edges, lines, corners, and textures to identify anomalies including uneven powder distribution, spatters, insufficient heat, and surface porosity. When Peregrine detects a potential quality issue, it automatically alerts the operator so adjustments can be made mid-build. Luke Scime, principal investigator for Peregrine at ORNL, described the core challenge the system addresses: "You're caring about things that occur on length-scales of tens of microns and happening in microseconds, and caring about that for days or even weeks of build time."
What's new here is the machine-agnostic design. Peregrine runs on standard cameras in the 4 to 20 megapixel range and a single high-powered laptop or desktop, and has been validated on seven different powder bed systems at ORNL, covering electron beam melting, laser powder bed, and binder jetting configurations. Printer manufacturers can install the software on existing hardware without custom integration work, and Peregrine's common image database transfers to each new machine to train new neural networks quickly.
The software feeds into ORNL's broader "digital thread" initiative, which builds a continuous data record from CAD design through feedstock selection, print build, and material testing. "Capturing that information creates a digital 'clone' for each part, providing a trove of data from the raw material to the operational component," said Vincent Paquit, who leads advanced manufacturing data analytics as part of ORNL's Imaging, Signals and Machine Learning group.
One active test case involves the Transformational Challenge Reactor program, which is pursuing the world's first additively manufactured nuclear reactor. For that application, the audit trail Peregrine generates matters as much as defect detection itself. "You could have a scenario in which the regulator will want detailed data on how a part was manufactured, and we can provide specs with the database built using Peregrine," Scime said. The work was funded by the Department of Energy's Advanced Manufacturing Office and Office of Nuclear Energy, and the results appear in the journal Additive Manufacturing.
