US researchers have used machine learning and computer vision to develop a tool for real-time quality assessment of parts being built in a 3D printer.
According to Oak Ridge National Laboratory, the software program, named Peregrine, can work on any powder bed system (such as laser bed powder fusion and electron beam melting), and can use standard cameras (between 4 and 20 megapixels) and a high-powered laptop or desktop computer. It is an alternative to expensive characterisation equipment.
ORNL researchers used a convolutional neural network (a type of deep learning popular in visual applications) to monitor the build at each layer. Anomalies — such as porosities or uneven powder or heat distribution — would see the system alert an operator.
“Capturing that information creates a digital ‘clone’ for each part, providing a trove of data from the raw material to the operational component,” said ORNL researcher Vincent Paquit.
“We then use that data to qualify the part and to inform future builds across multiple part geometries and with multiple materials, achieving new levels of automation and manufacturing quality assurance.”
The work is part of an effort by the lab to create a “digital thread” to collect and make sense of data at each step of production, from design to feedstock selection to build to testing.
Picture: DMG Mori
Subscribe to our free @AuManufacturing newsletter here.