(News from Nanowerk) Additive manufacturing, or 3D printing, can create custom parts for electromagnetic devices on demand and at low cost. These devices are very sensitive and each component requires precise fabrication. Until recently, however, the only way to diagnose printing errors was to manufacture, measure and test a device or use online simulation, both of which are computationally expensive and inefficient.
To remedy this, a research team co-led by Penn State created a first-of-its-kind methodology for diagnosing misprints with real-time machine learning. The researchers describe this framework — published in Additive manufacturing (“Mapping Geometric and Electromagnetic Feature Spaces with Machine Learning for Additively Fabricated RF Devices”) – as a critical first step towards correcting real-time 3D printing errors.
According to the researchers, this could make printing for sensitive devices much more efficient in terms of time, cost and computing bandwidth.
“A lot can go wrong during the additive manufacturing process for any component,” said Greg Huff, associate professor of electrical engineering at Penn State. “And in the world of electromagnetism, where dimensions are based on wavelengths rather than regular units of measurement, any small flaw can really contribute to large-scale system failures or degraded operations. If 3D printing a household item is like tuning a tuba – which can be done with wide adjustments – 3D printing devices working in the electromagnetic domain are like tuning a violin : small adjustments really matter.
In a previous project, researchers attached cameras to printheads, capturing an image each time something was printed. Although not the primary goal of this project, the researchers eventually curated a dataset that they could combine with an algorithm to classify types of printing errors.
“Generating the dataset and determining what information the neural network needed was central to this research,” said first author Deanna Sessions, who received her doctorate in electrical engineering from Penn State in 2021 and now works for UES Inc. as a contractor. for the Air Force Research Laboratory. “We use this information – from cheap optical images – to predict electromagnetic performance without having to do simulations during the manufacturing process. If we have pictures, we can tell if a certain item is going to be a problem. We already had these images and we said, “Let’s see if we can train a neural network to (identify errors that are creating performance issues).” And we found we could.
When the framework is applied to the print, it can identify errors during printing. Now that the impact of errors on electromagnetic performance can be identified in real time, the ability to correct errors during the printing process is much closer to becoming a reality.
“As this process is refined, it may start to create this type of feedback control that says, ‘The widget is starting to look like this, so I made this other tweak to let it work’, so that we can continue to use it,” Huff says.