Machine Learning System Provides "First-of-Its-Kind" Classification of 3D Printing Errors

from hackster.io

A team of researchers from Pennsylvania State University, the University of Dayton Researcher Institute, the Air Force Research Laboratory, and two private companies has come up with a framework designed to diagnose 3D printing errors via machine learning — the first, the team claims, of its kind.

"A lot of things can go wrong during the additive manufacturing process for any component," explains Greg Huff, associate professor of electrical engineering at Penn State, of the problem the team set out to solve. "And in the world of electromagnetics, where dimensions are based on wavelengths rather than regular units of measure, any small defect 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 broad adjustments — 3D-printing devices functioning in the electromagnetic domain is like tuning a violin: Small adjustments really matter."

With such tight tolerances in the parts being manufactured, the usual approach to getting things right is laborious: Making the device, measuring it, and testing it, then taking those readings and tweaking the next version, and the next, and continuing until the device passes muster. Simulation is an alternative, but computationally expensive — and slow.

Using a dataset of images captured by a camera fitted to a 3D print head, used in a previous project by the same team, the researchers were able to train an algorithm to classify various types of print errors — and to figure out how a 3D-printed electromagnetic device will perform.

"We’re using this information — from cheap optical images — to predict electromagnetic performance without having to do simulations during the manufacturing process," explains first author Deanna Sessions.

"If we have images, we can say whether a certain element is going to be a problem. We already had those images, and we said, 'Let’s see if we can train a neural network to [identify the errors that create problems in performance].' And we found that we could."

Identifying errors which could lead to an out-of-spec part is only part of the problem, though: The team is hopeful the system system can be adapted to provide live feedback during the print process, potentially allowing errors to be corrected before it's too late — automatically optimizing the print process as it's underway.

The team's work has been published under closed-access terms in the journal Additive Manufacturing.

Main article image, Huff and Sessions testing 3D printed devices, courtesy of Tyler Henderson/Penn State.