New Model Can Pinpoint Exactly Which Printer Made a 3D-Printed Part — Via Hidden "Fingerprints"
Differences between prints from even seemingly-identical printers can be recognized and used to confirm a part's origin, researchers show.
Researchers from the University of Illinois Urbana-Champaign, working with Sybridge Technologies, have developed a computer vision model that can work out where a 3D-printed part has come from — by looking for hidden "fingerprints" in the object.
"We are still amazed that this works: we can print the same part design on two identical machines — same model, same process settings, same material — and each machine leaves a unique fingerprint that the AI [Artificial Intelligence] model can trace back to the machine," says project lead Bill King, professor of mechanical science and engineering, of the team's work. "It's possible to determine exactly where and how something was made. You don't have to take your supplier's word on anything."
Anyone who has tried their hand at 3D printing may not be as surprised as King: even with the best ready-to-print hardware, it can be a finicky process requiring a range of adjustments before quality prints can be delivered. Coupled with manufacturing tolerances, it's easy to see why even printers that were neighbors on the assembly line will produce subtly different prints — and it's these subtle differences that make up the "fingerprints" tracked in the project.
Using a self-made dataset of 9,192 individual printed objects produced across three designs and four processes on a total of 21 printers, the team developed a computer vision model that they say has a 98 percent accuracy rate at determining which of the printers produced a given part — and without the need for the manufacturer to be involved.
"These manufacturing fingerprints have been hiding in plain sight," King claims. "There are thousands of 3D printers in the world, and tens of millions of 3D printed parts used in airplanes, automobiles, medical devices, consumer products, and a host of other applications. Each one of these parts has a unique signature that can be detected using AI. Our results suggest that the AI model can make accurate predictions when trained with as few as 10 parts. Using just a few samples from a supplier, it’s possible to verify everything that they deliver after."
The team's work has been published in the journal Advanced Manufacturing under open-access terms.