Direct ink writing is one of those 3D printing technologies that has a process that can vary significantly. It’s not very commercialized, and mostly reserved for research labs.
Now, a team of scientists and engineers at Lawrence Livermore National Laboratory (LLNL) in California have developed a camera-based inspection system that can monitor complex 3D-printed structures using AI and machine learning to measure small variations and potentially identify problems before a part leaves the printer. Described in npj Advanced Manufacturing, the system pairs printer-mounted cameras with machine learning and computer vision to turn thousands of in-print images into measurements and spatial maps of the deposited material.
Catching Problems Earlier
Direct ink writing can produce flexible cushions and pads that depend on strands only a fraction of a millimeter thick. Gaps, breaks or diameter changes can hurt performance.
Traditionally, inspection happens after printing, via X-ray imaging or mechanical testing. It’s costly, slow and only reveals failures after the fact.
The LLNL system adds an earlier check. A camera images each layer as it’s deposited, and software measures the newest strands, including filament diameter.
“It’s a first-pass check,” said project technical lead Brian Weston. “It allows us to see things before we do some very expensive tests and to fail parts earlier if we already know they have broken strands or other problems.”
LLNL researchers (from left) data scientist Michael Zelinski, engineer Hamed Ziad Ammar and principal investigator Brian Giera beside the DIW 3D printer. (Photo: Blaise Douros/LLNL)
How It Performs
- Trained on nearly 15,000 human-annotated images across several lattice geometries
- Across 55 parts, measurements typically landed within a few micrometers of human ones
- Analysis takes milliseconds versus 20 minutes to an hour by hand, roughly 100,000 times faster on average
- Reusable calibration and dataset work means far less training data for new cameras and parts
Seeing the Whole Print
On a cushion roughly 25 by 25 centimeters, researchers combined about 2,500 images from one layer into a map of the interior.
The map showed filament diameters drifting across the print, revealing a slight platform tilt that an overall average could have hidden.
Because the camera sits on the printer, it can also inspect parts too large for practical high-resolution CT.
What’s Next
Principal investigator Brian Giera said the approach can extend to other additive, subtractive and experimental systems. Near term, it could scrap heavily defective parts mid-print and flag when X-ray CT is warranted.
The capability is headed to the Kansas City National Security Complex for evaluation. Longer term, the data could feed digital twins linking a part’s structure to predicted performance.
“When the printer can inspect its own work, we can start thinking about the system making its own accept/reject calls,” Weston said. “If it sees a defect, maybe it can assess whether that makes the part nonconforming.” If you want to learn more about the research, find the study here.
What do you think of this research? Let us know in a comment below or on our LinkedIn or Facebook pages! Plus, don’t forget to sign up for our free weekly Newsletter to get the latest 3D printing news straight to your inbox. You can also find all our videos on our YouTube channel.
*Cover photo credit: Garry McLeod/LLNL