LLNL camera system measures DIW strand diameter to within a few micrometers, while printing
Lawrence Livermore National Laboratory built a camera-based inspection pipeline for direct ink writing that measures filament diameter as each layer goes down, rather than waiting for the finished part. The work, published in npj Advanced Manufacturing as “Scalable on-machine inspection of direct ink write additive manufacturing,” comes from a nine-person team led by Brian T. Weston and Brian Giera.
DIW pushes soft or paste-like material through a nozzle in strands sometimes a fraction of a millimeter thick, and small shifts in diameter, pitch, or strand angle change how a finished cushion behaves mechanically. Standard inspection means X-ray CT after the print finishes, which caps resolution against object size, or optical checks that see only the surface.
| Metric | Result |
|---|---|
| Training images | ~13,800 human-annotated |
| Test parts | 55 |
| Measurement accuracy | within a few micrometers of human measurements |
| Cushion footprint tested | ~25 x 25 cm |
| Images from one layer | ~2,420 |
A camera mounted on the printer captures each new layer, and image segmentation software isolates the newest strands so a computer vision algorithm can trace and measure them. On a 25 by 25 centimeter cushion, the team stitched roughly 2,420 images from a single layer into a spatial map of filament diameter across the whole part. That map showed a large-scale diameter pattern that a single average measurement would have hidden, one traced back to a slight tilt between the build platform and the nozzle.
The paper contrasts this with Ai Build’s AiMaker, which also inspects large-scale extrusion prints during the build but by comparing images to previous reference prints and flagging visual deviation, not by producing continuous diameter maps. The LLNL authors say a printer-mounted camera can cover parts larger than CT can practically resolve, though the paper does not give a specific size ceiling for either method or a cost figure for the camera system.