Cutting-edge research is at the core of everything we do.
A vision-based control system is reported, coupling active learning and uncertainty awareness with a foundation model to continually learn from errors during repeated builds of the same part.
We harnessed commonly available manufacturing process metadata alongside video to train deep learning regression models for closed-loop control and few-shot error detection in extrusion 3D printing.
We developed a generalisable deep learning model that can detect and correct a wide variety of 3D printing errors in real time and learn how to make parts from unseen materials.
Warp deformation is a commonly encountered error in additive manufacturing. We combined deep learning and expert heuristics to autonomously recognise and correct warp both in situ and for future prints.
