Asyril has kicked off a new collaboration with HEIA-FR institutes iCoSys and SeSi — co-supervised by researcher Wolf Beat (machine-learning) and Associate Professor (UAS) Jean-Luc Robyr (mechanical engineering), with scientific collaborators Vincent Magnin (ML, with Wolf Beat) and Sivanesan Nirosh (mechanical, with Jean-Luc Robyr) — to make its Asyfill hopper system smarter. The project, called FiLeML, is supported by the canton of Fribourg’s Push Innovation Voucher.
What’s Asyfill ?
It’s a parts hopper used in automated assembly. The hopper stores small parts in bulk (e.g., screws, caps, components) and releases just the right amount onto a feeding system so a robot and camera can see, pick, and place them. Think of it as the steady “parts supply” that keeps a line moving, fewer manual top-ups, more consistent flow.
Today, operators often rely on manual checks to see how full a parts tray is. FiLeML aims to change that with real-time, high-accuracy fill-level estimation so production lines spend less time waiting and more time running.
How it works
-
A small MEMS sensor captures the hopper’s vibratory signals.
-
Hybrid machine-learning models, trained on a mix of simulated and real data, estimate how full the tray is across different part types and hopper sizes.
-
The team is targeting sub-second results and a quick, under-3-minute calibration that non-experts can do on the shop floor.
FiLeML builds on outcomes from the earlier ModIA initiative (digital twin + ML). The current project is underway and runs through September 2025.
Why it matters
Fewer production interruptions, timely refill alerts, and better monitoring: added intelligence without added complexity for manufacturers.