A / VISUAL INTELLIGENCE
Examples → genAI → recognition
Teach the system
what quality means.
Quality control depends heavily on people’s experience. Our vision systems are Human-in-the-Loop: the operator shows examples, identifies issues, corrects and validates. The system learns those criteria and makes them repeatable.
From the few real samples we use generative AI to expand the data realistically: new variants of parts and defects, consistent with the process, which the expert confirms before they enter the system. So the model starts sooner and performs better, even when line images are scarce.
We do not program a machine for a closed list of defects. The expert transfers their knowledge of quality to the system. That is how we handle complex, variable or hard-to-formalise inspections that traditional vision struggles with.
B / SIGNAL INTELLIGENCE
Understand what your
machines are saying.
Every machine generates vibration, current, temperature, pressure, sound, motion, PLC data and other process information. These data are often collected but only partly used: they sit in logs or dashboards, without becoming a continuous reading of how the line is running.
In the signals we look for patterns, correlations, anomalies and changes in behaviour. We compare the current trend with the machine’s known regime, to see when something is drifting: wear, an unstable set-up, a different operating mode.
The goal is not to collect more data, nor to add sensors for their own sake. It is to turn machine signals into useful information that can be asked, explained and used to act before the problem becomes scrap or downtime.
