Use case
You may have operators who perform visual inspections on parts in production. You may have started to build up a small bank of images of good and bad parts. If this is the case, great, we can use it. And if not, it’s not too late to do it.
From these images, we can train a neural network capable of identifying visual defects in production in real time, with high-speed unit control.
This control mode has several advantages.
- It helps to identify defects as early as possible and avoid assembling failing parts with good parts, only to throw away the assembly.
- It allows you to identify process deviations very quickly, in order to get production back on track with minimal impact.
- It allows for automatic unit control with minimal cost.
- It eliminates subjectivity. Everything is defined by the choices made when classifying the images he learns about.
- It automatically enriches the defect library as it checks, allowing full traceability.
- It alerts you to the appearance of new potential defects and allows you to classify them according to the action to take. Following this, it updates itself to integrate this new information.