Specializing in the manufacture of rubber and/or PTFE membranes, Intercarat called on Psycle to automate the quality sorting of its production lines using vision. Thibault Lonpret, Robotics and Automation Project Manager, shares his feedback.
What were the main points of your vision specifications?
We wanted to be able to trace each defect back to its precise cause, so the vision software had to be able to tell one defect from another and assign a specific name to each defect. We also wanted to work with a vision expert dedicated to this task, to move forward on machine learning in parallel with building the line. This allowed us to be much more confident and efficient when we went into production, since most of the work had been done upstream and was monitored by experts.
Psycle is a very responsive partner that always provides detailed answers to questions and proactively came up with proposals throughout the project.
How is the Psycle system used today?
It is used both on our automated membrane production lines and by our finishing department, which until now inspected these products manually. In production, the idea is to eliminate any human operation: a robotic arm picks up the part, presents it for vision inspection, then sorts it according to whether or not it is compliant. The finishing team, for its part, continues to teach the system: it annotates images that it has either taken itself or extracted from the production database.
They didn't hesitate to add features to their software to meet our needs.
What were the main phases of the project?
We started with tests at Psycle to validate feasibility. Then Psycle provided us with hardware so we could build a prototype, which we fed for several months to reach the level of performance we wanted. Psycle then supported us during the installation and industrialization of the system on the lines. Finally, they continued their support by gradually giving us more and more rights on the tool and offering us custom features.