The challenge
What needed solving
The client made 69 gear variants distinguished by tooth count. Manual counting took 8–10 seconds per gear and produced a 9–12% false acceptance rate.
The solution
How Qualitas solved it
Cameras and controlled illumination capture each gear. A deep learning model on the Qualitas EagleEye Platform counts teeth, classifies the variant, and sends the result to a PLC.
Proof of concept
The system was validated across four gear variants and 1,200 products. The full manufacturing range included 69 variants by tooth count.
Results
In the proof of concept, teeth-counting accuracy reached 98%. Counting time fell from 8–10 seconds by hand to 1 second per gear, while image analysis took under 300 ms.
