Maddox AI enables reliable and fully automated visual inspection of filter components in high-volume automotive series production. Even defects that are difficult to detect and visually very similar to good parts can be identified precisely — while maintaining extremely short cycle times.
Product
Industry
Total savings / year
Our customer manufactures filter components for the automotive industry in very high volumes. With well over one million parts produced per day, quality inspection must maintain extremely short cycle times while also ensuring a detection rate of more than 99.9%.
Image capture is carried out using an industrial camera controller from Keyence. The images are then transferred to the Maddox AI system via FTP. The inspection result must subsequently be transmitted to the PLC so that defective parts can be reliably and automatically removed from the production line.
Before the introduction of Maddox AI, a rule-based image processing system was used. However, due to the high variability among good parts and the sometimes very similar visual appearance of good and defective parts, this system was unable to reliably distinguish between OK and NOK parts.
The result was a high false reject rate: many actually defect-free parts were incorrectly classified as scrap. This led to unnecessary material loss and reduced production efficiency.
The goal was therefore to significantly reduce the false reject rate without affecting the high-speed requirements of production.
Due to the high variability among good parts, the system was unable to reliably distinguish between OK and NOK parts. Result: High scrap costs
The AI-based inspection reliably distinguishes between real defects and permissible product variations. Result: Significantly reduced material and scrap costs
Several days of maintenance effort per month due to re-calibration of the camera system.
No recalibration effort and no downtime caused by the inspection system.
Due to the high variability among good parts, the system was unable to reliably distinguish between OK and NOK parts. Result: High scrap costs
Several days of maintenance effort per month due to re-calibration of the camera system.
The AI-based inspection reliably distinguishes between real defects and permissible product variations. Result: Significantly reduced material and scrap costs
No recalibration effort and no downtime caused by the inspection system.
Less pseudo scrap
Inference time per part
Inspected parts per day
Would you like to learn more about our solution? – Test for free now!
We are ISO 27001 certified.
