By retrofitting an existing Keyence system with Maddox AI, 100% inline inspection of strip metal parts becomes significantly more robust. Slug marks, scratches, impact damage, and missing punched holes are detected reliably, while false rejects and manual readjustment effort are significantly reduced.
Product
Industry
Total savings / year
Our customer manufactures strip metal parts and stamped metal components for the automotive industry and its suppliers. Multiple product variants must be inspected for typical defects that can occur during the stamping process. The main focus is on slug marks on the top side of the parts, scratches, impact damage, and incompletely punched hole patterns.
Quality control had to be performed as a 100% inline inspection directly on the production line. At the same time, the inspection system had to maintain a cycle time of less than one second per part, including image capture and AI evaluation. Another key requirement was robust detection across different variants, even when material surfaces changed, lot-to-lot variation occurred, or lighting conditions shifted slightly.
It was especially important that the existing inspection setup did not need to be replaced completely. For that reason, the application was implemented as a retrofit of the installed Keyence system: the existing camera infrastructure remained in place, while Maddox AI took over the inspection intelligence via an edge device.
The existing Keyence system was rule-based and heavily dependent on rigid parameters. As a result, surface variations and lot-to-lot fluctuations led to a high false reject rate.
Maddox AI replaces rule-based logic with an AI-powered inspection system. This enables robust defect detection across different product variants while reducing false rejects.
When product variants changed or production conditions shifted, regular manual readjustment was required. This increased operational effort during day-to-day production.
New NOK images can be annotated in self-service, used for retraining, and deployed directly to the edge device.
The existing Keyence system was rule-based and heavily dependent on rigid parameters. As a result, surface variations and lot-to-lot fluctuations led to a high false reject rate.
When product variants changed or production conditions shifted, regular manual readjustment was required. This increased operational effort during day-to-day production.
Maddox AI replaces rule-based logic with an AI-powered inspection system. This enables robust defect detection across different product variants while reducing false rejects.
New NOK images can be annotated in self-service, used for retraining, and deployed directly to the edge device.
Cycle time per part
Effort required for variant adjustments
Less false reject and less manual reinspection
Would you like to learn more about our solution? – Test for free now!
We are ISO 27001 certified.
