Case Study: SKF

Objective:

The customer wanted to detect manufacturing flaws in bearings.

Challenge:

The parts to be inspected had previously been through a process that might leave grease marks that were very similar to the flaws to be detected. The presence of these marks next to small flaws that might be accepted meant that the AI had to be able to detect very subtle differences, which occasionally not even the operators could agree on.

 

Solution:

This required developing a specific learning method to reduce the number of false positives (non-existent flaws) and different perspectives were used to obtain more reliable predictions.

“Working with ENAIA helped us resolve issues with automatic visual control applications that would have been practically impossible without the Artificial Intelligence and Deep Learning tools that ENAIA applied to our project. After a long process to select and filter images, ENAIA designed and programmed a neuronal network model that through ever more accurate learning managed to achieve a success ratio of 99.9998 %, which did not vary over time. ENAIA demonstrates a high professional and technological spirit that keeps them at the cutting edge of industry 4.0.”

Enrique Marzal Enciso

Manufacturing Engineer, SKF