Visual inspection of bearing rings

Situation

Japanese mechanical components manufacturer JTEKT group is renowned for their uncompromising approach towards the quality of its products. To protect this reputation, its plant in Olomouc, Czech Republic performs the inspection of every single piece of both the final product (bearing) and its components (e.g. bearing rings).

The inspection was originally performed manually – an employee had to check for multiple patterns, such as faulty laser engraving or a variety of barely visible surface defects on both diameters and faces of the product. Even a single overlooked defect could lead to damaged reputation, claims and fines by their customers.  

In order to make the quality inspection process more efficient, the company started searching for an automated inspection system. Considered systems, both the ones based on traditional machine vision and on deep learning, faced high amounts of false defect alarms connected with the presence of remnants of the liquid conservant on the products. The systems also had a tendency to overlook tiny defects, as these were as small as 0.5 mm. Tiny defects were also confused with dirt particles.

Solution

Cognexa delivered the Ring inspector – a fully automated machine consisting of mechanical, optical and software components seamlessly integrated into the assembly line. The system replaced the former human inspection of each part, automatically removing the defective parts from the assembly line.  

Ring inspector automatically inspects 100% of the bearing ring surface, looking for 15 types of defects on 14 different types of products. It also performs a robust inspection of various defects of laser engraved signature. The solution is based on deep learning and custom algorithms, and Cognexa was the only vendor able to detect even the smallest defects on extremely glossy surfaces with false indications.

Result

After the trial period the system has been in production since 03/2020. 2.5 years later, it still automatically performs the inspection for two production lines, benefiting the bottom line by reduction of human labor costs and decrease in customer claims. Being an integral part of the company’s digital transformation efforts, the system also archives the visual documentation and feeds live data about the quality of each produced part into the company’s quality management system. Our AI quality control solution increased the level of automation, saved 52 000 €/year and line productivity (parts per labor hour) raised by 100%.

Innovation sparked interest within the group and in the media:

• http://www.koyobearings.cz/aktuality/prvni-vyuziti-umele-inteligence-v-olomouckem-zavode-koyo-bearings/
• https://ct24.ceskatelevize.cz/ekonomika/3057770-olomoucka-firma-koyo-investovala-do-umele-inteligence-automat-kontroluje-vyrobena