98% accurate defect detection across all bearing variants
Kompass (AI) enabled Tigris
<0.5% false positives vs. >10% in legacy systems
Overview
NSK Bearing is a leading manufacturer of precision bearings catering to automotive and industrial applications. The production involves multiple SKUs each requiring a flawless surface finish and strict dimensional accuracy.
However, manual visual inspection posed significant challenges. Operators struggled to maintain consistency while inspecting high volumes, leading to undetected surface defects, high false positives, and reduced throughput. The need for a reliable, automated 360° inspection became essential to maintain quality, reduce human dependency, and support future production automation initiatives.
Opportunity
Ensuring defect-free bearing surfaces across multiple product variants presented unique challenges:
- Required complete 360° inspection including Inner diameter surface and outer diameter surface within a compact station footprint.
- Needed to accommodate 18 part SKUs at a common hardware setup.
- Aimed to achieve faster inspection cycles (<10 seconds per part) to align with production takt times.
- Enable traceability and defect highlighting for operator review and continuous improvement.
Jidoka’s Approach
• 360° Imaging with Triple-Camera Architecture
The bearing is loaded using a precision grip fixture designed to hold and place each variant securely. Three industrial cameras capture all critical surfaces:
- Top camera images the top face.
- Camera 2 captures the inner diameter as the part rotates 360°.
- Camera 3 captures continuous images of the outer diameter during the same rotation.
• Bottom-Face Inspection
Once top, ID and OD imaging are complete, the part is automatically flipped within the same station. The top camera then captures the bottom surface to ensure full-surface coverage.
• AI-Driven Defect Detection
Captured images are processed through Jidoka’s deep-learning Kompass engine, which classifies each bearing as OK or Not OK within 6 seconds. The system identifies common machining and surface defects such as dents, scratches, tool marks, or unwashed areas with 98% accuracy.
• Automated Decision and Handling
- Good parts are transferred to the linear conveyor for downstream operations.
- Defective parts are automatically removed via a pneumatic push rejection mechanism, ensuring seamless separation without disrupting the flow
• Smart Analytics & Visualization
Each inspection result is logged with visual defect highlights, enabling operator review, trend analysis, and traceability.