Background
Terumo, a leading global manufacturer of medical devices, needed an inspection system capable of guaranteeing zero defect escapes across large-format products with microscopic surface flaws. Manual inspection required six operators per line, with each part taking around five seconds to inspect — too slow and inconsistent for their growing production demands.
Solution
The project adopted an intermittent conveyor system where a robot arm carries the camera, coordinating its motion with the conveyor to capture images precisely during each stop.
The setup includes five inspection stations and eight cameras, working in combination with Nordbo’s AI visual inspection technology.
The system is compatible with parts up to 135 × 220 mm, and is designed to inspect more than 150 product variants, detecting a broad range of surface and assembly issues such as:
Hair and fiber contamination
Foreign bodies
Surface cuts, folds, and seals
Printing or character defects
Part presence or placement irregularities
This configuration enables stable, multi-angle imaging and consistent detection quality even with frequent product changeovers.
To meet the manufacturer’s zero-escape target, Nordbo Robotics developed a data-driven inspection process emphasizing image precision, AI consistency and system stability:
Optimised imaging strategy
Coordinating the robot arm’s motion with conveyor pauses ensured every camera captured sharp, well-lit images from the right angle and distance — critical for detecting microscopic irregularities.
Adaptive AI Learning
The AI models were trained using a diverse, balanced dataset representing multiple product types, defect categories and surface conditions. This approach allowed consistent detection performance across more than 150 variants without retraining for each one.
Workflow calibration
The system was fine-tuned for throughput and accuracy, ensuring reliable detection under real production conditions, while minimizing false calls.
Together, these optimisations provided stable and repeatable performance for complex, mixed-model medical manufacturing.
Results
Overkill rate: Reduced to 5–8%, below the contractual target (≤8%)
Inspection speed: Improved from 5 seconds manually to 1.6 seconds per part
Labor requirement: Reduced from 6 operators to 2.5 operators per line
Accuracy: Sustained zero-escape detection across different product types
Customer Feedback
“We are very satisfied with the AI inspection results. The system meets our accuracy expectations, and we’re already planning a second device,”
— Terumo Project Representative
Following the project’s success, the customer began planning a second AOI installation and evaluating Nordbo’s AI software platform for broader use across additional production lines.
Impact
Through a practical combination of robotics, optics and AI inspection, we were able to help Terumo achieve consistent, zero-escape results while reducing inspection time and manual effort.
Well-designed automation now supports the customer in meeting strict medical quality requirements and provides a solid foundation for future production scaling.Start writing here...