Customer Challenge
Amphenol Phoenix, one of the world’s leading interconnect product manufacturers, needed to improve its quality control process to keep up with increasing demand and strict quality requirements from its Fortune 500 customers, including Google, Apple, and Microsoft.
Their existing manual inspection workflow posed several challenges:
- Unstable results — quality varied due to the human factor.
- Labor shortage — finding and retaining skilled inspectors was difficult.
- Slow cycle time — each part took ~20 seconds to inspect manually.
- Low throughput — one operator could inspect ~2,000 parts per 10-hour shift.
These limitations constrained output and risked impacting both delivery times and competitiveness.
Nordbo Robotics’ Solution
To address Amphenol Phoenix’s challenges, Nordbo Robotics deployed its QC Platform, an AI-powered vision system for automated quality inspection.
The process began with a thorough analysis of defect types and categories. Several smaller AI models were built to test expected performance early on. After identifying 30+ defect classes, the main phase of data collection and annotation took place. Using Nordbo’s skilled annotation team, each image was labeled with high precision to ensure maximum model quality.
Once the primary model was built, it was performance-tested and tuned for accuracy and speed. During the first month of production, the system continued to learn: data and performance were closely monitored, and the model was updated to further improve results.
To meet the ambitious target of 3.5 seconds per part, Nordbo Robotics designed and built a customized inspection machine tailored specifically to Amphenol Phoenix’s product. This machine integrated seamlessly with the production line while ensuring optimal cycle times.
From a technical standpoint, the system featured:
The final solution combined advanced AI with robust hardware integration:
- Vibratory bowl feeder automatically fed parts into the system.
- 9 camera stations mounted in the machine, with several capturing multiple images per cycle.
- 21 high-resolution images processed per part cycle, ensuring comprehensive inspection coverage.
- Defect classification: trained on more than 30 defect categories, enabling precise differentiation between defect types.
- Operator interface: allowed staff to monitor live statistics, review historic data, and verify inspection trends.
- Sorting logic: all station results were combined at the final stage; each part was classified as OK or NG. NG parts were further sorted into 7 defect groups for easier downstream handling.
This combination of AI software + custom-built hardware enabled Amphenol Phoenix to achieve fast, consistent, and reliable defect detection — while freeing up operators from repetitive inspection tasks.
Measurable Results
The implementation of Nordbo’s QC Platform delivered significant improvements:
- Cycle time reduced from 20 seconds to 3.5 seconds per part.
- Throughput increased from ~4,800 pcs per laborer/day to 24,685 pcs — more than 1,100% improvement.
- Equivalent to 12 operators — reducing labor shortages and cost pressures.
- Higher accuracy in defect detection, ensuring consistent quality across batches.
- Boosted competitiveness — stronger market position thanks to reliable automation.
“With Nordbo Robotics’ QC Platform, we achieved results we didn’t think were possible. The system not only improved inspection speed but also gave us a level of accuracy and stability that manual inspection simply couldn’t deliver. This has allowed us to meet customer demands with confidence while reducing pressure on our workforce.”
— Quality Manager, Amphenol Phoenix
About Amphenol Phoenix
Founded in 2001, Amphenol Phoenix is part of Amphenol Corporation, one of the world’s largest manufacturers of interconnect products. The company serves a wide range of industries and supplies components to several Fortune 500 companies, including Google, Apple, and Microsoft.