Skip to content

Case Sharing: FATP Visual Inspection Solution for Electronics Manufacturing

Background and Customer Needs

The customer in this case is a large electronics manufacturing services (EMS) provider. Their primary requirement is to leverage AI-trained models to perform high-precision, automated visual inspection of product appearance, ensuring product quality. Key inspection items include:

  1. Screws: Verifying the presence and correct fastening.
  2. High-voltage warning labels: Ensuring correct placement and orientation.
  3. Product nameplate labels: Checking position and alignment.
  4. Vent valve: Confirming proper installation and absence of missing components.
  5. Serial number (SN) and QR code comparison: Ensuring consistency between the serial number and the QR code engraved on the product cover.

Common defects include:

  • Misplaced or incorrectly oriented labels
  • Missing screws
  • Missing vent valve plugs
  • Discrepancies between the serial number and the QR code

Failure to quickly identify these defects can compromise product quality, affect airtightness, and, in severe cases, lead to safety issues during use.

Challenges

  1. Complex inspection points: Multiple inspection areas are distributed across different sides of the product. Traditional 2D industrial cameras cannot efficiently capture multi-angle, multi-focus images.
  2. Large number of inspection items: With up to 28 components to inspect, a fast and accurate solution is urgently required.
  3. Labor constraints: Manual inspection is time-consuming, labor-intensive, and prone to subjective errors.

Solution & Key Technologies

To solve the above challenges, the TM AI Cobot integrates visual technology to automate the inspection process. The specific technical methods include:

Imaging Technology

Using the TM5-900 robotic arm, multi-angle image capture can be achieved to meet customer requirements. The arm is equipped with an autofocus 2D camera (EIH: color, supports autofocus) that provides precise positioning and captures clear inspection images through TMflow. The images taken by the TM5 EIH are stored in the AI AOI Edge for analysis.

AI-Powered Inspection

Using the TM AI+ Trainer, 400 images can be enhanced 10-fold and trained through 50 iterations within 35 minutes. This allows quick model retraining to address production line misjudgments or anomalies.
Leveraging an Intel I7 12700 CPU and Nvidia RTX3060 GPU, the system can complete AI inspection of 28 component positions within 30 seconds, meeting production cycle time (CT) requirements. This significantly enhances automation efficiency and reduces labor costs.

Inspection Workflow

  • Pass (OK): Inspection results are uploaded to the production system, and the product proceeds to the next station.
  • Fail (NG): Results are immediately flagged, prompting operators to address defective items.

Application Scenarios and Benefits

Application Scenarios:

Ideal for high-precision visual inspection tasks in manufacturing, including detecting missing components, identifying foreign objects, and verifying correct placement of parts.

Benefits:

  1. Automation: Replacing manual operations reduces labor costs.
  2. Accuracy: Achieves an inspection accuracy rate of over 99% with a false positive rate of less than 1%.
  3. Efficiency: Meets the demands of high-volume, high-cycle-time production lines.

Conclusion

The TM AI Cobot, with its innovative vision technologies and AI-based decision-making capabilities, provides customers with a highly efficient and reliable automated visual inspection solution. The inspection accuracy exceeds 99%, with a false positive rate of less than 1%, significantly improving efficiency and precision. It overcomes the limitations of traditional inspection methods, significantly improving product quality while reducing production costs, making it a benchmark application in modern smart manufacturing.