Shielding Case AOI & Assembly

Cosmetic Defects Inspection, 3D Metrology and Flexible Packaging of Shielding Case

Solution brief

The shielding case cosmetic defects, 3D metrology and flexible packaging solution are based on Dihuge TimesAI deep learning development platform, integrating CV+AI+Automation to achieve online real-time detection of shield in precision cell phone parts, defect removal, data analysis and statistics in a one-stop service, and can be continuously optimized iteratively.

Effectively solved four major difficulties in the field:

First, there is a large difference in the probability of occurrence of different types of defects, and the defect samples are unbalanced.

Second, there is very little data on serious functional defects.

Third, there is a significant difference in the shape and extent of the same type of defect.

Fourth, defects are small target defects.

Currently, there are more than 15 types of defects covered, covering almost all defects of shield products, such as: Metal surface impression, metal surface scratch, metal surface dirt, metal surface impression, black membrane impression, black membrane scratch, black membrane dirt, black membrane damage, membrane bubbles, black membrane offset position, hole deformation, hole burrs, cracks, poor planarity, etc.

Solution function

The shielding case defects, 3D metrology and flexible packaging solution based on AI vision can provide a one-stop service for online real-time detection, defective product removal, and data analysis and statistics of shield parts in precision products. By timely detecting batch defects and counting the distribution data of various defects, the production process is optimized, thereby achieving the effect of improving production yield; The detection system developed by this solution has excellent performance indicators and can be continuously updated iteratively. Its detection efficiency and accuracy are higher than manual detection, achieving the goal of reducing personnel and increasing efficiency.

Bright spot

1. Effectively solved three major difficulties in defect detection in the field (Unbalanced defect samples, small defects as targets, and different semantic levels of defect types) 2. Targeted design of a deep learning network architecture and model training method for shield components in precision parts, which has higher accuracy and better stability than conventional models 3. Excellent performance indicators, higher efficiency and accuracy than manual work, achieving the effect of reducing personnel and increasing efficiency 4. According to actual industrial production needs, combine product big data with industrial production to achieve the effect of optimizing production

Application scenario

Online real-time full inspection of shielding case

Product case

The top three companies in the global industry develop inspection solutions for mobile phone precision parts (stamping shield) based on Dihuge TimesAI platform, which is suitable for various line structures and sorting logic of users. There are nearly 20 types of defects detected: Metal surface impression, metal surface scratch, metal surface dirt, black membrane impression, black membrane scratch, black membrane dirt, black membrane damage, membrane bubbles, black membrane misalignment, hole deformation, hole burrs, cracks, poor planarity, etc. Defect detection performance indicators: Escape＜ 0.5%, overkill＜ 3%.

Scenario
Scenario
Scenario