IC Molding Cosmetic Defects Detection and Classification
Automatically identify and classify molding defects to increase yield and profitability
Graphical programming environment for deep learning-based industrial image analysis
Success or failure of an IC product hinges on the quality of the molding process which protects chips from the harm inflicted by external forces and moisture. Defects such as cracks, dilapidations, or voids may be embedded on the molding surface while a chip is being molded. Human inspection often misses very tiny cracks or low-contrast voids. It is also very challenging for conventional rule-based vision systems to detect the faulty region with clear defect definition. There are several types of defects such as cracks, jagged edges, and deformations. Many anomalies are also defects; however, rule-based vision systems cannot effectively differentiate a minor anomaly that is within tolerance from a clear defect that indicates the chip must be thrown out. The inability to classify defect patterns inhibits production teams from quickly understanding where there are potential issues.
Cognex Deep Learning tools help manufacturers identify and classify real molding defects. This advanced vision solution is trained using a series of images that represent both good and No Good (NG) results, enabling the software to omit anomalies that are within tolerance and flagging those that are truly significant defects. The Cognex location tool identifies the Region of Interest (ROI). Once the ROI is defined, the defect detection tool identifies the defect within that area. The classification tool then categorizes various types of defects. Using this information, production managers not only increase the yield of their finished ICs but also use the classification information to address production issues and fix them which increases profitability.