AI AOI Software System Solves Rare Defect Detection in Semiconductor Advanced Packaging

In the semiconductor advanced packaging process, the problem of rare defect and few-sample detection has always troubled enterprises. WeLinkirt's AI AOI software system provides an effective solution to this dilemma with advanced technologies.

User scenario: On the advanced packaging production line of a leading semiconductor manufacturer, in the chip packaging process, it is necessary to conduct an appearance defect inspection on the packaged chips. The inspection objects include rare defects such as scratches, cracks, and foreign objects on the surface of the chip package. Since the probability of such defects is extremely low, the number of samples is limited.

Pain points: Current industrial vision defect detection methods have deficiencies in end - to - end integrated operations, leading to quantification dilemmas. On the one hand, the miss-detection rate is relatively high, and rare defects are difficult to accurately identify, which affects product quality. On the other hand, false alarms are frequent, increasing the workload and cost of manual re-inspection. At the same time, traditional detection methods have a long change-over time and cannot quickly adapt to the detection requirements of different types of chips. There are also certain risks in data compliance, making it difficult to meet the enterprise's requirements for efficient and accurate detection.

Technical Principle

WeLinkirt's AI AOI software system uses a feature recognition algorithm based on a visual foundation model. This algorithm can deeply analyze the appearance features of chips and establish accurate defect judgment criteria through learning the features of good products. Its APDT positive-sample/few-sample learning mechanism is particularly crucial. With only 1-20 good samples, it can quickly learn the normal feature patterns of chips. When facing rare defects, based on the learned features, the system can keenly capture the differences from the normal pattern and accurately identify defects. The semantic false-alarm filtering function analyzes the defect semantics to eliminate false alarms caused by environmental interference and other factors, improving the accuracy of detection.

• The visual foundation model can automatically extract multi-dimensional features of the chip appearance, which is not affected by complex backgrounds and lighting conditions, enhancing the robustness of the system.
• APDT positive-sample/few-sample learning reduces the dependence on a large number of defect samples and has obvious advantages in rare-defect scenarios.
• Semantic false-alarm filtering is based on natural language processing technology to understand and judge defect information, effectively reducing the false-alarm rate.

WeLinkirt's Solution and Product

Centered on the AI AOI software system, it has the feature recognition ability of a visual foundation model, enabling 0-code automatic programming for a good product in 5 minutes, which greatly shortens the change-over time. Through APDT positive-sample/few-sample learning, only a small number of good samples are needed to complete model training. The semantic false-alarm filtering function further improves the detection accuracy. The system supports 100% local private deployment via SDK/API/Docker, ensuring that data does not leave the factory and meeting the enterprise's data compliance requirements. At the same time, it can be combined with the DaoAI 2D/3D AI AOI equipment to conduct a more comprehensive inspection of chips, such as the inspection of micron-level morphologies of hidden solder joints and coplanarity.

The AI AOI software system brings new breakthroughs to the rare defect detection in semiconductor advanced packaging with its high-efficiency, accuracy, and flexibility.

Quantitative results: After using the AI AOI software system, the chip defect detection rate reaches 98.5%, effectively reducing the miss-detection risk; the false-alarm rate is reduced by -75%, greatly reducing the workload of manual re-inspection; the change-over time is shortened from several hours to 5min, significantly improving the production efficiency of the production line.