In the semiconductor industry, the inspection of wafer edge defects is crucial. Traditional sampling inspection methods are difficult to meet the requirements of high - quality production. WeLinkirt provides an innovative solution.
User scenario: A leading semiconductor chip manufacturer needs to conduct quality inspections on wafers in a key process of the wafer manufacturing production line. Its products are wafers used for various high - performance semiconductor chips, and the inspection object is the edge part of the wafers. Minor defects on the wafer edge can affect the performance and reliability of the chips, so accurate detection of these defects is particularly important.
Pain points: The traditional inspection method uses a sampling inspection model, which has many quantitative dilemmas. Sampling inspection can only cover a part of the wafers, and the missed detection rate is as high as 5%. This means that a large number of defective wafers may flow into subsequent processes, increasing production costs and product risks. At the same time, the false alarm rate of manual inspection reaches 30%. Not only does it waste a lot of manpower and time for re - inspection, but it may also lead to good products being misjudged as defective. Moreover, when changing production to wafers of different specifications, it takes 30 minutes to readjust the inspection equipment and parameters, which seriously affects the production rhythm and efficiency. In addition, industry compliance requires 100% full inspection of wafers, and the traditional sampling inspection method is difficult to meet this requirement.
Technical principle
WeLinkirt uses advanced algorithms and imaging technologies to solve the above problems. In terms of algorithms, a feature recognition algorithm based on a visual basic model is used. This algorithm can quickly learn the normal and defective features of the wafer edge. Through APDT positive - sample/few - sample learning, the model can accurately identify various situations on the wafer edge with only 1 - 20 good wafers. This is because positive - sample learning allows the model to focus on normal features, and few - sample learning can quickly converge with a small amount of data, improving the generalization ability of the model. In terms of imaging, the self - developed 3D camera combined with 3D morphology reconstruction technology can obtain accurate 3D information of the wafer edge. For tiny defects hidden inside the edge, the 3D camera can capture images from different angles, and the 3D morphology reconstruction integrates these images into a complete 3D model, clearly showing the location and size of the defects.
- The feature recognition algorithm of the visual basic model can quickly learn normal and defective features, improving the accuracy of detection.
- APDT positive - sample/few - sample learning reduces the dependence on a large number of samples and shortens the model training time.
- The self - developed 3D camera provides multi - angle images, and the 3D morphology reconstruction technology generates an accurate 3D model, facilitating the detection of hidden defects.
WeLinkirt's solution and product introduction
WeLinkirt provides the DaoAI AI AOI software system and the DaoAI 2D / 3D AI AOI equipment. The DaoAI AI AOI software system has powerful feature recognition capabilities. One good wafer can complete 0 - code automatic programming in 5 minutes. Through APDT positive - sample/few - sample learning, it can quickly adapt to the inspection needs of wafers of different specifications. At the same time, the semantic false - alarm filtering function can effectively reduce the false - alarm rate. The DaoAI 2D / 3D AI AOI equipment uses a self - developed 3D camera and 3D morphology reconstruction technology, which can detect hidden solder joints, coplanarity, and micron - level morphology. In the implementation, the equipment is integrated into the wafer manufacturing production line, and the software system processes the data collected by the camera in real - time to conduct online 100% full inspection of the wafer edge.
WeLinkirt's solution has achieved a transformation from sampling inspection to online full inspection, bringing new breakthroughs to semiconductor wafer quality control.
Quantitative results: After adopting WeLinkirt's solution, the detection rate of wafer edge defects has reached 98%, and the missed detection rate has been reduced to <2%, greatly reducing the risk of defective wafers flowing into subsequent processes. The false - alarm rate has been reduced by - 70%, effectively saving manpower and time for re - inspection. The change - over time has been shortened from 30 minutes to 5 minutes, significantly improving the production rhythm and efficiency, and meeting the 100% full - inspection requirement of industry compliance.