In the semiconductor chip production process, accurate classification and detection of defects such as particles and scratches are crucial. WeLinkirt provides an efficient solution for the industry with advanced AI vision technology.
User Scenario: A leading semiconductor chip manufacturer needs to conduct strict inspections on the chip surface during the back - end packaging process of chip manufacturing. Its products are various high - performance chips, and the main inspection objects are particles and scratches on the chip surface. These particles may come from dust and impurities in the production environment, while scratches may occur during chip handling and processing. Accurately classifying and detecting these defects is crucial for ensuring the performance and reliability of the chips.
Pain Points: Traditional detection methods face many quantitative dilemmas. In terms of missed detections, due to the different shapes and sizes of particles and scratches, the missed detection rate of traditional detection methods is as high as 3%. This means that a large number of defective chips may enter the market, affecting the product reputation. The false alarm rate is also not to be underestimated, reaching 20%. Excessive false alarms lead to a large number of good products being misjudged, increasing the workload of re - inspection and production costs. In addition, manual inspection is not only inefficient but also costly in terms of labor. Moreover, a large amount of time is required to readjust the detection parameters during model change, and the model change time is as long as 30 minutes, seriously affecting production efficiency. For some tiny particles and scratches hidden inside the chips, traditional X-ray/CT detection methods are difficult to accurately identify, which cannot meet the requirements of high - precision detection.
Technical Principles
WeLinkirt uses advanced deep - learning algorithms and self - developed 3D camera imaging technology to solve these problems. The deep - learning algorithm is trained with a large amount of sample data and can learn the characteristic patterns of particles and scratches. For particles and scratches of different shapes and sizes, the algorithm can accurately classify and identify them. The self - developed 3D camera can perform three - dimensional topography reconstruction of the chip surface to obtain detailed information about the chip surface. This three - dimensional imaging method can detect particles and scratches hidden in tiny depressions or protrusions on the chip surface, which are often ignored by traditional 2D imaging technology. Through three - dimensional topography reconstruction, chip surface defects can be detected more comprehensively and accurately, improving the accuracy and reliability of detection.
- The deep - learning algorithm has a powerful feature - learning ability and can extract the key features of particles and scratches from complex images for accurate classification.
- The three - dimensional topography reconstruction technology of the 3D camera can provide three - dimensional information of the chip surface, making up for the deficiencies of 2D imaging and effectively detecting hidden defects.
- The adaptive adjustment ability of the algorithm can automatically optimize the detection parameters according to different chip types and detection requirements, improving the adaptability and accuracy of detection.
- By continuously learning and updating the model, it can continuously adapt to new types of particles and scratches and maintain the effectiveness of detection.
WeLinkirt's Solutions 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 is based on the feature recognition of the visual basic model and has a strong ability for positive - sample/few - sample learning. With only 1 - 20 good samples, it can achieve zero - code automatic programming in 5 minutes. At the same time, the system also has a semantic false - alarm filtering function, which can effectively reduce the false - alarm rate. The DaoAI 2D / 3D AI AOI equipment uses self - developed 3D cameras and three - dimensional topography reconstruction technology, which can detect hidden solder joints, coplanarity, and micron - level topography, and can perform high - precision detection on particles and scratches on the chip surface. In the implementation process, the system is first initialized, and relevant parameters are set according to the chip type and detection requirements. Then, a small number of good samples are used to train the system, allowing it to learn the characteristics of normal chips. During the production process, the equipment detects the chips in real - time, and the software system analyzes and classifies the detection results, marking the defective chips in time.
WeLinkirt's solution provides an efficient and accurate defect - detection method for semiconductor chip production.
Quantitative Results: By adopting WeLinkirt's solution, the semiconductor chip manufacturer has achieved remarkable results. In terms of the detection rate, it has increased from the original 97% to 99.2%, greatly reducing the missed - detection rate from 3% to <0.8%. The false - alarm rate has also been significantly reduced, from the original 20% to 3%, a reduction of - 85%, which has reduced a large amount of re - inspection workload and production costs. The model - change time has been shortened from the original 30 minutes to 5 minutes, improving production efficiency and enabling faster adaptation to the detection requirements of different chips.