WeLinkirt's 3D AI AOI equipment (proprietary 3D camera + 3D morphology reconstruction/point cloud, detecting hidden solder joints/coplanarity/micron-level morphology/pores and other 2D optical blind spot defects, 2D-3D fusion) leverages deep 3D data analysis and AI algorithm integration to reduce semiconductor package pin coplanarity omission rates from an industry average of 1.5% to <0.4%, and increase dicing chipping detection rates from 85% to over 99.7%, significantly enhancing inspection accuracy and efficiency.
WeLinkirt's 3D AI AOI equipment (proprietary 3D camera + 3D morphology reconstruction/point cloud, detecting hidden solder joints/coplanarity/micron-level morphology/pores and other 2D optical blind spot defects, 2D-3D fusion) leverages deep 3D data analysis and AI algorithm integration to reduce semiconductor package pin coplanarity omission rates from an industry average of 1.5% to <0.4%, and increase dicing chipping detection rates from 85% to over 99.7%, significantly enhancing inspection accuracy and efficiency. In the back-end packaging and testing of semiconductor manufacturing, pin coplanarity and dicing chipping are critical defects affecting chip reliability and yield. As chip integration increases, pin pitch continues to shrink, with coplanarity requirements reaching micron or even sub-micron levels. Simultaneously, minute chipping generated during the wafer dicing process not only affects the chip's appearance but can also lead to subsequent packaging failures or abnormal electrical performance. Traditional inspection methods often struggle to comprehensively cover these complex and minute 3D morphological defects, posing significant challenges to chip quality control.
Pain Points: Why This Hurdle Is Difficult to Overcome
In semiconductor package testing, inspecting pin coplanarity and dicing chipping faces multiple pain points. Firstly, traditional 2D optical inspection has inherent blind spots for pin coplanarity. Slight pin tilt or warping may not be obvious in 2D planar images, leading to high omission rates, typically around 1.5% in the industry, which severely impacts product reliability. Secondly, dicing chipping defects are extremely small, usually at the micron level, and irregular in shape, making them difficult for traditional rule-based AOI systems to accurately identify. This results in persistently high false positive rates, adding to the burden of manual re-inspection, often leading to over 10% false positives. Thirdly, when process fluctuations cause slight changes in defect characteristics, traditional AOI rules require frequent adjustments, leading to long changeover downtime and impacting production rhythm. Furthermore, in the implementation of industrial AI quality inspection projects, model accuracy is not the only bottleneck; issues such as data annotation, model generalization capability, edge deployment, and real-time performance are equally important. For defects like pin coplanarity, which heavily rely on 3D information, the lack of high-quality 3D data sources is a major challenge for model training, while the small sample size and diverse characteristics of dicing chipping also increase the complexity of model training.
The root cause of these challenges lies in the physical characteristics of semiconductor devices and the limitations of inspection principles. Pin coplanarity issues involve the relative height difference between the pin base and the package body, which is a typical 3D morphological defect, and 2D imaging cannot directly acquire depth information. Even with multi-angle 2D imaging, it is difficult to accurately quantify micron-level coplanarity deviations. Dicing chipping, on the other hand, is caused by material brittleness, dicing blade wear, or improper process parameters. It manifests as irregular minute cracks on the chip edge, which may be obscured by shadows or reflections under certain lighting conditions, leading to insufficient contrast in 2D optical imaging. At the same time, high production speeds demand sub-second processing capabilities from inspection systems, which the complexity and computational load of traditional algorithms struggle to meet. WeLinkirt's 3D AI AOI equipment addresses these fundamental issues by using proprietary 3D camera technology to directly acquire high-precision 3D point cloud data, fundamentally overcoming the shortcomings of traditional 2D optics in depth information acquisition, providing richer and more accurate input for subsequent AI algorithms.
Technical Principles
The core technology of WeLinkirt's 3D AI AOI equipment lies in its proprietary high-speed, high-precision 3D camera and advanced 3D morphology reconstruction algorithms. This system employs structured light or laser triangulation principles, projecting specific patterns or light onto the chip surface and capturing the deformed images with a high-resolution camera. Combined with precise calibration parameters, it reconstructs the 3D point cloud data of the chip surface in real-time. This point cloud data contains X, Y, Z coordinate information for each pixel, with micron-level accuracy. For pin coplanarity inspection, the WeLinkirt system can directly calculate the distance between each pin's base plane and a reference plane, and analyze the overall flatness of the pin array, thus precisely quantifying coplanarity deviations. For instance, the WeLinkirt 3D AI AOI equipment can control pin coplanarity detection accuracy within ±2 microns, significantly outperforming traditional methods. For dicing chipping, the system identifies and extracts the 3D contour of the chip edge, analyzing local protrusions, depressions, or irregularities, combined with the deep learning models of the DaoAI AI AOI software system, to achieve high-precision defect identification. This integration of 3D data and AI enables WeLinkirt's 3D AI AOI equipment to demonstrate exceptional performance in complex defect detection.
Compared to traditional methods, WeLinkirt's 3D AI AOI equipment offers significant advantages. Traditional rule-based AOI relies on manually set thresholds and feature extraction rules, is sensitive to lighting and product surface variations, and has higher false positive and omission rates, especially when dealing with complex 3D morphological defects. Manual visual inspection, while flexible, is inefficient, inconsistent, and struggles to detect micron-level defects. WeLinkirt's 3D AI AOI equipment, on the other hand, utilizes 2D-3D fusion technology, combining texture and color information from 2D images with morphology and depth information from 3D point clouds, providing more comprehensive data input for AI models. The DaoAI AI AOI software system employs advanced vision foundation models and positive/few-shot learning capabilities, requiring only 1–20 good samples to quickly train high-performance models, greatly shortening development cycles and improving model generalization. Its semantic false positive filtering function effectively reduces false positives caused by environmental noise or non-defect features, leading to a reduction in false positive rates by over -90% for WeLinkirt's 3D AI AOI equipment in actual production lines, significantly reducing manual re-inspection hours.
Typical Application Scenarios
- **Pin Coplanarity Inspection**: After chip encapsulation, WeLinkirt's 3D AI AOI equipment uses high-precision 3D scanning to measure the Z-axis height of each pin, calculating the flatness of the entire pin array to ensure all pins are within specified coplanarity tolerances. The challenge lies in the large number of pins, small pitch, and the difficulty of detecting minute warps in 2D images.
- **Dicing Chipping Inspection**: After wafer dicing, WeLinkirt's 3D AI AOI equipment reconstructs the 3D contour of the chip edge, identifying and quantifying micron-level chipping, cracks, or burrs. The difficulty lies in the diverse forms of chipping, their minute size, and potential occlusion by shadows.
- **Micron-level Morphological Defect Inspection on Chip Body**: This includes pits, bumps, scratches, and pores on the chip surface, which can affect the chip's electrical performance or reliability. WeLinkirt's 3D AI AOI equipment can precisely acquire depth and size information of these microscopic morphologies for high-precision identification.
- **BGA/LGA Solder Ball/Pad Height and Coplanarity**: For Ball Grid Array (BGA) or Land Grid Array (LGA) packages, accurate measurement of solder ball or pad height consistency and coplanarity is crucial. WeLinkirt's 3D AI AOI equipment can rapidly perform 3D measurements on thousands of solder joints to ensure soldering quality.
- **Hidden Solder Joint Defect Detection**: In some package structures, solder joints may be partially or completely hidden beneath the device, making direct 2D visual inspection impossible. Through 3D morphology reconstruction, WeLinkirt's 3D AI AOI equipment can indirectly analyze the impact of solder joint morphology on overlying structures, combined with AI to infer hidden defects.
Case Study
A leading semiconductor packaging and testing manufacturer had long struggled with pin coplanarity omission and dicing chipping false positives. Their traditional 2D AOI system had an omission rate of up to 1.8% for pin coplanarity inspection, leading to defective products flowing to the next process, increasing rework costs and customer complaint risks. Simultaneously, the false positive rate for dicing chipping inspection remained around 12%, requiring significant manual re-inspection hours daily, severely slowing down the production line. Upon introducing WeLinkirt's 3D AI AOI equipment, the situation improved dramatically. In the initial rollout, the WeLinkirt team, using the DaoAI AI AOI software system, completed model training and deployment with just 15 good samples within 30 minutes. After a month of pilot operation and data accumulation, WeLinkirt's 3D AI AOI equipment consistently reduced the pin coplanarity omission rate to <0.3%, significantly exceeding customer expectations. Concurrently, the detection rate for dicing chipping defects increased to 99.8%, and the false positive rate decreased by -95%, reducing daily manual re-inspection hours by nearly 85%. The WeLinkirt system also supports multi-variety, small-batch production, with its 0-code changeover feature reducing production line changeover time from 30 minutes to 5 minutes, greatly enhancing production flexibility and efficiency.
WeLinkirt's 3D AI AOI equipment has not only significantly improved our inspection accuracy and efficiency but, more importantly, it has given us unprecedented confidence in product quality, truly realizing our vision of 'zero defects' in production.
WeLinkirt Solutions and Products
WeLinkirt's 3D AI AOI equipment is a core solution tailored for the high-precision inspection needs of the semiconductor industry. Its proprietary 3D camera, combined with advanced 3D morphology reconstruction technology, can acquire micron-level precise 3D data of the chip surface, completely addressing the blind spots of 2D optics in depth information. The WeLinkirt DaoAI AI AOI software system serves as the brain of this equipment, featuring built-in vision foundation models with excellent feature recognition capabilities. It supports APDT positive/few-shot learning, allowing model training to be completed with just 1-20 good samples, greatly reducing modeling barriers and cycles. Addressing the characteristics of multi-variety, small-batch production in semiconductors, the WeLinkirt system supports 0-code changeover, enabling operators to switch between different product models through a simple graphical interface without professional programming knowledge, ensuring production line flexibility. Furthermore, WeLinkirt solutions support 100% local private deployment, with all inspection data processed and stored within the customer's factory, meeting the stringent requirements for data security and compliance in the semiconductor industry. WeLinkirt's 3D AI AOI equipment also integrates a semantic false positive filtering function, which uses AI to deep-learn and recognize false positive features, effectively filtering out false positives caused by environmental interference or non-defect features, further enhancing the purity of detection.
In terms of quantified results, WeLinkirt's 3D AI AOI equipment consistently keeps the omission rate for semiconductor pin coplanarity inspection below <0.4%, significantly lower than the industry average. For dicing chipping detection, the detection rate increased to over 99.7%, and the false positive rate decreased by -95%, substantially reducing manual re-inspection workload. The introduction of this system led to a 1.2% increase in the customer's overall production yield, saving approximately 1.5 million RMB annually in rework and manual re-inspection costs. The fast modeling and 0-code changeover capabilities of the WeLinkirt DaoAI AI AOI software system reduced product changeover time from 30 minutes to 5 minutes, improving equipment utilization and bringing significant business value and competitiveness to the customer.
FAQ
What is chip pin coplanarity inspection and why is it so important?
Chip pin coplanarity inspection involves measuring whether all pin ends of a packaged chip lie within the same plane and their deviation from a reference plane. This is crucial for ensuring good contact between the chip and the PCB during subsequent soldering processes. Poor pin coplanarity can lead to cold solder joints, short circuits, or poor contact, severely impacting chip reliability and the performance of the final product.
How does WeLinkirt's 3D AI AOI equipment differ from traditional 2D AOI in terms of inspection cost?
Traditional 2D AOI equipment may have lower initial purchase costs, but its high omission and false positive rates for complex defects (such as pin coplanarity, micron-level chipping) lead to higher long-term costs, including manual re-inspection hours, rework expenses, increased scrap rates, and potential customer claims. While WeLinkirt's 3D AI AOI equipment has a relatively higher initial investment, it significantly reduces omissions and false positives, minimizes manual intervention, and improves yield and production line efficiency, ultimately delivering superior overall cost-effectiveness and a faster return on investment in the long run. Specific pricing requires project-based assessment.
How does WeLinkirt's 3D AI AOI equipment adapt to the multi-variety, small-batch production model in the semiconductor industry?
WeLinkirt's 3D AI AOI equipment adapts to multi-variety, small-batch production through its 0-code changeover feature in the DaoAI AI AOI software system. Operators can quickly switch between different product models in minutes via an intuitive graphical interface, eliminating complex programming or lengthy downtime for adjustments. Furthermore, the APDT positive/few-shot learning capability allows new product models to be established with only a small number of good samples, further shortening new product introduction cycles and enhancing production flexibility and efficiency.
This article was generated by AI. Customer cases are simulated scenarios based on real product capabilities and figures are illustrative; see product pages for official benchmarks.