3D AI AOI Enables Zero - code Model Change for Multi - variety and Small - batch Chip Production

Wemio's 3D AI AOI device (self-developed 3D camera + 3D morphology reconstruction/point cloud, detecting hidden solder joints, coplanarity, micron-level morphology, pores and other defects in 2D optical blind spots, 2D-3D fusion) shortens the semiconductor chip model change time from several hours to 5 minutes through zero-code rapid model change technology, significantly improving production efficiency. In the semiconductor/chip industry, with the diversification of market demand, the multi-variety and small-batch production mode has gradually become the mainstream. A customer's chip production line mainly produces various types of chips, and the detection object is the micro-bumps on the chips. During the production process, problems such as micro-bump missing/bridging often occur, seriously affecting product quality and production efficiency.

Industry background and user scenario: In the semiconductor/chip industry, the rapid product replacement and the shift towards multi-variety and small-batch production have posed significant challenges. Wemio's 3D AI AOI device, with its unique zero-code rapid model change advantage, offers an effective solution for detecting defects such as micro-bump missing/bridging during chip production. As mentioned earlier, a customer's chip production line suffered from long model change times and low efficiency under the traditional production model, which severely restricted production progress and product quality. The application of this device has greatly improved the situation, significantly shortening the model change time and bringing remarkable economic benefits to the enterprise.

Pain points: Why is this hurdle so difficult to cross?

Multi - dimensional quantification of pain points: In traditional semiconductor chip production, multi-variety and small-batch production has brought numerous difficulties. In terms of the miss-detection rate, the traditional rule-based AOI detection method has a miss-detection rate of up to 5% for subtle defects such as micro-bump missing/bridging. This means that a large number of defective products may enter the market, causing reputational and economic losses to the enterprise. Regarding the false-alarm rate, the traditional method has a false-alarm rate of approximately 20%, resulting in a significant waste of manual re-judgment man-hours. Statistics show that manual re-judgment for each batch requires an additional 2-3 hours. The problem of model change downtime is also very serious. Each model change requires several hours of downtime, which greatly reduces the production rhythm and affects overall production efficiency. From the perspective of compliance risks, if defective products cannot be effectively detected, it may violate relevant industry standards and regulations, bringing potential legal risks to the enterprise. In terms of unit cost, due to problems such as manual re-judgment and model change downtime, the production cost per unit product has increased by about 15%.

Root - cause analysis: From a process perspective, the semiconductor chip production process is complex, and the size of micro-bumps is in the micron level. Traditional detection methods have difficulty accurately capturing their subtle changes. In terms of imaging, 2D optical detection has blind spots and cannot comprehensively obtain the 3D morphology information of micro-bumps, leading to miss-detection and false-alarms. The characteristics of chip materials may affect light reflection and refraction, interfering with the detection results. From the perspective of production rhythm, multi-variety and small-batch production requires frequent model changes. However, the traditional model change method requires manual code rewriting and parameter adjustment, which is time-consuming and labor-intensive. The release of the world's first industrial multi-modal large-model highlights the deficiencies of traditional detection methods in data processing and feature recognition, making it difficult to adapt to complex and changeable production requirements.

Technical principle

In - depth mechanism: Wemio's 3D AI AOI device uses a self-developed 3D camera for image acquisition. Through 3D morphology reconstruction and point-cloud technology, it can accurately obtain the 3D information of micro-bumps. In terms of algorithms, it combines the feature recognition of the visual basic model and uses the APDT positive-sample/few-sample learning technology. With only 1-20 good samples, it can quickly establish an accurate detection model. The device also uses semantic false-alarm filtering technology, which can effectively distinguish real defects from false-alarm information. The advantage of this technical mechanism is that it can comprehensively and accurately detect various defects of micro-bumps, even those hidden in the 2D optical blind spots.

Comparison with traditional methods: Compared with traditional rule-based AOI, rule-based AOI mainly relies on preset rules for detection. It is difficult to adapt to and accurately detect complex and changeable micro-bump defects. In contrast, Wemio's 3D AI AOI device has stronger adaptability and learning ability, and can automatically adjust the detection strategy according to different product models and defect types. Compared with manual visual inspection, manual visual inspection is subjective and prone to fatigue, with low detection efficiency and a high miss-detection rate. This device can perform 24-hour continuous detection with high detection accuracy and can quickly change models, greatly improving production efficiency. For example, when detecting micro-bump missing/bridging defects, the detection rate of this device can reach 99.2%, while the detection rate of traditional methods is only about 95%.

Typical application scenarios

• Micro - bump missing detection: Through 3D morphology reconstruction and point-cloud technology, the device can accurately obtain the shape and position information of micro-bumps. The difficulty lies in the tiny size of micro-bumps, which requires high-precision imaging and algorithms to accurately determine whether there is a missing bump. Wemio's 3D AI AOI device can effectively solve this problem with its self-developed 3D camera and advanced algorithms.
• Micro - bump bridging detection: Using the device's 2D-3D fusion technology, it comprehensively analyzes the connection between micro-bumps. The difficulty is that the bridging may be very subtle, and 2D optical detection is prone to missing it. The 3D detection ability of this device can clearly identify bridging defects.
• Coplanarity detection: Through 3D morphology reconstruction, the device measures the height and flatness of micro-bumps. The difficulty is to ensure that the measurement accuracy reaches the micron level, which the high-precision 3D camera and algorithms of this device can meet.
• Pore detection: Using the device's point-cloud technology, it analyzes the internal structure information of micro-bumps. The difficulty is that pores may be hidden inside and are not easy to find. The 3D detection method of this device can effectively detect pore defects.

Implementation case

Comparison before and after the implementation for an anonymous customer: A leading semiconductor chip manufacturer with a production line mainly engaged in multi-variety and small-batch chip production. Before using Wemio's 3D AI AOI device, the model change time was long, with each change taking 4-5 hours. The miss-detection rate was 5%, the false-alarm rate was 20%, and the manual re-judgment man-hours for each batch were about 3 hours. After the implementation of Wemio's 3D AI AOI device, the model change time was shortened to 5 minutes, the miss-detection rate was reduced to <0.8%, the false-alarm rate was reduced by -18%, and the manual re-judgment man-hours were almost zero. These data changes indicate that the device has greatly improved production efficiency and product quality.

The application of Wemio's 3D AI AOI device has brought a qualitative leap to the multi-variety and small-batch production in the semiconductor chip industry.

Wemio's solution and product

Product capabilities and implementation methods: Centered around the 3D AI AOI device, it has the ability of zero-code rapid model change. Through the feature recognition of the visual basic model, it can complete zero-code automatic programming in 5 minutes with one good sample. In terms of modeling, using the APDT positive-sample/few-sample learning technology, it can quickly establish an accurate detection model with only 1-20 good samples. During model change, there is no need for manual code writing, and the device automatically adjusts the detection parameters to achieve rapid model change. In terms of deployment, it supports multiple methods such as SDK / API / Docker and can be deployed 100% locally and privately to ensure data security. In terms of integration, the device can be seamlessly connected with other equipment on the production line to achieve efficient collaborative work. At the same time, the supporting DaoAI AI AOI software system can further optimize the detection process and improve detection accuracy.

Quantitative results and business value: After the implementation of Wemio's 3D AI AOI device, it has brought significant quantitative results to the enterprise. The model change time has been shortened from several hours to 5 minutes, greatly improving the production rhythm. The miss-detection rate has been reduced to <0.8%, and the false-alarm rate has been reduced by -18%, effectively improving product quality. The manual re-judgment man-hours are almost zero, saving a large amount of labor costs. At the same time, the production cost per unit product has been reduced by about 12%, enhancing the enterprise's market competitiveness.

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.