Application of AOI visual inspection in LED industry-Xianghongyu

1. The necessity of AOI technology being applied to LED detection

LED production involves complex processes such as evaporation, yellow light, and etching, which are prone to the formation of defects such as cracks, metal residues, cracks, and black spots on the chip surface1。 Traditional manual visual inspection is limited by human eye resolution and fatigue factors, making it difficult to meet the following challenges:

1. Miniaturization trend: LED lamp beads are reduced in size to 1-3 mm, making it difficult to observe subtle imperfections with the naked eye28。

2. High-speed production demand: The production line speed reaches hundreds of thousands of pieces per hour, and manual sampling cannot cover the full amount of testing28。

3. Complex defect types: including surface colloidal anomalies (more glue, less glue), bubbles, scratches, pin offsets, and dozens of defect types, which need to be accurately classified24。

AOI technology significantly improves yield and reduces costs through automated, data-based inspection processes. For example, after introducing AOI, the detection speed is increased by 500% and labor costs are reduced by 70%2。

2. Technical advantages of AOI system

1. High-precision imaging and light source design

• Multi-angle multispectral strobe light source and high-resolution camera (e.g., 4K/8K pixels) to capture small defects on the chip surface (down to 2μm)14。 For example, Han's AOI equipment clearly identifies wafer appearance defects through combined light source technology1。

• For different inspection scenarios (such as LED backlight plates), high-magnification lenses and rotating platforms are configured to achieve multi-angle imaging and cover complex defects such as warping and incomplete coding46。

2. Intelligent algorithms and autonomous learning

• AI models based on deep learning (such as Gechuang Dongzhi's "Tianshu System") automatically classify defect types through massive defect sample training, and the accuracy rate has increased from 85% to more than 90% manually57。

• The system has autonomous learning capabilities and can automatically update the model when encountering new defects, eliminating the need for algorithm engineers to intervene and shortening the development cycle47。

3. Efficient production collaboration

• The modular design integrates transmission, detection, and sorting functions, supports high-speed detection of 150,000-300,000 pieces per hour, and uploads data to the cloud in real time to achieve quality traceability24。

• The equipment is equipped with a pre-positioning function and manipulator to reduce downtime, and a single device can run continuously for more than 1.5 hours, supporting one person to manage multiple machines14。

3. Typical application scenarios and cases

1. LED chip inspection

• Front-end process: Detect process defects such as evaporation cracks and development residues. For example, Han's Semiconductor's AOI visual inspection machine (COW) can identify defects at the 2μm level, breaking the monopoly of foreign technology1。

• Wafer-level inspection: Gechuang Dongzhi's silicon wafer AOI equipment uses high-definition imaging and general anomaly localization technology to solve problems such as chemical residues and cracks, and improve semiconductor yields5。

2. LED lamp bead appearance detection

• The equipment cooperated by Dongsheng Intelligent and Ascend AI can detect more than 15 kinds of defects (such as bubbles and scratches) at a speed of 150,000 pieces per hour, and optimize the production process through cloud analysis2。

3. LED backlight panel detection

• The patented technology device identifies defects such as film paper lifting and pin offset through multi-camera collaboration, and the AI algorithm can adapt to different brightness scenarios and reduce manual parameter adjustment time46。

4. Micro LED mass detection

• For Micro LED wafers, traditional AOI can only detect surface defects, while Hymson and Fuzhou University developed non-contact electroluminescence technology (FED-NCEL) can non-destructively detect the performance of light-emitting layers and electrodes, filling the gap in the industry8。

4. Industry challenges and future trends

1. Technical bottlenecks

• AOI has limited ability to detect internal defects (such as abnormal light-emitting layers) and requires a combination of electroluminescence (EL) or photoluminescence (PL) techniques to achieve comprehensive detection8。

• Complex defects, such as similar texture interference, still require support from AI models with higher computing power7。

2. Development direction

• Multi-technology integration: For example, Dongsheng Intelligent combines AOI with Ascend AI hardware to improve algorithm efficiency2； Hymson integrates optical, mechanical and electrical testing to break through the problem of mass production of Micro LED8。

• Intelligent upgrade: Real-time linkage between inspection data and production system is realized through industrial Internet platforms (such as Gechuang Dongzhi) to optimize the yield of the whole process57。

summary

AOI visual inspection has become a key tool for improving quality and efficiency in the LED industry through three core capabilities: high-precision imaging, intelligent algorithms, and efficient collaboration. With the continuous breakthrough of AI and optical technology, AOI will develop in a more intelligent and full-process direction in the future, helping the LED industry move from "manufacturing" to "intelligent manufacturing". Enterprises need to choose suitable AOI solutions based on their own needs, such as Han's Semiconductor, Dongsheng Intelligence, Gechuang Dongzhi and other manufacturers have launched mature equipment for different scenarios125。