In the electronics/PCBA industry, the quality of SMT solder joints is crucial. WeLinkirt's 2D AI AOI equipment brings an accurate solution for SMT solder joint defect detection with advanced technology.
User scenario: The SMT production line of a leading electronics manufacturing company mainly produces various PCBA products. After the SMT placement process, it is necessary to inspect the solder joints on the PCB boards. The inspection objects include SMT solder joint defects such as cold solder joints, bridging, and insufficient solder. These defects may lead to unstable product performance and even failures during use, so the requirements for solder joint quality inspection are extremely high.
Pain points: Traditional SMT solder joint inspection methods have many quantitative dilemmas. First of all, the missed detection rate is relatively high. Some minor defects such as cold solder joints and insufficient solder are easily overlooked. According to statistics, the missed detection rate reaches about 2%, which may allow defective products to enter the market. Secondly, false alarms are serious, with a false alarm rate of about 30%. A large number of false alarms not only increase the workload of manual re-inspection but also reduce the production efficiency of the production line. In addition, traditional inspection methods require a lot of time for programming and debugging during model change, and the model change time is up to 30 minutes, which seriously affects the flexibility of production. In the current trend of pursuing precision in industrial vision quality inspection, these problems need to be solved urgently.
Technical principle
WeLinkirt's 2D AI AOI equipment uses high-resolution 2D imaging technology. The high-resolution camera it is equipped with can capture the subtle features of solder joints, providing clear and accurate image data for subsequent defect analysis. In terms of image processing, a deep learning algorithm is used for secondary image judgment. Through learning and training on a large number of solder joint images, the model can accurately identify different types of solder joint defects. This algorithm is effective because it can learn the differences between the normal features and defect features of solder joints. For example, for cold solder joint defects, the model can learn the differences in grayscale values, shapes, etc. of cold solder joints in the image compared with normal solder joints. At the same time, the semantic false alarm filtering algorithm can effectively filter false alarms based on the semantic information of solder joints, reducing unnecessary manual re-inspection.
- High - resolution 2D imaging technology provides clear images, laying the foundation for defect detection.
- The deep learning algorithm accurately identifies different types of solder joint defects through learning from a large amount of data.
- The semantic false alarm filtering algorithm filters false alarms based on semantic information, improving detection efficiency.
WeLinkirt's solution and product
Centered on the 2D AI AOI equipment, WeLinkirt provides a complete SMT solder joint inspection solution. The equipment has the ability of high-speed online full inspection and can inspect solder joints with micron-level accuracy, ensuring that no minor defects are missed. In terms of programming, paired with the DaoAI AI AOI software system, automatic programming without code can be completed for a good product in only 5 minutes, greatly shortening the model change time. At the same time, the APDT positive sample/few-shot learning function can complete model training with only 1-20 good products, improving the efficiency of establishing the detection model. In addition, the semantic false alarm filtering function further reduces the false alarm rate and improves the accuracy of detection.
The 2D AI AOI equipment brings an efficient and accurate solution for SMT solder joint inspection with high-resolution imaging and deep learning algorithms.
Quantitative results: After adopting WeLinkirt's solution, significant quantitative results have been achieved. First of all, the detection rate of solder joint defects has increased from about 98% to 99.4%, and the missed detection rate has been reduced to <0.6%. Secondly, the false alarm rate has been reduced by -63%, from about 30% to about 11.1%, greatly reducing the workload of manual re-inspection. In addition, the model change time has been shortened from 30 minutes to 5 minutes, improving the production flexibility and efficiency of the production line.