In the electronics industry, the quality inspection of passive components like MLCC is of crucial importance. Micro - cracks, as common defects, seriously affect the performance of components and the reliability of products. A leading electronics manufacturer faced problems such as low efficiency and low accuracy in micro - crack detection during the production process.
Industry background and customer pain points: In the electronics/PCBA industry, the production processes of passive components like MLCC have extremely high requirements for product quality. Especially in the component molding and packaging processes, micro - cracks are common and difficult - to - detect defects. Micro - cracks not only lead to unstable electrical performance of components but may also gradually expand during subsequent use, causing product failures. In the traditional inspection process, a leading electronics manufacturer mainly relied on manual visual inspection and traditional AOI equipment. Manual visual inspection is inefficient, and visual fatigue is likely to occur after long - term work, resulting in frequent missed and false inspections. Traditional AOI equipment has limited detection accuracy for micro - cracks, which cannot meet the needs of high - quality production. Meanwhile, the long model - changing time increases production costs and production cycles.
Technical principles
DaoAI uses advanced AI-AOI technology to solve the problem of micro - crack detection. In terms of imaging, high - resolution 2D/3D imaging equipment is used to capture the subtle features on the surface and inside of components. At the algorithm level, deep learning algorithms and the DaoAI World model are combined. The deep learning algorithm can accurately identify the features of micro - cracks by learning from a large number of micro - crack samples. Even micro - cracks at the micron level can be clearly distinguished. The DaoAI World model provides a general knowledge framework that can classify and analyze different types of micro - cracks, improving the accuracy and generalization ability of detection. In addition, the APDT positive - sample/few - sample learning technology only requires 1 - 20 good samples as positive samples to quickly train an efficient detection model, greatly reducing the workload of sample collection and annotation.
- High - resolution imaging: Use 2D/3D detection equipment to obtain clear component images for subsequent analysis.
- Deep learning algorithms: Identify crack features through learning and analysis of micro - crack samples.
- DaoAI World model: Provide a general knowledge framework to enhance the accuracy and generalization ability of detection.
- APDT positive - sample/few - sample learning: Reduce the workload of sample collection and annotation and quickly build a detection model.
- Semantic false - alarm filtering: Filter out non - crack interference information through image semantic analysis to reduce the false - alarm rate.
DaoAI's technology provides a high - precision and high - efficiency solution for the micro - crack detection of passive components like MLCC.
Solution implementation: To address the problems of this leading electronics manufacturer, DaoAI provided AI-AOI 2D/3D detection equipment and supporting AOI software. First, use high - resolution 2D/3D detection equipment to image passive components like MLCC and obtain detailed image information of the components. Then, use the deep learning algorithms and the DaoAI World model in the AOI software to analyze and process the images and identify micro - crack defects. At the same time, quickly complete the training and optimization of the model through the APDT positive - sample/few - sample learning technology. In addition, the software has the ability to change models in 5 minutes without coding. When producing components of different models, it can quickly adjust the detection parameters to meet new production needs.
Quantitative results: By using DaoAI's solution, this leading electronics manufacturer has achieved remarkable results. In terms of detection accuracy, the detection rate of micro - cracks has reached over 99%, and most micro - crack defects can be accurately identified. The false - alarm rate has been reduced by 80%, greatly reducing production delays and cost waste caused by false alarms. The missed - detection rate is controlled within 1%, effectively ensuring product quality. At the same time, the ability to change models in 5 minutes without coding significantly shortens the model - changing time, improves production efficiency, and reduces production costs.