Pole Tab Welding Quality Gate: Keep Burrs and Cold Welds Away from Thermal Runaway

In the production of new-energy batteries, the quality of pole tab welding directly affects the safety and reliability of battery cells. DaoAI provides an effective solution to the problem of pole tab welding detection with its unique AI-AOI automatic quality gate technology.

Scenario: In the current booming new-energy battery industry, the battery cell is the core component of the battery, and its quality is directly related to the performance and safety of the battery. As the channel for the internal current of the battery cell to flow out, the welding quality between the pole tab, the pole piece, and the cover plate is the key factor determining the current density distribution and the long-term reliability of the battery. Once there are problems such as burrs and cold welds in the pole tab welding, during the use of the battery, these defects may pierce the diaphragm to form an internal short-circuit or cause local overheating, which in turn may induce thermal runaway, bringing serious safety hazards to the use of the battery.

Pain Points: Why Is It Difficult?

From a quantitative perspective, traditional pole tab welding detection methods have many problems. Firstly, the missed detection rate is high. According to statistics, the missed detection rate of traditional visual inspection can reach 5% -10%, which means that 5-10 battery cells with welding defects may flow into the subsequent production process out of every 100 battery cells. Secondly, the false-positive rate is high, reaching 8% -15%. A large number of false positives not only increase the workload of manual re-checking but also reduce production efficiency. Thirdly, the model-changing time is long. When changing the welding detection of different models of pole tabs, the model-changing time of traditional methods may be as long as 30-60 minutes, seriously affecting the flexibility of production.

So, why is pole tab welding detection so difficult? On the one hand, customers do not want to make large-scale modifications to the existing welding production line, which limits the installation and use of detection equipment. On the other hand, the surface of the weld is a highly reflective metal, which makes it easy for traditional visual inspection to be interfered with by reflection during imaging, resulting in blurred images and difficulty in accurately identifying defects. Moreover, the burrs are small in size, usually in the micron-level, and the cold-weld features are hidden. Traditional visual inspection is difficult to capture these subtle changes, thus prone to false positives and missed detections.

Technical Principle

DaoAI's AI-AOI detection unit adopts a compact structure and can be directly embedded in the original workstation without major modifications to the existing production line. For highly reflective metal welds, it designs a multi-angle lighting and imaging system. Through lighting from different angles, the reflection can be effectively suppressed, making the image of the weld surface clearer for subsequent defect identification. Combined with a deep-learning model, the system can classify and judge key defects related to thermal runaway, such as burrs, misalignments, and cold welds. The deep-learning model has been trained with a large number of samples and can accurately identify different types of defect features, greatly improving the accuracy of detection.

Compared with traditional methods, DaoAI's technology has obvious advantages. Traditional visual inspection mainly relies on preset rules and features for judgment and is difficult to accurately identify complex defects and subtle changes. DaoAI's deep-learning model has a powerful learning ability and can continuously adapt to new defect features and welding morphologies. In addition, DaoAI also uses APDT positive-sample learning technology, which can adapt to the welding morphologies of different models of pole tabs. When changing models, it can quickly adjust the detection parameters, and the model-changing time can be shortened to less than 5 minutes, greatly improving the flexibility and efficiency of production.

Typical Application Scenarios

• Burr detection: The burrs are very small in size, usually between 10-50 microns. During detection, DaoAI's multi-angle lighting system can illuminate the weld from different angles, making the burrs more obvious in the image. The deep-learning model can accurately identify burrs by learning and analyzing the shape, size, and position of the burrs. The difficulty lies in how to clearly capture the tiny burrs on the highly reflective weld surface and distinguish the burrs from the normal weld texture.
• Cold - weld detection: The features of cold welds are hidden and difficult to observe directly with the naked eye. DaoAI's system analyzes the gray-scale value, texture, and other features of the weld area and uses the deep-learning model to judge whether there are cold welds. The difficulty lies in that the features of cold welds are not obvious and are easily confused with normal welding areas, requiring the model to have high sensitivity and accuracy.
• Pole - tab misalignment detection: Pole - tab misalignment can change the current path and cause local overheating. DaoAI uses image-recognition technology to compare the actual position of the pole tab with the standard position to judge whether the pole tab is misaligned. The difficulty lies in how to accurately determine the standard position of the pole tab and accurately identify the edge of the pole tab under different lighting conditions.
• Uneven weld detection: Uneven welds may lead to uneven current distribution and affect the performance of the battery. DaoAI's system analyzes the gray-scale value distribution of the weld to judge whether the weld is uniform. The difficulty lies in how to distinguish the normal gray-scale change of the weld from the uneven defects.

Implementation Case

A medium-sized battery cell factory with an annual production capacity of about 50 million battery cells on its existing welding production line mainly relies on sampling inspection and manual re-checking for pole tab welding detection, which is difficult to meet the customers' requirements for zero missed detection. When introducing DaoAI's AI-AOI automatic quality gate system, the implementation process was relatively smooth. The DaoAI team carried out compact equipment installation and debugging at the original workstation, minimizing the impact on the existing production line. Before the implementation, the missed detection rate of the factory was about 8%, the false-positive rate was about 12%, and the model-changing time was about 45 minutes. After the implementation, the missed detection rate was reduced to less than 1%, the false-positive rate was reduced to less than 3%, and the model-changing time was shortened to less than 5 minutes.

The quality gate only asks one question: Can this battery cell be put into the battery pack?

WeLinkirt's Solution and Product

WeLinkirt's DaoAI product takes the AI-AOI automatic quality gate as the core and provides a comprehensive solution for pole tab welding detection. The product embeds the detection unit at the original workstation to achieve 100% automatic judgment for each battery cell. It optimizes the lighting and imaging for highly reflective metal welds to suppress false positives caused by reflection. The deep-learning classification technology can accurately identify key defects such as burrs, misalignments, and cold welds. The APDT positive-sample learning technology enables the system to quickly adapt to multiple models of pole tabs, making the model-changing process more efficient. Unqualified products can be diverted in real-time to ensure that only qualified battery cells enter the subsequent production process.

Quantitative Results: The DaoAI system upgrades the original sampling inspection relying on manual labor to an automatic gate for piece - by - piece full inspection. The detection rate of key defects approaches over 99%, the missed detection rate is reduced to less than 1%, the false-positive rate is reduced by -9%, and the model-changing time is shortened from the original 30-60 minutes to less than 5 minutes. It reduces the safety risks related to thermal runaway from the source and greatly improves production efficiency and product quality.

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.