SkyVision Platform Enables Dual Inspection of Screw Torque and Vision

In the industrial screw-tightening process, ensuring the accuracy of screw torque and visual inspection is crucial. WeLinkirt's SkyVision zero-code video surveillance AI platform provides an efficient and reliable solution for this process with advanced technology.

User scenario: In the assembly line of a leading industrial manufacturing company, there is a crucial process of screw-tightening. The product of this process is large-scale mechanical equipment, and the inspection objects are the tightening status of screws, including whether the torque meets the standard, whether the screw appearance is damaged, and whether the installation is correct, which belongs to the typical category of industrial SOP / operating specifications.

Pain points: In previous inspections, there were many quantitative difficulties. The missed-detection rate reached about 3%, which means that for every 100 products produced, about 3 products with screw-tightening problems may not be detected. The false-alarm rate was also relatively high, about 15%. A large number of false-alarm messages increased the workload of operators. At the same time, due to the lack of native unified technology, the accuracy of behavior recognition was greatly reduced, resulting in low detection efficiency and high labor costs. Moreover, when the product was changed, it took a lot of time to reset the detection parameters.

Technical principle

The SkyVision zero-code video surveillance AI platform uses advanced computer vision algorithms and deep-learning technologies. In terms of imaging, high-definition industrial cameras are used to capture the image and video information of screws. Its hardware principle is based on an edge box, which can process data in real-time and issue alarms.

• In terms of algorithms, semantic understanding is carried out through the DaoAI World model. This model has a unified base and can perform accurate semantic analysis on the screw-tightening behavior and status, thus improving the accuracy of behavior recognition.
• The deep-learning model is trained with a large amount of data and can achieve in - field training of its own model within hours to quickly adapt to different product and detection requirements.
• The real-time computing ability of the edge box ensures that data can be quickly processed locally, realizing 100% local data storage without leaving the factory, which guarantees data security and real-time detection.

WeLinkirt's solution and product

Centered on the SkyVision zero-code video surveillance AI platform, this platform has the ability to train its own model within hours on - site without complex programming. Operators can quickly train the model according to different product requirements. For the dual inspection of screw torque and vision, the platform can accurately identify the screw-tightening behavior and status. At the same time, combined with the real-time alarm function of the edge box, once the screw tightening does not meet the specifications, an alarm will be issued immediately. In addition, the semantic understanding ability of the DaoAI World model further improves the accuracy of detection. In terms of matching, it can be combined with the DaoAI AI AOI software system for assistance, using the feature recognition of its visual basic model to improve the detection accuracy of the screw appearance.

The SkyVision platform makes industrial screw-tightening inspection more accurate and efficient.

Quantitative results: By using the SkyVision platform, the missed-detection rate is reduced to <0.6%, and the detection rate reaches 99.4%. The false-alarm rate is reduced by -63%, greatly reducing the ineffective workload of operators. The product change-over time is shortened from several hours to 5 minutes, improving production efficiency.