In the manufacturing of automotive parts, the surface quality of gears and shafts is of crucial importance. WeLinkirt provides an efficient surface defect detection solution for automotive parts manufacturers with advanced AI vision technology.
User Scenario: On the gear/shaft production line of a leading automotive parts manufacturer, various specifications of automotive gears and shafts are mainly produced. The detection objects are the surfaces of these gears and shafts, and the defects to be detected include surface scratches, cracks, sand holes, and impact damages. These defects may affect the service life and performance of gears and shafts, and further affect the overall quality of the vehicle.
Pain Points: The traditional manual inspection method is inefficient, with high labor costs, and is prone to missed detections and false alarms. Within the production line cycle, it is difficult for manual inspection to ensure 100% full inspection, and the missed detection rate is about 3%. At the same time, the false alarm rate is also relatively high, reaching 15%, which leads to a large number of products needing to be re-inspected, increasing production time and costs. In addition, when the specifications of the products being produced change, manual inspection requires retraining workers, and the change-over time is as long as 30 minutes, seriously affecting production efficiency.
Technical Principles
WeLinkirt uses advanced AI algorithms and imaging technologies to solve these problems. In terms of algorithms, a deep-learning - based visual foundation model is used, which has a powerful feature recognition ability. Through learning from a small number of positive samples (1-20), the model can quickly understand the features of good products and accurately distinguish between good products and defective products. In terms of imaging, the self-developed 3D camera can obtain the three-dimensional topography information of the gear and shaft surfaces, achieving high-precision imaging at the micron level. This is because 3D imaging can more comprehensively reflect the surface features of objects, and can also clearly detect defects hidden under the surface, such as sand holes. Through the three-dimensional topography reconstruction technology, the acquired 3D data can be processed and analyzed to accurately identify various types of defects.
- The deep-learning algorithm can automatically extract defect features, avoiding the limitations of traditional algorithms that require manual feature design, and improving the accuracy and generalization ability of detection.
- The high-precision imaging of the 3D camera can capture tiny surface defects, and can clearly image defects at the micron level, greatly improving the detection accuracy.
- The semantic false-alarm filtering technology can filter false alarms according to the semantic information of defects, reducing the false-alarm rate.
WeLinkirt Solutions and Product Introduction
WeLinkirt provides the DaoAI AI AOI software system and the DaoAI 2D / 3D AI AOI equipment. The DaoAI AI AOI software system has the ability of rapid programming. It can complete zero-code automatic programming for a good product in 5 minutes, and supports APDT positive-sample/few-sample learning. Only 1-20 good products are needed for model training. The system also has a semantic false-alarm filtering function, which can effectively reduce false alarms. The DaoAI 2D / 3D AI AOI equipment uses a self-developed 3D camera to achieve three-dimensional topography reconstruction, which can detect hidden solder joints, coplanarity, and micron-level topography. During the implementation process, the equipment is first installed on the production line to conduct on - line inspection of gears and shafts. The software system analyzes and processes the collected images and data to judge in real-time whether the products have defects.
WeLinkirt's solution effectively improves the efficiency and accuracy of automotive parts surface defect detection.
Quantifiable Results: After using WeLinkirt's solution, the detection rate has been increased to 99.4%, and the missed-detection rate has been reduced to <0.6%. The false-alarm rate has been reduced by -63%, greatly reducing the re-inspection volume. At the same time, the product change-over time has been shortened from 30 minutes to 5 minutes, significantly improving production efficiency.