0.3mm-level Surface Defect Detection for Stamped Sheet Metal: Reducing the Missed Detection Rate from 2.8% to 0.2%

The surface quality of stamped sheet metal parts directly affects the appearance quality of the whole vehicle. However, it is difficult to accurately detect tiny surface defects. WeLinkirt's DaoAI AI-AOI technology provides an effective solution to this problem. By optimizing the imaging and lighting schemes and using deep-learning models, it can stably detect 0.3mm - level subtle defects.

In industries such as automobile manufacturing, stamped sheet metal parts are indispensable and important components. Most of these stamped sheet metal parts are visible exterior coverings, such as car doors, hoods, and body side panels. Their surface quality directly determines the appearance quality of the whole vehicle. During the production process, various surface defects may occur on stamped sheet metal parts, such as scratches, dents, indentations, and pockmarks. Although these defects are tiny in size, they can have a serious impact on the appearance and quality of the product. Therefore, high-precision surface defect detection of stamped sheet metal parts is a key link to ensure product quality and corporate reputation.

Deep Dive into Pain Points: Why Is It Difficult?

From a quantitative perspective, there are many difficulties in detecting defects in stamped sheet metal parts. Firstly, the defect scale is extremely small, usually only a few tenths of a millimeter. For example, common scratches may be only 0.3mm or even smaller. Such tiny defects are very easy to be overlooked during the detection process. Secondly, the contrast is low. Since most stamped sheet metal parts are made of metal and have a reflective surface, it is very difficult for defects to form an obvious contrast with the normal area on the metal reflective surface, which further increases the detection difficulty. Moreover, the efficiency of manual visual inspection is low. Prolonged inspection work can easily cause fatigue to inspectors, leading to a significant increase in the missed-judgment rate. The actual measured missed-detection rate of a body parts factory is as high as 2.8%, which means that 28 defective products out of every 1000 may flow downstream or even to the client side.

The root cause of the difficulty in detecting these defects lies in their physical characteristics and the complexity of the detection environment. Tiny defects are almost imperceptible on the metal reflective surface, just like looking for an extremely subtle scratch on a mirror. It is very difficult for the naked eye to capture the signs of their existence. Moreover, the production environment of stamped sheet metal parts is usually complex. Factors such as light and dust can interfere with the detection results, further increasing the detection difficulty. In addition, manual inspection is highly subjective. There are differences in the inspection standards and judgment abilities of different inspectors, which also lead to frequent missed and false detections.

Technical Principle

DaoAI AI-AOI surface defect detection technology has been deeply optimized for metal reflective surfaces. In terms of imaging, a special optical imaging system is used. This system can effectively suppress the reflection on the metal surface, enhance the contrast between the defect and the normal area, and make tiny defects more clearly presented. At the same time, with a high-precision image sensor, it can capture images of 0.3mm - level subtle defects. In the lighting scheme, multi-angle and multi-spectral lighting technology is adopted. By irradiating light from different angles and spectra, the features of the defects are highlighted, and the recognizability of the defects is improved.

At the algorithm level, DaoAI AI-AOI uses an advanced deep-learning model. This model is trained with a large number of on - site samples and can accurately identify various types of defects such as scratches, dents, indentations, and pockmarks. During the training process, the model continuously learns and optimizes, and accurately extracts and classifies the features of different types of defects. Compared with traditional detection methods, traditional methods often rely on fixed rules and thresholds for detection, and have poor detection effects on tiny and complex defects. In contrast, the deep-learning model has stronger adaptability and flexibility, and can automatically adjust the detection strategy according to different detection scenarios and defect types, effectively improving the accuracy and stability of detection.

Typical Application Scenarios

• Surface inspection after the stamping process: Scratches and indentations may occur during the stamping process. When detecting, DaoAI AI-AOI uses the optimized imaging and lighting scheme to clearly capture the surface image. The difficulty lies in that there may be impurities such as oil stains on the surface after stamping, which affects the image quality and defect recognition.
• Dent detection after the forming process: Shallow dents may appear during the forming process. The system highlights the shadow features of the dents through multi-angle lighting and uses the deep-learning model for recognition. The difficulty is that the contrast of shallow dents is extremely low, and it is difficult to distinguish them from the normal surface.
• Pockmark detection after the welding process: Pockmarks may be generated at the welding site. By using multi-spectral lighting, the spectral differences between pockmarks and the surrounding area are highlighted to achieve accurate detection. The difficulty lies in the complex surface morphology of the welding area and many interference factors.
• Detection of subtle scratches after the grinding process: Subtle scratches may be left after grinding. The high-precision image sensor and optimized imaging algorithm can capture these subtle defects. The difficulty is that the scale of subtle scratches is extremely small and easy to be overlooked.

Implementation Case

A medium-sized body parts factory mainly supplies stamped sheet metal parts to several automobile manufacturing enterprises. The factory has a large production scale, with a daily output of thousands of pieces. Before introducing the DaoAI AI-AOI surface defect detection system, the factory always used manual visual inspection. However, the missed-detection rate was as high as 2.8%, resulting in a large number of defective products flowing downstream. This not only increased the rework and claim costs but also seriously affected the enterprise's quality reputation.

During the implementation process, the technical team of WeLinkirt first conducted a detailed investigation and analysis of the factory's production environment and detection requirements, and customized the imaging and lighting scheme according to the actual situation. Then, the deep-learning model was trained and optimized with on - site samples to ensure that the model could accurately identify various defects in the factory's products. After a period of debugging and optimization, the DaoAI AI-AOI system was officially put into operation.

After the system went live, the factory achieved full-scale surface defect inspection for stamped exterior parts. The missed-detection rate at the client side dropped from 2.8% to 0.2%. The appearance complaints and claims were significantly reduced, and the quality stability and delivery confidence were improved simultaneously.

WeLinkirt's Solution and Product

WeLinkirt's DaoAI AI-AOI surface defect detection solution is a complete system specifically designed for the surface defect detection of stamped sheet metal parts. The system includes an optimized imaging device, an advanced lighting system, and a powerful deep-learning algorithm model. The imaging device can capture high-precision images, providing an accurate data basis for defect detection. The lighting system enhances the feature presentation of defects through multi-angle and multi-spectral lighting. The deep-learning algorithm model can accurately identify and classify defects in the images.

This product has a high degree of flexibility and scalability, and can be customized according to the production needs and detection standards of different enterprises. At the same time, the system also supports continuous model iteration and optimization. With the continuous accumulation of on - site samples, the detection accuracy and stability of the model will be continuously improved, providing enterprises with a long-term and reliable surface defect detection solution.

Quantitative Results

By using the DaoAI AI-AOI surface defect detection system, the enterprise has achieved significant quantitative results. The detection accuracy has reached 97%, which means that the system can accurately identify most defective products. The missed-detection rate has dropped from 2.8% to 0.2%, greatly reducing the risk of defective products flowing to the client side. At the same time, the false-detection rate has been effectively controlled, reducing the ineffective rework caused by false detections and improving production efficiency. In addition, the appearance complaints and claims have been significantly reduced, and the enterprise's quality stability and delivery confidence have been greatly improved, winning an advantage for the enterprise in market competition.

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.