AI AOI Software System Enables Defect Detection on Textured Surfaces of Chemical Materials

With the advancement of industrial intelligence, the chemical materials industry has increasingly high requirements for product quality inspection. WeLinkirt's AI AOI software system plays a crucial role in the defect detection of textured surfaces of chemical materials, bringing new changes to the industry.

In the current trend of industrial development, intelligent manufacturing has become the goal pursued by various industries. The chemical/material industry, as a basic industry, its product quality directly affects many downstream fields. In the production process of chemical materials, materials with textured surfaces are widely used. For example, in the construction field, these materials can be used for wall decoration and floor paving; in the automotive industry, they can be used for interior and exterior parts. On the production line of a leading chemical material manufacturer, a large number of such chemical materials with textured surfaces are produced every day. To ensure that the product quality meets the standards, defect detection of the textured surfaces of the materials is required during the production process. The detection objects include various abnormal situations such as surface scratches, holes, and impurities.

Pain Points: Why Is It Difficult?

Traditional detection methods face many difficulties in the defect detection of textured surfaces of chemical materials. From the efficiency dimension, manual detection is extremely inefficient, and each detection cycle takes up to several hours. This is because manual detection requires inspectors to concentrate for a long time and check the material surface point by point, and human visual fatigue will intensify with the increase of detection time, further reducing the detection speed. From the cost dimension, the labor cost is high. Enterprises need to hire a large number of inspectors and provide them with training and welfare. At the same time, the accuracy of manual detection also has big problems. The missed detection rate is as high as 5%, which means that a considerable number of defective products may flow into the market, bringing potential quality risks to the enterprise. The false alarm rate also reaches 8%. Excessive false alarms will lead to unnecessary re-inspection work, wasting a lot of time and resources. In addition, with the increase of product types, the problem of long change-over time becomes more prominent. Traditional detection methods need to reset detection parameters and standards for different types of materials, and this process often takes professional personnel several hours or even days to complete, seriously affecting production efficiency. The root cause of these problems is that traditional detection methods mainly rely on human vision and experience, making it difficult to accurately analyze complex textured surfaces and quickly adapt to product diversification.

Take the tiny scratches on the textured surface as an example. It is very difficult for manual detection to find these subtle defects because the resolution of the human eye is limited, and the judgment standards of different inspectors also vary. For holes and impurities, they are also easily ignored or misjudged in the complex texture background. This makes traditional detection methods difficult to meet the requirements of high-quality production in the defect detection of textured surfaces of chemical materials, restricting the improvement of intelligent manufacturing level.

Technical Principle

WeLinkirt's AI AOI software system adopts the advanced feature recognition technology of the visual foundation model. This technology can accurately identify various features of the textured surface through in - depth analysis of a large amount of image data. Its principle is to train the model with a large amount of normal texture image data, so that the model can learn the characteristic patterns of normal textures. During the detection process, the system will compare the real-time collected images with the learned normal patterns, and once a difference is found, it is judged that there may be an abnormality. Compared with traditional methods, traditional methods mainly rely on fixed rules and templates for detection, which are often powerless for complex textured surfaces and tiny defects. The feature recognition technology of the visual foundation model can analyze the image from multiple dimensions, including the direction, density, and gray level of the texture, so as to achieve accurate detection of tiny defects.

In addition, the system also adopts the APDT positive sample/few-sample learning method. Traditional machine learning methods usually require a large amount of sample data for training, which not only increases the time and cost of sample collection, but also makes it difficult to obtain enough samples in practical applications. The APDT positive sample/few-sample learning method can quickly learn the characteristics of normal textures with only 1-20 good samples, greatly reducing the time and cost of sample collection. At the same time, the system also has a semantic false alarm filtering mechanism, which can perform a second screening of the detection results and remove false alarms caused by environmental interference and other factors, improving the detection accuracy. For example, in the production environment, changes in light and the influence of dust may cause false alarms. The semantic false alarm filtering mechanism can identify these interference factors and exclude them from the detection results.

Typical Application Scenarios

• Surface scratch detection: Scratches may appear as linear abnormalities on the textured surface. The system can accurately identify the position and length of scratches by analyzing the continuity and direction of the texture. The difficulty lies in that the scratches may be very subtle, with a small difference from the normal texture, and the system needs to have high-resolution image analysis capabilities.
• Hole detection: Holes appear as black circular or irregular shapes in the image. The system can detect the size and position of holes by analyzing the gray value of the image. The difficulty lies in that the complexity of the texture may mask the features of the holes, and the system needs to be able to extract the information of the holes from the complex background.
• Impurity detection: Impurities may have different colors and shapes, forming an obvious contrast with the surrounding texture. The system can identify the presence of impurities by analyzing the color and shape. The difficulty lies in that the size and color of impurities may be similar to the normal texture, and the system needs to have high-precision feature extraction capabilities.
• Texture inhomogeneity detection: Texture inhomogeneity is manifested as changes in the density and direction of the texture in local areas. The system can detect the areas of texture inhomogeneity by analyzing the statistical features of the texture. The difficulty lies in how to accurately define the normal variation range of the texture and avoid misjudgment.

Implementation Case

A large-scale chemical material production enterprise with a large production scale produces a large number of chemical materials with textured surfaces every day. Before introducing WeLinkirt's AI AOI software system, the enterprise had been using traditional manual detection methods and faced problems such as low detection efficiency, high missed detection rate, high false alarm rate, and long change-over time. During the implementation process, WeLinkirt's technical team first conducted in - depth research on the enterprise's production process and detection requirements, and then customized the system according to the actual situation of the enterprise. After a period of debugging and optimization, the system was officially put into operation.

After applying WeLinkirt's AI AOI software system, the detection effect of the enterprise has been significantly improved.

WeLinkirt's Solution and Product

WeLinkirt's AI AOI software system is the core to solve the problem of defect detection on textured surfaces of chemical materials. The system has unique capabilities: First, it supports 0-code automatic programming in 5 minutes with one good sample, greatly shortening the programming time and improving the detection efficiency. This means that even operators without professional programming knowledge can quickly complete the system setup and adjustment. Second, through APDT positive sample/few-sample learning, the model can be trained with only a small number of good samples, quickly adapting to the detection of different types of materials. The semantic false alarm filtering function effectively reduces the false alarm rate and improves the detection reliability. In addition, the system supports 100% local privatized deployment of SDK/API/Docker to ensure data security and meet the enterprise's requirement of data non-leaving the factory. At the same time, WeLinkirt's DaoAI 2D/3D AI AOI equipment can be used as a supporting device to provide more accurate image data for detection.

Quantitative results: After applying WeLinkirt's AI AOI software system, the detection effect of the manufacturer has been significantly improved. The detection rate has reached 98%, and the missed detection rate has been reduced to <2%, greatly improving the product quality. The false alarm rate has been reduced by -65%, reducing unnecessary re-inspection work. The change-over time has been shortened from several hours to 5 minutes, greatly improving the production efficiency.

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.