中国石油网消息：6月22日，在兰州石化3.5万吨/年特种丁腈橡胶装置生产A线上，一块崭新的橡胶产品刚刚下线。AI智能验胶系统瞬间开始工作：拍照、翻转、识别、记录……10秒就完成了一块橡胶的外观质量检验。“过去，我们要盯着每块胶仔细看，一天工作下来，眼睛都花了。现在有了智能验胶系统，轻松多了。”兰州石化橡胶部丁腈二区域员工李延军笑着说。

China Petroleum News: On June 22nd, on the production line A of the 35,000-ton/year special nitrile rubber unit at Lanzhou Petrochemical, a brand-new rubber product had just rolled off the line. The AI intelligent rubber inspection system immediately began to operate: taking pictures, flipping, recognition, recording... It completed the appearance quality inspection of a piece of rubber in just 10 seconds. "In the past, we had to closely examine each piece of rubber carefully. After a day of work, our eyes would get tired. Now with the intelligent rubber inspection system, it's much easier." Li Yanjun, an employee from the nitrile two area of the rubber department at Lanzhou Petrochemical, said with a smile.

AI智能验胶系统基于昆仑视觉大模型L3级别场景模型研发而成，集算法模型、系统集成、应用系统三大模块于一身，涵盖4个一级业务功能和10个二级业务功能。这是昆仑大模型首次在智能验胶系统应用，在深度适配石油行业场景应用的同时，保障了数据的安全性和合规性。

The AI intelligent rubber inspection system is developed based on the L3-level scene model of Kunlun Vision's large-scale model. It integrates three modules: algorithm model, system integration, and application system, covering 4 primary business functions and 10 secondary business functions. This is the first time that the Kunlun large-scale model has been applied in the intelligent rubber inspection system. While deeply adapting to the application scenarios of the petroleum industry, it also ensures the security and compliance of data.

系统应用前期，兰州石化技术团队收集了大量橡胶产品在各类光线、角度、环境下的图像和视频场景信息，经过数据准备、清洗、标注、检查与预处理，最终搭建出专属于丁腈橡胶缺陷识别的AI大模型。

In the early stage of system application, the technical team of Lanzhou Petrochemical collected a large amount of image and video scene information of rubber products under various light conditions, angles, and environments. After data preparation, cleaning, annotation, inspection and preprocessing, an AI large model specifically for the defect recognition of nitrile rubber was finally built.

当橡胶产品进入检测区，高清相机会立刻拍照，同时协同机械臂完成翻胶动作，确保块状橡胶六面检测全覆盖。经过图像分析，精准识别出那些肉眼极易错过的毫米级微小缺陷，如气孔、裂纹、杂质和颜色不均等。一旦发现问题，系统会立即发出声光警报，并将缺陷胶块自动推送至复检区，交由质检人员处理；合格品直接放行流转。整个过程无须人工搬运，大大减轻了质检人员的劳动强度。

When the rubber products enter the inspection area, the high-definition camera will immediately take a photo, and at the same time, the mechanical arm will cooperate to complete the flipping action to ensure that all six sides of the block-shaped rubber are inspected. Through image analysis, the system can accurately identify those minute defects that are easily overlooked by the naked eye, such as pores, cracks, impurities and color unevenness. Once a problem is detected, the system will immediately issue an audible and visual alarm and automatically push the defective rubber block to the re-inspection area, where it will be handled by the quality inspectors. The qualified products will be directly released for circulation. The entire process does not require manual handling, significantly reducing the labor intensity of the quality inspectors.

“自2025年5月投用以来，处理量稳定在135块/小时以上，缺陷胶的识别率和剔除率都达到100%，运行效果很理想。”兰州石化橡胶部丁腈二区域生产组副组长崔国锋自豪地介绍。

"Since its commissioning in May 2025, the processing capacity has remained above 135 pieces per hour. The identification rate and elimination rate of defective rubber have both reached 100%, and the operation effect is very satisfactory." Cui Guofeng, the deputy team leader of the Butadiene-Nitrile Two Area Production Team of the Rubber Department of Lanzhou Petrochemical, proudly introduced.

AI智能验胶系统不仅改变了丁腈橡胶生产线的质检方式，也为整个工业流程探索出可复制、可推广的模板。在橡胶、塑料、金属压延等表面缺陷检测需求突出的领域，可以直接移植，只需适配特定产品的缺陷特征和数据训练，就能快速部署应用。

The AI intelligent rubber inspection system not only revolutionized the quality inspection method of the nitrile rubber production line, but also provided a replicable and scalable template for the entire industrial process. In fields with prominent demands for surface defect detection, such as rubber, plastic, and metal extrusion, it can be directly applied. All that is needed is to adapt to the specific defect characteristics and data training of the products, and then it can be quickly deployed and put into use.

值得一提的是，随着大模型持续迭代，质量检验精度也不断提升。同时，AI智能验胶系统能及时保存质检信息，为橡胶产品质量分析提供了良好数据支撑。

It is worth noting that as large models continue to be refined, the accuracy of quality inspection has also been continuously improved. At the same time, the AI intelligent rubber inspection system can promptly save the inspection information, providing excellent data support for the analysis of rubber product quality.