首页|期刊导航|焊管|输送管焊缝缺陷动态检测技术的应用与探索

输送管焊缝缺陷动态检测技术的应用与探索OA

Application and Exploration of Dynamic Detection Technology for Weld Defects of Line Pipe

中文摘要英文摘要

针对油气输送管道焊缝缺陷检测的实时性与准确性需求,提出一种基于机器视觉的X射线数字动态成像(DR)人工智能识别方法.通过PCSDK技术采集平板探测器输出的原始图像数据,建立具有图像预处理及识别算法的视觉模型,对采集帧率、算力分配及图像分辨率进行协同优化,在保证检测精度的前提下实现系统资源的高效利用.采用动态视频序列抽帧技术构建缺陷语料库,结合改进型卷积神经网络(CNN)及时序特征融合策略实现焊缝裂缝、气孔等典型缺陷的智能识别.系统应用基于B/S架构,支持多并发识别,与专业评片软件协同作业,具备持续迭代功能.实际应用表明,当配置为 512×512像素分辨率、50帧/s帧率时,系统漏检率低于 5%,满足工业应用检测标准.该系统具有良好的工程适应性与推广应用前景,为工业管道焊缝缺陷的智能化检测提供了新的技术方案.

To meet the real-time and accurate detection requirements for weld defects in oil and gas pipelines,an artificial intelligence(AI)recognition method based on machine vision for X-ray digital radiography(DR)is proposed.The original image data output by the flat panel detector is collected through PCSDK technology,and a visual model with image preprocessing and recognition algorithms is established.The acquisition frame rate,computing power allocation,and image resolution are optimized collaboratively to achieve efficient system resource utilization while ensuring detection accuracy.Dynamic video sequence frame extraction technology is used to construct a defect corpus and is combined with an improved convolutional neural network(CNN)and a temporal feature fusion strategy to achieve intelligent recognition of typical defects such as weld cracks and porosity.The system is based on a B/S architecture,supports multiple concurrent recognitions,works in collaboration with professional film evaluation software,and has continuous iteration capabilities.Practical applications show that,when configured with a resolution of 512×512 pixels and a frame rate of 50 frames per second,the system has a missed detection rate of less than 5%,meeting the detection standards for industrial applications.The system has good engineering adaptability and promotion prospects,providing a new technical solution for intelligent detection of weld defects in industrial pipelines.

胡良波

中国石油集团宝石管业有限公司,陕西 宝鸡 721008

矿业与冶金

自动缺陷识别(ADR)动态检测实时图像采集人工智能机器视觉

automatic defect recognition(ADR)dynamic inspectionreal-time image acquisitionartificial intelligence(AI)machine vision

《焊管》 2026 (9)

60-66,7

国家科技重大专项"深海油气输送用管线钢及耐蚀合金特殊螺纹油套管的研制及应用课题1:深海油气输送用抗疲劳耐蚀管线钢制造与应用关键技术"(项目编号2025ZD0611602).

10.19291/j.cnki.1001-3938.2026.09.008

评论