融合动态评片知识AI模型的焊缝缺陷探测技术

doi:10.3901/JME.2025.06.024

机械工程学报 ›› 2025, Vol. 61 ›› Issue (6): 24-32.doi: 10.3901/JME.2025.06.024

• 仪器科学与技术 • 上一篇

扫码分享

融合动态评片知识AI模型的焊缝缺陷探测技术

姜洪权¹, 张毛杰¹, 程虎跃¹, 高建民¹, 姚欢², 高富超³, 郭晓峰³, 贺吉程¹

1. 西安交通大学机械制造系统工程国家重点实验室西安 710049;
2. 中国石油集团工程材料研究院有限公司西安 710077;
3. 国家管网集团西部管道有限责任公司乌鲁木齐 830013

收稿日期:2024-05-10 修回日期:2024-12-10 发布日期:2025-04-14
作者简介:姜洪权，男，1978年出生，博士，副教授，博士研究生导师。主要研究方向为质量大数据分析与应用技术，先进无损检测数据智能分析理论与方法。E-mail:hqjiang@mail.xjtu.edu.cn；张毛杰，男，2000年出生。主要研究方向为机器视觉与智能制造、无损检测与缺陷识别技术。E-mail:maojz7345@stu.xjtu.edu.cn；程虎跃(通信作者)，男，1998年出生，博士研究生。主要研究方向为机器视觉与智能制造、无损检测与缺陷识别技术。E-mail:3120101217@stu.xjtu.edu.cn
基金资助:
国家自然科学基金(52375513)、国家重点研发计划(2021YFF0602300)和国家天然气管网集团有限公司研究(20230382)资助项目。

Fusion of Dynamic Film Evaluation Knowledge AI Model for Weld Defect Detection

JIANG Hongquan¹, ZHANG Maojie¹, CHENG Huyue¹, GAO Jianmin¹, YAO Huan², GAO Fuchao³, GUO Xiaofeng³, HE Jicheng¹

1. State Key Laboratory for Manufacturing Systems Engineering, Xi'an Jiaotong University, Xi'an 710049;
2. Tubular Goods Research Institute of CNPC, Xi'an 710077;
3. West Pipeline Company of PipeChina, Urumqi 830013

Received:2024-05-10 Revised:2024-12-10 Published:2025-04-14

摘要/Abstract

摘要： 当前焊缝缺陷检测的方法本质上属于基于单张图像的“行业数据集+AI模型”范式，缺乏对无损检测领域知识的集成，对于大量对比度低、目标不显著的缺陷容易造成漏检、误检。针对这一问题，提出了一种融合无损检测领域动态评片知识的焊缝缺陷目标探测技术，即“行业数据集+AI模型+领域知识”的新范式。首先，基于人工探查缺陷“动态”过程知识，提出了一种基于对数变换的多图分解方法，实现单张图像“静态”分析向多张图像“动态”分析转变；其次，通过设计通道注意力机制模块，实现了多图特征的浅层融合；最后，以YOLOX为基础网络，并结合双向特征金字塔网络(Bidirectional feature pyramid network,BiFPN)，实现了多图特征的深度融合和缺陷目标探测。利用某企业管道环焊缝数据进行验证，结果表明所提方法的缺陷探测mAP达到了96.72%，相比于YOLOX提升了6.68%，尤其对未熔合和未焊透的探测能力有显著提升，提升了AI技术在无损检测领域的应用能力。

关键词: 焊缝缺陷, 射线图像, 缺陷探测, AI模型, 无损检测

Abstract: The current methods for detecting weld defects essentially belong to the paradigm of ‘industry dataset + AI model’ based on a single image, lacking integration of knowledge in the field of non-destructive testing. For a large number of defects with low contrast and insignificant targets, it is easy to cause missed and false detections. Aiming at this problem, a new paradigm of ‘industry dataset + AI model + domain knowledge’ is proposed for weld defect target detection technology that can integrate dynamic evaluation knowledge in the field of non-destructive testing. Firstly, based on the knowledge of the ‘dynamic’ process of manually detecting defects, a multi graph decomposition method based on logarithmic transformation is proposed to transform the ‘static’ analysis of a single image into the ‘dynamic’ analysis of multiple images; Secondly, by designing a channel attention mechanism module, shallow fusion of multi-image features was achieved; Finally, based on YOLOX network and combined with bidirectional feature pyramid network (BiFPN), deep fusion of multi graph features and defect target detection were achieved. Using data from a certain enterprise’s pipeline circumferential weld seam for verification, the results show that the proposed method achieves a defect detection mAP of 96.72%, which is 6.68% higher than YOLOX. Especially, it significantly improves the detection ability for incomplete fusion and incomplete penetration, enhancing the application ability of AI technology in non-destructive testing.

Key words: weld defects, X-ray images, defect detection, AI model, non-destructive testing

中图分类号:

TP391

姜洪权, 张毛杰, 程虎跃, 高建民, 姚欢, 高富超, 郭晓峰, 贺吉程. 融合动态评片知识AI模型的焊缝缺陷探测技术[J]. 机械工程学报, 2025, 61(6): 24-32.

JIANG Hongquan, ZHANG Maojie, CHENG Huyue, GAO Jianmin, YAO Huan, GAO Fuchao, GUO Xiaofeng, HE Jicheng. Fusion of Dynamic Film Evaluation Knowledge AI Model for Weld Defect Detection[J]. Journal of Mechanical Engineering, 2025, 61(6): 24-32.

参考文献

[1] WANG Xiaopeng，ZHANG Baoxin，CUI Jinhan，et al. Image analysis of the automatic welding defects detection based on deep learning[J]. Journal of Nondestructive Evaluation，2023，42(3)：82．
[2] 吴斌，杨敬，焦敬品，等. 奥氏体不锈钢小径管焊缝缺陷多模式超声复合全聚焦成像研究[J]. 机械工程学报，2020，56(14)：9-18. WU Bin，YANG Jing，JIAO Jingpin，et al. Study on ultrasonic multi-mode composite total focusing imaging method for welds defect inspection in smal-diameter tube of austenitic stainless steel[J]. Journal of Mechanical Engineering，2020，56(14)：9-18.
[3] ZHANG Baoxin，WANG Xiaopeng，CUI Jinhan，et al. Welding defects classification by weakly supervised semantic segmentation[J]. NDT and E International，2023，138：102899.
[4] YANG Wei，XIAO Yancai，SHEN Haikuo，et al. Generalized weld bead region of interest localization and improved faster R-CNN for weld defect recognition[J]. Measurement，2023，222：113619.
[5] 姜洪权，贺帅，高建民，等. 一种改进卷积神经网络模型的焊缝缺陷识别方法[J]. 机械工程学报，2020，56(8)：235-242. JIANG Hongquan，HE Shuai，GAO Jianmin，et al. An improved convolutional neural network for weld defect recognition[J]. Journal of Mechanical Engineering，2020，56(8)：235-242.
[6] ALAKNANDA，ANAD R S，KUMAR P. Flaw detection in radiographic weldment images using morphological watershed segmentation technique[J]. NDT & E International，2009，42(1)：2-8.
[7] ZAPATA J，VILAR R，RUIZ R. Performance evaluation of an automatic inspection system of weld defects in radiographic images based on neuro-classifiers[J]. Expert Systems With Applications，2011，38(7)：8812-8824.
[8] GAMDHA D，UNNIKRISHNAKURUP S，ROSE K J J，et al. Automated defect recognition on x-ray radiographs of solid propellant using deep learning based on convolutional neural networks[J]. Journal of Nondestructive Evaluation，2021，40(1)：18.
[9] WANG Donghuan，XIAO Hong，HUANG Shengqin. Automatic defect recognition and localization for aeroengine turbine blades based on deep learning[J]. Aerospace，2023，10(2)：178.
[10] 程松，杨洪刚，徐学谦，等. 基于YOLOv5的改进轻量型X射线铝合金焊缝缺陷检测算法[J]. 中国激光，2022，49(21)：136-144. CHENG Song，YANG Honggang，XU Xueqian，et al. Improved lightweight X-ray aluminum alloy weld defects detection algorithm based on YOLOv5[J]. Chinese Journal of Lasers，2022，49(21)：136-144.
[11] 高翔，李楷，衣正尧，等. 基于改进YOLOv5模型的船舶焊缝缺陷检测方法[J]. 大连理工大学学报，2023，63(04)：385-392. GAO Xiang，LI Kai，YI Zhengyao，et al. Ship welding seam defect detection method based on improved YOLOv5 model[J]. Journal of Dalian University of Technology，2023，63(04)：385-392.
[12] KWON J E，PARK J H，KIM J H，et al. Context and scale-aware YOLO for welding defect detection[J]. NDT and E International，2023，139：102919.
[13] GUO Runyuan，LIU Han，XIE Guo，et al. Weld defect detection from imbalanced radiographic images based on contrast enhancement conditional generative adversarial network and transfer learning[J]. IEEE Sensors Journal，2021，21：10844-10853.
[14] HU Jie，SHEN Li，ALBANIE S，et al. Squeeze- and-excitation networks[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2018：7132-7141.
[15] ZHENG Ge，LIU Songtao，WANG Feng，et al. YOLOX：Exceeding YOLO series in 2021[EB/OL]. [2021-08-06]. https://arxiv.org/abs/2107.08430.
[16] REDMON J，DIVVALA S，GIRSHICK R，et al. You only look once：Unified，real-time object detection[C]// Proceedings of the lEEE Conference on Computer Vision and Pattern Recognition，2016：779-788.
[17] ZHAO Zoupeng，HE Chen，ZHAO Guangming，et al. RA-YOLOX: Re-parameterization align decoupled head and novel label assignment scheme based on YOLOX[J]. Pattern Recognition，2023，140：109579.
[18] TAN Mingxing，PANG Ruoming，LE Q V. EfficientDet: Scalable and efficient object detection[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition，2020：10778-10787.
[19] 赵树煊，张洁，汪俊亮，等. 基于两阶段深度迁移学习的面料疵点检测算法[J]. 机械工程学报，2021，57(17)：86-97. ZHAO Shuxuan，ZHANG Jie，WANG Junliang，et al. Fabric defect detection algorithm based on two-stage deep transfer learning[J]. Journal of Mechanical Engineering，2021，57(17)：86-97.
[20] LU Yonghua，HUANG Yu，XU Jiajun，et al. Surface defect detection method of composite materials based on deep learning[J]. Transactions of Nanjing University of Aeronautics and Astronautics，2023，40(4)：487-499.

融合动态评片知识AI模型的焊缝缺陷探测技术

Fusion of Dynamic Film Evaluation Knowledge AI Model for Weld Defect Detection

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	胡朋, 赵瑞祥, 杨淑言, 段志荣, 解社娟, 陈振茂, 郑阳. 一种基于电磁-声复合检测方法的分时策略研究[J]. 机械工程学报, 2025, 61(2): 1-10.
[2]	张超, 王宏远, 季宏丽, 朱丽晨, 裘进浩. 金属-有机材料层合结构的脱粘损伤识别方法[J]. 机械工程学报, 2025, 61(2): 24-35.
[3]	梁鹏, 赵文卓, 武通海, 冉毅. 发电机转子小齿裂纹的超声回波特性及在机检测方法[J]. 机械工程学报, 2024, 60(8): 11-21.
[4]	涂洪铭, 伍剑波, 柯瑞, 夏慧, 陈笑天, MACIEJ Roskosz. 激励参数对Q235钢增量磁导率蝶形图的影响研究[J]. 机械工程学报, 2024, 60(2): 27-35.
[5]	何赟泽, 李响, 王洪金, 侯岳骏, 张帆, 牟欣颖, 刘浩, 程豪, 李时华, 李杰. 基于可见光和热成像的风机叶片全周期无损检测综述[J]. 机械工程学报, 2023, 59(6): 32-45.
[6]	沈功田, 金灿华, 张君娇, 胡斌, 刘渊, 王强. 无损云检测技术研究及仪器系统开发[J]. 机械工程学报, 2023, 59(4): 1-9.
[7]	朵文博, 李宏伟, 李帅兵, 杨栋, 卢保朋, 康永强, 曹炳磊. 基于太赫兹成像的层状复合绝缘结构内部分层缺陷SSA-CNN定量表征^*[J]. 电气工程学报, 2023, 18(3): 63-72.
[8]	沙经伟, 范孟豹, 曹丙花, 杨雪锋. 金属构件硬度的无损检测研究进展与展望[J]. 机械工程学报, 2023, 59(24): 1-17.
[9]	孙凤山, 范孟豹, 曹丙花, 刘林. 基于时频关键信息融合的热障涂层太赫兹准确测厚方法[J]. 机械工程学报, 2023, 59(14): 10-22.
[10]	陈国龙, 曹政, 张帅帅, 靳伍银. 基于科赫差测量平面涡流传感器的裂纹检测性能研究[J]. 机械工程学报, 2023, 59(10): 17-29.
[11]	王俊, 周正干, 王进, 刘雨生. 基于线阵和环阵换能器的CFRP层压板超声成像特性研究分析[J]. 机械工程学报, 2023, 59(10): 30-39.
[12]	黄松岭, 孙洪宇, 王珅, 赵伟, 彭丽莎. 压水堆核电站无损检测与状态监测研究综述[J]. 机械工程学报, 2022, 58(4): 1-13.
[13]	李伟, 邵鑫宇, 张伯莹, 袁新安, 殷晓康, 杨伟超. 交流电磁场和电磁超声复合无损检测技术研究[J]. 机械工程学报, 2022, 58(16): 153-159.
[14]	邓堡元, 何赟泽, 王洪金, 张宏, 杨渊, 马敏敏, 牟欣颖, 杨瑞珍. 光伏电池图像序列的深度学习检测方法[J]. 机械工程学报, 2021, 57(8): 98-106.
[15]	孙凤山, 范孟豹, 曹丙花, 叶波, 刘林. 基于几何纹理与Anscombe变换的蜂窝材料太赫兹图像降噪模型[J]. 机械工程学报, 2021, 57(22): 96-105.