Pipeline Girth Weld Defect Intelligent Identification of Digital Radiograph based on an Improved YOLOv5 Network

TANG Dayun; YUAN Jianwei; WU Di; HE Dongchang; ZHANG Peilei; PAN Bin; GONG Hao; JIN Hongfei

doi:10.7512/j.issn.1001-2303.2025.05.04

Important Issuse | Views : 0 下载量: 10 CSCD: 0

PDF
Export
Share
Collection
Album

Pipeline Girth Weld Defect Intelligent Identification of Digital Radiograph based on an Improved YOLOv5 Network
Vol. 55, Issue 5, Pages: 30-36(2025)
Received：29 August 2024，

Revised：2025-01-03，

Published：20 May 2025
DOI： 10.7512/j.issn.1001-2303.2025.05.04
稿件说明：

移动端阅览

汤大赟，袁健伟，吴頔，等.基于改进YOLOv5网络的管道环焊缝射线底片缺陷智能识别［J］.电焊机，2025，55（5）：30-36. DOI： 10.7512/j.issn.1001-2303.2025.05.04.

TANG Dayun, YUAN Jianwei, WU Di, et al.Pipeline Girth Weld Defect Intelligent Identification of Digital Radiograph based on an Improved YOLOv5 Network[J].Electric Welding Machine, 2025, 55(5): 30-36. DOI： 10.7512/j.issn.1001-2303.2025.05.04.

摘要

X射线数字辐射成像DR技术是环焊缝质量检测的必备手段，但现有的人工目视检测方法难以实现高精度的多种、小目标缺陷自动识别。为此，本研究提出了一种基于深度学习模型的管道环焊缝DR图像缺陷智能识别方法。首先针对工件缺陷复杂、射线噪声和散射等问题，采用对比度有限自适应直方图均衡化方法实现原始图像增强，并构建高质量的管道环焊缝DR图像数据集；其次，提出一种新颖的改进YOLOv5深度学习网络，采用C2F模块学习多种典型缺陷的特征分布，并结合多尺度特征融合检测手段来判断不同大小的目标缺陷；进一步采用均方误差损失函数提高对噪声的容忍性，并通过输出端实现对焊缝缺陷的自动检测。最终对管道环焊缝缺陷进行测试和验证，试验结果表明：通过改进YOLOv5深度网络模型，其检测结果的平均准确率为95.89%，漏检率为2.1%，能够有效准确地自动识别DR图像焊缝缺陷。

Abstract

Pipeline girth welds are a crucial component of oil and gas pipelines

significantly impacting the safety of natural gas and other transmissions. Currently

X-ray Digital Radiography (DR) technology has become an essential method for the quality inspection of girth welds. However

existing visual inspection methods struggle to achieve high-precision defect identification. Therefore

this paper proposes a deep learning-based algorithm for defect detection in pipeline girth welds using DR images. Firstly

to address issues such as complex workpiece defect structures and noise and scatter introduced by X-rays

guided filtering methods are used for data enhancement

and a pipeline girth weld DR dataset is created. Secondly

the backbone network of the improved YOLOv5 model

combined with the C2F (CSP Bottleneck with 2 convolutions) module

is employed to learn the defect feature distributions. The multi-scale feature fusion detection method is combined to judge the target defects of different sizes. The mean square error loss function is further used to improve the tolerance of noise

and the automatic detection of weld defects is realized through the output end. Finally

the common defects of pipeline girth welds were tested and verified

and the results showed that the average accuracy of detection results by the improved YOLOv5 deep network model proposed in this paper was 95.89%

and the missed detection rate was 2.1%

which could effectively and accurately automatically identify weld defects in DR Images.

关键词

Keywords

references

苟星禄，吴頔，董金枋，等 . 基于机器学习的铝合金脉冲激光焊接工艺及质量预测研究［J］. 电焊机， 2023 ， 53 （ 9 ）： 84 - 90 .

GOU X L ， WU D ， DONG J F ， et al . Quality Prediction of Pulse Laser Welding of Aluminum Alloy Based on Machine Learning ［J］. Electric Welding Machine ， 2023 ， 53 （ 9 ）： 84 - 90 .

李恩，何冰，江国焱，等 . 反应堆压力容器管座贯穿件与顶盖J型焊缝缺陷分析及改进［J］. 电焊机， 2024 ， 54 （ 4 ）： 77 - 83 .

LI E ， HE B ， JIANG G Y ， et al . Analysis and Improvement on the Penetration Testing Defects for J Type Joint of Head Cover and Pipe Seat in Reactor Pressure Vessel ［J］. Electric Welding Machine ， 2024 ， 54 （ 4 ）： 77 - 83 .

王飞，朱建江 . 机器视觉在复杂环境下焊接缺陷检测的应用研究［J］. 焊接技术， 2017 ， 46 （ 5 ）： 127 - 133 .

WANG F ， ZHU J J . Research on the Application of Machine Vision in Welding Defect Detection in Complex Environments ［J］. Welding Technology ， 2017 ， 46 （ 5 ）： 127 - 133 .

Li Y P ， Gao W X . Research on X-ray welding image defect detection based on convolution neural network ［J］. Journal of Physics：Conference Series ， 2019 ， 1237 （ 3 ）： 032005 .

Torkzadeh V ， Toosizadeh S . Automatic visual inspection system for quality control of the sandwich panel and detecting the dipping and buckling of the surfaces ［J］. Measurement and Control ， 2019 ， 52 （ 7-8 ）： 804 - 813 .

焦敬品，李思源，常予，等 . 集箱管接头内焊缝表面缺陷识别方法研究［J］. 仪器仪表学报， 2017 ， 38 （ 12 ）： 3044 - 3052 .

JIAO J P ， LI S Y ， CHANG M ， et al . Defect classification of weld surface in header pipe joint ［J］. Chinese Journal of Scientific Instrument ， 2017 ， 38 （ 12 ）： 3044 - 3052 .

Zapata J ， Vilar R ， Ruiz R . An adaptive-network-based fuzzy in-ference system for classification of welding defects ［J］. NDT & E International：Independent Nondestructive Testing and Evaluation ， 2010 ， 43 （ 3 ）： 191 - 199 .

Hou W ， Ye W ， Yi J ， et al . Deep features based on a DCNN model for classifying imbalanced weld flaw types ［J］. Measurement ， 2018 ， 131 ： 482 - 489 .

王春妮，李健，白真权，等 . 油气输送管道失效分析技术现状及研究进展［J］. 石油管材与仪器， 2024 ， 10 （ 1 ）： 1 - 8 .

WANG C N ， LI J ， BAI Z Q ， et al . Research Progress in Failure Analysis Technology for Oil and Gas Transmission Pipelines ［J］. Petroleum Tubular Goods & Instruments ， 2024 ， 10 （ 1 ）： 1 - 8 .

任文坚，王永红，李春凯，等 . 液体火箭发动机推力室熔焊缝数字化胶片图像缺陷识别方法研究［J］. 电焊机， 2023 ， 53 （ 1 ）： 9 - 14 .

REN W J ， WANG Y H ， LI C K ， et al . Research on Au

tomatic Detection Method of DR Digital imaging for Brazing Seam Defects of Thrust Chamber Body ［J］. Electric Welding Machine ， 2023 ， 53 （ 1 ）： 9 - 14 .

储钰昆，吴磊，杨洪刚，等 . 基于YOLOv8的铸件DR图像小目标缺陷检测［J］. 特种铸造及有色合金， 2025 （ 1 ）： 35 - 41 .

CHU Y K ， WU L ， YANG H G ， et al . Small Target Defect Detection in Casting DR Images Based on YOLOv8 ［J］. Special Casting & Nonferrous Alloys ， 2025 （ 1 ）： 35 - 41 .

曹帅，董立红，邓凡，等 . 基于YOLOv7-SE的煤矿井下场景小目标检测方法［J］. 工矿自动化， 2024 ， 50 （ 3 ）： 35 - 41 .

CAO S ， DONG L H ， DENG F ， et al . A small object detection method for coal mine underground scene based on YOLOv7-SE ［J］. Journal Of Mine Automation ， 2024 ， 50 （ 3 ）： 35 - 41 .

龙凌，陈浩，梁昊，等 . 基于轻量化YOLO网络的实时X射线焊缝缺陷检测［J］. 网络新媒体技术， 2023 ， 12 （ 2 ）： 30 - 38 .

LONG L ， CHEN H ， LIANG H ， et al . Real-time X-ray Weld Defect Detection Based on Lightweight YOLO Network ［J］. Network New Media Technology ， 2023 ， 12 （ 2 ）： 30 - 38 .

Alert me when the article has been cited

提交

Research on Quality Control Strategy of Automatic Assembly Welding of Guard Board

Typical applications of Artificial Intelligence Surrogate Models in Laser Metal Additive Manufacturing

Gas Shielded Metal Arc Welding with Misalignment-Gap Mode of Penetration Prediction and Feature Analysis

Research on the Application of Phased Array Ultrasonic Testing Technology in Weld Inspection of Bogie Frames

Research on Intelligent Evaluation Technology of Weld Quality based on Machine Vision and Deep Learning Neural Networks