浏览全部资源
扫码关注微信
收稿：2024-09-02，
修回：2025-09-11，
纸质出版：2026-01-20
移动端阅览
管道环焊缝缺陷是诱发管道失效的关键因素，传统数字射线照相无损检测技术（Digital Radiography non-destructive testing，DR）图像缺陷识别依赖人工评估，存在主观性强、识别效率低等问题。为解决这一难题，提出一种基于改进YOLOv8算法的DR图像缺陷智能识别方法。首先，针对DR图像噪声干扰与灰度分布不均的问题，系统对比中值滤波、均值滤波、高斯滤波及双边滤波的降噪效果，确定卷积核大小为3×3的中值滤波为最优降噪方案（PSNR=45.24 dB，SSIM=0.985）；然后，通过线性变换与伽马变换的组合增强策略，验证当线性变换参数a=0.7的线性变换、伽马变换参数 γ=2时，缺陷特征的辨识度显著提升。在此基础上，将全局注意力机制（Global Attention Mechanism，GAM）和双向特征金字塔网络（Bidirectional Feature Pyramid Network，BiFPN）融入YOLOv8算法框架，以提高环焊缝缺陷智能识别准确率及速度。实验结果表明，改进后的YOLOv8算法对裂纹、未熔合、未焊透等5类典型缺陷的识别准确率达83.1%，较原始YOLOv8算法提升10.7%；同时，模型的mAP@0.5与召回率分别达到74.7%和76.8%，较原算法提升 8.3%和14.4%。该方法有效突破了传统DR检测的智能化瓶颈，为管道环焊缝缺陷的高效、精准识别提供了可靠技术支撑，具有重要的工程应用价值。
Defects in pipeline girth welds are a key factor leading to pipeline failure. Digital Radiography (DR) non-destructive testing technology has become the mainstream method for detecting defects in girth welds due to its high recognition accuracy and convenient equipment deployment. However
traditional DR image defect identification relies on manual evaluation
which suffers from strong subjectivity and low efficiency. To address this challenge
an intelligent defect recognition method based on an improved YOLOv8 algorithm is proposed. Firstly
considering noise interference and uneven grayscale distribution in DR images
the denoising effects of median filtering
mean filtering
Gaussian filtering
and bilateral filtering were systematically compared. It was determined that median filtering with a convolution kernel size of 3×3 is the optimal denoising scheme (PSNR=45.24 dB
SSIM=0.985). Then
through a combined enhancement strategy using linear transformation and gamma correction
it was verified that when the linear transformation parameter a=0.7 and the gamma correction parameter γ=2
the distinguishability of defect features significantly improves. On this basis
the Global Attention Mechanism (GAM) and Bidirectional Feature Pyramid Network (BiFPN) were integrated into the YOLOv8 algorithm framework to enhance both the accuracy and speed of intelligent recognition of girth weld defects. Experimental results demonstrate that the improved YOLOv8 algorithm achieves an accuracy rate of 83.1% for identifying five typical types of defects
including cracks
lack of fusion
and incomplete penetration—an improvement of 10.7% over the original YOLOv8 algorithm. Additionally
the model's mAP@0.5 and recall rate reach 74.7% and 76.8%
respectively
representing improvements of 8.3% and 14.4% compared to the original algorithm. This method effectively overcomes the intelligence bottleneck of traditional DR inspection
providing reliable technical support for efficient and precise identification of pipeline girth weld defects
with significant engineering application value.
Alonso J ， Pavon S ， Vidal J ， et al . A New Insight on Phased Array Ultrasound Inspection in MIG/MAG Welding ［J］. Materials ， 2022 ， 15 （ 8 ）： 16 - 29 .
王乐乐 ， 李莉 ， 张斌 ， 等 . 中国油气储运技术现状及发展趋势 ［J］. 油气储运 ， 2021 ， 40 （ 09 ）： 961 - 972 .
WANG L L ， LI L ， ZHANG B ， et al . Current status and development trend of oil and gas storage and transportation technologies ［J］. Oil & Gas Storage and Transportation ， 2021 ， 40 （ 09 ）： 961 - 972 .
Chianese G ， Franciosa P ， Nolte J ， et al . Characterization of Photodiodes for Detection of Variations in Part-to-Part Gap and Weld Penetration Depth During Remote Laser Welding of Copper-to-Steel Battery Tab Connectors ［J］. Journal of Manufacturing Science and Engineering ；Transactions of the ASME， 2022 ， 144 （ 7 ）： 65 - 106 .
王维斌 ， 雷铮强 ， 杨辉 . 长输管道数字射线DR检测技术应用与展望 ［J］. 油气储运 ， 2020 ， 39 （ 12 ）： 1337 - 1343 .
WANG W B ， LEI Z Q ， YANG H . Application and prospect of digital radiographic detection technology in long-distance pipeline ［J］. Oil & Gas Storage and Transportation ， 2020 ， 39 （ 12 ）： 1337 - 1343 .
He C Y . Application of Intelligent Technology on Tunnel Steel Structure Factory ［J］. Mathematical Problems in Engineering ， 2022 （ 4 ）： 1 - 9 .
HU Y H ， LU L L ， ZHAN S F . Simulation and reliability evaluation of automated ultrasonic testing technology in semi-automatic welding of oil and gas pipelines ［J］. The International Journal of Advanced Manufacturing Technology ， 2023 ， 124 ： 4131 - 4141 .
Jayasudha J C ， Lalithakumari S . Weld defect segmentation and feature extraction from the acquired phased array scan images ［J］. Multimedia Tools and Applications 2022 ， 81 ： 31061 - 31074 .
Li S ， Gao J ， Zhou E ， et al . Deep learning-based fusion hole state recognition and width extraction for thin plate TIG welding ［J］. Welding in the World ， 2022 ， 66 ： 1329 - 1347 .
Liang D ， Wu Y ， Hu K ， et al . Weld seam track identification for industrial robot based on illumination correction and center point extraction ［J］. Journal of Advanced Mechanical Design，Systems，and Manufacturing ， 2022 ， 16 （ 3 ）： 49 - 85 .
Lim S J ， Kim Y L ， Cho S ， et al . Ultrasonic Inspection for Welds with Irregular Curvature Geometry Using Flexible Phased Array Probes and Semi-Auto Scanners：A Feasibility Study ［J］. Applied Sciences ， 2022 ， 12 （ 2 ）： 748 .
Lindgren E ， Zach C . Industrial X-ray Image Analysis with Deep Neural Networks Robust to Unexpected Input Data ［J］. Metals ， 2022 ， 12 （ 11 ）： 79 - 89 .
Nowroth C ， Gu T ， Grajczak J ， et al . Deep Learning-Based Weld Contour and Defect Detection from Micrographs of Laser Beam Welded Semi-Finished Products ［J］. Applied Sciences ， 2022 ， 12 （ 9 ）： 1125 - 1148 .
Sun H ， Ramuhalli P ， Jacob R E . Machine learning for ultrasonic nondestructive examination of welding defects： A systematic review ［J］. Ultrasonics ， 2023 ， 127 ： 106854 .
Wang W X ， Yamane S ， Wang Q ， et al . Visual sensing and quality control in plasma MIG welding ［J］. Journal of Manufacturing Processes ， 2023 ， 86 ： 163 - 176 .
Wang X P ， Wang X ， Zhang B X ， et al . Binary classification of welding defect based on deep learning ［J］. Science and Technology of Welding and Joining ， 2022 ， 27 （ 4 ）： 1 - 11 .
Wang X P ， Yu X H . Understanding the effect of transfer learning on the automatic welding defect detection ［J］. NDT & E International ， 2023 ， 134 ： 102784 .
Wang Z ， Gao W ， Song J . Applying SDR with CNN to Identify Weld Defect： A New Processing Method ［J］. Journal of Pipeline Systems Engineering and Practice ， 2023 ， 14 （ 2 ）： 256 - 269 .
Yang L ， Fan J ， Huo B ， et al . A nondestructive automatic defect detection method with pixelwise segmentation ［J］. Knowledge-Based Systems ， 2022 ， 242 ： 108338 .
Yang L ， Song S ， Fan J ， et al . An Automatic Deep Segmentation Network for Pixel-Level Welding Defect Detection ［J］. IEEE Transactions on Instrumentation and Measurement ， 2021 ， 71 ： 5003510 .
Yang Y ， He Y ， Guo H ， et al . Semantic segmentation supervised deep-learning algorithm for welding-defect detection of new energy batteries ［J］. Neural Computing and Applications ， 2022 ， 34 ： 19471 - 19484 .
Wen L ， Wang Y ， Li X . A New Cycle-consistent Adversarial Networks With Attention Mechanism for Surface Defect Classification With Small Samples ［J］. IEEE Transactions on Industrial Informatics ， 2022 ， 18 （ 12 ）： 8988 - 8998 .
相关作者
相关机构
公网安备11010802024621
