基于深度可分离残差网络的管道环焊缝AUT检测缺陷识别方法

doi:10.3901/JME.260494

摘要/Abstract

摘要： 全自动超声波检测技术被广泛地应用于管道环焊缝的检测，但受检测人员专业水平的影响，很容易造成误判和漏判。为了降低人为因素对缺陷识别精度和缺陷效率的影响，提出一种基于深度可分离卷积改进残差网络的管道环焊缝缺陷识别方法。将全自动超声波检测技术获得的管道环焊缝扫查信号进行预处理，构建图像数据集作为神经网络的输入。对损失函数加权，同时使用深度可分离卷积替换残差网络中残差模块中的第2个标准卷积，构造一种基于深度可分离卷积的缺陷识别模型DS-ResNet。使用改进的麻雀搜索算法对DS-ResNet中学习率、批大小和最大轮次等超参数进行自适应寻优。对提出的缺陷自动识别方法与优化前的网络进行对比测试，结果表明，提出的自动识别方法具有更高的准确率、精准率和调和平均值。为管道环焊缝缺陷智能识别提供了可行的研究思路。

关键词: 管道环焊缝, 缺陷识别, 全自动超声波检测, 麻雀搜索算法, 深度可分离卷积, 残差网络

Abstract: Automatic ultrasonic testing(AUT) technology is widely used for the inspection of pipeline girth welds. However, due to the influence of the proficiency level of the inspection personnel, it is prone to errors in judgment and missed detections. In order to reduce the influence of human factors on the accuracy and efficiency of defect identification, a defect identification method for pipeline girth weld based on depth separable convolution(DSC) improved ResNet is proposed. The scan signals of the pipeline girth weld obtained by AUT are preprocessed, and an image dataset is constructed as the input for the neural network. By weighting the loss function and replacing the second standard convolution in the residual modules of ResNet with DSC, a defect identification model called DS-ResNet is constructed based on DSC. The improved sparrow search algorithm(ISSA) is used to adaptively optimize hyperparameters such as learning rate, batch size, and maximum epochs in DS-ResNet. A comparative test is conducted between the proposed defect identification method and the network before optimization. The results indicate that the proposed identification method exhibits higher accuracy, precision, and F1-score. A feasible research approach is proposed for the intelligent identification of pipeline girth weld defects.

Key words: pipeline girth weld, defects identification, automatic ultrasonic testing, sparrow search algorithm, depth separable convolution, residual network

中图分类号:

TG441
TP181

梁士通, 焦敬品, 张响. 基于深度可分离残差网络的管道环焊缝AUT检测缺陷识别方法[J]. 机械工程学报, 2026, 62(10): 93-102.

LIANG Shitong, JIAO Jingpin, ZHANG Xiang. Defect Identification Method for AUT Inspection of Pipeline Girth Welds Based on Depth Separable Residual Network[J]. Journal of Mechanical Engineering, 2026, 62(10): 93-102.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.260494

http://www.cjmenet.com.cn/CN/Y2026/V62/I10/93

参考文献

[1] 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.
[2] 卞文彬，邓艾东，刘东川，等. 基于改进深度残差收缩网络的风电机组滚动轴承故障诊断方法[J]. 机械工程学报，2023，59(12)：202-214. BIAN Wenbin，DENG Aidong，LIU Dongchuan，et al. Fault diagnosis method of wind turbine rolling bearing based on improved deep residual shrinkage network[J]. Journal of Mechanical Engineering，2023，59(12)：202-214.
[3] DING L，WAN H P，LU Q B，et al. Using deep learning to identify the depth of metal surface defects with narrowband SAW signals[J]. Optics and Laser Technology，2022，157：108758.
[4] MA Q，DU G F，YU Z Y，et al. Classification of damage types in liquid-filled buried pipes based on deep learning[J]. Measurement Science and Technology，2022，34(2)：025010.
[5] XU H，YAN Z H，JI B W，et al. Defect detection in welding radiographic images based on semantic segmentation methods[J]. Measurement，2021，188：110569.
[6] LIANG S T，MA J. Compound fault diagnosis of gearbox based on RLMD and SSA-PNN[J]. Mathematical Problems in Engineering，2021，2021：3716033.
[7] ZHANG R，ZHAO N，FU L H，et al. Recognizing defects in stainless steel welds based on multi-domain feature expression and self-optimization[J]. Journal of Intelligent Manufacturing，2021，34(3)：1293-1309.
[8] JIA N，CHENG Y，LIU Y，et al. Intelligent fault diagnosis of rotating machines based on wavelet time-frequency diagram and optimized stacked denoising auto-encoder[J]. IEEE Sensors Journal，2022，22(17)：17139-17150.
[9] XIONG J Y，LIANG W，LIANG X B，et al. Intelligent quantification of natural gas pipeline defects using improved sparrow search algorithm and deep extreme learning machine[J]. Chemical Engineering Research & Design，2022，183：567-579.
[10] WANG J N，ZHU H Q，ZHANG Y J，et al. A novel prediction model for wind power based on improved long short-term memory neural network[J]. Energy，2023，265：126283.
[11] XUE J K，SHEN B，PAN A Q. A hierarchical sparrow search algorithm to solve numerical optimization and estimate parameters of carbon fiber drawing process[J]. Artificial Intelligence Review，2023，56(1)：1113-1148.
[12] BARANDELA R，RANGEL E，SANCHEZ J S，et al. Restricted decontamination for the imbalanced training sample problem[J]. Lecture Notes in Artificial Intelligence，2003，2905：424-431.
[13] BUDA M A，MAKI A，MAZUROWSKI M A. A systematic study of the class imbalance problem in convolutional neural networks[J]. Neural Networks，2018，106：249-259.
[14] PARK J，LEE S E，KIM H J，et al. System invariant method for ultrasonic flaw classification in weldments using residual neural network[J]. Applied Sciences Basel，2022，12(3)：1477.
[15] MUNIR N，KIM H J，PARK J，et al. Convolutional neural network for ultrasonic weldment flaw classification in noisy conditions[J]. Ultrasonics，2019，94：74-81.
[16] JIANG H Q，YANG D Y，ZHI Z L，et al. A normal weld recognition method for time-of-flight diffraction detection based on generative adversarial network[J]. Journal of Intelligent Manufacturing，2022，35(1)：217-233.
[17] LANG X M. Recognition method of pipeline weld defects based on auxiliary classifier generative adversarial networks[J]. IEEE Instrumentation & Measurement Magazine，2022，25(2)：69-77.
[18] MA Y X，LIU M Q，ZHANG Y，et al. Imbalanced underwater acoustic target recognition with trigonometric loss and attention mechanism convolutional network[J]. Remote Sensing，2022，14(16)：4103.
[19] ZHANG Y L，DING F，KWONG S，et al. Feature pyramid network for diffusion-based image inpainting detection[J]. Information Sciences，2021，572：29-42.
[20] SUN P J，LU Y D，ZHAI J C. Mapping land cover using a developed U-Net model with weighted cross entropy[J]. Geocarto International，2021，37(25)：9355-9368.
[21] ZHANG X. Analysis of typical defect atlas of AUT detection technology[J]. Henan Chemical Industry，2021，38(9)：57-59.
[22] FANG J K. Pipeline weld defect evaluation based on automated ultrasonic testing scan map[J]. Petroleum Engineering Construction，2010，36(4)：60-62，69.
[23] YANG H Y，CIFTCI U，YIN L J. Facial expression recognition by de-expression residue learning[C]// 31st IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，June 18-23，2018，Salt Lake City，America. NewYork：IEEE，2018：2168-2177.
[24] JIANG H Q，YANG D Y，ZHI Z L，et al. A normal weld recognition method for time-of-flight diffraction detection based on generative adversarial network[J]. Journal of Intelligent Manufacturing，2022，35(1)：217-233.
[25] HUA Z Y，ZHOU Y，HUANG H. Cosine-transform-based chaotic system for image encryption[J]. Information Sciences，2019，480：403-419.