基于ResGNNet多模态融合的油气管道缺陷等级磁记忆定量识别

doi:10.3969/j.issn.1004-132X.2025.09.027

摘要/Abstract

摘要：

针对油气管道磁记忆信号特征自动提取及缺陷等级定量识别难题，提出一种结合残差神经网络和图神经网络的多模态融合模型即ResGNNet模型。采用金属磁记忆检测仪采集L245N管线钢不同深度缺陷的磁记忆原始信号。为实现特征的自动提取，保留原始磁记忆信号的完整信息并考虑样本之间的相互关系，利用K近邻-动态时间规整将原始信号转换成节点图，并利用格拉姆角场将原始信号转换成二维图像。设计的图神经网络、残差神经网络可分别自动提取一维信号和二维图像的嵌入特征向量。融合多模态嵌入特征向量经多头自注意力机制加权筛选后，输入Softmax分级模块，完成缺陷等级识别。模型验证结果表明，管道缺陷等级定量识别的准确率达到93%。

关键词: 油气管道, 金属磁记忆技术, 缺陷等级, 图神经网络, 残差神经网络

Abstract:

Aiming at the problems of automatic extraction of magnetic memory signal features and quantitative identification of defect levels in oil and gas pipelines， a multimodal fusion model was proposed combining residual neural network and graph neural network（ResGNNet）. The original magnetic memory signals of defects of different depths on L245N pipeline steels were collected by metal magnetic memory detector. In order to realize automatic feature extraction， the complete information of the original magnetic memory signals was retained， and the relationship among samples was taken into account. The original signals were converted into a node graph by K nearest neighbor-dynamic time warping， and the original signals were converted into a 2D image by Gram angle field. The designed graph neural network and residual neural network may automatically extract the embedded feature vectors of 1D signals and 2D images respectively. The multimodal embedded feature vectors were fused， weighted and screened by multi-head self-attention mechanism， and then input into the Softmax classification module to complete the defect level identification. The model verification results show that the accuracy of quantitative identification of pipeline defect levels reaches 93%.

Key words: oil and gas pipeline, metal magnetic memory technology, defect level, graph neural network, residual neural network

中图分类号:

TG115.284

邢海燕, 武雪缘, 蔡智会, 赵力伟, 苏田, 韩晴. 基于ResGNNet多模态融合的油气管道缺陷等级磁记忆定量识别[J]. 中国机械工程, 2025, 36(09): 2150-2157.

Haiyan XING, Xueyuan WU, Zhihui CAI, Liwei ZHAO, Tian SU, Qing HAN. Quantitative Identification of Oil and Gas Pipeline Defect Levels Based on Magnetic Memory Using ResGNNet Multi-modal Fusions[J]. China Mechanical Engineering, 2025, 36(09): 2150-2157.

导出引用管理器 EndNote|Ris|BibTeX

链接本文: https://www.cmemo.org.cn/CN/10.3969/j.issn.1004-132X.2025.09.027

https://www.cmemo.org.cn/CN/Y2025/V36/I09/2150

图/表 19

图 1 预制不同深度缺陷及检测方向示意图

Fig.1 Schematic diagram of prefabricated defects of different depths and detection directions

图 2 缺陷检测示意图

Fig.2 Schematic diagram of defect detection

图3 同一缺陷8个检测方向的磁场切向分量

Fig.3 Tangential component of the magnetic field for the same defect in 8 detection directions

表1 同一缺陷8个检测方向的磁场切向分量峰峰值

Tab.1 Peak-to-peak values of the tangential component of the magnetic field for the same defect in 8 detection directions

检测方向	1	2	3	4
峰峰值H_pp/（A·mm^-1）	28.1	11.6	21.2	36.3
检测方向	5	6	7	8
峰峰值H_pp/（A·mm^-1）	26.5	9.8	22.9	34.5

图4 同一检测方向下不同深度缺陷的磁场峰峰值

Fig.4 Magnetic field peak-to-peak values for defects of different depths in the same detection direction

图5 ResGNNet模型架构

Fig.5 ResGNNet model architecture

图6 基于K-DTW的节点图构建

Fig.6 Constructed based on K-DTW node graphs

图7 图神经网络结构图

Fig.7 Structure of graph neural network structure

图8 GAF图像编码示意图

Fig.8 Schematic diagram of GAF image coding

图9 残差神经网络结构图

Fig.9 Structure of residual neural network

图10 多头自注意力机制原理图

Fig.10 Schematic diagram of the multiple self-attention mechanism

表 2 ResGNNet模型的主要参数

Tab.2 Main parameters of ResGNNet model

网络层	输入通道	输出通道	输出特征图尺寸
Conv2d	3	16	75×75×16
Maxpool			38×38×16
Resblock1	16	32	19×19×32
Resblock2	32	64	10×10×64
Resblock3	64	128	5×5×128
Avgpool			128
GConv1	320	320	320
GConv2	320	128	128
Concat			256
MultiHead			256
FC Layer			3
Softmax			3

图11 ResGNNet训练和测试准确率

Fig.11 ResGNNet training and testing accuracies

图12 ResGNNet训练损失曲线

Fig.12 ResGNNet training loss curve

表 3 不同检测方向缺陷等级识别准确率对比

Tab.3 Comparison of defect level identification accuracy in different detection directions

检测方向	1	2	3	4
准确率/%	92.85	86.66	94.66	100.00
检测方向	5	6	7	8
准确率/%	92.31	87.50	93.75	96.44

图13 训练前后样本的分布可视化

Fig.13 Visualization of sample distribution before and after training

图14 混淆矩阵

Fig.14 Mixing matrix

表4 模型分级识别性能评估 (%)

Tab.4 Model hierarchical recognition performance evaluation

	精确率	召回率	F1分数
Ⅰ级缺陷	95.74	93.75	94.74
Ⅱ级缺陷	91.67	91.67	91.67
Ⅲ级缺陷	84.62	91.67	88.00
宏平均	90.68	92.36	91.47

表 5 不同模型分级识别准确率 (%)

Tab.5 Different model classification recognition accuracy

准确率	ResNet	GNN	ResGNNet
训练集	90.12	93.26	99.99
测试集	72.85	78.96	93.33

参考文献 15

[1]	DUBOV A A. A Study of Metal Properties Using the Method of Magnetic Memory［J］. Metal Science and Heat Treatment， 1997， 39（9）：401-405.
[2]	苏三庆，高志刚，王威. 受拉钢丝应力状态的磁记忆信号峰峰值判别技术［J］. 西安建筑科技大学学报（自然科学版）， 2018， 50（2）：155-161.
	SU Sanqing， GAO Zhigang， WANG Wei. Discrimination of Stress Stage about Steel Wire under Tension Based on Peak-peak Difference Value of Magnetic Memory Signals［J］. Journal of Xi’an University of Architecture and Technology， 2018， 50（2）：155-161.
[3]	赵秉勋. 腐蚀-力-磁耦合模型及金属应力腐蚀的磁记忆方法研究［D］. 北京：北京交通大学， 2022.
	ZHAO Bingxun. Research on corrosion-stress-magnetic coupling model and stress corrosion of metal based on magnetic memory method［D］. Beijing：Beijing Jiaotong University， 2022.
[4]	湛立宁，卢俊文，王肖逸，等. 非接触式磁应力检测技术在埋地管道腐蚀评价中的应用［J］. 管道技术与设备， 2023（1）：46-50.
	ZHAN Lining， LU Junwen， WANG Xiaoyi， et al. Application of Non-contact Magnetic Stress Detection Technology in Corrosion Evaluation of Buried Pipelines［J］. Pipeline Technique and Equipment， 2023（1）：46-50.
[5]	焦杨. 非接触式磁记忆检测技术在埋地管道上的应用［J］. 全面腐蚀控制， 2017， 31（11）：23-25.
	JIAO Yang. Application of Non Contact Magnetic Memory Testing Technology in Buried Pipeline［J］. Total Corrosion Control， 2017， 31（11）：23-25.
[6]	石明江，陈瑞，冯林. 基于磁记忆的金属管道缺陷检测方法［J］. 电子测量与仪器学报， 2022， 36（1）：44-53.
	SHI Mingjiang， CHEN Rui， FENG Lin. Metal Pipeline Defect Detection Method Based on Magnetic Memory［J］. Journal of Electronic Measurement and Instrumentation， 2022， 36（1）：44-53.
[7]	万勇，王宇，杨勇，等. 管道缺陷类型多特征量阈值识别方法［J］. 油气储运， 2020， 39（3）：268-276.
	WAN Yong， WANG Yu， YANG Yong， et al. Multi-characteristic Threshold Identification Method for Pipeline Defect Types［J］. Oil & Gas Storage and Transportation， 2020， 39（3）：268-276.
[8]	王贵生，李炜，杨勇，等. 基于磁记忆信号特征的管道缺陷分类识别和分级识别方法［J］. 腐蚀与防护， 2022， 43（11）：68-73.
	WANG Guisheng， LI Wei， YANG Yong， et al. Identification Methods for Pipeline Defect Classification and Grading Based on Magnetic Memory Signal Characteristics［J］. Corrosion & Protection， 2022， 43（11）：68-73.
[9]	WAN Yuan， CHEN Xiaoli， SHI Ying. Adaptive Cost Dynamic Time Warping Distance in Time Series Analysis for Classification［J］. Journal of Computational and Applied Mathematics， 2017， 319：514-520.
[10]	KIPF T N， WELLING M. Semi-supervised Classification with Graph Convolutional Networks［J］. ArXiv e-Prints， 2016：arXiv：.
[11]	WANG Zhiguang， YAN Weizhong， OATES T. Time Series Classification from Scratch with Deep Neural Networks：a Strong Baseline［C］∥2017 International Joint Conference on Neural Networks. Anchorage， 2017：1578-1585.
[12]	HE Kaiming， ZHANG Xiangyu， REN Shaoqing， et al. Deep Residual Learning for Image Recognition［C］∥2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas， 2016：770-778.
[13]	SANTURKAR S， TSIPRAS D， ILYAS A， et al. How Does Batch Normalization Help Optimization？［J］. arXiv， 2018：1805.11604.
[14]	VASWANI A， SHAZEER N， PARMAR N， et al. Attention is All You Need［J］. arXiv， 2017：1706.03762.
[15]	廖柯熹，冷吉辉，何腾蛟，等. 一种埋地管道腐蚀缺陷检测方法［J］. 腐蚀与防护， 2021， 42（2）：52-55.
	LIAO Kexi， LENG Jihui， HE Tengjiao， et al. A Method for Detecting Corrosion Defects in Buried Pipelines［J］. Corrosion & Protection， 2021， 42（2）：52-55.