摘要：Pipeline defect detection holds significant importance in the domain of buried pipeline maintenance. Nevertheless, in the actual detection of buried pipelines, underground pipelines are intricate. Complex working conditions are prevalent, such as pipeline diameter changes, crossing, and connectors (valves, fittings), etc. These complex conditions can lead to abnormal fluctuation signals during the defect (metal loss type defects) detection process. As a result, the detection accuracy is reduced. Consequently, there is an urgent need for effective method to identify pipeline defects and complex conditions. To address these issues, an information fusion technique combining Transient Electromagnetic Method (TEM) signals and Weak Magnetic Detection Method (WMDM) signals is proposed. This technique detects buried pipelines and constructs one-dimensional datasets. On this basis, a variety of time–frequency analysis techniques are employed to transform one-dimensional data into two-dimensional images. Then, a Double-Branch Fusion network based on Triplet Attention (TA-DBF) is proposed. It aims to address the data imbalance problem between different complex working conditions in practical engineering. This model employs a multi-modal and multi-input strategy. Its double-branch network separately handles feature learning and classification tasks on input data. Finally, under varying imbalance factors, the performance of the model loss and information fusion technology are analyzed. Comparison and ablation experiments are carried out to evaluate the performance of the model. When the imbalance factor is 0.1, the PrecisionMA, RecallMA and F1-scoreMA of TA-DBF reach 0.93. The results indicate that this method can effectively identify the complex working conditions of buried pipelines. ? 2025

ISSN号：0263-2241

卷、期、页：卷255

发表日期：2025-11-01

期刊分区（SCI为中科院分区）：二区

收录情况：SCI（科学引文索引印刷版）,EI（工程索引）,SCIE（科学引文索引网络版）

发表期刊名称：Measurement: Journal of the International Measurement Confederation

参与作者：Zhao, Yixin

通讯作者：宋福霖,Miao, Xingyuan

第一作者：赵弘

论文类型：期刊论文

论文概要：宋福霖,赵弘,Miao, Xingyuan,Zhao, Yixin，Identification method of buried pipeline complex conditions based on TA-DBF model and information fusion technique，Measurement: Journal of the International Measurement Confederation，2025，卷255

论文题目：Identification method of buried pipeline complex conditions based on TA-DBF model and information fusion technique