焊管2025，Vol.48Issue(5)：27-35,9.DOI:10.19291/j.cnki.1001-3938.2025.05.004
管道环焊缝超声相控阵检测图谱智能识别研究
Research on Pipeline Circumferential Weld Phased Array Ultrasonic Testing Images Intelligent Recognition
王波 1谢建桥 2张辉宇 2刘钊 2张如韵 3杨新基 3董绍华3
作者信息
- 1. 中石化新疆新春石油开发有限责任公司,新疆 塔城 834700
- 2. 中石化胜利海上石油工程技术检验有限公司,山东 东营 257001
- 3. 中国石油大学(北京),北京 102249
- 折叠
摘要
Abstract
To enhance the intelligent detection of circumferential weld defects in oil and gas pipelines and develop safer,more efficient pipeline networks,a deep learning-based intelligent recognition method for ultrasonic phased array images was proposed.A dataset containing various typical defects in pipeline circumferential weld ultrasonic phased array images was first constructed,with data quality optimized through image enhancement and noise reduction techniques.Subsequently,the deep learning model architecture was improved by integrating target detection principles with defect imaging characteristics to address classification and localization requirements.Experimental results demonstrate that the proposed method achieved an identification accuracy of 0.963 and a localization precision of 0.86 for defects in circumferential weld ultrasonic phased array images.This approach effectively enables automated intelligent detection of circumferential weld defects and provides robust technical support for intelligent defect evaluation,thereby advancing safety assessments in oil and gas pipeline systems.关键词
超声相控阵检测/缺陷识别/卷积注意力模块/残差学习/特征金字塔Key words
phased array ultrasonic testing/defect identification/convolutional attention module/residual learning/feature Pyramid分类
金属材料引用本文复制引用
王波,谢建桥,张辉宇,刘钊,张如韵,杨新基,董绍华..管道环焊缝超声相控阵检测图谱智能识别研究[J].焊管,2025,48(5):27-35,9.
