摘要
小直径管道阵列结构焊接位置易产生泄漏缺陷,且管道阵列排布方式复杂。针对能够实现小直径管道阵列结构逐管快速扫查装置控制的自动化检测软件系统尚未开发,制约其缺陷检测精度这一问题,设计了一种非接触式快速扫查检测软件系统。该软件系统整合了小直径管道检测过程中的能量激励模块、信号接收模块、数据存储模块以及数据处理模块。软件人机交互UI界面中具有串口通信、硬件控制和结果展示等单元,操作者只需进行简单的按键操作即可实现密集小直径弯管阵列快速检测,并实时显示管道的检测时域信号与各管道健康状态。软件系统验证实验结果表明,该系统对各弯管检测准确率可达87.0%,所设计软件系统能够推动检测自动化的发展,并对激光超声导波技术及深度学习技术的工业化应用起到促进作用。
The welding positions of small-diameter pipeline array structures are prone to leakage defects,and the layout of pipeline arrays is complex.The lack of an automated testing software system that can control the rapid scanning of small-diameter pipeline array structures on a per-pipe basis has hindered the accuracy of defect detection.A non-contact rapid scanning testing software system is designed.The software system integrates the energy excitation module,signal reception module,data storage module,and data processing module in the smalldiameter pipeline detection process.The software’s human-machine interaction UI interface includes serial communication,hardware control and result display units.The operator can perform simple button operations to achieve rapid detection of dense small-diameter bent pipe arrays and display the real-time detection time-domain signal and the health status of each pipeline.The experimental results of the software system verification experiment show that the accuracy of the system for detecting each bent pipe can reach 87.0%.The system can promote the development of automated testing and facilitate the industrial application of laser ultrasonic guided wave technology and deep learning technology.
作者
胡广泽
黄刘伟
洪晓斌
Hu Guangze;Huang Liuwei;Hong Xiaobin(Dalian University of Technology,School of Control Science and Engineering,Dalian,Liaoning 116081,China;South China University of Technology,School of Mechanical and Automotive Engineering,Guangzhou 510630,China)
出处
《机电工程技术》
2023年第10期99-102,共4页
Mechanical & Electrical Engineering Technology
基金
广东省重点领域研发计划项目(2019B010154002)。
关键词
小直径管道
无损检测
导波
自动化控制
深度学习
small diameter pipeline
non destructive testing
guided wave
automation control
deep learning