›› 2015, Vol. 28 ›› Issue (11): 13-.

• 论文 • 上一篇    下一篇

基于小波能量系数和神经网络的管道缺陷识别

姜银方,郭华杰,陈志伟,杜斌,刘秋阁   

  1. (1.江苏大学 机械工程学院,江苏 镇江 212013;
    2.江苏省特种设备安全监督检验研究院 镇江分院,江苏 镇江 212009)
  • 出版日期:2015-11-15 发布日期:2015-12-15
  • 通讯作者: 郭华杰(1988—),男,硕士研究生。研究方向:超声导波技术在管道无损的检测。E-mail:ghj0712@126.com
  • 作者简介:姜银方(1962—),男,教授。研究方向:激光技术,板料塑性成形理论,表面工程技术等。
  • 基金资助:

    江苏省特检院2012年度科技基金资助项目(KJ(Y)2012049)

Pipeline Defect Identification Based on Wavelet Energy Coefficients and Neural Network

JIANG Yinfang,GUO Huajie,CHEN Zhiwei,DU Bin,LIU Qiuge   

  1. (1.School of Mechanical Engineering,Jiangsu University,Zhenjiang 212013,China;
    2.Zhenjiang Sub-branch,Jiangsu Special Equipment Safety Supervision Inspection Institute,Zhenjiang 212009,China)
  • Online:2015-11-15 Published:2015-12-15

摘要:

利用基于小波能量系数的BP神经网络方法对管道焊缝和管道凹槽进行分类识别。建立了导波检测系统,采集了管道凹槽缺陷和焊缝的多组检测信号样本,从信号样本中提取出小波能量系数,并将小波能量系数应用于BP神经网络的训练与识别。结果表明,该方法对管道缺陷的识别准确率较高,且识别效果稳定,在随机抽取信号样本进行的5次试验中,对焊缝和凹槽的最低识别准确率分别为92%和98%,最高识别准确率均为100%。

关键词: 超声导波, 小波能量系数, 神经网络, 管道, 缺陷识别

Abstract:

The method wavelet energy coefficients combined with BP neural network is used to distinguish pipeline grooves from welds.A guided wave detection system was established,a set of test samples of pipeline grooves from welds were collected,and wavelet energy coefficients were extracted from test samples and applied to the training and recognition of BP neural network.Results show that the identification accuracy of pipeline defects of this method is high and stable with a minimum identification accuracy of 92% and 98% for weld and groove respectively,and a highest recognition accuracy of 100% in the 5 experiments on randomly chosen samples.

Key words: ultrasonic guided wave;wavelet energy coefficient;neural network;pipeline;defect identification

中图分类号: 

  • TP18