• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2017, Vol. 53 ›› Issue (12): 70-77.doi: 10.3901/JME.2017.12.070

• 承压设备的无损检测与评价 • 上一篇    下一篇

基于模糊加权马尔科夫链的焊缝隐性损伤 磁记忆特征参数定量预测

邢海燕, 孙晓军, 王犇, 葛桦, 党永斌, 喻正帅   

  1. 东北石油大学机械科学与工程学院 大庆 163318
  • 出版日期:2017-06-20 发布日期:2017-06-20
  • 作者简介:

    邢海燕,女,1971年出生,博士,教授。主要研究方向为石

    E-mail:xxhhyyhit@163.com

  • 基金资助:
    * 国家自然科学基金(11272084,11072056,11472076)、中石油科技创新基金(2015D-5006-0602)、东北石油大学研究生创新科研(YJSCX2016-024NEPU)和黑龙江省博士后科研启动基金(LBH-Q13035)资助项目; 20160727 收到初稿,20170214 收到修改稿;

Quantitative MMM Characteristic Parameter Prediction for Weld Hidden Damage Status Based on the Fuzzy Weighted Markov Chain

XING Haiyan, SUN Xiaojun, WANG Ben, GE Hua, DANG Yongbin, YU Zhengshuai   

  1. School of Mechanical Science and Engineering, Northeast Petroleum University, Daqing 163318
  • Online:2017-06-20 Published:2017-06-20

摘要:

针对磁记忆技术在焊缝隐性损伤状态定量评价与预测中的难题,建立基于无偏灰色模糊加权马尔科夫链的焊缝隐性损伤状态磁记忆特征参数定量预测模型。以Q235焊接试件为试验材料,拾取焊缝纵向和横向磁记忆信号分布,与同步X射线检测结果进行对比得出:在宏观损伤时满足传统的磁记忆检测判据,即切向分量Hp(x)具有极大值,法向分量Hp(y)过零值;而在隐性损伤时并没有出现传统判据特征,说明Hp(x)与Hp(y)并不能准确地判断隐性损伤状态和程度。为此引入正交矢量合成梯度比Kr,可以较为敏感地反映隐性损伤状态,但敏感性同时带来局部振荡波动性,造成对隐性损伤后续发展状态的磁记忆定量预测困难。因此,在无偏灰色预测理论和马尔科夫链的基础上,结合模糊隶属函数建立模糊加权马尔科夫链磁记忆定量预测模型,验证结果表明:无偏灰色理论结合模糊加权马尔科夫链后的定量预测模型,其最大相对误差从 38.492 5%降到5.046 4%,为实际工程中焊缝隐性损伤未来发展状态的定量预测与维修策略选择提供了新的思路。

关键词: 磁记忆, 马尔科夫链, 模糊隶属度, 无偏灰色, 隐性损伤

Abstract:

In order to solve the difficulty in quantitatively predicting the evolution of hidden damage for welded joints by applying metal magnetic memory (MMM) technology, a quantitative MMM prediction model is presented based on unbiased gray theory and fuzzy weighted Markov theory. Steel Q235 welded plate specimens are tested by TSC-4M-8 MMM Instrument. By contrasting the result of X ray synchronous detection, the distribution of MMM tangential and normal signals are obtained. When the macroscopic defects come into being, the signals meet the conventional MMM criterion which the tangential signalHp(x) occurs peak value and the normal signalHp(y) has zero value. However, at the stage of hidden damage, the signals don’t conform to the traditional MMM criteria, which means thatHp(x) andHp(y) can’t accurately judge the appearance and degree of hidden damage. For this reason, the ratio of orthogonal gradient vector sum (Kr) is introduced, which can more sensitively reflect the existence of the hidden damage. Based on unbiased gray fuzzy weighted Markov chain, the quantitative MMM prediction model is established. The results show the maximum relative error of the model is 5.046 4% decreased from 38.492 5%, which provides a novel way for the quantitative prediction of the future development of the weld hidden damage and the maintenance strategy selection in practical engineering.

Key words: fuzzy membership, Markov chain, MMM, unbiased gray, hidden damage