作者：朱红秀1, 刘欢1, 李宏远1, 黄松岭2, 苏志毅2
作者单位：1. 中国矿业大学(北京)机电与信息工程学院, 北京 100083;
2. 清华大学电机系电力系统国家重点实验室, 北京 100084
关键词：漏磁;缺陷分类;BP神经网络;收敛速度
摘要：
缺陷准确量化是管道漏磁检测领域中长期存在的一个难点, 而对缺陷进行科学的分类是实现准确量化的重要前提。针对不同形态的缺陷, 分析其特征参数对漏磁信号的影响因素, 建立用于缺陷分类的BP神经网络模型, 设计改进的Levenberg-Marquardt算法用于网络训练, 并利用Ansoft Maxwell 3D建立仿真缺陷数据作为样本进行测试。结果表明:改进后的神经网络系统可实现对缺陷的有效分类, 改善传统算法分类精度低、误差大的缺点, 收敛速度大幅度提高。该方法已成功应用于大型油气田软件工程领域, 为实现缺陷准确量化提供基础和依据。
Research on the classification and identification of oil and gas pipeline corrosion defects
ZHU Hongxiu1, LIU Huan1, LI Hongyuan1, HUANG Songling2, SU Zhiyi2
1. School of Mechanical Electronic & Information Engineering, China University of Mining & Technology (Beijing), Beijing 100083, China;
2. State Key Lab of Power Systems, Department of Electrical Engineering, Tsinghua University, Beijing 100084, China
Abstract: Accurate quantification of defects is a difficult problem in the field of magnetic flux leakage inspection for a long time, and scientific classification of defects is a basic and important prerequisite for it. For different forms of defects, analyses the shape parameter factors influencing magnetic flux leakage signal, a classification method based on modified BP neural network is proposed, Improved Levenberg-Marquardt method is also employed as the training algorithm and simulation defect data from Ansoft Maxwell 3D is used as a sample for testing. The test result has proved that the improved neural network system could effectively achieve the classification of defects with high classification accuracy and small errors; in particular, the convergence speed has greatly improved. This method has been successfully applied to large oil and gas fields of software engineering; it has provided the foundation and basis for achieving an accurate quantification which is of great significance.
Keywords: magnetic flux leakage;classification of defects;back-propagation neural networks;convergence rate
2015, 41(6): 91-95 收稿日期: 2014-10-11;收到修改稿日期: 2014-12-19
基金项目: 国家"863"重大项目(2011AA090301)国家重大科学仪器设备开发专项(2013YQ140505)
作者简介: 朱红秀(1970-),女,河北遵化市人,副教授,博士,主要从事机电一体化、测试技术的教学与应用研究工作。
参考文献
[1] 黄辉, 何仁洋, 熊昌胜, 等. 漏磁检测技术在管道检测中的应用及影响因素分析[J]. 管道技术与设备, 2010(3):17-19.
[2] 吴欣怡, 赵伟, 黄松岭. 基于漏磁检测的缺陷量化方法[J].电测与仪表, 2008, 45(5):20-22.
[3] 黄松岭, 李路明, 鲍晓宇, 等. 管道漏磁检测中的信号处理[J].无损检测, 2000(2):55-57.
[4] Joshi A, Udpa L, Udpa S, et al. Adaptive wavelets for characterizing magnetic flux leakage signals from pipeline inspection[J]. Magnetics IEEE Transactions, 2006, 42(10):3168-3170.
[5] 崔伟, 黄松岭, 赵伟. 基于RBF网络的漏磁检测缺陷定量分析方法[J]. 清华大学学报:自然科学版, 2006(7):1216-1218.
[6] 田凯, 孙永泰, 高慧, 等. 贝叶斯算法BP神经网络缺陷量化研究[J]. 中国测试, 2014(3):93-97.
[7] 吴欣怡, 黄松岭, 赵伟. 使用改进型BP神经网络量化裂纹漏磁信号[J]. 无损检测, 2009(8):603-605.
[8] Fan J. The modified Levenberg-Marquardt method for nonlinear equations with cubic convergence[J]. Mathematics of Computation, 2012, 81(277):447-466.
