摘要
由于漏磁信号与缺陷轮廓的非线性关系,由管道漏磁信号描述管道缺陷的几何特征一直是管道漏磁检测的难点.本文采用小波基函数神经网络的方法,建立了由管道缺陷的漏磁信号到缺陷截面轮廓图的网络映射.算法中应用迭代自组织数据分析(ISODATA)动态聚类的算法使得基函数中心的选取更加合理,经过多层分辨率的训练,网络输出表明,该网络可以较准确反映出缺陷的几何特征,为管道缺陷的特征提取提供一种可行的方法.
Because of the nonlinear relationship between the magnetic flux leakage (MFL) signals and profiles of defects, it is difficult to describe the characters of defects in buried pipelines by pipeline MFL inspection signals. In this paper,a net mapping from pipeline MFL inspection signals to profiles of defects is established by using the wavelet basis function neural network method, in which centers of basis functions are selected using iterative self-organizing data analysis techniques (ISODATA) dynamic clustering algorithm. After this multi-resolution wavelet basis function neural network is trained, the output indicates that this net can accurately reflect the characters of defects,therefore it can be a feasible method to extract the characters of pipeline defects.
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2005年第5期395-399,共5页
Journal of Tianjin University：Science and Technology
基金
国家自然科学基金资助项目(69974025)
关键词
埋地管道
漏磁检测
小波基函数神经网络
ISODATA算法
缺陷
buried pipeline
magnetic flux leakage detection
wavelet basis function neural network
ISODATA algorithm
defect
