基于蚁群神经网络的油气套管裂缝缺陷检测  

Detection in Cracked Defects of Oil-Gas Casing Based on Ant Colony Neural Network

作　　者：黄博[1] 师奕兵[1] 张伟[1] 卢涛[1,2] 

机构地区：[1]电子科技大学自动化工程学院,四川成都610054 [2]中海油田服务股份有限公司技术中心,北京101149

出　　处：《测控技术》2011年第3期98-102,共5页Measurement & Control Technology

基　　金：国家863计划资助项目(2006AA06Z222);教育部新世纪优秀人才支持计划资助项目(NECET-05-0804)

摘　　要：提出一种基于蚁群神经网络的漏磁信号定量分析方法。首先,通过有限元仿真,分析和确定了漏磁信号中能够反映裂缝缺陷参数的各个特征量;其次,通过对蚁群算法原理的研究,建立了以漏磁信号为对象的神经网络模型;最后,模拟实际检测中不同缺陷信号的特征量作为网络输入,并对网络性能进行测试。实验结果验证了蚁群神经网络的可行性,并且能明显改善神经网络的收敛速率和准确性,有效提高了漏磁信号定量识别的效率和质量。A new method on quantitative analysis of magnetic flux leakage signal based on ant colony neural network is proposed.Firstly,the parameters of the magnetic flux leakage signal which can reflect the various characteristics of cracked defects are determined by finite element method（FEM） simulation.Secondly,based on the study in the principle of ant colony algorithm,the neural network model is established for magnetic flux leakage signals processing.Finally,through the simulated working environment,the performance of the new neural network is tested with the different signal feature as input.Experimental results obviously proved the feasibility of ant colony neural network,verified the increases on the convergence rate and the accuracy of the neural network,also indicated that the efficiency and quality of quantitative identification of the magnetic flux leakage signals had been greatly improved.

关 键 词：蚁群算法 神经网络 漏磁检测 套管 

分 类 号：TG115.28[金属学及工艺—物理冶金] TP183[金属学及工艺—金属学]