基于RBF网络的漏磁检测缺陷定量分析方法  

Quantitative analysis method for MFL testing for oil and gas pipelines based on RBF neutral network

作　　者：崔伟[1] 黄松岭[1] 赵伟[1] 

机构地区：[1]清华大学电机工程与应用电子技术系电力系统及大型发电设备安全控制和仿真国家重点实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2006年第7期1216-1218,1222,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(50305017)

摘　　要：为了正确评估油气管道的使用寿命和安全状况,需根据漏磁检测信号特征对缺陷进行准确的定量分析。提出一种基于径向基函数(RBF)神经网络、用于定量分析油气管道缺陷的迭代方法,给出了具体的算法步骤,并采用自适应学习机制来训练网络,既加快了该算法的收敛速度,又避免了陷入局部最小值问题。仿真结果表明:该方法不仅训练速度明显快于普通反向传播(BP)网络,而且最大量化误差仅为0.26%。该方法有助于提高漏磁检测的准确度,可为油气管道的安全评估提供可靠的依据。In order to properly evaluate the service lifetime and security status of oil and gas pipelines, the defects should be quantitatively analyzed based on the features of the magnetic flux leakage （MFL） signals. An iterative quantitative analysis method based on radial-basis function （RBF） neutral network was put forward in this paper, and the detailed procedures of the iterative algorithm were introduced. And, by using self-adaptive mechanism to train the network, the convergence time was shortened and the local minimum in the error surface was effectively avoided. The results of simulation show this method has much faster training speed than standard backward propagation （BP） network, and the maximal error of quantification is only 0. 26%. This method can be used to increase the quantification accuracy of MFL detection, and provide reliable basis for the security evaluation of oil and gas pipelines.

关 键 词：油气管道 漏磁检测 径向基函数神经网络 缺陷 

分 类 号：TE973.6[石油与天然气工程—石油机械设备]