基于模糊RBF神经网络的管道泄漏检测方法  

Detection method of pipeline leak based on fuzzy RBF neural network

作　　者：李炜[1] 张美玲[1] 毛海杰[1] 

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050

出　　处：《兰州理工大学学报》2009年第6期79-83,共5页Journal of Lanzhou University of Technology

基　　金：教育部"春晖计划"(Z2005-1-62001)

摘　　要：针对模糊BP神经网络在管道泄露检测与估计中存在网络构建训练速度慢、易陷入局部最优等问题,提出将模糊RBF神经网络方法应用于管道的泄漏检测与估计.首先依据管道泄漏时流量、压力的变化机理,将采集到的实际运行中管道内的流量差与压力差信号模糊化后作为RBF神经网络的输入,以泄漏尺寸大小的置信度作为网络的输出,并结合专家先验知识所得的模糊规则,构建管道泄漏检测的模糊RBF神经网络.进而以实际管道运行数据对其进行离线仿真测试,仿真结果表明模糊RBF神经网络克服了模糊BP神经网络的不足,提高了泄漏估计的精度,使网络构建更加高效、优化.Aimed at the problems of slowness of network formation training and easiness of falling into local optimal case when the fuzzy BP neural network was used for detection and estimation of pipeline leak, a fuzzy RBF neural network method was presented for pipeline leak detection and estimation. First of all, based on the mechanism of flow and the pressure changes of pipeline leakage, the differential signal of the flow and pressure gathered from the actually operating pipeline was converted into fuzzy one and taken as the RBF neural network input. Then the confidence level of leak age dimension was taken as the network output, and finally a priori knowledge of experts was incorporated into the formation of the fuzzy rules, resulting in a fuzzy pipeline leak detection RBF neural network. Then with the actual operation data was used for their off-line simulation test and the simulation result showed that the fuzzy RBF neural network overcame the shortcomings of the fuzzy BP neural network to improve the accuracy of the leak estimation, making the network more efficient and optimized.

关 键 词：管道 泄漏检测 模糊RBF神经网络 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]