基于奇异值特征的管道压力波识别方法  

A method to identify the pipeline pressure wave based on singular value characteristics

作　　者：王明达[1] 张来斌[1] 梁伟[1] 费盼峰[1] 

机构地区：[1]中国石油大学(北京)机电工程学院

出　　处：《石油机械》2009年第12期68-71,117,共4页China Petroleum Machinery

基　　金：国家863计划项目"基于双扭环机制的输油管线泄漏诊断的新装置与方法研究"(2008AA06Z209);教育部新世纪优秀人才支持计划资助项目"成品油输送管线泄漏的精确识别与定位方法"(NCET-05-0110)

摘　　要：在输油管道泄漏诊断过程中,提取压力波动信号的哪种特征参数作为识别依据是提高管道泄漏诊断准确率的关键。为此提出一种基于奇异值分解的压力波动信号特征提取方法,并通过构建BP神经网络对其分类,判断管道所处工况状态。在不影响识别效果的前提下,对奇异值向量进行适当降维压缩,能达到消除压力信号噪声和减少识别运算量的目的。试验表明,采用引入"动量项"的BP神经网络压力波识别分类器识别准确率高。基于奇异值特征与BP神经网络的管道压力波识别方法运算速度快,能满足管道泄漏监测在线运行的要求,有较高的应用价值。In diagnosis of oil pipeline leakage,the characteristic parameter of pressure fluctuation signal to be extracted as identificational basis is the key to improvement of pipeline leakage diagnosis accuracy. Accordingly,a method to extract pressure fluctuation signal which is based on singular value decomposition is put forward. And it is classified by establishing the BP neural network to make a judgement of the operating state of the pipeline. In the condition that the identificational effect is not disturbed,the dimension reduction and compression of the singular value vector is conducted,which can achieve the purpose of removing pressure signal noise and reduce identificational calculations. The test proves that the identificational classifier adopting 'momentum' as a key parameter of the BP neural network pressure wave has a high accuracy of identification. There is a fast calculation speed for the pipeline pressure wave identification method based on the singular value and the BP neural network. It can satisfy the online operation requirement of pipeline leakage monitoring and thus is of great value for application.

关 键 词：奇异值特征 管道压力波 Hankel矩阵构造 BP神经网络 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TQ021.1[自动化与计算机技术—计算机科学与技术]