动态聚类算法在管道泄漏检测中的应用  

Application of ISODATA method in pipeline leakage detection

作　　者：胡瑾秋[1] 张来斌[1] 王朝晖[1] 梁伟[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》2007年第2期31-35,72,共5页China Petroleum Machinery

基　　金：国家自然科学基金资助项目(50105015;50375103);教育部霍英东青年教师教育基金资助项目(91051);北京市科技新星计划资助项目(2003B33)

摘　　要：为了提高负压波法检测管道泄漏的准确性和灵敏度,正确区分不同工况下的压力波形,通过对负压波的产生及模式识别中无监督学习方法的研究,采用ISODATA动态聚类算法对负压波波形进行分类,判断管道所处的工况状态。为了加快分类速度,利用K-L变换对上述聚类算法进行优化,消除负压波特征参数间的相关性,降低了特征空间的维数,同时减少了计算时间。实验结果验证了该优化算法的有效性和准确性。Pipeline transportation has been widely used in every walk of life because of its special merits.However,increasingly serious leakage problem has not only affected the normal operation,but also made huge loss to human life and wealth.Based on the analysis of occurrence of negative pressure waves and unsupervised learning of pattern recognition,this paper makes use of ISODATA method to classify the negative pressure waves,determine the working condition of the pipeline,and carry out the optimization by K-L transformation,which eliminates the correlativity of each characteristic parameters,reduces the dimensions of characteristic space and accelerates the calculation.The experimental results have testified the effectiveness and accuracy of this method.

关 键 词：负压波法 ISODATA动态聚类算法 泄漏检测 K-L变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]