长输天然气管道的海量音波数据挖掘方法研究  

Research on the data mining method for massive acoustic waves in the long-distance transmission gas pipeline

作　　者：梁伟[1] 张来斌[1] 郭存杰 

机构地区：[1]中国石油大学(北京)机械与储运工程学院 [2]中石油北京天然气管道有限公司

出　　处：《石油机械》2012年第1期4-8,95,共5页China Petroleum Machinery

基　　金：国家自然科学基金项目"油气储运‘设备链’瞬态过程的耦变规律及其故障的自组织诊断方法"(51005247);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);北京市科技新星计划项目"油气生命线关键设备耦变故障的自组织诊断方法"(2010B068);北京市教委共建项目"天然气主管线泄漏检测系统研究"

摘　　要：长输天然气管道的音波检测技术受高采样频率影响,多通道音波时间序列为海量数据,如何快速有效进行数据挖掘是面临的一大难题。为此,对系统产生的海量音波数据进行了数据挖掘方法研究。介绍了基于自底向上法的分段线性表示的原理,利用该方法对音波序列进行数据分割,突出信号拐点信息,同源聚类,减小了由自定义长度分割带来的统计误差,实现了不同信号源数据的有效分割。同时,运用分裂基算法取代基2FFT对音波序列进行快速频率抽取,与基2FFT相比,分裂基算法在保持频谱图信息完整性的同时,减少了约33%的实数乘法计算量,提高了系统分析速度。试验结果证明了这2种方法的有效性和实用性。The technology of sound wave detection of the long-distance transmission gas pipeline is affected by high sampling frequency.The time series of multi-channel sound waves are characterized by massive data and it is a difficult task to mine the data quickly and effectively.Therefore,the method to mine the massive sound wave data produced by the system was studied.The principle of the bottom-up method-based piecewise linear representation was introduced.The method was used to conduct a data partitioning of the acoustic wave sequence and foreground the inflection point information of signals.Through the homologous clustering method,the statistic errors caused by user-defined length segmentation were reduced and the effective partitioning of data with different signal sources was achieved.Meanwhile,the split radix algorithm was adopted to replace the 2FFT radix to carry out the fast frequency-decimal operation of sound wave series.Compared with the 2FFT radix,the split radix algorithm reduces 33% of the calculation of real number multiplication and improves the speed of the system analysis while it keeps the completeness of the spectrogram information.

关 键 词：天然气管道 音波 数据挖掘 分段线性表示 分裂基 

分 类 号：TE973.6[石油与天然气工程—石油机械设备]