基于神经网络的长距离油气管道安全预警系统  

An artificial neural networks based long-distance safety monitoring system for buried oil pipelines

作　　者：王立[1] 王耀辉[2] 肖昕璐[2] 闫湖[1] 郭澎[2] 史广顺[1] 王庆人[1] 

机构地区：[1]南开大学机器智能研究所,天津300071 [2]南开大学现代光学研究所,天津300071

出　　处：《高技术通讯》2008年第7期719-724,共6页Chinese High Technology Letters

基　　金：863计划(2006AA06Z22)资助项目

摘　　要：提出了一种基于分布式光纤传感和人工神经网络判别的长距离输油管道安全预警系统。该系统利用光纤传感器收集管道周围土壤的振动信号,通过神经网络判断是否存在针对管道的破坏性行为和判别破坏性行为的类别,实现对油气管道的长距离安全预警。系统在预处理阶段对信号大幅度降维,降低数据处理的时间复杂度,以满足实时性的要求。在识别阶段则采用人工神经网络模型,包括反向传播(BP)网络和支持向量机(SVM)。试验结果表明,这两种神经网络模型对打夯、镐刨、电钻三类破坏行为的识别率分别达到96.5和97.1%,均优于以往文献中的报道。The paper proposes a general long-distance oil pipeline monitoring system based on the distributed optical fiber sen- sots and artificial neural networks （ANN）. This system uses fiber sensors for collection of vibration signals of the soil around pipelines, and then, applies artificial neural networks, including the back-propagation （BP） network and the support vector machine （SVM）, for detecting if there exist damaging actions to pipelines according to the collected signals and determines if there is a need to give an alarm. During pre-processing, the system concentrates on reducing the signal dimension so as to reduce the time complexity and then meet the real-time requirement. The experiments show that these two neural network models are able to identify damaging actions with the accuracy rates of 96.5 % and 97.1%, respectively, much better than that of other reports ever since.

关 键 词：傅立叶分析 小波分析 人工神经网络 支持向量机 管道安全 第三方破坏 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]