事件驱动型SVM预测供水管网管道结构状况  

Event-driven SVM for predicting structural condition of water supply pipelines

作　　者：周念清[1] 柳娅楠 郑茂辉[2] 李浩楠 ZHOU Nianqing;LIU Yanan;ZHENG Maohui;LI Haonan(School of Civil Engineering,Tongji University,Shanghai 200092,China;Shanghai Institute of Disaster Prevention and Relief,Tongji University,Shanghai 200092,China)

机构地区：[1]同济大学土木工程学院,上海200092 [2]同济大学上海防灾救灾研究所,上海200092

出　　处：《给水排水》2021年第3期144-149,共6页Water & Wastewater Engineering

基　　金：国家重点研发计划资助项目(2016YFC0802400,2017YFC0803300)。

摘　　要：结合网络核密度估计(NetKDE)和支持向量机(SVM)算法构建了一个新型的事件驱动的供水管道状况预测模型。以上海市杨浦区供水管网为例,应用NetKDE方法对管网历史事件分布进行分析计算,得到的事件核密度用以表征管网中各管段结构安全状况,并利用自然分界法划分状况等级。以状况等级为响应特征,用径向基(RBF)核函数建立供水管网SVM预测模型,通过网格搜索和交叉验证方法确定最优的模型参数,然后利用优化后的模型进行分类预测。结果表明,该模型总体分类准确率达0.91,能够有效识别结构安全状况较差的管段,对于供水管网安全隐患排查和风险防控具有重要的现实意义。Combining Network kernel density estimation(NetKDE)and Support vector machine(SVM)algorithms,a new event-driven model is constructed for predicting structural conditions of water supply pipelines in this paper.Taking the water supply network in Yangpu district of Shanghai as an example,the NetKDE method is utilized to analyze the distribution of historical network events.The event kernel density value is calculated to characterize the structural safety status of each pipe section in the network,and the status grade is divided by natural break method.Then a SVM prediction model is established by the Radial basis function(RBF),and the optimal model parameters are determined through grid search and cross-validation methods.Finally,the structural conditions of water supply pipelines are predicted with the optimized model.The experimental results show that the overall accuracy rate is 0.91,and it can effectively identify the pipe sections with poor structural safety,which has important practical significance for the detection and prevention of potential safety hazards of water supply network.

关 键 词：供水管网 结构状况 网络核密度估计 支持向量机 事件驱动 

分 类 号：TU990[建筑科学—市政工程]