APP下载

基于粒子群优化的SVM供水管道泄漏诊断方法

2018-04-04王学渊陈志刚钟新荣卢宁

现代电子技术 2018年7期

王学渊 陈志刚 钟新荣 卢宁

摘 要: 供水管道泄漏会造成水资源浪费和经济损失,传统支持向量机泄漏诊断模型中存在参数选择不确定的问题,导致其分类结果不稳定。提出将粒子群进化算法应用于泄漏诊断支持向量机模型中的参数优化选择,利用粒子群群体智能优化搜索从全局迅速地迭代出合理的支持向量机的惩罚参数以及核参数,使建立的PSO?SVM管道泄漏诊断模型达到最优。实验测试表明,通过结合粒子群算法全局搜索收敛速度快的优点,有效地解决了支持向量机模型中两个重要参数优化选择的问题,提升了支持向量机分类的准确率和效率。

关键词: 供水管道; 泄漏诊断; 支持向量机; 粒子群算法; 参数优化; PSO?SVM

中图分类号: TN911.1?34; TP274.2 文献标识码: A 文章编号: 1004?373X(2018)07?0156?04

PSO?SVM based leakage diagnosis method of water supply pipeline

WANG Xueyuan1, 2, CHEN Zhigang1, 2, ZHONG Xinrong3, LU Ning1, 2

(1. School of Mechanical?electronic and Vehicle Engineering, Beijing University of Civil Engineering and Architecture, Beijing 100044, China;

2. Beijing Engineering Research Center of Monitoring for Construction Safety, Beijing 100044, China;

3. Changqing Downhole Technology Company, Xian 710000, China)

Abstract: The leakage of water supply pipeline will cause the water resource waste and economic losses. The traditional leakage diagnosis model based on support vector machine (SVM) has the problem of uncertain parameter selection, which may cause the unstable classification result. The particle swarm optimization (PSO) algorithm is proposed for parameter optimization selection in leakage diagnosis model based on SVM. The particle swarm intelligent optimization search is used to quickly iterate the reasonable penalty parameter and kernel parameter of SVM in the overall situation, so as to make the pipeline leakage diagnosis model based on PSO?SVM optimal. The experimental results show that, in combination with the fast convergence speed of PSO global search, the method can solve the two important parameters optimization selection problem in SVM model, and improve the accuracy and efficiency of SVM classification.

Keywords: water supply pipeline; leakage diagnosis; SVM; PSO; parameter optimization; PSO?SVM

0 引 言

随着我国城市化水平不断提高,供水管道长度呈现加速增长趋势,但供水管网道的漏损现象却十分普遍。对于城市供水管道的漏失诊断,如何从复杂干扰中识别泄漏是整个技术的关键难点[1]。针对这一关键点,文献[2]提出基于神经网络的管道泄漏检测识别方法,很大程度上提升了泄漏检测的准确率。但是,神经网络存在训练样本需求多、收敛速度慢等问题,文献[3]提出基于结构风险最小化的SVM理论应用于管道泄漏诊断领域,它在小样本及非线性的情形下有独特优势,可以实现对管道泄漏进行比较精确的诊断。但有研究表明,在成熟的条件下核函数对SVM识别性能的影响不大,其中对SVM性能起到关键作用的因素[4]是核参数和惩罚因子[C。……

登录APP查看全文