一种基于瞬态响应数据的水管泄漏检测算法  

A water pipe leakage detection algorithm based on transient response data

作　　者：胡怀雯[1] Hu Huaiwen(Hunan Mechanical Electrlcal Polytechnic,Changsha 410151,China)

机构地区：[1]湖南机电职业技术学院,湖南长沙410151

出　　处：《无线互联科技》2023年第17期135-137,共3页Wireless Internet Technology

基　　金：湖南省教育科学“十三五”规划2019年度教育财建研究专项课题,项目名称:基于物联网的学校后勤水电节能精准管理研究,项目编号:CJ193851。

摘　　要：文章详细分析了水管泄漏瞬态响应模型,一些算法根据水头的峰值和相位可以获取管道泄漏的信息,但是这些算法对数据噪声敏感;另一些算法要求在检测时计算出管道系统的频率响应图,导致检测工作烦琐,实时性不高。因此文章根据现有的瞬态响应模型,利用深度学习网络提取瞬态响应数据特征,检测水管模型的泄漏位置与泄漏量。实验证明,相较于传统的瞬态响应方法,深度神经网络能在一定程度上克服数据噪声带来的影响,同时也不需要每次检测时都测量出频率响应图,只要在训练时加入各种频率的数据即可获得相应频率下的管道特征,增加了算法的实时性。This text provides a detailed analysis of the transient response model for detecting leaks in pipelines.Some algorithms can obtain information about pipeline leaks based on the peak value and phase of the water head.However,these algorithms are sensitive to data noise.Other algorithms require calculation of the frequency response graph of the pipeline system during inspections,which leads to tedious work and low real-time performance.Therefore,this paper analyzes in detail the transient response model of water pipes during leakage,and uses deep learning networks to extract data characteristics from the transient response to detect the leakage position and flow rate of the water pipe model.The experimental results show that compared with the traditional method based on transient response,deep neural networks can overcome the influence of data noise to some extent.Additionally,there is no need to measure the frequency response diagram each time during detection,simply adding data of various frequency terms during training will gain the corresponding frequency-dependent features of the pipeline,thus increasing the real-time performance of the algorithm.

关 键 词：瞬态响应 数据噪声 深度学习 泄漏检测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]