【作者】 刘浩；

【导师】 刘文中；

【作者基本信息】 华中科技大学 ， 新能源科学与工程， 2020， 硕士

【摘要】 管道漏失检测的主要目的是对供水管网,工业系统中工作管道等进行故障排查,以防止管道泄漏而造成资源浪费,或者影响正常的生产过程。管道漏失检测的精确性直接影响到故障排查的准确度,因此,管道漏失检测技术具有重要意义。本文通过对漏失前后管道振动加速度数据的采集及分析,利用支持向量机,和参数优化算法对采集的数据完成漏失检测模型的训练与测试,找出管道漏失检测效果较好的模型。主要内容包括管道振动加速度数据采集和分析,以及利用支持向量机实现管道漏失检测。管道振动加速度数据采集部分,设计了一种基于物联网技术的振动加速度测量装置,采集管道振动加速度信号,并通过Lo Ra(Long Range Radio远距离无线电技术)将多个测量装置采集的数据上传至本地网关,经网关汇总到云平台中存储。考虑到物联网设备计算处理能力有限,Lo Ra进行数据传输前需对数据进行压缩处理。压缩算法通过对信号的分析,选择了自适应差分脉冲编码调制压缩算法(ADPCM)。完成数据采集后,通过对管道漏失前后的振动加速度信号进行分析,找出管道泄漏前后,加速度信号主要的参数变化,便于后续选择合适的特征量用于支持向量机训练管道漏失检测模型。在支持向量机(SVM)进行管道漏失检测部分,介绍了支持向量机原理,以及不同的参数优化算法,比较了原始支持向量机模型与不同优化算法优化后的支持向量机模型进行管道漏失检测的结果,找出了检测效果较好的管道漏失检测模型。对训练得到的管道漏失检测模型进行测试验证,结果表明,无论是对于振动加速度原始数据,还是经过压缩算法编解码后的数据,利用粒子群算法对原始支持向量机模型进行参数优化后得到的管道漏失检测模型,检测效果均能达到最好,对原始数据的检测准确率为98%,对压缩解压缩后数据的检测准确率为90.07%,较好的实现了管道漏失检测。更多还原

【Abstract】 The main purpose of pipeline leak detection is to troubleshoot water supply networks,working pipelines in industrial systems,etc.to prevent pipeline leakage from causing waste of resources or affecting normal production processes.The accuracy of pipeline leak detection directly affects the accuracy of troubleshooting.Therefore,pipeline leak detection technology is of great significance.In this thesis,through the collection and analysis of pipeline vibration acceleration data before and after the leak,the support vector machine and parameter optimization algorithms are used to complete the training and testing of the leak detection model to find the model with the better pipeline leak detection effect.The main contents include the acquisition and analysis of pipeline vibration acceleration data,and the realization of pipeline leakage detection using support vector machines.In the pipeline vibration acceleration data acquisition part,a vibration acceleration measurement device based on the Internet of Things technology is designed to collect pipeline vibration acceleration signals and upload the data collected by multiple measurement devices to the local through Lo Ra(Long Range Radio long-range radio technology).The gateway is aggregated into the cloud platform for storage through the gateway.Considering that the computing and processing capabilities of Io T devices are limited,Lo Ra needs to compress the data before data transmission.The compression algorithm selects the adaptive differential pulse code modulation compression algorithm(ADPCM)by analyzing the signal.After the data collection is completed,the vibration acceleration signals before and after the pipeline leakage are analyzed to find out the main parameter changes of the acceleration signals before and after the pipeline leakage,which is convenient for the subsequent selection of appropriate feature quantities for support vector machine training pipeline leakage detection models.In the support vector machine(SVM)pipeline leak detection part,the principle of support vector machine and different parameter optimization algorithms are introduced.The original support vector machine model and the support vector machine model optimized by different optimization algorithms are compared for pipeline leak detection.As a result,the better pipeline leak detection model was found.The pipeline leak detection model obtained by training is tested and verified,and the results show that whether it is the original data of vibration acceleration or the data encoded and compressed by the compression algorithm,the original SVM model is optimized by using the particle swarm optimization algorithm.The leakage detection model can achieve the best detection effect.The detection accuracy of the original data is 98%,and the detection accuracy of the compressed and decompressed data is 90.07%.The pipeline leakage detection is well realized.更多还原