【作者】 陈萍；

【导师】 林伟国；

【作者基本信息】 北京化工大学 ， 检测技术与自动化装置， 2009， 硕士

【摘要】 目前,管道运输在国民经济的发展中发挥着越来越重要的作用,然而,由于管道自然老化腐蚀,施工造成的断裂,人为的偷油破环,管道运行的可靠性和安全性不能得到很好地保障,因此,在经济发展中,确保管道安全可靠的运行有着迫切的需求。先进的管道泄漏检测技术是管道正常运行和安全防护的关键手段。在实际的管道泄漏检测中,要做到准确的从含有噪声的泄漏信号中提取真实的泄漏信号,增强泄漏信号,其关键就在于泄漏信号特征的提取。但是,由于埋地管道的管径和材质,以及不同的埋设条件,管道泄漏信号具有不同的时、频率特征,管道泄漏信号的特征不但容易受到不同工况条件下噪声的干扰,而且还在管道中散射和反射,另外,传感器的非线性误差,变送器模数转换的量化误差都会影响管道泄漏信号的特征。正是由于以上原因,在复杂工况条件下,管道泄漏信号的特征提取变得异常困难,这也是在管道泄漏检测领域中众多学者研究的热点之一。本文以基于压电式动态压力传感器的管道泄漏信号暂态过程为研究对象,分析不同工况下首、末站泄漏信号的差异,介绍基于神经网络的管道泄漏信号特征提取的方法。提出把动态压力信号做正负区间划分,把相邻区间信号的累加值差分、均值差分和峰值差分作为泄漏信号识别的三个特征量。同时通过对泄漏信号特征量做相对量化,给出泄漏信号特征量的横向、纵向合理性的评判依据,并以此评判依据对实际泄漏信号中提取得到的特征量进行检验,验证特征量提取方法的合理性。最后,给出以首站、末站特征量共同作为输入的神经网络管道泄漏诊断模型及其训练、检验结果。管道泄漏监测系统的稳定性一方面取决于管道泄漏信号检测算法的准确性,另一方面还取决于管道泄漏监测系统程序设计和管道泄漏监测系统所依托的操作系统的稳定性。在提出基于神经网络的管道泄漏信号特征提取方法的基础上,本文还设计一套基于Linux操作系统的管道泄漏监测系统。该管道泄漏监测系统很好地利用Linux操作系统运行性能稳定的特点,以Qt作为Linux平台下的开发语言,通过TCP/IP网络来接收管道数据采集终端采集的管道动态压力信号,利用Sqlite数据库来对管道动态压力信号进行存储。该管道泄漏监测系统包括实时数据接收、历史数据查询、数据库存储等模块。更多还原

【Abstract】 Nowadays, pipeline is one of the most important tools in industrial transportations and has been playing a more important role on national economy growth.Due to the erosion of pipeline, human-intended damage, the safety of pipeline transportation is in danger. So it is most urgent and important to keep pipeline in a well working state.The techniques for pipeline leak detection are the key to protect the industrial transportations from damages. Many ideas and ways have been developed to detect pipeline leakage, while the key to leakage signal detection is leakage signal feature extraction. As the difference of pipeline size and local natural condition, pipeline leakage signal has different spectrum features, and this leakage signal feature is also affected by dynamic working condition of pipeline transportation, besides that, non-linear errors of sensors and quantization error of AD converter also introduce the difficulty to extract leakage signal features. The difficulty of leakage signal feature extraction has draw great attention of many researchers. In this paper, piezoelectric dynamic pressure based leak signal transient is chosen as research object, differences between leak signals of upstream and downstream under different operational conditions are analyzed, leakage signal enhancement using wavelet decomposition is introduced. A method of positive and negative interval divisions of dynamic pressure signal is proposed, differences of weighted signal sums, signal mean values, signal peaks of every two consecutive intervals are selected as three features of leak signal, Feasibility criteria for lengthwise and breadth wise evaluation of leak signal features are present, evaluation is done with features extracted from field data and feasibility is verified. At last, neural network based leak diagnose model with both features from upstream and downstream and its training, testing results are delivered.In addition, with the novel pipeline leakage signal extraction method introduced above, a pipeline leakage detection system based on Linux operating system is developed. The front-end of this system is programmed by Qt, a well-developed open-source graphic user interface tool. Pipeline signal is received using TCP/IP network from remote terminal unit. Powered with great SQLITE database, all the pipeline signal data can be stored easily and is convenient to use for future analysis. This pipeline leakage detection system has the following functional blocks: live signal receiving, wavelet analysis, historical signal search and database storage for pipeline signal.更多还原