长距离铁精矿输送管道泄漏检测研究

发布时间：2018-02-15 03:03

本文关键词： 矿浆管道泄漏检测压力信号去噪敏感奇异值分解小波近似熵极限学习机　出处：《昆明理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着管道运输的广泛应用,管道的安全运行也得到了人们的关注。管道发生泄漏时会造成非常严重的后果,包括经济和生产的损失、自然资源的浪费和环境因素的破坏等,尤其是后两者是无法补偿的,迫切地需要能够对管道泄漏进行准确及时检测的方法。管道泄漏检测作为管道安全运行重要构成部分成为管道运输中的重要研究内容。管道泄漏检测属于实际应用问题,管道若发生泄漏事故,此段管道要报废进行更换,因此管道泄漏检测需借助管道泄漏实验系统进行研究工作,浆体管道泄漏实验系统为管道泄漏检测提供实验数据基础。为了给下一步管道泄漏检测提供良好基础,需要对压力信号进行预处理,抑制管道压力信号中噪声干扰。在管道的实际工业运行中,要根据实际生产需求对管道矿浆的输送量进行工况调整,此时管道的压力波变换与管道发生泄漏时产生的负压波相似,如何排除工况调整干扰准确的对管道进行泄漏检测避免错报、漏报具有重要的意义。论文的主要研究工作如下:(1)针对矿浆管道特殊运输方式和复杂输送机理的问题,通过理论研究与实际经验设计并搭建了矿浆管道泄漏实验系统。理论计算出合理的管材类型和管道壁厚从而克服了管材磨蚀问题,设计了动力系统、混浆及清洗系统和测量采集系统。通过实验证明了管道泄漏系统的可行性,并为下一步的研究工作提供有效的实验数据。(2)针对管道压力泄漏信号去噪的问题,采用基于敏感因子奇异值分解的管道泄漏压力信号去噪的方法。该方法首先对原始信号构造Hankel矩阵再进行SVD分解,将分解后得到的分量信号利用敏感因子找出敏感分量,最后通过定位因子选择敏感分量所对应的奇异值进行信号重构,并用该方法对矿浆管道泄漏实验系统中采集到的压力信号进行降噪处理,作为信号的预处理为管道泄漏检测提供良好的基础。(3)针对从管道泄漏检测中的非线性和非平稳压力信号中提取泄漏特征难的问题,采用一种小波分解近似熵和极限学习机相结合的管道泄漏检测方法。首先对管道压力信号进行小波分解,选取含有主要特征的前3层分量,将前3层分量的近似熵和峭度值作为特征向量,最后通过极限学习机对特征向量进行识别分类。基于小波变换近似熵和极限学习机相结合的方法能有效准确的进行管道泄漏能识别。通过理论计算与实际经验相结合设计了浆体管道泄漏实验验系统,并为管道泄漏检测提供实验数据;将敏感因子引入传统的奇异值分解,通过信号重构有效的抑制管道噪声干扰,为管道泄漏检测提供良好基础;采用近似熵与小波变换结合的方法,提取工况调整状态、泄漏状态和正常运行状态时特征向量,通过极限学习机有效准确进行识别分类,为管道准确的泄漏检测提供新方法,具有一定理论与实际意义。
[Abstract]:With the wide application of pipeline transportation, people also pay attention to the safe operation of pipeline. The leakage of pipeline will cause very serious consequences, including the loss of economy and production, the waste of natural resources and the destruction of environmental factors, etc. In particular, the latter two are irreparable. It is urgent to be able to detect pipeline leakage accurately and timely. As an important part of pipeline safe operation, pipeline leakage detection is an important research content in pipeline transportation. Pipeline leakage detection is a practical application problem. In the event of pipeline leakage accident, the pipeline has to be scrapped and replaced, so the pipeline leakage detection needs to be studied with the pipeline leakage experimental system. The slurry pipeline leak experiment system provides the experimental data basis for pipeline leakage detection. In order to provide a good basis for pipeline leakage detection in the next step, it is necessary to preprocess the pressure signal. In the actual industrial operation of the pipeline, it is necessary to adjust the transportation rate of the pipeline slurry according to the actual production demand. At this time, the pressure wave transformation of pipeline is similar to the negative pressure wave generated by pipeline leakage. How to eliminate the interference of adjustment of working conditions and accurately detect the pipeline leakage to avoid misreporting, The main research work of this paper is as follows: 1) aiming at the problems of special transportation mode and complex conveyer mechanism of slurry pipeline, Based on the theoretical research and practical experience, the experimental system of slurry pipeline leakage is designed and built. The reasonable pipe type and pipe wall thickness are calculated theoretically, thus the problem of pipe abrasion is overcome, and the power system is designed. The feasibility of pipeline leakage system is proved by experiments, and effective experimental data is provided for further research. The method of pipeline leakage pressure signal denoising based on sensitivity factor singular value decomposition (SVD) is adopted. Firstly, the original signal is constructed by Hankel matrix and then decomposed by SVD, and the sensitive component is found by using the sensitivity factor. Finally, the singular value corresponding to the sensitive component is selected by the location factor to reconstruct the signal, and the pressure signal collected in the slurry pipeline leakage experiment system is de-noised by the method. As the preprocessing of the signal, it provides a good foundation for pipeline leakage detection. Aiming at the problem that it is difficult to extract leakage characteristics from nonlinear and non-stationary pressure signals in pipeline leakage detection, A method of pipeline leakage detection based on wavelet decomposition approximate entropy and ultimate learning machine is adopted. Firstly, the pressure signal of pipeline is decomposed by wavelet, and the first three layers with main characteristics are selected. Using the approximate entropy and kurtosis of the first three layers as eigenvector, Finally, the eigenvector is recognized and classified by the extreme learning machine. Based on the combination of wavelet transform approximate entropy and ultimate learning machine, the pipeline leakage energy can be identified effectively and accurately. Through theoretical calculation and practical experience, the pipeline leakage energy can be identified effectively and accurately. Combined with the design of slurry pipeline leakage test system, It also provides experimental data for pipeline leakage detection, introduces the sensitive factor into the traditional singular value decomposition, effectively suppresses the pipeline noise through signal reconstruction, and provides a good foundation for pipeline leakage detection. The method of combining approximate entropy with wavelet transform is used to extract the characteristic vectors of operating condition adjustment state, leakage state and normal running state, and to identify and classify effectively and accurately through the ultimate learning machine, which provides a new method for accurate leak detection of pipeline. It has certain theoretical and practical significance.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TD50

【参考文献】