改进LMD算法在管道泄漏中的应用研究
发布时间：2018-11-04 17:03
【摘要】：针对天然气管道泄漏检测过程中难以提取泄漏特征信息及泄漏定位精度低的问题,本文将局部均值分解(Local Mean Decomposition,LMD)算法应用于管道泄漏检测中,实现管道泄漏信号的分解、特征提取以及泄漏定位。首先,介绍了局部均值分解的理论算法,并将其应用在信号分解中。LMD作为处理非平稳随机信号的一种有效手段,其具备信号分解的自适应特性和完备性。但由于算法本身的影响,易产生模态混叠。针对模态混叠现象,本文利用总体局部均值分解算法(Ensemble Local Mean Decomposition,ELMD),借助辅助噪声技术,抑制LMD分解过程中模态混叠的问题。其次,有用信息在传输过程中往往会受到各种噪声的影响和干扰,使信号源中的有用信号被削减或混淆。为增强有用信号,抑制噪声干扰,保障后续提取的特征值能够代表信号特征,需要对采集的原始信号进行降噪预处理。同时为避免小波分解过程存在的技术漏洞,提出基于小波包的ELMD谱峭度联合降噪算法。该算法在ELMD分解出的各有效PF(Product Function)分量的基础上,利用谱峭度最优参数及小波包能量分布确定信号重构节点,完成对各PF分量的信号降噪。降噪后的各PF分量可表征原始信号在不同尺度下的特征。再次,通过分析管道信号的特点,对时频分析理论进行研究,提出基于时频域的自适应最优核(Adaptive Optimal Kernel,AOK)谱熵参数,定量描述信号的时频特性。对各PF分量中提取相应AOK参数用于初步判定管道是否发生泄漏及工况种类,其针对管道正常运行、管道泄漏、管道敲击具有很好的区分度,具有良好的识别准确率。最后,介绍了基于ELMD多尺度相关的管道泄漏检测算法。通过经ELMD分解得到的PF分量与互相关算法的融合,得到不同特征尺度下的时延差,从而完成对管道泄漏的定位。该算法相比直接利用原始信号进行相关计算得到的定位结果更加精确,有助于管道泄漏定位精度的提升。
[Abstract]:In view of the difficulty of extracting leakage characteristic information and the low accuracy of leak location in the process of natural gas pipeline leakage detection, this paper applies the local mean decomposition (Local Mean Decomposition,LMD) algorithm to pipeline leakage detection to realize the decomposition of pipeline leakage signal. Feature extraction and leak location. Firstly, this paper introduces the theoretical algorithm of local mean decomposition, and applies it to signal decomposition. As an effective method to deal with non-stationary random signals, LMD has the adaptive property and completeness of signal decomposition. However, due to the influence of the algorithm itself, it is easy to produce modal aliasing. Aiming at the phenomenon of modal aliasing, the problem of modal aliasing in the process of LMD decomposition is suppressed by means of the total local mean decomposition algorithm (Ensemble Local Mean Decomposition,ELMD) and the auxiliary noise technique. Secondly, the useful information is often affected and interfered by various noises in the transmission process, which reduces or confuses the useful signals in the signal source. In order to enhance the useful signal suppress the noise interference and ensure that the extracted eigenvalue can represent the signal feature it is necessary to pre-process the original signal. In order to avoid the technical loophole in the wavelet decomposition process, a ELMD spectral kurtosis joint denoising algorithm based on wavelet packet is proposed. On the basis of the effective PF (Product Function) components decomposed by ELMD, the optimal parameters of spectral kurtosis and the energy distribution of wavelet packets are used to determine the reconstructed nodes of the signal, and the signal denoising of each PF component is completed. Each PF component after denoising can characterize the characteristics of the original signal at different scales. Thirdly, by analyzing the characteristics of pipeline signals, the time-frequency analysis theory is studied, and an adaptive optimal kernel (Adaptive Optimal Kernel,AOK spectral entropy parameter based on time-frequency domain is proposed to quantitatively describe the time-frequency characteristics of the signals. The corresponding AOK parameters are extracted from each PF component to determine whether there is leakage or not and the working conditions of the pipeline are preliminarily determined. It has a good degree of distinction and good recognition accuracy for the normal operation of the pipeline, pipeline leakage and pipe percussion. Finally, the pipeline leak detection algorithm based on ELMD multi-scale correlation is introduced. Through the fusion of the PF component obtained by ELMD decomposition and the cross-correlation algorithm, the delay difference of different characteristic scales is obtained, and the location of pipeline leakage is completed. The algorithm is more accurate than that obtained by correlation calculation using the original signal directly, and it is helpful to improve the location accuracy of pipeline leakage.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7
本文编号：2310549
[Abstract]:In view of the difficulty of extracting leakage characteristic information and the low accuracy of leak location in the process of natural gas pipeline leakage detection, this paper applies the local mean decomposition (Local Mean Decomposition,LMD) algorithm to pipeline leakage detection to realize the decomposition of pipeline leakage signal. Feature extraction and leak location. Firstly, this paper introduces the theoretical algorithm of local mean decomposition, and applies it to signal decomposition. As an effective method to deal with non-stationary random signals, LMD has the adaptive property and completeness of signal decomposition. However, due to the influence of the algorithm itself, it is easy to produce modal aliasing. Aiming at the phenomenon of modal aliasing, the problem of modal aliasing in the process of LMD decomposition is suppressed by means of the total local mean decomposition algorithm (Ensemble Local Mean Decomposition,ELMD) and the auxiliary noise technique. Secondly, the useful information is often affected and interfered by various noises in the transmission process, which reduces or confuses the useful signals in the signal source. In order to enhance the useful signal suppress the noise interference and ensure that the extracted eigenvalue can represent the signal feature it is necessary to pre-process the original signal. In order to avoid the technical loophole in the wavelet decomposition process, a ELMD spectral kurtosis joint denoising algorithm based on wavelet packet is proposed. On the basis of the effective PF (Product Function) components decomposed by ELMD, the optimal parameters of spectral kurtosis and the energy distribution of wavelet packets are used to determine the reconstructed nodes of the signal, and the signal denoising of each PF component is completed. Each PF component after denoising can characterize the characteristics of the original signal at different scales. Thirdly, by analyzing the characteristics of pipeline signals, the time-frequency analysis theory is studied, and an adaptive optimal kernel (Adaptive Optimal Kernel,AOK spectral entropy parameter based on time-frequency domain is proposed to quantitatively describe the time-frequency characteristics of the signals. The corresponding AOK parameters are extracted from each PF component to determine whether there is leakage or not and the working conditions of the pipeline are preliminarily determined. It has a good degree of distinction and good recognition accuracy for the normal operation of the pipeline, pipeline leakage and pipe percussion. Finally, the pipeline leak detection algorithm based on ELMD multi-scale correlation is introduced. Through the fusion of the PF component obtained by ELMD decomposition and the cross-correlation algorithm, the delay difference of different characteristic scales is obtained, and the location of pipeline leakage is completed. The algorithm is more accurate than that obtained by correlation calculation using the original signal directly, and it is helpful to improve the location accuracy of pipeline leakage.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN911.7
【参考文献】
中国期刊全文数据库 前10条
1 田野;;基于次声波的输气管道泄漏监测系统[J];油气田地面工程;2016年10期
2 张晓威;刘锦昆;陈同彦;季文峰;冯新;;基于分布式光纤传感器的管道泄漏监测试验研究[J];水利与建筑工程学报;2016年03期
3 孙洁娣;肖启阳;温江涛;王飞;;改进LMD及高阶模糊度函数的管道泄漏定位[J];仪器仪表学报;2015年10期
4 孙洁娣;肖启阳;温江涛;王飞;;基于LMD包络谱熵及SVM的天然气管道微小泄漏孔径识别[J];机械工程学报;2014年20期
5 王保群;林燕红;焦中良;;我国天然气管道现状与发展方向[J];国际石油经济;2013年08期
6 孟令雅;付俊涛;李玉星;刘翠伟;刘光晓;;输气管道泄漏音波信号传播特性及预测模型[J];中国石油大学学报(自然科学版);2013年02期
7 刘小龙;王华;赵淑娥;陈建军;魏军;刘强;;自适应最优核时频分布在地震储层预测中的应用[J];中南大学学报(自然科学版);2012年08期
8 方亮;苏旭;赵晓龙;;天然气长输管道泄漏检测技术进展[J];化工装备技术;2012年03期
9 何存富;郑兴强;骆建伟;杭利军;吴斌;;消偏型Sagnac光纤管道泄漏检测系统及其稳定性研究[J];中国激光;2012年02期
10 范玉生;;小波和小波包变换在心电信号去噪中的应用[J];重庆科技学院学报(自然科学版);2010年01期
中国硕士学位论文全文数据库 前6条
1 张冉;城市管道燃气防泄漏监测技术研究[D];东华理工大学;2016年
2 胡月;基于负压波原理的输油管线泄漏监测技术研究[D];长春理工大学;2016年
3 薄瑞瑞;基于LMD的振动信号处理及故障特征提取研究[D];内蒙古大学;2015年
4 段乐峥;基于HHT的供水管道泄漏检测研究[D];厦门大学;2014年
5 刘盈;基于次声波的煤气管道泄漏监测系统研究[D];电子科技大学;2010年
6 王久龙;基于红外成像技术的埋地管道泄漏定位实验研究[D];大庆石油学院;2008年
，本文编号：2310549
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2310549.html
