管道漏磁检测数据特征提取及特征分析方法研究

发布时间：2018-03-08 04:03

本文选题：异常判断　切入点：特征提取　出处：《东北大学》2014年硕士论文　论文类型：学位论文

【摘要】：管道运输是一种极为重要的运输方式,在工业生产和经济发展中发挥着越来越重要的作用。但是,随着管线运行时间的增加,管道内外壁难免会发生一定程度的腐蚀。由于腐蚀而产生的缺陷往往会导致灾难性事故的发生。因此,利用漏磁方法对管道进行缺陷检测具有十分重要的意义。漏磁检测的核心就是通过传感器检测到的漏磁信号,来对缺陷的位置、形状做出准确的判断。因此,对于漏磁信号中异常数据的判别和特征提取是整个漏磁检测过程中十分关键的一个环节。针对这一重要环节,本文提出了对异常数据进行判别、特征提取和对数据特征进行分析的方法。主要对以下三个问题进行了研究：对于单个缺陷异常数据的判别和特征提取、对于位置复杂缺陷漏磁检测数据的处理方法以及异常数据特征与缺陷形状之间的联系。首先,在差分超限法和阈值法的基础上提出幅值差分法对检测数据进行分析,提取出异常部分。对提取出的异常数据进行特征提取。主要提取的特征有,轴向峰谷值、谷谷值、拐点间距、波形面积、波形能量等。其中针对拐点间距较难提取的问题,本文提出了一种基于小波变换的异常数据拐点检测方法。管道内外壁中不只存在单个缺陷,同时存在很多位置较为复杂的缺陷。针对这一问题,提出了对不同类型管道焊缝进行判断检测的方法。并解决了焊缝中存在缺陷的判别问题。另外,解决了缺陷簇中距离过近缺陷产生的相互覆盖的异常数据的处理问题。首先对每个缺陷产生的漏磁数据进行分离,并应用对数据进行拟合的方法预测填补数据遗失部分,以达到能够进行特征提取的目的。最后,通过直观分析,分别得到针对缺陷长、宽、深变化较大的数据特征。利用主元分析法对多个特征进行分析处理。将多个特征形成的多维数据进行降维,得到较少的几个主成分,利用主成分来代替多个特征数据。同时,分别获得对缺陷的长、宽、深评价作用比较大的数据特征。
[Abstract]:Pipeline transportation is an extremely important mode of transportation, which plays a more and more important role in industrial production and economic development. A certain degree of corrosion will inevitably occur in the inner and outer walls of the pipeline. Defects caused by corrosion will often lead to catastrophic accidents. It is very important to use magnetic flux leakage method to detect defects in pipelines. The core of magnetic leakage detection is to make accurate judgment on the position and shape of defects through magnetic leakage signals detected by sensors. The discrimination and feature extraction of abnormal data in magnetic flux leakage signal is a key link in the whole process of magnetic flux leakage detection. In view of this important link, this paper proposes to distinguish the abnormal data. The methods of feature extraction and data feature analysis. The following three problems are studied: the discrimination and feature extraction of single defect anomaly data, The processing method of magnetic flux leakage detection data with complex position defects and the relationship between abnormal data characteristics and defect shape are discussed. Firstly, the amplitude difference method is proposed to analyze the detection data on the basis of differential over-limit method and threshold method. The main features extracted are: axial peak-valley value, valley value, inflection point spacing, waveform area, waveform energy, etc. Among them, it is difficult to extract the inflexion distance. In this paper, a method of detecting inflexion of abnormal data based on wavelet transform is proposed. There are not only a single defect in the inner and outer wall of the pipeline, but also a lot of defects in the complex position. This paper puts forward the method of judging and detecting the weld seam of different types of pipeline, and solves the problem of judging the defects in the weld seam. In this paper, we solve the problem of dealing with the abnormal data overlying each other in the defect cluster. Firstly, the magnetic flux leakage data generated by each defect is separated, and the method of fitting the data is used to predict and fill the missing part of the data. In order to achieve the goal of feature extraction. Finally, through intuitive analysis, the defect length, width, The principal component analysis method is used to analyze and deal with multiple features. The multidimensional data formed by multiple features are dimensionally reduced, and a few principal components are obtained, and principal components are used to replace multiple feature data. At the same time, The length, width and depth of the defect are obtained respectively.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U178

【相似文献】