【作者】 李莺莺；

【导师】 靳世久；

【作者基本信息】 天津大学 ， 精密仪器及机械， 2006， 博士

【摘要】 油气管道在线检测技术是保障管道运输安全的重要手段,管道在线检测的原理有多种,漏磁法和超声波法的应用最为广泛。新兴的电磁超声技术作为一种激发原理特殊的超声,因不需要耦合介质等优点,引起了无损检测领域的高度重视,本论文以管道检测的漏磁技术和电磁超声技术为重点进行研究。针对缺陷漏磁检测定量化、智能化的难题,结合油气管道检测现场的实际需求,通过理论分析和大量实验,系统分析了管道缺陷漏磁检测技术。首先根据漏磁检测的工作原理,引出基于解析法的偶极子管道缺陷理论模型,并分别以有限长矩形、无限长矩形和锥形缺陷为例做了相应的漏磁场分布研究。针对磁偶极子模型的不足,将有限元法应用到缺陷漏磁场分析中,实现了常见管道样本缺陷的漏磁场仿真。研究了漏磁信号特征的有关影响因素,例如缺陷的外形尺寸、传感探头提离距离、检测仪移动速度、管壁磁化水平、磁铁形状和管道内应力等,得到了各种因素对漏磁检测的影响规律,为相关补偿提供了理论指导。研究了感应线圈漏磁检测信号的小波去噪方法,并详细阐述了应用插值技术去除坏死通道的影响的方法。引入缺陷漏磁信号的模式识别方法,根据实测信号典型管道附件的信号特征,对环焊缝、直焊缝和螺旋焊缝进行了模式识别研究,同时根据缺陷尺寸参数识别结果采用最大安全工作压力方法进行了管道缺陷安全性评价研究。为实际应用的管道检测设备编制了漏磁检测数据分析系统软件,实现了缺陷漏磁场数据的显示,缺陷及焊缝的自动识别量化和评估。论文研究了漏磁检测的缺陷定量分析问题,将神经网络和模式识别方法应用到缺陷漏磁检测中,通过实验和仿真结合的方法建立了缺陷特征样本库,采用BP网络建立了缺陷特征量提取的网络模型;采用RBF网络建立缺陷轮廓的网络映射,并在此基础上提出一种收敛速度更快,精度更易控制的小波基函数神经网络缺陷识别算法。电磁超声技术可以进行油气管道应力腐蚀裂纹的探测。本文推导了电磁超声激发和接收过程的模式方程,并采用有限元的方法建立了EMAT换能器脉冲电涡流的有限元模型,分析了电涡流的趋肤效应,及各种因素对电涡流分布的影响,建立了感应涡流在静磁场作用下,受到洛仑兹力作用激发和接收超声波的模型。电磁超声信号由于受到噪声的污染质量较差,课题采用电磁超声探伤仪获取实测信号,并针对接收信号属于非稳态时变信号的特点,提出采用一种非线性自适应的时域信号处理方法进行信号去噪。更多还原

【Abstract】 Oil and gas pipeline in-line inspection is an important mean to ensure pipeline transmission safety. There are many principles of pipeline in-line inspection, and Magnetic flux Leakage (MFL) and ultrasonic method are used widely. Rising electromagnetic acoustic (EMA) technique, as a type of ultrasonic with special excitating principle, attracts highly attention in Nondestructive Testing (NDT) field, because of its need-no-couplant merit. MFL and EMA techniques are researched in this dissertation.Concentrating on the difficult problems that pipe defects are not evaluated quantitatively and intelligently by MFL inspection, pipeline defects MFL inspection technology is analyzed systematically, by theoretical analysis and experiments or testing. Firstly, based on MFL inspection principle, this dissertation presented dipole pipeline defect model based on analytical method, and had analyzed finite-length rectangle defect, infinite-length rectangle defect, and cone defect magnetic flux field distribution. Aimed at the lack of magnetic dipole model, finite-element method (FEM) is applied to defect leak magnetic field analysis, and the distribution of ordinary defect MFL field is simulated. The affecting factors of MFL signal features, such as defect geometry patameters, lift-off value, the speed of testing device, pipe magnetization degree, permanent magnet shape, operating pressure, are researched, then got some important law and offer theorical direction for defect signal compenstation. Research induce coil MFL inspection signal wavelet de-noise method, and explain the way to wipe off the affect of bad channel signals by interpolation. The pattern recognition method of pipe MFL signals is put forward, then girth welds, straight welds and spiral welds recognition is researched based on the feature of typical pipeline accessories practical signals. At the same time, by virtue of recognition of defect parameters, the maximum safe working pressure method is adopted to rank defects. MFL inspection data analysis software for the using pipeline inspection device is designed in this dissertation, to display MFL data and identify, quantize and evaluate pipeline defects and welds. This dissertation researched pipeline defect inspection quantitative analysis problem, applied neural network and pattern recognition methods to MFL defect inspection, and established the defect feature sample store by experiment and simulation method. The application of BP neural network is discussed firstly to extract defect feature, and established a net mapping更多还原