【作者】 谢林涛；

【导师】 谢品芳；

【作者基本信息】 重庆大学 ， 电气工程， 2009， 硕士

【摘要】 瞬变电磁法(Transient Electromagnetic Method)是一种时间域电磁法,是在没有一次场背景下观测研究二次场,大大简化了对地质对象产生异常的研究,具有简单易行、信息丰富、精度较高的特点。但是,后期数据处理量大,正反演计算复杂,很难做到实时成像,用于现场指导。随着瞬变电磁技术的不断发展,对它的要求也越来越高,如何快速成像成为瞬变电磁技术发展的热门方向。本文通过介绍瞬变电磁探测方法,讨论了几种常用装置在应用中的特点,以中心回线方式装置为例,讨论了瞬变电磁成像中数据处理的过程,论文主要做了以下工作:①分析了实际发射电流非阶跃下降的情况,提出了“先校正后计算”的数据预处理方法。由W-TEM发射机,讨论了中心回线方式下发射电流阶跃和斜阶跃下降瞬变响应的关系,结合斜阶跃下降电流对测量结果的校正方法对模型进行仿真实验,提出了短延时发射机情况下瞬变电磁响应数据的简单较正,避开了复杂的较正公式,为快速计算电阻率提供前提。②讨论了瞬变电磁响应常规的数值计算方法,提出线性加速和对分预估计两种改进的视电阻率数值计算方法。前一种属于变步长迭代法,通过分析文献中计算视电阻率的常用方法,提出的加速变步长的计算方法;另一种通过分析响应公式中核函数的特点,针对仿真中遇到的迭代初始值选取不定的问题,提出的适合本文应用的对分预估计电阻率计算方法。③把人工神经网络应用到本文所研究的课题中,提出基于自变量和瞬变场参数为输入的两种神经网络训练方法。借用神经网络处理不确定性非结构化信息的特点来拟合二次涡流曲线,讨论了输入与输出向量并分析了它们对训练结果的影响;把神经网络引入到常规视电阻率计算中,选择样本数据,比较各种训练算法的精度和收敛性,确定隐含层单元数,实现基于神经网络的数据处理方法。最后用理论模型和实例计算证明了数值计算和神经网络计算的有效性,并对算法做了对比研究。④分析了其它装置下瞬变响应的特点,把本文所研究的方法进一步推广。从响应公式入手,讨论了各参数变量对工程测量数据的影响,分析了数值计算中的关键问题,最后,从本文研究角度出发,对计算提出了可行性的建议。更多还原

【Abstract】 Being a method of time-domain electrical magnetic surveying, the Transient Electromagnetic Method (TEM) that is used to observe and investigate the secondary field can greatly simplify investigation of anomaly of buried geologic bodies on condition that the primary field does not exist. It has some advantages, such as concise, precise, be rich in information and so on. On the other hand , people can not get the picture timely for guiding the job being done, because there is a large quantity of data and the processing is quite complex in latter time. The requirement to transient electromagnetic technology is becoming high with its continuous development. How to image fast becomes a hot direction of TEM.In this thesis, by introducing the transient electromagnetic method of detection, the characteristics of different receiver devices are discussed in application. Data-processing is discussing by taking the example of central-loop device in the process of TEM imaging. The main tasks of the thesis are presented as follows:①The condition of transmitter current wave decay in non-step in actual experiment is analyzed. The pretreatment method of“correction first, then calculation”is presented. Considering of W-TEM transmitter, the relationship between ramp turn-off and step turn-off transmitter current under central-loop device is discussed. Simulation testing to model is done combining with calibration method about turn-off effect from transmitter current. A simple method calibrating TEM response data, which can avoid the complicated mathematical treatment process and provide prerequisite for calculation resistivity ,is given under short decay turn-off current.②Conventional method of calculating resistivity in TEM response is analyzed. Two other fast improved methods of calculating resistivity, which are linear accelerated iterative algorithm and bisection pre-estimate method, are given. The former one is belonging to variable step-size iterative. Improved variable step-length numerical calculation is presented based on analysis of apparent resistivity methods commonly used. The latter one that suit the request in this thesis is bisection pre-estimate of numerical method, it also solves the problem of selecting initial value encountered in Newton’s method. Simulation tests proved that it is valuable in engineering applications.③Artificial neural network (ANN) is used for the research in this thesis. Two methods based on independent variable and transient field parameters are given. Response curves of TEM are simulated by ANN. Input and output vectors and their impact on the training results are discussed. Choosing sample data, comparing the precise and convergent of different training algorithms, determining the number of hidden layer, at last, calculating resistivity using ANN is realization. The practicability of numerical calculation and ANN calculation algorithm are proved with theoretical models and field data experiments. Comparative study between them is also done.④Characteristic about transient response in other devices is analyzed. The method discussed in this thesis is extended. Variable parameters on the impact of engineering measurements and key issues in numerical calculation are discussed. Feasible suggestion about calculation is presented from the point of view in this thesis.更多还原