应用BP网络作TEM一维反演

瞬变电磁法（ＴＥＭ）原始数据（ｄＢ／ｄｔ）要经过视电阻率（ρｓ）公式计算，然后利用适当的反演方法拟合求取层参数，利用人工神经网络直接将归一化后的ｄＢ／ｄｔ数据与地层参数之间的隐含关系表现在ＢＰ网络中，通过一定数量样本学习后的网络，只要输入原始ｄＢ／ｄｔ数据即可求出地层参数，该网络只要经过合理的参数选择和学习，其反演的效率和精度都能达到了极好的水平。

THE APPLICATION OF BP NEURAL NETWORK TO TEM 1-D INVERSION

Before the firsthand information (d B /d t ) measured by Pulse Transient Electromagnetic Method is applied to inversion of the geoelectric parameters It must be firstly transformed by the apparent resistivity formula or some other means With the introduction of Back-Propagation Neural Network, the concealed relation between the data d B /d t and the inversed results can be shown The neural network having learnt from a number of samples can carry on the caculation of the layer parameters, only if the multi-channel data are input So long as the network goes through reasonable parameters choosing and sample studying, its inverse efficiency and accuracy would approach to a very high level