基于神经网络的时间域直升机电磁数据电导率深度成像

采用G-S变换以及高斯数值积分法，形成了时间域直升机的航空电磁响应正演样本集，分析了飞机测量过程中吊舱高度变化对电磁响应的影响，并将吊舱高度的变化等效成电导率为零的假层厚度的变化，以去除高度计等的影响.以假层半空间模型为基础，研究了基于人工神经网络的电导率深度成像算法，通过分析两个三层模型的电导率深度成像结果得出，神经网络方法计算时间域航空电磁探测的视电导率精度较高，特别是对高阻层的视电导率计算.

Research on conductivity depth imaging of time domain helicopter-borne electromagnetic data based on neural network

One-dimensional time domain helicopter-borne electromagnetic responses are computed using G-S transform and Gauss numerical integral. The variation of electromagnetic responses due to bird height is analyzed, and then the variation of bird height is transformed to the variation of pseudo-layer thickness of zero conductivity to eliminate the error of the altimeter. Based on the pseudo-layer model, conductivity depth imaging is conducted using artificial neural network while the apparent height is introduced as the output of ANN. CDI obtained by artificial neural network is much closer to the true model than that by lookup table method, especially for resistive layers.