摘要
针对传统井地电阻率法反演方法存在依赖初始模型、反演时间较长等问题,提出一种基于改进Efficient Net反演方法。该方法首先通过Gmsh构造不同形状和不同电阻率的井地地电模型,通过正演得到视电阻率数据,组成数据集;然后在经典的Efficient Net网络基础上进行改进得到一种新的反演网络Efficient Net-B 0,并使用上述数据集训练该网络;最后将视电阻率数据输入到训练好的网络中,直接得到反演结果。实验结果表明,该方法能快速、准确地反演出异常体的位置、形态和电阻率值。
Addressing the issues of dependence on initial models and long inversion times associated with traditional borehole-to-surface resistivity inversion methods,an improved EfficientNet inversion method is proposed.This approach begins by constructing borehole-to-surface geoelectric models with varying shapes and resistivities using Gmsh,and generating apparent resistivity data through forward modeling to form a dataset.Subsequently,an updated inversion network named EfficientNet-B0 is derived from the classical EfficientNet network and trained using the aforementioned dataset.Finally,apparent resistivity data is input into the trained network,allowing for the direct acquisition of inversion results.Experimental results demonstrate that this method can quickly and accurately invert the location,shape,and resistivity values of anomalous bodies.
作者
余祥庆
Yu Xiangqing(Electronics and Information School of Yangtze Univesity,Jingzhou,Hubei,China,434023)
出处
《仪器仪表用户》
2025年第1期59-61,共3页
Instrumentation
