.png){kind=icon}
Return
Mixture density network efficient inversion method for TEM data with IP effect
.png){kind=icon}
PRE
AI
DOI：10.6038/cjg2025t0161.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Transient Electromagnetic Method (TEM) is a geophysical technique known for its deep investigation capability and high efficiency, widely used in engineering exploration. However, many complex geological formations contain polarizable media, which introduce Induced Polarization (IP) effects that can distort TEM data. Direct inversion of TEM data with IP effects exacerbates the inversion's multi-solution problem and may even lead to erroneous results. To address these issues, we proposed a rapid inversion method for IP-affected TEM resistivity based on mixture density network. This method uses a linear combination of Gaussian kernels to model the probability distribution of any complex geological resistivity structure. It can quickly extract resistivity information from TEM data in sign reversal and obtain the posterior probability density statistical parameters of the resistivity model, thereby providing uncertainty information about the model. Simulation and field results show that this method can output statistical parameters for 894 survey points in 2.53 seconds. The average relative error of the maximum a posteriori model is only 0.0153. This method offers a novel solution for efficient and accurate TEM imaging of complex geological structures.
Keywords:
Induced polarization effect
Transient electromagnetic method
Uncertainty inversion
Journal
C
IF:
1.4
Papers:
205
Citations:
0