Using logarithmic multi-channel seismoelectric spectral ratios to estimate porosity and permeability

In this study, we devise a methodology which takes the data of logarithmic multi-channel seismoelectric spectral ratios (LMC-SESRs) as input for a broad learning (BL) plane neural network, aiming to concurrently assess porosity and permeability that are crucial hydrological parameters. We compare the sensitivity of LMC-SESRs data to porosity and permeability for a multilayer model. The results demonstrate that LMC-SESRs data exhibit sensitivity to both porosity and permeability, with a more pronounced sensitivity to porosity. Subsequently, we conduct network training and testing for porosity and permeability reconstruction using both LMC-SESRs and non-logarithmic data as inputs for the BL neural network. The results of testing dataset reveal that using LMC-SESRs data yields better reconstructions of porosity and permeability compared to using non-logarithmic data. After that, we perform simultaneous inversion of porosity and permeability for a multilayer model, validating the effectiveness of our method. Noise resistance tests are also carried out, demonstrating that the proposed method exhibits a good anti-noise ability.