Abstract:The application of the Cole-Cole model within time-domain induced polarization (TDIP) forward field modeling shows that the model parameters can characterize time-varying states of the TDIP field and support observed data analysis. The Cole-Cole model contains real and imaginary parts, and it requires a frequency-to-time conversion for TDIP forward modeling. However, the TDIP field is usually expressed by a real number, and its intuitive time-varying states field intensity increases with charging time. Therefore, the forward model should be constructed in a simpler form. We have aimed to develop a forward model using mathematical functions not based on physical principles. The Weibull (WB) growth model, which is primarily used to describe the time-varying curve features in regression analysis, is introduced into the basic algorithm of the TDIP forward model. Subsequently, a forward expression of the TDIP effect is established. Based on the time-varying shape and scale parameters, this expression describes the time-varying rate and relaxation states of the TDIP fields. Furthermore, based on the extensively used conjugate gradient optimization, an apparent WB parameter scheme is initiated to calculate the spectral parameters that represent the relaxation and time-varying rate obtained from the multi-time-channel TDIP data. Finally, this scheme is applied to interpret the different simulated and actual TDIP data. The results demonstrate that the WB growth model can be used for the TDIP forward model without involving physical principles, the model parameters without specific physical significance can be used to represent the time-varying states of TDIP fields, and apparent WB parameters can be used to discern different TDIP observed data. The setting of the TDIP forward model and model parameters can actually be more flexible and diverse, so as to obtain simpler forward expressions and ensure a highly efficient inverse solution.

Data-driven modeling of time-domain induced polarization

Processing Time-Domain Induced Polarization Data Using Variational Mode Decomposition

Bayesian inference of petrophysical properties with generative spectral induced polarization models

Predictive Modeling with Multiresolution Pyramid VAE and Industrial Soft Sensor Applications.

Targeted Analysis of High-Risk States Using an Oriented Variational Autoencoder

Time-domain solution of Cole-Cole model with induced polarization method

Physics and Chemistry from Parsimonious Representations: Image Analysis via Invariant Variational Autoencoders

Feature-based Magnetotelluric Inversion by Variational Autoencoder Using a Subdomain Encoding Scheme

Variational Autoencoding Molecular Graphs with Denoising Diffusion Probabilistic Model

Deep generative learning for exploration in large electrochemical impedance dataset

Industrial Data Modeling with Low-Dimensional Inputs and High-Dimensional Outputs

Modeling conditional distributions of neural and behavioral data with masked variational autoencoders

Weighted Laterally Constrained Inversion of Airborne Transient Electromagnetic Data with IP Effects

Learning identifiable and interpretable latent models of high-dimensional neural activity using pi-VAE

$Φ$-DVAE: Physics-Informed Dynamical Variational Autoencoders for Unstructured Data Assimilation

On the use of time series to improve signal processing of electrical data

TimeVAE: A Variational Auto-Encoder for Multivariate Time Series Generation

Time-domain Induced Polarization Forward Modeling and an Apparent Spectral Parameter Solution Based on the Weibull Growth Model

Scalable Modeling of Spatiotemporal Data using the Variational Autoencoder: an Application in Glaucoma

Predictive variational autoencoder for learning robust representations of time-series data

Unsupervised Intense VSP Coupling Noise Suppression With Iterative Robust Deep Learning