Miguel A. Corrales

selected publications

1. 

   Annealed Stein Variational Gradient Descent for Improved Uncertainty Estimation in Full-Waveform Inversion

   Miguel Corrales, Sean Berti, Bertrand Denel, Paul Williamson, Mattia Aleardi, and Matteo Ravasi

   arXiv preprint arXiv:2410.13249 2024
2. 

   Regularization strategies for Seis2Rock-based petrophysical inversion of prestack seismic data

   Miguel Corrales, Juan Romero, Nick Luiken, Hussein Hoteit, and Matteo Ravasi

   In Fourth International Meeting for Applied Geoscience & Energy 2024
3. 

   Seis2rock: A data-driven approach to direct petrophysical inversion of pre-stack seismic data: Earth and Space Science

   M Corrales, H Hoteit, and M Ravasi

   2024
4. 

   A Wasserstein GAN with Gradient Penalty for 3D Porous Media Generation.

   M. Corrales, M. Izzatullah, H. Hoteit, and M. Ravasi

   Nov 2022

   Abs HTML

   Linking the pore-scale and reservoir-scale subsurface fluid flow remains an open challenge in areas such as oil recovery and Carbon Capture and Storage (CCS). One of the main factors hindering our knowledge of such a process is the scarcity of physical samples from geological areas of interest. One way to tackle this issue is by creating accurate, digital representations of the available rock samples to perform numerical fluid flow simulations. Recent advancements in Machine Learning and Deep Generative Modeling open up a new promising avenue for generating realistic digital rock samples at low cost. This is particularly the case for Generative Adversarial Networks (GANs) due to their ability to learn complex high-dimensional distributions and produce high-quality samples. The present study introduces a novel Wasserstein GAN with gradient penalty (WGAN-GP) to generate 3D high-quality porous media samples. Moreover, a comprehensive set of evaluation metrics inspired by the geometry and topology of the structure and the fluid flow properties is established to assess the quality of the generative process.

5. 

   Plug and Play Post-Stack Seismic Inversion with CNN-Based Denoisers

   J. Romero, M. Corrales, N. Luiken, and M. Ravasi

   Nov 2022

   Abs HTML

   Seismic inversion is the prime method to estimate subsurface properties from seismic data. However, such inversion is a notoriously ill-posed inverse problem due to the band-limited and noisy nature of the data. Consequently, the data misfit term must be augmented with appropriate regularization that incorporates prior information about the sought-after solution. Conventionally, model-based regularization terms are problem-dependent and hand-crafted; this can limit the modeling capability of the inverse problem. Recently, a new framework has emerged under the name of Plug-and-Play (PnP) regularization, which suggests reinterpreting the effect of the regularizer as a denoising problem. Convolutional neural networks-based denoisers are state-of-the-art methods for image denoising: their adoption in the PnP framework has led to algorithms with improved capabilities over classical regularization in computer vision and medical imaging applications. In this work, we present a comparison between standard model-based and data-driven regularization techniques in post-stack seismic inversion and give some insights into the optimization and denoiser-related parameters tuning. The results on synthetic seismic data indicate that PnP regularization using a bias-free CNN-based denoiser with an additional noise map as input can outperform standard model-based methods.

6. 

   Bayesian RockAVO: Direct petrophysical inversion with hierarchical conditional GANs

   M. Corrales, M. Izzatullah, M. Ravasi, and H. Hoteit

   Aug 2022

   Abs HTML

   Reservoir characterization is a critical component in any oil and gas, geothermal, and CO2 sequestration project. A fundamental step in the process of characterizing the subsurface is represented by the inversion of petrophysical parameters from seismic data. However, this problem suffers from various uncertainty sources originating from inaccuracies in the measurements, modeling errors, and complex geological processes. Moreover, the non-linearity of the rock-physics model and Zoeppritz equation that constitute the modelling operator, further complicates the inversion process. In this work, we propose a novel data-driven approach where well-log information is used to obtain optimal basis functions that link band-limited petrophysical reflectivities to pre-stack seismic data. Subsequently, the inversion of such band-limited reflectivities for petrophysical parameters is framed in a Bayesian framework where a generative adversarial network is used to produce a geologically realistic prior distribution. The trained prior distribution is updated using the Stein Variational Gradient Descent and a set of representative solutions is produced that is consistent with the uncertainties in the data and the nonlinear operators.