Publications

ARMA conference poster for a year-round grad intern from CSM

We show that the reflection interference fringe (RIF) is formed on a screen far away from the microdroplets placed on a prism-based substrate, which have low contact angles and thin droplet heights, caused by the dual convex-concave profile of the droplet, not a pure convex profile. The geometric formulation shows that the interference fringes are caused by the optical path difference when the reflected rays from the upper convex profile at the droplet-air interface interfere with reflection from the lower concave profile at oblique angles lower than the critical angle. Analytic solutions are obtained for the droplet height and the contact angle out of the fringe number and the fringe radius in RIF from the geometric formulation. Furthermore, the ray tracing simulation is conducted using the custom-designed code. The geometric formulation and the ray tracing show excellent agreement with the experimental observation in the relation between the droplet height and the fringe number and the relation between the contact angle and the fringe radius. This study is remarkable as the droplet's dual profile cannot be easily observed with the existing techniques. However, the RIF technique can effectively verify the existence of a dual profile of the microdroplets in a simple setup. In this work, the RIF technique is successfully developed as a new optical diagnostic technique to determine the microdroplet features, such as the dual profile, the height, the contact angle, the inflection point, and the precursor film thickness, by simply measuring the RIF patterns on the far-field screen.

SSA oral presentation

Abstract not provided.

This study introduces the Progressive Improved Neural Operator (p-INO) framework, aimed at advancing machine-learning-based reduced-order models within geomechanics for underground resource optimization and carbon sequestration applications.The p-INO method transcends traditional transfer learning limitations through progressive learning, enhancing the capability of transferring knowledge from many sources.Through numerical experiments, the performance of p-INO is benchmarked against standard Improved Neural Operators (INO) in scenarios varying by data availability (different number of training samples).The research utilizes simulation data reflecting scenarios like single-phase, two-phase, and two-phase flow with mechanics inspired by the Illinois Basin Decatur Project.Results reveal that p-INO significantly surpasses conventional INO models in accuracy, particularly in data-constrained environments.Besides, adding more priori information (more trained models used by p-INO) can further enhance the process.This experiment demonstrates p-INO's robustness in leveraging sparse datasets for precise predictions across complex subsurface physics scenarios.The findings underscore the potential of p-INO to revolutionize predictive modeling in geomechanics, presenting a substantial improvement in computational efficiency and accuracy for large-scale subsurface simulations.

Uncertainty quantification (UQ) plays a vital role in addressing the challenges and limitations encountered in full-waveform inversion (FWI). Most UQ methods require parameter sampling which requires many forward and adjoint solves. This often results in very high computational overhead compared to traditional FWI, which hinders the practicality of the UQ for FWI. In this work, we develop an efficient UQ-FWI framework based on unsupervised variational autoencoder (VAE) to assess the uncertainty of single and multi-parameter FWI. The inversion operator is modeled using an encoder-decoder network. The input to the network is seismic shot gathers and the output are samples (distribution) of model parameters. We then use these samples to estimate the mean and standard deviation of each parameter population, which provide insights on the uncertainty in the inversion process. To speed up the UQ process, we carried out the reconstruction in an unsupervised learning approach. Moreover, we physics-constrained the network by injecting the FWI gradients during the backpropagation process, leading to better reconstruction. The computational cost of the proposed approach is comparable to the traditional autoencoder full-waveform inversion (AE-FWI), which is encouraging to be used to get further insight on the quality of the inversion. We apply this idea for synthetic data to show its potential in assessing uncertainty in multi-parameter FWI.

This study introduces the Progressive Improved Neural Operator (p-INO) framework, aimed at advancing machine-learning-based reduced-order models within geomechanics for underground resource optimization and carbon sequestration applications.The p-INO method transcends traditional transfer learning limitations through progressive learning, enhancing the capability of transferring knowledge from many sources.Through numerical experiments, the performance of p-INO is benchmarked against standard Improved Neural Operators (INO) in scenarios varying by data availability (different number of training samples).The research utilizes simulation data reflecting scenarios like single-phase, two-phase, and two-phase flow with mechanics inspired by the Illinois Basin Decatur Project.Results reveal that p-INO significantly surpasses conventional INO models in accuracy, particularly in data-constrained environments.Besides, adding more priori information (more trained models used by p-INO) can further enhance the process.This experiment demonstrates p-INO's robustness in leveraging sparse datasets for precise predictions across complex subsurface physics scenarios.The findings underscore the potential of p-INO to revolutionize predictive modeling in geomechanics, presenting a substantial improvement in computational efficiency and accuracy for large-scale subsurface simulations.

Characterizing the subsurface properties such as permeability and thermal conductivity is important for stimulation planning and heat production in enhanced geothermal systems (EGS). Data assimilation methods, such as the Kalman-type methods, are widely used for characterization by assimilating observed dynamic data such as pressure and temperature. However, these approaches only perform well when the parameters follow a Gaussian distribution. The geothermal sites are usually highly heterogeneous with non-Gaussian distributed complex structures such as faults and fractures, which are difficult to characterize. Over the past few years, emerging deep generative models and their impressive applications in different tasks have provided a solution to produce images with complicated features. In this work, we use the Wasserstein Generative Adversarial Network (WGAN), a deep generative model, to generate fractured images from the low-dimensional and Gaussian distributed latent space. The ensemble method, a Kalman-type data assimilation method, is then applied to the latent variables to characterize the permeability fields of a fractured geothermal site using temperature data. A synthetic two-dimensional example is presented to show the performance of our approach.

Uncertainty quantification (UQ) plays a vital role in addressing the challenges and limitations encountered in full-waveform inversion (FWI). Most UQ methods require parameter sampling which requires many forward and adjoint solves. This often results in very high computational overhead compared to traditional FWI, which hinders the practicality of the UQ for FWI. In this work, we develop an efficient UQ-FWI framework based on unsupervised variational autoencoder (VAE) to assess the uncertainty of single and multi-parameter FWI. The inversion operator is modeled using an encoder-decoder network. The input to the network is seismic shot gathers and the output are samples (distribution) of model parameters. We then use these samples to estimate the mean and standard deviation of each parameter population, which provide insights on the uncertainty in the inversion process. To speed up the UQ process, we carried out the reconstruction in an unsupervised learning approach. Moreover, we physics-constrained the network by injecting the FWI gradients during the backpropagation process, leading to better reconstruction. The computational cost of the proposed approach is comparable to the traditional autoencoder full-waveform inversion (AE-FWI), which is encouraging to be used to get further insight on the quality of the inversion. We apply this idea for synthetic data to show its potential in assessing uncertainty in multi-parameter FWI.

Characterizing the subsurface properties such as permeability and thermal conductivity is important for stimulation planning and heat production in enhanced geothermal systems (EGS). Data assimilation methods, such as the Kalman-type methods, are widely used for characterization by assimilating observed dynamic data such as pressure and temperature. However, these approaches only perform well when the parameters follow a Gaussian distribution. The geothermal sites are usually highly heterogeneous with non-Gaussian distributed complex structures such as faults and fractures, which are difficult to characterize. Over the past few years, emerging deep generative models and their impressive applications in different tasks have provided a solution to produce images with complicated features. In this work, we use the Wasserstein Generative Adversarial Network (WGAN), a deep generative model, to generate fractured images from the low-dimensional and Gaussian distributed latent space. The ensemble method, a Kalman-type data assimilation method, is then applied to the latent variables to characterize the permeability fields of a fractured geothermal site using temperature data. A synthetic two-dimensional example is presented to show the performance of our approach.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Induced seismicity is an inherent risk associated with geologic carbon storage (GCS) in deep rock formations that could contain undetected faults prone to failure. Modeling-based risk assessment has been implemented to quantify the potential of injection-induced seismicity, but typically simplified multiscale geologic features or neglected multiphysics coupled mechanisms because of the uncertainty in field data and computational cost of field-scale simulations, which may limit the reliable prediction of seismic hazard caused by industrial-scale CO2 storage. The degree of lateral continuity of the stratigraphic interbedding below the reservoir and depth-dependent fault permeability can enhance or inhibit pore-pressure diffusion and corresponding poroelastic stressing along a basement fault. This study presents a rigorous modeling scheme with optimal geological and operational parameters needed to be considered in seismic monitoring and mitigation strategies for safe GCS.