We propose the use of balanced iterative reducing and clustering using hierarchies (BIRCH) combined with linear regression to predict the reduced Young's modulus and hardness of highly heterogeneous materials from a set of nanoindentation experiments. We first use BIRCH to cluster the dataset according to its mineral compositions, which are derived from the spectral matching of energy-dispersive spectroscopy data through the modular automated processing system (MAPS) platform. We observe that grouping our dataset into five clusters yields the best accuracy as well as a reasonable representation of mineralogy in each cluster. Subsequently, we test four types of regression models, namely linear regression, support vector regression, Gaussian process regression, and extreme gradient boosting regression. The linear regression and Gaussian process regression provide the most accurate prediction, and the proposed framework yields R2 = 0.93 for the test set. Although the study is needed more comprehensively, our results shows that machine learning methods such as linear regression or Gaussian process regression can be used to accurately estimate mechanical properties with a proper number of grouping based on compositional data.
Injecting CO2 into a deep geological formation (i.e., geological carbon storage, GCS) can induce earthquakes along preexisting faults in the earth's upper crust. Seismic survey and regional geo-structure analysis are typically employed to map the faults prone to earthquakes prior to injection. However, earthquakes induced by fluid injection from other subsurface energy storage and recovery activities show that systematic evaluation of the potential of induced seismicity associated with GCS is necessary. This study mechanistically investigates how multiphysical interaction among injected CO2, preexisting pore fluids and rock matrix alters stress states on faults and which physical mechanisms can nucleate earthquakes along the faults. Increased injection pressure is needed to overcome capillary entry pressure of the fault zone, driven by the contrast of fluids' wetting characteristics. Accumulated CO2 within the reservoir delays post shut-in reduction in pressure and stress fields along the fault that may enhance the potential for earthquake nucleation after terminating injection operations. Elastic energy generated by coupled processes transfers to low-permeability or hydraulically isolated basement faults, which can initiate slip of the faults. Our findings from generic studies suggest that geomechanical simulations integrated with multiphase flow system are essential to detect deformation-driven signals and mitigate potential seismic hazards associated with CO2 injection.
The earth texture with complex morphological geometry and compositions such as shale and carbonate rocks, is typically characterized with sparse field samples because of an expensive and time-consuming characterization process. Accordingly, generating arbitrary large size of the geological texture with similar topological structures at a low computation cost has become one of the key tasks for realistic geomaterial reconstruction and subsequent hydro-mechanical evaluation for science and engineering applications. Recently, generative adversarial neural networks (GANs) have demonstrated a potential of synthesizing input textural images and creating equiprobable geomaterial images for stochastic analysis of hydrogeological properties, for example, the feasibility of CO2 storage sites and exploration of unconventional resources. However, the texture synthesis with the GANs framework is often limited by the computational cost and scalability of the output texture size. In this study, we proposed a spatially assembled GANs (SAGANs) that can generate output images of an arbitrary large size regardless of the size of training images with computational efficiency. The performance of the SAGANs was evaluated with two and three dimensional (2D and 3D) rock image samples widely used in geostatistical reconstruction of the earth texture and Lattice-Boltzmann (LB) simulations were performed to compare pore-scale flow patterns and upscaled permeabilities of training and generated geomaterial images. We demonstrate SAGANs can generate the arbitrary large size of statistical realizations with connectivity and structural properties and flow characteristics similar to training images, and also can generate a variety of realizations even on a single training image. In addition, the computational time was significantly improved compared to standard GANs frameworks.
Kim, Sung E.; Seo, Yongwon; Hwang, Junshik; Yoon, Hongkyu Y.; Lee, Jonghyun
Stochastic network modeling is often limited by high computational costs to generate a large number of networks enough for meaningful statistical evaluation. In this study, Deep Convolutional Generative Adversarial Networks (DCGANs) were applied to quickly reproduce drainage networks from the already generated network samples without repetitive long modeling of the stochastic network model, Gibb’s model. In particular, we developed a novel connectivity-informed method that converts the drainage network images to the directional information of flow on each node of the drainage network, and then transforms it into multiple binary layers where the connectivity constraints between nodes in the drainage network are stored. DCGANs trained with three different types of training samples were compared; (1) original drainage network images, (2) their corresponding directional information only, and (3) the connectivity-informed directional information. A comparison of generated images demonstrated that the novel connectivity-informed method outperformed the other two methods by training DCGANs more effectively and better reproducing accurate drainage networks due to its compact representation of the network complexity and connectivity. This work highlights that DCGANs can be applicable for high contrast images common in earth and material sciences where the network, fractures, and other high contrast features are important.
Subsurface energy activities such as unconventional resource recovery, enhanced geothermal energy systems, and geologic carbon storage require fast and reliable methods to account for complex, multiphysical processes in heterogeneous fractured and porous media. Although reservoir simulation is considered the industry standard for simulating these subsurface systems with injection and/or extraction operations, reservoir simulation requires spatio-temporal “Big Data” into the simulation model, which is typically a major challenge during model development and computational phase. In this work, we developed and applied various deep neural network-based approaches to (1) process multiscale image segmentation, (2) generate ensemble members of drainage networks, flow channels, and porous media using deep convolutional generative adversarial network, (3) construct multiple hybrid neural networks such as convolutional LSTM and convolutional neural network-LSTM to develop fast and accurate reduced order models for shale gas extraction, and (4) physics-informed neural network and deep Q-learning for flow and energy production. We hypothesized that physicsbased machine learning/deep learning can overcome the shortcomings of traditional machine learning methods where data-driven models have faltered beyond the data and physical conditions used for training and validation. We improved and developed novel approaches to demonstrate that physics-based ML can allow us to incorporate physical constraints (e.g., scientific domain knowledge) into ML framework. Outcomes of this project will be readily applicable for many energy and national security problems that are particularly defined by multiscale features and network systems.
Greater utilization of subsurface reservoirs perturbs in-situ chemical-mechanical conditions with wide ranging consequences from decreased performance to project failure. Understanding the chemical precursors to rock deformation is critical to reducing the risks of these activities. To address this need, we investigated the coupled flow-dissolution- precipitation-adsorption reactions involving calcite and environmentally-relevant solid phases. Experimentally, we quantified (1) stable isotope fractionation processes for strontium during calcite nucleation and growth, and during reactive fluid flow; (2) consolidation behavior of calcite assemblages in the common brines. Numerically, we quantified water weakening of calcite using molecular dynamics simulations; and quantified the impact of calcite dissolution rate on macroscopic fracturing using finite element models. With microfluidic experiments and modeling, we show the effect of local flow fields on the dissolution kinetics of calcite. Taken together across a wide range of scales and methods, our studies allow us to separate the effects of reaction, flow, and transport, on calcite fracturing and the evolution of strontium isotopic signatures in the reactive fluids.
Using a combination of geospatial machine learning prediction and sediment thermodynamic/physical modeling, we have developed a novel software workflow to create probabilistic maps of geoacoustic and geomechanical sediment properties of the global seabed. This new technique for producing reliable estimates of seafloor properties can better support Naval operations relying on sonar performance and seabed strength, can constrain models of shallow tomographic structure important for nuclear treaty compliance monitoring/detection, and can provide constraints on the distribution and inventory of shallow methane gas and gas hydrate accumulations on the continental shelves.