Rock fragmentation requires a significant amount of energy and poses a considerable economic challenge for the mining industry. Microwave treatment has emerged as a viable technology with the potential utility to decrease the strength of rocks prior to excavation. This study presents a coupling of finite element (FE) and discrete element methods to investigate the electromagnetic-thermal-mechanical responses on microwave-treated rocks in terms of fracture formation, coalescence and propagation. The developed numerical models amenably predict the experimental results in terms of heat absorption, surface temperature, and the timing of cracking or spallation. The percentage differences between the experimentally measured and numerically predicted cracking/spallation times was below 12% across the investigated microwave powers. Numerical results further indicated that the onset of fracture nucleation consistently occurred at an absorbed energy of approximately 5.5 kWh/t, irrespective of the applied microwave power. The developed model was subsequently employed to examine the effect of microwave power on the fracture development over a range of power levels up to 100 kW. The findings indicated that, for a given absorbed energy, higher power levels promote more pronounced fracture development, and beyond a certain thermal power density, further increases in forward power yield minimal gains in the percentage of microwave-induced fractures. In addition, the effect of distance between the rock and microwave antenna on the percentage of induced fractures was explored.
The safe and efficient operation of enhanced geothermal systems (EGS) is challenged by the dynamic evolution of reservoir matrix failure and fracture deformation. Factors including fracture number, injection parameters and rock properties affect both matrix and fracture in EGS while their relative influence degree and key mechanism remains unclear. This study numerically investigates the multistage evolution of matrix failure and fracture deformation using two-dimensional fully coupled thermo-hydro-mechanical (THM) models based on data from the GR1 Well at the Qiabuqia geothermal site. Models with variations in fracture number, injection parameters (injection temperature and pressure) and rock properties (elastic modulus and Poisson's ratio) were established to simulate EGS operation. Our simulation results reveal that over the 30-year EGS operation, matrix failure exhibits a three-stage dynamic evolution, while fracture deformation process can be divided into two distinct stages. The sensitivity of effect factors on reservoir evolution was ranked, indicating that injection pressure exerts the most significant influence which exceeding 44MPa substantially increases matrix failure and fracture deformation. Variations in fracture number exert a complex influence as increased number amplifying matrix failure while constraining extreme fracture deformation. The influence of injection parameters on fracture deformation is reflected in differences in the amplitude of aperture evolution, whereas rock properties govern the initial fracture aperture. Research into influencing mechanisms elucidates the primary way governing matrix failure and fracture deformation is the alteration of effective stress in reservoir. The research results contribute guidance for the construction and safe operation of EGS project with similar geological conditions.
Hydraulic fracturing often generates fractures at various inclinations, particularly gently dipping fractures that propagate along bedding planes in shale gas-bearing formations. However, how such low inclinations affect proppant embedding-depth distributions and fracture-surface roughness remains unclear. To understand these effects, a self-developed apparatus was used to conduct single-fracture conductivity tests on multilayer-propped shale fractures at closure pressures of 20 and 40MPa and inclination angles ranging from 0° to 3°. Three-dimensional surface scanning was used to determine the average proppant embedding depth, embedding-depth distribution, and joint roughness coefficient (JRC). On this basis, Hertzian contact theory was extended from a single spherical particle to a multilayer proppant packing system. The fractal dimension (Ds) of the embedding-depth distribution was then introduced to develop a JRCDs model accounting for closure pressure and fracture inclination angle. The results showed that higher closure pressure increased both the average proppant embedding depth and fracture-surface roughness. As the inclination angle increased, the average embedding depth initially decreased and then changed only slightly. Compared with the JRCαn model based on average embedding depth, the JRCDs model reduced the MAE, RMSE, and MAPE by 63.4%, 60.9%, and 58.0%, respectively. This theoretical framework provides a useful basis for optimizing hydraulic-fracturing parameters in shale gas production.
Predicting the onset of reservoir-scale instabilities, such as compaction bands, remains a significant challenge because they often nucleate under ”subcritical” conditions, well before standard linear stability limits are reached. This paper identifies the Level-3 HMC (Hydro-Mechanical-Chemical) Gateway as the primary engine driving this process. We characterise a universal five-node Hasse trajectory that describes the structural evolution of a reservoir: from grain-scale chemical fluctuations to an active energy-pumping gateway, and finally to a ”frozen” macroscopic hierarchy. Using the Cooper Basin (Australia) and the Heyuan Reef (China) as a conjugate pair, we demonstrate that the former is an incipient, ”mid-cascade” system. We show that the irregular localisation in the Cooper Basin indicates a transitional dynamic regime in which fluid and solid pressure oscillations evolve toward a stationary mode. Because this state is confined to a remarkably narrow parameter range, it allows for identification of deep reservoir properties such as the chemical dissolution kinetics (Damköhler number, Da) from the observed style of dynamics. In contrast, Heyuan represents the ”post-cascade” state where the chemical fuel is exhausted, leaving behind a rhythmic Turing pattern which does not allow inversion of Da but contains information about the self-diffusion processes. This framework suggests that field observations are not static states but transient snapshots of a self-organising geodynamic engine, offering a new path for high-fidelity reservoir characterisation.
Underground hydrogen storage (UHS) in porous reservoirs is governed by coupled multiphase flow, heat transfer, poromechanical stress redistribution, geochemical partitioning, and microbially mediated hydrogen consumption. A thermo–hydro–mechanical–chemical–biological (THMCB) benchmarking framework is developed and tested using an illustrative Ketzin-referenced case. The objective is to provide a consistent basis for nondimensionalization, process screening, pore-to-reservoir upscaling, and simulator verification. The main contribution is the integration of established THMCB equations within one scale-consistent benchmarking workflow. A thermodynamically consistent reservoir-scale THMCB model is formulated. It is used to analyze cyclic operation using hydraulic and thermal penetration behavior together with control-volume dissolution–consumption dynamics. The numerical results for repeated operating cycles use a 15-state control-volume ODE system. In the central Ketzin-referenced benchmark, the imposed flow range remains Darcy–capillary dominated, hydraulic perturbations penetrate the reservoir thickness more rapidly than thermal oscillations, and gas–liquid hydrogen mass transfer is fast relative to the cycle time (Damt∼106–107). The broad Forchheimer-coefficient sweep also identifies a non-Darcy near-well sector. Biogeochemical sweeps indicate sulfate-reduction dominance under the adopted chemical and microbial assumptions. The 50-year benchmark compares the characteristic length and time scales of gas–liquid hydrogen mass transfer, reaction, heat transport, poroelastic response, and periodic convergence. The 50-year calculation reports hydrogen return, reaction loss, and cycle-end repeatability using molar amounts per unit pore volume. These values are benchmark outputs and do not predict UHS performance at Ketzin. The model provides analytical tests for model simplification, calibration, and verification in porous-reservoir UHS .
Understanding subsurface conditions is critical in both natural and anthropogenic environments, enabling accurate risk assessments and informed construction practices. Adequate characterization of geological and engineering properties is particularly important in urban areas where the subsurface interacts with a complex infrastructure system. The Chicago Loop district exemplifies these interactions, presenting a dense concentration of high-rise buildings, complex subsurface infrastructure, and continuous upgrades to the built environment. With reference to this case study, through this work we make two primary contributions: (1) Develop an integrated framework for characterizing the subsurface using geological and geotechnical data synthesized from borehole drilling records, recent geological interpretations, and historical datasets of subsurface conditions; (2) Construct detailed 3-D geologic and geotechnical models of the subsurface that represent the complex and spatially variable nature of the properties of key geologic and engineering units. Together, these contributions provide an example of how unifying the current understanding of the subsurface between the engineering and geoscience technical fields improves future analyses of subsurface behavior. Our novel approach has significant implications for underground urban planning, resource management, and hazard mitigation.
Heterogeneous shale failure is controlled by the coupled effects of mineralogical heterogeneity and pore-fracture architecture, but their cross-scale influence on damage evolution and brittleness remains difficult to quantify. In this study, high-resolution CT imaging, SEM-EDS mineral mapping, and nanoindentation testing were integrated to construct a microstructure-informed multi-component digital rock model. The actual CT-derived pore geometry and SEM-EDS-informed mineral distribution were incorporated into a PFC-based discrete element model, and phase-specific mechanical properties were assigned according to nanoindentation results. The coupled effects of porosity and stiff-mineral content on contact-force evolution, displacement localization, strength, and stiffness were then investigated. The results show that increasing porosity interrupts continuous force-chain pathways, promotes displacement localization, and reduces both peak strength and elastic modulus. In contrast, increasing stiff-mineral content enhances load-transfer efficiency by forming a mineral-controlled load-bearing framework, while the deformability mismatch between stiff minerals and the soft matrix promotes local stress concentration and deformation incompatibility. Based on the simulation and experimental results, a statistical damage constitutive model was established by incorporating porosity-dependent equivalent strength, multicomponent weighted strength, and equivalent elastic modulus. The predicted peak strength and elastic modulus show good agreement with PFC simulation results and laboratory measurements within the tested range, with relative residuals generally constrained within ±10% for strength and ±8% for stiffness. Furthermore, an energy-based brittleness index was used to characterize the brittle-ductile response under different porosity and mineral-content conditions. This study provides a microstructure-informed framework for linking mineral-pore heterogeneity, damage evolution, and brittleness characterization in heterogeneous shale.
Understanding contact characteristics in rock fractures is essential for predicting fluid flow and heat transfer in geothermal and other subsurface energy systems. This study systematically investigates how contact obstacles, characterised by contact area ratio (c) and spatial distribution mode (uniform versus normal), govern coupled thermo‑hydraulic processes in single fractures. A three‑dimensional parallel‑plate fracture model incorporating discrete contact elements is developed to isolate contact effects while eliminating geometric confounders such as roughness and aperture variability. Finite‑element simulations are performed for contact ratios ranging from c= 0% to 80% under different contact distributions. The results show that increasing contact area ratio markedly reduces hydraulic aperture and flow channel connectivity, leading to pronounced changes in flow regime and heat transfer behaviour. A critical threshold at c≈ 40% is identified, beyond which flow transitions from channelised to highly tortuous patterns, significantly delaying thermal breakthrough. Moreover, contact distribution exerts a strong control on flow path geometry: uniform contact distributions promote relatively straight preferential pathways and more efficient advective heat transport, whereas normal distributions induce higher tortuosity and enhanced conductive heat exchange with the fracture walls. Based on these mechanisms, a modified analytical solution for temperature evolution is proposed by introducing a hydraulic aperture formulation that explicitly accounts for both contact ratio and flow tortuosity. The analytical predictions show excellent agreement with numerical results. These findings provide mechanistic insights into contact‑controlled thermo‑hydraulic behaviour in fractured rock and offer general implications for the design and optimisation of geothermal and other geo‑energy extraction systems.
Pore volume compressibility is a key geomechanical parameter in geology and reservoir engineering, as it strongly affects hydrocarbon production in sedimentary porous rocks. Most available approaches rely on empirical relations or isotropic assumptions, limiting their applicability to anisotropic pore systems. In this work, pore volume compressibility is estimated using a micromechanical model based on the Mori–Tanaka–Benveniste (MTB) scheme for both isotropic and Vertically Transversely Isotropic (VTI) rocks, accounting for porosity, mineralogy, pore geometry, and pore fluid. The results reproduce the characteristic exponential decrease in pore compressibility with porosity reported in the literature and highlight two kinds of pore compressibility as functions of uniform volumetric strain and uniform normal stress in VTI media, strongly controlled by pore shape (aspect ratio). Oblate pores (α<1) exhibit higher compressibility than prolate pores (α>1), whereas compressibility values converge to a unique value for spherical pores (α=1). The estimated pore compressibilities are applied to hydrocarbon volume estimation using the General Material Balance Equation (GMBE) for a volumetric, undersaturated reservoir. When oblate shapes are present with aspect ratios less than approximately 0.1, disruptive behavior occurs between the two compressibilities, directly impacting hydrocarbon estimates. Finally, a first prediction of the anisotropic pore compressibilities of two well-characterized formations, Bazhenov and Niobrara, is given. The proposed mechanical framework provides a physically consistent tool for estimating pore compressibility based on the fundamental concepts of mechanics and physics of anisotropic rocks and improves hydrocarbon volume estimation through micromechanical modeling.
Site-specific fracture network characterization in deep underground rock masses is essential for safety assessment of subsurface engineering systems, yet available data are typically restricted to one-dimensional borehole observations. Genetic discrete fracture network (DFN) models simulate fracture nucleation, growth, and arrest through physically motivated rules, and previous studies have demonstrated their ability to reproduce realistic network topologies. However, applying these models to specific sites requires inverse calibration of the governing growth parameters, which cannot be directly measured in situ. Existing calibration approaches—based mainly on expert-guided manual trial-and-error or trace-based optimization using two-dimensional outcrop data—are impractical when only deep borehole data are available. To address this gap, this study proposes a borehole-conditioned genetic DFN modeling framework comprising two integrated components: (i) a generation algorithm that incorporates during-generation P10 feedback control, set-specific nucleation termination, and inter-set arrest interactions to condition the network on target borehole observations; and (ii) a surrogate-based parameter optimization methodology that calibrates growth parameters against additional metrics, including the fracture size distribution exponent, neighboring-borehole P10, and volumetric fracture intensity (P32). The framework is validated using conceptual models and field borehole data from the KAERI Underground Research Tunnel (KURT). The target-borehole P10 was reproduced with high accuracy, and the optimization results showed metric-dependent agreement with the target metrics. Minor residual discrepancies, likely associated with local geological heterogeneity and sampling bias, were mitigated through an auxiliary post-processing step affecting less than 1% of the generated fractures. The framework extends automated genetic DFN calibration to borehole-constrained environments for deep underground applications.
Existing pre-drilling pore pressure prediction methods have significant errors and cannot effectively guide drilling engineering design and drilling operations, leading to frequent incidents such as well leaks and overflows. To achieve accurate pre-drilling prediction of pore pressure in the upper clastic rock layer of the Penglai gas field, a new pore pressure prediction model was established using porosity-corrected acoustic travel time data. Considering the characteristic of well-developed fractures in the area, a 3D (three-dimensional) pore pressure prediction model based on a sequential Gaussian co-Kriging 3D interpolation algorithm was established, and a 3D pore pressure data volume is constructed to achieve pre-drilling prediction of pore pressure. The results show that: Based on actual drilling results, compared to models that do not account for porosity effects, the average absolute error of the corrected model was reduced from 12.92% to 3.52%; the sequential Gaussian co-Kriging interpolation algorithm simultaneously considers the randomness of fracture-fault zones and the continuity of non-fracture-fault zones; The pore pressure values obtained from the 3D data set show good agreement with actual drilling data, with errors ranging from ±10%. The proposed workflow is intended for development drilling in mature oil and gas fields where calibration wells are available, enabling pore-pressure prediction for planned wells prior to drilling rather than for frontier exploration without well control.
Accurate permeability prediction is important for geological CO2 sequestration, as it directly determines the injection rate, plume migration behavior, and storage capacity. However, conventional empirical correlations such as Kozeny-Carman equation, have low predictive accuracy in heterogeneous sedimentary formations. This study develops and validates machine learning (ML) models for permeability prediction in the Mesozoic reservoir succession of Petrel Sub-basin, Australia, estimated storage capacity of 15.9 Gt CO2. Nine ML algorithms were trained (linear models, ensemble methods, support vector regression, and neural networks) on core samples from four exploration wells with depths varying from 761 to 3043 m and permeability varying from 0.00002 to 497 mD. Gaussian noise augmentation and Leave-One-Out Cross-Validation (LOOCV) addressed small-sample limitations. Boosting-based ensemble methods significantly outperformed traditional empirical correlations, with the highest accuracy (R2: 0.9945, RMSE: 6.83 mD) obtained by AdaBoost. LOOCV evaluation gave more conservative estimates with Gradient Boosting R2 of 0.7855, which gives realistic performance expectations for geologically similar settings. Gaussian Process regression produced 91.7 and 77.8 % coverage of reservoir facies (>1 mD) and seal units (<0.01 mD), respectively, with interval widths ranging between 2.7 and 6.8 log10 units. Permutation importance supported SHAP rankings even with the small sample size. SHAP determined the engineered depth-porosity product as the most significant predictor, which incorporates coupled burial-compaction effects. Neural networks performed poorly (R2 = 0.8411), indicating that traditional ML is better at data-limited subsurface characterization. The framework is applicable for injectivity, seal-integrity, and probabilistic capacity screening for carbon storage appraisal, under similar settings and input-parameter ranges.
Salt cavern gas storage is frequently subjected to irregular and high-frequency injection–withdrawal cycles across multiple time scales, making long-term stability prediction critical for safe underground energy storage. In many regions, however, salt cavern projects remain at early operational stages, resulting in limited long-term monitoring data. Moreover, monitoring datasets from different regions and salt formations are often not interoperable, which restricts the generalizability and adaptability of conventional machine-learning models. To overcome these constraints, this study proposes a zero-shot learning framework for long-term stability prediction of salt caverns under data-scarce conditions. The model integrates fundamental design parameters, operational conditions, and salt-rock geomechanical properties, and augments them with six physics-informed semantic features to improve cross-domain generalization. The framework predicts the maximum displacement of the cavern under unknown working conditions. On the test set, the proposed approach achieves an R²of 0.9, showing a marked improvement over models without semantic features. Among the evaluated architectures, the Transformer provides the best performance (R²= 0.9845; RMSE=0.0617). These findings demonstrate that zero-shot learning enables long-term stability prediction for newly constructed salt cavern gas storage under data-scarce conditions, thereby offering a practical tool to support the safe operation and maintenance of underground gas storage facilities.
Safe carbon storage and sequestration in deep geological formations require a thorough understanding of caprock integrity to prevent potential CO2 leakage into the atmosphere or groundwater. Caprock characteristics can vary significantly depending on the geological setting and environmental conditions. In this work, a Kingsport formation limestone rock sample was experimentally injected with supercritical carbon dioxide (scCO2) to determine the viability of the rock as a sealing layer for CO2 storage. The mineralogy of the rock was identified using high-speed coupled micro-X-ray fluorescence/ micro-X-ray diffraction and electron probe micro-analysis. Calcite was the dominant mineral constituent with K-feldspar veins, very fine quartz grains dispersed throughout the rock, and a few 750 μm to 300 μm dolomite inclusions. The tested sample was initially subjected to subsurface-like conditions by application of a 0.1 M NaCl solution under a confining pressure of 30 MPa at 70 °C. Afterward, scCO2 was applied in a flow experiment at a pressure gradient of 1.3 MPa. Minimal penetration of scCO2 after prolonged exposure (about 500 h) was observed due to extremely low permeability of the rock estimated to be ⁓1.6 nano-Darcy. No visible pore structure was detected in pre- and post-testing samples at the resolution of micro-computed tomography imaging. The results indicate very low porosity and undetectable chemical interaction between scCO2 and the limestone sample, indicating a high sealing efficiency of the rock and its suitability for use as a caprock.
Freeze-thaw (F-T) cycles play a pivotal role in rock degradation and expedite rock fracturing in cold regions, thereby leading to catastrophic failures in rock engineering. Therefore, it is of great significance to understand the impact of F-T cycles on rock fracturing and to formulate a theoretical framework for rock engineering projects in cold regions. This study employed experimental testing, discrete element method (DEM) modeling and micromechanical theoretical analysis on notched semi-circular bend (NSCB) specimens of red sandstone to assess the fracturing mechanism under Mode-I loading. A linear parallel bond model (LPBM) within the DEM framework was calibrated to replicate the experimental load-displacement response, acoustic emission (AE) counts, and macroscopic fracture trajectory. Results reveal that increasing F-T cycles significantly diminishes peak load and fracture toughness (KIC) and leads to earlier microcrack initiation, as evidenced by simulated AE counts. The analysis of internal stress distribution and pore structure indicates that frost cracking promotes the transition of micropores to macropores, thereby greatly altering water retention behavior and accelerating rock degradation. A micromechanical model is adapted to correlate pore-scale properties with fracture development. The results show that swelling pressure induced by ice formation within the pores increases the stress intensity factor (SIF) (Kcdf) at crack tips. When Kcdf exceeds the material's KIC, subcritical crack growth is initiated. The validated DEM model also demonstrates its reliability in capturing micromechanical damage mechanisms associated with F-T cycles. The obtained results may provide insightful theoretical and experimental framework for evaluating the structural stability and predicting the fracture hazard in rock engineering operations in cold climates.
A major challenge in the implementation of Engineered Barrier Systems (EBS) for the geological disposal of nuclear waste is monitoring their response to thermal and hydraulic loading. This applies to mock-up scale tests designed to validate the EBS concept via 'wired' local sensors, but also to future full-scale EBSs to be monitored 'wireless' via electrical geophysical techniques for EBS validation during the operational stage. To underpin the monitoring of EBS, this study investigates experimental techniques for measuring the volumetric water content, suction, and electrical conductivity of bentonite-based materials. It is demonstrated that Time Domain Reflectometry (TDR) probes can be successfully used to monitor volumetric water content provided that i) the measured apparent relative dielectric permittivity is corrected for the effect of clay electrical conductivity and ii) the calibration equation accounts for the effect of bulk density and adsorbed water. It was also shown that the electrical conductivity inferred from TDR and direct current four-point method measurements may differ substantially, and such a difference can be associated with the frequency-dependent contribution of free and adsorbed water to bulk electrical conductivity. Finally, encapsulated Micro-Electro-Mechanical Systems (MEMS) sensors were demonstrated to measure total suction with satisfactory accuracy over the full relative humidity range.
Accurate estimation of the maximum horizontal stress (SH) is essential for wellbore stability analysis, reservoir development, and subsurface energy applications. Conventional breakout-based methods often rely on elastic assumptions, leading to substantial inaccuracies in formations exhibiting inelastic deformation. This study presents an integrated experimental-numerical-machine learning framework for SH estimation that explicitly accounts for inelatic rock behavior. Elastoplastic parameters for four lithologies (shale, two limestones, and sandstone) were derived from single-and multi-stage pseudo-triaxial compression tests. A finite element model incorporating the Drucker-Prager failure criterion was developed and validated against laboratory data. The validated model was then used to generate a dataset of 2188 synthetic cases linking in situ stress conditions to breakout width and depth. Three transparent Group Method of Data Handling (GMDH) correlations were trained to estimate SH from breakout geometry and rock properties. The best-performing model achieved R2 = 0.98 and RMSE = 5.89 MPa. Independent validation using modified thick-wall cylinder experiments demonstrated strong agreement between measured, numerically simulated, and machine learning-predicted SH values. In contrast, elastic-based approaches produced significantly overestimated failure zones and unrealistic stress predictions. The proposed framework provides a computationally efficient, field-applicable tool for stress estimation that incorporates inelastic deformation. This approach improves the reliability of breakout-based stress analysis in both brittle and ductile formations and offers practical implementation potential for petroleum, geothermal, and subsurface engineering applications.
