Understanding the spatial variations of in-situ stress is crucial for optimizing subsurface resource development and mitigating geological hazards. However, due to the spatial sparsity of conventional stress measurement methods (e.g., hydraulic fracturing and overcoring), critical stress variations might be overlooked. To enhance the continuity of stress profiling, we utilize borehole cross-sectional ellipticity resulting from stress-induced borehole deformation to obtain continuous and abundant stress information. We developed a workflow to extract borehole cross-sectional ellipticity from the acoustic televiewer (ATV) logs, which provide high-resolution measurements of borehole cross-sectional geometry. Based on a borehole array in the Bedretto Underground Laboratory in Switzerland, we characterized the stress variations within a hectometer-scale granitic rock mass using borehole cross-sectional ellipticity. We found that each borehole exhibited significant variations in cross-sectional ellipticity along its depth, indicating strong heterogeneity of the stress field. Subsequently, we employed a grid search algorithm to invert the continuous stress variations along each borehole based on its crosssectional ellipticity. The results indicate that the rock mass is generally in a normal faulting stress regime, but the stress orientations and relative stress magnitudes vary significantly. The causes of such stress variations could be related to local stress concentrations caused by fractures or stress perturbations resulting from fault slip. Our work provides a quantitative characterization of continuous stress variations within rock masses without direct stress measurements, which is beneficial to various geoscientific and subsurface engineering applications.

Borehole breakouts are stress-induced spalling on the borehole wall, which are identifiable in acoustic image logs as paired zones with near-symmetry azimuths, low acoustic amplitudes, and increased borehole radius. Accurate breakout characterization is crucial for in-situ stress analysis. In recent years, deep learning has been introduced to automate the time-consuming and labor-intensive breakout picking process. However, existing approaches often suffer from misclassification of non-breakout features, leading to high false positive rates. To address this limitation, this study develops a deep learning framework, termed Breakout-picker, with a specific focus on reducing false positives in automatic breakout characterization. Breakout-picker reduces false positives through two strategies. First, the training of Breakout-picker incorporates negative samples of non-breakout features, including natural fractures, keyseats, and logging artifacts. They share similar characteristics with breakouts, such as low acoustic amplitude or locally enlarged borehole radius. These negative training samples enables Breakout-picker to better discriminate true breakouts and similar non-breakout features. Second, candidate breakouts identified by Breakout-picker are further validated by azimuthal symmetry criteria, whereby detections that do not exhibit the near-symmetry characteristics of breakout azimuth are excluded. The performance of Breakout-picker is evaluated using three acoustic image log datasets from different regions. The results demonstrate that Breakout-picker outperforms other automatic methods with higher accuracy and substantially lower false positive rates. By reducing false positives, Breakout-picker enhances the reliability of automatic breakout characterization from acoustic image logs, which in turn benefits in-situ stress analysis based on borehole breakouts.

Fracture characterization from acoustic borehole image (ABI) logs is essential for subsurface geological investigation, as natural fractures (NFs) and drilling-induced tensile fractures (DITFs) provide critical constraints on rock properties and in-situ stress orientation. Deep-learning-based approaches have been proposed for automated fracture characterization, but they typically rely on task-specific architectures trained from scratch. Such models require substantial annotated data to perform reliably, yet obtaining expert-labeled borehole image samples is costly and time-consuming, leaving most practical datasets severely limited in size. In this study, we present Fracture-picker, a framework that adapts a state-of-the-art vision foundation model (VFM) DINOv3, pretrained on large-scale satellite imagery via self-supervised learning, to the task of fracture characterization from ABI logs. The framework integrates fracture-targeted image enhancement, fine-tuning of the DINOv3 backbone, and post-processing for automated extraction of fracture orientation with per-fracture confidence rating. We trained Fracture-picker on a dataset of only 80 annotated ABI image samples and validate it through blind well tests on unseen boreholes. Experimental results demonstrate that the adapted VFM outperforms task-specific models, including UNet, YOLOv8, and Swin Transformer trained from scratch under the same data-limited conditions, indicating that pretrained visual representations can effectively compensate for the scarcity of domain-specific annotations. Our findings establish that adapting pretrained VFMs to borehole image interpretation offers a superior alternative to training task-specific models from scratch when annotated data are limited

Profiling the in situ stress along deep boreholes is crucial for understanding crustal mechanics and facilitating subsurface exploration and developments. Existing approaches for borehole stress profiling often avoid natural fractures, because the stress field near fractures is challenging to measure and interpret. Here we present a novel approach to profile the in situ stress, primarily relying on the natural fractures intersected by deep boreholes. The critically-stressed natural fractures feature signatures identifiable on temperature log, while others do not. An abundant and diverse set of classified natural fractures is utilized for stress inversion. We illustrate this novel approach by re-visiting the KTB borehole data set. The natural fracture classification facilitated a two-stage stress inversion that efficiently profiles both the in situ stress magnitude and orientation. The inverted stress matches well with independent borehole observations that were costly to conduct, and provides additional information on the crustal strength variations. The inversion is further experimented on subsets of natural fractures along the KTB borehole in order to capture stress heterogeneity over smaller length scales. The limitation and scale-dependence of this novel approach is examined by integrating the fracture distribution and stress heterogeneity. Profiling in situ stress via natural fractures is feasible and complementary to existing approaches, and can offer new insights on the characteristics of crustal stress, its spatial heterogeneity, and its interactions with geological discontinuities.

My research focuses on crustal stress inversion and the analysis of stress evolution in earthquake sequences using focal mechanism data. To address the limitations of conventional approaches in nodal-plane selection, focal-mechanism uncertainty, and structurally complex regions, I developed the CDL framework to improve the robustness and physical consistency of stress inversion. CDL integrates mechanical plausibility, slip-direction consistency, and nodal-plane uncertainty into a unified framework, allowing spatiotemporal stress variations to be resolved without prescribing the true fault plane in advance. This approach is applicable to major earthquake sequences, regional seismicity, and complex plate-boundary environments, roviding a useful tool for investigating postseismic stress adjustment, fault interaction, and the stress conditions associated with earthquake generation.

The strength criterion is essential for understanding rock engineering under diverse geological conditions and stress paths. However, traditional rock strength criteria, such as the Mohr-Coulomb (M-C) and Hoek-Brown (H-B) criteria, do not fully capture the failure characteristics of rock under complex three-dimensional stress conditions and lack physical significance in describing microscopic deformation and fracture mechanisms. Inspired by the Wiebols-Cook (W-C) criterion, we propose a novel three-dimensional strength criterion that explicitly considers the Coulomb frictional slip of micro-fractures during loading and uses the accumulated macroscopic additional strain energy caused by that slip as the threshold for the rock criterion. The proposed criterion is based on an equivalent continuum model of a three-dimensional discrete fracture network, making it adaptable to both intact rock and fractured rock masses at various scales. By adjusting the parameters of the fracture network, we analyze the effects of intermediate principal stress, fracture scale effects, and heterogeneity in the friction coefficient on rock strength. The strength predictions based on the proposed criterion show a high degree of agreement with true triaxial experimental results, demonstrating their accuracy and applicability. The influence of the intermediate principal stress arises from the non-uniform strain energy induced by micro-fractures and Coulomb slip. Furthermore, the increase in fracture scale and friction coefficient heterogeneity leads to a decrease in rock strength, with the fracture scale effect having a more significant impact. This study establishes a clear connection between macroscopic strength characteristics and microscopic physical mechanisms, offering new insights into cross-scale mechanical properties from intact rock to fractured rock masses. Additionally, it provides a quantitative basis for assessing rock mass stability in deep-earth engineering exploration, development, and operations.

Jiang, S.^*, Ma, X.^† and Mukuhira, Y., Kilometer-scale crustal stress inverted through diverse natural fractures - synthetic tests and real borehole applications, Rock Mechanics Bulletin. [DOI]

Natural fractures in the Earth's crust provide valuable insights into the crustal stress. Field studies have shown that critically stressed fractures are typically hydraulically conductive, whereas non-critical fractures rarely exhibit measurable flow. To capture fracture criticality and associated uncertainty, this study adopts a probabilistic criticality model rather than a binary Coulomb criterion. We develop a nonlinear inversion method to estimate kilometer-scale crustal stress by integrates hydraulic conductivity characteristics and orientation data from natural fractures in deep boreholes. The inversion employs the Neighborhood Algorithm to optimize the likelihood between calculated critical probability and hydraulic conductivity, yielding estimates of stress orientation and relative magnitude. To evaluate performance, we constructed a crustal rock mass model and generated synthetic datasets with varied fracture number, the ratio of hydraulically conductive to non-conductive fracture, and orientation diversity. We further applied the method to four kilometer-scale scientific boreholes: Cajon Pass, Long Valley, NTS, and KTB. The inverted stress states are consistent with independent stress measurements, confirm the accuracy of proposed stress inversion method and the correlation between fracture criticality probability and hydraulic conductivity. Due to its minimal data requirements and straightforward implementation, the proposed stress inversion method is well-suitable for various geological environments.

The characteristics of the shallow in-situ stress field are of great significance to the construction, safe operation and maintenance of underground engineering. Among various in-situ stress measurement methods, the hydraulic fracturing method has become one of the most widely used engineering techniques due to its simple operation, stable test results, and no requirement for pre-acquired rock mechanical parameters. Taking two engineering boreholes in south-central Guangdong as the research objects, this paper conducts systematic analysis based on hydraulic fracturing in-situ stress test data and ultrasonic imaging logging data, and accurately obtains the magnitude and spatial orientation of in-situ stress at the two borehole sites. On this basis, the stability of fractures around the boreholes is evaluated. The results show that the fractures in one borehole present a high risk of slip instability. It is necessary to emphatically monitor the dynamic variation of in-situ stress in this area during engineering construction, so as to ensure the overall stability and safety of underground engineering structures.

The borehole cross-section ellipticity analysis method can efficiently obtain continuous in-situ stress information and compensate for the lack of stress data in non-hydraulic fracturing intervals. In this paper, the QSZK borehole near the Shandong segment of the Tan-Lu Fault Zone is taken as the research object. Based on five sets of repeated ultrasonic borehole imaging data from this borehole, the cross-section ellipticity analysis method is adopted to interpret continuous ellipticity parameters such as the orientation of the minor axis and the axial ratio. The results show that the orientation of the minor axis of the borehole cross-section is consistent with the strike of hydraulic fractures within the studied interval. The overall dominant orientation of the minor axis is concentrated in the nearly EW direction, which is basically consistent with the measured in-situ stress direction of the borehole and the direction of the regional tectonic stress field. This study verifies the applicability of the cross-section ellipticity analysis method in in-situ stress research, which can provide an effective auxiliary means for borehole-scale in-situ stress field analysis.

The Changning shale gas development area, located in the southern Sichuan Basin, is one of the earliest shale gas fields in China to enter commercial production. Complex geological structures and long-term fluid injection have led to frequent seismicity and a highly heterogeneous stress field. In this study, using data from 17 seismic stations successively deployed by the University of Science and Technology of China between January 2019 and June 2022, we jointly utilize S/P amplitude ratios and Z‑component waveform information to invert for a large number of focal mechanism solutions spanning the three‑and‑a‑half‑year observation period, substantially expanding the focal mechanism catalog for the Changning area. After applying quality control to these focal mechanisms, a sliding‑window stress inversion in both space and time is performed on them to resolve the local spatiotemporal variations in the orientations of the three principal stresses and the shape ratio ( $\phi = \frac{S_2 - S_3}{S_1 - S_3}$ ) against the background of an overall NWW‑SEE trending tectonic stress field in the Changning area.

High-quality acoustic televiewer (ATV) images are essential for continuous borehole-wall interpretation, including fracture identification, structural analysis, and subsequent automated processing. However, vertical groove artifacts often introduce local brightness and pseudocolor distortions in ATV images, producing stripe-like anomalies that may obscure geological textures and affect both visual interpretation and machine-based analysis. Unlike ordinary image inpainting problems, the groove region is not a completely missing area, but a polluted observation that still preserves useful borehole-wall information. To address this problem, we formulate groove suppression as mask-constrained residual restoration. The proposed framework takes the artifact image, a masked-context image, and the groove mask as a seven-channel input, and predicts an RGB correction residual that is added only inside the detected groove region. This design suppresses the groove artifact while structurally preserving pixels outside the mask and retaining as much observed geological texture as possible. Because real ATV groove images usually lack clean references, we further develop a palette-guided synthetic groove generation strategy based on palette-coordinate compression to construct paired training data. Mask-focused luminance and correction losses are used to encourage sufficient restoration of the narrow groove band. Experiments on synthetic paired ATV data show that the method substantially reduces mask-inside RGB and luminance errors, while comparisons with ordinary LaMa and traditional interpolation/inpainting baselines demonstrate the advantage of treating the groove as a polluted observation rather than a missing hole. For real ATV images without ground truth, proxy metrics and visual inspection are jointly used to assess groove suppression, mask-confined correction, and possible residual artifacts. This study provides a controllable image-restoration workflow that improves the reliability of ATV image interpretation and supports downstream automated analysis of borehole-wall structures.

This research focuses on the core scientific and engineering demands for the efficient development of unconventional oil and gas reservoirs, concentrating on two core directions: precise characterization of the stress field and regulation of the expansion laws of hydraulic fracturing fractures. It conducts systematic research by reviewing the theoretical systems and applicable boundaries of mainstream stress calculation models such as isotropic (ISO), transversely isotropic (VTI), and viscoplastic stress relaxation (VSR) models. It quantitatively analyzes the control mechanisms of key factors such as pore pressure, effective stress coefficient, anisotropy of elastic parameters, and creep parameters on the prediction results of stress. It clarifies the model adaptation criteria under different rock types and geological backgrounds: in organic-rich clay-rich shale reservoirs, the VSR model can more accurately reproduce the real stress state formed by long-term geological evolution, while the traditional linear elastic model still has stable applicability and engineering application value in other rock types reservoirs. Based on this, by combining the measured data from typical oil and gas blocks at home and abroad and the numerical simulation of hydraulic fracturing, the study clarifies the core controlling effects of inter-layer stress differences, reservoir thickness, and stress state of the formation on the expansion behavior of hydraulic fractures, and establishes a correlation method for precise stress characterization and optimized design of fracturing. The research results can provide systematic theoretical support and technical guidance for engineering applications such as optimizing fracturing schemes, well pattern deployment, risk prevention and control (induced earthquakes), and underground gas storage of unconventional oil and gas reservoirs.

The stress relaxation process of rocks is critical for controlling the time-dependent stability of underground engineering. However, existing stress relaxation models largely rely on macroscopic phenomenological laws, making it difficult to explicitly couple the statistical characteristics of microscopic fractures with macroscopic stress relaxation. We established a stress relaxation model based on Discrete Fracture Network (DFN) statistical characterization and macro-microscopic iteration. Using a homogenization method, the model achieves the cross-scale projection of fracture frictional slip response. Under constant horizontal stress and constant vertical strain boundary conditions, the model implements a self-adjustment process involving microscopic critical fracture identification, macroscopic stress relaxation, and damage accumulation through discrete iterative time steps. We specifically demonstrate the effects of different friction coefficients and principal stresses on the degree of stress relaxation, relaxation modulus, and damage evolution characteristics of the model. Furthermore, we reveal the fundamental mechanism by which the intermediate principal stress indirectly regulates the stress relaxation rate by adjusting the critical fracture ratio. Additionally, the model's stress relaxation results are highly consistent with experimental results from stress relaxation triaxial tests. This model quantitatively confirms and extends the concept that macroscopic stress relaxation is driven by internal stress drop release within rock masses, and provides a theoretical basis for large-scale underground rock mass stress evolution.

Continuous well logs provide high-resolution vertical information on lithology, mineral composition, pore structure, and elastic properties, but they do not directly measure in-situ stress. DFIT-derived stress measurements provide more direct stress constraints, yet they are sparse and subject to interpretation uncertainty. To address this mismatch between continuous indirect log information and sparse direct stress constraints, a physics-informed probabilistic workflow is developed for reconstructing continuous minimum horizontal stress, $ S_{h\mathrm{min}} $ , profiles. The workflow integrates well-log sequence representation learning with posterior updating from sparse DFIT-derived stress constraints, while incorporating geomechanical bounds associated with the frictional lower limit, the vertical-stress upper bound, and stress-relaxation concepts. By learning lithologic and stratigraphic controls on $ S_{h\mathrm{min}} $ and explicitly accounting for stress-measurement uncertainty, the proposed framework enables small-sample, multivariate, interpretable, and uncertainty-aware prediction of continuous $ S_{h\mathrm{min}} $ profiles.