Accurate identification of lithology is considered very important in oil and gas exploration because it has a direct impact on the evaluation and development planning of any reservoir. In complex reservoirs where extreme class imbalance occurs, as critical minority lithologies cover less than 5%, the identification accuracy is severely constrained. Recent deep learning methods include convolutional neural networks, recurrent architectures, and transformer-based models that have achieved substantial improvements over traditional machine learning approaches in identifying lithology. These methods demonstrate great performance in catching spatial patterns and sequential dependencies from well log data, and they show great recognition accuracy, up to 85–88%, in the case of a moderate imbalance scenario. However, when these methods are extended to complex reservoirs under extreme class imbalance, the following three major limitations have been identified: (1) single-scale architectures, such as CNNs or standard Transformers, cannot capture thin-layer details less than 0.5 m and regional geological trends larger than 2 m simultaneously; (2) generic imbalance handling techniques, including focal loss alone or basic SMOTE, prove to be insufficient for extreme ratios larger than 50:1; and (3) conventional Transformers lack depth-dependent attention mechanisms incorporating stratigraphic continuity principles. This paper is dedicated to proposing a geological-constrained multi-scale Transformer framework tailored for 1D well-log sequences under extreme imbalance larger than 50:1. The systematic approach addresses the extreme imbalance by deep-feature fusion and advanced class-rebalancing strategies. Accordingly, this framework integrates multi-scale convolutional feature extraction using 1 × 3, 1 × 5, 1 × 7 kernels, hierarchical attention mechanisms with depth-aware position encoding based on Walther’s Law to model long-range dependencies, and adaptive three-stage class-rebalancing through SMOTE–Tomek hybrid resampling, focal loss, and CReST self-training. The experimental validation based on 32,847 logging samples demonstrates significant improvements: overall accuracy reaches 90.3% with minority class F1 scores improving by 20–25% percentage points (argillaceous siltstone 73.5%, calcareous sandstone 68.2%, coal seams 65.8%), and G-mean of 0.804 confirming the balanced recognition. Of note, the framework maintains stable performance even when there is extreme class imbalance at a ratio of up to 100:1 with minority class F1 scores above 64%, representing a two-fold improvement over the state-of-the-art methods, where former Transformer-based approaches degrade below. This paper provides the fundamental technical development for the intelligent transformation of oil and gas exploration, with extensive application prospects.

This study develops a progressive damage failure criterion to address limitations of traditional instantaneous strength criteria that cannot capture damage evolution or quantify accumulation mechanisms in coalbed methane (CBM) wellbore collapse analysis. A damage variable D defines four evolutionary stages—initiation, propagation, acceleration, and coalescence—coupled with stress redistribution and drilling fluid invasion effects to enable quantitative collapse period prediction. Three-dimensional numerical simulations using FLAC3D 7.0 (Itasca Consulting Group, Minneapolis, MN, USA) reveal that stress anisotropy controls directional damage initiation, while increasing the horizontal stress ratio K substantially reduces the safe collapse period and narrows the safe drilling fluid density window. Horizontal wells exhibit significantly higher collapse pressure requirements than vertical wells, providing a quantitative basis for trajectory optimization. Model predictions show good agreement with published experimental results, with maximum deviations of ≤10% for the collapse period and ≤9% for damage depth. Field applications demonstrated notably fewer wellbore stability incidents and improved drilling efficiency compared to conventional design approaches, validating the practical effectiveness of the proposed methodology for unconventional resource development.

The stability of deep marine shale wellbores is influenced by both bedding anisotropy and drilling fluid intrusion. Existing models fail to adequately account for the coupled effects of intrusion depth and strength degradation. This study, targeting Longmaxi Formation shale, established a collapse pressure prediction model incorporating drilling fluid intrusion depth through direct shear tests and nuclear magnetic resonance (NMR) techniques. Experimental results indicate that shear strength reaches its minimum at β = 45°, decreasing by approximately 60% compared to β = 0° or 90°. Intrusion causes exponential decay in bedding plane strength, with the cohesion degradation coefficient λ_c = 0.158 mm⁻¹ significantly exceeding the internal friction angle degradation coefficient λ_φ = 0.089 mm⁻¹. Sensitivity analysis indicates that bedding angle and invasion depth rank third (±3%) and fourth (±1.5%), respectively, in influencing collapse pressure. Field validation confirmed excellent model prediction accuracy (R² = 0.956; RMSE = 0.55 MPa; MAPE = 1.05%), with all errors below 4%. This model accurately predicts the time-varying characteristics of collapse pressure, providing a theoretical basis for optimizing the design of drilling fluid density.

Deep coal-rock mining faces complex environments with high in-situ stress and strong heterogeneity, posing severe threats to drilling engineering safety. This study systematically investigated the influence mechanism of heterogeneity on damage evolution paths and its application in drilling risk assessment through multi-scale characterization, laboratory experiments, numerical simulation, and comprehensive evaluation. A quantitative evaluation methodology was established using CT scanning, acoustic testing, SEM, and XRD techniques, incorporating 6 primary and 15 secondary indicators to achieve heterogeneity classification ( $$ I_{H} $$ =0.24–0.78). The methodology was validated using 14 sampling points from three coalfields (Ordos Basin, Qinshui Basin, and Huainan Mining Area) spanning different coal ranks and tectonic settings. Through 210 triaxial compression tests with synchronous AE monitoring on 168 specimens from 14 sampling points, three mechanisms controlling damage evolution were revealed: stress concentration, crack propagation path alteration, and energy dissipation differentiation. Three typical damage evolution paths were identified: progressive type (characterized by gradual and continuous damage accumulation), step-wise type (featuring intermittent damage jumps at discrete stress levels), and burst type (exhibiting abrupt damage surge near peak stress) ( $$ \sigma _{{\text{int} }} /\sigma _{c} = 0.67 - 0.45I_{H} $$ , $$ R^{2} = 0.91 $$ , n = 42, p < 0.001). A numerical model considering heterogeneity was constructed with high accuracy (error < 5%), and the damage depth prediction equation was established ( $$ D_{d} = 0.17e^{{2.68I_{H} }} $$ , $$ R^{2} = 0.968 $$ , n = 14, p < 0.001), revealing damage depth increases from 0.32 m to 1.18 m (269% increase). A risk assessment framework based on damage evolution paths was developed with tiered management strategies. Field applications across 28 drilling operations in three mining areas achieved 63% reduction in wellbore instability, 45% decrease in drilling fluid loss, and 55% reduction in pipe sticking risk, providing scientific basis for deep coal seam drilling engineering.

Ultra-deep well drilling faces challenges of high temperature, high pressure, and narrow density windows, with rock thermal damage significantly affecting wellbore stability. Through high-temperature triaxial compression experiments on granite, sandstone, and shale, we revealed that within 100 °C–300 °C, elastic modulus decreased by 15%–55%, compressive strength decayed by 20%–65%, and cohesion degraded over 50%. A thermal damage constitutive model and time-dependent wellbore stability analysis model were established using thermo-hydro-mechanical coupling theory. Results show thermal stress reaches 30–50 MPa, and thermo-mechanical coupling increases tangential stress concentration by 27%–40%. Under typical conditions, density window width narrows from 0.26 g cm ⁻³ initially to 0.18 g cm ⁻³ after 24 h (31% reduction) and 0.12 g cm ⁻³ after 72 h (54% reduction). A staged dynamic density control optimization method was proposed. Field application in Tarim Basin (7850 m depth, 210 °C) shortened drilling cycle by 15%, reduced non-productive time by 54%, increased safety margin by 62%, and improved wellbore stability score from 65 to 88 points, validating the method’s effectiveness.