Accurate identification and quantitative characterization of fractures in hot dry rock (HDR) reservoirs are of great significance for the efficient development of geothermal resources. To address the challenges associated with precise fracture recognition and quantification, an improved fracture identification model named TSD-Unet is proposed in this study. This model is constructed by integrating the Spatial Group-wise Enhance (SGE) module and Dynamic Snake Convolution (DSC) module into the U-Net architecture. This integration enables the TSD-Unet model to extract both spatial and multi-morphological features of fractures from granite failure images. The SGE and DSC modules, inserted after the convolutional layers, allow the model to effectively combine spatial and morphological features of fractures. Ablation experiments and multi-model comparison experiments were conducted using a granite fracture image dataset. The comparison results demonstrate the competitiveness of the TSD-Unet model in segmentation performance, achieving an accuracy (Acc) of 62.88 % and an intersection over union (IoU) of 46.06 %. Compared to the traditional U-Net model, TSD-Unet shows improvements of 8.02 % in Acc and 9.51 % in IoU. Based on the segmentation results and the proposed feature computation method, quantitative analyses were performed on fracture characteristics such as length, area, average width, and maximum width, revealing that the results based on TSD-Unet closely match actual conditions. This research provides a precise and efficient method for intelligent fracture identification and feature extraction in HDR reservoirs, offering significant theoretical guidance for improving the efficiency of geothermal resource exploitation.

To address the insufficient understanding of coupled gas, dust, and airflow transport in coal mine excavation roadways and the lack of quantitative criteria for regulating ventilation parameters, a gas-dust-airflow coupling analysis framework integrating mechanism analysis with response characterization was established. A theoretical gas-dust-airflow coupling model was first developed, and CFD simulations and mechanism analyses were performed within a unified parameter space designed by response surface methodology (RSM). The CFD results elucidate an airflow-dominated asymmetric coupling pattern, with dust and gas contributing via weak feedback and indirect modulation, respectively. A reduction of approximately 55% in airflow velocity resulted in a 60-65% decrease in turbulent diffusivity, causing maximum dust and gas concentrations to increase by 149.16%-179.07% and 144.64%-152.28%, respectively. Regression models for the maximum gas concentration at the heading corner and the maximum dust concentrations at the driver position and pedestrian breathing height were established using RSM, enabling the prediction of coupling characteristics. The response surface results reveal spatial heterogeneity: the corner gas concentration exhibited a non-coupled nonlinear response dominated by independent effects of gas source intensity and airflow velocity; the driver position dust concentration is governed primarily by airflow-dust interaction; whereas the dust concentration at breathing height is controlled by the combined gas-dust-airflow coupling. Finally, the minimum required ventilation velocity was determined using a grid optimization algorithm, and a ventilation regulation map was constructed. This work clarifies the gas-dust-airflow coupling mechanism from theory, transport, and characteristics perspectives, providing support for rapid parameter determination and intelligent ventilation control under coupled conditions.

Deep shale reservoirs exist in complex thermo-pressure environments, making it difficult to quantitatively characterize their microscopic mechanical behaviors and anisotropic evolution patterns. In this study, molecular dynamics simulations combined with Perl scripts are employed to quantify the tensile mechanical responses of single shale minerals, kerogen and multi-component systems under thermo-pressure conditions, and reveal the underlying microscopic mechanisms. The results show that the mechanical stability of shale ranks as brittle minerals > clay minerals > organic matter. The degree of thermally induced strength degradation follows kaolinite > kerogen > montmorillonite > brittle minerals; kaolinite loses up to 62% strength along the [010] crystal direction, and quartz only loses 2% along the [100] crystal direction. The influence degree of pressure is ranked as clay minerals, organic matter and brittle minerals. The strength of kaolinite rises by up to 90.6% along the [100] crystal direction under pressure, with quartz merely increasing by 0.97% along the [010] crystal direction. In composite models, rigid quartz and kaolinite restrict the plastic flow of kerogen via interfacial confinement. Under thermo-pressure coupling, kerogen enhances thermal softening along the [001] crystal direction without affecting rigid bond-dominated [100]/[010] regions. Quartz enhances directional stability through pressure hardening, and kaolinite induces interlayer anisotropy. The kerogen-quartz-kaolinite composite system undergoes a critical shift from the balance between pressure hardening and thermal softening to thermal softening predominance along the [100] crystal direction. Deformation along the [001] crystal direction is always controlled by thermal softening, while those along the [100]/[010] crystal directions are dominated by pressure hardening. This study is the first to quantitatively characterize the anisotropic evolution of tensile mechanical properties from single minerals, kerogen to multi-component composite systems under temperature-pressure coupling, and reveals the microscopic mechanism of rigid phase interfacial constraint and component synergistic effect.

Excessive dust and gas concentrations are critical risk factors in mining disasters, making the optimization of ventilation systems essential. Frequent threshold exceedances of dust, gas, or both in heading faces necessitate elucidation of the impact of airflow on their transport behavior to improve ventilation strategies. However, conventional experimental approaches, such as scaled models and field tests, fail to capture the complexity of real-world conditions due to limitations, including independent analyses of dust and gas, similarity errors, spatial constraints, and harsh field conditions. To address these limitations, a full-scale variable-frequency ventilation experimental system was developed to simultaneously monitor dust and gas transport in heading faces. The system comprises a variable-frequency ventilation control system, dust generation system, gas supply system, and data acquisition system. Validation through experiments on variable-frequency ventilation control, gas transport, and dust dispersion demonstrated high reliability, with R2 values ranging from 0.64 to above 0.98 (average R2 approximate to 0.80). On this basis, full-scale experiments on coupled airflow-gas-dust transport were conducted. The results indicate that dust modulates the turbulence structure, thereby affecting gas diffusion; gas influences both airflow and dust through its physical properties and transport behavior; and airflow variations influence the transport and concentration distribution of both phases. These interactions establish a dynamic equilibrium. This system overcomes traditional limitations, providing high-fidelity data and theoretical insights for advancing coupled dust-gas ventilation control, intelligent ventilation strategies, and energy-efficient operations in heading faces.

Open-pit mining in the high-altitude areas of the western Sichuan Plateau triggers fluctuations in groundwater levels, causing rocks to undergo freeze–thaw cycles at different immersion depths and posing threats to the stability of slopes and the integrity of groundwater systems. In this study, nuclear magnetic resonance (NMR) technology, scanning electron microscopy (SEM), and permeability tests were adopted to systematically investigate the influence of freeze–thaw cycles on the pore structure and permeability of slate under different immersion depths. Additionally, the distinguishing characteristics of freeze–thaw damage under various immersion states was revealed. By integrating the fractal dimension, which characterizes the complexity of pore structure, into the modeling process, a permeability model incorporating porosity, fractal dimension, number of freeze–thaw cycles, and immersion depth was established. The results showed that freeze–thaw cycles significantly influenced the pore structure and permeability of rock samples. Specifically, as the number of freeze–thaw cycles increased, both the porosity and permeability exhibited an upward trend, with these trends becoming more pronounced as immersion depth raised. Notably, under conditions of moderate to full immersion, the rock was more prone to developing internal microcracks. Furthermore, a fractal dimension-adjusted permeability model was proposed. The established model integrated four key influencing factors, namely initial porosity, fractal dimension, number of freeze–thaw cycles, and immersion depth, achieving a remarkable coefficient of determination (R2) of 0.925. The established model allowed for the estimation of permeability changes in rock subjected to varying immersion depths and different numbers of freeze–thaw cycles, serving as a reference for mitigating rock degradation and protecting groundwater resources.

The evolutionary process of rock deformation and instability progresses through distinct stages: closure, initiation, development, coalescence, rupture, and slip of internal cracks. Accurate identification of these stages and their precursory information is crucial for analyzing rock failure mechanisms and defining critical engineering parameters. This study employed Digital Image Correlation (DIC) and Acoustic Emission (AE) techniques to record the evolution of the apparent deformation field and internal microfracturing processes, respectively, during uniaxial compression tests on fractured rock specimens. A comprehensive dataset characterizing deformation and instability was established using six quantified feature parameters derived from these techniques (AE1, AE2, yxy1, yxy2, yxy3, yxy4; 5,088 samples). Parameters exhibiting low correlation were identified via Pearson correlation matrices. Unsupervised clustering using the K-means++ algorithm was then applied to generate state labels, and the correspondence between these labeled clusters and the physical stages of deformation and instability was investigated. Subsequently, an XGBoost machine learning model was trained using the cluster-derived labels to construct a state identification model for fractured rock mass. The model’s identification capability was evaluated under different feature parameter combinations. Key results demonstrate: Clustering analysis revealed three distinct clusters corresponding to the stable state (encompassing the compaction stage, elastic deformation stage, and stable crack expansion stage), unstable state (the unstable crack expansion stage), and fracture-instability state (the post-peak damage stage) under uniaxial compression. The identification model trained on the multi-parameter feature combination achieved high test accuracies: 99.87% (stable state), 97.98% (unstable state), and 98.84% (fracture-instability state), with overall accuracy (99.49%), precision (97.17%), and recall (98.90%). Compared to models trained on single-parameter or single-type feature parameter combinations, the model exhibited superior and stable identification performance, minimal variance in parameter contribution rates, and strong generalization capability. These findings validate the feasibility of using machine learning to identify the rock deformation-instability evolution process and provide a novel analytical framework for detecting precursory characteristics and deciphering the evolutionary pathway of rock instability.

In engineering practice, structures such as bridges, tunnels, and embankments often undergo cyclic loading during their service life. Such loading typically has the characteristics of periodicity and intermittency, and different types of loading can cause different degrees of damage, affecting the safety and stability of engineering structures. Rock and concrete, as the main raw materials in structural engineering, often require a large number of laboratory tests to accurately determine their mechanical properties. The repeatability and ease of operation of the sample preparation process need to be given special consideration, and the emergence of 3D printing technology can effectively solve these problems. Based on this, 3D-printed cement-based rock-like materials were prepared using 3D printing equipment, and a continuous and intermittent cyclic loading and unloading test scheme was designed. The test samples were tested using a pressure testing machine, a piezoelectric acoustic emission system, and a digital image correlation (DIC) device to analyze the evolution of damage in the 3D-printed samples under different types of loading. The study shows that the compressive strength of the samples increases first and then weakens as the cyclic upper limit stress increases, and the trend is further exacerbated by intermittent cyclic loading compared to continuous cyclic loading. With 60 % of the compressive strength as the dividing line for damage degree, the Kaiser effect is significant under low stress cycling, while the acoustic emission quiet period in high stress cycling is broken, and the Filicity effect gradually becomes obvious. At the same time, the trend of acoustic emission phenomena with the increasing upper limit stress of the cycling also shows an upward trend. Based on the acoustic emission AFRA value, a new k value can be quantitatively obtained by quasi-definition, which can provide the proportion of tensile and shear cracks in each stage of the sample under different loading forms during loading. In addition, the vertical displacement trend of the sample under loading can be quantitatively obtained by the virtual extensometer in DIC. The research results can provide theoretical support for the safety and stability of structural engineering.

Local intelligent ventilation based on proportional-integral-derivative (PID) control, which is regulated considering global gas concentration, is an effective way to accurately control gas concentration. However, the traditional monitoring system, as the source of PID process data, has a sparse distribution of monitoring points, making it difficult to capture the overall gas concentration across the working face. To obtain the global gas concentration and improve the regulation precision, a new PID automatic control method complemented by computational fluid dynamics (CFD) for gas concentration in the tunneling face is proposed. Firstly, the traditional monitoring system is complemented by CFD simulation as the source of process variable data for the PID controller model. Secondly, the PID controller model utilizes the maximum gas concentration from the CFD simulation to regulate the air velocity. The gas concentration field after the ventilation control is calculated through the CFD simulation. Two steps are repeated for all time steps in the time loop until the end of simulation time is reached. The gas source term setting, maximum gas concentration monitoring, and PID controller model are programmed by user defined function (UDF) program and embedded with the CFD model. The research tackles PID process variable data input errors caused by sparse monitoring points, enabling real-time analysis and precise control of local intelligent ventilation simulation systems. Additionally, the ventilation power and facility control schemes are proposed. The CFD-based method offers a new way for testing, developing, and optimizing local intelligent ventilation systems before construction, with potential applications in tunneling faces.

Shale oil and gas-important unconventional hydrocarbon resources-exhibit huge exploration potential. The tensile mechanical properties of shale are crucial for enhancing oil and gas recovery. Herein, molecular dynamics simulations were performed to investigate shale tensile properties. The influence of the mineral composition and pore structure of shale on its tensile mechanical properties and underlying microscopic mechanisms were studied. Results revealed that compared to single-mineral models, composite models generally exhibited considerably lowered tensile strengths (reductions of 47.1 %-98.8 %). However, in the 001 direction, the kerogen-montmorillonite model showed 10.1 % higher tensile strength than montmorillonite. This composite model also showed higher tensile strength than kerogen, exhibiting increases of 157.3 %-1859.6 % in all crystal directions. However, the kerogen-calcite model exhibited 31 % lower tensile strength in the 001 direction than kerogen. An increase in the pore size from 2 to 10 nm resulted in Young's modulus reductions for all minerals by 1.5 %-99.3 % in all crystal directions. Kaolinite exhibited the smallest reduction in Young's modulus in the 100 Crystal directions (1.5 %) and the largest in the 001 Crystal direction (99.3 %). The geometrical control mechanism of pore morphology on damage resistance is decoded, the resistance of different pore shapes to damage varies, demonstrating orientation-dependent failure patterns where slit-shaped pores provide maximum stability in 100 and 010 Crystal directions while triangular configurations dominate in 001 Crystal direction. These findings establish structure-property relationships at the atomic scale that advance fundamental understanding of shale's failure mechanics, while providing engineering-relevant guidelines for predicting hydraulic fracture propagation and optimizing reservoir stimulation strategies.

The existing mine safety evaluation methods adopt relatively simple evaluation indicators and cannot quantitatively analyze the uncertainty degree of various indicators, resulting in significant limitations in practical engineering applications. To address these problems, a comprehensive mine safety evaluation method based on entropy weight and set pair analysis method is proposed. First, a comprehensive mine safety evaluation index system consisting of 19 indicators was constructed based on four factors: human, machine, environment, and management. Then, the entropy weight method was used to automatically determine the weights according to the degree of dispersion of the safety evaluation data, which could effectively avoid subjective interference while ensuring both simplicity of calculation and objectivity and accuracy. Finally, expert scoring was used to obtain the scores of each safety evaluation indicator, and set pair analysis method was applied to calculate the degree of connection between the indicator scores and different risk levels. The weighted average method was further adopted to obtain the average connection degree between each evaluation object and different risk levels. Based on the maximum membership degree theory, the final mine safety risk level was determined. The application results showed that the calculated risk level of Xiaobaodang Mine using this method was low, which was consistent with the actual situation. The main unsafe factors of the mine were high worker fatigue rate during actual operations, low compliance rate of technical standards, slow equipment updates, and incomplete emergency management measures and untimely safety inspections. The method can provide a reference for improving mine safety in the future.

Understanding the dynamic mechanical behavior and particle breakage of granular materials is essential for revealing failure mechanisms in geotechnical systems. To minimize the influence of particle shape and assess the dynamic mechanical properties, energy absorption, and gradation evolution of granular materials, a 37-mmdiameter split Hopkinson pressure bar was employed to conduct uniaxial impact tests at strain rates of 103 s- 1 on three narrow-graded groups of smooth, spherical quartz glass beads. The influence of strain rate and particle size were analyzed. The main findings are as follows: The quartz glass beads exhibited a trend of strain rate sensitivity and particle size dependency: pre-consolidation pressure increased with strain rate but decreased with increasing particle size. The dynamic compression curves of medium and coarse sand displayed a plateau stage, whereas fine sand exhibited an exponential hardening trend. A comprehensive analysis of the damage variable, and energy absorption ideality revealed that a higher strain rate enhanced energy dissipation, particle breakage and delayed the energy-absorption stage. The particle size distributions of fragmented medium and coarse sand followed a fractal distribution, while fine sand progressively approached the fractal distribution with strain rate rose. A modified two-parameter Weibull distribution effectively characterized particle breakage behavior verified by SEM images.

The fracturing mechanisms and failure characteristics of geothermal reservoir rocks subjected to thermal shock induced by liquid nitrogen are of critical importance for enhancing the efficiency of geothermal energy extraction. In this study, granite specimens were heat-treated at temperatures ranging from 25 degrees C to 850 degrees C, followed by rapid cooling using liquid nitrogen (LN2). A combination of uniaxial compression testing, acoustic emission (AE), digital image correlation (DIC), and scanning electron microscopy (SEM) was employed to systematically investigate the evolution of damage and energy transformation under varying thermal shock conditions. Furthermore, a novel damage criterion based on energy storage efficiency was proposed to assess and predict the post-shock instability and failure behaviour of the rock. The results demonstrate that uniaxial compressive strength declines significantly with increasing heat-treatment temperature, with the most substantial degradation occurring between 550 degrees C and 850 degrees C. The stress-strain response evolves from a single peak to multiple peaks and subsequently returns to a single peak as temperature increases. Abrupt rises in cumulative AE counts, alongside highly localised strain concentrations observed through DIC, effectively capture the initiation and propagation of cracks. An integrated analysis of AE parameters, including energy, amplitude, and event localisation, reveals the spatiotemporal evolution of microcracks within the specimens. Energy analysis indicates that both the total input energy and elastic strain energy at peak stress decrease monotonically with rising temperature, while the dissipated energy exhibits a non-linear trend, initially increasing and then decreasing. The proposed criterion based on energy storage efficiency proves effective in predicting the onset of instability in thermally shocked granite, providing a robust theoretical foundation for assessing the stability of geothermal reservoirs and guiding practical engineering applications.

An innovative energy-absorbing and bearing structure was proposed, which incorporated the coupling of glass microspheres with a metal tube. Glass microsphere-filled steel tube (GMFST) column, consisting of external steel tube and inner glass microspheres, was expected to give full play to the energy-absorbing and load-bearing capacities of the particle while restricting particle flow from collapsing, thereby enhancing the overall structural strength. Four groups of steel tubes and the GMFST specimens were designed and subjected to axial compression tests at four different loading rates to investigate the performance of the structure. These tests aimed to analyze the deformation mode, mechanical response, and energy absorption capacity of the GMFST columns under quasi-static to low-speed compression conditions. The results indicated that the deformation process and failure mode of GMFST columns were similar to those of hollow steel tubes, albeit with a different post-buckling mode. Filling the steel tubes with glass microspheres reduced the load fluctuation range, moderated load–displacement curves, and exhibited a strain rate strengthening effect. The GMFST columns demonstrated superior energy absorption capacity, with significant increases in crush force efficiency, the averaged crush force, and the total absorbed energy, particularly in terms of subsequent support capacity. The load-increasing reinforcement properties enabled GMFST columns to overcome the limitations associated with the unstable post-buckling path of energy‑absorbing damping structure, exhibiting outstanding load-bearing performance and stability in the later stages. The results provided valuable guidelines for designing and engineering high-performance GMFST columns, serving as a new type of energy-absorbing and supporting structure.

Rock brittleness and brittle failure play significant roles in deep underground space stability, unconventional gas extraction drilling, hydraulic fracturing efficiency, and coal mining disaster prevention, and these rock engineering processes involve cyclic loading conditions. However, nearly all the existing rock brittleness indices (BIs) are conducted with monotonic loading, and the variation in brittleness with cyclic loading has not been comprehensively studied. This study rearranged energy evolution and conversion based on stress–strain curves to investigate the existing energy-based rock BIs. Monotonic and cyclic triaxial compression experiments were conducted on red sandstone specimens to investigate the differences in the mechanical properties. This paper proposes a BI based on post-peak energy conversion to evaluate rock brittleness under monotonic and cyclic loadings. This index is defined as the ratio of the extra inputted/released energy to the fracture energy. The results showed that the variations in mechanical parameters in the two types of experiments were consistent at the pre-peak stage and different at the post-peak stage. The value of the proposed BI revealed that rock brittleness under cyclic loading was lower, which conformed with the qualitative analysis of stress–strain curves and macroscopic failure plane. The proposed BI strongly correlated with the mechanical parameters under monotonic and cyclic loading. The applicability verification, compared with the existing energy-based BIs for different rock types, confining pressures, and inclination angles of bedding planes, showed that the novel BI exhibited an excellent linear correlation with the strength parameters and had broad applicability.

Rock is a widely used engineering material, and accurate understanding of its internal microcrack evolution process during loading can provide a theoretical basis for preventing instability and failure in rock engineering. The rock pressure test system and acoustic emission equipment were used to carry out uniaxial loading and acoustic emission monitoring tests on fractured red sandstone. Based on the RA and AF values of acoustic emission and the fractal theory, the internal crack patterns and evolution rules of red sandstone with different crack angles were explored. The results show that the compressive strength of red sandstone varies with the crack Angle in the shape of “U”, and there is an obvious stress drop after the elastic stage of 30° and 45° crack red sandstone, which has a good correspondence with the acoustic emission ringing count rate. During the loading process, the damage mainly caused by tension cracks first appears inside the rock, and then the tension cracks increase steadily and the shear cracks gradually increase, indicating that the rock enters the stage of fracture instability development. With the development of shear cracks, the peak strength is finally reached and the instability failure occurs. During the loading process of 30° fracture red sandstone, the corresponding stress drop phenomenon of “shear crack group” appears. According to the D/S statistical analysis of different crack samples based on acoustic emission ringing count, the fractal dimension presents a decreasing-rising-decreasing trend with the increase of crack inclination, which corresponds to the change of the complexity of crack development in rock.

Abstract Rock engineering achieves the secondary stress balance through rock mass structure adjustment, where energy conversion is throughout and associated closely with rock deformation and damage. In this study, a series of triaxial compression tests were conducted on red sandstone to investigate these features. The results showed that the damage state of red sandstone specimens presented five stages under different confining pressure, corresponding to the multistage evolution characteristic of the energy conversion. In the case of the dissipation energy conversion ratio (η), it showed five stages: a gradual increase, decreasing gradually and reaching a minimum value, increasing gradually, increasing with growth rate, and accelerated growth, therein the strong nonlinearity reflected the stability and instability of the internal structure of the rock and had the basic characteristics of the mutation theory, therefore the damage state warning model was established on just that. The relation between the η and time fitted by a four‐rank potential function had a fitting parameter (R2) larger than 0.9, and the bifurcation set of the η calculated by the damage state warning model had twice stages less than 0. The second stage, which occurred near the minimum value of the η and run through the plastic deformation stage, could be used to predict rock damage and fracture, and it was proven feasible by acoustic emission (AE) precursor and better than AE warning. This research can enrich the methods for identifying rock damage state and provide reference for revealing the occurrence and development mechanism of various rock instability disasters.

The nonlinearity of the constitutive relation for rocks becomes more prominent with a more complex physical-mechanical environment and mechanical behavior. The accurate establishment of the constitutive relation affects the determination of rock deformation and damage state from physical features. In this study, a novel statistical damage constitutive model for rocks is proposed based on quantified energy conversion. The novelty of the model is that the nature of rock damage before and after damage stress is considered. In the constitutive model, the evolution characteristics of energy conversion show a five-stage evolution with a ‘spoon’ form and correspond to the rock deformation and damage process, which can be fitted with the modified GaussAmp function; the damage variable is deduced by the Weibull distribution with energy conversion as the distribution variable, which presents a monotonic decrease caused by initial defects before the σ cd and shows a ‘S’ shape caused by nascent cracks after the σ cd . Furthermore, triaxial test data of three types of rocks under different confining pressures were used to verify the proposed model, and the results were in good agreement with the test data in most cases. The characteristics of the crack closure stage, peak stress, residual strength, and stress drop process are controlled by the model parameters, which can be determined using experimental data. As these parameters definitely have a physical meaning and a relation to the confining pressure, the proposed model has the potential to be used in rock engineering.

为研究人工土石混合边坡中块石块度及空间分布对边坡稳定性的影响,用Python语言编写脚本生成不同块石块度、块石间距、块石轮廓等参数构成土石混合边坡模型,实现土石混合边坡块石块度及分布的随机生成.将模型导入Abaqus软件中生成土石混合边坡的数值计算模型,对土石混合边坡的稳定性进行计算,得到了块石的不同块度大小及不同空间分布对土石混合边坡稳定性的影响规律.研究表明:在含石量相同、块石块度范围为 10～50 cm的条件下,随着块石块度增加,土石混合边坡的安全系数提升,塑性破坏带变宽、变长,边坡稳定性增加;当块度为10～20 cm与 30～40 cm的块石分层分布在土石边坡中且块度大的块石邻近边坡下部时,安全系数较高,土石混合边坡的塑性破坏带分布面积广,边坡稳定性提升.研究结果可为人工土石混合边坡施工提供参考.

In order to obtain the real material parameters of heterogeneous rock, the material parameters of red sandstone specimens under uniaxial compression tests are inverted based on the Digital Image Correlation (DIC) method and the Finite Element Model Updating (FEMU) method. The DIC method is employed to calculate the displacement field of red sandstone specimens during uniaxial compression loading. Concurrently, a uniaxial compression elastic–plastic finite element numerical model with non-uniform material parameters is developed based on the FEMU method. The model adopts the Mohr–Coulomb yield criterion and adjusts the boundary conditions in real-time to maintain consistency with the test. The vertical displacement field of the numerical model is juxtaposed with that of the test to construct the objective function. Optimization is achieved using the Artificial Fish Swarm algorithm, which enables the acquisition of the non-uniform distribution and evolution process of the material parameters of specimens at different loading moments. The results indicate that this method can spatially obtain the non-uniform distribution field of material parameters and temporally track the evolution of material parameters during the loading process. This research lays a solid foundation for enhancing the accuracy of intelligent coal mining and dynamic disaster monitoring and early warning in coal mines.

为探究裂隙岩体表观变形场非均匀变形特征并判识其损伤不稳定发展状态,预制含不同角度裂隙的砂岩试件并开展单轴压缩试验,联合AE和DIC方法,获取试件应力-应变曲线和表观变形场演化过程,并训练基于AdaBoost、RF和LightGBM算法的状态判识模型.结果表明:各裂隙砂岩试件应力-应变曲线均经历压密阶段、弹性变形阶段、新生裂纹稳定发展阶段、新生裂纹不稳定发展阶段和峰后破裂阶段,其表观剪应变场γxy呈非均匀演化特征,促使试件表观应变分区异化,表现出数值差异和空间汇聚两个特征;γxy应变场非均匀变形空间指标和程度指标均呈现三阶段演化特征,并在第一次演化阶段转变中蕴含着损伤破裂的开始.四特征参数组合训练的判识模型具有较高的泛化能力和较强的判识能力,其次是双特征参数组合,单一特征参数训练的判识模型效果最差;多特征参数组合下训练的LightGBM模型中各单一特征参数的贡献率更为平均,即LightGBM模型对裂隙砂岩状态的判识能力最强、鲁棒性好.