Potassium feldspar is a crucial rubidium-bearing mineral within granite-hosted rubidium deposits in China. However, the precise magmatic-hydrothermal stages and mechanisms responsible for its rubidium enrichment are not well understood. The Huashi rubidium deposit, located in eastern Hebei and characterized by distinct lithological zoning within its ore-hosting granite, serves as an exemplary natural laboratory for investigating this issue. This study examines the Madi pluton, which is the host intrusion of the Huashi deposit, by integrating field geology and petrography with in situ major- and trace-element analyses of two genetically distinct types of potassium feldspar, magmatic(Kfs1) and hydrothermal(Kfs2), using EPMA, LA-ICP-MS, and TIMA. By combining these analyses with mineral paragenesis, we aim to elucidate the stages and mechanisms of rubidium enrichment. The Madi pluton exhibits a distinct lithological zonation, consisting of pale-red, medium- to fine-grained granite, gray-white porphyritic alkali feldspar granite, and amazonite-bearing alkali feldspar granite. Significant rubidium mineralization is predominantly found within the latter two lithological units. The crystallization of magmatic K-feldspar(Kfs1) occurred at elevated temperatures(approximately 711 ℃) under conditions of minimal fractionation, resulting in a relatively low rubidium(Rb) content. Considering that the Madi pluton is characterized as a peraluminous granite, its progenitor magma is typically sourced from hydrous metasedimentary regions. During the ascent of the magma and subsequent crystallization differentiation, volatile-rich fluids reached saturation and exsolved at an earlier stage. In the magmatic-hydrothermal transition phase, the early exsolution of chlorine-rich fluids facilitated the extraction and mobilization of Rb, leading to the alteration of Kfs1 to albite and the formation of Rb-enriched muscovite. Microscopically, the Madi pluton exhibits a gradual transition between potassium feldspar Kfs1 and potassium feldspar Kfs2, with albite developing along their contact boundary. Electron probe microanalysis(EPMA) indicates substantial differences in the Al₂O₃, K₂O, and Na₂O contents between the two feldspars, corroborating that Kfs2 was formed through the hydrothermal dissolution-reprecipitation process of Kfs1. Furthermore, the presence of “myrmekitic texture” at the interface between zinnwaldite and albite further corroborates the involvement of dissolution-reprecipitation processes. During the hydrothermal stage, fluorine- and rubidium-rich fluids facilitated the replacement of Kfs1 through dissolution-reprecipitation, resulting in the formation of microcline(Kfs2) with significant rubidium enrichment(~5 239.38×10-6). This process was accompanied by the precipitation of minerals such as topaz and fluorite. The enrichment of rubidium in potassium feldspar at the Huashi deposit predominantly occurred during the hydrothermal stage, primarily driven by the dissolution-reprecipitation of earlier magmatic Kfs1 by F-rich and Rb-rich hydrothermal fluids. The exsolution of fluids and subsequent autometasomatism during the magmatic-hydrothermal transition played a crucial role in the pre-concentration and mobilization of rubidium. Conversely, early magmatic potassium feldspar contributed minimally to the mineralization process. This study provides valuable insights into the metallogenic mechanisms of granite-hosted rubidium deposits and offers guidance for the exploration of similar deposits.
Indium (In) is a significant rare and dispersed metal. The Tieshilong deposit, situated at the northern terminus of the eastern slope of the central segment of the Nanling Mountains, exemplifies a typical hydrothermal polymetallic deposit enriched with In, Pb, and Zn. Despite its significance, a comprehensive understanding of the occurrence state, primary carrier minerals, and enrichment mechanisms of indium within this deposit remains insufficient. This study employs field investigation, petrographic analysis, and micro-area analytical techniques such as scanning electron microscopy with energy dispersive X-ray spectroscopy (SEM-EDS) and laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) to elucidate the occurrence state and enrichment mechanisms of indium in the deposit. Findings indicate that indium is predominantly concentrated in sphalerite, with concentrations ranging from 2.8×10-6 to 1 124.7×10-6 and an average of 318×10-6, reaching over 1 050×10-6 in certain deeper samples, thus demonstrating a significant enrichment potential. Within sphalerite, indium primarily incorporates into the lattice through isomorphic substitution, exhibiting a positive correlation with Cu and a negative correlation with Zn. This suggests that indium predominantly enters the sphalerite lattice via the coupled substitution mechanism Cu⁺+In³⁺↔2Zn²⁺. When the cadmium (Cd) content ranges from 0.2% to 0.6%, there is a significant increase in the indium (In) content within sphalerite, demonstrating a distinct “indium window effect. ”Microstructural analysis reveals that indium is locally concentrated at the edges and fractures of sphalerite crystals and is frequently associated with cloudy chalcopyrite. This suggests that subsequent hydrothermal activity has imposed a superimposed and remodeling effect on the early indium-rich sphalerite. It is comprehensively concluded that indium in the Tieshilong deposit is not predominantly enriched during the early cassiterite stage. Instead, it primarily incorporates into the sphalerite lattice and achieves enrichment during the medium- to high-temperature Zn-Pb-Cu sulfide mineralization stage. Sphalerite serves as the principal carrier mineral for indium in this region. The coupled substitution of Cu-In, the specified Cd content range, and subsequent hydrothermal remodeling collectively govern the enrichment and localized redistribution of indium. The development of deep indium-rich sphalerite presents promising exploration prospects.
The Xuefeng arcuate tectonic belt is situated in the southwestern segment of the Jiangnan orogenic belt, within the transitional zone between the Yangtze Paraplatform and the South China Caledonian fold system. Renowned for its abundant gold resources, this belt is often referred to as the “golden belt” of Hunan Province. The Qingshuitang gold deposit is located in the central-western region of this tectonic belt, positioned between the Anhua-Xupu-Tongdao and Xinhua-Wuyang-Chengbu thrust nappe deep faults. Through comprehensive field geological investigations and ore microscopy, this study delineates the geological characteristics of the deposit. The exposed strata in the area predominantly comprise the Yanmenzhai Formation of the Qingbaikouan Banxi Group, along with the Chang’an Formation and the Nantuo Formation of the Nanhua System. The ore bodies are primarily controlled by northwest-trending extensional-shear faults and manifest as stratoid and lenticular formations. The ore minerals are predominantly composed of native gold, accompanied by pyrite and arsenopyrite as the primary gold-bearing minerals, while quartz serves as the principal gangue mineral. Alterations in the wall rock include silicification and pyritization. The mineralization process can be categorized into a sedimentary-regional metamorphic phase and a hydrothermal filling-replacement phase, with the latter representing the primary stage of gold enrichment. Based on this framework, in-situ sulfur isotope analyses of pyrite from the Qingshuitang gold deposit were performed using laser ablation multi-collector inductively coupled plasma mass spectrometry(LA-MC-ICP-MS). The findings indicate that the δ³⁴S values of gold-bearing hydrothermal pyrite range from -21.5‰ to -0.2‰, which are consistent with the crustal sulfur domain. However, these values are markedly different from those of sedimentary pyrite in the adjacent wall rocks(the Nantuo Formation of the Nanhua system), which exhibit δ 34S values ranging from 17.2‰ to 23.9‰, with no overlap between the two groups. This suggests that the sulfur within the ore-forming system did not originate from the host strata. Instead, the primary source of the ore-forming materials, specifically sulfur, was metamorphic fluids emanating from deep crustal rocks during regional metamorphism. Given the occurrence of hydrothermal chlorite within the ore veins and the presence of a NW-trending extensional-shear fault system, which constitutes the principal ore-hosting structure, it can be deduced that the ore-forming fluids of the Qingshuitang gold deposit were predominantly derived from metamorphic fluids produced through the devolatilization of deep crustal rocks during regional metamorphism.
Carlin-type gold deposits represent significant low-temperature hydrothermal gold deposits on a global scale. Traditionally, these deposits are characterized by an elemental assemblage of Au-As-Sb-Hg-Tl, with the high-temperature element tungsten (W) historically considered irrelevant to their mineralization processes. However, recent studies have identified widespread anomalous enrichment of tungsten in classic carlin-type gold concentration regions, such as Nevada (USA), Yunnan-Guizhou-Guangxi (China), and West Qinling (China). In certain deposits, the paragenesis of gold and tungsten or the formation of independent scheelite orebodies has been observed, offering critical insights and enhancements to the conventional low-temperature mineralization theory. This paper provides a comprehensive review of global research advancements, elucidates the temporal-spatial distribution and shared metallogenic characteristics of carlin-type gold deposits, and synthesizes the enrichment context, temporal-spatial patterns, elemental coupling relationships, and occurrence states of tungsten. The enrichment mechanism of tungsten is reconstructed by examining the material source, fluid migration, precipitation and enrichment processes, and metallogenic epochs. The existing challenges and research directions are examined with respect to the mineralization process, microscopic mechanisms, and key controlling factors. The findings indicate that tungsten (W) in carlin-type gold deposits predominantly exists in three forms: fine-grained scheelite, W-rich hydrothermal rutile, and as an isomorph in pyrite and arsenopyrite. Tungsten exhibits a strong association with gold orebodies and zones of intense alteration. It primarily originates from deep magmatic and crust-mantle mixed fluids, with additional contributions from W-rich strata. Tungsten migrates stably in the forms of heteropolytungstates, thiotungstates, and tungstate ions. The mixing of fluids results in rapid decreases in temperature and pressure, increases in pH, and decreases in oxygen fugacity. These processes, when combined with reactions involving carbonate wall rocks, collectively facilitate the decomposition and precipitation of tungsten complexes. Tungsten and gold are synchronous, cogenetic, and enriched through the same mechanisms. Tungsten serves not only as an important genetic tracer indicator but also holds significant potential for associated resource utilization in certain regions. Furthermore, it can be employed as an effective prospecting indicator for carlin-type gold deposits.The results can provide a reference for improving the Au-W coupled metallogenic theory and guiding comprehensive exploration and efficient resource utilization.
The Yuka gold deposit is situated within the western segment of the North Qaidam metallogenic belt, to the south of the North Qaidam deep fault, with its principal ore-controlling factors yet to be fully elucidated. Through comprehensive field geological investigations and an integration of existing research, this study identifies the Lower Volcanic Formation of the Ordovician Tanjianshan Group as both the host stratum and a significant source bed. Faults F2 and F12 are recognized as critical ore-transporting structures within the Yuka area, serving as primary conduits for fluid migration. The NW- and NNW-oriented secondary faults, which originate from F2 and F12, play a pivotal role in determining the distribution of ore deposits and orebodies, thus representing the most essential ore-controlling factors in the region. The intrusion of gabbro is posited to have contributed to the development of brittle-ductile shear structures. Furthermore, hydrothermal activity linked to the intrusion of intermediate to acidic dykes is suggested to have facilitated the mobilization and migration of gold (Au), culminating in its enrichment and mineralization. Silicification and pyritization exhibit the closest spatial association with gold mineralization. In contrast, prevalent alterations in the region, such as carbonatization, chloritization, and sericitization, not only facilitate fluid migration by providing advantageous channels but also contribute essential iron-bearing chemical traps for gold mineralization, thereby serving as critical sites for gold precipitation. The secondary ductile shear zones with NW-trending and nearly SN-trending orientations, along with silicification and pyritization, are identified as the most significant and direct prospecting indicators in the Yuka area. Carbonatization, chloritization, and sericitization, in conjunction with the Lower Volcanic Formation of the Tanjianshan Group, serve as indirect prospecting indicators. Furthermore, this paper examines the mineralization model of gold deposits in the Yuka area, based on an analysis of ore-controlling factors and prospecting indicators, as well as previous research findings, with the objective of providing a reference for gold exploration in this and similar deposits.
Achieving a breakthrough in mineral exploration at the Malianjing copper mine has proven challenging due to the area’s complex geological settings and limited deep detection capabilities. To address these challenges, a multi-stage geophysical approach was employed, informed by geochemical anomaly verification from previous exploration phases. Initially, a mineralization-favorable zone covering 3.5 km² was delineated within the working area, utilizing integrated geochemical and geological criteria. Within this target zone, induced polarization (IP) mid-gradient measurements were conducted as a rapid reconnaissance technique to identify polarizable anomalies potentially indicative of sulfide mineralization. Based on the distribution of IP mid-gradient anomalies, a bilateral three-electrode IP sounding profile and an audio-frequency magnetotelluric (AMT) sounding profile were deployed to characterize the electrical structure and polarizability response from shallow to moderate depths. A distinct ore-induced anomaly zone was effectively delineated using the IP intermediate gradient method. In the western section of this anomaly, bipole-dipole IP sounding profile surveys were conducted to investigate the deep extension of the mineralization anomaly, utilizing the superior depth resolution offered by the bipole-dipole array. Conversely, in the eastern section, audio-frequency magnetotelluric(AMT) measurements were employed to further delineate the spatial distribution of the anomaly across a broader depth range, thus providing essential geophysical evidence for examining the ore belt architecture and deep ore-controlling structures. To corroborate the geophysical interpretations, a drilling program was strategically designed and executed at the 600 m line position of the IP sounding profile. The borehole reached a terminal depth of 256.46 meters. Comprehensive core logging and assay results identified the presence of two copper ore bodies, each exhibiting an apparent thickness of 1 meter, and one copper mineralized body also with an apparent thickness of 1 meter. The copper grades in the two ore bodies were determined to be 0.96% and 0.22%, respectively, thereby confirming the presence of economically viable mineralization.The spatial positioning and depth of the intersected bodies exhibited a favorable correlation with the delineated induced polarization (IP) anomalies, particularly aligning with the high-chargeability and moderately low-resistivity signatures anticipated by the combined interpretative approach. This successful validation underscores the efficacy of integrating high-power IP surveys, which offer substantial penetration depth and a robust signal-to-noise ratio, with audio-frequency magnetotellurics(AMT), which provides reliable resistivity imaging of deep conductive structures, for exploration in this mining area. This integrated methodology not only bolsters confidence in guiding subsequent borehole placement but also serves as an effective model for geophysical exploration of copper deposits in this region, as well as in adjacent mining areas facing similar geological and exploration challenges.
To examine the dynamic mechanical properties, energy dissipation mechanisms, and damage evolution of sandstone under impact loading, as well as to develop a dynamic damage constitutive model that accurately describes its complete response process, dynamic compression tests were performed on sandstone specimens utilizing a 50 mm split Hopkinson pressure bar (SHPB) system at impact pressures of 0.1, 0.2, 0.3, 0.4, 0.5 MPa. The experiments yielded dynamic stress-strain curves, strain-rate time-history curves, and energy evolution characteristics across varying impact pressure levels. Based on these findings, a piecewise statistical damage constitutive model was formulated using a three-parameter Weibull distribution. The results indicate that sandstone exhibits a pronounced strain-rate effect under impact loading. As impact pressure increases, the average dynamic compressive strength rises from 81.07 MPa to 137.24 MPa, the dynamic failure strain increases from 0.7247×10⁻² to 1.1379×10⁻², and the dynamic elastic modulus increases from 18.65 GPa to 22.35 GPa. These observations suggest that sandstone fails at higher stress levels and undergoes greater deformation under elevated strain-rate conditions. The dynamic stress-strain response typically progresses through distinct stages: compaction, quasi-elastic deformation, nonlinear damage accumulation, and rapid post-peak softening. These stages reflect the sequential closure, propagation, and coalescence of internal microcracks, culminating in overall failure. Energy analysis reveals that incident, reflected, and absorbed energies all increase with rising impact pressure, whereas transmitted energy remains relatively low. The increase in absorbed energy correlates strongly with the fragmentation evolution of the specimens, transitioning from blocky to pulverized failure. This suggests that absorbed energy is predominantly expended in crack initiation, propagation, fragment separation, and frictional dissipation. Utilizing continuous damage mechanics, the entire dynamic compression process of sandstone is categorized into the compaction, quasi-elastic, and damage stages. Model parameters are determined through a combination of a peak-anchoring strategy and the nonlinear least-squares method. The findings suggest that the proposed model effectively characterizes the pre-peak nonlinear response, peak strength attributes, and initial post-peak softening behavior of sandstone across varying strain rates. This research offers an empirical foundation and theoretical framework for analyzing dynamic responses, characterizing damage evolution, and developing related constitutive models for sandstone subjected to impact loading.
As mining operations progress to greater depths, the surrounding geological formations are exposed to elevated temperatures, variable pressures, and significant disturbances from mining activities. This results in increased heterogeneity of the medium and substantial noise interference on-site, which undermine the precision and stability of microseismic localization algorithms. To enhance the reliability of microseismic localization in complex deep mining environments, this study introduces a high-precision localization method utilizing the XGBoost machine learning algorithm and its hybrid variants. Initially, a comprehensive multi-source dataset was developed by integrating both publicly available and real-world mining data. A systematic data preprocessing workflow—comprising steps of cleaning and filtering, denoising, resampling, and normalization—was implemented to enhance signal quality. Subsequently, XGBoost was employed as the core model, with its optimal configurations identified through hyperparameter optimization and sensitivity analysis. The proposed method was evaluated against several traditional algorithms, including Multi-Layer Perceptron(MLP), AdaBoost, Bayesian Ridge(BR), Decision Tree, K-Nearest Neighbors(KNN), and Lasso, to assess its comparative performance. In conclusion, three strategies for fusion enhancement were proposed: XGB-Select for feature selection, STL-Feat and SVR-Feat for pre-processing feature enhancement, and STL-Res and SVR-Res for residual fitting correction. The findings indicate that XGBoost achieves lower average localization errors and higher positioning success rates(PSR) under conditions of noise disturbance and wave velocity perturbation, demonstrating superior localization accuracy and robustness compared to traditional algorithms in complex nonlinear environments. Among the integrated models, the STL-Res model is identified as the most favorable approach due to its exceptional adaptive capability. XGB-Select offers significant improvements in training efficiency with only a slight compromise in accuracy. The SVR-Feat model effectively extracts nonlinear features in environments free from interference or with low noise, making it particularly suitable for scenarios with sparse sensor deployment, where it performs well under low sensor density; however, its localization accuracy is notably sensitive to noise disturbances. This research contributes a reference technical solution for disaster risk assessment in deep mines.
The precise determination of crown pillar thickness is a crucial concern in deep underground mining, especially under intricate geological and hydrogeological conditions, such as mining beneath surface infrastructure, water bodies, or during the transition from open-pit to underground operations. The crown pillar functions as a vital structural component that separates underground stopes from the overlying strata and surface environments. Insufficient thickness can result in plastic penetration, roof instability, water ingress, or surface subsidence, while excessive thickness leads to substantial ore loss and diminished economic efficiency. Consequently, identifying an optimal thickness that balances safety and resource recovery remains a fundamental engineering challenge. This study proposes a multi-method combined approach to optimize the crown pillar thickness for a complex iron-polymetallic mine utilizing staged open-stoping with subsequent backfill. Initially, several traditional theoretical approaches—including the material mechanics method, span-to-thickness ratio method, load transfer line intersection method, Prussian arch theory, structural mechanics method, and Lupenite theory—were employed to estimate a preliminary range of safe thicknesses under varying safety factors. The theoretical findings suggested that, when the safety factor exceeds 1.5, the minimum required thickness should not be less than 18 meters. Subsequently, a three-dimensional numerical model of the mining area was developed using the Mohr-Coulomb constitutive model. The staged mining and backfilling sequence was simulated in four steps to capture the progressive deformation and plastic evolution of the crown pillar. Four thickness schemes(15 m, 20 m, 25 m, and 30 m) were comparatively analyzed. The results indicate that vertical displacement exhibits a funnel-shaped distribution, whereas horizontal displacement follows an “S”-shaped pattern, with maximum deformation occurring in the thickest portion of the orebody. When the thickness is 15 meters, plastic zones penetrate through the pillar, indicating structural instability. For thicknesses of 20 meters and above, plastic zones remain unconnected, thereby maintaining overall stability. Additionally, both maximum vertical and horizontal displacements decrease exponentially with increasing thickness. The most significant reduction in deformation is observed when the thickness increases from 15 meters to 20 meters, suggesting the presence of a critical interval for stability enhancement. Beyond 20 meters, the marginal improvement in stability gradually diminishes, while ore loss continues to increase. Through the integration of theoretical calculations and numerical simulations, the optimal crown pillar thickness for the studied mine was determined to be 20 meters. This research not only establishes a systematic framework that combines analytical and numerical approaches but also introduces a deformation attenuation-based optimization criterion, providing quantitative guidance for crown pillar design in similar complex mining conditions.
As mineral resource extraction extends to deeper subterranean levels, the substantial accumulation of waste rock presents significant ecological and safety challenges. Cemented paste backfill technology, which utilizes solid waste, not only manages deep rock strata but also enables large-scale waste disposal. In Southwest China, the conversion of vanadium-titanium magnetite waste rock into manufactured sand as a substitute for natural aggregates in backfilling offers a promising sustainable solution. However, accurately predicting the uniaxial compressive strength(UCS)—a critical evaluation metric—remains challenging due to the intricate internal hydration mechanisms and highly variable data. Traditional statistical models, such as the response surface methodology(RSM), often exhibit physical inaccuracies, predicting implausible negative strength values in extreme boundary conditions characterized by low cement-to-sand ratios and early curing ages. To address these limitations and elucidate the nonlinear evolution mechanisms of UCS, this study conducted a comprehensive 64-group full-matrix cross-experiment. Manufactured sand sourced from a vanadium-titanium magnetite mine in Yunnan was employed as the primary backfill aggregate. The experimental framework identified solid mass concentration, cement-to-sand ratio, and curing age as the primary controlling variables. We conducted a systematic analysis of the multifactorial coupling effects through single-factor evaluation and three-dimensional response surface analysis. Subsequently, we developed an advanced intelligent prediction model to address the structural limitations of traditional polynomial regression. The discrete experimental data were processed using a novel logarithmic space transformation and Min-Max normalization. A back propagation(BP) neural network with a 3-8-1 topology was constructed and integrated with a genetic algorithm(GA) to globally optimize the initial network weights and thresholds, thereby avoiding local optima. The experimental results indicated that the uniaxial compressive strength(UCS) of the cemented backfill exhibits a continuous nonlinear decrease with a reduction in the cement-to-sand ratio. In contrast, the UCS demonstrates a pronounced peak effect in response to variations in mass concentration, achieving maximum structural integrity at an optimal concentration of 82%, where the internal hydration environment and aggregate skeleton are most fully developed. The multi-dimensional coupling analysis demonstrated that the interaction between the cement-to-sand ratio and curing age exerts the most significant influence, while the interaction involving mass concentration is relatively weaker. Importantly, the comprehensive validation of the proposed GA-BP neural network model across the full sample resulted in a high coefficient of determination(R²) of 0.9147 and a root mean square error(RMSE) of 0.685 MPa. Ultimately, the newly developed GA-BP framework, enhanced by logarithmic spatial transformation, effectively mitigated predictive distortions in regions characterized by extremely low values. This framework achieved a high-precision, unbiased nonlinear mapping of uniaxial compressive strength(UCS) across a diverse array of complex mix proportions. This advanced modeling approach offers a robust scientific foundation and technical reference for the large-scale utilization of mine solid waste resources and the automated design of backfill proportions.
Cemented whole tailings backfill(CWTB) represents a pivotal technology in the pursuit of solid waste resource utilization and environmentally sustainable mining practices. Investigating the strength characteristics of CWTB is critically important for ensuring mine safety. This study addresses the ambiguous strength evolution of backfill under low-temperature conditions prevalent in cold-region mines, as well as the limitations in accuracy associated with traditional prediction methods. Uniaxial compressive strength(UCS) tests were conducted on CWTB samples from Yulong Mining Co., Ltd. in Inner Mongolia, considering the combined effects of four factors: curing temperature (ranging from 5 ℃ to 20 ℃), cement-to-tailings ratio (ranging from 1∶4 to 1∶10), slurry mass concentration(ranging from 62% to 68%), and curing age (ranging from 3 to 28 days). Utilizing the results from orthogonal experiments, analysis of variance(ANOVA) was applied to quantitatively assess the significance of each factor. Furthermore, a multivariate linear regression model incorporating two-factor interaction effects was developed through response surface methodology(RSM) to elucidate the coupled influence mechanisms of these multiple factors. A support vector regression(SVR) model was developed using four factors as inputs and unconfined compressive strength(UCS) as the output variable. The model’s critical hyperparameters, namely the penalty factor (C) and the kernel parameter(g), were globally optimized through a genetic algorithm(GA), yielding a high-precision GA-SVR strength prediction model. The findings suggest that, within the experimental parameters, the cement-tailings ratio is the most influential factor in strength development, followed by curing age and temperature, while the impact of slurry mass concentration is comparatively minor. Low temperatures significantly hinder the hydration process. However, this negative effect can be effectively mitigated by optimizing the cement-tailings ratio and prolonging the curing period. The GA-SVR model demonstrated coefficients of determination(R²) of 0.999 and 0.998 for the training and testing datasets, respectively. The root mean square error(RMSE) and mean absolute error(MAE) were significantly reduced in comparison to the traditional support vector regression(SVR) model, indicating a substantial enhancement in predictive performance. These findings further confirm the efficacy of genetic algorithm(GA) optimization in improving both fitting accuracy and generalization capability. Based on engineering requirements, the proposed model determined optimal cement-to-tailings ratios of 1∶4.3 and 1∶8.8 for a curing temperature of 15 ℃, a slurry concentration of 66%, and target compressive strengths of ≥2.4 MPa at 3 days and ≥0.6 MPa at 7 days, respectively. This optimization strategy can reduce the costs of cementitious materials by 7.36% and 8.88%, respectively, while significantly decreasing the experimental workload. The high-precision predictive model developed in this study offers a reliable quantitative tool for the strength design and optimization of CWTB in cold-region mines, with significant theoretical and practical implications for engineering.
The regulation of material levels and discharge flow rates in slurry mixing equipment is crucial for maintaining filling body production quality and ensuring the safety of transportation pipelines. In a particular mining operation utilizing the paste pump filling process, there is potential for further optimization of the structural parameters of the mixing and feeding system. Through theoretical calculations and numerical simulations, the mechanism by which the pipe clamp valve controls the discharge flow rate is elucidated. It is observed that when the pipe clamp valve is opened to a degree of 25% to 30%, the discharge flow rate ranges from 92.0 m³/h to 109.4 m³/h. The constriction of the flow channel results in significant pressure loss of the slurry, thereby facilitating control over the discharge flow rate. However, continuous high-speed erosion and negative pressure due to insufficient valve opening pose safety risks, such as excessive wear and potential pipe rupture. By removing the pipe clamp valve and implementing a reverse “Z” shape reverse slope discharge structure, with a reverse slope height of 1.2 m and an inner discharge pipe diameter ranging from 150 mm to 140 mm, the discharge flow rate is adjusted to between 106.6 and 84.3 m³/h, while maintaining a stable material level at 1.8 m. Through theoretical calculations, numerical simulations, and semi-industrial tests, the pressure variation patterns along the pipeline were confirmed to be consistent, and the data closely aligned, demonstrating that the model and analysis are both accurate and reliable. The implementation of a reverse slope structure effectively eliminates the need for pipe clamp valves and fault points, thereby enhancing operational safety, reducing the depth of the pump pit, and decreasing the construction costs associated with the civil engineering of the filling preparation workshop. This study provides valuable insights and guidance for practical engineering applications.
Goafs are characterized by significant concealment, complex morphology, and indistinct boundaries. Inaccurate identification of their occurrence conditions can adversely affect hazard risk assessment and the development of control measures, potentially leading to engineering disasters such as collapses, water inrushes, and mud inrushes. Consequently, precise detection of goafs is of paramount importance for ensuring mine safety and for the prevention and control of disasters. This paper provides a systematic review of the advancements in goaf detection technologies, summarizing the technical principles, applicable conditions, advantages, and limitations of the predominant methods for both exposed and concealed goafs. Additionally, it examines the adaptability of various detection methods under different geological conditions and explores the characteristics of multi-method joint applications. The findings indicate that a high-precision detection system for exposed goafs has been developed, utilizing technologies such as 3D laser scanning, UAV laser scanning, and RTK-GNSS. In contrast, concealed goafs are primarily investigated through electromagnetic, electrical, seismic, and microgravity methods, as well as muon detection and drilling verification. There is an observable trend towards the integration of multi-physics collaborative detection and joint inversion techniques. It is evident that single-method approaches are insufficient for the accurate identification of goafs in complex geological settings. Consequently, the fusion of multi-source information, intelligent interpretation, and digital modeling are anticipated to be pivotal directions for future advancements. This study offers a valuable reference for the selection of goaf detection technologies, their engineering applications, and future research endeavors.
As shallow mineral resources become depleted, mining operations are increasingly extending into deeper strata. This expansion often results in ground pressure disturbances caused by goafs, which can lead to deformation and failure of vertical shafts, thereby impeding safe production in deep mining environments. The majority of existing research focuses on the disturbance characteristics of shafts within single mining areas, with limited studies systematically examining the coordinated evolution of shaft displacement, stress, and plastic zones under varying mining sequences of upper and lower superimposed orebodies. Utilizing the Xincheng gold mine in Shandong Province as a case study, this paper investigates the upper orebody located at depths of -335 to -480 meters and the unmined lower orebody at depths of -680 to -885 meters, where an upward layered filling mining method with a layer thickness of 3 meters is employed. A high-precision three-dimensional numerical model is developed using the integrated 3DMine-Rhino-Griddle-FLAC3D modeling approach, featuring dimensions of 3 519.52 m×2 102.50 m×576.40 m and comprising over 2.16 million tetrahedral elements. The rock mechanical parameters are refined according to the Hoek-Brown criterion, and displacement constraint boundaries are established to accurately simulate in-situ stress conditions. Three mining strategies have been developed: (1) concurrent extraction of both upper and lower orebodies, (2) exclusive extraction of the upper orebody, and (3) sequential extraction of the lower orebody following the complete extraction and backfilling of the upper section. Displacement monitoring lines have been installed along the shaft lining, and cross-sectional analyses are conducted to evaluate the differential impacts of these mining methods on shaft stability. The shaft is categorized into near-surface, mid-deep, and deep sections based on the intensity of disturbance. The near-surface section experiences minimal mining disturbance, with maximum deformation not exceeding 2 mm. The mid-deep section is identified as the primary disturbance zone; concurrent mining results in compounded disturbances, with a maximum horizontal displacement of 5.3 mm. In contrast, sequential mining induces significant vertical displacement, whereas mining solely the upper orebody results in the least deformation. The deep section exhibits the most pronounced deformation variations. Following the extraction of the lower orebody, the maximum three-dimensional displacement reaches 4.0 mm, with the shaft deviating approximately 2.5 mm from the goaf. Stress concentrations are predominantly located on both sides of the orebodies and at the shaft bottom, with peak principal stress reaching up to 61.7 MPa. The sequential extraction of the lower orebody induces localized tensile stress of 0.57 MPa in proximity to the shaft. This stress can be mitigated through the concurrent excavation of both upper and lower orebodies. The predominant mechanism of rock mass failure is shear damage, with plastic zones primarily developing in the hanging wall of the orebodies. Across all three mining schemes, no mining-induced plastic failure is observed around the shaft, and only minor plastic bands are present in the original excavation zone of the shaft. Mining exclusively the upper orebody results in the smallest extent of the plastic zone. A comprehensive analysis of displacement, stress, and plastic zone evolution indicates that mining solely the upper orebody induces the least disturbance and preserves overall shaft stability. Simultaneous mining results in significant stress concentration without causing evident instability of the shaft. Sequential mining from the upper to lower orebodies generates the greatest disturbance; however, the shaft's response remains manageable. This study confirms that upward layered filling mining can effectively mitigate ground pressure damage to shafts during deep mining operations. It establishes a comprehensive technical framework encompassing 3D geological modeling, numerical simulation, surrounding rock response analysis and shaft stability evaluation. The research provides theoretical support and engineering reference for mining sequence optimization, shaft safety protection and ground pressure control in deep mines with multi-level superimposed orebodies.
Mine ventilation is fundamental to ensuring safety in underground mining operations, as its dependability has a direct impact on personnel safety, operational continuity, and emergency response capabilities. Traditional ventilation systems have predominantly depended on manual expertise and periodic adjustments, which present notable limitations. These limitations include delayed responses to dynamic disturbances, excessive energy consumption, and inadequate resilience under abnormal conditions such as fires or gas outbursts. These challenges are further intensified in deep mining environments, where intricate geological conditions and high-stress settings necessitate more responsive and adaptive ventilation solutions. The ongoing advancement of intelligent mines is instigating a paradigm shift in ventilation systems, transforming them from passive safety infrastructure into intelligent hubs that integrate real-time sensing, autonomous control, and production coordination. In this context, this review offers a systematic examination from technology, engineering, and management perspectives across four critical dimensions: The intelligent sensing of the ventilation environment involves comprehensive multi-parameter monitoring of airflow, gas concentrations, and equipment status. Real-time ventilation network calculations facilitate precise estimations of airflow distribution under changing boundary conditions. Intelligent decision-making and regulation entail the dynamic control of fans, dampers, and air doors. Ventilation visualization and simulation utilize digital twin technologies to support scenario-based analysis and predictive reasoning. Despite these significant advancements, several bottlenecks persist. Foremost among these is the incomplete closure of the “sensing-decision-control” loop, where sensing data are not seamlessly translated into automated control actions. Multi-system integration remains challenging due to heterogeneous communication protocols and data silos, and traditional management models often fail to meet the operational requirements of intelligent systems. To address these gaps, a development framework is proposed comprising three synergistic components: (1) an integrated technical architecture focused on “sensing interconnection-simulation and prediction-collaborative decision-making”; (2) unified data standards to facilitate interoperability across subsystems, including ventilation, production, and safety monitoring; and (3) a standardized“operational condition-specific strategy” management system that supports precise regulation under diverse mining scenarios. Future directions for intelligent mine ventilation highlight the convergence of advanced intelligent sensing, digital twin modeling, and autonomous control technologies. Achieving this integration will facilitate a fundamental shift from experience-driven operations to data-driven decision-making, and from passive response mechanisms to proactive pre-control capabilities. These advancements will provide crucial support for developing safe, efficient, and low-carbon intelligent mines, ensuring that ventilation systems evolve in alignment with the increasing complexity of underground mining operations.
The truck dispatching system in open-pit mining is crucial for optimizing production efficiency and resource utilization. Nonetheless, the operational environment is characterized by dynamic and uncertain factors, such as variations in equipment operation cycles and the heterogeneous distribution of ore grades. These factors can compromise the effectiveness of predetermined dispatching plans if timely and adaptive decision-making is not implemented, thereby adversely affecting production targets and ore quality control. To address these challenges, this study proposes an optimization model(DRL model) for open-pit mine truck dispatching utilizing the Double Deep Q-Network(DDQN)algorithm. Initially, a high-fidelity simulation environment for trucks and loaders in open-pit mines was developed using the AnyLogic platform, providing a realistic simulation of the loading, transportation, and unloading processes, as well as accounting for random interference factors. Subsequently, a multi-objective reward function was designed, incorporating key performance indicators such as production achievement rate, grade stability, and equipment utilization rate. The Double Deep Q-Network (DDQN) algorithm was employed to train the agent in learning the optimal dispatching strategy, addressing the continuous decision-making challenge within a high-dimensional state space. By decoupling action selection from target value computation, DDQN effectively mitigates the overestimation bias inherent in traditional Deep Q-Networks(DQN), thereby resulting in a more stable training process that is particularly well-suited for high-dimensional dynamic dispatching scenarios. Experimental results demonstrate that, in comparison to both conventional dispatching strategies and the standard DQN approach, the proposed model effectively adapts to production fluctuations, significantly reduces the deviation between the actual workload and the target values at each loading point and crushing station, maintains ore grade stability around the target level of 66%(with fluctuations controlled within ±5%), and exhibits rapid convergence and robust performance. This research achieves dual collaborative optimization of open-pit mine truck dispatching in terms of both quantity and quality, thereby providing effective decision support for the intelligent production of mines.
Underground mining operations take place in enclosed and intricate environments, where miners are exposed to numerous hazards, including roof collapses, sidewall spalling, falling objects, and fires. Consequently, the standardized use of essential personal protective equipment (PPE) is vital for ensuring both operational safety and facilitating emergency evacuations. The advancement of underground video surveillance systems has enabled vision-based automatic supervision of key PPE, thereby enhancing regulatory efficiency. Nevertheless, challenges such as target occlusion, back-facing targets, and the simultaneous presence of multiple individuals within the same frame in underground video monitoring often result in a high false-alarm rate when alarms are generated solely based on single-frame detection outcomes. To mitigate this issue, a compliance determination and alarm methodology for monitoring the wearing of key PPE in underground mine videos is proposed. Utilizing the DsDPM66_PPE dataset, three detectors were developed to identify miners, helmets, and self-rescuers. The detected PPE items were then associated with the corresponding miners, and continuous individual tracks were established by integrating head, chest, and waist discriminability conditions. Negative evidence was systematically accumulated across PPE-related tracks, and corresponding alarms were activated once the accumulated negative evidence surpassed a predetermined threshold. The findings indicated that the mAP50 values for the three target detectors each achieved 0.993, while the mAP50-95 values were 0.920, 0.880, and 0.846, respectively. A total of 2 603 tracks were generated from 66 videos, among which 18 were identified as alarm tracks, resulting in an alarm precision of 94.4%. The optimal performance was observed when K_SELF_NEG was set to 5. The proposed methodology effectively mitigates false alarms arising from missed detections, occlusion, and short-term invisibility, thereby providing technical support for violation screening and assisted review in the safety supervision of critical PPE in underground mining environments.
The global mining industry is currently advancing its transition towards digitalization and intelligence, with smart mining emerging as the central focus of industry development. Surface transportation, serving as a crucial link between underground mining and surface processing, significantly influences the overall production efficiency and operational costs of mining operations. However, the scheduling of mine vehicles is complicated by multi-source and multi-sink constraints, heterogeneous fleets, and intricate dynamic interdependencies. Traditional scheduling approaches, which largely depend on manual expertise or simplistic heuristic rules, often fail to accommodate dynamic changes in such complex environments, resulting in suboptimal system efficiency and limited adaptability to disruptions. To overcome the limitations of traditional methods, particularly their inadequacy in managing high-dimensional state spaces and sparse reward signals, this paper introduces an innovative intelligent scheduling approach (HRLGNN) that integrates graph neural networks (GNN) with hierarchical reinforcement learning (HRL). Initially, the entire scheduling system is rigorously formalized as a heterogeneous graph structure, consisting of ore bin nodes, crusher nodes, and vehicle nodes. This modeling approach effectively captures the topological relationships and state information inherent in the transport network. Subsequently, a feature extraction network utilizing graph attention networks (GAT) is developed. By computing attention coefficients, this network facilitates adaptive weighting of neighboring nodes, thereby enabling an in-depth exploration of spatiotemporal feature correlations among heterogeneous entities. Furthermore, to address the challenge of dimensionality in large-scale scheduling, a two-tier cascading decision-making framework is established based on the actor-critic architecture. The high-level agent is tasked with prioritizing vehicle selection from the idle fleet, while the low-level agent manages specific destination assignments, such as selecting optimal ore bins or crushers. The model is trained using the proximal policy optimization (PPO) algorithm to ensure robust and stable policy updates. A series of comprehensive simulation experiments were executed across a spectrum of application scenarios, encompassing both small-scale and large-scale instances, to evaluate the proposed algorithm in comparison to four deep reinforcement learning (DRL) baselines and four heuristic rules. The experimental findings indicate that the HRLGNN algorithm demonstrates superior convergence speed and stability. Quantitatively, the proposed method attained optimal outcomes in 14 out of 16 test instances, with an average gap ratio of merely 0.05 across all instances, thereby significantly surpassing the performance of the comparative algorithms. Furthermore, the simulation results in real-world scenarios reveal that the proposed method effectively minimizes the maximum makespan, enhances vehicle utilization rates, and reduces vehicle queuing times. This method is capable of holistically considering various factors in the scheduling process, thereby improving overall operational efficiency. The proposed methodology addresses the challenges associated with sparse rewards and intricate system coupling in mine vehicle scheduling with notable efficacy. By decomposing the joint decision-making problem and employing graph-based inductive biases, the method significantly enhances overall operational efficiency. This research not only delivers a high-precision, responsive solution for mine logistics but also contributes novel theoretical insights and methodologies applicable to a wider range of intelligent manufacturing logistics optimization issues.
Online Submission
Peer Review
Office Work
Editor-in-Chief