Aiming at the high temperature-pollution coupling problem caused by complex diagonal ventilation in the large-scale mining of dense stope groups of deep well in metal mines, a four-dimensional analysis framework of "orthogonal test and control variable method, Fluent numerical simulation, Ventsim ventilation solution and multiple linear regression" was constructed, which revealed the key role of the selective closing of the wind door in the cross-cut in the reconstruction of the wind path. The results show that the importance of the influence of various factors on the temperature field of the ore drawing roadway is as follows: the surrounding rock temperature>the inlet wind velocity of the 2^# cross-cut>the inlet wind temperature>the inlet wind velocity of the 0^# cross-cut. The optimization scheme of closing the wind door of 0^# cross-cut roadway and constructing 2^# inlet and 0^# outlet through the wind path is proposed. The scheme increases the wind volume of the ore drawing roadway from 10.8 m³/s to 30.7 m³/s, increases the regional wind speed from 0.1 m/s to 0.6 m/s, and reduces the average temperature in the roadway to 26.03 ℃. The optimization scheme effectively solves the problem of local crosswind, improves the efficiency of pollutant discharge, and improves the underground working environment. It provides a theoretical basis and engineering practice example for the regulation of "precise minimally invasive surgery" for the ventilation system of dense stope groups in similar deep wells.

Roadways are prone to deformation and failure under the influence of high in-situ stress as mining progresses to deep underground areas. Cracks in shotcrete, mesh fragmentation and weathering spalling of surrounding rock occur in the heavy haulage roadway of a crushing chamber in an ultra-deep metal mine. In-situ stress and rock mechanical parameters were measured by the hydraulic fracturing method and rock mechanics tests. Deformation characteristics of the roadway were analyzed via deformation monitoring, borehole photography and ground pressure monitoring, and the failure mechanism of surrounding rock was explored based on monitoring data. A combined support scheme was proposed, consisting of 30 mm thick initial shotcreting, 5 m long anchor cable for active support, steel mesh, U29 steel arch for passive support and 120 mm thick secondary shotcreting. Engineering practice shows that the bearing capacity of the surrounding rock is enhanced after reinforcement, no deformation and failure are observed, and the surrounding rock maintains a relatively stable state.

Aiming at the problems that the traditional engineering analogy method is easily affected by subjective evaluation, and that the simple numerical simulation method involves complex modeling and heavy workload in the optimization of stope structure parameters, an artificial intelligence stope parameter optimization method based on DBO-MLP-SVM is proposed. The Dung Beetle Optimizer (DBO) was used to optimize the performance of Multilayer Perceptron (MLP) and the Support Vector Machine (SVM). The two were fused to overcome the defect that MLP is easy to fall into a local optimum, and the relevant mine stope parameters and stope stability classification data were collected to train the proposed model. Taking a limestone mine in southwest China as an example, the stability classification under different pillar widths was obtained by substituting the model, and the stope displacement, pillar deformation and surface settlement were analyzed and verified by FLAC3D numerical simulation. The results show that the DBO-MLP-SVM model predicts that the stope is stable when the pillar width is greater than 8 m, and the optimal stope structure parameter is pillar width×spacing = 8 m×8 m, which is consistent with the numerical simulation results. Compared with the traditional engineering analogy method, the production capacity of the stope after intelligent collaborative optimization increased by 51.80%, and the loss rate and dilution rate changed little, which effectively balances production efficiency and safety. This method provides an efficient and accurate technical approach for the optimization of stope structure parameters, which can serve as a reference for the optimization of mine stope parameters.

The extraction of overlain resources shall be conducted on the premise of service shaft safety. Taking an iron mine in Anhui Province as the research background, the disturbance influence exerted by the extraction of overlain resources on No.2 service shaft was studied. Mine geological surveys and analyses on physical and mechanical parameters of rock mass were carried out. Two mining schemes were determined by means of the empirical analogy method. Numerical simulations of the two mining schemes were performed via multi-software collaboration. The stability of the service shaft after the extraction of overlain resources was analyzed, and the feasible scheme was selected. Finally, the stress state around the service shaft under horizontal stress disturbance was analyzed by adopting the stress concentration principle of circular holes in thin plates. The research results show that both the two-step mining and skip-one-mine-one schemes for overlain resources are feasible. From the perspective of improving resource utilization, the scheme of mining rooms in the first step and extracting pillars in the second step with timely filling is preferred. The stress around the service shaft induced by mining disturbance is lower than twice the in-situ stress, and the shaft remains in a stable state. The research results provide technical support for the safe extraction of overlain resources around service shafts.

After a certain iron mine in Henan Province was mined by sublevel caving method, due to the lack of forced caving measures, the thickness of the upper overburden strata became insufficient as the stope advanced to depth, resulting in the collapse of the goaf roof and the remaining point pillars from the original room-and-pillar method in the -40 m sublevel, which seriously threatened the safe mining of the stope. Using Mathews stability graph method and numerical simulation, the stability of the goaf was analyzed and studied. It was found that there was a large area of caving risk in the roof of the goaf. Therefore, it was suggested that the -72 m sublevel stop using the caving mining method, that the area above -72 m should be sealed off, and the area above -90 m should be mined by the filling mining method instead. Considering the mining conditions of ore body and the production capacity requirements of the mine, three technically feasible mining schemes were proposed: the sublevel open stoping with subsequent backfill method, the panel mechanized upward horizontal cut-and-fill stoping method, and the upward horizontal cut-and-fill stoping method perpendicular to the orebody strike. Taking into account the technical and economic indicators and mining safety, it was recommended that the mine adopt the panel mechanized upward horizontal cut-and-fill stoping method. After the mine transitioned from the caving mining method to the filling mining method, the ore dilution rate decreased to 10%, the loss rate decreased to 8%, and the development-to-production ratio increased to 7.47 m/kt. The annual output met the mine's design and production requirements, and within the remaining service life, the mine economy benefit could increase by 49.476 7 million yuan.

Ore and rock are distributed alternately in Las Bambas Copper Mine, with varying hardness and diverse lithology across zones. Single blasting parameters are hardly applicable to the complex geological conditions. To achieve proper ore fragmentation and reduce waste rock dilution during blasting, research was conducted on ore-rock separation controlled blasting technology and fragmentation prediction for open-pit bench blasting. Rock mass blastability was classified based on the Rock Mass Index (RMI). In-situ retention blasting and cast blasting were proposed to distinguish between ore and rock, thereby achieving zoned accumulation of broken materials. Schemes of multi-point initiation with differentiated borehole pattern parameters and dense empty hole blasting were designed. Blasting data of Ferrobamba mining area were used for training based on XGBoost model, and ore blasting fragmentation of 25 sample groups was predicted and verified through field tests. The results show that with the multi-point initiation scheme using differentiated borehole pattern parameters, a hole depth of 16.5 m, borehole pattern parameters of 9.0 m×8.0 m for rock area and 6.5 m×5.5 m for ore area, distinct ore-rock boundary grooves are formed after blasting, and ore and rock accumulate separately. The characteristic parameter D₈₀ of blasting fragmentation distribution is 127 mm, while the predicted D₈₀ value is 124 mm with an error of only about 2.36%. The proportion of fragments sized smaller than 25.6 mm reaches 36.7%, which satisfies the requirements of the concentrator.

To address the prevalent engineering challenges of safety production monitoring and early warning in non-coal mines, such as the fragmentation of multi-source heterogeneous data, lagging risk identification, and insufficient regulatory collaboration, a "One Map" system for safety production early warning in non-coal mines based on multi-source data fusion technology was constructed. Relying on core technologies including multi-source data fusion, real-time status monitoring, and safety risk four-color map visualization, the system designed a dual-track architecture that couples government supervision terminal and enterprise supervision terminal collaboratively. The government terminal focused on macro-level compliance supervision and emergency linkage command, while the enterprise terminal emphasized micro-level closed-loop rectification of risks and hidden dangers. Through a data sharing and collaboration mechanism, a new integrated closed-loop supervision model was formed. The engineering application results of this "One Map" collaborative early warning system in three mining enterprises in Lu'an City show that the system improves the safety supervision efficiency at the government terminal by approximately 22% and achieves a risk assessment accuracy of over 90%. At the enterprise supervision terminal, the hidden danger handling time is reduced by 30.5%, and the accident rate is decreased by 15%. The data utilization rate at the multi-source data collaborative application terminal is increased by 20%. The research provides an effective technical implementation path for intelligent safety supervision in non-coal mines and holds reference value for promoting the digital transformation of the industry.

Intelligent early warning of mine safety risk is the key technical means to ensure mine safety production. In the face of challenges such as complex mining environment, strong multi-factor coupling of disasters, and large amount of multi-source heterogeneous data, systematic research was carried out on the typical application scenarios and fusion path of mathematical methods. Starting from three typical application scenarios: dynamic prediction of disaster risk, real-time perception and intelligent control of human-machine-environment total factors, and emergency decision-making optimization of accident chain traceability and simulation, the evolution rules of mine safety early warning technology "from single to multiple, from traditional to intelligent" "from discrete monitoring to closed-loop contro" and "from experience-driven to intelligent assistance" were sorted out. On this basis, the mathematical methods supporting the above application scenarios were summarized, including basic statistics and spatio-temporal analysis, machine learning and deep learning, complex system modeling, mechanism-data hybrid driving model and digital twin model. The collaborative framework of application scenarios and mathematical methods was constructed, and the selection and combination of mathematical methods in different application scenarios were clarified. The research results can provide technical reference and method support for the transformation of mine safety risk early warning to "active prevention and control".

Aiming at the stability control of surrounding rock in gently inclined orebody stopes and safety risks of surface waste dumps under the coupling effect of surface load and mining disturbance, mechanical parameters of key rock strata were measured via indoor rock mechanics tests. Combined with physical similarity simulation adopting Digital Image Correlation (DIC) technology and three-dimensional numerical simulation, fracture evolution features as well as response laws of stress and displacement fields of stope surrounding rock under dual disturbance were systematically analyzed. Test results show that mining disturbance induces stress redistribution in surrounding rock. Superimposed surface load increases the maximum vertical displacement of stope roof directly beneath the waste dump by 12.86%, and a subsidence center consistent with the theoretical boundary of mining pressure relief arch forms on the surface. Tensile stress concentrates in the roof, floor and upper surrounding rock of the stope, while compressive stress concentrates in inter-level roof and floor pillars. Backfill mining weakens stress concentration and deformation transmission, reducing the compressive stress on inter-level roof and floor pillars by 61.67% and roof tensile stress by 10%-15%. It effectively alleviates the unloading effect and improves the stability of stope-surface system. Field monitoring verifies the effectiveness of the adopted stability control measures, providing a scientific basis for safe mining of similar surface structure-stope coupling systems.

The distribution law of in-situ stress at working faces is crucial for the safe production of metal mines. Aiming at the complex problem of in-situ stress distribution at working faces under fault structure conditions in a spodumene ore body, the stress distribution law was explored by combining numerical simulation and field measurement. Based on measured data, the initial state of underground engineering under the real in-situ stress field was simulated using the rapid stress boundary method. The results show that stress concentration exists at the upper part of the hanging wall and the lower part of the footwall of each fault, and stress distribution at other positions presents obvious hierarchical characteristics. The inversion data of numerical simulation show that abrupt changes occur in stress contours on both sides of the faults. Specifically, the stress rises from 7 MPa to 20 MPa in areas with high fault density. The measured stress values are consistent with numerical simulation results in fault-free zones. It indicates that the rapid stress boundary method can accurately reflect the actual stress state of underground rock mass and help compensate for the limitation that measuring points must avoid adverse structural planes such as faults.