To address the frequent occurrence of surrounding rock instability such as roof fall and rib spalling in roadways with weak interlayers at a working face of Wengfu Phosphate Mine, field investigations and FLAC3D numerical simulations were conducted, and the surrounding rock control effects of conventional bolt-mesh-shotcrete support and the composite support technology of "bolt-mesh-shotcrete+steel arch+advanced support" were systematically compared. The research results show that the weak interlayer contains 3.9% kaolinite, and expansion and softening after water immersion are the main causes of surrounding rock deterioration. The vertical displacement of the roadway is reduced from -8.33 mm to -6.67 mm by the composite support technology, with a decrease of 19.93%. The horizontal displacements in X and Y directions are decreased from 2.61 mm and -1.64 mm to 1.27 mm and -0.58 mm, with reductions of 51.34% and 64.63% respectively. The stress concentration is remarkably alleviated, and the total number of failure units is lowered by 13.82%. It is shown by field application that the average monthly accident rate is reduced from 0.6 times to 0.27 times, with a decrease of 55.0%. The large deformation of surrounding rock under multi-thin interbed conditions is effectively controlled by the composite support technology. A theoretical basis and key technical paradigm are provided for the safe excavation of roadways containing dense weak interlayers.

The engineering disturbance in the process of mining often leads to a variety of fractures in the rock, which affects its mechanical properties. Using indoor uniaxial compression test, combined with XTDIC and acoustic emission monitoring system, the effects of fracture position, fracture dip angle and rock combination on the mechanical behavior and failure mode of layered composite rock samples with single fracture were investigated. The results show that with the increase of fracture dip angle, the peak stress of fracture rock samples gradually increases from 13.19-25.34 MPa to 33.89-42.99 MPa, the elastic modulus shows a non-monotonic change trend of increasing first and then decreasing, the deterioration coefficient gradually decreases from 0.32-0.66 to 0.06-0.20, and the strength deterioration effect of fracture on sandstone is more significant than that of limestone. The fracture position dominates the evolution of the failure mode. When the fracture is located in the upper part of the rock sample, the tensile failure is dominant. When the fracture is located in the middle of the rock sample, the failure mode is regulated by the fracture dip angle. Under the low fracture dip angle, the tensile failure is still dominant, and under the high fracture dip angle, it is transformed into tensile-shear mixed failure. The analysis of acoustic emission parameters (RA/AF value) shows that when the fracture dip angle is low, the priority cracks are mainly tensile cracks, accounting for 81.69%-88.41%. With the increase of fracture dip angle, the proportion of shear cracks increases, and the failure mode gradually transitions from tensile failure to multi-tensile and less-shear failure, and finally forms a tensile-shear mixed failure with tensile cracks accounting for 56.33%-62.73%.

To address the problem that low-energy acoustic emission (AE) signals account for an excessively high proportion during the fracture process of water-soaked coal-rock, high-energy critical information to be easily masked, resulting in insufficient accuracy in failure precursor identification, synchronous uniaxial compression and acoustic emission tests were conducted on coal-rock specimens under natural conditions and after 120 h of water-soaked. Based on the energy contribution rate and damage stress, a novel AE signal screening method was developed to extract dominant signals during coal-rock failure and investigate their precursory characteristics. The results show that water-soaked significantly decreases the uniaxial compressive strength of coal-rock. Compared with specimens under natural conditions, the average uniaxial compressive strength of water-soaked coal-rock decreases by 14.79%. Cracks exhibite multidirectional propagation behavior, thereby aggravating the final failure of the coal-rock. The cumulative AE count of coal-rock specimens under natural conditions reach 4×10⁶, while that of water-soaked specimens decrease to 2×10⁶. Moreover, the evolution of the cumulative AE count shows good agreement with the stress-strain behavior of the coal-rock. The AE dominant signal extraction method based on the energy contribution rate and damage-stress time window effectively identifies dominant signals during the failure process of water-soaked coal-rock. Compared with coal-rock under natural conditions, the extracted dominant signals exhibite lower energy thresholds but higher signal proportions. Furthermore, the extracted dominant signals show strong correspondence with the failure precursors of water-soaked coal-rock. These findings provide a theoretical basis for engineering stability prediction.

To study the adsorption and diffusion laws of CH₄ and CO₂ in the coal molecular structure model, the adsorption and diffusion processes of CH₄ and CO₂ in the coal molecular structure model under different temperatures, pressures and moisture contents were analyzed using the grand canonical ensemble Monte Carlo method and molecular dynamics simulation method. The results show that with increasing temperature or moisture content, the adsorption capacity of CH₄ and CO₂ decreases to a certain extent, while the adsorption and diffusion capacity of CO₂ is better than that of CH₄. With increasing pressure, the adsorption capacity of the two gases increases, but it tends to saturate, exhibiting typical saturated adsorption characteristics. The Langmuir fitting results show that the increase of temperature or moisture content leads to the decrease of adsorption parameters a and b, and the parameter value of CO₂ is always higher than that of CH₄, indicating that CO₂ and coal show higher adsorption affinity. The increase of temperature can significantly enhance the kinetic energy of the two gas molecules and reduce their local enrichment in the coal molecular structure model, thus improving the diffusion performance. The mean square displacement of CO₂ at each temperature is greater than that of CH₄, indicating that the diffusion ability of CO₂ is stronger and the spatial distribution is more uniform. With increasing moisture content, the average mass density of the two gases in the coal molecular structure model decreases. Although the presence of water molecules does not change the overall trend of the gas adsorption curve, the adsorption saturation point of the gas is significantly reduced. The high moisture content condition (6.55%) will significantly inhibit the gas diffusion behavior. Water molecules fill the pores and interact with coal molecules, resulting in restricted gas diffusion paths, weakening the degree of gas migration, and limiting the gas diffusion rate. The diffusion coefficients of CH₄ and CO₂ decrease to 0.063×10^-8 m²/s and 1.33×10^-8 m²/s, respectively. The research results provide a theoretical basis for improving coalbed methane recovery rate and optimizing CO₂ sequestration strategy.

As a key equipment in the underground coal mining transportation system, the scraper conveyor's bearings operate for extended periods under extreme conditions such as high temperature, high impact loads, dust pollution, and continuous operation, resulting in a high risk of failure. Existing fault diagnosis methods still face numerous challenges in feature extraction and classification accuracy when dealing with complex working conditions and multiple failure modes. To address this, a bearing fault diagnosis method based on a dual-branch convolutional neural network was proposed. Firstly, one-dimensional vibration signals were converted into two-dimensional time-frequency images using continuous wavelet transform to enhance the time-frequency feature representation capability of the input data. Subsequently, a dual-branch convolutional neural network architecture was constructed, in which one branch employed a dilated convolutional residual module to strengthen local feature extraction, while the other branch introduced a global attention mechanism to optimize global feature learning, and feature fusion was applied to improve the model's diagnostic capability. Finally, experimental validation was conducted on the CWRU and PU bearing datasets. The results show that the proposed method can achieve fault diagnosis accuracy rates of 99.78% and 97.94% on the two datasets, respectively, and outperforms comparison models such as DCRB-CNN, GAM-CNN and GAM-ResNet in overall performance, demonstrating significant diagnostic capability and strong generalization ability. Moreover, bearing fault diagnosis experiments were carried out using real-world coal mine field data, further validating the method's practicality and potential for engineering deployment under strong noise and non-stationary operating conditions. The research results provide an efficient and reliable deep learning solution for the intelligent fault diagnosis of complex coal mining equipment such as scraper conveyors.

To address the issues of low efficiency and insufficient accuracy in traditional mining equipment anomaly detection, a mining equipment anomaly detection model integrating bidirectional long short memory network and an improved subtraction average optimizer was proposed. Attention mechanism was introduced to achieve dynamic weighting of key features. Convolutional neural network was combined to extract local temporal features, and guided regularized random forest was employed to perform feature selection, thereby reducing data dimensionality and suppressing noise interference. The improved subtraction average optimizer introduced chaotic mapping initialization and golden-sine step-size adaptive adjustment strategy, which significantly enhanced the convergence speed and training stability of the model. The results show that on the mining equipment multisensor dataset and the mining machinery fault simulation dataset, the accuracy of the proposed model reaches 94.17% and 95.48%, respectively, and its comprehensive performance surpasses that of existing mainstream detection models. In ablation experiments, when all key modules are integrated, the accuracy, precision, recall, and F₁ score of the model reach 98.25%, 97.62%, 98.05% and 97.84%, respectively. In field tests on key equipment such as belt conveyors and crushers at a large open-pit mine, the detection accuracy of the model increases to 97.32%, with an average detection time of only 0.39 s. This indicates that the model can detect abnormal states of mining equipment efficiently and accurately, providing reliable technical support for ensuring mining production safety.

To address the problems that existing foreign object detection algorithms for underground mine conveyor belts mainly model cross-channel dependencies through channel dimensionality reduction, resulting in high computational overhead and limited detection accuracy, a foreign object detection method based on attention-scale fusion and cross-spatial learning, termed ES-YOLO, was proposed. Firstly, an Efficient Multi-Scale Attention (EMA) module was introduced into the backbone network. By reshaping channels into the batch dimension and grouping channel dimensions into multiple sub-features, spatial semantic features were evenly distributed across different feature groups. Subsequently, a Scale Sequence Feature Fusion (SSFF) module was introduced to improve the network's capability for multi-scale feature extraction. By horizontally stacking feature maps at different scales and applying 3D convolution to capture scale-sequence features, the module effectively integrated high-dimensional semantic information from deep feature maps with detailed information from shallow feature maps. Furthermore, a Dimension-Aware Selective Integration (DASI) module was introduced to adaptively assign weights to high-dimensional and low-dimensional feature maps, thereby improving the model's ability to detect small objects. Experimental results demonstrate that, compared with the baseline YOLOv8n model, the proposed method increases the mean average precision by 1.5 percentage points to 89.7%. The class-wise mean average precision for large ore blocks improve by 4.1 percentage points, while the model parameters are reduced by 18%. Compared with mainstream models, the proposed method has significant advantages in detection accuracy and lightweight, making it more suitable for real-time foreign object detection in underground conveyor belts in mines.

Aiming at the problems of insufficient accuracy and weak environmental adaptability of traditional coal quantity detection devices due to the dynamic change of coal volume and the harsh working conditions such as high dust and high water mist in coal mines, a coal quantity detection method based on LiDAR and visual image fusion for extreme working conditions was proposed. Firstly, the joint calibration of LiDAR and visual camera was carried out, the mapping relationship was established and the coordinate system was unified. Then, the sampling data was obtained by a specific strategy, and the coal region was identified by Poisson reconstruction algorithm, spatial filtering strategy and structural similarity index. Finally, the depth image completion network was used to repair the depth map, and the coal amount was calculated based on the repair results. The results show that the maximum detection error of the fusion method of LiDAR and visual camera is only 2.90%, and the error decreases further with the increase of coal volume. Compared with mainstream depth completion algorithms such as ENet, the proposed algorithm achieves the best core accuracy index and meets the industrial real-time requirements. This method is basically consistent with the fluctuation of the detection curve of the contact electronic belt scale, which verifies its reliability. This method effectively overcomes the limitations of a single sensor, significantly improves the accuracy of coal quantity detection, and can operate stably in harsh environments. It provides a new technical solution for the accurate detection of coal quantity of belt conveyor and holds great significance for promoting the intelligent and efficient production of coal mines.

Aiming at the problems of poor scheduling scheme, low utilization rate and high cost of unmanned trucks in large open-pit coal mines, an optimization method based on digital twin simulation was proposed, and a digital twin system framework for unmanned trucks was constructed. The digital twin simulation model was built by AnyLogic software to realize the real-time mapping of physical and virtual space. A mathematical model aiming at minimizing transportation cost, minimizing total transportation time and maximizing truck utilization was established, and a Quantum Enhanced Particle Swarm Optimization (QEPSO) algorithm was designed to solve the model. The verification results of an open-pit coal mine in Xinjiang show that compared with the traditional particle swarm optimization algorithm, the improved particle swarm optimization algorithm and the genetic algorithm, the maximum completion time of the QEPSO algorithm is reduced by 5.84%-11.34%, the total transportation cost is reduced by 5.12%-10.49%, and the truck utilization rate is increased by 4.32%-9.14%, with a faster convergence speed and better stability. The digital twin scheduling scheme shortens the task completion time by 73 min compared with the traditional scheme, effectively improves the transportation efficiency. The study provides a feasible technical path and method support for the intelligent scheduling of unmanned trucks in open-pit coal mines.

Production scheduling of underground metal mines is confronted with severe challenges caused by multi-source uncertainties, such as geological conditions, equipment status, and safety risks, while traditional deterministic scheduling models struggle to guarantee system stability and operational efficiency. To address this issue, a Monte Carlo simulation-based scheduling optimization model(MC-MOO) for underground metal mines was proposed. Firstly, probability distribution models were established for multi-source uncertainty factors. Subsequently, Monte Carlo scenario simulation was integrated with the NSGA-Ⅱ multi-objective optimization algorithm to develop a closed-loop framework of "probabilistic modeling-scenario simulation-multi-objective optimization", enabling the comprehensive optimization of scheduling cost, production stability, and equipment load balancing. The case study results show that compared with the traditional deterministic scheduling models, the total scheduling cost of the MC-MOO model is reduced by 5.97%, the average monthly output is increased by 16.53%, the standard deviation of output is reduced by 41.82%, and the equipment load balance is increased by 23.61%. It provides an effective modeling and solving paradigm for mine intelligent scheduling under uncertain environment.