Abstract:Underground coal mine fires generate smoke that spreads along roadways,and its flow state is easily affected by internal obstacles,forming local smoke stagnation zones,increasing the accumulation risk of toxic gases,and seriously threatening the safe evacuation of personnel and rescue decision-making.Existing studies mainly focus on the effects of roadway structural variations or ventilation systems,while systematic investigations on the influences of obstacle geometric shape,height,and distance are lacking.To address this issue,this study used the Particle Image Velocimetry(PIV)experimental method to investigate the effects of three key factors—obstacle shape,obstacle height,and the distance between the obstacle and the fire source-on the evolution of recirculation vortices and velocity distribution.The results showed that the fixed geometric leading edge of a square obstacle forced strong flow separation of smoke at sharp corners,forming structurally stable and large-scale recirculation vortices,with a maximum reverse smoke velocity of-0.027 m/s and a recirculation vortex height of 184 mm,both significantly higher than those of a circular obstacle.Therefore,compared with circular obstacles,square obstacles posed a higher risk due to their stronger smoke retention capacity and were therefore considered to require priority consideration in ventilation and evacuation design.An obstacle with a height of 100 mm mainly formed small-scale,high-velocity recirculation vortices,whereas a 200 mm-high obstacle caused the recirculation vortex height to expand to 233 mm,while the maximum reverse smoke velocity decreased to-0.015 m/s,forming large-scale,low-velocity recirculation vortices.Thus,although tall obstacles weakened vortex flow velocity,the large smoke stagnation zones they formed increased the space for toxic gas accumulation and the risk of personnel entrapment.A critical distance of 200-300 mm existed between the fire source and the obstacle,at which the velocity reached its peak and the vortex structure was the most stable.

Abstract:Coal-rock property identification during cutting mainly includes the recognition of process signals generated during cutting,infrared imaging,image feature recognition,and reflectance spectrum identification.Owing to the complex environment of roadway excavation,noise generated during the operation of a roadheader-bolter integrated machine easily overwhelms the acoustic signals produced by cutting,while vibration signals are also susceptible to interference from the machine's own vibrations.Methods based on infrared imaging and image recognition are easily affected by high dust concentration and low illumination during excavation,resulting in poor performance in practical working faces.Meanwhile,identification methods based on a single sensor are limited by environmental complexity and the identification range.To address these issues,a multi-source information fusion identification method based on acoustic and vibration signals was proposed.First,considering the complex underground environment with strong noise interference,the Variational Mode Decomposition(VMD)algorithm was used to decompose acoustic and vibration signals,and a Shannon entropy-based mode selection criterion was proposed to reconstruct and denoise the signals,thereby obtaining effective acoustic and vibration signals.Then,a Markov Transition Field(MTF)was employed to transform the acoustic and vibration signal feature sequences into two-dimensional images for feature fusion.Finally,a Convolutional Neural Network(CNN)combined with a Convolutional Block Attention Module(CBAM)was introduced to perform spatial feature learning on the two-dimensional images,enabling automatic extraction of multi-scale features and precise enhancement of key features,and thus achieving accurate identification of coal-rock properties during cutting.Experimental results showed that the proposed method achieved an identification accuracy of 99.404 8％,which was significantly higher than that of traditional CNN models.

Abstract:The trajectory tracking control accuracy of coal mine anchor bolt drilling booms has an important impact on the efficiency and safety of support operations.To address the insufficient trajectory tracking accuracy caused by the nonlinearity,time delay,and time-varying parameters of the hydraulic system of underground coal mine anchor bolt drilling booms,an Iterative-Learning-Compensation-Based Sliding-Mode Active Disturbance Rejection Control(ILC-SMADRC)algorithm was proposed.A six-degree-of-freedom kinematic model of the anchor bolt drilling boom was established based on the D-H parameter method,and smooth trajectories were planned using quintic polynomial interpolation.A high-order Extended State Observer(ESO)was designed to achieve accurate estimation of system states and disturbances.A novel sliding mode reaching law constructed based on the Rayleigh distribution function and the tanh function was proposed to effectively suppress high-frequency chattering.Iterative learning compensation was incorporated to improve trajectory tracking accuracy and robustness,and system stability was proven using the Lyapunov criterion.Simulation results showed that,compared with conventional PID and Sliding Mode Control(SMC)methods,the cumulative absolute error of the ILC-SMADRC algorithm was reduced by 83.5％and 59.2％,respectively,with the error of joint 6 reduced by 96.8％compared with SMC,significantly enhancing the motion control accuracy and stability of the anchor bolt drilling boom under complex underground operating conditions.

Abstract:In response to persistent dense noise anomalies,instantaneous impulse anomalies,and missing anomalies in sensor data caused by the complex underground environment of coal mines,existing data anomaly detection methods have difficulty adapting to nonlinear time-series fluctuations,exhibit high false alarm rates when processing high-frequency data,and rely heavily on large amounts of labeled samples.To address these issues,a coal mine sensor data anomaly detection method was proposed.First,Z-score normalization was used to eliminate dimensional differences in sensor data.Second,proximity-based anomaly detection methods—the distance-based K-Nearest Neighbors(KNN)algorithm and the density-based Local Outlier Factor(LOF)algorithm-were used to perform preliminary anomaly screening and assign anomaly labels.Meanwhile,temporal features including lag features,statistical features,differential features,Fast Fourier Transform(FFT)features,and time-encoding features were extracted using dual-scale sliding windows and concatenated to form a feature matrix.Then,the feature matrix and corresponding anomaly labels were used to construct the sample set required for the eXtreme Gradient Boosting(XGBoost)model.The sample set was divided into training,validation,and test sets according to the temporal order of the data,and the XGBoost model was trained using the training set after negative-sample undersampling.Finally,the trained XGBoost model was used to compute the anomaly probability of each sample in the validation set,and a Precision-Recall(PR)curve was plotted.The anomaly probability that maximized the F1 score was selected as the anomaly decision threshold,and sample points with anomaly probabilities greater than or equal to the threshold were labeled as anomalies,thereby outputting sensor anomaly data.Experimental results show that the proposed method has high anomaly detection accuracy and can maintain stable detection performance under different data distributions and noise environments,demonstrating good generalization capability.

Abstract:To address the degradation in diagnostic performance caused by missing sample labels of key components-such as bearings and gearboxes-in mine hoisting machines under complex operating conditions,a fault diagnosis technology for critical hoisting machine components based on multi-scale feature transfer learning was proposed.A Lightweight Fault Diagnosis Model Based on Domain Adversarial and Multi-Scale Time-Frequency Feature Extraction(DAMSF-LFDM)was constructed.A Serpentiform Wavelet Coefficient Matrices(SWCMs)representation was proposed.By combining wavelet packet transform,piecewise aggregate approximation,and serpentiform reorganization,a multi-scale time-frequency feature matrix was constructed to fully capture the internal correlation characteristics of vibration signals across different frequency bands.A Multi-Scale Residual Ghost Convolution Block(MRGCB)was proposed.It employed multiple parallel convolutional layers to effectively extract deep features of the input data at different scales,thereby strengthening the model's ability to capture multi-scale information.To extract personalized fault features from SWCMs and perform adaptive fusion,a Fused Multi-Scale Fault Feature Extraction Module(FMFFEM)was introduced.Feature fusion was carried out via summation,and an adaptive feature-weight allocation mechanism was incorporated to complete the fused extraction of features from different frequency bands.By integrating multi-level maximum mean discrepancy loss with a domain adversarial mechanism,a deep transfer diagnosis network based on the domain adversarial mechanism was established,improving the model's adaptability across operating conditions.Experimental results demonstrated that the DAMSF-LFDM model significantly outperformed the comparative models overall,achieving the highest fault diagnosis accuracy across different transfer tasks.The average cross-condition accuracies on the SEU dataset and the MFS-RDS dataset reached 98.67％and 99.80％,respectively.

Abstract:Existing control methods for negative pressure of gas extraction boreholes exhibit delayed responses to changing operating conditions and lack adaptive capability and dynamic feedback control,making it difficult to achieve precise control of borehole negative pressure.To address these problems,an intelligent control method for negative pressure in gas extraction boreholes based on Particle Swarm Optimization(PSO)and Back Propagation(BP)was proposed.A coal seam gas-air migration model was derived,and on this basis,COMSOL numerical simulation software was used to obtain gas extraction datasets under different extraction conditions.A PSO algorithm was introduced to optimize the initial weights of the BP algorithm,improving the reliability of negative pressure prediction for gas extraction.Taking gas extraction flow rate or gas extraction volume fraction as the target value,the PSO-BP algorithm predicted the corresponding extraction negative pressure,and the valve opening was adjusted to make the borehole negative pressure reach the predicted value,thereby achieving precise control of gas extraction boreholes.The results showed that,compared with Extreme Learning Machine(ELM),Temporal Convolutional Network(TCN),and Support Vector Machine(SVM)algorithms,the BP algorithm more accurately captured the variation patterns of gas extraction data characteristics.The PSO-BP algorithm achieved better performance than the BP algorithm in terms of Mean Squared Error(MSE),Root Mean Squared Error(RMSE),Mean Absolute Error(MAE),Mean Bias Error(MBE),Mean Absolute Percentage Error(MAPE),and Coefficient of Determination(R2).After on-site implementation of intelligent borehole negative pressure control,both gas extraction volume fraction and gas extraction flow rate increased compared with those before implementation.

Abstract:Fault classification methods for gas drainage pumps based on single-component features fail to consider the interactions among different components,making it difficult to accurately capture the essential characteristics of faults and thereby limiting their accuracy and reliability in gas drainage pump fault classification tasks.To address this issue,a gas drainage pump fault classification model based on fused multi-component joint features using Graph Sampling and Aggregation with a Hierarchical Attention Mechanism(GraphSAGE-HAT)was proposed.First,vibration data from the bearings,impeller,and pump casing of the gas drainage pump were collected using acceleration sensors to construct a dataset containing features from three components,and the dataset was preprocessed using the Min-Max normalization method.Second,the preprocessed dataset was transformed into a homogeneous graph using the K-Nearest Neighbor(KNN)algorithm to facilitate graph feature learning by graph neural network algorithms.Then,a hierarchical attention mechanism(HAT)was introduced to construct the GraphSAGE-HAT algorithm,in which HAT performed hierarchical weighted aggregation of intra-node component features as well as node and neighboring-node features in the homogeneous graph,effectively capturing inter-component correlation features and inter-node data structural characteristics.Finally,the aggregated features were fed into a fully connected layer to achieve gas drainage pump fault classification.Experimental results showed that,compared with single-component features,the model combining fused multi-component joint features with the GraphSAGE algorithm improved classification accuracy by 9.24％.With the further introduction of HAT,the model based on fused multi-component joint features and the GraphSAGE-HAT algorithm achieved an accuracy of 98.08％.

Abstract:With increasing mining depth and increasingly complex geological conditions,existing studies on the distribution of advanced abutment pressure in working faces are difficult to be fully applicable to hard-roof conditions.To further reveal the distribution and evolution characteristics of advanced abutment pressure in hard-roof working faces,taking the 8105 working face of Majiliang Coal Mine of Datong Coal Mine Group Co.,Ltd.as the engineering background,theoretical analysis,numerical simulation,and field measurements were comprehensively adopted,and the influences of different working face advance distances,mining speeds,and mining thickness on the distribution of advanced abutment pressure were analyzed.The results showed that the advanced abutment pressure in hard-roof working faces exhibited obvious zoning characteristics,which were successively divided,from near to far from the working face,into a pressure relief zone,a stress concentration zone,a secondary pressure relief zone,and an in-situ stress zone.With the increase in working face advance distance,the cantilever length of the main roof increased,the peak value of advanced abutment pressure in the stress concentration zone increased,the valley value of advanced abutment pressure in the secondary pressure relief zone decreased,and the range of the secondary pressure relief zone expanded.When the main roof fractured,a roof"rebound-retraction"effect occurred,causing the peak value of advanced abutment pressure to decrease,the valley value to increase,and the range of the secondary pressure relief zone to shrink.Increasing the mining speed or the mining thickness both led to an increase in the peak value of advanced abutment pressure,an expansion of the secondary pressure relief zone,and an increase in the pressure relief magnitude,thereby making the zoning characteristics of advanced abutment pressure in the working face more pronounced.

Abstract:Under severe strata pressure,the output signals of hydraulic support pin shaft sensors exhibit strong nonstationarity and time-varying characteristics.A single neural network architecture is unable to simultaneously account for multiscale spatial feature extraction and long-term temporal dependency modeling,and it lacks an adaptive weight allocation mechanism during multiscale feature fusion,which limits the generalization performance of error prediction models.To address these issues,a hybrid neural network CNN-LSTM-SAtt that integrated a Convolutional Neural Network(CNN),a Long Short-Term Memory Network(LSTM),and a self-attention mechanism(SAtt)was proposed and was applied to load error prediction of hydraulic support pin shaft sensors.First,a combined method of Variational Mode Decomposition(VMD),Fast Fourier Transform(FFT),and Hilbert Transform(HT)(VMD-FFT-HT)was adopted to construct multidomain features.Then,CNN was used to extract deep spatial morphological features in the frequency domain and time-frequency domain,while LSTM was employed to capture the long-term temporal evolution patterns of time-domain signals.Finally,SAtt was introduced to dynamically assign weights to multidomain features according to signal fluctuation characteristics,thereby establishing a high-precision nonlinear mapping between the sensor load response signal and the excitation signal.The results of five typical loading experiments of pin shaft sensors conducted using a force standard machine indicated that the predicted values of the CNN-LSTM-SAtt model can effectively correct the error components in the load response signals of pin shaft sensors.Compared with traditional models and single neural network models,this model exhibits significant advantages in both prediction accuracy and generalization capability,enabling effective prediction of load errors of hydraulic support pin shaft sensors under complex working conditions.

Abstract:Under uniaxial compression,the strength,deformation,and energy evolution characteristics of bedded sandstone depend not only on the inherent properties of the sandstone matrix but are also strongly controlled by factors such as bedding dip angles,bonding strength of bedding planes,and bedding density,exhibiting pronounced anisotropic behavior.At present,systematic studies on the mechanical properties and energy dissipation characteristics of bedded sandstone during uniaxial compression have mainly focused on a single bedding angle or a single energy parameter,lacking a synergistic analysis of energy parameters and mechanical parameters,as well as failure types and characteristics,under different bedding dip angles.To address this issue,bedded sandstone was taken as the research object,and uniaxial compression tests were conducted to systematically obtain stress-strain curves of bedded sandstone with five bedding dip angles of 0,30,45,60,and 90°.The mechanical properties,including uniaxial compressive strength,elastic modulus,and Poisson's ratio,were analyzed,the dominant failure types and characteristics were observed,and the evolution laws of input energy,elastic energy,and dissipated energy were calculated based on the principle of energy conservation.The results showed that the bedding dip angle significantly influenced the mechanical properties of sandstone.The peak strength and elastic modulus exhibited a U-shaped variation with bedding dip angle,reaching the minimum at 45° and the maximum at 90°.Poisson's ratio showed a single-peak pattern,attaining the maximum value of 0.32 at 45°.The peak strain reached the maximum at 45° and the minimum at 90°.The failure types and characteristics evolved with bedding dip angle from through-bedding failure to along-bedding shear combined with local through-bedding failure,and finally to along-bedding shear slip.Specimens with bedding dip angles of 0° and 90° were dominated by through-bedding brittle failure,whereas those with 45° and 60° were dominated by along-bedding shear plastic failure.The input energy,elastic energy,and dissipated energy were all regulated by bedding dip angle.Specimens with bedding dip angles of 0° and 90° showed a higher proportion of elastic energy,which was released intensively at failure,while specimens with bedding dip angles of 30-60° exhibited a higher proportion of dissipated energy with a continuously increasing trend.

Abstract:To address the problems of poor spatial continuity and difficulty in achieving real-time dynamic feedback in traditional evaluation methods for coal mine roof hydraulic fracturing-such as borehole observation,microseismic monitoring,acoustic emission,or roadway deformation detection-the high-density electrical method combined with electrical resistivity tomography is innovatively adopted to realize dynamic,quantitative,and three-dimensional visualization monitoring of fracture development and fracturing fluid migration during the fracturing process.Taking the 112205 working face of No.1 Coal Mine of Shaanxi Xiaobaodang Mining Co.,Ltd.as the engineering background,continuous apparent resistivity surveys were carried out before,during,and after the staged fracturing of the XBD-02L horizontal well.Based on resistivity variations,a quantitative evaluation index ρSL for fracturing effectiveness was proposed.The detection results showed that the effective influence range of fracturing reached up to 200 m in the horizontal direction and 57 m in the vertical direction,and the fracturing fluid exhibited a dynamic diffusion-loss process.Multi-source verification of the high-density electrical results was conducted by integrating geophysical and engineering data:audio-frequency electrical penetration imaging demonstrated that resistivity anomaly zones were highly consistent with water-rich structures,and mine pressure monitoring data indicated that the periodic weighting interval in the fractured area was reduced by 39.35％,showing a significant mine pressure mitigation effect.Based on high-density electrical detection,a dynamic visualization monitoring and quantitative evaluation system for the entire process of coal mine roof hydraulic fracturing is established,which provides key technical support for optimizing hydraulic fracturing design and ensuring safe and efficient coal mining.

Abstract:Manual ash content detection of flotation tailings in coal preparation plants has a low degree of automation and cannot meet the requirements of online,rapid,and accurate detection.Applying a robotic manipulator to flotation tailings ash content detection improves detection efficiency and safety.To address the problems of motion stuttering and insufficient compliance of the robotic manipulator during operation,an improved Task-Joint Space Dynamic Adaptive Compliance Control(TJS-DACC)algorithm was proposed.In this algorithm,a reinforcement learning framework was introduced into TJS-DACC,and the response speed and acceleration of the manipulator end effector were comprehensively considered to construct a multi-objective fused reward function.Meanwhile,penalty and loss functions were designed,and an optimization model for the interpolation weight factor"α"was established to achieve adaptive fusion of task-space and joint-space control of the manipulator.Matlab simulation experiments and physical platform experiments were conducted to verify that,when controlled by the improved TJS-DACC algorithm,the sampling efficiency of the flotation tailings ash content detection manipulator increased by 26.13％and 15.03％,respectively,compared with those under the traditional joint-space PID algorithm and the TJS-DACC algorithm.Moreover,the trajectory was continuous and smooth,joint coordination was strong,and no emergency stops,stuttering,or impact phenomena occurred,indicating that the control performance is superior to that of the comparison algorithms.

Abstract:The support load of hydraulic supports in fully mechanized mining faces of shallow-buried coal seams exhibits strong spatial non-stationarity and regional differences.At present,most studies only predict the load of a single support,and the experimental design ignores the correlations with adjacent supports,making it difficult to accurately predict support loads in different regions of the working face.To address this issue,a hydraulic support load zoning prediction method based on spatiotemporal feature fusion was proposed.The distribution characteristics of roof deflection and support load in the working face were analyzed based on elastic foundation beam theory.Combined with field-measured support load data,the K-means++clustering algorithm was used to divide hydraulic supports into regions and to verify the rationality of the obtained zoning.Through feature engineering,the load data of six adjacent supports were introduced as exogenous variables into an XGBoost model.By integrating the advantages of LSTM in capturing long-term dependencies and nonlinear responses,an XGBoost-LSTM combined prediction model was constructed.By optimizing the input window length and output step size parameters,accurate prediction of representative support loads in each region was achieved.The results showed that the support load exhibited a distribution trend of"larger in the middle and smaller at both ends",and the supports were divided into five typical regions.When the window length was 60 and the output step size was 6,the model achieved the best performance in multi-step prediction tasks.The coefficient of determination of the prediction results in each region increased to 0.947,while the root mean square error(RMSE)and mean absolute error(MAE)decreased to 0.303 MPa and 0.152 MPa,respectively.Compared with the suboptimal XGBoost model,RMSE and MAE were reduced by 44.4％and 35.4％,respectively,and the prediction accuracy was higher during critical stages such as periodic weighting.

Abstract:At present,smart mine construction generally faces problems such as imbalance between investment and output,lack of standards,low accuracy of core disaster early warning systems,insufficient technology integration,system silos,and insufficient utilization of data resources,which seriously restrict the in-depth application of intelligent technologies in mine safety prevention and control,production optimization,and operational decision-making.This study introduces the theoretical development of smart mine construction and the current status of technology integration and data governance,and points out that although smart mine construction has continuously evolved from digital mines and perceptual mines to intelligent mines and has achieved significant progress in technical architecture and system applications,it is still at the stage of system intelligence and has not yet realized truly comprehensive intelligence.Existing systems have prominent problems,including difficulties in comprehensive perception,deep interconnection,business collaboration,and data sharing,as well as limitations in intelligent decision-making.Based on an analysis of the core problems currently existing in smart mine construction in terms of technology integration,data governance,and ecosystem building,this study highlights the urgent need to promote smart mine upgrading and transformation through optimized technology integration and data governance,and proposes a three-stage upgrade pathway for smart mine construction,consisting of system integration and intelligent safety system development,full-process intelligent enhancement covering coal mining,tunneling,electromechanical systems,transportation,ventilation,and operation management,and the third stage focusing on the construction of large mining models,ecosystem collaboration,and emergent intelligence.

Abstract:Current studies on surrounding rock control in dynamic pressure roadways mainly focus on roadways under single-or two-directional free-surface conditions,while relatively few studies have applied the combined control technology of roof-cutting pressure relief and full anchor-cable support to dynamic pressure roadways with three-directional free surfaces.Taking the three-directional free-surface dynamic pressure roadway of the Ji15-23100 working face in Pingmei No.4 Mine as the engineering background,the surrounding rock stress distribution characteristics before and after roof-cutting pressure relief were analyzed.The results showed that before excavation,the plastic zone of the three-directional free-surface dynamic pressure roadway had already exceeded the anchorage range of the original support system,and roof cutting could improve the surrounding rock stress environment and reduce the peak supporting stress.The original roadway support scheme suffered from insufficient support length and low pretension force,failing to effectively anchor the plastic zone,which resulted in large roof subsidence and severe sidewall damage in the three-directional free-surface dynamic pressure roadway.Based on the above analysis,a coordinated control technology of"roof-cutting pressure relief+full anchor-cable support"was proposed.High-strength anchor cables were used to form a superimposed beam structure to transfer roof loads to deep,stable rock strata,while roof-cutting blasting was employed to weaken key roof strata and interrupt stress transfer paths.Field monitoring results indicated that after implementing the proposed coordinated control technology,the maximum roof subsidence was reduced to 184 mm,and the maximum deformation of the coal pillar rib and solid coal rib reached 149 mm and 122 mm,respectively.These results verify the effectiveness of the proposed technology in controlling the surrounding rock of three-directional free-surface dynamic pressure roadways.

Abstract:In response to the problems of excessive noise,large drift,and numerous missing values in the monitoring data under complex underground working conditions of mining drilling machines,a bidirectional temporal convolutional joint generative adversarial interpolation network with embedded spatio-temporal attention(BiTCGAIN-STA)was designed.The bidirectional temporal convolutional network(BiTCN)is used to capture the temporal dependency between previous and subsequent time sequences,and the spatio-temporal attention(STA)mechanism is employed to adaptively allocate time and channel weights.Through generative adversarial training,the distribution consistency and diversity of the interpolated samples are improved.At the same time,real data fine-tuning is performed on the target domain to enhance the transfer robustness.A bearing fault diagnosis model based on adaptive weighted fusion and the Informer network is proposed.The Informer long sequence feature extraction network is used to deeply represent the fused signals,thereby improving the ability to identify weak fault features.Experimental results show that,under different missing rates,the root mean square error(RMSE)of the BiTCGAIN-STA model is significantly higher than that of mainstream models such as Mean,MICE,and GAIN,achieving high-quality data reconstruction.The bearing fault diagnosis model has an identification accuracy of 99.87％for weak faults,significantly higher than models such as Transformer and graph neural network(GNN).

Abstract:Existing dynamic point cloud removal methods are mostly designed for structured scenarios.When applied to unstructured coal mine environments characterized by non-uniform point cloud distribution,large spatial structural differences,and large-scale dynamic occlusions,these methods tend to cause ground truncation and local holes in the global map,thereby damaging map continuity.To address these issues,an adaptive cascaded dynamic point cloud removal method for unstructured coal mine scenarios was proposed.The proposed method first converted three-dimensional point clouds into the two-dimensional image domain using multi-resolution depth projection and rapidly generated prior labels of dynamic regions by constructing depth consistency constraints;then,the space was discretized using a ring-sector strategy,and the decision threshold of the Scan Ratio Test(SRT)was adaptively adjusted based on the prior information of dynamic regions to determine whether dynamic point clouds existed within each grid cell;subsequently,secondary verification was performed on grid cells containing dynamic point clouds by incorporating a height lower-bound check,and Regional Ground Plane Fitting(R-GPF)was applied to recover ground point clouds missing due to dynamic occlusion;finally,an accurate and continuous static point cloud map was obtained.Experimental results on public datasets and simulated coal mine scenarios showed that the proposed method achieved good mapping consistency,effectively removed dynamic point clouds while preserving the integrity of vertical geometric structures and the continuity of ground surfaces in dynamic neighborhoods,effectively suppressed false ghosting caused by mapping pose drift,avoided ground holes and wall fractures,and maintained the continuity of static point cloud structures.The Static Accuracy(SA)of the proposed method was higher than that of the ERASOR method,the Dynamic Accuracy(DA)was higher than that of the Removert method,and both the Average Accuracy(AA)and Harmonic Accuracy(HA)were optimal,effectively balancing dynamic point removal accuracy and static point fidelity.

Abstract:Existing vision-based drill rod counting methods for underground coal mines mostly rely on single or local visual features.When state transitions such as drilling advance,drilling retreat,and short pauses occur during drilling operations,accompanied by common underground interferences including complex illumination,target occlusion,and target jitter,the counting features may exhibit abnormal fluctuations,leading to miscounting or missed counts.To address these issues,a drill rod counting method based on underground coal mine drilling rig operation state identification was proposed.First,a lightweight object detection model was constructed by introducing a Large Separable Kernel Attention(LSKA)module,a Generalized Efficient Layer Aggregation Network(GELAN),and a DynamicHead into the YOLOv11 model to stably detect key components of the drilling rig(the chuck and the gripper).Then,the DeepSORT algorithm was used to continuously track the chuck and the gripper,obtain the relative distance variation curve between them,and smooth the curve using Kalman filtering to suppress noise.Finally,trough positions were extracted from the smoothed curve;drilling advance and drilling retreat states were identified according to the spacing between adjacent troughs,and troughs under the drilling advance state were accumulated to achieve drill rod counting.Experimental results showed that the improved YOLOv11 model achieved an mAP@0.5 of 0.966 and accurately detected targets under complex conditions such as complex illumination,target occlusion,and target jitter.The proposed method achieved an average accuracy of 97.97％,meeting the accuracy requirements for automatic drill rod counting in underground environments.

Abstract:With the increasing complexity of coal spontaneous combustion prediction scenarios and the growth of data volume,existing coal spontaneous combustion temperature prediction methods based on neural network models are unable to accurately capture key disaster-inducing precursors under specific scenarios,resulting in a significant decline in prediction accuracy and reliability in practical applications,which fails to meet the requirements of coal spontaneous combustion prediction under the development of intelligent coal mines.To address these issues,by investigating the variation patterns of multiple characteristic indicators of coal spontaneous combustion under ventilation influence,a high-temperature point prediction model for coal spontaneous combustion based on a Particle Swarm Optimization(PSO)-optimized BP neural network(PSO-BP)was proposed.Five factors,including the concentrations of O2,CO,CO2,CH4,and the height from the air inlet,were selected as the input parameters of the model,and temperature values were taken as the prediction output.Experimental results showed that as coal temperature increased,the oxidation rate of coal samples accelerated,and the spontaneous combustion process of coal exhibited a three-stage evolution characteristic.The O2 concentration showed a stage-wise decreasing trend,while the concentrations of CO,CO2,and CH4 exhibited exponential variations with ventilation time.The high-temperature zone dynamically migrated toward the air inlet side along the direction opposite to the airflow.After parameter optimization,the Mean Absolute Error(MAE)of the PSO-BP model was reduced by 1.2％compared with that before optimization,the Mean Bias Error(MBE)was reduced by 58.5％,and the Root Mean Square Error(RMSE)was reduced by 18.6％.Data from the goaf on the intake and return air sides of the No.1 East coal seam in a mine of the Baojishan mining area of Gansu Jingmei Energy Company Ltd.were further selected to validate the model.The results showed that the PSO-BP model achieved the minimum error indices across different coal seams,and the fluctuation ranges of its error indices were smaller than those of the BP neural network model across different coal seams.

Abstract:Existing underground coal mine pipeline installation equipment mostly adopts manual remote control operation,with inflexible movements and low pipeline handling efficiency.To address this problem,a pipeline installation robot based on binocular vision and grasp-release trajectory planning was designed to achieve automatic pipeline recognition and grasping and placing.The kinematic model of the manipulator of the pipeline installation robot was established and solved using the DH method,and the transformation matrices of the binocular camera and the gripper relative to the base coordinate system of the robot manipulator were derived.The hue,saturation,and brightness ranges of the pipelines to be recognized were determined through experiments.A simulated underground coal mine tunneling working-face environment was constructed,and experimental tests of the pipeline installation robot were carried out.The results showed that the robot realized automatic recognition,grasping,and installation of underground pipelines.The maximum deviation between the measured and theoretical values of the gripper coordinates was within 40 mm,and the maximum deviations of yaw,pitch,and roll angles were within 1°,meeting the pipeline positioning accuracy requirements.Moreover,the time required for pipeline grasping,lifting,and installation was greatly shortened compared with the manual method,and the installation efficiency was significantly improved.