Abstract:Aiming at the unfavorable environmental factors such as low brightness of artificial light source and dense distribution of coal gangue in coal mine underground,which lead to the problems of feature extraction difficulty and low localization accuracy of existing coal gangue detection algorithms,a real-time segmentation method of coal gangue based on adaptive low illumination image enhancement is proposed,which consists of CG-IENet(Coal Gangue Image Enhancement Network)and CG-TSNet(Coal Gangue Target Segmentation Network).The CG-IENet is based on the CycleGAN network model framework,which optimizes the design of the encoder,feature extrac-tion network and decoder in the generator network through the attention mechanism,and adopts PatchGAN as the discriminator network to achieve adaptive picture quality enhancement of low illumination coal gangue images,so as to better retain the surface texture information of the coal gangue.The CG-TSNet is based on the DeepLabV3+network model framework,which introduces the lightweight backbone network MobileNetV2 and the neck network SPS-ASPP module in the encoder,and combines the feature fusion module,the attention mechanism,and the dynamic sampling in the decoder,which improves the model segmentation accuracy and reduces the model complex-ity effectively at the same time,it enables faster processing speeds and lower energy consumption for easy deployment of edge-based com-puting devices.The experimental results show that CG-IENet improves 47.86%,61.20%,14.30%,31.77%and 23.68%on average in the peak signal-to-noise ratio,structural similarity index,information fidelity criterion,entropy and gray mean,respectively,compared with the other low illumination image enhancement models,and the mean chromaticity error is as low as 1.18.The three-dimensional gray scale distribution graph and gray scale histogram are used to visually compare and analyze the various models,which finally shows that CG-IENet has the best enhancement effect,which can better improve the brightness of the image to avoid color distortion while retaining the detailed information of the image.CG-TSNet has the highest mean intersection over union,mean pixel accuracy,pixel accuracy,and F1-Score(F1)metrics,91.76%,95.84%,96.85%,and 95.04%,respectively,and the best overall detection performance when compared to the other four semantic segmentation models,such as HRNet,UNet,and PSPNet,on the basis of the guaranteed model memory of 4.567 MB.The model can be adapted to the complex working conditions in underground coal mines,thus realizing accurate and efficient sorting of coal gangue.

Abstract:To address the high deployment cost and limited disaster resilience of downhole wireless positioning systems,an active down-hole target positioning method is proposed based on an improved ConvNeXt network and panoramic vision technology.The method con-sists of two stages:Offline training and online positioning.In the offline training stage,the downhole environment is divided into regions,and panoramic image data are collected to construct a multi-scale training dataset through data augmentation.Based on the improved Con-vNeXt network,an area positioning model and a fine positioning model are trained separately,where the area model provides coarse-grained position estimation and the fine model achieves precise localization.During the online positioning stage,a panoramic camera mounted on a mobile device continuously captures environmental images.After image stitching and enhancement preprocessing,the area positioning model first determines the target's regional range,and the corresponding fine positioning model is then invoked for real-time localization.Experimental results demonstrate that the combination of 0.5 m area positioning and 0.1 m fine positioning achieves an accur-acy of 94%.With the integration of the CoordAttention module,the accuracy further increases to 95%,representing a 14.67%improve-ment over conventional single-step methods.Compared with MobileNetV3(91%)and ResNet(92.33%),the improved ConvNeXt net-work exhibits superior performance.Furthermore,the proposed method maintains strong robustness against typical downhole disturbances such as illumination variation,motion blur,personnel movement,and dust interference,achieving 95.33%accuracy on the 616 roadway dataset collected from the Yangchangwan Coal Mine of the National Energy Group Ningxia Coal Industry Company Limited.The study provides a low-cost,robust,and easily deployable positioning solution for intelligent downhole mobile platforms such as autonomous vehicles and robots.

Abstract:As the coal industry accelerates its transition towards intelligent transformation,video image-based object detection and hazard identification technologies,with advantages of non-contact,visualization,and high real-time performance,are increasingly becoming a key support for enhancing mine safety and production efficiency.This study survey focuses on the innovation and application requirements of video image intelligent analysis technologies in the context of the coal industry's intelligent transformation.It conducts a systematic study on the current state of intelligent construction in complex mining environments,potential safety hazards,existing advanced target detec-tion and hazard identification methods,as well as technical challenges,supported by practical application cases.The study first examines the current state of construction of mine video surveillance systems,the diversity and complexity of safety hazards,and elucidates the ne-cessity and application value of machine vision technology in underground environments;Subsequently,it compares and analyzes the dif-ferences and application value between traditional image processing methods and deep learning-based object detection and recognition technologies in terms of detection accuracy,robustness,and computational efficiency.The study focuses on exploring the adaptability and limitations of mainstream detection models in severe underground mine scenarios such as low illumination,high dust and fog,noise inter-ference,and occlusion.On this basis,combined with the construction and application cases of existing mine video datasets,the role of datasets in model training,performance evaluation,and generalization capability enhancement is analyzed,and the deployment perform-ance of these methods in typical tasks such as coal flow transportation monitoring,violation behavior recognition,operation equipment hazard detection and safety warning in fully mechanized mining faces is further discussed.Finally,the survey summarizes the challenges that current technologies still face,including limited adaptability to complex environments,data imbalance and annotation difficulties,con-straints of real-time computing resources,and instability in long-term behavior tracking.Future research directions are outlined,including generalization learning,multi-modal complementary fusion,large-scale semantic modeling,edge-cloud collaborative computing,and data privacy and security protection.Through systematic review and in-depth analysis,this study provides structured theoretical references and technical route support for the continuous innovation and practical implementation of mine video image intelligent analysis technology,thereby enhancing coal mine safety monitoring,hazard warning,and emergency response systems toward higher levels of intelligence and intrinsic safety.

Abstract:Thermodynamic disasters(fires and explosions)are the most serious disaster in coal mine,so it is necessary to establish a fire characteristic parameter monitoring system for monitoring.At present,the goaf wireless communi-cation monitoring system has the prob-lem that the communication equipment is limited due to the goaf collapse,and the fire characteristic parameters cannot be monitored stably.In order to improve the stability of the wireless communication monitoring system,the reasons for the limited wireless penetration communication in the goaf environment are analyzed,and the antenna performance of the communication equipment is confirmed to be an important factor of the shadow communication performance after the goaf collapse.Three key technologies for improving the stability of wireless penetration communication are further studied:antenna environment adaptation,antenna performance detection and automatic op-timization of communication parameters.In order to improve the adaptability of the antenna environment,a reconfigurable antenna is de-signed to ensure that the adaptive antenna can maintain high radiation efficiency under different environments such as unburied and buried.A performance detection method based on return loss detection is applied,which can directly measure and evaluate the antenna perform-ance at fixed transmission frequency.The automatic optimization of communication parameters can be carried out according to the work-ing environment,such as adjusting the transmission power and spread spectrum factor of LoRa,so as to balance the power consumption and performance of the communication system.Finally,a set of communication equipment was designed with optimized antenna to en-hance the stability of wireless penetration communication.Ground and underground tests in coal mines were conducted on the communica-tion equipment with and without antenna optimization.The ground test results show that the performance of the optimized wireless com-munication equipment is superior to that of the unoptimized equipment in different burial medium test environments,and it can meet the stable communication requirement within 10 m.The actual underground test results indicate that this research technology can effectively adapt to the situation where the monitoring of fire characteristic parameters in the goaf is restricted after the goaf collapses.It can signific-antly enhance the communication signal strength and achieve stable communication within 10 m even when buried in the actual goaf,ulti-mately achieving stable monitoring of fire characteristic parameters in the goaf.

Abstract:Coal mine safety production video analysis and recognition technology is a core technical support for ensuring the intelligent construction of coal mines and the high-quality development of the coal industry in China.To effectively address the impact of complex underground environments such as low illumination,high dust,and non-uniform lighting on the quality of video surveillance images,and to improve the real-time performance and accuracy of safety hazard identification,image enhancement technology has become a key link in the process of coal mine video AI recognition.This paper systematically elaborates on the urgent needs and development status of im-age enhancement technology in the context of intelligent coal mine construction,analyzes the multi-factor coupling causes of underground image degradation and their constraints on intelligent analysis performance,and proposes an image enhancement technology system framework that runs through all levels of the system in combination with the"perception-edge-cloud"3-level collaborative intelligent video system architecture for coal mines.Focusing on typical underground imaging challenges,it sorts out and reviews the principles,ad-vantages,disadvantages,and representative models of traditional image enhancement methods such as histogram equalization,wavelet transform,and Retinex,deep learning-based enhancement methods such as super-resolution reconstruction,low-light enhancement,and defogging and dust removal,as well as multi-modal fusion enhancement technologies that integrate infrared/laser/millimeter wave with visible light,clarifying the technical characteristics and applicable scenarios of various methods.At the same time,combined with typical application scenarios of"human-machine-environment-management"such as mine personnel monitoring,equipment status monitoring,and operation process supervision,it demonstrates the practical application effects of targeted image enhancement technologies in improv-ing target recognition,defect detection accuracy,and operation monitoring clarity.Finally,in view of the existing problems in current tech-nologies,such as insufficient environmental dynamic adaptation capability,limited edge computing power,lack of high-quality real data-sets,and limited processing effects on multi-factor coupling degradation,it points out the future development directions of image enhance-ment technology,including collaborative optimization of lightweight edge computing models and hardware,continuous deepening of re-search on enhancement algorithms based on advanced network architectures such as GAN and Transformer,exploring the combination with large models to achieve active intelligent perception and semantic understanding,promoting the engineering application of cross-modal fusion technology,and ultimately forming a robust visual enhancement capability to support high-precision intelligent perception of the entire"human-machine-environment"domain and collaborative management and control of hazard sources underground.

Abstract:To address the challenges of coal-rock image segmentation under complex underground mining conditions—such as low illu-mination,high noise,and motion blur—this paper proposes an improved semantic segmentation model named Agent Swin-UNet,which integrates an Agent Attention mechanism into the Swin-UNet(Sliding Window Transformer U-Net)framework.The model adopts Swin Transformer as the backbone network,leveraging its hierarchical Window Multi-Head Self-Attention(W-MSA/SW-MSA)mechanism to establish long-range illumination dependencies,thereby alleviating detail loss in dark regions and degradation of local features.An Agent Attention Module is embedded into the skip connections between the encoder and decoder.This module introduces a triple-cooperative mechanism that employs agent tokens to realize an"aggregation-broadcast"style of feature interaction,reducing computational complex-ity from O(N2)to O(Nn)while preserving global semantic modeling capability and significantly improving computational efficiency.By incorporating spatially-aware bias,the model enhances its adaptability to noise distribution and effectively suppresses unstructured inter-ference,while the integration of depthwise separable convolution(DWC)strengthens local texture reconstruction,improving boundary de-lineation and fine-detail recovery.To mitigate the severe foreground-background imbalance inherent in coal-rock imagery,a composite loss function combining cross-entropy,Dice,and multi-scale structural similarity(MS-SSIM)losses is designed.This hybrid supervision optimizes the training process from multiple perspectives—classification consistency,regional overlap,and structural similarity—enhan-cing semantic coherence and boundary completeness under class-imbalance conditions.Experiments on the Shaanxi-Shanxi-Hebei Struc-tural Coal Dataset demonstrate that Agent Swin-UNet achieves 91.26%mIoU and 88.81%mPA on the standard test set,outperforming Segmenter,DeepLabv3,and the baseline Swin-UNet.Under noise interference with an intensity of 0.05,its mIoU remains 84.14%,indic-ating excellent noise robustness.Ablation studies further confirm that the Agent Attention Module is the principal source of performance improvement,particularly in high-noise environments(>0.05).The proposed method provides a robust and efficient solution for rapid coal-rock segmentation and intelligent excavation in complex underground environments.

Abstract:As one of the key parameters of roadway support,preload plays a decisive role in active support,and studies at home and abroad have shown that improving preload can effectively control roadway deformation and reduce the occurrence of roofing accidents.When the anchor cable is unloaded after the tensioning is completed,the retraction of the anchor cable leads to a large loss of preload,and the actual applied preload is much lower than the design value,and it is difficult to play the role of active support of the prestressed anchor cable.In order to reduce the loss of preload,a high preload stress uniform anchor cable was developed,and the components included high preload stress uniform locks,steel strands,pallets,support frames and other components.The high preload stress evenly distributed anchor cable can reduce the retraction of the anchor cable caused by the lock clamp and the retraction of the anchor cable caused by the damage and deformation of the surrounding rock,and reduce the loss of preload during the tensioning process.Its installation process is the pre-load of the rotating nut after the secondary tensioning,the first tensioning is exactly the same as the traditional anchor cable tensioning pro-cess,the secondary tensioning is realized by the support frame assistance,the nut is rotated after the gap is stretched,and the anchor cable is retracted when the anchor cable is stretched for the first time.In the process of the test,Digital Image Correlation(DIC)technology was used to study the"load-displacement"distribution law of the ordinary nut and the stress-uniformly distributed lock nut during the static load tensile process of the anchor cable.The field measurement and laboratory measurement show that the preload loss of the mine anchor cable is about 35%in the tensioning process,the preload loss of the high preload stress uniform load-bearing anchor cable during the ten-sioning process is only 10%,and the preload loss is reduced by 25%,and the stress uniformly distributed lock nut can alleviate the stress concentration of the thread pair in the tensile process and improve the bearing capacity of the lock thread.The research results have been applied in deep roadways such as Zhuji Mine,Paner Mine and Dingji Mine,and the control effect on the surrounding rock of the roadway is good.

Abstract:Aiming at the problems that the traditional contact deviation detection method in coal mines cannot detect the deviation amount,and the two-dimensional image detection method is affected by the harsh underground environment,a quantitative detection method for belt conveyor deviation in coal mines based on three-dimensional point cloud is proposed.To address the low efficiency of deviation de-tection caused by the large amount of point cloud data,the point cloud data of the belt conveyor is collected through the ROI(Region of Interest)local area sampling method,and the voxel centroid downsampling of the point cloud data is carried out to streamline the point cloud data.In order to filter out the noise points and redundant information points such as idler rollers and the frame in the point cloud data,an improved point cloud segmentation method combining Fast Euclidean Clustering(FEC)and region growing is proposed to achieve the filtering of redundant point cloud information such as idler rollers and the fast and accurate segmentation of the point cloud data on the conveyor belt surface.To solve the problems of low accuracy and efficiency in extracting the edge point information of the 3D point cloud of the conveyor belt,an adaptive RANSAC(Random Sample Consensus)inner point extraction method is proposed to intro-duce SPRT(Sequential Probability Ratio Test)and AIC(Akaike Information Criterion)to obtain the optimal model parameters of the con-veyor belt edge line,realizing the efficient and accurate fitting of the conveyor belt edge line.Aiming at the problem of precise quantitat-ive detection of conveyor belt deflection,a quantitative deflection detection method based on the spatial position deviation between the center line of the conveyor belt and the center line of the frame is proposed to accurately detect the deviation direction and amount of the conveyor belt.The experimental platform for belt conveyor deviation detection is built,and experiments are designed in two different ex-perimental environments of normal lighting and simulated fog and dust to verify the quantitative deviation detection method for belt con-veyors.Experimental results demonstrate that the proposed method achieves high-precision and real-time detection of conveyor belt devi-ation under various conditions.Under normal lighting,the average measurement error is 0.05 cm with a detection speed of 0.551 seconds per frame in the unloaded state,while under loaded conditions,the error increases to 0.12 cm and the detection speed decreases to 0.729 seconds per frame.The method maintains an error below 0.103 cm at low belt speeds and within 0.12 cm at high speeds,fulfilling practic-al accuracy requirements.Even in simulated fog and dust environments with severe point cloud loss,the system achieves an average error of 0.11 cm and a detection speed of 0.565 seconds per frame.Overall,the maximum average error across all test scenarios is 0.12 cm,and the fastest detection speed reaches 0.551 seconds per frame,both of which satisfy the real-time and accuracy demands of conveyor belt de-viation monitoring in coal mine production.

Abstract:Point cloud semantic segmentation in coal mine roadways is a critical technology for scene understanding in underground coal mine environments.However,the coexistence of multi-scale objects and imbalanced class samples in mine point clouds lead to poor seg-mentation performance for small-scale targets.To address the issue,a direction-guided grouping method for semantic segmentation of point clouds in coal mine roadways is introduced.Leveraging the directional distribution of typical small-scale objects(e.g.,pipes and cables),a Direction-Guided Grouping(DIG)strategy is proposed to optimize the shape of grouping envelopes in point cloud segmentation networks,thereby increasing the proportion of small-scale object points within groups and enhancing feature extraction.Based on this strategy,two directional guided grouping methods are proposed:one is Ellipsoid Query-based Direction-Guided Grouping(DIG-EQ),which preserves stronger local neighborhood relationships,and the other is Space-Filling Curve-based Direction-Guided Grouping(DIG-SFC),which offers higher computational efficiency.Both methods can be flexibly adapted to different network architectures and signific-antly improve the performance of small-scale object recognition.To mitigate class imbalance,a hybrid loss function is employed to im-prove sensitivity to underrepresented categories.A semantic segmentation dataset of coal mine roadway point clouds collected from differ-ent coal mine roadways is constructed for evaluation.The results show that the proposed method attains mean Intersection-over-Union(mIoU)of 61.84%,69.49%and 76.63%on PointNet++,PointNeXt-L,and Point Transformer V3 backbones,respectively,representing im-provements of 15.83%,6.25%and 0.35%over baseline models.For small-scale categories,mIoU reach 22.28%,28.61%and 47.30%,with relative gains of 29.99%,42.34%and 5.33%.These results demonstrate the effectiveness of the proposed approach in coal mine roadway scenarios.

Abstract:As a temporary storage facility in the coal mining and transportation process,coal bunkers are prone to safety accidents such as collapse and breach if their conditions are not properly monitored during long-term use.Due to the working environment of coal bunkers,traditional monitoring methods for coal bunkers have problems such as difficult perception and poor accuracy.A coal bunker condition monitoring method based on 4D millimeter-wave radar is proposed.By installing 4D millimeter-wave radar on the top of the coal bunker,it scans the velocity point cloud inside the coal bunker from top to bottom.The static point cloud and the moving point cloud are separated according to the Doppler velocity.For the separated dynamic point cloud,the velocity of the dynamic point cloud is calculated to obtain the coal feeding state of the coal bunker.For the separated static point cloud,calculate the farthest point of the static point cloud to obtain the remaining coal stacking height of the coal bunker.Then,use the Alpha shape algorithm to envelope the separated static point cloud to obtain the internal fitting contour of the coal bunker,realize the visualization of the coal stacking state of the coal bunker,and finally cal-culate the distance from the contour point cloud to the center of the coal bunker under different coal bunker heights.The contour curves composed of the minimum distances from the coal bunker contour to the center point of the coal bunker at different heights were obtained.The contour curves composed of the minimum distances from the coal bunker at different heights were transformed into image features by using Gram angular field transformation.The improved Dense Net neural network model was utilized for machine learning to achieve in-telligent recognition of the wall-hanging state of the coal bunker.At the same time,a coal bunker monitoring system was developed based on Unity and applied in industry.The results show that the coal bunker condition monitoring method based on 4D millimeter-wave radar has a good effect and can meet the synchronous monitoring of the coal pile height,coal feeding status and coal bunker wall connection status of the coal bunker.Compared with lidar,the maximum accuracy deviation of 4D millimeter-wave radar point cloud in perceiving the coal bunker contour is 0.22 m.This technology provides a new method for intelligent perception of coal bunkers.

Abstract:In the confined space of mining faces,where sensor deployment is restricted,developing a coal-rock structural perception sys-tem based on a single ground-penetrating radar(GPR)device holds significant engineering value.The accurate identification of rough coal-rock boundaries on exposed coal walls represents a critical challenge in constructing such a system.To address the limitations of single-do-main or homogeneous multi-feature representations in fully capturing the electrical differences between coal and rock,and to overcome the accuracy degradation of the conventional Fuzzy C-Means(FCM)algorithm caused by its equal-weighting strategy,a rough coal-rock boundary identification method is developed by integrating multi-domain robust features of radar images with an improved FCM al-gorithm.Twelve electromagnetic features capable of characterizing coal-rock electrical differences are first extracted from the time do-main,frequency domain,time-frequency domain,and wavelet domain,and their effectiveness is validated through forward simulations.Subsequently,three rough-surface models with varying root-mean-square heights are constructed to compute the mean coefficient of vari-ation and Pearson correlation coefficients among features.Seven features—envelope area,pulse width,spectral centroid,phase variation rate,mean instantaneous frequency,mean instantaneous phase,and scale energy ratio—are selected to establish a multi-domain robust fea-ture space for identifying transition zones of rough coal-rock interfaces.Based on the FCM framework,position encoding,L1-norm dis-tance metric,fuzziness criterion,label median filtering,and ground-truth judgment strategies are introduced to enhance the perception of spatial continuity in coal-rock distributions.Furthermore,an attention mechanism is incorporated to dynamically adjust feature weights,enabling adaptive clustering of coal-rock transition regions.Finally,the spatial distribution characteristics of the coal-rock interface within the transition zone are utilized to achieve precise boundary identification.Experimental results demonstrate that the proposed method ef-fectively identifies rough coal-rock boundaries on exposed coal walls,achieving a recognition error of 2.6%.

Abstract:Foreign object detection on mine conveyor belts is of significant importance for ensuring coal mine safety production and en-hancing automation levels.However,traditional machine vision methods are limited by low detection accuracy,high missed detection rates,and inadequate performance in identifying multi-scale and small-sized foreign objects,due to challenges such as insufficient under-ground illumination,strong point light interference,and complex backgrounds.Additionally,the limited availability of foreign object samples constrains model training effectiveness.To address these issues,an improved YOLOv7 foreign object detection algorithm is pro-posed,which integrates a feature enhancement structure and a focused linear attention mechanism.First,a feature enhancement module with a multi-scale parallel architecture and switchable atrous convolution is introduced in the backbone network to expand the receptive field and improve small target feature extraction.Second,a focused linear attention mechanism is incorporated to suppress complex back-ground interference while enhancing attention to critical foreign object regions.Third,an efficient dynamic serpentine convolution module is designed to adapt to the geometric characteristics of strip-shaped objects and strengthen semantic representation during feature fusion.Furthermore,a direction-aware SIOU loss function is employed to improve bounding box fitting accuracy and training convergence effi-ciency.Finally,a transfer learning strategy is implemented,where pre-training on generic datasets is leveraged before fine-tuning for coal mine conveyor belt detection tasks to mitigate small-sample training challenges.Experimental results on the CUMT-BELT dataset demon-strate that the proposed algorithm achieves a Precision of 89.3%,which represents a 9.3%improvement over the baseline YOLOv7 model.The method is also shown to significantly outperform other mainstream detection approaches in precision,recall,and mAP@0.5 metrics,while particularly demonstrating superior robustness and adaptability in multi-scale,small-target,and strip-shaped object detection tasks,thereby validating its effectiveness and practicality in complex mining scenarios.

Abstract:Solid backfill coal mining,as a green mining method that balances resource recovery and ecological protection,relies on coal and gangue separation as a core process for the efficient operation of integrated underground mining,selection,and backfill technology.However,coal and gangue identification,as a key technology for precise coal and gangue separation,faces challenges such as difficulties in feature extraction and vague boundary positioning in the complex underground working conditions.To address this,a method for recog-nizing the transient falling images of coal and gangue separation in underground solid backfill coal mining using Multimodal Large Lan-guage Models(MLLM)was proposed,with the underground coal and gangue separation in solid backfill coal mining as the research back-ground.First,an experimental platform for capturing transient falling images of coal and gangue separation in underground solid backfill coal mining was independently designed and built to simulate the complex underground conditions of low illumination and high dust.High-speed cameras were used to capture transient falling images of coal and gangue under different conditions.The collected images were pre-processed using optimized algorithms to enhance the brightness of low-illumination images and improve the quality of images in dusty en-vironments.The images were then annotated and augmented to construct a dataset for training and testing coal and gangue identification models.Subsequently,to address the shortcomings of the traditional SegFormer model in boundary recognition of coal and gangue images,an ECA was introduced and the loss function was optimized to construct the ECSegFormer model.Furthermore,MLLM was integrated in-to the ECSegFormer model to form the MLLM-ECSegFormer architecture.The MLLM Qwen-VL(7B)was used to extract the center co-ordinates of coal and gangue targets,and a spatial attention mask was generated through a Gaussian heatmap,which was then incorporated into the ECSegFormer encoder in stages to achieve dynamic interaction between multimodal prior knowledge and image features.The ex-perimental results showed that after the integration of the multimodal large language model,the performance of all classical image recogni-tion models was significantly improved.Specifically,the MLLM-ECSegFormer achieved an MIoU of 95.50%,an MPA of 98.92%,and an accuracy rate of 98.87%,significantly outperforming classical image recognition models in terms of recognition accuracy,model complex-ity,and recognition efficiency.Compared with classical image recognition models,the MLLM-ECSegFormer demonstrated stronger edge recognition continuity under complex conditions.Particularly in scenarios with dust interference and irregular shapes of coal and gangue,the segmentation accuracy of the target area was significantly better than that of traditional models.The research findings provided a new method for precise identification of coal and gangue,enhance the intelligence level of solid backfill coal mining technology,and are of great significance for the green and intelligent mining of coal resources.

Abstract:In view of the poor applicability of the traditional roadway portal hydraulic support due to its shortcomings such as bulky bot-tom beam,large space occupation and inconvenient installation and dismantling.Through theoretical analysis,model test,and numerical simulation,the supporting performance of roadway pile foundation portal hydraulic support was studied.Based on analyzing and summar-izing the advantages and disadvantages of traditional roadway portal hydraulic supports,a new type of roadway hydraulic support based on the stable structure system of support-pile foundation-surrounding rock was proposed.A simulation test model of support performance was established,and the digital speckle correlation method was used as the observation method to compare and analyze the support perform-ance of traditional portal hydraulic support and pile foundation portal hydraulic support in terms of horizontal displacement of surrounding rock,bottom drum value,and maximum shear strain.A numerical simulation model of support performance was constructed.The displace-ment of surrounding rock of roadway under different pile lengths,different pile diameters,and different pile foundation angles was calcu-lated,and the influencing factors of roadway pile foundation gantry hydraulic support performance were analyzed.The results show that:①The comparative analysis of the similar simulation test results shows that the deformation of the lower rock layer of the surrounding rock of the roadway under the support of the roadway pile foundation portal hydraulic support is significantly reduced.The maximum hori-zontal displacement,bottom drum value,and maximum shear strain are reduced by 21%,12%,and 32.6%,respectively.②The numerical simulation results show that the horizontal displacement of the surrounding rock of the roadway diffusion decreases from the sidewall of the roadway under the roadway pile foundation portal hydraulic support.The pile foundation has a partitioning effect on the roadway bot-tom drum,and the bottom drum only appears between the piles,and the bottom drum value decreases with the increase of the distance from the rock layer to the bottom surface of the roadway.The plastic deformation of the surrounding rock of the roadway is distributed in a butterfly shape,mainly concentrated at the top of the roadway,on both sides,and between the piles,and the plastic deformation area at the bottom of the roadway is separated by the pile foundation.③The analysis of the influencing factors indicates that increasing the length and diameter of the pile foundation can reduce displacement at the base of the roadway and effectively control bottom drum effects.However,the benefits of the pile foundation diminish once its length surpasses a certain point.Additionally,increasing the angle of the pile foundation can lead to greater displacement at the roadway bottom and worsen the bottom drum effects,which is detrimental to con-trolling the deformation of the surrounding rock.

Abstract:Tunnel Boring Machine(TBM)play a crucial role in coal mine rock roadway excavation.However,the processes of equipment lowering,transportation,and underground assembly still face urgent technical challenges such as high difficulty in construction manage-ment,significant safety risks,and low operational efficiency.Considering the specific requirements of coal mine roadways for TBM cut-terheads,a pure bolted design concept for cutterheads is proposed,and a combined method of theoretical analysis,ANSYS numerical sim-ulation,and on-site testing is adopted for force analysis.Four load condition models,including maximum thrust,upper-soft and lower-hard strata,turning correction,and stuck-release mode,are established.Various finite element mesh sizes are selected for stress calculation re-spectively.Combined with specific working condition parameters and boundary conditions,force loading is applied to the cutterhead,cla-rifying the maximum stress,maximum displacement,and shear force borne at the segmented positions of the cutterhead under various working conditions.The maximum working load of a single bolt is calculated and its strength check is completed.Through mechanical cal-culation and bolt layout optimization analysis,the optimal segmentation form and bolt distribution are determined,and a pure bolted struc-ture cutterhead suitable for coal mine roadways is developed.The results show that the pure bolted structure increases the underground as-sembly efficiency of the cutterhead by 30%and no bolt failure occurs during the excavation period;the pure bolted structure is reasonably designed,with good safety,reliability,and convenience in disassembly,assembly,and maintenance;the underground application has veri-fied the effectiveness of the established load condition models,bolt calculation methods,and preload verification methods.

Abstract:Compared with loose blasting,throwing blasting has higher vibration intensity and lower frequency,which has a greater impact on the safety production of open-pit mines.To accurately predict the vibration velocity of throwing blasting,literature review data and on-site monitoring data were used to analyze the characteristics of throwing blasting vibration velocity related data.Pearson correlation ana-lysis was used to clarify the key influencing factors of throwing blasting vibration velocity,and a prediction index system for throwing blasting vibration velocity was constructed.We separately constructed a genetic algorithm optimized least squares support vector machine(GA-LSSVM)model and an Elman neural network optimized adaptive enhancement algorithm(Elman Adaboost)model,and integrated the two using the inverse error method to form an integrated prediction model for throwing blasting vibration velocity.We also proposed evaluation criteria and testing methods for the models.The results indicate that the distance from the blasting center,height difference,number of rows,total charge,explosive consumption per unit,and hole spacing are the main factors affecting the vibration velocity of throwing blasting.Optimize to determine a maximum iteration count of 100,activation function of Relu,random number seed of 42,num-ber of neurons of 30,and data allocation ratio of 8∶2.Compared with the single model,the integrated prediction model can overcome the limitations of the traditional single prediction model,have better information capture ability,and improve the robustness and prediction ac-curacy of the vibration speed prediction model.The evaluation index determination coefficient(R2),root mean square error(RMSE),and mean absolute error(MAE)of this model are 0.957,4.382,and 2.173,respectively.Compared with GA-LSSVM and Elman Adaboost mod-els,R2 has increased by 5.51%and 12.34%,RMSE has increased by 15.88%and 25.63%,and MAE has increased by 35.99%and 33.34%,re-spectively.

Abstract:Understanding the fracture movement laws of rock strata plays an important role in the control of mine pressure and the preven-tion and control of dynamic disasters such as strong mine seismic events in stope.The fracture modes of thin rock strata and thick rock strata are obviously different,and in the production practice,due to the complexity of geological conditions,rock strata thickness often dis-plays"variable-thickness"characteristics.To systematically analyze fracture modes under varying thickness conditions,this study estab-lished a structure mechanical model of variable-thickness roof strata in the stope based on the geological setting of 30202 longwall face in Muduchaideng Coal Mine,and the influence mechanism of factors such as strata thickness gradient,unsupported span length,and mechan-ical strength on fracture modes were investigated,revealing the mechanical principles governing different fracture behaviors.The results show that with the thickness gradient increase,the dominant stress acting on the strata transitions from tensile stress to shear stress,shift-ing the fracture mode from"tensile fracture"to"shear fracture",and the unsupported span length of rock strata shows an increase with the increase of tensile strength and shear strength of rock strata,and shows different growth trends under different thickness gradients.Com-bined with the numerical simulation test,it is obtained that under the geological conditions of 30202 longwall face,the critical thickness of the rock strata in which the fracture mode of the thick and hard rock strata evolves from"tensile fracture"to"shear fracture"is 16.5 m,which reveals the different of fracture mode of rock strata with different thickness under specific geological conditions.Based on the re-sponse law of the seismic source fracture mechanism and the structure formed by the fracture of thick and hard rock strata,it is verified that the fracture modes of rock strata with different thicknesses are different.At the same time,based on the correlation between the on-site microseismic monitoring data and the structure formed by the fracture of thick and hard rock strata,through the inversion analysis of the whole process of overburden structure evolution and mine earthquake activity under thick and hard rock strata,the relationship between different fracture modes of rock strata and energy release is clarified.The research results can provide theoretical and technical support for the prevention and control of dynamic disasters.

Abstract:With the rapid advancement of intelligent coal mine construction,the volume of underground operational video data has surged dramatically.Current video processing and storage methods predominantly rely on single-scene video analysis and raw-format storage techniques,which face critical limitations:monolithic scene models lead to incomplete information descriptions,and constrained storage capacity results in short data retention periods.To address the practical need for comprehensive yet low-cost semantic analysis of under-ground videos,this paper proposes a novel coal mine video captioning method integrating working-condition complexity metric-based ad-aptive keyframe extraction and multimodal semantic modeling,achieving optimal computational parsing and natural language description of underground video content.First,a complexity metric assignment method is designed based on the distinctive features of underground working conditions.Building on this,a dynamic frame-sampling frequency algorithm is proposed to minimize computational overhead while ensuring robust key information capture.Subsequently,a Multimodal Large Language Model(MLLM)-based technical framework is developed,incorporating four core modules:adaptive keyframe extraction,large-model-driven visual-semantic feature extraction,prompt engineering and text encoding,and multimodal fusion and text decoding.This framework enables efficient,low-cost generation of natural language descriptions for full-scene underground video information.Comparative experiments demonstrate that the proposed meth-od achieves a key information capture rate of 95.4%while reducing computational resource consumption to 1.5%of traditional dense-sampling approaches.These results validate its viability as a technical solution for high-fidelity,cost-effective semantic analysis of under-ground videos.

Abstract:Mine dust is one of the important reasons that endanger mine personnel and equipment.Mine dust concentration detection is an important work content of coal safety management.Due to the complexity of the coal mine environment,the traditional mine dust concen-tration sensor has a small detection range and a high false alarm rate.The existing video dust concentration detection algorithm cannot eliminate the influence of luminosity on dust concentration detection and cannot locate the unevenly distributed dust concentration,result-ing in low detection accuracy and limited detection range.In view of the above problems,an intelligent detection technology of mine dust concentration based on dark channel prior is proposed.An intelligent detection test platform for dust concentration was built to collect dust images under different luminosity.The dark channel prior algorithm was used to extract the dark channel image of the dust image,and the image transmittance was calculated.Pearson correlation coefficient method was used to calculate the correlation between image transmit-tance and dust concentration under different luminosity.An ambient light detection algorithm is designed to detect the ambient luminosity,and a mathematical model of dust concentration and image transmittance under different luminosity is established to calculate the dust con-centration value.Aiming at the problem of uneven distribution of dust concentration in the environment,a dust area location algorithm is proposed to distinguish and process the dust concentration in different areas.In order to verify the practicability and accuracy of the al-gorithm in this paper,the algorithm is used to detect the images of different dust concentrations under different luminosity,and compared with the dust concentration sensor and the existing video detection algorithm.The results show that the proposed intelligent detection al-gorithm of dust concentration can effectively reduce the influence of environmental luminosity on the detection of dust concentration,and can effectively locate and calculate the uneven distribution of dust concentration.The ambient light detection algorithm can detect the lu-minosity from low to high under different dust concentrations,and the accuracy is 92.4%.The detection accuracy of dust concentration is 93.36%,and the average error rate is 6.64%.The proportion of the intersection and union ratio(IoU)greater than 0.5 between the dust con-centration positioning area and the actual dust area was 66.7%.The dust image of Jinjitan Coal Mine is selected to verify the feasibility of the algorithm in this paper,which effectively improves the problem of dust area location and concentration detection.The research results provide technical support for intelligent detection of mine dust concentration image,and have important practical significance for accurate detection and early warning of mine dust concentration.

Abstract:Stress redistribution of deep excavation rock mass is a complex mechanical problem of multiple stresses.First,the mechanical model of surrounding rock at the moment of excavation of deep high-stress soft rock roadways was established based on the superposition principle of elastic theory,and the variation rules of tangential stress σ1 and radial stress σ3 were studied.Then,the supporting mechanical model of deep high-stress soft rock roadways was constructed,and the supporting effect of surrounding rock was revealed.Next,the re-sponse characteristics of σ1 and σ3 of surrounding rock and corresponding failure mechanisms under traditional support and compensation support were investigated through physical model tests.Finally,the applicability of compensation support technology was verified by nu-merical simulation and field engineering application.Mechanical analysis results indicated that the stress redistribution in deep high-stress soft rock roadways shows the trend of radial pressure relief and tangential pressure increase,σ1 becomes twice of the original and σ3 de-creases to 0 at the moment of excavation.After supported the excavated rock mass,the σ3 of roadway edge is the support resistance of sur-rounding rock,and σ1 decreases with the increase of support resistance.The model test results showed that surrounding rock deformation of compensation support reduces by 73.7%,the destroyed area reduces by 88.3%,the crack length reduces by 11.0%.The σ3 of the shal-low surrounding rock increases by 68.3%,and the peak value of σ1 decreases by 18.2%.The compensation effect of traditional support is weak,the σ3 attenuates greatly and the σ1 concentrates highly,which causes the crack to continue to open and extend in depth,resulting in the surrounding rock expansion deformation.Compensation support gives full play to the three-axis strength of the surrounding rock and mobilizes the self-bearing capacity of the deep rock mass by NPR cable.The compensation degree of σ3 is high and σ1 is significantly re-duced,thus restraining the propagation and penetration of crack and realizing the self-stability of surrounding rock.After the compensa-tion support technology is used in soft rock roadways,the NPR cables achieve constant resistance,the deformation of surrounding rock and the damage degree of support are significantly reduced,which indicates that this technology has a good control effect on deep soft rock engineering with high in situ stress.