2024 Vol. 50, No. 5

Academic Column of Editorial Board Member
Research on automatic detection and alarm methods for coal mine rock burst and coal and gas outburst accidents
SUN Jiping, CHENG Jijie
2024, 50(5): 1-5, 13. doi: 10.13272/j.issn.1671-251x.18188
<Abstract>(165) <HTML> (50) <PDF>(57)
Abstract:
The automatic perception and alarm method for coal mine rock burst and coal and gas outburst is an effective measure to timely detect accidents and emergency rescue, reduce casualties, avoid or reduce secondary accidents such as gas and coal dust explosions, and curb delayed, missed, and concealed reporting of accidents. It is difficult to perceive coal mine rock burst accidents, and there is currently no automatic detection and alarm method for coal mine rock burst accidents. Coal mine rock burst accidents are mainly discovered manually. At present, there are only automatic alarm methods for coal and gas outbursts based on methane, wind speed, and direction sensors. There are problems such as slow response speed and inability to detect significant increases in methane concentration before methane sensor damage. A method for image perception and alarm of coal mine rock burst and coal and gas outburst has been proposed. Based on the image features of coal mine rock burst and coal and gas outburst temperature, color, depth, burial, etc., the method recognizes coal mine rock burst and coal and gas outburst. Based on the changes in gas concentration in the roadway space and mining face, the method distinguishes between rock burst and coal and gas outburst. If the gas concentration rapidly increases over a large area, it is judged as coal and gas outburst, otherwise it is judged as rock burst. This method has the advantages of intuitiveness, fast response speed, non-contact, wide monitoring range, simplicity and reliability, and can intuitively record the real situation of coal mine rock burst and coal and gas outburst. When the coal mine rock burst and coal and gas outburst accidents are alarmed, the personnel on duty in the control room can immediately confirm the accident through video recording and carry out emergency rescue in a timely manner. A method has been proposed to reduce the impact of coal mine rock burst and coal and gas outbursts on image perception, including multi point arrangement of cameras, setting of cameras at higher positions, timely transmission of video data, and multi-point arrangement of methane sensors.
Path planning of coal mine underground robot based on improved artificial potential field algorithm
XUE Guanghui, WANG Zijie, WANG Yifan, LI Yanan, LIU Wenhai
2024, 50(5): 6-13. doi: 10.13272/j.issn.1671-251x.2024030014
<Abstract>(151) <HTML> (37) <PDF>(27)
Abstract:
Path planning is one of the key technologies that urgently need to be solved in the application of coal mine robots in narrow underground roadways. A path planning method for coal mine robots based on improved APF algorithm is proposed to address the issues of traditional artificial potential field (APF) algorithms that planning paths in narrow roadway environments may be too close to the roadway boundary, as well as the possibility of unreachable targets and path oscillations near obstacles. Referring to the relevant provisions of the Coal Mine Safety Regulations, the boundary potential field between the two sides of the roadway is established. The robot's path is planned as much as possible in the middle of the roadway to improve the safety of robot travel. The method introduces regulatory factors into the repulsive potential field of obstacles to solve the problem of unreachable targets. The method introduces corner constraint coefficients to smooth the planned path, reduce oscillations, improve planning efficiency, and ensure the safety of the planned path. The simulation results show that when the target point is very close to the obstacle, the improved APF algorithm can successfully plan a path that can reach the target point. The improved APF algorithm reduces the planning cycle by an average of 14.48% compared to traditional algorithms. The cumulative value of steering angle reduces by an average of 87.41%, and the sum of absolute curvature values is reduced by an average of 78.09%. The results indicate that the improved APF algorithm plans smoother paths, shorter path lengths, and has higher planning efficiency and safety.
Overview
Current research status and development trends of deep well rescue technology and equipment
WEN Hu, HOU Zongxuan, ZHENG Xuezhao, CAI Guobin, YAN Ruijin
2024, 50(5): 14-22, 35. doi: 10.13272/j.issn.1671-251x.18175
<Abstract>(147) <HTML> (39) <PDF>(36)
Abstract:
Deep well rescue technology refers to the key technologies involved in various aspects of rescuing trapped personnel during the process of deep well accident rescue. It mainly includes environmental detection technology, life detection technology, deep well rapid demolition technology, emergency communication network construction technology, and other auxiliary technologies to ensure the smooth progress of deep well accident rescue. Deep well rescue equipment refers to necessary hardware equipment and software platforms during the implementation of deep well rescue technology. Studying deep well rescue technology and equipment is crucial for ensuring the safety of trapped personnel and reducing accident losses. The current research status of deep well rescue equipment and key technologies is analyzed. It is pointed out that existing rescue technologies and equipment cannot fully meet the complex and changing environmental requirements. There are problems such as insufficient research on the universality and specificity of rescue equipment, the need to improve the intelligence level of rescue equipment, difficulty in meeting the needs of complex rescue environments with network collaboration capabilities, and insufficient innovation research on rescue equipment. In order to solve the above issues, the development trend of deep well rescue equipment and technology is discussed. ① Deep well rescue equipment should be divided into specialized and universal categories based on different rescue scenarios. A single equipment should develop towards multifunctionality, high reliability, and high mobility. ② Rescue equipment is intelligent, precise, and self decision-making, achieving a rescue mode of intelligent equipment as the main focus and personnel as the auxiliary. ③ It is suggested to build a rapid networking, multi-mode networking, and integrated rescue network platform. ④ Although TDLAS integration does not currently meet rescue standards, its high resolution, high sensitivity, and integrability will play an important role in the future, achieving high integration, lightweight, and efficiency of environmental monitoring equipment.
Achievements of Scientific Research
Intelligent shearer cutting control based on process driven technology
ZHENG Chuang, LI Danning, FENG Yinhui
2024, 50(5): 23-27, 150. doi: 10.13272/j.issn.1671-251x.2023090017
<Abstract>(93) <HTML> (35) <PDF>(20)
Abstract:
The traditional shearer cutting control lacks analysis of the state of the shearer drum, resulting in low quality of cutting template generation. It does not fully consider the undulation of the working face and geological environmental conditions, which makes it impossible to obtain the optimal cutting path. Relying on the control unit of the shearer itself cannot adjust the height of the drum in a timely manner. In order to solve the above problems, a process driven intelligent shearer cutting control scheme is proposed. According to the hydraulic support number of the working face, real-time collection of corresponding drum cutting height data is carried out. Combined with historical data of drum cutting height, real-time data is processed to generate a shearer cutting template that conforms to the trend of the working face roof and floor curve. Based on realistic data from the roof and floor of the working face and manual coal cutting experience, the method plans the cutting path of the shearer and performs real-time intervention to achieve adaptive coupling between the cutting height of the shearer drum and the curve of the roof and floor of the working face. By editing the coal mining process and setting the cutting template data, a coal mining process table file is formed. The cutting height of the shearer drum is adjusted accordingly to achieve adaptive height adjustment control of the shearer. The intelligent shearer cutting control scheme based on process driven technology is applied to the 43207 working face of Yujialiang Coal Mine in Shendong Coal Group. It achieves unmanned and normalized coal mining operations. The number of personnel in the production team working face is reduced from 3 to unmanned in the middle of the working face, and the automatic coal cutting rate of the shearer is over 97%.
Research on visual recognition technology for appearance defects of steel wire rope in mine hoist
WANG Guofeng, WANG Shoujun, TAO Rongying, LI Nan, LUO Ziqiang
2024, 50(5): 28-35. doi: 10.13272/j.issn.1671-251x.2024010080
<Abstract>(190) <HTML> (36) <PDF>(40)
Abstract:
A visual recognition method for appearance defects of mine hoist steel wire ropes based on computer vision and deep learning is proposed to address the problems of difficult deployment for detecting multiple steel wire ropes, low image acquisition quality of steel wire ropes, poor adaptability and accuracy of visual detection methods. Firstly, an online monitoring system for the steel wire rope of the mine hoist is constructed. Secondly, the steel wire rope images are collected by the ground mobile inspection platform and the underground intrinsic safety high-speed camera, and a steel wire rope image dataset is established. Considering the effects of underground dust, susceptibility of camera lenses to contamination, uneven lighting, and high light reflection of steel wire ropes, image denoising methods based on Retinex algorithm and homomorphic filtering are used to denoise the steel wire rope images. The processing results show that the automated multi-scale Retinex with color restoration (AutoMSRCR) algorithm based on color gain weighting is the optimal solution. The defect detection process is based on convolutional neural networks, and a defect detection model based on YOLOv5s is constructed. In order to reduce the influence of human factors and the workload of parameter tuning, a Focus structure is added to YOLOv5s for optimization. The improved YOLOv5s model is used as a pre training model for steel wire rope defect detection to further reduce the memory usage of the model and improve the loading and detection speed of the model. The experimental results show that the proposed method has detection errors of 1.61% and 1.35% for wire breakage at 2 positions of the steel wire rope, and detection errors of 2.43%, 3.44%, 2.11%, and 3.39% for wear at 4 positions of the steel wire rope. In response to the problem that the detection precision of the original steel wire rope safety monitoring system for the main shaft hoist of Gubei Coal Mine, Huaihe Energy Holding Group, cannot meet the on-site requirements, the proposed method is adopted to improve the original system. The on-site application results show that the accuracy of wire rope breakage detection is increased from 80% to 96%, the damage positioning error is reduced from 500 mm to within 300 mm. The damage positioning accuracy is increased from 75% to 98%, the real-time detection rate of damage is increased from 76% to 90%, and the tail rope distortion detection rate is increased from 70% to 85%.
Analysis and Research
Research on coal gangue detection in coal preparation plant based on YOLOv5s-FSW model
YAN Bijuan, WANG Kaimin, GUO Pengcheng, ZHENG Xinxu, DONG Hao, LIU Yong
2024, 50(5): 36-43, 66. doi: 10.13272/j.issn.1671-251x.2023100090
<Abstract>(153) <HTML> (47) <PDF>(31)
Abstract:
A coal gangue detection method in coal preparation plant based on YOLOv5s-FSW model is proposed to address the problems of insufficient feature extraction, large parameter quantity, low detection precision, and poor real-time performance in existing coal gangue detection models. This model is improved on the basis of YOLOv5s. Firstly, the C3 module in the Backbone section is replaced with a FasterNet Block structure, which improves detection speed by reducing the number of model parameters and computation. Secondly, in the Neck section, a parameter free SimAM attention mechanism is introduced to enhance the model's attention to important targets in complex environments, further improving the model's feature extraction capability. Finally, in the Prediction layer, the CIoU bounding box loss function is replaced with Wise-IoU, and the model focuses on ordinary quality anchor boxes to improve convergence speed and bounding box detection precision. The results of the ablation experiment indicate that compared with the YOLOv5s model, The mean average precision (mAP) of the YOLOv5s-FSW model has been improved by 1.9%, the model weight has been reduced by 0.6 MiB, the number of parameters has been reduced by 4.7%, and the detection speed has been improved by 19.3%. The comparative experimental results show that the YOLOv5s-FSW model has a mAP of 95.8%, which is 1.1%, 1.5%, and 1.2% higher compared to the YOLOv5s-CBC, YOLOv5s-ASA, and YOLOv5s-SDE models, respectively, and compared to YOLOv5m, YOLOv6s improved by 0.3%, 0.6% respectively. The detection speed of the YOLOv5s-FSW reaches 36.4 frames per second, which is 28.2% and 20.5% higher than the YOLOv5s-CBC and YOLOv5s-ASA models, respectively. Compared to YOLOv5m, YOLOv6s and YOLOv7, the detection speed of the YOLOv5s-FSW has increased by 16.3%, 15.2%, and 45.0%, respectively. The visualization experiment results of the thermal map show that the YOLOv5s-FSW model is more sensitive to the target feature areas of coal gangue and has higher attention. The detection experiment results show that in complex scenes with dim environments, blurred images, and mutual occlusion of targets, the YOLOv5s-FSW model has a higher confidence score for coal gangue target detection than the YOLOv5s model, and effectively avoids the occurrence of false positives and missed detection.
Research on online detection of particle size in fine-grained coal classification overflow
SUN Haozhi, MA Jiao, SHI Changliang, WANG Hanlu
2024, 50(5): 44-51, 59. doi: 10.13272/j.issn.1671-251x.2024040010
Abstract:
Real time online detection of the particle size of the overflow in the selection and classification of fine-grained coal can be carried out, and the classification parameters can be adjusted to reduce the content of coarse particles in the overflow and improve the total clean coal recovery rate. The current research generally limits the detection of overflow particle size to around 180 μm, and the upper limit of slurry volume concentration is 10%. It cannot meet the requirements of overflow particle size detection for fine-grained coal classification cyclones with coarse particle size, wide particle size range, and high volume concentration. A set of ultrasonic online particle size detection system has been developed to improve the upper limit of coal particle size and slurry volume concentration detection. Based on the ultrasonic attenuation model, a coal particle size detection model suitable for on-site conditions of fine-grained coal classification with coal particle size of 44.5-600 μm and slurry volume concentration of 0-40% is constructed. A coal particle size distribution prediction model is established using a BP neural network optimized by particle swarm optimization algorithm, achieving the prediction of the particle size distribution of the overflow slurry in a fine-grained coal classification cyclone. The simulation results based on the coal particle size detection model show that the ultrasonic attenuation value decreases first and then increases with the increase of coal particle size, and increases with the increase of ultrasonic frequency and slurry volume concentration. The ultrasonic online particle size detection system and coal particle size distribution prediction model are respectively used to detect the distribution of overflow particle size (actual value is 150.0, 215.0, 315.0 μm) in a hydraulic classification cyclone of a certain mine. The results show that the relative errors of the measurement values of the detection system are 10.87%, 9.81%, 8.48%, and the relative errors of the predicted values of the prediction model are 9.27%, 6.05%, and 6.92%. It indicates that the research have achieved accurate detection of overflow particle size in fine-grained coal classification.
Research on coal gangue recognition algorithm based on HGTC-YOLOv8n model
TENG Wenxiang, WANG Cheng, FEI Shuhui
2024, 50(5): 52-59. doi: 10.13272/j.issn.1671-251x.2024030064
<Abstract>(156) <HTML> (28) <PDF>(22)
Abstract:
The existing deep learning based coal gangue recognition methods have problems in complex working conditions such as low lighting, high noise, and motion blur in coal mines, such as low precision of coal gangue recognition, easy omission of small target coal gangue, large model parameter and computational complexity, and difficulty in deploying to devices with limited computing resources. A coal gangue recognition algorithm based on the HGTC-YOLOv8n model is proposed. The method replaces the backbone network of YOLOv8n with HGNetv2 network, effectively extracts multi-scale features to improve coal gangue recognition performance and reduces model storage requirements and computational resource consumption. The method embeds a Triplet Attention mechanism module in the backbone network to capture interaction information between different dimensions. The method enhances the extraction of target features in coal gangue images, and reduces the interference of irrelevant information. The method selects the content aware reassembly of features(CARAFE) to improve the upsampling operator of YOLOv8n neck feature fusion network, utilizing contextual information to enhance perceptual field of view and improve the accuracy of small target coal gangue recognition. The experimental results show the following points.① The average precision of the HGTC-YOLOv8n model is 93.5%, the parameters number of the model is 2.645×106, the number of floating-point operation is 8.0×109, and the frame rate is 79.36 frames/s. ② The average precision of the YOLOv8n model has increased by 2.5% compared to the YOLOv8n model, and the number of parameters and floating-point operations have decreased by 16.22% and 10.11%, respectively. ③ The comparison results with the YOLO series models show that the HGTC-YOLOv8n model has the highest average precision, the least number of parameters and floating-point operations, fast detection speed, and the best overall detection performance. ④ The coal gangue recognition algorithm based on the HGTC-YOLOv8n model has improved the low precision of coal gangue recognition and the easy omission of small target coal gangue under complex working conditions in coal mines. The method meets the requirements of real-time detection of coal gangue images.
Multi step prediction of dense medium clean coal ash content based on time series alignment and TCNformer
WANG Jun, WANG Ranfeng, WEI Kai, HAN Jie, ZHANG Qian
2024, 50(5): 60-66. doi: 10.13272/j.issn.1671-251x.2023090007
Abstract:
Due to the different positions of various sensors during the dense medium separation process, there is a time lag between the main process parameters of dense medium separation and ash content, which affects the results of clean coal ash content. The grey prediction method based on regression models lacks the utilization of time series information and cannot capture the dynamic features of the dense medium production process over time. The time series based ash prediction method fails to fully consider the time dependence relationship between the main process parameters of ash content and dense medium separation. In order to solve the above problems, a multi step prediction method for dense medium clean coal ash content based on time series alignment and TCNformer is proposed. The method quantifies the lag step between the main process parameters of ash content and dense medium separation through lag correlation analysis. The method moves the main process parameters of dense medium separation in the time dimension accordingly, aligning the time series of the main process parameters of ash content and dense medium separation, and eliminating the time lag between the main process parameters of ash content and dense medium separation. On the basis of the Transformer model, a time convolutional network (TCN) is introduced to extract features, and the unidirectional encoder is extended to a bidirectional encoder to construct the TCNformer model for multi-step prediction of clean coal ash content. The sequence of process variables corresponding to the grey data at future moments obtained from the time series alignment is used as an input to the decoder to improve the model prediction precision. The experimental results show that the average absolute error of this method is 0.157 9%, the root mean square error is 0.215 2%, and the average Pearson correlation coefficient is 0.505 1, which can effectively improve the precision of predicting clean coal ash content.
Multi-personnel underground trajectory prediction method based on Social Transformer
MA Zheng, YANG Dashan, ZHANG Tianxiang
2024, 50(5): 67-74. doi: 10.13272/j.issn.1671-251x.2023110084
<Abstract>(98) <HTML> (24) <PDF>(17)
Abstract:
Currently, in the prediction methods of underground personnel trajectories in coal mines, Transformer not only has lower computational complexity compared to recurrent neural network(RNN) and long short-term memory (LSTM), but also effectively solves the problem of long-term dependence caused by gradient disappearance when processing data. But when multi personnel are moving simultaneously in the environment, the Transformer's prediction of the future trajectories of all personnel in the scene will have a significant deviation. And currently, there is no model in the field of underground multi personnel trajectory prediction that simultaneously uses Transformer and considers the mutual influence between individuals. In order to solve the above problems, a multi personnel underground trajectory prediction method based on Social Transformer is proposed. Firstly, each individual is independently modeled to obtain their historical trajectory information. Feature extraction is performed using a Transformer encoder, followed by a fully connected layer to better represent the features. Secondly, an interactive layer based on graph convolution is used to connect each other, allowing spatially close networks to share information with each other. This layer calculates the attention that the predicted object allocates to its neighbors when influenced by them, extracts their motion patterns, and updates the feature matrix. Finally, the new feature matrix are decoded by the Transformer decoder to output predictions of future position information. The experimental results show that the average displacement error of Social Transformer is reduced by 45.8% compared to Transformer. Compared with other mainstream trajectory prediction methods such as LSTM, S-GAN, Trajectoron++, and S-STGCNN, the prediction errors are reduced by 67.1%, 35.9%, 30.1%, and 10.9%, respectively. This can effectively overcome the problem of inaccurate prediction trajectories caused by mutual influence among personnel in the underground multi personnel scenario of coal mines and improve prediction precision.
Research on high-precision coal flow detection of belt conveyors based on machine vision
JI Xianliang, ZHANG Wenjie, WANG Yuqiang, LIU Yong, TIAN Zuzhi, FU Zheng
2024, 50(5): 75-83. doi: 10.13272/j.issn.1671-251x.2024030028
<Abstract>(135) <HTML> (36) <PDF>(27)
Abstract:
In response to the problems of missing image details and poor fitting effect in multiple fractures or areas with large fracture spacing in existing machine vision based coal flow detection methods for belt conveyors, a high-precision coal flow detection system for belt conveyors based on machine vision is proposed. It is based on the principle of direct beam oblique collection laser triangulation. The line laser emitter is arranged directly above the measurement position of the belt conveyor and vertically irradiates the coal pile. The coal pile moves uniformly with the belt conveyor, and a camera at an oblique angle is used to capture real-time images of the surface of the coal pile containing laser stripes. The method calibrates the coal flow detection system, including camera internal parameter calibration and laser plane calibration, to obtain the height information of the coal pile. The processing of laser stripe images on coal flow cross-sections is carried out. The gray center of gravity method and regional skeleton method are compared and analyzed from multiple perspectives such as extraction precision and algorithm real-time performance. Based on the comparison results, the regional skeleton method is selected to extract the center of laser stripes. Aiming at the problem of poor fitting effect of laser stripe fracture repair using image dilation operation, the least squares method is proposed as the laser stripe fracture repair algorithm. Compared with closed operations, the least squares method has better smoothing effect and higher precision in fitting processing. The method establishes a coal flow cross-sectional area calculation model. By calculating the cross-sectional area of the coal pile at each frame, the coal flow volume at different belt speeds can be obtained. The experimental results show that when the belt speeds are 0.25, 0.5, and 1 m/s respectively, the detection system errors are relatively small, with maximum errors of 2.78%, 3.61%, and 3.89%. It verifies that the coal flow detection system has high accuracy.
Research on information model of coal mine fully mechanized mining equipment based on industrial Internet
PAN Wenlong, LI Shengjun, GAO Quanjun, YANG Luyu, LIU Qingfu, ZHANG Heming
2024, 50(5): 84-92. doi: 10.13272/j.issn.1671-251x.2024010022
<Abstract>(128) <HTML> (21) <PDF>(30)
Abstract:
The equipment for coal mine fully mechanized working faces comes from different manufacturers, with inconsistent interfaces, different data systems and business logic, resulting in data barriers and slow data exchange between systems. Based on the industrial Internet architecture, an intelligent fully mechanized mining technology architecture including equipment layer, access layer, edge layer, PaaS layer and application layer is proposed. Based on this architecture, coal mine fully mechanized mining equipment is treated as the overall data object, and a method for constructing an information model of coal mine fully mechanized mining equipment is designed. Four key elements and modeling rules, including attributes, methods, events, and alarms are defined to achieve seamless communication between various physical entities and heterogeneous systems. This means defining, describing, and associating information resources of fully mechanized mining equipment, providing a complete and unified data object expression, description, and operation model. A modeling element optimization mechanism based on importance and semantic similarity is proposed to address the problem of excessive attribute elements in the information model of fully mechanized mining equipment. A fully mechanized mining equipment information model for the 81004 working face of No.1 Mine of Huayang New Materials Technology Group Co., Ltd. is established using the above method. The operation status of the electric motors of the fully mechanized mining equipment is evaluated. The results show that based on this model, all electric motors in the working face could be monitored for starting times in a short period of time, multiple motor equipment starting power balance, and the operating efficiency. Analysis results are generated to provide data support for decision-making work.
SEI based intelligent monitoring video transmission method for coal mines
CHEN Jia, WANG Qi, WANG Peng
2024, 50(5): 93-98. doi: 10.13272/j.issn.1671-251x.2023100025
<Abstract>(58) <HTML> (21) <PDF>(14)
Abstract:
Currently, there is a high latency problem in the transmission of video surveillance data in coal mines, and the main cause of video transmission delay is encoding delay. In order to solve the above problems, a intelligent monitoring video transmission method for coal mines based on media supplemental enhancement information(SEI) without video encoding is proposed. This method caches a copy of the compressed video frame obtained by demultiplexing the video stream, and decodes the compressed video frame to obtain the decoded video frame. The method stores the AI model analysis results in the decoded video frame through SEI, writes the custom SEI into the network extraction layer unit corresponding to the compressed video frame copy of the decoded video frame based on the timestamp correspondence. The method multiplexes the compressed video frame copy to achieve real-time transmission of coal mine intelligent monitoring videos. Experimental testing of this method is conducted on a 24 core CPU. The results show that for videos with a resolution of 1280×720, the overall CPU utilization rate for video processing using this method decreases from 24.7% to 36.3% when using traditional methods to 20.3% to 23.9%. The end-to-end delay decreases from 1946 ms to 345 ms. For videos with a resolution of 1920×1080, the overall CPU utilization rate for video processing using this method decreases from 29.2% to 41.8% using traditional methods to 18.5% to 26.3%. The end-to-end latency decreases from 6204 ms to 479 ms. This method reduces the transmission delay of coal mine intelligent monitoring videos by avoiding the video encoding process, saves CPU or GPU resources required for video encoding, and reduces the hardware cost of the intelligent video monitoring system.
A pose recognition method for warehouse cleaning robots based on extended Kalman filtering
LI Guihu, GAO Guijun, LI Junxia, JIA Xuefeng
2024, 50(5): 99-106. doi: 10.13272/j.issn.1671-251x.2024020004
<Abstract>(77) <HTML> (24) <PDF>(12)
Abstract:
The lighting intensity of coal mine water storage roadways is uneven and the structured features are obvious. Traditional vision based robot pose recognition methods are not accurate. The single robot positioning techniques such as Adaptive Monte Carlo localization (AMCL) method have significant cumulative errors in the output pose information with the long-term operation of the cleaning robot. It is easy to encounter situations where the coal slurry is not cleaned thoroughly and collides with both sides of the roadway. In order to solve the above problem, a multi-sensor fusion clearance robot pose recognition method based on extended Kalman filtering is proposed. Firstly, the method builds a multi-sensor fusion algorithm framework and establishes models for odometer, inertial measurement devices, and LiDAR data acquisition. Secondly, based on the principle of extended Kalman filtering, an observation equation is established using the angle information of the inertial measurement device. Combined with the odometer pose information, the first fusion of the clearance robot pose matrix is obtained. Then, the position information of the lidar is iterated with the previous pose matrix to obtain the second fused clearance robot pose matrix. Finally, the complementary filtering algorithm is used to process the pose matrix of the clearance robot after two fusion and output the final pose matrix of the clearance robot. The experimental results show that the maximum position error in linear pose recognition is 0.04 m, and the maximum attitude angle error is 0.05 rad. The maximum position error in the simulated roadway experiment is 0.1 m, and the maximum attitude angle error is 0.085 rad. Compared with the AMCL method, the pose recognition method of the warehouse cleaning robot based on extended Kalman filtering shows significant effectiveness in reducing the cumulative error during the operation of the warehouse cleaning robot.
Multi sensor adaptive fusion SLAM method for underground mobile robots in coal mines
MA Aiqiang, YAO Wanqiang
2024, 50(5): 107-117. doi: 10.13272/j.issn.1671-251x.2024050031
<Abstract>(120) <HTML> (27) <PDF>(18)
Abstract:
Mobile robots based on simultaneous localization and mapping (SLAM) technology can quickly, accurately, and automatically collect spatial data for spatial intelligent perception and environmental map construction. It is the key to achieving intelligent and unmanned coal mines. However, the current multi sensor fusion SLAM method in coal mines suffers from degradation and failure in robot front-end pose estimation, as well as insufficient precision in back-end fusion. This study proposes a LiDAR-visual-IMU adaptive fusion SLAM method for underground mobile robots in coal mines. The method clusters and segments LiDAR point cloud data, extracts line and surface features, and uses IMU pre integration state for distortion correction. The method uses image enhancement algorithm based on adaptive Gamma correction and contrast limited adaptive histogram equalization (CLAHE) to process low light images, and then extracts visual point and line features. The method provides initial pose values for LiDAR feature matching and visual feature tracking using IMU pre integration state. The pose of the mobile robot is obtained by matching the line and surface features of adjacent frames of LiDAR. Then, visual point and line feature tracking is performed to calculate the LiDAR, visual, and IMU pose changes. The stability of the front-end odometer is detected by setting dynamic thresholds, and the optimal pose is adaptively selected. The method constructs residuals for different sensors, including point cloud matching residuals, IMU pre integration residuals, visual point line residuals, and edge residuals. In order to balance precision and real-time performance, a sliding window based joint nonlinear optimization of multi-source data for laser point cloud features, visual features, and IMU measurements is implemented to achieve continuous and reliable SLAM in coal mines. Experimental verification is conducted on the effects before and after image enhancement. The results show that the image enhancement algorithm based on adaptive Gamma correction and CLAHE can significantly improve the brightness and contrast of the backlight and lighting areas, increase the feature information in the image, and significantly improve the quality of feature point extraction and matching. It achieves a matching success rate of 90.7%. To verify the performance of the proposed method, experimental verification is conducted in narrow corridor and coal mine roadway scenarios. The results show that the root mean square error of the proposed method in narrow corridor scenarios is 0.15 m, and the consistency of the constructed point cloud map is high. The root mean square error of positioning in the coal mine roadway scenario is 0.19 m. The constructed point cloud map can truly reflect the underground environment of the coal mine.
A positioning solution method for roadheader under optical target occlusion conditions
WANG Pengpeng, LI Rui, LIU Xin, LI Xiang, FU Changliang
2024, 50(5): 118-124. doi: 10.13272/j.issn.1671-251x.2023110001
Abstract:
In order to solve the problem of interruption in the positioning of roadheader in the case that the optical target is blocked under the current commonly used integrated navigation positioning of roadheader based on "inertial navigation+visual measurement+optical target", a positioning solution method for roadheader under optical target occlusion occlusion is proposed. Firstly, the method collects images of an optical target composed of four rectangular distributed target points in unblocked conditions, obtains the pixel coordinates of the imaging spot of the target points in the camera, and constructs a rectangle. Then, the method expands and constructs an auxiliary rectangular area box according to a certain proportion. Secondly, the method collects images of partially blocked target points, obtains the pixel coordinates of the imaging spot of the unblocked target points in the camera. The method determines the corresponding relationship between the unblocked target points and the imaging spot based on the Euclidean distance between the imaging spot of the target points and the vertex of the auxiliary rectangular area box, thereby determining the blocked target points. Thirdly, using the known geometric dimensions of the target and the target attitude information provided by inertial navigation, the method establishes the corresponding relationship between the projected target point and the imaging spot, and then solves for the pixel coordinates of the spot corresponding to the blocked target point. Finally, the spatial coordinates of the center position of the optical target are obtained using the perspective-N-point (PNP) algorithm to achieve the positioning solution of the roadheader. The experimental results show that when the optical target is blocked, by calculating the pixel coordinates of the light spot corresponding to the blocked target point, the problem of interruption in the positioning of the roadheader can be solved. It ensures the real-time positioning of the roadheader, and the positioning error meets the actual positioning requirements of the roadheader.
Comparative study on stress acoustic emission changes in damage and failure of raw coal and briquette
WANG Linzhi, LIU Dongmei, WANG Shuaiqi, CAO Kuo, GAO Linsheng
2024, 50(5): 125-134. doi: 10.13272/j.issn.1671-251x.2024050017
Abstract:
When studying the relationship between acoustic emission features and coal samples and fractures, both raw coal and briquette can be used as experimental samples. Most coal seams have soft materials, making it difficult to manufacture standard raw coal samples. Therefore, it is common to use briquettes as research samples in experiments. However, coal briquettes change the original structure of coal, affecting its physical and mechanical properties. The applicability of using briquettes instead of raw coal as experimental samples has always been a focus of academic discussion. And currently, there is relatively limited research on the differences in acoustic emission features between raw coal and briquette in pseudo triaxial compression experiments. In order to solve the above problems, pseudo triaxial compression acoustic emission experiments are conducted on raw coal and briquette, with a focus on discussing and analyzing mechanical properties, fracture modes, spatiotemporal evolution of acoustic emission, frequency band energy distribution, nonlinear features, and other aspects. The results show that the acoustic emission energy released during the loading process and the total peak stress energy are closely related to the strength of the coal sample. The raw coal mainly exhibits a mixed failure mode of shear and tension, while the briquette mainly exhibits a tensile axial crack failure mode. The acoustic emission positions of coal samples correspond to their macroscopic fracture morphology, but their occurrence time and spatial distribution are different. In the pre peak loading stage, the acoustic emission signal of raw coal is relatively small, while the acoustic emission response of briquette is intense and reaches its maximum value at the peak stress moment. Through wavelet packet analysis, it is found that the energy distribution of the acoustic emission frequency band of briquette is smaller than that of raw coal. The acoustic emission signals of raw coal are mainly concentrated in the frequency range of 10-120 kHz, while the acoustic emission signals of briquette only jump in the frequency range of 0-100 kHz, indicating that the micro fracture scale of briquette is larger than that of raw coal. 90% of the waveform energy of raw coal and briquette is active at 0-150 kHz. When the loaded sample approaches instability failure, i.e. around 99% of the peak stress, the Hurst index of the acoustic emission signals of raw coal and briquette are both greater than 0.5. It indicates a long-term correlation between the acoustic emission time series and the loading process.
Experimental study on the mechanical and acoustic emission features of frozen single fractured sandstone under drop hammer impact
HE Xinyao, CHANG Yuan, REN Fuqiang
2024, 50(5): 135-141, 156. doi: 10.13272/j.issn.1671-251x.2023110021
Abstract:
Mining rock masses in high-altitude cold regions can experience instability due to low temperature environments and dynamic load disturbances. Existing research mostly focuses on the static features of fractured sandstone under different freezing temperatures. Considering the influence of engineering excavation, further research is needed to investigate the mechanical and acoustic emission features of frozen fractured sandstone under dynamic loads. Therefore, a drop hammer impact test is conducted on frozen single fractured sandstone. The mechanical and acoustic emission features of frozen single fractured sandstone are analyzed using acoustic emission monitoring technology. The experimental results show the following points. ① An increase in the inclination angle of the crack will cause an increase in the rebound amplitude of the strain time curve before the peak strain. The crack will change from being distributed on both sides of the crack to being distributed on both ends of the crack. After the drop height of the hammer increases, the strain time curve shows a significant bimodal rebound before the strain peak, and the damage is significantly intensified. A decrease in freezing temperature will lead to an earlier onset of strain peak and an increase in strain peak. ② The propagation of microcracks has stage features, corresponding to strong microcracking activity at the peak strain and accompanied by intense energy release. ③ The activity of microcracking activity increases first and then decreases with the increase of inclination angle of the crack. The drop height of the hammer increases, and the intensity of microcracking activity gradually decreases. The decrease in freezing temperature leads to an earlier occurrence of microcracking activity. ④ Micro cracks are mainly tensile cracks, corresponding to macroscopic failure modes. ⑤ The sharp increase in entropy value is a precursor to sandstone failure and can be used as a warning indicator for dynamic instability of sandstone.
A multi-modal detection method for holding ladders in underground climbing operations
SUN Qing, YANG Chaoyu
2024, 50(5): 142-150. doi: 10.13272/j.issn.1671-251x.2024010068
<Abstract>(94) <HTML> (22) <PDF>(15)
Abstract:
Currently, most research on recognizing unsafe behaviors of underground personnel focuses on improving precision through computer vision. However, underground areas are prone to occlusion, unstable lighting, and reflection, making it difficult to accurately recognize unsafe behaviors using computer vision technology alone. Especially, similar actions such as climbing ladders and holding ladders during climbing operations are easily confused during the recognition process, posing safety hazards. In order to solve the above problems, a multi-modal detection method for holding ladders in underground climbing operations is proposed. This method analyzes surveillance video data from two modalities: visual and audio. In terms of visual modality, the YOLOv8 model is used to detect the presence of ladder. If there is a ladder, the position coordinates of the ladder are obtained, and the video segment is put into the OpenPose algorithm for pose estimation to obtain the features of various skeletal joint points of the human body. These skeletal joint point sequences are then placed into improved spatial attention temporal graph convolutional networks(SAT-GCN) to obtain human action labels and their corresponding probabilities. In terms of audio modality, the PaddlePaddle automatic language recognition system is used to convert speech into text, and the bidirectional encoder representations from transformers (BERT) model is used to analyze and extract the features of text information, so as to obtain the text label and its corresponding probability. Finally, the information obtained from the visual and audio modalities is fused at the decision-making level to determine whether there is a dpersonnel holding ladders for underground climbing operations. The experimental results show that in action recognition based on skeleton data, the optimized SAT-GCN model improves the recognition precision of three types of actions: holding, climbing, and standing by 3.36%, 2.83%, and 10.71%, respectively. The multi-modal detection method has a higher recognition accuracy than the single modal method, reaching 98.29%.
A maintenance guidance system for coal mine electromechanical equipment based on improved YOLOv5s
XU Jun, ZHAO Xiaohu, HOU Nianqi, WANG Jie, LIU Yulin
2024, 50(5): 151-156. doi: 10.13272/j.issn.1671-251x.2023090069
Abstract:
In order to solve the problems of large workload and low versatility of QR code labelling and complex implementation and difficult deployment of existing no-registration recognition methods in the auxiliary maintenance of coal mine electromechanical equipments, a coal mine electromechanical equipments maintenance guidance system based on improved YOLOv5s is proposed. The system consists of a equipment no-registration recognition module, a fault maintenance guidance module, and a remote expert access guidance module. The equipment no-registration recognition module collects images of faulty equipments through the camera on HoloLens glasses, and analyzes and processes them through an improved YOLOv5s image recognition algorithm to recognize the faulty equipment model. The fault maintenance guidance module automatically matches and calls the preset mixed reality disassembly and assembly model based on the model of the faulty equipment, forming a maintenance guidance solution. The remote expert access guidance module achieves interaction between remote experts and on-site maintenance personnel through audio and video sessions, virtual annotation, and other methods. In order to ensure an immersive experience for users when using mixed reality equipment, ShuffleNetV2 is used to replace the Backbone in YOLOv5s to obtain the YOLOv5s-SN2 network, which reduces the number of model parameters and computational overhead. The experimental results show that YOLOv5s-SN2 has a slight decrease in precision compared to YOLOv5s, but the number of floating-point operations per second (FLOPS) has decreased from 16.5×109 to 7.6×109, and the number of parameters has decreased from 15.6×106 to 8.2×106. Among the YOLO series models, YOLOv5s-SN2 has the best performance. Taking the three leaf Roots blower as an example to verify the overall effectiveness of the system, the results show that YOLOv5s-SN2 can quickly recognize the motor model, call the matching virtual model and maintenance process. The remote experts can assist on-site personnel in electromechanical equipment maintenance through methods such as audio and video access and annotation.