Accurate initial alignment for shearers in low-speed operation within underground environment remains a critical challenge. Conventional initial alignment methods meet accuracy requirements but suffer from slow convergence, limited accuracy, and heading angle estimation errors due to Earth's rotation, gyroscopic drift, and dependence on external references. This paper proposes a novel alignment method that integrates Ultrawideband (UWB) positioning data with odometer measurements for improved alignment accuracy. The method incorporates external parameters such as lever arm offsets and installation angles into an optimal fusion algorithm. Additionally, a robust ranging strategy based on the two-stage iteratively reweighted least squares (TS-IRLS) method is developed to mitigate the effects of non-line-of-sight (NLOS) errors and improve UWB positioning accuracy. Finally, simulation and field test, conducted on a mobile platform mimicking shearer kinematics in a simulated tunnel environment show that the proposed UOBOBA method improves alignment accuracy by 40.35% in the East and 60.75% in the North directions, with pitch and roll errors within 0.02 degrees and heading errors within 0.35 degrees. The method achieves rapid convergence within 70 s, demonstrating its effectiveness in improving alignment performance under low-speed conditions and indicating that the scheme provides a promising solution for the navigation of underground mining equipment.

Accurate positioning of coal shearers is a critical prerequisite for realizing intelligent and autonomous operations in underground coal mines. However, existing localization methods face challenges, including non-line-of-sight errors, measurement outliers, and long-term inertial heading drift. To tackle these challenges, this study proposes an adaptive maximum correntropy Kalman filter fusion method, which integrates inertial navigation systems (INSs), ultra-wideband (UWB) measurements, and odometer data. Specifically, a sliding-window adaptive kernel bandwidth estimation scheme is designed to embed the maximum correntropy criterion into a square-root form of the Kalman filter (AMCC-SRCKF), improving robustness against nonGaussian noise. In addition, a geometric yaw constraint strategy based on dual UWB tags is introduced to suppress INS heading drift. The fusion framework also incorporates a robust weighted total least squares method, which accounts for covariance uncertainties in anchor node coordinates and range measurements, thereby enhancing the reliability of UWB positioning under complex conditions. The simulation and industrial experimentation results demonstrate that the proposed AMCC-SRCKF method significantly improves both positioning accuracy and consistency compared to existing methods, making it well suited for precise shearer positioning in underground coal mines.

Accurate identification of coal-rock characteristics is crucial for optimizing drilling parameters, improving operational efficiency, and ensuring safety in downhole pressure relief drilling with anti-impact drilling robots (AIDR). This study proposes a multi-attention-based methodology to enhance the speed and accuracy of coal-rock recognition using measurement-while-drilling (MWD) data. Coal-rock specimens with varying hardness were prepared to simulate actual downhole conditions, and MWD data were collected using a test bench. Combined with drilling conditions, these data were used to classify coal-rock properties and develop a comprehensive drilling measurement database. To improve recognition accuracy, the raw data were processed through low-pass filtering and normalization before being input into the ML-Multi-Head Attention network for training. Leveraging the self-attention mechanism, the model effectively captures global dependencies and long-term temporal patterns, addressing the complexity of dynamic MWD data under unknown coal-rock conditions. Furthermore, the multilayer perceptron enhances feature extraction through nonlinear mapping and classification, improving overall model performance. Finally, through transfer learning, the model was applied to actual AIDR operations. Experimental results demonstrate that the proposed method achieves 98.6% identification accuracy and a detection speed of 0.045 seconds, enabling efficient and precise differentiation of coal-rock drilling conditions.

Vacuum contactors frequently malfunction owing to substantial loads and strong vibrations in industrial equipment of mining, metallurgy, electric power, transport, etc., thus, causing reduction in productivity, damage to equipment, and even personal injury. However, it is challenging to diagnose the status of the vacuum contactor in real time, due to the unclear nature of the vacuum contactor failure mechanism, the strong dynamic autocorrelation between variable data, and the presence of data redundancy across operating conditions. Therefore, an online fault diagnosis method is proposed for vacuum contactors, consisting of mechanism analysis, feature mining, and data-driven diagnosis. Initially, a dual-coil electromagnet equivalent model is established for the vacuum contactor to explore the fault mechanism, supporting algorithm design. Then, Dynamic Fault-relevant Component Analysis (DFRCA) is proposed for fault diagnosis of vacuum contactor, which handles data dynamics and remove redundant information that is irrelevant to diagnosis. Finally, the probability of fault occurrence is quantified by DFRCA-driven Bayesian theory. The experimental results demonstrate the effectiveness and superiority of the proposed diagnostic method.

As the demand for efficient and automated coal mining continues to rise, accurately identifying the mixing ratio of coal and gangue becomes a crucial step in enhancing the intelligence of the top-coal mining process. Currently, the main challenges stem from the background noise. These factors complicate the effective collection of key signals. Additionally, existing identification methods often rely on single-signal analysis, which limits their accuracy and robustness. To address these issues, this paper presents a novel method for coal-gangue recognition based on key time-frequency information from electromagnetic waves combined with decision fusion. First, a propagation model for electromagnetic waves in coal-gangue mixtures is established. This model reveals the key time-frequency domain information during the propagation process and is subjected to numerical simulation. Subsequently, a coal-gangue identification model is constructed based on the feature extraction of key time-frequency domain information. To further enhance identification accuracy, a decision fusion method based on an improved Analytic Hierarchy Process (IAHP) is designed. Finally, a simulation experimental platform for top-coal mining is built to validate the proposed method. Experimental results indicate that this method achieves a high level of accuracy in coal and gangue identification, providing effective technical support for automated coal mining.

With the advancement of fully mechanized coal mining, demand for high precision localization in underground environments has grown substantially, as both autonomous equipment and human operators rely on accurate positional data for safety and efficiency. Ultra wideband (UWB) technology is a preferred solution in Global Navigation Satellite System (GNSS) denied settings due to its high ranging accuracy and strong multipath resistance. However, non-line-of-sight (NLoS) propagation caused by obstacles and spatially varying accuracy requirements remain key challenges. This study proposes an adaptive UWB anchor deployment framework that integrates bidirectional long short term memory (BiLSTM) based region prediction, Cramér–Rao lower bound (CRLB) based accuracy modeling, and particle swarm optimization (PSO). Monte Carlo simulations using real world datasets under both line-of-sight (LoS) and mixed LoS/NLoS conditions were conducted, with performance evaluated via the root error bound (REB). Results show that the method achieves median REB reductions of 23.94

Under complex coal-rock formations and variable load conditions, accurate recognition of drilling operation states is essential for the safety and efficiency of hydraulic drilling. To overcome the limited robustness of purely mechanistic or purely data-driven approaches in strongly nonlinear drilling environments, this study proposes a hybrid framework that integrates a physics-based mechanistic drilling dynamics model with a Transformer-long short-term memory data-driven model under a Bayesian inference scheme. The mechanistic model is established through analyses of drill bit-coal-rock interactions and drill string dynamics, while key parameters are estimated using an improved INIR-Harris hawks optimization algorithm, supporting reliable characterization of drilling instability states. The hybrid model is validated on a multivariate time-series dataset under different drilling types and geological conditions, achieving over 98% recognition accuracy with a false alarm rate (FAR) below 1.5% across multiple drilling types, and over 97% accuracy with a FAR below 2% under varying geological conditions. These results demonstrate the effectiveness, robustness, and engineering applicability of the proposed method for hydraulic drilling state monitoring.

The anti-punching drilling robot is essential for pressure relief operations in high-stress mines. Its accuracy in identifying coal and rock properties directly impacts drilling efficiency and pressure relief effectiveness. This paper addresses the current reliance on manual experience for coal-rock drilling state recognition, which suffers from low accuracy, long response times, and inability to meet unmanned drilling and pressure relief requirements. Based on a one-dimensional convolutional neural network (1DCNN) and long short-term memory (LSTM) combined with simulation experiments, a coal-rock properties recognition method for the drilling process is proposed. The model's recognition accuracy is enhanced by incorporating a convolutional block attention mechanism (CBAM), and the improved dung beetle optimization (IDBO) algorithm is employed to further optimize the model's hyperparameters, determining the optimal network parameter combination. Coal and rock drilling simulation test bench is constructed, featuring six types of representative coal and rock test blocks. Four categories of sensor signals, including rotation speed, rotation torque, feed speed, and feed pressure, are collected to conduct corresponding comparative testing and analysis. Results demonstrate that the proposed method achieves high coal-rock drilling recognition accuracy of 97.00%, significantly outperforming 1DCNN, 1DCNN-LSTM, logistic regression, support vector machines (SVM), decision trees, random forests, K-means clustering, and Transformer approaches.

Conveyor belt deviation is one of the common factors leading to belt conveyor failure. Real-time detection of conveyor belt deviation is crucial for realizing intelligent detection of the belt conveyor. This paper proposes a Lightweight and Efficient Conveyor Belt Deviation Detection algorithm based on YOLOv7 for edge computing devices. This algorithm incorporates the Mixed Local Channel Attention (MLCA), ELAN-Depthwise Separable Convolution (ELAN-DW), and SPP-Lightweight (SPPL) modules, along with a dynamic decoupled head (D-Dhead) that utilizes both channel and spatial attention mechanisms. According to the self-built dataset, the model’s test accuracy is 0.936. The results show that the algorithm has strong feature extraction capabilities, high real-time performance, and high detection accuracy in scenarios with blurring and insufficient lighting. The algorithm proposed in this paper provides a reference for research on intelligent detection technology for conveyor belt deviation and offers a new solution for edge intelligent fault diagnosis method of belt conveyor, which has important engineering application value.

Coal-rock characteristic identification is a crucial technology for realizing shearers intelligent control. To enhance the intelligence level of shearers, this paper presents a novel approach for identifying coal-rock characteristics using the temperature field of the coal wall. First, we introduce an enhanced You Only Look Once (YOLO) model, termed Temperature sensitive region-YOLO (TSR-YOLO), specifically designed to extract temperature-sensitive regions within the coal wall temperature field. In terms of structural design, TSR-YOLO innovatively incorporates the Cross Stage Partial FasterNet (C3k2-FasterNet) into the backbone network to accelerate feature extraction and devises the Cross-Stage Partial Kolmogorov-Arnold Network (C3k2-KAN) to enhance detailed feature representation. In the bottleneck network, it integrates the Dynamic Convolution (DynamicConv) module to capture broader and more complex feature, as well as the Variational Overlapping Vision-Generalized Spatial Cross-Stage Partial (VoV-GSCSP) module to enhance computational efficiency and optimize feature extraction performance. Subsequently, we propose a coal-rock characteristic identification method utilizing the ConvNeXt model. To validate its effectiveness, we conduct ablation and comparative experiments using experimental data obtained from infrared images of the coal wall during the shearer cutting process. The results indicate that proposed approach achieves a mean Average Precision (mAP) reaching 98.1% and an inference speed of 2 ms per image in identifying temperature-sensitive regions of the coal wall temperature field, surpassing other comparative models. Furthermore, the accuracy of coal-rock property identification reaches 97.6%. This study presents a new approach to coal-rock characteristic identification methods.

Accurate detection of pressure-relief boreholes is crucial for evaluating drilling quality and monitoring safety in coal mine roadways. Nevertheless, the highly challenging underground environment—characterized by insufficient lighting, severe dust and water mist disturbances, and frequent occlusions—poses substantial difficulties for current object detection approaches, particularly in identifying small-scale and low-visibility targets. To effectively tackle these issues, a lightweight and robust detection framework, referred to as YOLO-DFBL, is developed using the YOLOv11n architecture. The proposed approach incorporates a DualConv-based lightweight convolution module to optimize the efficiency of feature extraction, a Frequency Spectrum Dynamic Aggregation (FSDA) module for noise-robust enhancement, and a Biformer (Bi-level Routing Transformer)-based routing attention mechanism for improved long-range dependency modeling. In addition, a Lightweight Shared Convolution Head (LSCH) is incorporated to effectively decrease the overall model complexity. Experimental results on a real coal mine roadway dataset demonstrate that YOLO-DFBL achieves an mAP@50:95 of 78.9%, with a compact model size of 1.94 M parameters, a computational complexity of 4.7 GFLOPs, and an inference speed of 157.3 FPS, demonstrating superior accuracy–efficiency trade-offs compared with representative lightweight YOLO variants and classical detectors. Field experiments under challenging low-illumination and occlusion environments confirm the robustness of the proposed approach in real mining scenarios. The developed method enables reliable visual perception for underground drilling equipment and facilitates safer and more intelligent operations in coal mine engineering.

Path tracking control methods for mobile robots show poor applicability and limited performance in coal mine underground with complex scene interference, such as slip and unstructured pavement impacts. Therefore, a kinematic model considering slip is used as the basis, combined with the lateral response in the dynamics model, to build dynamics-kinematics hybrid model with the rotational acceleration as the virtual control input, which utilizes generalized disturbances to represent the uncertainty induced by robot slippage. Then, the extended state observer (ESO) is proposed to estimate the traveling state and generalized disturbance, which addresses the difficulty of directly measuring velocity and disturbance information. Furthermore, the hierarchical control strategy is proposed based on path tracking controller and track speed controller. The path tracking controller is developed for realizing the mobile robot kinematic path tracking, and the track speed controller is proposed for the dynamics track velocity control, which possesses excellent interference immunity and traveling control accuracy. Finally, the experimental study on the path tracking control of mobile robot is carried out in the simulated coal mine underground. The results show that the proposed method can well realize the traveling control under the complex scene interference. The RMS values of trajectory tracking errors are 5.86 cm, 8.22 cm, and 0.0972 rad, and the chattering phenomenon is obviously improved.

The liquid cooling plate is critical for ensuring the safe and efficient operation of transformers, particularly in confined environments. In recent years, researchers have made significant strides in enhancing transformer cooling efficiency to address challenges posed by aging infrastructure. However, conventional serpentine-channel cooling plates suffer from significant limitations in thermal performance and flow efficiency. This study proposes a density-based topology optimization framework to overcome fixed-configuration constraints, utilizing the SIMP (Solid Isotropic Material with Penalization) method where material density serves as a continuous design variable. By penalizing intermediate densities through the power-law relation, the method drives solutions toward near-binary material distributions while maximizing heat exchange and minimizing flow resistance. The results provide a novel approach for enhancing heat dissipation in transformer cooling systems, offering potential improvements in both energy efficiency and operational safety.

Drilling conditions serve as reliable indicators for evaluating drilling system performance. The recognition methodology forms a critical foundation for intelligent drilling and plays a significant role in preventing borehole accidents such as pipe sticking. This study proposes a recognition method for typical drilling conditions based on bit motion characteristics, developing intelligent algorithms that utilize fundamental motion parameters and drive signals as inputs to achieve accurate condition identification. Firstly, micro-drilling experiments were conducted under various pressure and rotational speed combinations. The variation characteristics of drilling speed, rotational speed, torque, and drilling pressure during the process were analyzed. Based on rotational and feed motion features at different stages, four typical drilling conditions were defined: stable drilling, slowing-down drilling, intermittent jamming, and pipe sticking. Subsequently, using actual coal mine drilling data, feature indicators for recognition models were selected through collective statistical characteristics and rolling statistical properties. Recognition models employing support vector machine (SVM), convolutional neural network (CNN), and long short-term memory (LSTM) networks were constructed and trained. Test results demonstrate that the LSTM-based model achieved the highest recognition accuracy (84% overall). It attained 100% accuracy in identifying slowing-down drilling and intermittent jamming conditions, though recognition accuracy for stable drilling remained relatively low. This research contributes to enhanced intelligent recognition of drilling conditions and provides theoretical support for advancing intelligent drilling technologies.

The environment of underground coal mines is characterized by low-illumination intensity and high dust concentration, causing captured coal-gangue images to suffer from problems such as low contrast, blurred details, and color distortion. These degradation issues severely restrict the performance of automatic coal-gangue recognition and sorting systems. To address these challenges, this paper proposes a coal-gangue image enhancement algorithm based on an improved HVI-CIDNet model. Targeting the uneven lighting and high-dust environment, the proposed method utilizes the HVI color space to effectively decouple the intensity and color information of coal-gangue images, constructing a dual-branch decoupled deep learning network. A learnable dual-residual cross-attention block is designed to optimize feature fusion. Furthermore, by introducing a frequency domain enhancement block and optimizing the branch structure, the network synergistically enhances the recovery of weak texture details on coal-gangue surfaces. Simultaneously, a dual color space constraint loss is integrated to ensure the stability of enhancement effects under complex environmental conditions. Experimental results demonstrate that the proposed algorithm achieves a peak signal-to-noise ratio of 32.498 dB, structural similarity of 0.933, and learned perceptual image patch similarity of 0.126 on a self-constructed coal-gangue dataset, outperforming mainstream comparative methods in all metrics. Moreover, the enhanced images significantly improve the accuracy of downstream detection models, achieving average detection accuracies of 89.2% and 95.6% on vision transformer and YOLOv12 models, respectively. This research provides an effective solution for coal-gangue image enhancement under complex working conditions.

Accurate recognition of coal-rock properties (CRPs) while drilling is a critical prerequisite for ensuring the intelligent control and stable operation of antipunching drilling robot. The primary challenges in industrial applications are potential safety hazards and single drilling mode in coal mine roadways, resulting in samples with insufficient number, unbalanced distribution, and background interference. However, most current CRPs recognition methods rely on sufficient and balanced samples, limiting their application efficiency and scope. To this end, we propose a selective kernel transformer with model-data fusion loss (SKformer-MFL) model with data correction, which generates simulated signals via electromagnetic simulation model to assist the CRPs recognition while drilling under limited samples. First, a coal-rock drilling model based on electromagnetic simulation is established to generate simulated signals as the primary part of training samples. Then, the simulated signals are corrected using denoising diffusion probabilistic model improved with vector-quantized variational autoencoder to minimize the feature discrepancies between simulated and real signals. Finally, the corrected high-fidelity signals are used to train the proposed SKformer-MFL model, where a novel loss is built to handle the imbalanced training dataset through an adaptive weighting mechanism. This model can simultaneously capture long-range correlated features and local mutation multiscale features of the input signals, thereby improving the model’s recognition capability and generalization performance. The experimental results indicate that the proposed method can produce high-fidelity electromagnetic simulation signals and achieve excellent recognition performance for CRPs while drilling under limited samples.

Coal mine drilling rigs serve as critical equipment for exploration and disaster prevention in coal mining operations. Ensuring stability and safety during the drilling process is paramount, as a rational and efficient control scheme contributes significantly to the prevention of severe accidents. However, these rigs are frequently subjected to the complex and variable conditions of the underground coal mine environment. Most existing studies rely heavily on predetermined assumptions, failing to adapt to the dynamic changes of the system in real time. To address this challenge, this paper proposes a self-healing control method based on reinforcement learning. Specifically, mathematical models of key system components are established, and a comprehensive evaluation index for the drilling process is constructed. Subsequently, a Deep Deterministic Policy Gradient (DDPG) control strategy is designed to achieve real-time optimization of the drilling process and active fault avoidance. Experimental results demonstrate that the proposed method yields superior control performance and possesses significant engineering application value

Aiming at the problem that drilling parameters are difficult to adjust in time for the driller due to the complex geological environment in underground coal mines, a drilling parameter control method based on online identification of drillability and multi-objective optimization of drilling parameters is proposed. A drillability grade identification model is established, with rotational speed and torque as input parameters, which can accurately identify the current drilling state. A multi-objective optimization model of the optimal drilling parameters is established with the mechanical specific energy and drilling speed prediction model as the objective functions, and the NSGA-II algorithm and TOPSIS algorithm are used for solutions and decision-making. A fuzzy PID controller is established. For the control of rotational speed parameters and drilling pressure parameters, the advantages and disadvantages of the fuzzy PID control method and the traditional PID control method are compared through simulation and experiments. A control method based on the drillability identification model and the multi-objective optimization model is established. According to the different drillability grades, the drilling parameters are adjusted in time to ensure the normal drilling state. By constantly approaching the optimal parameters through the drilling parameters, the drilling efficiency is improved. Through experimental verification, this control method effectively prevents the occurrence of drilling speed reduction and intermittent sticking and can adjust the drilling parameters to continuously optimize.

At the top-coal caving face, accurately and quickly identifying the content of coal and gangue is an important prerequisite for intelligent and efficient mining. Data scarcity has been caused by the limited conditions in the mine, restricting the recognition of the coal-gangue mix ratio at the top-coal caving face. Therefore, this article proposes a method for identifying the coal-gangue mix ratio under insufficient sample conditions, based on improved auxiliary classifier generative adversarial network (IACGAN) and parallel coordinate and squeeze-and-excite attention (PCSA)-MobileNetV3. First, the coal-gangue vibration data are mapped into 2-D vibration spectral images. Then, the IACGAN is enhanced with a self-attention (SA) mechanism and residual structure to generate diverse, high-quality training samples despite sample deficiency. These generated images are then added to the dataset to expand the dataset size. Finally, the PCSA mechanism is integrated into the MobileNetV3 model to classify coal-gangue vibration spectral images based on the learned spatial and spectral features. The experimental results indicate that the proposed method outperforms superior recognition performance and effectively assists in the mix ratio recognition of coal and gangue under a small sample.

As the core equipment in fully mechanized mining faces, the shearer's autonomous positioning technology plays a crucial role in achieving automation in coal mining. However, due to the harsh conditions in fully mechanized mining faces, the accuracy of the shearer's integrated positioning system deteriorates significantly during short-term sensor failures. To address this issue, this paper presents a non-holonomic constraint algorithm that Integrates shearer cutting information. Experimental results demonstrate that the proposed method maintains high accuracy during short-term failures of the integrated positioning system.