Xiaofeng Yi

Magnetic resonance technology, owing to its direct qualitative and quantitative detection capability for water, shows broad application prospects in detecting hidden water hazards in tunnels. However, when applied to water hazards located on tunnel sidewalls, conventional magnetic resonance methods are constrained by the quasi-full-space geometry of tunnels, making it difficult to achieve spatial localization of the target water bodies. To address this bottleneck, this paper proposes a magnetic resonance localization method based on dual-antenna cooperative detection, aiming to overcome the quasi-full-space limitation and achieve effective localization of sidewall water hazards. First, the theoretical model of the tunnel magnetic resonance dual-antenna cooperative localization detection is derived. Second, to enhance detection depth and efficiency, the antenna configuration suitable for sidewall detection is optimized through comparative analysis. On this basis, using the optimized antenna structure, the magnetic resonance signal responses of hidden water bodies under different occurrence conditions are systematically simulated and comparatively analyzed. Meanwhile, field experiments are conducted to validate the practical performance of the proposed cooperative detection method. The results demonstrate that the proposed method can effectively improve spatial localization accuracy, providing a feasible technical approach for detecting water hazards in tunnel engineering

To address the narrow applicability, low programmability, and bulky size of traditional centralized DC electrical resistivity instruments for detecting disaster-inducing water-bearing formations in coal mine floors, this paper designs an STM32-based intrinsically safe mining electrical resistivity monitoring system. The system enables one-time deployment and long-term monitoring, meeting underground operational needs. With rich peripherals, low power consumption, and high programmability, the STM32 is ideal for intrinsically safe system design. Taking intrinsic safety as the core prerequisite, the hardware design of the host and electrode converter is completed. The host handles signal transmission and data acquisition, comprising the STM32 control chip, 24-bit AD acquisition circuit, and H-bridge transmitting circuit. The electrode converter manages electrode state switching, featuring the STM32 control chip and solid-state relay array circuit. Experimental results confirm that the system accurately acquires subsurface electrical information, characterizes geological structures, and complies with intrinsic safety standards.

In the process of long-term monitoring of the coal seam floor of a coal mining face using electrical resistivity tomography (ERT), the data loss caused by electrode disconnection adversely affects early warning of water inrush and prevents the identification of hidden dangers, hindering safe production. Due to the particularity of the monitored environment, the maintenance of offline electrodes may not be timely. Therefore, how to deal with the loss of measured data caused by electrode disconnection has become a problem that must be solved in the long-term monitoring process. In this paper, we analyze the effect of electrode disconnection on the measured data. Then, the principle of the general regression neural network (GRNN) algorithm is introduced. The missing values in the measured data are predicted using the GRNN algorithm. The results of verification experiments conducted in a water tank show that when the original data integrity is 82.96%, the predicted data accuracy reaches 91.46%, and when the original data integrity is only 55.56%, the predicted data accuracy still reaches 82.45%. Finally, actual applications of the proposed method are carried out on coal mining faces. A set of data with an integrity of 73.8% is predicted. Compared with the measured data when all the electrodes are online, the accuracy of the predicted data is 85.18%. The accuracy of the data predicted using the proposed method is 14.99% higher than that of the data predicted using the commonly used mean value interpolation method.

Water inrush of coal seam floor is usually accompanied by the upward extension of fractures in coal seam floor and the filling of fractures by confined water. Time-lapse electrical resistivity tomography (TL-ERT) is sensitive to low-resistivity anomalous bodies such as water, so it can be used to realize dynamic tracking of the upward extension process of water-bearing fractures. However, conventional TL-ERT methods cannot determine which points in the measured profile are the risk points of water inrush, so a naive Bayes classification algorithm based on pseudo-random matrix is proposed to solve this problem. By using this method, the probability contour map of the measured coal seam floor can be obtained, and the measured point whose probability value exceeds the threshold value of water inrush risk will be warned. Physical simulations show that the probability contour map generated by the proposed method can effectively display high-risk areas. The actual monitoring is carried out on a coal mining face, and the threshold value of water inrush risk is calculated as 0.45 according to the water inrush coefficient. Since there is no measured point whose probability value exceeds the threshold value, there is no need to issue water inrush early warning.

Accurate subsurface velocity modeling is of fundamental scientific and practical significance for seismic data processing and interpretation. However, conventional depth-domain methods still face limitations in physical consistency and inversion accuracy. To overcome these challenges, this study proposes a deep learning-based seismic velocity modeling approach in the time domain. The method establishes an end-to-end mapping between seismic records and velocity models directly in the time domain, reducing the nonlinear complexity of mapping time-domain data to depth-domain models and improving prediction stability and accuracy. Synthetic aquifer velocity models were constructed from representative stratigraphic features, and multi-shot seismic records were generated through forward modeling. A U-Net network was employed, taking multi-shot seismic records as input and time-domain velocity fields as output, with training guided by a mean squared error (MSE) loss function. Experimental results show that the proposed strategy outperforms conventional depth-domain approaches in aquifer structure identification, velocity recovery, and interlayer contrast depiction. Quantitatively, significant improvements in MSE, peak signal-to-noise ratio, and structural similarity index indicate higher reconstruction reliability. Overall, the results confirm the effectiveness and potential of the proposed time-domain framework for aquifer velocity inversion and its promise for intelligent seismic velocity modeling.

The increasing prevalence of retinal diseases fuels the demand for automated diagnostic technologies. However, the sparsity and multiscale features of retinal optical coherence tomography (OCT) images, as well as the high demands on the applicability of detection algorithms in practical settings, pose great challenges to the accurate and robust retinal OCT image classification algorithms. To address these issues, this article introduces a frequency-aware network (FANet) for retinal OCT-based disease classification, leveraging complementary spatial- and frequency-domain information. FANet leverages a multiscale Laplacian pyramid (MSLP) to capture fine-grained details in the spatial domain and a harmonic feature enhancement (HFE) module to model broader contextual information in the frequency domain. These diverse features are integrated through a feature fusion unit (FFU) to form a comprehensive feature representation. Our experiments demonstrate that FANet outperforms existing state-of-the-art (SOTA) models, achieving a 2.3% accuracy improvement on our retinal OCT dataset while maintaining low computational complexity. Furthermore, the extended experiments comprehensively demonstrate FANet’s robustness to noise interference, strong generalization capability across different datasets, adaptability to imbalanced multiclass classification scenarios, and sensitivity to varying amounts of training data.

Optical Coherence Tomography (OCT) is a high-resolution 3D imaging technique widely used in medical diagnostics. However, its increasing resolution leads to massive data volumes, creating challenges in storage, transmission, and real-time access—issues that are further intensified in Internet of Things (IoT) environment. In such settings, where portable OCT devices operate at the edge in remote or resource-limited environments, efficient and secure data handling is critical for real-time diagnostics, remote consultations, and cloud integration. Existing compression methods often struggle to balance efficiency, image quality, and the preservation of diagnostic features, limiting their applicability in constrained IoT infrastructures. To address these challenges, we propose OCT-CSNet, a deep compressed sensing (CS) model tailored for OCT images. By integrating deep learning with traditional CS theory, our model leverages non-parametric prior features to guide both compression and reconstruction stages. In the compression stage, the dual-path compression module efficiently captures the inherent sparsity and structural coherence of OCT images while preserving critical differential information. In the reconstruction stage, we fuse the Proximal Gradient Descent (PGD) algorithm with a deep neural network, incorporating data-driven priors into the iterative optimization process. Extensive experiments show that OCT-CSNet can improve the reconstruction quality under different compression ratios while retaining semantic features that are critical to downstream tasks. In addition, its high computational efficiency and low memory footprint also provide potential for its application in IoT-driven imaging systems.

The limited space within the tunnel constrains the size of the antenna for NMR detection, thereby significantly impacting the signal-to-noise ratio (SNR) of NMR signals. Insufficient SNR data poses substantial challenges to obtaining reliable NMR signals. The paper presents a novel approach to address the challenge of strong background noise in tunnel environments and low SNR data by incorporating the ground multi-channel remote reference denoising method into tunnel NMR advance detection. Specifically designed for narrow tunnels, a multi-channel non-coaxial and non-coplanar remote reference denoising method is proposed. Firstly, the effectiveness of the non-coaxial, non-coplanar remote reference denoising method is verified in the laboratory environment. Secondly, the correlation between the detector antenna and the reference antenna is calculated theoretically to ensure the significant correlation between the detector antenna and the reference antenna. Finally, two processing methods of reference denoising and non-reference denoising are carried out respectively by combining the tunnel detection data. By comparing the inversion results and the engineering construction results, the effectiveness of non-coaxial and non-coplanar remote reference denoising methods in tunnel NMR detection is proved, which provides relevant research support for expanding the application of tunnel NMR detection technology.

Water damage seriously threatens the safe production of coal mines, so it is necessary to carry out advanced detection to determine the hydrogeological situation, and the preliminary survey often involves the drilling of on-site drill holes in the tunnel. The use of directional drill holes, combined with advanced geophysical prospecting technology, enables advanced water disaster detection with long distance and high precision and is independent of the tunnel environment influence. The transient electromagnetic method (TEM) is highly sensitive to low-resistivity anomalies and plays a crucial role in water damage detection. To address the size limitation of borehole detection, in this study, a small rectangular multi-turn loop borehole advanced detection method was developed (borehole TEM, BTEM) to detect low-resistivity anomalies within 10 m of the borehole in the radial direction. To satisfy the size requirements for borehole detection and the detection distance, a small rectangular multi-turn loop device with a width of 6 cm and a length of 50 cm was designed. To resolve the issue of self-inductance and mutual inductance enhancement caused by multi-turn coils, a uniform full-space low-resistivity abnormal body model was established using the Ansys Maxwell software, and we analyzed the vertical magnetic field component of the rectangular multi-turn small loop at different time points and the transient electromagnetic response of the different turns. Then, we determined the appropriate parameters for the transmitting and receiving device. The developed method was applied to several different experimental scenarios to obtain the electrical distribution of the anomalous body in front of the device, and the measured data were inverted and interpreted to obtain the apparent resistivity-depth profile. The results demonstrate that the inversion results align well with the actual situation, confirming the effectiveness of the BTEM. This research offers a potential solution for borehole advance detection and provides a solid theoretical foundation for further studies.