Dynamic monitoring system for mine tunneling process

Abstract

The invention relates to the technical field of mine tunneling, and particularly discloses a dynamic monitoring system for a mine tunneling process, which acquires time sequence data of slope dependent variable and surface subsidence through a sensor group deployed at a key position in a mine, and acquires point cloud data of a tunneling roadway and a mining surface through three-dimensional laser scanning. The data are transmitted to a processing center. And in a processing center, analyzing the data based on a modal prior cross-domain fine-grained interaction analysis technology so as to determine a dynamic monitoring result of the mine tunneling process. And finally, a dynamic monitoring result is sent to terminal equipment of a manager to be displayed, and real-time monitoring information is provided. Therefore, comprehensive analysis is carried out by integrating various data information, more comprehensive and deep understanding of the mine environment is provided, and the identification of potential safety hazards or abnormal conditions is more accurate, so that the safety and reliability of mine operation are effectively improved.

Description

Technical Field

[0001] The present application relates to the field of mine excavation technology, and more specifically, to a dynamic monitoring system for a mine excavation process. Background Art

[0002] Mining is a complex and high-risk engineering activity. Especially during the excavation process, due to the uncertainty of geological conditions, changes in rock mass structure and the influence of the external environment, it is easy to cause safety hazards such as slope instability and surface subsidence. In order to ensure the safety and efficiency of mining operations, a real-time dynamic monitoring system is particularly important.

[0003] However, the traditional method of monitoring the mining process usually requires staff to regularly go to the site to measure and record, which is not only time-consuming but also labor-intensive. Especially in large or complex mines, it is difficult to quickly obtain comprehensive data. In addition, the time interval for manual monitoring is long and cannot provide immediate information feedback; even if automated sensors are used, if there is a lack of efficient communication means and data analysis capabilities, it is difficult to ensure the speed of data transmission and processing, resulting in delayed response. In addition, the data collected by traditional monitoring systems are usually isolated, and there are no effective tools for comprehensive analysis. It is difficult to discover the inherent connection between different parameters and trend prediction, which limits the overall evaluation and early warning capabilities of mining conditions.

[0004] Therefore, an optimized dynamic monitoring system for the mine excavation process is desired. Summary of the invention

[0005] The present application provides a dynamic monitoring system for the mine excavation process, which can make up for the limitation of insufficient information of a single data type when evaluating the mine condition by integrating multiple data information for comprehensive analysis. This multi-source data analysis method provides a more comprehensive and in-depth understanding of the mine environment, making it more accurate to identify potential safety hazards or abnormal conditions, thereby effectively improving the safety and reliability of mine operations.

[0006] According to one aspect of the present application, a dynamic monitoring system for a mine excavation process is provided, comprising:

[0007] A mine status data acquisition module, which is used to collect a time series data set of slope strain and a time series data set of surface subsidence through a sensor group deployed at multiple key locations inside the monitored mine;

[0008] A mine tunnel laser scanning module is used to perform three-dimensional laser scanning on the tunnel and mining surface of the monitored mine to obtain three-dimensional laser scanning point cloud data of the mine;

[0009] A data transmission module, used to transmit the time series data set of the slope strain, the time series data set of the surface subsidence and the three-dimensional laser scanning point cloud data of the mine to a data processing center;

[0010] A mine data analysis module is used to perform cross-domain fine-grained interactive analysis based on modal priors on the time series data set of the slope strain, the time series data set of the surface subsidence and the mine three-dimensional laser scanning point cloud data in the data processing center to determine the dynamic monitoring results of the mine excavation process;

[0011] The mine excavation process monitoring module is used to transmit the dynamic monitoring results of the mine excavation process to the terminal equipment of the management personnel for display.

[0012] In the above-mentioned dynamic monitoring system for mine excavation process, the mine data analysis module includes:

[0013] A three-dimensional voxel encoding unit, used for performing three-dimensional voxel encoding on the three-dimensional laser scanning point cloud data of the mine to obtain three-dimensional voxel grid features of the mine;

[0014] A time series feature interactive fusion unit, used for performing time series feature analysis and interactive fusion on the time series data set of the slope strain and the time series data set of the ground subsidence to obtain a slope strain-ground subsidence time series feature interactive fusion representation;

[0015] A cross-domain fine-grained aggregation analysis unit is used to perform cross-domain fine-grained aggregation analysis on the three-dimensional voxel grid features of the mine and the interactive fusion representation of the slope strain-surface subsidence time series features to obtain a cross-modal fine-grained aggregation representation of the mine excavation state;

[0016] A mine excavation process dynamic monitoring unit is used to dynamically monitor the mine excavation process based on the cross-modal fine-grained aggregation representation of the mine excavation state to determine whether there is any abnormality in the excavation process of the monitored mine.

[0017] In the above-mentioned system for dynamic monitoring of mine excavation process, the three-dimensional voxel encoding unit includes:

[0018] A voxel distance encoding subunit, used for performing voxel distance encoding on the mine three-dimensional laser scanning point cloud data to obtain mine three-dimensional voxel grid data;

[0019] The mine three-dimensional feature extraction subunit is used to obtain a mine three-dimensional voxel grid feature map as the mine three-dimensional voxel grid feature by passing the mine three-dimensional voxel grid data through a mine three-dimensional feature extractor based on 3D-CNN.

[0020] In the above-mentioned dynamic monitoring system for mine excavation process, the time series feature interactive fusion unit includes:

[0021] A time series causal correlation feature extraction subunit is used to respectively pass the time series data set of the slope strain and the time series data set of the surface subsidence through a time series encoder based on a void causal convolutional neural network model to obtain a time series causal correlation feature vector of the slope strain and a time series causal correlation feature vector of the surface subsidence;

[0022] The feature interaction fusion processing subunit is used to obtain a slope strain-surface subsidence time series feature interaction fusion representation vector as the slope strain-surface subsidence time series feature interaction fusion representation vector by means of a feature interaction fusion module based on an X-linear attention bilinear pooling network.

[0023] In the above-mentioned dynamic monitoring system for mine excavation process, the cross-domain fine-grained aggregation analysis unit includes:

[0024] A modal kernel feature extraction subunit, used for performing modal kernel feature extraction on the three-dimensional voxel grid feature of the mine to obtain the three-dimensional voxel grid kernel feature of the mine;

[0025] A fine-grained feature aggregation description processing subunit is used to perform fine-grained feature aggregation description processing on the three-dimensional voxel grid features of the mine based on the interactive fusion representation of the slope strain-surface subsidence time series features and the three-dimensional voxel grid kernel features of the mine to obtain a cross-modal fine-grained aggregation representation of the mine excavation state.

[0026] In the above-mentioned dynamic monitoring system for mine excavation process, the modal kernel feature extraction subunit is used to: input the mine three-dimensional voxel grid feature map into the modal kernel feature extraction network to obtain the mine three-dimensional voxel grid kernel feature encoding vector as the mine three-dimensional voxel grid kernel feature.

[0027] In the above-mentioned dynamic monitoring system for mine excavation process, the fine-grained feature aggregation description processing subunit includes:

[0028] The priori cluster center of the mine excavation state determines the secondary subunit, which is used to determine the priori cluster center encoding vector of the mine excavation state based on the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature encoding vector;

[0029] A mine three-dimensional voxel grid feature expansion secondary subunit is used to perform feature decoupling and feature flattening processing on the mine three-dimensional voxel grid feature map to obtain a set of mine three-dimensional voxel grid feature vectors;

[0030] The cross-domain fine-grained aggregation description processes the secondary sub-unit, which inputs the set of the prior cluster center encoding vector of the mine excavation state and the three-dimensional voxel grid feature vector of the mine into the cluster center-based cross-domain fine-grained aggregation descriptor to obtain the cross-modal fine-grained aggregation representation vector of the mine excavation state as the cross-modal fine-grained aggregation representation of the mine excavation state.

[0031] In the above-mentioned dynamic monitoring system for mine excavation process, the mine excavation state prior clustering center determines the secondary subunit, which is used to: cascade the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature coding vector, and linearly transform the cascaded mine excavation state feature vector to obtain the mine excavation state prior clustering center coding vector.

[0032] In the above-mentioned dynamic monitoring system for the mine excavation process, the cross-domain fine-grained aggregation description processing secondary subunit is used to:

[0033] Using a semantic measurement module of an inverse hyperbolic cosine function, a semantic distance measurement is performed between each mine three-dimensional voxel grid feature vector in the set of mine three-dimensional voxel grid feature vectors and the mine excavation state priori clustering center encoding vector to obtain a set of mine excavation state semantic measurement coefficients;

[0034] Using a binary function to perform clustering judgment on the set of semantic measurement coefficients of the mine excavation state to obtain a set of semantic clustering coefficients of the mine excavation state;

[0035] Based on the set of semantic clustering coefficients of the mine excavation state, the set of prior clustering center encoding vectors of the mine excavation state and the set of three-dimensional voxel grid feature vectors of the mine are aggregated to obtain a cross-modal fine-grained aggregated representation vector of the mine excavation state.

[0036] In the above-mentioned dynamic monitoring system for mine excavation process, the dynamic monitoring unit of mine excavation process is used to: pass the cross-modal fine-grained aggregation representation vector of the mine excavation state through a classifier-based dynamic monitor of mine excavation process to obtain monitoring results, and the monitoring results are used to indicate whether there are abnormalities in the excavation process of the monitored mine.

[0037] The present application provides a dynamic monitoring system for the mine excavation process, which collects time series data of slope strain and surface subsidence through sensor groups deployed at key locations inside the mine, as well as point cloud data of excavation tunnels and mining surfaces obtained through three-dimensional laser scanning, and these data are transmitted to the processing center. At the processing center, the above data are analyzed based on the cross-domain fine-grained interactive analysis technology of modal prior to determine the dynamic monitoring results of the mine excavation process. Finally, the dynamic monitoring results are sent to the terminal device of the manager for display, providing real-time monitoring information. In this way, by integrating multiple data information for comprehensive analysis, it is possible to make up for the limitation of insufficient information of a single data type when evaluating the mine condition. This multi-source data analysis method provides a more comprehensive and in-depth understanding of the mine environment, making it more accurate to identify potential safety hazards or abnormal conditions, thereby effectively improving the safety and reliability of mine operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings of the embodiments of the present application are briefly introduced below. Obviously, the drawings described below only relate to some embodiments of the present application, and are not intended to limit the present application.

[0039] Figure 1 A schematic block diagram of a system for dynamically monitoring a mine excavation process according to an embodiment of the present application.

[0040] Figure 2 This is a schematic block diagram of a mine data analysis module in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application.

[0041] Figure 3 This is a schematic diagram of data flow of a mine data analysis module in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application.

[0042] Figure 4 A schematic block diagram of a three-dimensional voxel encoding unit in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application.

[0043] Figure 5 It is a schematic block diagram of a time series feature interactive fusion unit in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application.

[0044] Figure 6 This is a schematic block diagram of a cross-domain fine-grained aggregation analysis unit in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application.

[0045] Figure 7 This is a schematic block diagram of a fine-grained feature aggregation description processing subunit in a dynamic monitoring system for a mine excavation process according to an embodiment of the present application. DETAILED DESCRIPTION

[0046] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present application.

[0047] In view of the above technical problems, in the technical solution of this application, a dynamic monitoring system for the mine excavation process is proposed, such as Figure 1 As shown, the system for dynamic monitoring of mine excavation process includes: a mine status data acquisition module 10, which is used to collect the time series data set of slope strain and the time series data set of surface subsidence through the sensor group deployed at multiple key positions inside the monitored mine; a mine excavation tunnel laser scanning module 20, which is used to perform three-dimensional laser scanning on the excavation tunnel and mining face of the monitored mine to obtain the mine three-dimensional laser scanning point cloud data; a data transmission module 30, which is used to transmit the time series data set of slope strain, the time series data set of surface subsidence and the mine three-dimensional laser scanning point cloud data to the data processing center; a mine data analysis module 40, which is used to perform cross-domain fine-grained interactive analysis based on modal prior on the time series data set of slope strain, the time series data set of surface subsidence and the mine three-dimensional laser scanning point cloud data at the data processing center to determine the dynamic monitoring result of the mine excavation process; a mine excavation process monitoring module 50, which is used to transmit the dynamic monitoring result of the mine excavation process to the terminal device of the management personnel for display.

[0048] Exemplarily, in the mine status data acquisition module 10, a time series data set of slope strain and a time series data set of ground subsidence are collected by a sensor group deployed at multiple key locations inside the monitored mine. It should be understood that during the mine excavation process, due to the complexity and uncertainty of geological conditions, the rock structure may change, resulting in safety hazards such as slope instability or ground subsidence. In order to ensure the safety and efficiency of mine operations, these potential risks must be monitored in real time. Slope strain reflects the changes in stress inside the rock mass, while ground subsidence reflects the surface displacement. Together, the two provide key information about mine stability. Therefore, obtaining a time series data set of these two parameters can help us understand the dynamic changes of the mine more comprehensively and detect anomalies in a timely manner.

[0049] In one embodiment, in order to obtain a time series data set of slope strain and a time series data set of surface subsidence, it is first necessary to select the appropriate sensor type and installation location. Usually, sensors are installed in key parts of the mine, such as the top and bottom of the slope and the surface areas that may be affected. These sensors can be strain gauges, inclinometers, settlement plates and other equipment, which can continuously record the data of slope strain and surface subsidence. In order to ensure the quality and representativeness of the data, the arrangement of sensors should cover different geological units within the entire monitoring range and take into account the influencing factors of different engineering activities. For example, increase the density of sensors in places where large deformations are expected or in areas where problems have occurred historically.

[0050] Exemplarily, in the mine excavation tunnel laser scanning module 20, the excavation tunnel and mining face of the monitored mine are subjected to three-dimensional laser scanning to obtain the mine three-dimensional laser scanning point cloud data. It should be understood that the three-dimensional laser scanning of the excavation tunnel and mining face of the monitored mine is carried out to obtain the mine three-dimensional laser scanning point cloud data because this technology can provide high-precision and comprehensive spatial information, which is essential for understanding the internal structure of the mine and its changes over time. In mining operations, understanding the geometry of the rock mass, fault distribution, fissure development, etc. is the basis for ensuring safe and efficient production. The data obtained by three-dimensional laser scanning can accurately depict the complex topographic features inside the mine, including the shape of the tunnel, the state of the roof and side walls, and the specific position and angle of the mining face. This not only helps engineers to more intuitively grasp the actual situation on site, but also provides solid data support for a series of engineering activities such as subsequent risk assessment, support design, and blasting planning.

[0051] In one embodiment, in order to achieve three-dimensional laser scanning, it is first necessary to select a suitable scanning device and determine the optimal scanning path. Usually, a ground-based three-dimensional laser scanner or a mobile lidar system carried by a drone is used. The former is suitable for fine scanning at fixed locations, while the latter is more suitable for rapid coverage of large areas. In actual operation, technicians will set up multiple scanning stations in key areas according to the specific layout and work requirements of the mine to ensure that the entire monitoring range is covered without blind spots. During the scanning process, the equipment will emit a laser beam and receive the reflected signal, record the spatial coordinates and reflection intensity of each point, and form a dense and accurate point cloud data set.

[0052] Exemplarily, in the data transmission module 30, the time series data set of the slope strain, the time series data set of the surface subsidence, and the three-dimensional laser scanning point cloud data of the mine are transmitted to the data processing center. It should be understood that these different types of data provide different perspectives on the state of the mine: slope strain reflects the changes in the internal rock structure, surface subsidence reflects the surface displacement, and three-dimensional laser scanning captures the specific form of the topography. A single type of data may not provide enough information to accurately assess the status of the mine. Therefore, by integrating multiple data signals and transmitting them to the data processing center for comprehensive analysis, the limitations of different data types can be compensated, the perception of the overall structure of the mine can be enhanced, and strong support can be provided for subsequent risk prediction and management decisions.

[0053] In one embodiment, in order to transmit these data to the data processing center, a stable and reliable communication network needs to be built first. Inside the mine, due to the complex geographical environment and possible signal shielding problems, a combination of wired and wireless methods is usually adopted. For example, optical fiber cables are laid between fixed monitoring sites to ensure high-speed and stable connections; for mobile or difficult-to-wire locations, wireless sensor networks (WSNs) or industrial-grade wireless communication technologies such as LoRa, ZigBee, etc. are used to ensure that data can be transmitted smoothly in complex underground environments. In addition, considering the particularity and safety requirements of mining operations, all communication equipment must have explosion-proof and anti-interference characteristics to adapt to harsh working conditions.

[0054] Exemplarily, in the mine data analysis module 40, at the data processing center, the time series data set of the slope strain, the time series data set of the surface subsidence, and the mine three-dimensional laser scanning point cloud data are subjected to cross-domain fine-grained interactive analysis based on modal priors to determine the dynamic monitoring results of the mine excavation process. It should be understood that in the above-mentioned system for dynamic monitoring of mine excavation process, the process of comprehensively analyzing the time series data set of the slope strain, the time series data set of the surface subsidence, and the mine three-dimensional laser scanning point cloud data at the data processing center to determine the dynamic monitoring results of the mine excavation process is crucial, because each data type provides a different perspective on the state of the mine. For example, the slope strain reflects the stress changes inside the rock mass, the surface subsidence reflects the surface displacement, and the three-dimensional laser scanning can capture the specific form of the topography. A single type of data may not provide enough information to accurately assess the mine condition, and the fusion of multiple data can make up for the limitations of different data types. Therefore, combining this information for analysis can achieve a more comprehensive understanding of the mine environment and help to more accurately identify potential safety hazards or abnormal conditions.

[0055] Specifically, the technical concept of this application is to collect slope strain and surface subsidence in real time, obtain three-dimensional laser scanning point cloud data of the mine, and then introduce data processing and analysis algorithms based on artificial intelligence and deep learning in the back end to capture the slope strain-surface subsidence time series feature interaction representation and mine three-dimensional voxel grid features, and then perform cross-domain fine-grained interactive analysis based on modal priors on these two different types of features about the mine excavation status, so as to extract the cross-modal fine-grained aggregate representation semantics of the mine excavation status, thereby realizing dynamic monitoring and anomaly detection of the mine excavation process. In this way, the real-time and intelligent level of mine excavation process monitoring can be improved, providing strong support for mine safety management and operation.

[0056] Specifically, in the technical solution of this application, if Figure 2 and Figure 3 As shown, the mine data analysis module 40 includes: a three-dimensional voxel encoding unit 41, which is used to perform three-dimensional voxel encoding on the three-dimensional laser scanning point cloud data of the mine to obtain the three-dimensional voxel grid characteristics of the mine; a time series feature interactive fusion unit 42, which is used to perform time series feature analysis and interactive fusion on the time series data set of the slope strain and the time series data set of the surface subsidence to obtain the slope strain-surface subsidence time series feature interactive fusion representation; a cross-domain fine-grained aggregation analysis unit 43, which is used to perform cross-domain fine-grained aggregation analysis on the three-dimensional voxel grid characteristics of the mine and the slope strain-surface subsidence time series feature interactive fusion representation to obtain a cross-modal fine-grained aggregation representation of the mine excavation state; a mine excavation process dynamic monitoring unit 44, which is used to perform dynamic monitoring of the mine excavation process based on the cross-modal fine-grained aggregation representation of the mine excavation state to determine whether there is any abnormality in the excavation process of the monitored mine.

[0057] In one embodiment, Figure 4 As shown, the three-dimensional voxel encoding unit 41 includes: a voxel distance encoding subunit 411, which is used to perform voxel distance encoding on the mine three-dimensional laser scanning point cloud data to obtain mine three-dimensional voxel grid data; a mine three-dimensional feature extraction subunit 412, which is used to pass the mine three-dimensional voxel grid data through a mine three-dimensional feature extractor based on 3D-CNN to obtain a mine three-dimensional voxel grid feature map as the mine three-dimensional voxel grid feature.

[0058] Exemplarily, in the voxel distance encoding subunit 411, since the original mine three-dimensional laser scanning point cloud data is usually very dense and large, directly processing it may result in excessive consumption of computing resources. Through voxelization (i.e., voxel distance encoding), the amount of data can be effectively reduced while retaining the key features of the topography. It is worth mentioning that here, each voxel represents a spatial unit of fixed volume, which helps to simplify subsequent analysis and modeling. At the same time, the voxel grid helps to better capture the shape, boundaries and other geometric characteristics of objects, such as the spatial characteristics of mine tunnels and mining faces, because it is easier to define and calculate local features such as curvature and normals at the voxel level. In addition, for certain specific application scenarios, such as rock structure analysis or excavation face stability assessment, voxelization can help highlight those areas that are meaningful for analysis.

[0059] Exemplarily, in the mine three-dimensional feature extraction subunit 412, the mine three-dimensional voxel grid data is passed through a mine three-dimensional feature extractor based on 3D-CNN to obtain a mine three-dimensional voxel grid feature map as the mine three-dimensional voxel grid feature. It should be understood that the mine three-dimensional voxel grid data is passed through a mine three-dimensional feature extractor based on 3D-CNN (three-dimensional convolutional neural network) to obtain a mine three-dimensional voxel grid feature map as the mine three-dimensional voxel grid feature, in order to automatically learn and extract representative geometric and semantic features from complex three-dimensional point cloud data. The mining environment usually contains a large number of complex and irregular structures, such as tunnels, mining faces, and rock formations. The distribution of these structures in three-dimensional space is crucial to understanding the internal conditions of the mine. Traditional manual feature design methods are difficult to capture such rich information, while 3D-CNN based on deep learning can automatically identify feature patterns that are valuable for mine status assessment. Specifically, 3D-CNN is a convolutional neural network architecture specifically designed for processing three-dimensional data. It can capture local and global spatial relationships at multiple scales, thereby effectively characterizing the characteristics of the internal structure of a mine. When the three-dimensional voxel grid data is input into 3D-CNN, the network applies convolution operations layer by layer, gradually extracts feature representations at different levels, and finally outputs a compact and expressive feature map, namely the three-dimensional voxel grid feature map of the mine. In this process, 3D-CNN learns how to distinguish different geological features through the training phase, such as distinguishing rock types, identifying cracks or fault locations, etc. At the same time, due to its multi-layer structure, 3D-CNN can also capture detailed changes at different scales, which helps to more accurately describe the subtle differences inside the mine. In addition, 3D-CNN also has good generalization capabilities and can provide accurate and reliable feature extraction results even in the face of unseen data. Therefore, processing the three-dimensional voxel grid data of the mine through 3D-CNN can not only improve the degree of automation of feature extraction, but also enhance the understanding of the mine environment, providing a solid foundation for subsequent risk assessment, anomaly detection and safety management decisions.

[0060] In one embodiment, Figure 5As shown, the time series feature interaction fusion unit 42 includes: a time series causal association feature extraction subunit 421, which is used to pass the time series data set of the slope strain variable and the time series data set of the surface subsidence through a time series encoder based on a void causal convolutional neural network model to obtain a slope strain variable time series causal association feature vector and a surface subsidence time series causal association feature vector; a feature interaction fusion processing subunit 422, which is used to pass the slope strain variable time series causal association feature vector and the surface subsidence time series causal association feature vector through a feature interaction fusion module based on an X-linear attention bilinear pooling network to obtain a slope strain variable-surface subsidence time series feature interaction fusion representation vector as the slope strain variable-surface subsidence time series feature interaction fusion representation.

[0061] Exemplarily, in the temporal causal association feature extraction subunit 421, the temporal data set of the slope strain and the temporal data set of the surface subsidence are respectively encoded in a temporal encoder based on a void causal convolutional neural network model to extract the temporal association features of the slope strain and the temporal association features of the surface subsidence, respectively, so as to obtain the temporal causal association feature vector of the slope strain and the temporal causal association feature vector of the surface subsidence. Through the processing of the temporal encoder based on the void causal convolutional neural network model, the receptive field can be expanded without increasing the number of parameters by using the void convolution, and the characteristics of the causal convolution ensure that when processing the data at each time point, only the information of the previous time step is considered, and the data at the future time is not used. In this way, the time long-term dependence and temporal causal relationship in the temporal data set of the slope strain and the temporal data set of the surface subsidence can be captured respectively, which is very important for building a physically reasonable model, especially when analyzing geological processes such as slope deformation, which follow a strict causal order. At the same time, for slope and surface stability monitoring, this helps to understand the impact of historical data on current conditions and predict future trends.

[0062] Exemplarily, in the feature interaction fusion processing subunit 422, it is considered that slope stability and surface settlement are two closely related but independent phenomena. The time series characteristics of slope strain reflect the changes in the internal rock structure, while the time series characteristics of surface subsidence more reflect the impact of the external environment on the ground. There is a complex interaction between the two, and these relationships are crucial for predicting potential safety risks. In some cases, changes in a small range of the slope may not immediately cause a significant response on the surface, but may cause large-scale surface subsidence after a long period of accumulation. Therefore, in order to more comprehensively and effectively capture the interactive information and correlation between the time series characteristics of slope strain and the time series characteristics of surface subsidence, in the technical solution of the present application, the time series causal association feature vector of the slope strain and the time series causal association feature vector of the surface subsidence are further fused through a feature interaction fusion module based on an X-linear attention bilinear pooling network to obtain a slope strain-surface subsidence time series feature interaction fusion representation vector. In particular, through the feature interaction fusion module based on the X-linear attention bilinear pooling network, the nonlinear correlation between the temporal characteristics of the slope strain and the temporal characteristics of the surface subsidence can be captured by calculating the outer product between the temporal causal correlation feature vector of the slope strain and the temporal causal correlation feature vector of the surface subsidence, so as to generate a richer and more representative feature representation without increasing too much computational burden. This not only enhances the learning ability of the model, but also improves its prediction accuracy for future events, such as discovering risk factors that may lead to landslides or collapses in advance.

[0063] Exemplarily, in the cross-domain fine-grained aggregation analysis unit 43, the three-dimensional voxel grid feature of the mine and the interactive fusion representation of the slope strain-surface subsidence time series feature are subjected to cross-domain fine-grained aggregation analysis to obtain a cross-modal fine-grained aggregation representation of the mine excavation state. It should be understood that since the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid feature map respectively represent different state representations of the mine excavation process, the slope strain and surface subsidence provide internal rock structure changes and surface subsidence information in the time dimension, while the mine three-dimensional voxel grid feature reflects the spatial structure of the mine. The information levels and types contained in the state data of the mine excavation process of different modes are different. For example, the slope strain-surface subsidence time series feature interactive fusion representation can reflect the time series trend of the mine parameter data that changes dynamically, while the mine three-dimensional voxel grid feature is good at describing the internal static geometry of the mine. Therefore, in order to be able to perform information fusion of the two different modal data at a fine-grained level, thereby enhancing the comprehensive understanding and perception of the overall structure of the mine and providing a basis for the subsequent dynamic monitoring of the mine excavation process, in the technical solution of the present application, the interactive fusion representation of the three-dimensional voxel grid characteristics of the mine and the slope strain-surface subsidence time series characteristics is further subjected to cross-domain fine-grained aggregation analysis to obtain a cross-modal fine-grained aggregation representation of the mine excavation state.

[0064] In one embodiment, Figure 6 As shown, the cross-domain fine-grained aggregation analysis unit 43 includes: a modal kernel feature extraction subunit 431, which is used to perform modal kernel feature extraction on the three-dimensional voxel grid feature of the mine to obtain the three-dimensional voxel grid kernel feature of the mine; a fine-grained feature aggregation description processing subunit 432, which is used to perform fine-grained feature aggregation description processing on the three-dimensional voxel grid feature of the mine based on the interactive fusion representation of the slope strain-surface subsidence time series feature and the three-dimensional voxel grid kernel feature of the mine to obtain the cross-modal fine-grained aggregation representation of the mine excavation state.

[0065] In one embodiment, the modal kernel feature extraction subunit is used to: input the mine 3D voxel grid feature map into the modal kernel feature extraction network to obtain the mine 3D voxel grid kernel feature encoding vector as the mine 3D voxel grid kernel feature. Specifically, the process can be expressed as follows:

[0066]

[0067] Where F2 is the three-dimensional voxel grid feature map of the mine, decouple(·) is the feature decomposition, x1, x2, x i 、x j and x nare the first, second, i-th, j-th and n-th mine three-dimensional voxel grid feature vectors in the set of mine three-dimensional voxel grid feature vectors obtained after the mine three-dimensional voxel grid feature map is subjected to feature decoupling and flattening processing, KFDN(·) represents the kernel feature extraction operation, ||·||1 represents the one-norm of the feature vector, M represents the number of vectors in the set of mine three-dimensional voxel grid feature vectors minus one, D i is the semantic difference value of the mine three-dimensional voxel grid corresponding to xi, exp(·) represents an exponential function with the natural constant e as the base, n is the number of vectors in the set of the mine three-dimensional voxel grid feature vectors, and v2 is the mine three-dimensional voxel grid kernel feature encoding vector.

[0068] In one embodiment, Figure 7 As shown, the fine-grained feature aggregation description processing subunit 432 includes: a secondary subunit 4321 for determining a priori clustering center of mine excavation state, which is used to determine the priori clustering center encoding vector of mine excavation state based on the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature encoding vector; a secondary subunit 4322 for expanding the mine three-dimensional voxel grid feature, which is used to perform feature decoupling and feature flattening processing on the mine three-dimensional voxel grid feature map to obtain a set of mine three-dimensional voxel grid feature vectors; a secondary subunit 4323 for cross-domain fine-grained aggregation description processing, which inputs the priori clustering center encoding vector of mine excavation state and the set of mine three-dimensional voxel grid feature vectors into a cross-domain fine-grained aggregation descriptor based on the cluster center to obtain a cross-modal fine-grained aggregation representation vector of mine excavation state as the cross-modal fine-grained aggregation representation of the mine excavation state.

[0069] In one embodiment, the mine excavation state priori cluster center determination secondary subunit 4321 is used to: cascade the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature encoding vector, and linearly transform the cascaded mine excavation state feature vector to obtain the mine excavation state priori cluster center encoding vector. Specifically, the process can be expressed as follows:

[0070] v c =W c [v1; v2]+b c

[0071] Wherein, v1 is the slope strain-surface subsidence time series feature interactive fusion representation vector, [·;·] is a cascade operation, and W c is the weight transformation matrix, b c is the bias vector, v c It is the prior cluster center encoding vector of mine excavation status.

[0072] In one embodiment, the cross-domain fine-grained aggregation description processing secondary subunit 4323 is used to: use the semantic measurement module of the inverse hyperbolic cosine function to perform semantic distance measurement on each mine three-dimensional voxel grid feature vector in the set of mine three-dimensional voxel grid feature vectors and the mine excavation state priori clustering center encoding vector to obtain a set of mine excavation state semantic measurement coefficients. Specifically, the process can be expressed by the formula:

[0073]

[0074] Among them, ||·|| 2 is the square of the vector norm, arccosh is the inverse hyperbolic cosine function, r i (x i ,v c ) is x i With v c Semantic measurement coefficient of mine excavation status between.

[0075] A binary function is used to perform clustering judgment on the set of semantic measurement coefficients of the mine excavation state to obtain a set of semantic clustering coefficients of the mine excavation state. Specifically, the process can be expressed by the formula:

[0076]

[0077] Among them, τ is the preset threshold, a k (x i ) is the i-th mine excavation state semantic clustering coefficient in the set of mine excavation state semantic clustering coefficients.

[0078] Based on the set of semantic clustering coefficients of the mine excavation state, the prior clustering center encoding vector of the mine excavation state and the set of the mine three-dimensional voxel grid feature vector are aggregated to obtain the cross-modal fine-grained aggregate representation vector of the mine excavation state. Specifically, the process can be expressed by the formula:

[0079]

[0080] Wherein, V is the cross-modal fine-grained aggregation representation vector of the mine excavation state.

[0081] In summary, through the processing of cross-domain fine-grained aggregation analysis, the cross-modal pseudo-clustering center between the slope strain-surface subsidence time series feature interactive fusion information and the mine three-dimensional voxel grid feature can be used as an anchoring bridge to integrate data features of different modalities to achieve deeper information extraction and comprehensive feature representation of mine excavation status. In this way, fine-grained information fusion and aggregation are performed at different feature levels to generate a comprehensive representation of the complex semantic structure between the input three-dimensional voxel grid features of the mine and the slope strain-surface subsidence time series feature interactive fusion representation, which helps to better predict potential risk events, such as landslides or collapses, thereby providing strong support for dynamic monitoring and anomaly detection and safety management of mine excavation processes. Compared with the traditional cross-domain interactive analysis network, the cross-domain fine-grained aggregation analysis processing of this application can effectively find a mapping method by introducing a modal kernel feature extraction mechanism, while maintaining the unique attributes of each modality of the three-dimensional voxel grid feature of the mine and the slope strain-surface subsidence time series feature interactive fusion representation of the two, so that data from different modalities can be reasonably compared and combined in a common space, thereby generating a comprehensive mine excavation state feature representation. This process not only improves the intelligence level of the monitoring system, but also brings new possibilities for mine safety management and decision support, especially providing strong technical support in early warning and risk assessment.

[0082] Exemplarily, in the mine excavation process dynamic monitoring unit 44, the mine excavation process is dynamically monitored based on the cross-modal fine-grained aggregation representation of the mine excavation state to determine whether there is an abnormality in the excavation process of the monitored mine. In one embodiment, the mine excavation process dynamic monitoring unit is used to: pass the cross-modal fine-grained aggregation representation vector of the mine excavation state through a classifier-based mine excavation process dynamic monitor to obtain a monitoring result, and the monitoring result is used to indicate whether there is an abnormality in the excavation process of the monitored mine. In other words, the multi-dimensional cross-modal fine-grained aggregation feature representation of the mine excavation state is used to perform classification processing, so as to perform dynamic monitoring of the mine excavation process, thereby detecting whether there is an abnormality in the excavation process of the monitored mine. In this way, intelligent technology can be used to realize dynamic monitoring and anomaly detection of the mine excavation process, which helps to improve the real-time and intelligent level of mine excavation process monitoring and provide strong support for mine safety management and operation.

[0083] In one embodiment, a support vector machine (SVM) is used as the classifier-based dynamic monitor of the mine excavation process. SVM is a supervised learning model, which is particularly suitable for classification problems in high-dimensional space and has good performance for binary classification problems. It maximizes the interval between samples of different categories by finding an optimal hyperplane, thereby achieving effective classification of new samples. In this application, the cross-modal fine-grained aggregation representation vector of the mine excavation state is input into the SVM-based dynamic monitor of the mine excavation process. This vector contains information such as the three-dimensional voxel grid features of the mine after data processing, the interactive fusion representation of the slope strain-surface subsidence time series features, etc. This information comprehensively reflects the current state of the mine. Then, SVM is used to distinguish between normal and abnormal states to obtain monitoring results.

[0084] Preferably, taking into account that the slope strain-surface subsidence temporal feature interactive fusion representation vector and the mine three-dimensional voxel grid feature map respectively represent the temporal interactive fusion features of the slope strain and the surface subsidence and the local three-dimensional model encoding features of the mine three-dimensional model, when performing cross-domain fine-grained aggregation analysis based on modal prior, modal differences and feature encoding features will result in insufficient modal prior coverage and lead to insufficient long-distance fine-grained aggregation representation of the mine excavation state cross-modal fine-grained aggregation representation vector, thereby reducing the expression effect of the mine excavation state cross-modal fine-grained aggregation representation vector and affecting the accuracy of the monitoring results obtained by the mine excavation process dynamic monitor based on the classifier.

[0085] Therefore, in one example, when the cross-modal fine-grained aggregated representation vector of the mine excavation state passes through a classifier-based mine excavation process dynamic monitor to obtain a monitoring result, the cross-modal fine-grained aggregated representation vector of the mine excavation state is optimized, and the optimization includes the following steps:

[0086] Arranging the eigenvalues ​​of the cross-modal fine-grained aggregate representation vector of the mine excavation state in ascending order to form a cross-modal fine-grained aggregate representation sequence encoding vector of the mine excavation state;

[0087] In response to the absolute value of the difference between the i-th eigenvalue and the i+1-th eigenvalue of the cross-modal fine-grained aggregate representation sequence encoding vector of the mine excavation state being less than or equal to the distance difference hyperparameter ε, the weighted sum between the i-th eigenvalue and the i+1-th eigenvalue is calculated as the optimized i+1-th eigenvalue v′ i+1 =ω1×v i +ω2×v i+1 , where v i and v i+1They respectively represent the i-th eigenvalue and the i+1-th eigenvalue of the cross-modal fine-grained aggregate representation sequential encoding vector of the mine excavation state, and ω1 and ω2 respectively represent the first weight hyperparameter and the second weight hyperparameter;

[0088] In response to the absolute value of the difference between the i-th eigenvalue and the i+1-th eigenvalue of the cross-modal fine-grained aggregate representation sequential encoding vector of the mine excavation state being greater than the distance difference hyperparameter ε:

[0089] Calculate the square root of the sum of squares of all eigenvalues ​​of the cross-modal fine-grained aggregate representation vector of the mine excavation state Among them, v1, v2, v n They respectively represent the first, second and nth eigenvalues ​​of the cross-modal fine-grained aggregate representation vector of the mine excavation state;

[0090] The square root root is multiplied by 2 and then divided by the square of the length of the cross-modal fine-grained aggregate representation vector of the mine excavation state to obtain the cross-modal fine-grained aggregate representation space primitive value of the mine excavation state: base = 2×root / L 2 , where L represents the length of the cross-modal fine-grained aggregation representation vector of the mine excavation state;

[0091] After multiplying the cross-modal fine-grained aggregate representation space primitive value of the mine excavation state by the i-th eigenvalue, the weighted subtraction between the product and the i+1th eigenvalue is calculated to obtain the optimized i+1th eigenvalue v i+1 ′=ω3×v i ×base-ω4×v i+1 , where ω3 and ω4 represent the third weight hyperparameter and the fourth weight hyperparameter, respectively;

[0092] Based on v′1=v1, where v′1 represents the first eigenvalue of the optimized cross-modal fine-grained aggregate representation vector of the mine excavation state, the i+1th eigenvalue v of the combined optimization is i+1 ′ is used to obtain the optimized cross-modal fine-grained aggregation representation vector of mine excavation status.

[0093] In this way, in order to solve the problem that the feature set of the cross-modal fine-grained aggregated representation vector of the mine excavation state has insufficient global interactive response representation capability due to the long distance exceeding the predetermined local distribution interval threshold under the predetermined feature value sequence distribution, the high-dimensional feature space primitive representation of the cross-modal fine-grained aggregated representation vector of the mine excavation state based on self-inner product fusion is used to capture the complex structure of the global network interaction of its feature values, so as to reconstruct the interactive response relationship between the feature values ​​of the cross-modal fine-grained aggregated representation vector of the mine excavation state by simulating the scale-based high-dimensional feature space potential primitive, so as to realize the encoding reconstruction of the real sequence distribution behavior of the cross-modal fine-grained aggregated representation vector of the mine excavation state under the long distance, improve the encoding expression effect of the cross-modal fine-grained aggregated representation vector of the mine excavation state, and improve the accuracy of the monitoring results obtained by the dynamic monitor of the mine excavation process based on the classifier. In this way, intelligent technology can be used to realize dynamic monitoring and anomaly detection of the mine excavation process, which is helpful to improve the real-time and intelligent level of mine excavation process monitoring and provide strong support for mine safety management and operation.

[0094] Exemplarily, in the mine excavation process monitoring module 50, the dynamic monitoring results of the mine excavation process are transmitted to the terminal device of the manager for display. It should be understood that in the mining environment, the geological conditions are complex and changeable, and potential safety hazards such as slope instability and surface subsidence may occur at any time. The data collected by the sensor group deployed at multiple key locations inside the mine and the point cloud data obtained by three-dimensional laser scanning of the excavation tunnel and mining surface are generated after the cross-domain fine-grained interactive analysis based on modal prior by the processing center. Dynamic monitoring results of the mine excavation process. These results not only contain detailed information on the current mine status, but also can predict possible problems in the future, providing a scientific basis for managers.

[0095] In one embodiment, in order to achieve this goal, an efficient and reliable data transmission and display system needs to be established. First, after the data processing center completes the analysis, the dynamic monitoring results will be transmitted to the terminal devices used by each management level through a secure and stable communication network. This usually involves the use of wired or wireless communication technologies, such as fiber optic networks for high-speed connections between fixed sites, while Wi-Fi, 4G / 5G mobile networks or industrial-grade wireless protocols (such as LoRa, ZigBee) are suitable for a wider coverage range. Considering the uniqueness and safety requirements of the mining operating environment, all communication equipment must have strong anti-interference capabilities and good explosion-proof performance, and a redundant mechanism should be provided to ensure the continuity and reliability of data transmission. Furthermore, an alarm function can also be integrated to promptly notify relevant personnel to take measures to prevent accidents when abnormalities are detected in the excavation process of the monitored mine.

[0096] In summary, according to the embodiment of the present application, a dynamic monitoring system for the mine excavation process is explained, which collects time series data of slope strain and surface subsidence through sensor groups deployed at key positions inside the mine, as well as point cloud data of excavation tunnels and mining surfaces obtained through three-dimensional laser scanning, and these data are transmitted to the processing center. At the processing center, the above data are analyzed based on the cross-domain fine-grained interactive analysis technology of modal prior to determine the dynamic monitoring results of the mine excavation process. Finally, the dynamic monitoring results are sent to the terminal device of the manager for display, providing real-time monitoring information. In this way, by integrating multiple data information for comprehensive analysis, it is possible to make up for the limitation of insufficient information of a single data type when evaluating the mine condition. This multi-source data analysis method provides a more comprehensive and in-depth understanding of the mine environment, making it more accurate to identify potential safety hazards or abnormal conditions, thereby effectively improving the safety and reliability of mine operations.

[0097] The prefixes such as "first" and "second" used in the embodiments of the present application are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described objects. The use of prefixes such as ordinal numbers used to distinguish description objects in the embodiments of the present application does not constitute a limitation on the described objects. For the description of the described objects, please refer to the description of the context in the embodiments, and the use of such prefixes should not constitute an unnecessary limitation.

[0098] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0099] In the various embodiments of the present application, unless otherwise specified or logically conflicting, the terms and / or descriptions between the various embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationships.

[0100] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0101] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0102] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A dynamic monitoring system for a mine excavation process, characterized in that: include: A mine status data acquisition module, which is used to collect a time series data set of slope strain and a time series data set of surface subsidence through a sensor group deployed at multiple key locations inside the monitored mine; A mine tunnel laser scanning module is used to perform three-dimensional laser scanning on the tunnel and mining surface of the monitored mine to obtain three-dimensional laser scanning point cloud data of the mine; A data transmission module, used to transmit the time series data set of the slope strain, the time series data set of the surface subsidence and the three-dimensional laser scanning point cloud data of the mine to a data processing center; A mine data analysis module is used to perform cross-domain fine-grained interactive analysis based on modal priors on the time series data set of the slope strain, the time series data set of the surface subsidence and the mine three-dimensional laser scanning point cloud data in the data processing center to determine the dynamic monitoring results of the mine excavation process; The mine excavation process monitoring module is used to transmit the dynamic monitoring results of the mine excavation process to the terminal equipment of the management personnel for display.

2. The dynamic monitoring system for mine excavation process according to claim 1 is characterized in that: The mine data analysis module comprises: A three-dimensional voxel encoding unit, used for performing three-dimensional voxel encoding on the three-dimensional laser scanning point cloud data of the mine to obtain three-dimensional voxel grid features of the mine; A time series feature interactive fusion unit, used for performing time series feature analysis and interactive fusion on the time series data set of the slope strain and the time series data set of the ground subsidence to obtain a slope strain-ground subsidence time series feature interactive fusion representation; A cross-domain fine-grained aggregation analysis unit is used to perform cross-domain fine-grained aggregation analysis on the three-dimensional voxel grid features of the mine and the interactive fusion representation of the slope strain-surface subsidence time series features to obtain a cross-modal fine-grained aggregation representation of the mine excavation state; A mine excavation process dynamic monitoring unit is used to dynamically monitor the mine excavation process based on the cross-modal fine-grained aggregation representation of the mine excavation state to determine whether there is any abnormality in the excavation process of the monitored mine.

3. The dynamic monitoring system for mine excavation process according to claim 2 is characterized in that: The three-dimensional voxel encoding unit comprises: A voxel distance encoding subunit, used for performing voxel distance encoding on the mine 3D laser scanning point cloud data to obtain mine 3D voxel grid data; The mine three-dimensional feature extraction subunit is used to obtain a mine three-dimensional voxel grid feature map as the mine three-dimensional voxel grid feature by passing the mine three-dimensional voxel grid data through a mine three-dimensional feature extractor based on 3D-CNN.

4. The dynamic monitoring system for mine excavation process according to claim 3 is characterized in that: The temporal feature interactive fusion unit comprises: A time series causal correlation feature extraction subunit is used to respectively pass the time series data set of the slope strain and the time series data set of the surface subsidence through a time series encoder based on a void causal convolutional neural network model to obtain a time series causal correlation feature vector of the slope strain and a time series causal correlation feature vector of the surface subsidence; The feature interaction fusion processing subunit is used to obtain a slope strain-surface subsidence time series feature interaction fusion representation vector as the slope strain-surface subsidence time series feature interaction fusion representation vector by means of a feature interaction fusion module based on an X-linear attention bilinear pooling network.

5. The dynamic monitoring system for mine excavation process according to claim 4 is characterized in that: The cross-domain fine-grained aggregation analysis unit includes: A modal kernel feature extraction subunit, used for performing modal kernel feature extraction on the three-dimensional voxel grid feature of the mine to obtain the three-dimensional voxel grid kernel feature of the mine; A fine-grained feature aggregation description processing subunit is used to perform fine-grained feature aggregation description processing on the three-dimensional voxel grid features of the mine based on the interactive fusion representation of the slope strain-surface subsidence time series features and the three-dimensional voxel grid kernel features of the mine to obtain a cross-modal fine-grained aggregation representation of the mine excavation state.

6. The dynamic monitoring system for mine excavation process according to claim 5 is characterized in that: The modal kernel feature extraction subunit is used to: input the mine 3D voxel grid feature map into a modal kernel feature extraction network to obtain a mine 3D voxel grid kernel feature encoding vector as the mine 3D voxel grid kernel feature.

7. The dynamic monitoring system for mine excavation process according to claim 6 is characterized in that: The fine-grained feature aggregation description processing subunit includes: The priori cluster center of the mine excavation state determines the secondary subunit, which is used to determine the priori cluster center encoding vector of the mine excavation state based on the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature encoding vector; A mine three-dimensional voxel grid feature expansion secondary subunit is used to perform feature decoupling and feature flattening processing on the mine three-dimensional voxel grid feature map to obtain a set of mine three-dimensional voxel grid feature vectors; The cross-domain fine-grained aggregation description processes the secondary sub-unit, which inputs the set of the prior cluster center encoding vector of the mine excavation state and the three-dimensional voxel grid feature vector of the mine into the cluster center-based cross-domain fine-grained aggregation descriptor to obtain the cross-modal fine-grained aggregation representation vector of the mine excavation state as the cross-modal fine-grained aggregation representation of the mine excavation state.

8. The dynamic monitoring system for mine excavation process according to claim 7 is characterized in that: The mine excavation state prior clustering center determines a secondary subunit, which is used to: cascade the slope strain-surface subsidence time series feature interactive fusion representation vector and the mine three-dimensional voxel grid kernel feature coding vector, and linearly transform the cascaded mine excavation state feature vector to obtain the mine excavation state prior clustering center coding vector.

9. The dynamic monitoring system for mine excavation process according to claim 8, characterized in that: The cross-domain fine-grained aggregation description processing secondary subunit is used to: Using a semantic measurement module of an inverse hyperbolic cosine function, a semantic distance measurement is performed between each mine three-dimensional voxel grid feature vector in the set of mine three-dimensional voxel grid feature vectors and the mine excavation state priori clustering center encoding vector to obtain a set of mine excavation state semantic measurement coefficients; Using a binary function to perform clustering judgment on the set of semantic measurement coefficients of the mine excavation state to obtain a set of semantic clustering coefficients of the mine excavation state; Based on the set of semantic clustering coefficients of the mine excavation state, the set of prior clustering center encoding vectors of the mine excavation state and the set of three-dimensional voxel grid feature vectors of the mine are aggregated to obtain a cross-modal fine-grained aggregated representation vector of the mine excavation state.

10. The dynamic monitoring system for mine excavation process according to claim 9, characterized in that: The mine excavation process dynamic monitoring unit is used to: pass the cross-modal fine-grained aggregation representation vector of the mine excavation state through a classifier-based mine excavation process dynamic monitor to obtain a monitoring result, and the monitoring result is used to indicate whether there is an abnormality in the excavation process of the monitored mine.

Images