Multi-dimensional coal mine risk identification method and system

Abstract

The invention discloses a multi-dimensional coal mine risk identification method and system, and relates to the technical field of electric digital data processing, and the method comprises the steps: obtaining spatial geographic data, basic attribute data, monitoring data and tectonic geological data of a coal mine region, constructing a geographic information model according to a base-mining area-coal mine three-level spatial logic, and carrying out the layered presentation; receiving text search, map selection and layer switching interaction, and generating a target object set based on spatial position relation and geocoding analysis; executing at least one of space statistics, time sequence comparison, emergency rescue path analysis and knowledge base retrieval construction according to a preset dimension for the target object set, and generating risk identification result data; and displaying the result data and the spatial object in an associated manner, and outputting a map view or a statistical chart. Through the technical scheme of the invention, integrated management and collaborative analysis of multi-source data are realized, and the risk identification efficiency and the visual decision support capability are improved.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and in particular to a multi-dimensional coal mine risk identification method and a multi-dimensional coal mine risk identification system. Background Technology

[0002] Coal mines are a crucial foundation for my country's energy supply, and their safety levels directly impact the industry's high-quality development and the safety of miners' lives. Risk identification, as a preliminary step in the coal mine disaster prevention and control chain, requires timely identification of high-risk areas and mines under complex geological conditions and multi-source monitoring data constraints, providing support for safety management measures such as water control, outburst prevention, and collapse prevention.

[0003] Existing coal mine risk identification systems are mostly formed by gradually adding functions to traditional geographic information platforms or monitoring platforms, resulting in problems such as scattered data types and fragmented systems. On the one hand, spatial geographic data, basic coal mine attribute data, online monitoring data, and structural geological information are usually scattered across different business systems, lacking an integrated organization and display based on the three-level spatial logic of "base-mining area-coal mine," making it difficult to achieve timely, step-by-step risk focus and horizontal comparison from macro-regions to specific mines. On the other hand, most systems still rely primarily on empirical threshold judgments based on single dimensions or a few indicators. Spatial statistical analysis, risk time-series evolution comparison, emergency rescue path planning, and risk Q&A based on professional structural knowledge are often implemented in different tools or subsystems, lacking a unified entry point and collaborative analysis mechanism. This leads to a long risk identification process chain, high human involvement, and an easy omission of key risk clues.

[0004] In terms of information display and interaction, existing systems generally use a single-layer map overlaid with multiple layers to present coal mine-related information. They lack refined search, map selection, and layer switching mechanisms tailored to specific target areas and coal mines. Users find it difficult to promptly align quantitative analysis results with specific spatial locations, resulting in low efficiency for cross-time and cross-regional comparative analysis. Overall, current technologies cannot yet achieve integrated management of coal mine risks based on multi-source data, multi-dimensional risk analysis targeting specific areas and coal mines, and intuitive visualization output on a unified geographic information platform. This makes it difficult to support the precise, real-time, and intelligent requirements for coal mine risk identification under complex operating conditions. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a multi-dimensional coal mine risk identification method and system. By integrating spatial geographic data, coal mine attribute data, monitoring data, and structural geological data on a geographic information platform and constructing a three-tiered spatial model of base-mining area-coal mine, it achieves integrated management and hierarchical display of multi-source data in the coal mine area, providing a unified data foundation for risk identification. Through spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a structural knowledge base, based on target object sets obtained through text search, map selection, or layer switching, it comprehensively depicts the spatial distribution, temporal evolution, and emergency accessibility of risks in target areas and target coal mines, improving the accuracy and timeliness of risk identification. By associating risk identification results with spatial objects and outputting them in the form of map views and / or statistical charts, it achieves intuitive expression and multi-dimensional comparison of risk information, supporting management and regulatory bodies to quickly identify high-risk areas and formulate targeted prevention and control measures.

[0006] To achieve the above objectives, the present invention provides a multi-dimensional coal mine risk identification method, comprising: Spatial geographic data, basic attribute data, monitoring data, and structural geological data covering the coal mining area are acquired on the geographic information platform. A geographic information model of the coal mining area is constructed based on the three-level spatial logic of base-mining area-coal mine, and the above data are presented in hierarchical order on the geographic information display interface. The system receives user interaction operations on the geographic information display interface targeting a target area and / or a target coal mine, and parses the interaction operations into a set of target objects based on spatial location relationships and geocoding. The interaction operations include at least one of text search, map selection, or layer switching. For the target object set, a multi-dimensional risk analysis process is executed according to a preset multi-dimensional analysis dimension to generate risk identification result data related to the target object set. The multi-dimensional risk analysis process includes at least one of spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base. In the geographic information display interface, the risk identification result data is associated with the spatial objects in the data model and displayed, and the coal mine risk identification results are output in the form of map view and / or statistical chart.

[0007] In the above technical solution, preferably, the spatial statistical analysis in the multidimensional risk analysis process includes: Upon receiving a spatial statistical analysis instruction from a user for the target area, the target area is determined by at least one of the following methods: polygon drawing, rectangular selection, or administrative region selection. Based on the ray method, the topological relationship between the center point coordinates of each coal mine in the system and the boundary of the target area is determined to see if the center point of the coal mine falls within the target area, and the set of coal mines within the target area is obtained by filtering. Based on the coal mine set within the target area and the monitoring point data corresponding to the target area, calculate at least one or more of the following spatial statistical indicators: the number of coal mines within the target area, the total designed production capacity of coal mines within the target area, the total average water inflow of coal mines within the target area over a preset period, the proportion of high-risk mines obtained based on the coal mine risk level identification, and the monitoring point density obtained based on the number of monitoring points and the area of ​​the target area. The spatial statistical indicators are then written into the risk identification result data.

[0008] In the above technical solution, preferably, the time-series change comparison analysis in the multidimensional risk analysis process includes: Upon receiving a time-series change comparison analysis instruction for the target object set, load two sets of risk-related layers formed at the first and second time points for the same target area and / or target coal mine set; The two sets of risk-related layers are spatially aligned based on a unified spatial reference coordinate system to ensure that the coordinates of the same geographical location are consistent in the two sets of layers. An image segmentation algorithm is used to extract the contour information of the target region of a preset type from the two sets of risk-related layers to obtain the contour of the first target region and the contour of the second target region. Based on the intersection-union ratio (IUU) calculation rule, the similarity between the outline of the first target region and the outline of the second target region is calculated. When the similarity is lower than a preset threshold, it is determined that there is a significant difference between the two sets of risk-related layers at the corresponding locations. The difference regions are spatially labeled and described in the geographic information display interface, and the difference labeling information is written into the risk identification result data.

[0009] In the above technical solution, preferably, when performing a time-series change comparison analysis of two sets of risk-related layers, a roll-up interactive method is used to display the layer comparison, the roll-up interactive method including: When loading two sets of risk-related layers that need to be compared, unify the two sets of layers to the same spatial reference coordinate system and set the initial transparency of the two sets of layers; Add vertical and / or horizontal roller shutter strips to the geographic information display interface. When the user drags the roller shutter strip, the first group of layers is made visible on one side of the roller shutter strip and the second group of layers is made visible on the other side of the roller shutter strip, depending on the position of the roller shutter strip within the longitude or latitude range of the current view. The visibility range of the two groups of layers is adjusted in real time. A gradient transition area is set on both sides of the roller shutter strip to reduce the impact of the layer segmentation boundary on the visual effect, and to keep the two sets of layers and difference annotation information updated synchronously during the movement of the roller shutter strip.

[0010] In the above technical solution, preferably, the emergency rescue path analysis in the multidimensional risk analysis process includes: It acquires real-time geographic information, including road congestion status, weather information, and underground passage access status; accident site information, including accident location coordinates, risk impact range, and rescue needs; and rescue resource information, including the location of rescue teams, the location of rescue equipment storage points, and the location of medical points. Outlier identification and standardization preprocessing are performed on the above multi-source data to construct a multi-dimensional data source for path planning; A rescue road network model is constructed based on the multidimensional data source. For each candidate road segment in the rescue road network, multidimensional attribute values ​​including road segment length, road segment travel time, road segment safety factor and resource accessibility factor are calculated. A path cost function is constructed based on each attribute value and the corresponding weight coefficient. A path search algorithm based on the path cost function is used to determine the path with the minimum comprehensive cost from the candidate paths from the rescue starting point to the accident location in the rescue road network as the emergency rescue route. The emergency rescue route is written into the risk identification result data and displayed in the geographic information display interface. Furthermore, the system reacquires the multi-source dynamic data within a preset time interval and updates the path cost of the candidate road segments. When the path cost change rate of the current emergency rescue route exceeds a preset threshold, the system re-executes the path search to update the emergency rescue route and outputs the updated route information and resource allocation suggestions.

[0011] In the above technical solution, preferably, the knowledge retrieval based on the constructed knowledge base in the multidimensional risk analysis process includes: A coal mine-specific structural knowledge base is constructed, which includes structured structural unit attribute data, unstructured geological exploration reports and historical risk cases, and knowledge graph data constructed based on expert experience. When a text query request for the target object set is received, the natural language question input by the user is semantically encoded to obtain a question vector. The similarity is calculated with the candidate data vector in the coal mine-specific structural knowledge base. A preset number of candidate results are selected from the candidate data with a similarity of not less than a preset threshold. Based on the candidate results, an answer containing structural unit attributes, risk characteristics and prevention and control suggestions is generated. Upon receiving a construction query request triggered by map point selection, the corresponding construction unit identifier is queried based on the coordinates of the selected location. A standardized query request is automatically generated and input as a natural language question into the knowledge retrieval process based on the construction knowledge base for semantic encoding and retrieval processing. The generated answer content is then associated with spatial objects in the target object set and displayed. Construction risk knowledge is written into the risk identification result data.

[0012] In the above technical solution, preferably, the parsing process of the text search operation in the interactive operation includes: Construct a multi-level administrative region geocoding database covering coal mining areas. Record the name, administrative division code, geometric center coordinates and boundary coordinates of each administrative region in the geocoding database, and establish mapping relationships for commonly used abbreviations and synonyms. The search text entered by the user is segmented and stop words are filtered. Based on the trained classification model, the search intent is identified and distinguished into administrative region search or coal mine name search. In scenarios such as administrative region search or coal mine name search, precise matching is first performed in the geocoding database. If precise matching fails, the similarity between the search text and the names of each administrative region in the geocoding database is calculated based on the edit distance. A set of candidate administrative regions is determined in the administrative regions with similarity not lower than a preset threshold, and a preset number of candidate results are returned for the user to choose from. After determining the target administrative region or target coal mine, the map zoom level is calculated based on its boundary range, and the center of the map view is located to the geometric center coordinates of the target area. The target administrative region or target coal mine is then added to the target object set.

[0013] In the above technical solution, preferably, the multi-dimensional coal mine risk identification method further includes a scene freezing and backtracking process, which specifically includes: Upon receiving a scene save command, the zoom level, center coordinates, visible layer identifiers, labeled object information, and save time of the current map view are collected. The above parameters are organized into structured keyframe data according to preset fields and stored in the data storage system. At the same time, a thumbnail of the current view is generated and an index relationship between labels and keyframe identifiers is established. Upon receiving a scene backtracking instruction, the zoom level and center coordinate parameters in the target keyframe data are read from the data storage system. Within a preset total transition time, the zoom level and center coordinates of the current view are gradually adjusted using time interpolation. After the interpolation adjustment is completed, the visible layers and annotation information recorded in the target keyframe data are loaded to restore the corresponding scene view. The restored scene is then added to the risk identification result data for subsequent analysis.

[0014] In the above technical solution, preferably, the specific process of constructing a geographic information model of the coal mine area based on the three-level spatial logic of base-mining area-coal mine and presenting the above data in a hierarchical manner includes: The correlation at the base level should include at least macro data such as administrative boundaries, main transportation lines, major water systems and mining area distribution; the correlation at the mining area level should include at least meso data such as geological structure, aquifer boundaries, mine location and risk zoning; and the correlation at the coal mine level should include at least micro data such as the location of mining face, monitoring point location, old working water area and water-rich anomaly area. A preset zoom level threshold is set for the base level, mining area level, and coal mine level. Based on the relationship between the current map zoom level and the zoom level threshold, the current display level is determined, and a progressive loading strategy is used to load data elements of the target level in batches during the level switching process. After the level switch is completed, the focus offset is calculated based on the coordinates of the core spatial object of the target display level and the coordinates of the current view center. Within the preset focus transition time, the view center is interpolated and updated according to the smooth transition function, and the map view is smoothly moved to the geometric center position of the core spatial object.

[0015] This invention also proposes a multi-dimensional coal mine risk identification system, which applies the multi-dimensional coal mine risk identification method disclosed in any of the above technical solutions, including: An integrated data modeling unit is used to acquire spatial geographic data, basic attribute data, monitoring data, and structural geological data covering the coal mining area. Based on the three-level spatial logic of base-mining area-coal mine, a geographic information model of the coal mining area is constructed, and the above data is presented hierarchically in the geographic information display interface. An interactive operation parsing unit is used to receive interactive operations from users on the geographic information display interface for target areas and / or target coal mines, and parse the interactive operations into a set of target objects based on spatial location relationships and geographic codes; A multidimensional risk analysis unit is used to perform a multidimensional risk analysis process for the target object set according to a preset multidimensional analysis dimension. The multidimensional risk analysis process includes at least one of spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base, to generate risk identification result data related to the target object set. The risk object display unit is used to associate and display the risk identification result data with the spatial objects in the geographic information model of the coal mine area in the geographic information display interface, and output the coal mine risk identification result in the form of map view and / or statistical chart. The integrated data modeling unit, interactive operation parsing unit, multidimensional risk analysis unit, and risk object display unit are configured to work together to execute the multidimensional coal mine risk identification method.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) By acquiring spatial geographic data, coal mine basic attribute data, monitoring data and structural geological data on a unified geographic information platform, and constructing a coal mine regional geographic information model based on the three-level spatial logic of base-mining area-coal mine and presenting it in a hierarchical manner, the integrated organization and hierarchical display of multi-source heterogeneous data have been realized, which has improved the efficiency of risk factor focusing and information consistency from macro-region to specific mine.

[0017] (2) By receiving interactive operations such as text search, map selection and / or layer switching on the geographic information display interface, and parsing the interactive operations into a set of target objects based on spatial location relationships and geographic codes, the calculable selection and positioning of target areas and / or target coal mines is realized, reducing the operational costs and risk of misselection caused by repeated screening and manual alignment.

[0018] (3) By executing a multidimensional risk analysis process for the target object set according to a preset multidimensional analysis dimension and generating risk identification result data, the unified scheduling and result aggregation of at least one of the following analysis capabilities—spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base—is realized, thereby improving the analysis efficiency of risk identification and the reusability of the result output.

[0019] (4) By displaying the risk identification results data in relation to the spatial objects in the data model in the geographic information display interface and outputting them in the form of map view and / or statistical charts, the linkage presentation of risk results and spatial location is realized, which enhances the intuitiveness of risk identification results and the ability to assist decision-making. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a multi-dimensional coal mine risk identification method disclosed in an embodiment of the present invention; Figure 2 This is a schematic diagram of an emergency rescue route planning process disclosed in one embodiment of the present invention; Figure 3 This is a schematic diagram of the architecture of a multi-dimensional risk identification system for coal mines disclosed in one embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] The present invention will now be described in further detail with reference to the accompanying drawings: like Figure 1 and Figure 2 As shown, the multi-dimensional coal mine risk identification method provided by the present invention runs in a geographic information platform environment.

[0023] First, spatial geographic data (including administrative boundaries, transportation, water systems, topography, etc.), basic coal mine attribute data (including mine name, designed capacity, risk level, etc.), monitoring data (including water inflow, roof displacement, monitoring point coordinates, etc.), and structural geological data (including faults, aquifer boundaries, old workings and water-prone areas, areas with multiple water hazards, etc.) covering the target coal mine area are accessed into the geographic information platform. A geographic information model of the coal mine area is constructed according to the three-level spatial logic of "base-mining area-coal mine". The spatial objects of the base layer, mining area layer and coal mine layer are associated with the corresponding attribute data. The geographic information display interface displays the data in layers according to the map zoom level, realizing a step-by-step focus from the regional base to the specific mine, and then to the working face and monitoring points.

[0024] Based on the aforementioned data model, the method receives user interactions on the geographic information display interface, including locating administrative regions or coal mines through text search, defining target areas by selecting boxes, drawing polygons, or selecting single points on a map, and selecting different types of risk layers through layer switches. Based on spatial location relationships and geocoding, the user interactions are parsed into one or more target objects, such as target administrative regions, target mining areas, target coal mines, and their surrounding risk zones, forming a set of target objects.

[0025] For the target object set, multi-dimensional analysis dimensions are preset based on the business needs of coal mine safety management. In this embodiment, these include spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base. When the user selects spatial statistical analysis, the number of coal mines, production capacity, hydrological monitoring indicators, and risk levels within the target area are quantitatively analyzed. When the user selects temporal change comparison analysis, spatial differences in risk-related layers at different times are identified. When the user selects emergency rescue path analysis, emergency path planning is performed based on accident location, road network, and rescue resources. When the user selects knowledge retrieval based on a constructed knowledge base, structured answers are generated for specified structural units or text questions. The analysis results are uniformly organized into risk identification result data and associated with the target objects.

[0026] Finally, in the geographic information display interface, the risk identification results data are bound to the spatial objects in the geographic information model of the coal mine area. The map view is displayed in the form of layer overlay, feature highlighting, and annotation information window. The corresponding statistical charts, lists and text descriptions are output in the side panel or pop-up window, so that users can complete the browsing, analysis and decision support of coal mine risks in a single interface.

[0027] In this implementation, the geographic information model of the coal mine area uses an intelligent scaling algorithm to achieve progressive focusing of map data and automatically filter out information from non-target areas. The intelligent scaling algorithm switches the display level by judging the relationship between the current map scaling level Z and thresholds Z1 and Z2. The switching formula is as follows: Where Z is the current map zoom level (the larger the value, the more microscopic the view), and Z1 and Z2 are the level switching thresholds (which can be customized according to the actual scale of the coal mine).

[0028] During the switching process, a progressive loading algorithm is used to control the data display pace. The loading progress formula is: in, P ( t )for t Data loading progress at any given time (0≤ P ( t (≤1), k=0.8 is the loading rate coefficient, ensuring that the hierarchical data switching is completed within 1.5 seconds.

[0029] Target region focusing is achieved through a smooth movement algorithm, and the focusing offset is calculated as follows: in,(X center , Y center ) represents the coordinates of core elements at the target level (such as the center of the mining area or the coal mine entrance). X current , Y current ) represents the current view center coordinates, T=1000ms is the focus transition duration, and sin(πt / 2T) is the smooth transition function to ensure smooth and natural view movement.

[0030] In this implementation, a unified geographic information platform, three-level spatial logic modeling, interactive target object parsing, and multi-dimensional risk analysis process are used to achieve the visualization and multi-perspective identification of various types of hidden dangers such as coal mine water hazards and structural risks. This solves the problems of fragmented information display, single analysis dimension, and insufficient decision support in traditional solutions, improves the efficiency and accuracy of coal mine risk identification, and provides a data foundation for subsequent early warning and emergency response.

[0031] In the above embodiments, preferably, the spatial statistical analysis in the multidimensional risk analysis process includes: After receiving a spatial statistical analysis command from a user for a specific target area, the user can define the boundary of the target area on the geographic information display interface by drawing polygons, selecting rectangles, or selecting administrative regions. The platform abstracts the boundary of the target area into a polygonal region based on a unified spatial reference coordinate system.

[0032] For each registered coal mine, the coordinates of its geometric center point are obtained, and the ray casting method is used to determine whether the center point falls inside the polygon of the target area. A point-polygon relationship judgment is performed on all coal mines to filter out the set of coal mines within the target area. For the monitoring point data of the target area, relevant monitoring points within the area (such as water inflow monitoring points, surface subsidence monitoring points, etc.) are filtered using the same spatial judgment method to ensure that the statistical range is consistent with the spatial boundary of the area.

[0033] Based on this, the system automatically calculates multidimensional spatial statistical indicators for the target area by combining the coal mine set and the monitoring point set within the target area. For example: it counts the number of coal mines and sums the designed production capacity of each mine within the target area to obtain the total production capacity; it sums the average water inflow of each coal mine over a recent period (e.g., the last 7 days or the last 30 days) to obtain the total water inflow; it calculates the proportion of high-risk mines based on the risk level (high, medium, low) of the coal mines; and it calculates the monitoring point density based on the number of monitoring points and the area of ​​the target area. These statistical results are written into the risk identification results data and displayed in the map view as annotations, thematic maps, or statistical panels.

[0034] Specifically, three flexible region definition methods are available to meet the needs of different analysis scenarios: Polygon drawing: Users can manually draw any polygon on the map, and the system records the boundary coordinates in WKT (Well-KnownText) format to generate regional geometric objects; Rectangular selection: The user drags the mouse to form a rectangular area, and the system automatically calculates the coordinates of the diagonal vertices of the rectangle to generate the area range; Administrative Region Selection: Select a preset administrative region (province / city / county) from the drop-down menu to directly access the administrative region boundary data from the database. After the region is defined, the formula for determining whether a coal mine is located within the target region using the ray casting method is as follows: in, P =( X p , Y p ( ) represents the coordinates of the coal mine's center point. This algorithm can accurately screen coal mines within the region, with an accuracy rate of ≥99.5%.

[0035] Based on the filtered regional coal mine data, the following core quantitative indicators are automatically calculated, covering key dimensions such as production capacity, risk, and hydrology: Number of coal mines: in, M This represents the total number of coal mines within the system. P i For the first i Coordinates of the center point of each coal mine; Total capacity: in, Q i For the first i The designed production capacity of a coal mine (unit: 10,000 tons / year), if the coal mine is in a state of shutdown. Q i =0; Total water inflow: in, W i For the first i The average water inflow of the coal mine over the past 7 days (unit: m³) 3 / d), the data comes from the real-time monitoring system; Percentage of high-risk mines: in, Risk ( P i The risk level is indicated by (high risk = 1, medium risk = 0.5, low risk = 0), which is derived from a comprehensive assessment of water hazard and geological structure risks. Monitoring point density: in, K This represents the total number of monitoring points within the system. S i For the first i Coordinates of each monitoring point S area Area of ​​the target region (unit: km²) 2 This reflects the coverage of regional risk monitoring.

[0036] This implementation method enables rapid and automated statistical analysis of multi-dimensional indicators such as the number of coal mines, production capacity, hydrological data, and risk levels within any user-specified spatial range. This avoids manual querying and aggregation, significantly improving the efficiency of regional risk quantification analysis. Furthermore, the statistical results are displayed in conjunction with map spatial objects, facilitating the identification of risk-concentrated areas and areas with weak monitoring capabilities, thus supporting regulatory authorities in developing differentiated regulatory strategies.

[0037] In the above embodiments, preferably, the time-series change comparison analysis in the multidimensional risk analysis process is used to identify changes in the risk layer of the same target area at different times.

[0038] Upon receiving a command to perform a comparative analysis of temporal changes for a set of target objects, the system first loads two sets of risk-related layers for the same target area and / or the same coal mine set, generated at the first and second time points, based on user selection or system presets. Examples include a layer showing the distribution of water-rich anomaly zones, surface subsidence range, or roof displacement grading for the same working face on different dates. To ensure spatial consistency, both sets of layers are converted to a unified spatial reference coordinate system, and coordinate accuracy is uniformly processed, ensuring consistent coordinates for the same geographical location in both sets of layers.

[0039] Based on spatial alignment, image segmentation or region extraction algorithms are used to extract the contour information of the target region from two sets of layers. For example, the contour lines or hierarchical regions of the water-rich anomaly area or the roof anomaly area can be rasterized or vectorized to obtain the contours of the first target region and the second target region. Subsequently, the spatial similarity between the two sets of region contours is calculated based on the intersection-union ratio rule, and the ratio of the intersection area to the union area is used as the similarity index. When the similarity is lower than a preset threshold, it is determined that there is a significant change at that location.

[0040] For areas identified as changing, spatial annotations are made on the geographic information display interface using highlighted outlines, semi-transparent overlays, or flashing annotations. Difference descriptions are generated in the side panel, such as "the area of ​​the water-rich anomaly zone on the southeast side of a certain working face has expanded" or "the range of the high-value area of ​​roof displacement in a certain monitoring area has shrunk." The above difference annotation information is written into the risk identification result data to provide a basis for subsequent risk assessment.

[0041] In this implementation, automated layer space alignment and regional contour comparison enable quantitative identification and visualization of changing areas between risk layers at different times. This reduces the workload of manual map comparison, improves the efficiency of identifying risk evolution trends (such as water hazard expansion and abnormal area migration), and supports managers in identifying areas of heightened risk in advance and taking targeted prevention and control measures.

[0042] In the above embodiments, preferably, when performing a time-series change comparison analysis of two sets of risk-related layers, a roll-up interactive method is used to display the layer comparison, which is used to perform an interactive visual comparison of the two sets of risk layers.

[0043] When performing a time-series change comparison analysis of two sets of risk-related layers, unify the two sets of layers to the same spatial reference coordinate system and set the initial transparency so that both sets of layers can be displayed within the current view area. Add an interactive roll-up bar to the geographic information display interface. The roll-up bar can be set to vertical mode (sliding along the longitude direction) and / or horizontal mode (sliding along the latitude direction).

[0044] When a user drags the shutter strip in vertical shutter mode, the layer at the first moment is set to be visible on one side of the shutter strip, and the layer at the second moment is set to be visible on the other side, using the position of the shutter strip within the longitude range of the current view as the boundary. The visibility range of each layer is updated in real time as the shutter strip moves. In horizontal shutter mode, similar control is applied based on the position of the shutter strip within the latitude range of the current view. Narrow gradient transition areas are preset on both sides of the shutter strip, allowing the two sets of layers to transition with a gradual change in transparency at the boundary, avoiding obvious splicing marks.

[0045] During the movement of the roller shutter strip, the identified areas of difference and the corresponding difference annotation information are kept updated in sync with the current view, so that when the user drags the roller shutter strip, they can intuitively compare the risk distribution differences between the two sets of layers at the same position and learn about the specific changes through the difference annotation text.

[0046] During implementation, spatial alignment (accuracy ≤ 1m) is achieved based on the WGS84 coordinate system, with a default layer transparency of 0.9; when the user drags the roller shutter bar, the visible range of Layer A in vertical roller shutter mode is [ X min , X s The visible range of LayerB is []. X s , X max In the formula: X min , X maxThe longitude boundary of the current view; the visible range of Layer A in horizontal roll-up mode is [ Y min , Y s The visible range of LayerB is []. Y s , Y max In the formula: Y min , Y max This represents the latitudinal boundary of the current view. The target contour is extracted using the U-Net image segmentation method; the similarity between the two sets of regions is calculated using the Intersection over Union (IoU) algorithm, as shown in the following formula: In the formula: S intersect The area of ​​intersection of the two groups of regions. S union The area of ​​the union is used. When IoU < 0.9, it is considered a significant difference. When IoU < 0.9, the difference region is automatically labeled and a description is generated.

[0047] Specifically, areas of difference are marked with red dashed boxes, and a description of the difference is generated (such as "The water-rich anomaly area in 2024 expanded by 15% compared to 2023, with the main expansion direction being northeast"). The marking information moves synchronously with the roller shutter strip to ensure that users can view the details of the difference in real time.

[0048] In this implementation, two sets of risk-related layers are compared and displayed in the same view through the interactive roller shutter method, which greatly reduces the cognitive burden of users switching between different views, significantly improves the intuitiveness and efficiency of risk change identification, and makes it easier for technicians to quickly locate risk change areas and trends in daily analysis and training.

[0049] In the above implementation, preferably, the emergency rescue path analysis in the multidimensional risk analysis process is used to plan the optimal rescue route from the rescue starting point to the accident location in the event of an accident or in a drill scenario.

[0050] First, real-time geographic information, including road congestion status, weather information, and underground passage access status, is collected. Accident site information (including accident location coordinates, accident type, risk impact range, and rescue needs) and rescue resource information (including the location of rescue teams, the location of rescue equipment storage points, and the location of medical treatment points) are also collected.

[0051] Outlier identification and standardization preprocessing are performed on the above multi-source data. For example, statistical methods are used to remove obviously abnormal sensor readings, and attribute values ​​of different dimensions are normalized to a unified range to construct a multi-dimensional data source suitable for path planning.

[0052] Based on multidimensional data sources, a rescue road network model covering both surface roads and underground passages is constructed, representing roads or passages as sets of road segments. For each candidate road segment, multidimensional attribute values ​​are calculated, including segment length, predicted travel time, segment safety factor, and resource accessibility factor: travel time can be estimated based on current speed and congestion status; the safety factor can be attenuated based on the spatial distance between the road segment and the risk area; and the resource accessibility factor can be weighted based on the presence of resources such as medical points, equipment depots, and refuge chambers near the road segment. A path cost function is constructed based on the above attributes and their corresponding weight coefficients.

[0053] Employ path search algorithms based on path cost functions (such as improved Dijkstra's or A*). The algorithm searches for a set of candidate paths between the rescue starting point and the accident location, calculates the comprehensive cost of each path, selects the path with the lowest comprehensive cost as the current emergency rescue route, and overlays this route on the geographic information display interface as a polyline or arrow, while providing key node prompts and road segment attribute descriptions. Subsequently, dynamic data is reacquired and the cost of each road segment is updated at preset time intervals. When the rate of change of the total cost of the current emergency rescue route exceeds a preset threshold, the path search is re-executed and the emergency route and resource allocation suggestions are updated.

[0054] In the implementation process, the formula for constructing the path cost function using the improved Dijkstra algorithm is as follows: in, e For road segments in the path, L ( e The length of the road segment is (normalized to 0-1, calculated by the ratio of the actual length of the road segment to the maximum length of the road segment in the area). T ( e The time taken to travel on the road segment (including congestion and delays, standardized to 0-1) is the time it takes for the road segment to travel. T ( e = Actual travel time / Smooth travel time); S ( e The safety factor for a road section (valued between 0 and 1) is calculated by avoiding high-risk areas, such as areas affected by sudden water inrush or collapsed sections, based on the distance from high-risk areas; the closer the distance, the better. S ( e (the smaller) R ( eThe resource accessibility coefficient (value 0-1; the richer the surrounding rescue resources, the better) is. R ( e The larger the area, the better, especially if there are medical facilities nearby. R ( e (Increase by 0.2); ω1=0.2, ω2=0.4, ω3=0.3, ω4=0.1 are weighting coefficients, which can be dynamically adjusted according to the type of rescue (such as flood rescue, roof collapse rescue). For example, in flood rescue, the weighting coefficient can be increased by 0.2. S ( e The algorithm assigns a weight (ω3=0.4) to each road segment, prioritizing the avoidance of flooded sections. It calculates the overall cost of each segment and selects the "minimum cost path" from the rescue starting point to the accident site as the optimal rescue route. The path calculation time is ≤3 seconds, meeting real-time requirements.

[0055] The path cost is recalculated every 30 seconds, based on the path cost change rate. It automatically updates the planned route and outputs the optimal route, navigation instructions, and resource allocation suggestions.

[0056] In this implementation, considering multiple factors such as distance, time, safety, and resource accessibility, dynamic optimization of emergency rescue routes is achieved. Compared with a single shortest path or shortest time path scheme, it can better meet the requirements of "safety first, efficiency in balance" in complex coal mine rescue scenarios, shorten the time for rescue forces to reach the accident site, and reduce secondary risks during the rescue process.

[0057] In the above implementation, preferably, the knowledge retrieval based on the constructed knowledge base in the multidimensional risk analysis process is used to support intelligent risk question answering in both text and map point selection methods.

[0058] First, a coal mine-specific structural knowledge base is constructed, which uniformly manages structured structural unit attribute data (such as fault names, strikes, dip angles, water content, and stability assessments), unstructured geological exploration reports and historical risk cases, as well as knowledge graph data based on expert experience. Unstructured text is segmented, entity extracted, and relation identified to form indexable knowledge units. Representation vectors or index features are generated for each knowledge unit, constructing a knowledge index structure oriented towards question-and-answer retrieval.

[0059] When a user inputs a natural language question via text query, the input text undergoes word segmentation, stop word filtering, and semantic encoding to obtain a question vector. This vector is then compared with candidate data vectors in the knowledge base, and several results are selected from those with similarity scores not lower than a preset threshold. Based on these candidate data, a structured answer containing structural unit attributes, typical risk characteristics, and prevention and control suggestions is generated according to a preset template. This answer is displayed in text and list formats in the sidebar of the geographic information interface, while relevant structural units or risk areas are highlighted on the map.

[0060] When a user selects a construction unit in the construction layer by clicking on a map, the method queries the corresponding construction unit identifier based on the coordinates of the selected location, automatically generates a standardized query request, and performs the same semantic retrieval and answer generation process as input for a natural language question or parameterized query. Finally, the answer content is associated with and displayed with the selected spatial object.

[0061] Specifically, a dedicated knowledge base for coal mine geological tectonics is constructed based on "structured data - unstructured data - expert experience," covering all dimensions of geological structural information: Structured data includes tectonic unit attributes (name, type, scale, formation age), geological characteristics (lithology, water content, stability coefficient), and regional distribution patterns (such as fault strike and collapse column distribution density), and is stored in a relational database. Unstructured data includes geological exploration reports, historical risk cases (such as details of a water inrush accident caused by a fault), and expert analysis documents. These are converted into text data using OCR technology and indexed in full text using Elasticsearch. Expert Experience: Experience and knowledge from coal mine geology experts (such as "methods for determining the water content of a certain structural unit") are collected through interviews and questionnaires. Knowledge graph technology is used to construct the relationship between "structural unit - risk characteristics - remediation suggestions." The knowledge base employs a maintenance mechanism of "monthly updates + real-time supplementation," regularly crawling the latest geological research results in the industry using web crawling technology, while also supporting manual supplementation by experts to ensure the timeliness and completeness of the knowledge base.

[0062] During implementation, the DeepSeek public API and a coal mine-specific knowledge base were integrated. The knowledge base was built based on structured data, unstructured data, and expert experience, and was maintained using a "monthly update + real-time supplement" mechanism.

[0063] Natural language question answering uses the BERT model to parse the question intent, extracts core elements, and then uses a vector matching algorithm to search the knowledge base. The similarity formula is: In the formula: Q Let the problem vector be...D j For the first in the knowledge base j Given a data vector, set a similarity threshold of 0.6, and return the top 5 most similar data. Structured answers are generated using a fine-tuned GPT-3.5 model (generation time ≤ 5 seconds); map point-based question and answer uses a spatial coordinate matching algorithm to obtain the structural unit attribute ID, automatically generates standardized questions and sends them to the DeepSeek API, and displays the answers in a pop-up window, including structural unit attributes, risk cases and prevention and control suggestions.

[0064] In this implementation, by constructing a dedicated knowledge base and a semantic retrieval mechanism, intelligent question-and-answer and visual presentation of coal mine structural risk knowledge are realized. This reduces the workload of frontline personnel in reviewing lengthy geological reports and manually summarizing key risk points, significantly lowers the threshold for acquiring professional knowledge, and improves the efficiency and accuracy of structural risk assessment and prevention and control plan formulation.

[0065] In the above embodiments, preferably, the parsing process of the text search operation in the interactive operation is used to convert the characters entered by the user through the search box into a set of target objects.

[0066] First, a multi-level administrative geocoding database covering coal mining areas is constructed. Entries are created for each administrative region in this database, recording its name, administrative division code, boundary geometric information, and geometric center coordinates. Mapping relationships are also established for commonly used abbreviations, former names, or aliases. For registered coal mines, the mine name, aliases, and information about the administrative region to which they belong are recorded and linked to the geocoding database.

[0067] When a user enters search text in the search box of the geographic information display interface, the search text is segmented and filtered for stop words. A trained text classification model is used to determine the type of search intent, such as distinguishing between administrative region search or coal mine name search.

[0068] In the administrative region search scenario, an exact match is first performed in the geocoding database. If no exact match is found, the similarity between the search text and the names of each administrative region is calculated, for example, based on edit distance or other string similarity metrics. A set of administrative regions with a similarity of no less than a threshold is then selected, sorted by similarity, and a preset number of candidate options are returned for the user to choose from. A similar matching process is performed in the coal mine name search scenario.

[0069] Once the target administrative region or target coal mine is determined, the appropriate zoom level of the map is calculated based on its boundary range or mining area. The center of the map view is located at the geometric center coordinates of the region or coal mine. The relevant boundaries or markers are highlighted on the map, and the target administrative region or coal mine is added to the target object set for subsequent analysis processes such as spatial statistical analysis, time series comparison, emergency route planning, and knowledge retrieval.

[0070] During implementation, precise matching and redirection are achieved based on natural language parsing and geocoding mapping technologies. A multi-dimensional mapping relationship database is constructed by integrating data from four levels of administrative regions in coal mining areas across the country, supporting synonym and abbreviation matching. After the user enters the search text T, intelligent parsing is achieved through the following steps: Step 1: After the user inputs text T, it is preprocessed by jieba word segmentation and the Naive Bayes classifier is used for intent recognition. T is then split into a keyword sequence (e.g., "Shanxi Province Datong City Mining Area" is split into "Shanxi Province", "Datong City", and "Mining Area"), and stop words (e.g., "of" and "located in") are removed. Step 2: Intent recognition. A trained Naive Bayes classifier is used to determine the type of user request (administrative region search / coal mine name search / risk area search), with a classification accuracy of ≥95%. Step 3: Fuzzy matching. If exact matching fails (e.g., no completely matching name is found when "Datong Mining Area" is entered), the similarity is calculated using the edit distance algorithm. The formula is: in, T i For the first in the database i The name of an administrative region, Edit ( T , T i ) for T Convert to T i The minimum number of character insertions, deletions, or replacements required. Len (•) represents the string length. Set the similarity threshold. Sim 0 = 0.7, when Sim ( T , T i )≥ Sim When the value is 0, the Top 3 matching results are returned for the user to choose from.

[0071] When switching views, the map zoom level Z is adjusted by the length L of the diagonal of the bounding rectangle. The calculation formula is as follows: in, Z min =3、Z max =12 is the upper and lower limits of the scaling level. L min =1000m L max =1000000m is the diagonal length threshold for the corresponding zoom level. Simultaneously, the map center point is located to the geometric center of the administrative region. X avg , Y avg ), , , (( X i , Y i (where n is the coordinate of the vertex of the administrative region boundary and n is the number of vertices), to achieve precise "one-step" jump.

[0072] In this implementation, the precise mapping of user text input to spatial target objects is achieved through geocoding database, intent recognition, and fuzzy matching strategies. This significantly improves the search and location efficiency of coal mines and administrative regions, avoids the inefficient operation of users manually dragging and zooming on the map to find target areas, and helps to quickly focus on areas of risk concern.

[0073] In the above embodiments, preferably, the multi-dimensional coal mine risk identification method also includes a scene freezing and backtracking process, which is used to save key spatial views during the analysis process and restore them when needed.

[0074] After a user completes a risk analysis or locates a key area (such as a high-risk work surface and its surrounding water-rich anomaly area), they can trigger a scene freeze process via the "Save Scene" option on the interface. The method reads a set of key parameters from the current map view, including the current map zoom level, view center coordinates, the set of currently visible layer identifiers, information on drawn labeled objects, and the scene save time. These parameters are organized into structured keyframe data according to preset fields, stored using formats such as JSON, and indexed with user-defined or automatically generated tags. A thumbnail of the current view is also generated for list display. The keyframe data can be stored in a dedicated data storage device and supports retrieval by time or tag.

[0075] When a user subsequently issues a "scene rewind" command and selects a keyframe, the method reads the zoom level and center coordinate parameters of that keyframe from storage. Within a preset transition time (e.g., one to two seconds), it gradually adjusts the zoom level and center coordinates of the current view using an interpolation algorithm, allowing the view to smoothly transition to the position and scale corresponding to the target keyframe. After the transition is complete, the corresponding layer is loaded based on the visible layer identifiers recorded in the keyframe, and the labeled objects recorded in the keyframe (e.g., high-risk area boundaries, path results, text annotations, etc.) are restored. The restored scene is then added to the risk identification result data for direct use by the user in meetings, training sessions, or subsequent analyses.

[0076] Specifically, the keyframe data structure defines that each scene keyframe is stored in JSON format and contains the following core fields to ensure the integrity and reusability of view information: {"sceneID":"SC20240520001", / / Unique identifier for keyframes} "sceneName":"Water-rich abnormal zone at working face No. 2", / / User-defined name "zoomLevel": 11.5, / / Map zoom level "centerX":112.356, / / Longitude of the center point "centerY":37.892, / / latitude of the center point "visibleLayers":["waterLayer","faultLayer"], / / List of visible layers "annotations":[{"id":"A01","content":"Abnormal water inflow point"}], / / annotation information "saveTime":"2024-05-2014:30:25", / / save time "tags":["Water Hazard Risk","Workface No. 2"] / / Custom tags (for retrieval)}.

[0077] During implementation, the scene freezing process supports one-click saving of key parameters of the current view to generate structured keyframes. These keyframes are stored in JSON format and include core fields such as sceneID, zoomLevel, centerX, centerY, and visibleLayers. After the user selects a target keyframe, a progressive adjustment algorithm is used to achieve smooth view rewinding. The parameter change formula is as follows: in, t For transition time 0≤ t ≤ T , T =1000 ms For the total transition time, ( Z target , X target , Y target () represents the target parameters in the keyframe. After the transition is complete, the system automatically loads the visible layers and annotation information recorded in the keyframe, achieving "one-click restoration" of the view.

[0078] In this implementation, the analysis scene is frozen and traced back by using keyframes, which reduces the operational costs for users to repeatedly zoom, drag and select layers in complex map environments. This allows important risk scenes and analysis views to be saved for a long time and reproduced quickly, improving the traceability of risk assessment and the efficiency of collaborative communication.

[0079] In the above embodiments, preferably, a geographic information model of the coal mining area is constructed based on the three-level spatial logic of base-mining area-coal mine and the above data is presented in a hierarchical manner to support the construction of geographic information model and the switching of view levels.

[0080] During the data modeling phase, based on the actual organizational structure and spatial distribution characteristics of coal mine safety supervision, spatial objects are divided into three layers: the base layer, the mining area layer, and the coal mine layer. The base layer is associated with macro-level data such as administrative boundaries, major transportation routes, major water systems, and the scope of major energy bases; the mining area layer is associated with meso-level data such as specific mining area boundaries, the division of multiple coalfield blocks, the distribution of geological structures, aquifer boundaries, and risk zoning; and the coal mine layer is associated with micro-level data such as the location of each coal mine shaft, the location of mining faces, the layout of monitoring points above and below ground, old working water areas, and water-rich anomaly areas. All of these multi-layered data are uniformly registered in the same spatial reference coordinate system, and their respective layers, visibility rules, and relationships with other layers are recorded in the data structure.

[0081] During the map display phase, preset zoom level thresholds are set for the base layer, mining area layer, and coal mine layer. For example, at a smaller zoom level, only base layer elements are displayed; at a medium zoom level, mining area layer elements are displayed; and at a larger zoom level, coal mine and working face related elements are displayed.

[0082] Based on the relationship between the current map zoom level and the threshold values ​​of each level, the target level to be displayed is determined, and a progressive loading strategy is adopted: when the view switches from the base level to the mining area level or from the mining area level to the coal mine level, the data of the target level is loaded in blocks according to the spatial range in order to control the front-end rendering pressure and network traffic.

[0083] Meanwhile, after the level switch is completed, the focus offset is calculated based on the coordinates of the core spatial objects in the target level (such as the center of the target mining area or the target coal mine shaft) and the coordinates of the current view center. Within a preset time period, the view center is interpolated and updated through a smoothing function, so that the map view naturally moves to the geometric center of the core spatial objects, improving the user's spatial perception of key risk areas.

[0084] In this implementation, hierarchical data modeling defines the core data elements and accuracy standards for each level, ensuring the orderly connection of information from the macro to the meso to the micro levels. Base level (Level 1): Covers provincial / regional coal mine management units, including macro information such as administrative boundaries, transportation arteries, major water systems, and mining area distribution. The data accuracy is 1:100,000. It is collected through provincial geographic information databases and satellite remote sensing images, retaining only key elements that affect regional risks (such as the distribution of large aquifers). Mining Area Level (Level 2): ​​Focuses on the scope of a single mining area, including mesoscopic information such as geological structures (faults, collapse columns), aquifer boundaries, mine locations, and risk zones. The data accuracy is 1:50,000. It is constructed by mining area 3D seismic reports and geological exploration data, and automatically hides redundant information in the base level that is not related to mining area risks. Coal Mine Level (Level 3): Refined to the internal workings of the coal mine, including micro-information such as the location of the mining face, monitoring points (water pressure and water inflow sensors), old workings with water, and water-rich anomaly areas. The data accuracy is 1:10000. It is generated from the coal mine mining plan and underground measured data, and only elements related to the real-time risks of the coal mine are loaded.

[0085] In this implementation, a hierarchical visualization of the coal mining area from macro to micro is achieved through a three-level spatial logic modeling and hierarchical switching strategy. This effectively avoids the "information overload" caused by overlaying too many elements in a single view, helping users quickly focus on risk-related mining areas and coal mines, and providing an intuitive spatial basis for subsequent statistical analysis, trend comparison and path planning.

[0086] like Figure 3 As shown, the present invention also proposes a multi-dimensional coal mine risk identification system, which applies the multi-dimensional coal mine risk identification method disclosed in any of the above embodiments, including an integrated data modeling unit, an interactive operation parsing unit, a multi-dimensional risk analysis unit, and a risk object display unit. Each unit can be deployed on the server and client in the form of software modules or services.

[0087] The integrated data modeling unit is used to acquire spatial geographic data, basic attribute data, monitoring data, and structural geological data covering the coal mining area from external business platforms, monitoring terminals, and historical archives. It performs format conversion, coordinate unification, and quality verification on the data. Based on the three-level spatial logic of base-mining area-coal mine, it constructs a geographic information model of the coal mining area and presents the above data in a hierarchical manner in the geographic information display interface, supporting multi-scale map browsing.

[0088] The interactive operation parsing unit is used to receive user interactive operations on the geographic information display interface for target areas and / or target coal mines, including text search for administrative regions or coal mines, selection of target areas through maps, and switching of layers, etc., and parses the interactive operations into a set of target objects based on spatial location relationships and geocoding, which serve as input for subsequent analysis.

[0089] The multidimensional risk analysis unit is used to perform a multidimensional risk analysis process for a set of target objects according to a preset multidimensional analysis dimension. The multidimensional risk analysis process includes at least one of spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base, and generates risk identification result data related to the set of target objects. The risk object display unit is used to associate and display the risk identification results data with spatial objects in the geographic information model of the coal mine area in the geographic information display interface. It outputs the analysis results through map highlighting, thematic rendering, differential area annotation, path drawing and statistical chart panel, and supports advanced interactions such as roll-up comparison and scene freeze-frame retrospection, so as to realize the integrated closed loop from data collection, analysis and calculation to decision support.

[0090] The integrated data modeling unit, interactive operation parsing unit, multidimensional risk analysis unit, and risk object display unit are configured to work together to execute a multidimensional coal mine risk identification method.

[0091] During implementation, the system environment includes hardware, software, and network environments. The hardware environment includes a server cluster (3 main servers + 2 backup servers, computing power ≥ 20 TOPS), edge computing nodes (response latency ≤ 300ms), interactive terminals, and data acquisition devices. The software environment uses CentOS 8.5 (server-side) and Windows 11 Professional (desktop terminal) operating systems, integrating ArcGIS Pro 3.1 geographic information engine, a hybrid database architecture of "MySQL 8.0 + MongoDB 6.0 + Redis 7.0", Spring Cloud Alibaba backend framework, Vue 3.0 frontend framework, and multiple AI algorithm libraries. The network environment adopts a hybrid underground architecture of "industrial Ethernet + 5G wireless communication" (bandwidth ≥ 1Gbps, latency ≤ 20ms), with an enterprise dedicated line (bandwidth ≥ 500Mbps) above ground, supporting VPN remote access.

[0092] In this implementation, the integration of multi-source spatial data, interactive analysis, multi-dimensional risk analysis and visualization are organically combined through integrated hardware and software configuration. This fully implements the methods and steps of the above implementation, significantly improving the automation and intelligence level of coal mine risk identification and decision support, and adapting to the coal mine safety management needs under complex working conditions such as multi-aquifer interaction and dense geological structures.

[0093] According to the multi-dimensional coal mine risk identification system disclosed in the above embodiments, the functions to be implemented by each module correspond to the steps of the multi-dimensional coal mine risk identification method disclosed in the above embodiments. In the implementation process, the operation is carried out with reference to the above embodiments, and will not be repeated here.

[0094] The multi-dimensional coal mine risk identification method and system disclosed in the above embodiments will be specifically described through the following examples.

[0095] Example 1: Application of Water Hazard Risk Prevention and Control in a Certain Haizi Coal Mine This coal mine is a large underground mine with a mining depth of 900-1150m. It mainly faces the risk of combined water hazards from Jurassic sandstone water and old workings water. Multiple water-bearing faults are distributed within the mining area, making it crucial to prevent water inrush at the working face during mining operations. Currently, the mine has implemented this system for training and use to assist frontline personnel in quickly understanding risk distribution and improving the efficiency of emergency response plan development. The specific application process is as follows: Map navigation and search positioning: Frontline technicians can use the three-level navigation function of "base (the coal base) - mining area (the Haizi mining area) - coal mine (target coal mine)" to focus on the No. 3 mining face from the macro area within 10 seconds. At the same time, by entering "the No. 3 working face of the Haizi coal mine" for precise search, the system will jump to the target area within 1 second and clearly display the distribution of faults around the working face, the location of monitoring points and the distribution range of old working water, helping technicians to quickly establish spatial risk awareness; Scene freezing and statistical calculation: For high-risk areas around Working Face No. 3, technicians can save the current view as a keyframe with one click (named "Fault Affected Area of ​​Working Face No. 3"). When conducting team briefings and plan discussions, they can revert to this scene within 1 second without repeatedly adjusting map parameters. By selecting a 3km radius around the working face, the system automatically calculates the distribution information of 8 water pressure monitoring points, 3 water inflow monitoring points, and 2 emergency material reserve points within the area, providing data support for risk assessment and material allocation plan formulation. Roller shutter analysis and risk prediction: Technicians load the water-rich anomaly zone layers from the two most recent geophysical explorations of the working face. Through the vertical roller shutter comparison function, they can intuitively observe the changing trend of the anomaly zone range. The system automatically marks the difference areas to help determine the development direction of the water-rich anomaly and provide a reference for adjusting the mining pace and optimizing the monitoring plan. Emergency rescue route pre-planning: Combining the layout of underground passages and the distribution of surface roads in the coal mine, technicians simulated a sudden water seepage accident at the No. 3 working face. Based on the preset location of the rescue base and the distribution of material storage points, the system planned three emergency routes, marking the estimated travel time, key nodes and resource accessibility of each route, providing a scientific basis for the development of targeted emergency response plans. Intelligent Q&A on tectonic maps: When technicians select the F3 fault near the working face on the map, the system immediately returns the basic geological information of the fault (lithology: siltstone, water content: medium, structural scale: strike length 3.2km) and related risk prevention and control suggestions ("It is recommended to increase the density of monitoring points around the fault, and focus on monitoring water pressure changes"), which helps frontline personnel quickly grasp the risk characteristics of the tectonic unit and improve the professionalism of on-site prevention and control.

[0096] Application feedback: Through learning and using this system, front-line technicians have significantly improved their understanding of the spatial distribution and key influencing factors of coal mine water hazard risks. The time for risk assessment and contingency plan formulation has been shortened by more than 40% compared with traditional methods. The convenience of obtaining key information such as geological structure and monitoring data has been greatly improved, laying a solid foundation for risk prevention and control in subsequent actual production.

[0097] Example 2: Learning and Application of Multi-Type Risk Supervision in a Mine This coal mine is an expanded and renovated mine with a mining area of ​​18 km². 2The mine comprises multiple mining faces and primarily faces roof fall and water hazard risks arising from geological structures, necessitating integrated monitoring of multiple areas and types of risks. Currently, this system has been implemented in the mine for daily risk monitoring training, assisting managers and frontline personnel in improving their risk identification and handling capabilities. The specific application process is as follows: Map navigation and statistical analysis: Managers can quickly view the overall distribution of the coal mine at the base level, switch to the mining area level to analyze the macroscopic distribution characteristics of faults and aquifers within the mining area, and then focus on a single mining face to view the layout of monitoring points and mining progress; by selecting the entire mining area on a monthly basis, the system automatically calculates quantitative indicators such as the monitoring point coverage rate and the proportion of high-risk areas for the 10 mining faces within the mining area, helping managers quickly grasp the key points of risk supervision for the entire mine; Roller shutter analysis and trend judgment: Technicians loaded roof displacement monitoring data layers and water-rich anomaly zone layers from different periods. By comparing the roller shutters, they could intuitively observe the data change trends. The system helped identify that the roof displacement of the two working faces in the southeast showed a slow increasing trend, and the water-rich anomaly zone in the western working face slightly expanded. This provided a reference for taking preventive measures in advance and optimizing the monitoring plan. Intelligent Q&A on Geotectonics Map: When managers are concerned about risk management in areas with dense faults within the mining area, they can input "the main risk types and key prevention and control points of the dense fault area in this mine" into the system. The system will return structured analysis results within 5 seconds, clarifying that the core risks in the area are fault water conduction and roof instability, and providing targeted prevention and control suggestions. At the same time, the system will mark the extent of the dense fault area on the map to assist in the development of differentiated supervision plans. Emergency Drill Support Planning: In emergency drills for roof collapse accidents organized in coal mines, technicians use this system to simulate accident scenarios. Based on the underground passage conditions and the deployment location of rescue forces, the system plans the optimal rescue route and marks key risk points and precautions during the rescue process, providing technical support for improving the relevance of drills and optimizing emergency response procedures.

[0098] Application Feedback: Through learning and using this system, the risk supervision efficiency of coal mine managers has been significantly improved, enabling them to quickly integrate multi-dimensional risk data and accurately identify key areas for supervision; frontline personnel have had greater access to professional geological information and emergency response suggestions, and their business capabilities have been effectively assisted, providing strong support for coal mines to establish a scientific and efficient risk supervision system and enhance their safety production assurance capabilities.

[0099] The core value of this invention lies in constructing a coal mine risk identification system based on "multi-dimensional data intelligent fusion, dynamic interactive analysis, and AI-driven decision support." The implementation process can be tailored to specific coal mine geological conditions, risk types, and management needs, and is not limited to fixed parameters and configurations. Functional adaptation: For coal mines with prominent water hazard risks, the weight of "avoiding waterlogged sections" in emergency rescue route planning can be strengthened, and the water-rich anomaly zone identification algorithm in the curtain analysis can be optimized; for coal mines with complex geological structures, the knowledge base of intelligent question-and-answer on tectonic maps can be expanded, and the stability assessment index of tectonic units can be added. Hardware compatibility: Supports access to existing sensors and monitoring equipment in coal mines. Data integration can be achieved simply by adapting the protocol, without the need for complete replacement, thus reducing transformation costs. The server cluster supports horizontal scaling. When a coal mine adds a new working face or expands the monitoring scope, computing power can be increased by adding node servers. Software Expansion: Adopting a modular design, it can add special function modules (such as gas risk identification module and roof risk early warning module), and coordinate with existing functions through API interface to meet personalized needs.

[0100] The effectiveness of this system relies on the coordinated operation of seven functional modules: map navigation and search positioning as the foundation, enabling precise information focusing; scene capture and statistical calculation as the core, supporting efficient analysis; rolling screen analysis and emergency planning as the key, enabling dynamic risk management; and intelligent question answering as a supplement, lowering the barrier to information access. These modules support each other, forming a closed loop of "data-analysis-decision," driving the transformation of coal mine risk identification from "experience-driven" to "data-driven and AI-driven."

[0101] The above are merely preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-dimensional coal mine risk identification method, characterized in that, include: Spatial geographic data, basic attribute data, monitoring data, and structural geological data covering the coal mining area are acquired on the geographic information platform. A geographic information model of the coal mining area is constructed based on the three-level spatial logic of base-mining area-coal mine, and the above data are presented in hierarchical order on the geographic information display interface. The system receives user interaction operations on the geographic information display interface targeting a target area and / or a target coal mine, and parses the interaction operations into a set of target objects based on spatial location relationships and geocoding. The interaction operations include at least one of text search, map selection, or layer switching. For the target object set, a multi-dimensional risk analysis process is executed according to a preset multi-dimensional analysis dimension to generate risk identification result data related to the target object set. The multi-dimensional risk analysis process includes at least one of spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base. In the geographic information display interface, the risk identification result data is associated with the spatial objects in the data model and displayed, and the coal mine risk identification results are output in the form of map view and / or statistical chart.

2. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, Spatial statistical analysis in the multidimensional risk analysis process includes: Upon receiving a spatial statistical analysis instruction from a user for the target area, the target area is determined by at least one of the following methods: polygon drawing, rectangular selection, or administrative region selection. Based on the ray method, the topological relationship between the center point coordinates of each coal mine in the system and the boundary of the target area is determined to see if the center point of the coal mine falls within the target area, and the set of coal mines within the target area is obtained by filtering. Based on the coal mine set within the target area and the monitoring point data corresponding to the target area, calculate at least one or more of the following spatial statistical indicators: the number of coal mines within the target area, the total designed production capacity of coal mines within the target area, the total average water inflow of coal mines within the target area over a preset period, the proportion of high-risk mines obtained based on the coal mine risk level identification, and the monitoring point density obtained based on the number of monitoring points and the area of ​​the target area. The spatial statistical indicators are then written into the risk identification result data.

3. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, The time-series change comparison analysis in the multidimensional risk analysis process includes: Upon receiving a time-series change comparison analysis instruction for the target object set, load two sets of risk-related layers formed at the first and second time points for the same target area and / or target coal mine set; The two sets of risk-related layers are spatially aligned based on a unified spatial reference coordinate system to ensure that the coordinates of the same geographical location are consistent in the two sets of layers. An image segmentation algorithm is used to extract the contour information of the target region of a preset type from the two sets of risk-related layers to obtain the contour of the first target region and the contour of the second target region. Based on the intersection-union ratio (IUU) calculation rule, the similarity between the outline of the first target region and the outline of the second target region is calculated. When the similarity is lower than a preset threshold, it is determined that there is a significant difference between the two sets of risk-related layers at the corresponding locations. The difference regions are spatially labeled and described in the geographic information display interface, and the difference labeling information is written into the risk identification result data.

4. The multi-dimensional coal mine risk identification method according to claim 1 or 3, characterized in that, When performing a time-series change comparison analysis of two sets of risk-related layers, a roll-up interactive method is used to display the layer comparison. This roll-up interactive method includes: When loading two sets of risk-related layers that need to be compared, unify the two sets of layers to the same spatial reference coordinate system and set the initial transparency of the two sets of layers; Add vertical and / or horizontal roller shutter strips to the geographic information display interface. When the user drags the roller shutter strip, the first group of layers is made visible on one side of the roller shutter strip and the second group of layers is made visible on the other side of the roller shutter strip, depending on the position of the roller shutter strip within the longitude or latitude range of the current view. The visibility range of the two groups of layers is adjusted in real time. A gradient transition area is set on both sides of the roller shutter strip to reduce the impact of the layer segmentation boundary on the visual effect, and to keep the two sets of layers and difference annotation information updated synchronously during the movement of the roller shutter strip.

5. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, The emergency rescue path analysis in the multidimensional risk analysis process includes: It acquires real-time geographic information, including road congestion status, weather information, and underground passage access status; accident site information, including accident location coordinates, risk impact range, and rescue needs; and rescue resource information, including the location of rescue teams, the location of rescue equipment storage points, and the location of medical points. Outlier identification and standardization preprocessing are performed on the above multi-source data to construct a multi-dimensional data source for path planning; A rescue road network model is constructed based on the multidimensional data source. For each candidate road segment in the rescue road network, multidimensional attribute values ​​including road segment length, road segment travel time, road segment safety factor and resource accessibility factor are calculated. A path cost function is constructed based on each attribute value and the corresponding weight coefficient. A path search algorithm based on the path cost function is used to determine the path with the minimum comprehensive cost from the candidate paths from the rescue starting point to the accident location in the rescue road network as the emergency rescue route. The emergency rescue route is written into the risk identification result data and displayed in the geographic information display interface. Furthermore, the system reacquires the multi-source dynamic data within a preset time interval and updates the path cost of the candidate road segments. When the path cost change rate of the current emergency rescue route exceeds a preset threshold, the system re-executes the path search to update the emergency rescue route and outputs the updated route information and resource allocation suggestions.

6. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, The knowledge retrieval based on the constructed knowledge base in the multidimensional risk analysis process includes: A coal mine-specific structural knowledge base is constructed, which includes structured structural unit attribute data, unstructured geological exploration reports and historical risk cases, and knowledge graph data constructed based on expert experience. When a text query request for the target object set is received, the natural language question input by the user is semantically encoded to obtain a question vector. The similarity is calculated with the candidate data vector in the coal mine-specific structural knowledge base. A preset number of candidate results are selected from the candidate data with a similarity of not less than a preset threshold. Based on the candidate results, an answer containing structural unit attributes, risk characteristics and prevention and control suggestions is generated. Upon receiving a construction query request triggered by map point selection, the corresponding construction unit identifier is queried based on the coordinates of the selected location. A standardized query request is automatically generated and input as a natural language question into the knowledge retrieval process based on the construction knowledge base for semantic encoding and retrieval processing. The generated answer content is then associated with spatial objects in the target object set and displayed. Construction risk knowledge is written into the risk identification result data.

7. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, The parsing process for the text search operation in the interactive operation includes: Construct a multi-level administrative region geocoding database covering coal mining areas. Record the name, administrative division code, geometric center coordinates and boundary coordinates of each administrative region in the geocoding database, and establish mapping relationships for commonly used abbreviations and synonyms. The search text entered by the user is segmented and stop words are filtered. Based on the trained classification model, the search intent is identified and distinguished into administrative region search or coal mine name search. In scenarios such as administrative region search or coal mine name search, precise matching is first performed in the geocoding database. If precise matching fails, the similarity between the search text and the names of each administrative region in the geocoding database is calculated based on the edit distance. A set of candidate administrative regions is determined in the administrative regions with similarity not lower than a preset threshold, and a preset number of candidate results are returned for the user to choose from. After determining the target administrative region or target coal mine, the map zoom level is calculated based on its boundary range, and the center of the map view is located to the geometric center coordinates of the target area. The target administrative region or target coal mine is then added to the target object set.

8. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, It also includes a scene freeze and rewind process, which specifically includes: Upon receiving a scene save command, the zoom level, center coordinates, visible layer identifiers, labeled object information, and save time of the current map view are collected. The above parameters are organized into structured keyframe data according to preset fields and stored in the data storage system. At the same time, a thumbnail of the current view is generated and an index relationship between labels and keyframe identifiers is established. Upon receiving a scene backtracking instruction, the zoom level and center coordinate parameters in the target keyframe data are read from the data storage system. Within a preset total transition time, the zoom level and center coordinates of the current view are gradually adjusted using time interpolation. After the interpolation adjustment is completed, the visible layers and annotation information recorded in the target keyframe data are loaded to restore the corresponding scene view. The restored scene is then added to the risk identification result data for subsequent analysis.

9. The multi-dimensional coal mine risk identification method according to claim 1, characterized in that, The specific process of constructing a regional geographic information model of coal mining areas based on a three-level spatial logic of base-mining area-coal mine and presenting the above data hierarchically includes: The correlation at the base level should include at least macro data such as administrative boundaries, main transportation lines, major water systems and mining area distribution; the correlation at the mining area level should include at least meso data such as geological structure, aquifer boundaries, mine location and risk zoning; and the correlation at the coal mine level should include at least micro data such as the location of mining face, monitoring point location, old working water area and water-rich anomaly area. A preset zoom level threshold is set for the base level, mining area level, and coal mine level. Based on the relationship between the current map zoom level and the zoom level threshold, the current display level is determined, and a progressive loading strategy is used to load data elements of the target level in batches during the level switching process. After the level switch is completed, the focus offset is calculated based on the coordinates of the core spatial object of the target display level and the coordinates of the current view center. Within the preset focus transition time, the view center is interpolated and updated according to the smooth transition function, and the map view is smoothly moved to the geometric center position of the core spatial object.

10. A multi-dimensional coal mine risk identification system, characterized in that, The multi-dimensional coal mine risk identification method as described in any one of claims 1 to 9 includes: An integrated data modeling unit is used to acquire spatial geographic data, basic attribute data, monitoring data, and structural geological data covering the coal mining area. Based on the three-level spatial logic of base-mining area-coal mine, a geographic information model of the coal mining area is constructed, and the above data is presented hierarchically in the geographic information display interface. An interactive operation parsing unit is used to receive interactive operations from users on the geographic information display interface for target areas and / or target coal mines, and parse the interactive operations into a set of target objects based on spatial location relationships and geographic codes; A multidimensional risk analysis unit is used to perform a multidimensional risk analysis process for the target object set according to a preset multidimensional analysis dimension. The multidimensional risk analysis process includes at least one of spatial statistical analysis, temporal change comparison analysis, emergency rescue path analysis, and knowledge retrieval based on a constructed knowledge base, to generate risk identification result data related to the target object set. The risk object display unit is used to associate and display the risk identification result data with the spatial objects in the geographic information model of the coal mine area in the geographic information display interface, and output the coal mine risk identification result in the form of map view and / or statistical chart. The integrated data modeling unit, interactive operation parsing unit, multidimensional risk analysis unit, and risk object display unit are configured to work together to execute the multidimensional coal mine risk identification method.

Images