A mine disaster emergency command method and device based on a knowledge graph and a medium

Abstract

The application discloses a mine disaster emergency command method and device based on a knowledge graph, and a medium, relates to the technical field of emergency decision-making, and comprises the following steps: constructing a candidate causal event pair based on a disaster event sequence, forming a local causal fragment set through event type matching and correlation degree calculation, and mapping the local causal fragment set to a disaster knowledge graph to generate a disaster causal chain set and a disaster situation subgraph; searching a disposal path similar to a current disaster structure from historical disposal records and emergency plans according to the disaster causal chain set, and screening a candidate emergency strategy set in combination with resource availability; and constructing an anti-fact scenario set based on the disaster situation subgraph, performing disaster evolution deduction on the anti-fact scenario set, calculating personnel risk indexes, equipment loss indexes and production capacity influence indexes of each emergency strategy, forming a scheme evaluation result set, and screening and recommending an emergency scheme. The application realizes interpretable modeling of disaster evolution, and improves the scientificity and adaptability of mine emergency response.

Description

Technical Field

[0001] This invention relates to the field of emergency decision-making technology, and in particular to a knowledge graph-based method, equipment, and medium for emergency command in mine disasters. Background Technology

[0002] In the field of mine safety production and emergency management, the rapid development of IoT, big data, and artificial intelligence technologies has enhanced the real-time acquisition and fusion analysis capabilities of multi-source heterogeneous monitoring data, driving the evolution of disaster early warning and emergency response mechanisms towards intelligence. In recent years, knowledge graph technology, due to its advantages in semantic modeling, relational reasoning, and knowledge reuse, has been gradually introduced into industrial safety scenarios to construct disaster evolution logic models and auxiliary decision support systems. Some studies attempt to structurally associate historical accident cases, emergency plans, and real-time sensor data to form static or semi-dynamic knowledge representation systems to support the retrieval and recommendation of emergency plans.

[0003] Current mainstream mine emergency command technologies still have significant limitations. First, most systems rely on preset rules or statistical models for emergency response, failing to effectively integrate the deep causal logic between real-time monitoring data and historical handling experience. This results in recommended strategies being out of sync with the actual situation on-site, leading to insufficient adaptability. Second, in the emergency plan evaluation phase, qualitative or simplified quantitative indicators are generally used, lacking a multi-dimensional comprehensive benefit calculation mechanism based on counterfactual reasoning. This makes it impossible to accurately predict the comprehensive impact of different strategies on personnel safety, equipment protection, and production recovery, easily causing secondary losses or resource misallocation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a knowledge graph-based emergency command method for mine disasters to address the problems of poor adaptability of emergency strategies and one-sided assessments caused by the lack of real-time causal modeling and counterfactual multi-dimensional deduction capabilities in existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a mine disaster emergency command method based on knowledge graphs, which includes, Collect multi-source mine monitoring data, perform time synchronization and spatial coding, and generate disaster event sequences; Based on disaster event sequences, candidate causal event pairs are constructed. Local causal fragment sets are formed through event type matching and correlation calculation, and mapped to the disaster knowledge graph to generate a disaster causal chain set and a disaster situation subgraph. Based on the disaster causal chain set, the response paths similar to the current disaster structure are retrieved from historical response records and emergency plans, and a candidate emergency strategy set is formed by combining resource availability screening. A counterfactual scenario set is constructed based on the disaster situation subgraph. The system performs disaster evolution simulations on a set of counterfactual scenarios, calculates personnel risk indicators, equipment loss indicators, and production capacity impact indicators for each emergency strategy, forms a set of scheme evaluation results, and selects and recommends emergency schemes. Command instructions are issued according to the recommended emergency plan, data on the execution process is collected, execution result records are generated, and the results are written into the disaster event sequence.

[0007] As a preferred embodiment of the knowledge graph-based mine disaster emergency command method of the present invention, the specific steps for generating the disaster event sequence are as follows: Spatial location identification and coding, quality inspection and missing data completion are performed on multi-source mine monitoring data to generate complete monitoring data for multi-source mines; Disaster characteristics are extracted from complete monitoring data of multi-source mines and combined into disaster event units; Successive disaster event units are grouped into a disaster event sequence according to their chronological order.

[0008] As a preferred embodiment of the knowledge graph-based mine disaster emergency command method of the present invention, the step of constructing candidate causal event pairs based on disaster event sequences and forming a set of local causal fragments through event type matching and relevance calculation is as follows: Based on the disaster event sequence, disaster event units are labeled with event type to generate disaster event sequences with event type identifiers; Candidate causal event pairs are constructed based on disaster event sequences with event type identifiers, and a set of candidate causal event pairs is generated. For each candidate causal event pair, a correlation calculation is performed based on the set of candidate causal event pairs to form a set of local causal fragments.

[0009] As a preferred embodiment of the knowledge graph-based mine disaster emergency command method of the present invention, the specific steps for generating the disaster causal chain set and the disaster situation subgraph are as follows: Map local causal fragment sets to a disaster knowledge graph to generate an updated disaster knowledge graph; Extract the disaster causal chain set from the updated disaster knowledge graph and generate the corresponding disaster situation subgraph for the region.

[0010] As a preferred embodiment of the knowledge graph-based emergency command method for mine disasters described in this invention, the specific steps for retrieving a handling path similar to the current disaster structure from historical handling records and emergency plans based on the disaster causal chain set are as follows: Extract current disaster structural feature descriptions from disaster causal chain sets and disaster situation subgraphs; Based on historical response records and emergency plans, a historical response path database will be constructed. Based on the current disaster structure characteristics description, similar structural response paths are retrieved from the historical response path database to form a set of similar structural response paths.

[0011] As a preferred embodiment of the knowledge graph-based emergency command method for mine disasters described in this invention, the specific steps for constructing the counterfactual scenario set are as follows: By filtering the set of structurally similar handling paths based on resource availability, a set of candidate emergency strategies is generated; Based on the set of candidate emergency strategies, a set of counterfactual scenarios is constructed using the disaster situation subgraph as a basis.

[0012] As a preferred embodiment of the knowledge graph-based mine disaster emergency command method of the present invention, the specific steps for screening and recommending emergency plans are as follows: Based on each counterfactual scenario subgraph in the counterfactual scenario set, construct the initial state of disaster evolution and the emergency intervention sequence; The initial state of disaster evolution and the emergency intervention sequence are used to perform time-step disaster evolution simulation on the counterfactual scenario subgraph to generate the disaster evolution trajectory; Based on the disaster evolution trajectory, statistical indicators of personnel risk, equipment loss, and production capacity impact are used to generate an emergency strategy evaluation indicator set; Based on the emergency strategy evaluation index set and combined with multi-objective evaluation rules, a set of scheme evaluation results is obtained, and emergency schemes are selected and recommended.

[0013] As a preferred embodiment of the knowledge graph-based mine disaster emergency command method of the present invention, the specific steps for generating execution result records are as follows: Convert the sequence of events in the recommended emergency response plan into command instructions one by one; Execute command instructions and collect data on gas concentration, surrounding rock displacement, roadway deformation, ventilation parameters, equipment operating conditions, and personnel location to generate execution process data; Calculate the deviation rate between the execution process data and the expected target, and generate execution result records.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the knowledge graph-based mine disaster emergency command method described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the knowledge graph-based mine disaster emergency command method described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by dynamically mapping disaster event sequences to a knowledge graph, constructing causal chains and situational subgraphs, interpretable modeling of disaster evolution is achieved, and multi-dimensional evolutionary deduction is performed based on counterfactual scenarios to quantitatively evaluate the comprehensive impact of various emergency strategies, thereby selecting the best command plan that takes into account personnel safety, equipment protection and production capacity recovery, and improving the scientificity and adaptability of mine emergency response. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a knowledge graph-based emergency command method for mine disasters.

[0019] Figure 2 A flowchart generated for a disaster event sequence.

[0020] Figure 3 A flowchart for constructing and mapping a set of local causal fragments.

[0021] Figure 4 A flowchart for constructing counterfactual scenarios and evaluating contingency plans. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a mine disaster emergency command method based on knowledge graph, including the following steps: S1: Collect multi-source mine monitoring data, perform time synchronization and spatial coding, and generate disaster event sequences; S1.1: Multi-source mine monitoring data includes gas concentration monitoring, surrounding rock displacement monitoring, roadway deformation monitoring, ventilation parameter monitoring, equipment operating condition monitoring, and personnel location monitoring; Furthermore, monitoring commands are simultaneously initiated for the gas concentration monitoring device, surrounding rock displacement monitoring device, roadway deformation monitoring device, ventilation parameter monitoring device, equipment condition monitoring device, and personnel location monitoring device inside the mine at fixed sampling cycles. At the trigger time of the sampling cycle, the gas concentration monitoring data, surrounding rock displacement monitoring data, roadway deformation monitoring data, ventilation parameter monitoring data, equipment condition monitoring data, and personnel location monitoring data are read, and all monitoring data are uniformly attached to the same time coordinate.

[0026] S1.2: Spatial location identification coding, quality inspection and missing data completion are performed on multi-source mine monitoring data to generate complete monitoring data for multi-source mines; Furthermore, based on the mine's underground zoning, roadway numbering, equipment location numbering, and personnel location numbering, a unique spatial coordinate identifier is attached to each gas concentration monitoring, surrounding rock displacement monitoring, roadway deformation monitoring, ventilation parameter monitoring, equipment operating condition monitoring, and personnel location monitoring. This spatial coordinate identifier includes the monitoring point number, roadway number, and depth coordinates. For each monitoring quantity, amplitude validity, rate of change rationality, and monitoring continuity are verified. Monitoring quantities with amplitudes exceeding the monitoring device's range are discarded. For monitoring quantities with missing continuity, linear interpolation is used to complete the data from two consecutive sampling periods at the same monitoring point, ensuring that each time and spatial coordinate identifier has a complete monitoring quantity, outputting a complete multi-source mine monitoring quantity.

[0027] S1.3: Extract disaster characteristics from complete monitoring data of multi-source mines and combine them into disaster event units; Furthermore, the average value, rate of change, and maximum deviation of the gas concentration monitoring, surrounding rock displacement monitoring, roadway deformation monitoring, ventilation parameter monitoring, equipment operating condition monitoring, and personnel location monitoring data in the complete monitoring data of multi-source mines are calculated within a fixed-length statistical window. The average value of each monitoring data is used as a stable reference value, the rate of change of each monitoring data is used as a trend characteristic value, and the maximum deviation of each monitoring data is used as an anomaly degree characteristic value. The stable reference value, trend characteristic value, and anomaly degree characteristic value are combined into a disaster feature vector under the same time coordinate and the same spatial coordinate identifier. The disaster feature vector and the corresponding time coordinate and spatial coordinate identifier are packaged together to generate a single-structure disaster event unit.

[0028] S1.4: Combine consecutive disaster event units into a disaster event sequence according to chronological order; Furthermore, based on the time coordinates contained in each disaster event unit, all disaster event units are sorted in order of time coordinates from earliest to latest, and the sorted disaster event units are sequentially connected to form a disaster event sequence with a continuous time structure, wherein each disaster event unit retains its original time coordinates, spatial coordinate identifiers and disaster feature vectors in the disaster event sequence.

[0029] S2: Based on disaster event sequences, candidate causal event pairs are constructed. Local causal fragment sets are formed through event type matching and correlation calculation, and mapped to the disaster knowledge graph to generate a disaster causal chain set and a disaster situation subgraph. S2.1: Based on the disaster event sequence, the disaster event units are labeled with event types to generate a disaster event sequence with event type identifiers; Furthermore, based on the stable reference value, trend characteristic value, and anomaly degree characteristic value contained in the disaster event unit, the stable reference value, trend characteristic value, and anomaly degree characteristic value are compared with the event type discrimination rules item by item. When the combination of stable reference value, trend characteristic value, and anomaly degree characteristic value meets the conditions defined in the event type discrimination rules, the disaster event unit is labeled as the corresponding event type. All disaster event units that have completed event type labeling are arranged in the original time coordinate order to obtain a disaster event sequence with event type identification.

[0030] It should be noted that the event type discrimination rule is set by identifying the combination pattern of stable reference value, trend characteristic value and abnormality degree characteristic value within the same statistical window. The setting steps are as follows: determine the variation range of each monitoring quantity under normal conditions based on the statistical distribution of complete monitoring quantities of multi-source mines in the stable operating stage, and then extract the characteristic change trend based on the monitoring quantity change pattern before the occurrence of various disaster events in historical disaster records. This forms a multi-condition combination composed of three types of features: "offset direction of stable reference value", "change direction of trend characteristic value" and "deviation degree of abnormality degree characteristic value". Disaster event units that meet a certain combination pattern are labeled as the corresponding event type.

[0031] Example: When the trend characteristic value shows a continuous upward pattern and the deviation direction of the anomaly degree characteristic value is consistent with the change pattern before the historical gas anomaly event, the disaster event unit can be marked as a gas anomaly event.

[0032] S2.2: Construct candidate causal event pairs based on the disaster event sequence with event type identifiers, and generate a set of candidate causal event pairs; Furthermore, the disaster event sequence with event type identifiers is traversed sequentially according to the time coordinates of each disaster event unit. A finite time range is set after the time coordinates of each disaster event unit, and within the finite time range, disaster event units with the same or adjacent roadway numbers and event type identifiers are searched from the disaster event sequence. When the disaster event units satisfy the temporal sequence relationship and spatial proximity relationship, the earlier disaster event unit is taken as the causal event, and the later disaster event unit is taken as the result event. The causal event and the result event are combined into candidate causal event pairs, and the causal event identifier and the result event identifier are recorded. All candidate causal event pairs are combined into a candidate causal event pair set.

[0033] It should be noted that the finite time range is set by statistically analyzing the time delay characteristics of changes in monitored quantities during disaster propagation. The setting steps are to extract continuous disaster event units from historical disaster records, calculate the time interval distribution between adjacent events, and then select the time interval range that can cover the main propagation path from the time interval distribution as the finite time range, so that the causal event and the result event maintain the actual disaster propagation possibility in terms of time sequence.

[0034] Example: When the time interval between consecutive events exhibits a "short-term concentrated" pattern in the historical geopressure event chain, this time interval can be used as a finite time range to search for disaster event units that may become the resulting events from the disaster event sequence.

[0035] S2.3: Based on the set of candidate causal event pairs, perform correlation calculation for each candidate causal event pair to form a set of local causal fragments; Furthermore, for each candidate causal event pair, the disaster feature vector corresponding to the causal event and the disaster feature vector corresponding to the result event are compared one by one using the correlation calculation method. The comparison results are summarized into a correlation value according to the correlation calculation method. At the same time, the time interval between the causal event and the result event is calculated based on their time coordinates. Then, the correlation value and the time interval are jointly judged according to the correlation establishment condition and the propagation time condition. When the correlation value meets the correlation establishment condition and the time interval meets the propagation time condition, the candidate causal event pair is recorded as a local causal segment. The causal event identifier, result event identifier, correlation value and time interval are recorded for the local causal segment. All candidate causal event pairs that meet the conditions form a set of local causal segments.

[0036] A correlation coefficient calculation method is used to compare each feature-corresponding item one by one. The comparison results are then summarized into a correlation coefficient value, expressed as follows: ; ; ; ; in, The index indicates the monitoring quantity type. The correlation value of a single monitoring quantity The index indicates the monitoring quantity type. The stable reference value normalized difference index The index indicates the type of monitoring quantity. The trend characteristic value normalized difference index, The index indicates the monitoring quantity type. The normalized difference index of the abnormality characteristic value. This indicates that the disaster event unit corresponding to the outcome event in the candidate causal event pair is in the monitoring quantity type index. Stable reference value at that time This indicates that the disaster event unit corresponding to the causal event in the candidate causal event pair is in the monitoring quantity type index. Stable reference value at that time The index indicates the type of monitoring quantity. The normalization coefficient of the stable reference value, This indicates that the disaster event unit corresponding to the outcome event in the candidate causal event pair is in the monitoring quantity type index. Trend characteristic value of time, This indicates that the disaster event unit corresponding to the causal event in the candidate causal event pair is in the monitoring quantity type index. Trend characteristic value of time, The index indicates the monitoring quantity type. The trend eigenvalue normalization coefficient, This indicates that the disaster event unit corresponding to the outcome event in the candidate causal event pair is in the monitoring quantity type index. The abnormality characteristic value at that time This indicates that the disaster event unit corresponding to the causal event in the candidate causal event pair is in the monitoring quantity type index. The abnormality characteristic value at that time The index indicates the monitoring quantity type. The normalization coefficient of the abnormality characteristic value. The index indicates the type of monitoring quantity. The stable reference value weighting coefficient, The index indicates the monitoring quantity type. The trend characteristic value weighting coefficient, The index indicates the monitoring quantity type. The weighted coefficients of the abnormality characteristic values, The index indicates the monitoring quantity type. The correlation normalization coefficient, Indicates the index of monitoring quantity type; It should be noted that the correlation condition is set by conducting a joint distribution analysis of the normalized difference indices of various monitoring quantities. The setting steps are as follows: based on the difference in disaster characteristic vectors between the actual causal events and the resulting events in historical disaster propagation records, the combination intervals of the normalized stable reference value difference index, trend characteristic value difference index, and anomaly degree characteristic value difference index in the actual causal relationship are statistically analyzed. The difference combination interval that can reflect the actual causal propagation trend is defined as the correlation condition, so that the correlation value can accurately reflect the synchronicity of characteristic changes during the disaster propagation process.

[0037] Example: When the trend characteristic values ​​of multiple monitoring quantities in a real disaster propagation chain change in the same direction, the "same direction of difference" can be used as part of the condition for the correlation to be established, so that the correlation determination can screen out candidate causal event pairs with consistent change patterns.

[0038] The propagation time condition is set by statistically analyzing the time delay distribution between causal events and result events in the disaster propagation path. The setting steps are to extract consecutive event pairs from the historical disaster propagation chain, record the propagation time interval between events, and select a range that can cover the actual propagation delay from the propagation time interval distribution as an acceptable propagation time condition, so that the judgment result can reflect the real time correlation between events in the disaster propagation process.

[0039] Example: When the propagation time intervals between adjacent events in a historical ventilation anomaly event chain are concentrated, this distribution range can be used as a propagation time condition to ensure that the temporal correlation of candidate causal event pairs is consistent with the actual propagation characteristics of ventilation anomalies.

[0040] It should also be noted that the correlation calculation method is a mathematical method that measures the degree of correlation between causal events and result events based on the changing patterns of stable reference values, trend characteristic values, and anomaly degree characteristic values ​​in disaster feature vectors. By calculating the difference between the three feature items of the disaster feature vectors corresponding to the causal event and the disaster feature vectors corresponding to the result event under each type of monitoring quantity, performing normalization processing, and summarizing according to the weighted combination rule, a correlation value representing the degree of synchronization of changes in the multidimensional monitoring quantity characteristics of the two disaster event units is generated. The higher the correlation value meets the conditions for the establishment of correlation, the more consistent the causal event and the result event are in the changing trends of multi-source monitoring quantities, thus having a higher probability of causal association. This method can integrate the characteristic changes of different types of monitoring quantities while maintaining a unified mathematical structure, making the screening process of local causal fragments computable and stable.

[0041] S2.4: Map the set of local causal fragments to the disaster knowledge graph to generate an updated disaster knowledge graph; Furthermore, for each causal event identifier and result event identifier in the local causal fragment set, an event node search is performed in the disaster knowledge graph according to the event type identifier and spatial coordinate identifier. When there is an event node in the disaster knowledge graph that matches the event type identifier and spatial coordinate identifier, a directed relation edge is added between the corresponding event nodes, and the relevance value and time interval are written into the directed relation edge. When there is no event node in the disaster knowledge graph that matches the event type identifier and spatial coordinate identifier, an event node is added, and a directed relation edge is established between the added event nodes, and the relevance value and time interval are recorded. The updated disaster knowledge graph is obtained by performing a mapping operation on all local causal fragments.

[0042] It should be noted that, before the disaster knowledge graph is updated, it needs to be built with a data structure that is deterministic, reproducible and entirely derived from the existing production data and historical records of the mine. Its construction process does not rely on dynamic inference or causal fragments generated in subsequent steps. Instead, it extracts "disaster-related entities, disaster-related attributes and disaster-related relationships" based on the existing information of the mine to form an initial graph. The establishment method is as follows: Accident reports, hidden danger investigation records, production scheduling records, equipment maintenance records, and roadway distribution structure data accumulated in the mine over a long period of time are classified and organized. The names of disaster types, monitoring quantities, equipment, roadway numbers, key location numbers, and response actions appearing in the data are extracted as entity nodes. Then, based on the temporal descriptions, operational descriptions, and causal descriptions appearing in the data, the accident occurrence conditions, abnormal monitoring quantity patterns, equipment failure triggering conditions, and sequential relationships between different disasters are extracted as relation edges. After forming entity nodes and relation edges, regional subgraphs are created using roadway numbers and monitoring point numbers as spatial indexes. Entity nodes and relation edges in the same region are combined into a spatial local structure. Finally, all entity nodes and all relation edges constitute the initial structure of the disaster knowledge graph, which serves as the prerequisite for subsequent local causal fragment mapping and disaster causal chain generation.

[0043] S2.5: Extract the set of disaster causal chains from the updated disaster knowledge graph and generate a disaster situation subgraph for the corresponding region; Furthermore, event nodes in the updated disaster knowledge graph that do not appear as any result event nodes are taken as the starting event nodes of the disaster causal chain. Depth-first traversal is performed based on the time interval and relevance value in the directed relation edges, and the event path from the starting event node to the ending event node is recorded. Each event path is taken as a disaster causal chain, and all event paths form a disaster causal chain set. At the same time, a subgraph containing event nodes and directed relation edges within the space range corresponding to the current monitoring area and the spatial coordinates adjacent to the current monitoring area is extracted from the updated disaster knowledge graph as the disaster situation subgraph.

[0044] S3: Based on the disaster causal chain set, retrieve the disposal path similar to the current disaster structure from historical disposal records and emergency plans, and form a candidate emergency strategy set by combining resource availability screening. Construct a counterfactual scenario set based on the disaster situation subgraph. S3.1: Extract the current disaster structure feature description from the disaster causal chain set and disaster situation subgraph; Furthermore, from the disaster causal chain set, select several disaster causal chains that contain the spatial coordinate identifiers corresponding to the current monitoring area and have the highest overlap with the node set of the disaster situation subgraph. Record the event type sequence, spatial coordinate identifier sequence, and adjacent event time interval sequence for each selected disaster causal chain according to the event sequence. Combine the disaster feature vectors corresponding to each event node in the disaster situation subgraph to construct the current disaster structure feature description. The current disaster structure feature description includes the event type sequence, spatial coordinate identifier sequence, time interval sequence, and risk level feature quantity.

[0045] S3.2: Construct a historical response path database based on historical response records and emergency plans; Furthermore, by breaking down the event narrative structure of the accident process text, the operational sequence information of the on-site dispatch records, and the handling action rules of the emergency plan into uniformly formatted action units, the accident process text, on-site dispatch records, and handling process descriptions in the emergency plan from historical handling records are structurally extracted. The action units are then recombined according to time sequence, spatial coordinate identifiers, and triggering conditions, restoring each historical handling process into a handling event sequence. Each handling event in the handling event sequence includes the handling action type, the spatial coordinate identifier of the target, the triggering conditions, and the execution sequence number. The handling event sequence is aligned with the event type sequence, spatial coordinate identifier sequence, and time interval information in the corresponding disaster description to form a historical handling path feature description. All historical handling path feature descriptions are stored as a historical handling path library.

[0046] S3.3: Based on the current disaster structure characteristics description, retrieve structurally similar disposal paths from the historical disposal path database to form a set of structurally similar disposal paths; Furthermore, the event type sequence, spatial coordinate identifier sequence, and time interval sequence in the current disaster structural feature description are compared with the corresponding sequences in the feature description of each historical treatment path in the historical treatment path database. The structural similarity between the current disaster structural feature description and the historical treatment path feature description is calculated based on event type matching rules, spatial location matching rules, and time sequence matching rules. When the structural similarity meets the structural similarity judgment criteria, the corresponding historical treatment path is included in the set of structurally similar treatment paths. The structural similarity judgment criteria can be determined by statistically analyzing the structural similarity distribution of successful treatment samples in historical treatment paths, selecting a lower limit that covers most successful treatment samples and excludes obviously failed treatment samples. The adjacent spatial coordinate identifier range in the spatial location matching rules can be combined with the influence distances of roadway division, gas diffusion, and surrounding rock damage. Several adjacent spatial units not exceeding a given physical distance are considered to be in the same adjacent range; as long as the spatial coordinate identifier falls within this range, it is determined that the spatial location is similar.

[0047] Based on event type matching rules, spatial location matching rules, and temporal sequence matching rules, the structural similarity between the current disaster structural feature description and the historical response path feature description is calculated. The expression is as follows: ; in, Indicates the first Structural similarity index between the characteristics of historical disaster response paths and the characteristics of current disaster structures. This represents the result calculated based on the event type matching rules. The matching metric of historical handling paths along the event type sequence dimension. This represents the first result calculated based on spatial location matching rules. The matching metric of historical disposal paths in the spatial coordinate identifier sequence dimension. This represents the result calculated based on the time sequence matching rules. The matching metric of historical processing paths in the dimension of time interval sequence. Indicates the first Structural similarity normalization coefficients describing the characteristics of historical disposal paths; It should be noted that the number calculated based on the event type matching rules is... Matching metrics of historical handling paths on the event type sequence dimension It is obtained by statistically analyzing the proportion of the same event type sequence in the current disaster structure feature description and the event type sequence in the historical response path feature description, and by comprehensively considering the positional consistency relationship; the first is calculated based on the spatial position matching rule. Matching measure of historical treatment paths in spatial coordinate identifier sequence dimension It is obtained by statistically analyzing the proportion of overlap between the spatial coordinate identifier sequence in the current disaster structure feature description and the spatial coordinate identifier sequence in the historical disposal path feature description within the same roadway number and adjacent spatial coordinate identifier range, and by comprehensively considering spatial proximity relationships; the first is calculated based on the time sequence matching rule. Matching metrics of historical processing paths in the time interval sequence dimension This is obtained by comparing the proportion of time interval sequences in the current disaster structure feature description and the historical response path feature description that are consistent in both the chronological order of events and the trend of time interval changes; the first Structural similarity normalization coefficients of historical disposal path feature descriptions It is used for , and The superposition results are standardized in terms of dimensions and compressed in terms of scope, so that the descriptions of historical handling paths of different lengths and number of events are comparable in the evaluation of structural similarity.

[0048] The event type matching rule compares the event type sequence in the current disaster structure feature description with the event type sequence in the historical response path feature description sequentially from a fixed starting point. It counts the number of locations with the same event type and belonging to the key disaster event type, and uses the ratio of this number to the total number of compared locations as the event type matching metric. The spatial location matching rule aligns two spatial coordinate identifier sequences sequentially according to the event sequence, determining whether the roadway numbers are consistent and whether the spatial coordinate identifiers are within a pre-defined adjacent spatial range (based on the roadway geometry, gas diffusion distance, surrounding rock failure impact zone, and the actual distribution of spatially related events in historical disasters, determining the physical distance between the target spatial coordinate identifier and the target spatial coordinate identifier). Several adjacent spatial units not exceeding a given safety impact radius are uniformly classified into the same proximity range. The number of locations meeting the conditions is counted, and the ratio of this number to the total number of aligned locations is used as a spatial location matching metric. The temporal sequence matching rule compares two time interval sequences in the event sequence to determine whether the order of events is consistent and whether the time interval change trends are in the same direction. The number of event pairs meeting the conditions is counted, and the ratio of this number to the total number of event pairs participating in the comparison is used as a temporal sequence matching metric. These three matching metrics, as components of structural similarity evaluation, are used to measure the degree of consistency between the response path and the current disaster structure in three dimensions: event type, spatial distribution, and temporal evolution.

[0049] S3.4: Filter the set of structurally similar disposal paths based on resource availability to generate a set of candidate emergency strategies; Furthermore, based on the equipment operating condition monitoring data, ventilation parameter monitoring data, and personnel location monitoring data in the complete monitoring data of multi-source mines, a current resource availability description is constructed. The current resource availability description includes a list of available equipment, the range of available ventilation capacity, and the number of personnel that can be deployed. The current resource availability description is then compared item by item with the equipment resources, ventilation conditions, and personnel configuration requirements required for each disposal path in the set of structurally similar disposal paths. When the resource conditions required for a disposal path are within the allowable range of the current resource availability description, the disposal event sequence corresponding to the disposal path is solidified into an emergency strategy. All emergency strategies that meet the resource availability conditions are combined into a candidate emergency strategy set.

[0050] Furthermore, the specific process for constructing a current resource availability description based on equipment condition monitoring, ventilation parameter monitoring, and personnel location monitoring from the comprehensive multi-source mine monitoring data is as follows: Within the time range corresponding to the current disaster structure characteristic description, equipment condition monitoring data is extracted from the comprehensive multi-source mine monitoring data according to equipment identification, and the current characteristics, vibration characteristics, and temperature characteristics are compared with the allowable operating range of the equipment item by item. Equipment that meets the allowable operating range is recorded as available equipment resources. Ventilation parameter monitoring data is extracted according to ventilation facility identification, and the wind speed monitoring data and wind pressure monitoring data are compared with the rated capacity range of the ventilation facilities item by item. Ventilation facilities that can meet emergency ventilation needs are recorded as available ventilation resources. Personnel location monitoring data is extracted according to spatial coordinate identification, and the number of on-duty personnel and job types that can be deployed are statistically analyzed in conjunction with job configuration data. Personnel who can participate in emergency response are recorded as available personnel resources. Available equipment resources, available ventilation resources, and available personnel resources are combined to form the current resource availability description.

[0051] S3.5: Based on the set of candidate emergency strategies, construct a set of counterfactual scenarios using the disaster situation sub-graph as a foundation; Furthermore, for each emergency strategy in the candidate emergency strategy set, according to the type of handling action and the spatial coordinates of the target recorded in the sequence of handling events, the event node or adjacent event node under the corresponding spatial coordinates is found in the disaster situation sub-graph. The handling event is then attached as an intervention event to the corresponding event node or disaster causal chain path, forming a counterfactual scenario sub-graph containing the disaster event evolution path and the emergency handling intervention path. The counterfactual scenario sub-graph corresponding to each emergency strategy is then included in the counterfactual scenario set.

[0052] S4: Conduct disaster evolution simulations on the counterfactual scenario set, calculate personnel risk indicators, equipment loss indicators, and production capacity impact indicators for each emergency strategy, form a set of scheme evaluation results, and screen and recommend emergency schemes; S4.1: Construct the initial state of disaster evolution and emergency intervention sequence based on each counterfactual scenario subgraph in the counterfactual scenario set; Furthermore, the current disaster causal chain path and the event node set and corresponding disaster feature vector are read from the counterfactual scenario subgraph. The disaster feature vector, time coordinate and spatial coordinate are identified as the initial state of disaster evolution. The sequence of disposal events attached to the disaster causal chain path is identified as the emergency intervention sequence. The disposal events are numbered and arranged according to the execution order in the disposal event sequence to form the initial state of disaster evolution and emergency intervention sequence for a single simulation.

[0053] S4.2: Use the initial state of disaster evolution and emergency intervention sequence to perform time-step disaster evolution simulation on the counterfactual scenario subgraph to generate disaster evolution trajectory; Furthermore, the time axis is discretized according to the time intervals along the disaster causal chain path. At each time step, the disaster feature vector is updated based on the propagation direction and relevance value of the directed relation edges in the disaster knowledge graph. Simultaneously, based on the execution time of the response events in the emergency intervention sequence, the corresponding response actions are applied to the disaster feature vector under the spatial coordinates of the response target. The stability reference value, trend feature value, and anomaly degree feature value are adjusted. After completing the propagation and intervention processes, the disaster feature vector, time coordinates, and spatial coordinates of the current time step are recorded. The records from all time steps are combined sequentially to form the disaster evolution trajectory.

[0054] S4.3: Based on the disaster evolution trajectory, statistical indicators of personnel risk, equipment loss, and production capacity impact are generated to produce an emergency strategy evaluation indicator set; Furthermore, based on the personnel location monitoring characteristics and the risk level characteristics in the disaster feature vector at each time step in the disaster evolution trajectory, the number of personnel and their dwell time within the hazardous spatial coordinate identification area are statistically analyzed to construct a personnel risk index (based on spatial coordinate identification areas where the risk level characteristics have been consistently high in historical disaster records, and combined with hazardous work areas listed in the mine safety regulations (such as high-concentration gas accumulation roadways, concentrated areas of roof fracture zones, or local dead-end roadways with insufficient ventilation, etc.), the spatial coordinate identification areas are clustered to obtain hazardous spatial coordinate identification areas); based on the equipment condition monitoring characteristics and the abnormality level characteristics in the disaster feature vector in the disaster evolution trajectory, the number of equipment in faulty or over-limit states and their corresponding durations are statistically analyzed to construct an equipment loss index; based on the equipment condition monitoring characteristics related to production capacity and the operating status characteristics in the disaster feature vector in the disaster evolution trajectory, the number of restricted production links and their corresponding time ranges are statistically analyzed to construct a production capacity impact index. The three types of indicators are categorized according to emergency strategy identification to form personnel risk indicators, equipment loss indicators, and production capacity impact indicators corresponding to the emergency strategies.

[0055] It should be noted that the rated power, rated speed, rated load, and other operating parameters of key equipment related to production capacity are read, and indicators such as power, current, speed, and start-up / shutdown status within the current sampling period are extracted from the equipment operating condition monitoring data. The current indicators are compared with the corresponding rated parameters to calculate the operating condition utilization rate. Based on the degree of influence of each piece of equipment in the process flow on the overall production capacity, the production capacity contribution coefficient is given. The product of the operating condition utilization rate and the production capacity contribution coefficient of each piece of equipment is summed and normalized to obtain the operating status characteristic quantity under the current time coordinate and spatial coordinate identification. This allows the operating status characteristic quantity to reflect whether the production capacity is restricted and the degree of restriction.

[0056] S4.4: Based on the emergency strategy evaluation index set and combined with multi-objective evaluation rules, obtain the scheme evaluation result set and screen and recommend emergency schemes; Furthermore, based on multi-objective evaluation rules, personnel risk indicators, equipment loss indicators, and production capacity impact indicators are converted into evaluation quantities with unified dimensions. A comprehensive evaluation result is calculated for each emergency strategy. The comprehensive evaluation result, along with the emergency strategy number and the sequence of events to be handled, is recorded as the scheme evaluation result. The scheme evaluation results corresponding to all emergency strategies are combined into a scheme evaluation result set. The scheme evaluation result set is sorted according to the order of the comprehensive evaluation results. At least one emergency strategy is selected from the top-ranked emergency strategies as a recommended emergency plan. The recommended emergency plan is simultaneously output to the emergency command link and returned to the behavior record used to update the subsequent counterfactual scenario construction and disaster knowledge graph.

[0057] It should be noted that the multi-objective evaluation rule is set by uniformly normalizing and weighting personnel risk indicators, equipment loss indicators, and capacity impact indicators within the same batch of emergency strategies. Specifically, in a single scheme evaluation, the maximum and minimum values ​​of personnel risk indicators, equipment loss indicators, and capacity impact indicators corresponding to all emergency strategies are first statistically analyzed within that batch. A linear normalization mapping function is then constructed for each type of indicator, mapping the minimum and maximum values ​​of that type of indicator to a dimensionless evaluation quantity using linear interpolation. This ensures that the smaller the original value of an emergency strategy, the smaller the corresponding dimensionless evaluation quantity after normalization. Based on the principle of prioritizing personnel safety, followed by equipment integrity, and then capacity recovery, fixed priority weights are assigned to the personnel risk evaluation quantity, equipment loss evaluation quantity, and capacity impact evaluation quantity. The three types of dimensionless evaluation quantities are weighted with their corresponding priority weights to obtain the comprehensive evaluation quantity for each emergency strategy. In subsequent steps, all emergency strategies are ranked according to the magnitude of the comprehensive evaluation quantity, and recommended emergency schemes are selected accordingly. The calculation process and ranking results of the comprehensive evaluation quantity together constitute the pre-set multi-objective evaluation rule.

[0058] S5: Issue command instructions based on the recommended emergency plan, collect data on the execution process, generate execution result records, and write them into the disaster event sequence; S5.1: Convert the sequence of events in the recommended emergency response plan into command instructions one by one; Furthermore, the sequence of events to be handled is extracted from the recommended emergency response plan. Each event is then structurally transformed according to its execution order to generate corresponding command instructions. These instructions include time coordinates, spatial coordinate identifiers, action type, resource requirements, and expected target values. The time coordinates are obtained by adding a fixed sampling period to the time coordinates of the latest event in the disaster event sequence; the spatial coordinate identifiers are directly taken from the node spatial coordinate identifiers in the disaster situation sub-graph associated with the event; the action type is consistent with the semantic content of the event; the resource requirements are matched with the equipment, personnel, and material information needed for the action based on the current resource availability description; and the expected target values ​​are extracted from the event, containing quantitative indicators corresponding to the action type, including the target parameter name, target value, and time requirement for achievement.

[0059] S5.2: Execute command instructions and collect data on gas concentration, surrounding rock displacement, roadway deformation, ventilation parameters, equipment operating conditions, and personnel location as execution process data; Furthermore, after receiving the command instructions, the execution unit initiates the corresponding handling operation based on the spatial coordinates and handling action type specified in the instructions, and simultaneously triggers the data acquisition process of the corresponding monitoring devices. Gas concentration monitoring data is collected by gas sensors deployed at designated spatial coordinates at fixed sampling intervals; surrounding rock displacement monitoring data is collected by surrounding rock displacement sensors; roadway deformation monitoring data is collected by roadway deformation monitoring devices; ventilation parameter monitoring data is collected by wind speed and wind pressure sensors in the ventilation system; equipment operating condition monitoring data is collected by equipment operating status monitoring devices; and personnel location monitoring data is collected by personnel positioning terminals. All monitoring data are accompanied by spatial coordinates consistent with the command instructions and time coordinates aligned with the fixed sampling intervals, and the execution process data is output.

[0060] S5.3: Calculate the deviation rate between the execution process data and the expected target, and generate execution result records; Furthermore, monitoring quantities with the same names as the expected target values ​​in the execution process data are extracted as actual execution values ​​and compared with the expected target values ​​to obtain the deviation rate. If the expected target value is a time-constrained target, the deviation rate is calculated as the ratio of the actual completion time to the expected achievement time. The risk level classification rules can be determined by statistically analyzing the deviation rate distribution of samples with labeled disposal effects in historical or simulated disposal data. For example, in a certain type of gas anomaly disposal scenario, disposal actions with a deviation rate less than 0.1 can be selected as low risk, 0.1 to 0.3 as medium risk, and greater than 0.3 as high risk, thus obtaining the risk level corresponding to different deviation rate ranges.

[0061] S5.4: Record the execution results in the disaster event sequence according to the time coordinate from earliest to latest to generate a historical event sequence; Furthermore, the execution result records are sorted according to their time coordinates and appended to the end of the disaster event sequence in ascending order of time coordinates. During the appending process, all fields in the execution result records are retained, including event sequence ID, time coordinate, spatial coordinate identifier, action type, actual execution value, expected target value, deviation rate, and risk level. After the appending operation is completed, the disaster event sequence contains the original event records and the newly added execution result records, forming a continuous, complete, and unmodifiable historical event sequence. This disaster event sequence, containing the original disaster event units and the newly added execution result records, is called the historical event sequence and is used for long-term storage and statistical analysis. In subsequent calculations based on the disaster event sequence, only records containing disaster feature vector fields are selected as disaster event units to participate in the construction of candidate causal event pairs and the formation of local causal fragments. The execution result records are only used as supplementary information for statistically correcting rule parameters and evaluation thresholds, without changing the structure of the disaster event units.

[0062] This embodiment also provides a computer device applicable to the knowledge graph-based mine disaster emergency command method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the knowledge graph-based mine disaster emergency command method proposed in the above embodiment.

[0063] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0064] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the knowledge graph-based mine disaster emergency command method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0065] In summary, this invention achieves interpretable modeling of disaster evolution by dynamically mapping disaster event sequences to knowledge graphs, constructing causal chains and situational subgraphs, and conducting multi-dimensional evolutionary deductions based on counterfactual scenarios to quantitatively assess the comprehensive impact of various emergency strategies. This allows for the selection of command schemes that balance personnel safety, equipment protection, and production capacity recovery, thereby improving the scientific nature and adaptability of mine emergency response.

[0066] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A mine disaster emergency command method based on knowledge graph, characterized in that: include, Collect multi-source mine monitoring data, perform time synchronization and spatial coding, and generate disaster event sequences; Based on disaster event sequences, candidate causal event pairs are constructed. A set of local causal fragments is formed through event type matching and relevance calculation, and then mapped to a disaster knowledge graph to generate a set of disaster causal chains and a disaster situation subgraph. The specific steps are as follows. Based on the disaster event sequence, disaster event units are labeled with event type to generate disaster event sequences with event type identifiers; Candidate causal event pairs are constructed based on disaster event sequences with event type identifiers, and a set of candidate causal event pairs is generated. Furthermore, disaster event units are searched from the disaster event sequence, with earlier disaster event units being regarded as causal events and later disaster event units as resultant events, and a candidate causal event pair set is formed. Based on the set of candidate causal event pairs, the relevance of each candidate causal event pair is calculated to form a set of local causal fragments; Furthermore, for each candidate causal event pair, the disaster feature vector corresponding to the causal event and the disaster feature vector corresponding to the result event are compared one by one using the correlation calculation method, and the results are summarized into a correlation value. At the same time, a joint judgment is made based on the time interval. When the correlation value meets the correlation condition and the time interval meets the propagation time condition, the candidate causal event pair is recorded as a local causal segment and a set of local causal segments is formed. Map local causal fragment sets to a disaster knowledge graph to generate an updated disaster knowledge graph; Extract disaster causal chain sets from the updated disaster knowledge graph and generate disaster situation subgraphs for the corresponding regions; Furthermore, the event nodes in the updated disaster knowledge graph that do not appear as any result event nodes are taken as the starting event nodes of the disaster causal chain. The time interval and relevance value in the directed relation edges are used as the traversal basis to perform a depth-first traversal and record the event path from the starting event node to the ending event node. All event paths are combined into a disaster causal chain set. The event nodes and directed relation edges are extracted from the updated disaster knowledge graph as a subgraph of the disaster situation. Based on the disaster causal chain set, the response paths similar to the current disaster structure are retrieved from historical response records and emergency plans, and a candidate emergency strategy set is formed by combining resource availability screening. A counterfactual scenario set is constructed based on the disaster situation subgraph. The system performs disaster evolution simulations on a set of counterfactual scenarios, calculates personnel risk indicators, equipment loss indicators, and production capacity impact indicators for each emergency strategy, forms a set of scheme evaluation results, and selects and recommends emergency schemes. Command instructions are issued according to the recommended emergency plan, data on the execution process is collected, execution result records are generated, and the results are written into the disaster event sequence.

2. The mine disaster emergency command method based on knowledge graph as described in claim 1, characterized in that: The specific steps for generating the disaster event sequence are as follows: Spatial location identification and coding, quality inspection and missing data completion are performed on multi-source mine monitoring data to generate complete monitoring data for multi-source mines; Disaster characteristics are extracted from complete monitoring data of multi-source mines and combined into disaster event units; Successive disaster event units are grouped into a disaster event sequence according to their chronological order.

3. The mine disaster emergency command method based on knowledge graph as described in claim 1, characterized in that: The specific steps for retrieving similar response paths to the current disaster structure from historical response records and emergency plans based on the disaster causal chain set are as follows: Extract current disaster structural feature descriptions from disaster causal chain sets and disaster situation subgraphs; Based on historical response records and emergency plans, a historical response path database will be constructed. Based on the current disaster structure characteristics description, similar structural response paths are retrieved from the historical response path database to form a set of similar structural response paths.

4. The mine disaster emergency command method based on knowledge graph as described in claim 3, characterized in that: The specific steps for constructing the counterfactual scenario set are as follows: By filtering the set of structurally similar handling paths based on resource availability, a set of candidate emergency strategies is generated; Based on the set of candidate emergency strategies, a set of counterfactual scenarios is constructed using the disaster situation subgraph as a basis.

5. The mine disaster emergency command method based on knowledge graph as described in claim 4, characterized in that: The specific steps for selecting and recommending emergency plans are as follows. Based on each counterfactual scenario subgraph in the counterfactual scenario set, construct the initial state of disaster evolution and the emergency intervention sequence; The initial state of disaster evolution and the emergency intervention sequence are used to perform time-step disaster evolution simulation on the counterfactual scenario subgraph to generate the disaster evolution trajectory; Based on the disaster evolution trajectory, statistical indicators of personnel risk, equipment loss, and production capacity impact are used to generate an emergency strategy evaluation indicator set; Based on the emergency strategy evaluation index set and combined with multi-objective evaluation rules, a set of scheme evaluation results is obtained, and emergency schemes are selected and recommended.

6. The mine disaster emergency command method based on knowledge graph as described in claim 5, characterized in that: The specific steps for generating the execution result record are as follows: Convert the sequence of events in the recommended emergency response plan into command instructions one by one; Execute command instructions and collect data on gas concentration, surrounding rock displacement, roadway deformation, ventilation parameters, equipment operating conditions, and personnel location to generate execution process data; Calculate the deviation rate between the execution process data and the expected target, and generate execution result records.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the knowledge graph-based mine disaster emergency command method as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the knowledge graph-based mine disaster emergency command method as described in any one of claims 1 to 6.

Images