A knowledge graph-based mine sealed space multi-parameter big data monitoring system

Abstract

The application relates to the technical field of knowledge graphs, and particularly provides a mine-used closed space multi-parameter big data monitoring system based on a knowledge graph, which comprises a space entity correlation embedding subsystem used for receiving collected real-time monitoring data streams, forming a unified data frame, binding each monitoring data point with a corresponding mine-used closed space entity, and dynamically updating a mine-used closed space knowledge graph; a correlation abnormal mode extraction subsystem performs clustering analysis on abnormal data nodes in an abnormal propagation subgraph, according to the space-time distance and correlation strength between the abnormal data nodes, merges nodes with similar abnormal characteristics and existing propagation relations into the same abnormal mode; and a risk situation deduction subsystem deduces adjacent space entities and parameter change ranges possibly affected in a future time period by simulating the diffusion process of abnormal parameters in a risk transmission network. The application enhances the systematicness, foresight and intelligent level of mine safety monitoring.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and in particular to a multi-parameter big data monitoring system for confined spaces in mines based on knowledge graphs. Background Technology

[0002] As mining operations in my country extend to deeper and more geologically complex areas, safety monitoring in confined spaces such as goafs, ventilation roadways, and gas extraction areas faces increasingly severe challenges. Traditional monitoring methods typically rely on single or limited sensors, primarily collecting isolated parameters, lacking the ability to fuse and analyze multi-source heterogeneous data and dynamically correlate them. In practical applications, the mining environment is highly dynamic and involves multiple coupled factors; a single parameter cannot comprehensively reflect the true safety status of a confined space, easily leading to delayed warnings or false alarms, thus affecting the timeliness and accuracy of safety production decisions. Furthermore, existing systems mostly employ threshold alarm mechanisms, lacking in-depth analysis of historical and real-time data, failing to achieve early risk prediction and causal tracing, thus hindering the improvement of intelligent safety management in mines.

[0003] Existing technology 1, Chinese patent application number 202512003003.1, discloses a digital gas parameter early warning system for mines based on gas analysis. This system includes data acquisition and feature extraction, spatial topology association, gas state intent analysis, risk inference based on knowledge graphs, and generation of graded early warnings to achieve intelligent monitoring of mine gas safety. It dynamically associates gas behavior patterns with the spatial structure of roadways to form a networked situational awareness, and uses a disaster knowledge graph to perform deep semantic association and probabilistic reasoning on multi-source precursor information, thereby generating comprehensive risk conclusions with causal logic. While this system achieves a leap from discrete point alarms to spatial situational assessment, and from simple threshold judgments to intelligent causal inference, improving the accuracy of early warnings and decision support capabilities, it suffers from limitations. The monitoring parameters are singular in dimension, making it difficult to comprehensively reflect the comprehensive risk status of confined spaces. It primarily focuses on the analysis and early warning of gas parameters, failing to systematically integrate and associate various heterogeneous parameters such as environmental temperature and humidity, air pressure fluctuations, and equipment status, resulting in limited perception capabilities for complex risks such as fires, asphyxiation, and explosions. Anomaly detection lacks spatiotemporal correlation and propagation path analysis, making it difficult to pinpoint the root cause and spread trend of risks. Although it achieves correlation between gas behavior patterns and spatial structures, its anomaly identification is mainly based on point threshold judgment or local pattern analysis, failing to fully construct and utilize the topological and dependency network between monitoring points, between monitoring points and spatial entities, and between entities. Therefore, it is difficult to track the dynamic propagation process of anomalous parameters along paths such as ventilation networks and equipment connections, and cannot effectively reveal the correlation characteristics and diffusion subgraphs of anomaly clusters, limiting the accuracy of risk tracing and situation prediction. Risk extrapolation relies on predefined disaster knowledge graphs and probabilistic reasoning, resulting in insufficient adaptability to real-time dynamic evolution characteristics. Its risk extrapolation is based on static or semi-static disaster knowledge graphs for semantic correlation and probabilistic calculations, failing to deeply integrate the dynamic characteristics such as the anomaly evolution rate, direction, and path expansion trend reflected in real-time monitoring data streams. This leads to limited real-time simulation and forward-looking early warning capabilities of the extrapolation model for sudden and cascading risks.

[0004] Prior art two, Chinese patent application number 202210191298.3, discloses a multi-parameter monitoring system for confined spaces in mines. The system includes a confined space robot brain, a multi-parameter alarm for confined spaces, and network transmission equipment. The confined space robot brain is connected to the multi-parameter alarm for confined spaces via the network transmission equipment. The multi-parameter alarm for confined spaces is installed in the confined area to be monitored underground in a coal mine, realizing automatic and continuous monitoring of environmental parameters in the confined space underground. This application can relieve personnel of the heavy task of on-site inspection, not only eliminating errors and human negligence in manual inspection, but also greatly improving monitoring efficiency. This application can realize real-time and continuous monitoring of environmental parameters in confined spaces, unaffected by the arrangement and frequency of inspection personnel, and the data has strong real-time performance. While automated and continuous intelligent monitoring avoids the adverse effects of human, machine, and environmental factors on closed-loop management caused by on-site measurements, data remains disconnected from physical entities, forming data silos and lacking spatial and structured integration. Although multi-parameter automatic continuous acquisition is achieved, monitoring data still exists as discrete points, without systematic binding and association with spatial entities such as roadways, working faces, ventilation facilities, and electromechanical equipment. The lack of a unified spatial coordinate system and entity attribution information prevents the construction of a digital model reflecting the actual physical structure and topological relationships underground, limiting the overall analysis and collaborative utilization of the data. The functionality is limited to real-time monitoring and threshold alarms, lacking intelligent analysis and risk prediction capabilities: it mainly completes data collection, transmission, and over-limit alarms, belonging to the perception layer function; it does not introduce intelligent analysis modules such as knowledge graphs, abnormal pattern mining, and risk inference, and cannot extract related abnormal patterns from massive monitoring data, nor can it perform quantitative simulation and graded early warning of future risk situations based on the current state and historical patterns. Overall, it is in the stage of monitoring and alarm rather than intelligent perception-analysis-early warning; the system closed loop is incomplete, lacking a complete technical chain from anomaly identification to decision support: it has not designed subsequent subsystems such as anomaly correlation analysis, risk propagation simulation, and early warning output, and has not formed a complete closed-loop system from data fusion, anomaly extraction, situation inference to early warning generation, thus failing to provide in-depth decision support information for safety management.

[0005] Current technologies 1 and 2 have shortcomings in terms of the systematic nature, forward-looking perspective, and intelligence level of mine safety monitoring, which need further improvement. Therefore, this invention provides a multi-parameter big data monitoring system for confined spaces in mines based on knowledge graphs. Summary of the Invention

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] One aspect of the present invention provides a knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines, comprising:

[0008] The spatial entity association embedding subsystem receives real-time monitoring data streams collected by various sensors deployed within the mine's enclosed space. These data streams include parameters such as gas component concentration, ambient temperature and humidity, and air pressure fluctuations. The subsystem aligns these data streams according to timestamps and spatial coordinates, forming a unified data frame. Based on a pre-constructed mine enclosed space entity model containing spatial entities such as roadways, working faces, ventilation facilities, and electromechanical equipment, along with their topological relationships, the subsystem binds each monitoring data point to its corresponding mine enclosed space entity through spatial coordinate matching within the data frame. The bound real-time monitoring data undergoes parameter type identification and numerical quantization, extracting feature vectors for each data point containing information such as parameter type, value, time, and the entity it belongs to. These feature vectors are used as new nodes and connected to existing mine enclosed space entity nodes via edges representing monitoring and influence relationships, dynamically updating the mine enclosed space knowledge graph and generating a new graph structure containing real-time monitoring data.

[0009] An anomaly pattern extraction subsystem receives dynamically updated knowledge graphs of mine confined spaces; it traverses all monitoring data points in the knowledge graph, extracts parameter values, and compares them with the normal fluctuation range of these parameters in historical periods or adjacent spatial regions, marking anomaly data nodes that exceed thresholds; starting from these anomaly data nodes, it performs a breadth-first search along the edges of spatial adjacency, ventilation impact, and equipment connection in the mine confined space knowledge graph to trace possible propagation paths of anomaly parameters, forming an anomaly propagation subgraph; it performs cluster analysis on the anomaly data nodes in the anomaly propagation subgraph, grouping nodes with similar anomaly characteristics and propagation relationships into the same anomaly pattern based on the spatiotemporal distance and correlation strength between them; each anomaly pattern contains a set of anomaly parameters, its spatial entity, and the correlation paths between them; and it outputs a set of anomaly patterns and their corresponding subgraph structures.

[0010] The risk situation simulation subsystem receives a set of abnormal patterns and their subgraph structures, and calls up snapshots of the mine confined space knowledge graph from multiple historical time points. It compares the current abnormal pattern subgraph with the normal evolution trajectory of the same spatial region in the historical mine confined space knowledge graph, calculating the rate and direction of change of the attribute values ​​of the current abnormal nodes, as well as the expansion trend of the abnormal propagation path. Based on the evolutionary characteristics of the rate and direction of change and the expansion trend, it constructs a risk transmission network using the dependencies of ventilation and electromechanical systems between entities in the mine confined space knowledge graph. By simulating the diffusion process of abnormal parameters in the risk transmission network, it predicts the range of changes in adjacent spatial entities and their parameters that may be affected in the future. Finally, it synthesizes the simulation results of each abnormal pattern, generates a risk level assessment and early warning area identifier, and outputs them to the monitoring terminal.

[0011] This invention uses knowledge graphs to materialize and dynamically fuse multi-source heterogeneous monitoring data, enabling the extraction and propagation tracking of abnormal patterns. Based on this, it completes the simulation and deduction of risk situations and graded early warning. It enhances the systematicness, foresight, and intelligence of mine safety monitoring, and provides integrated technical support for the early detection, accurate location, and proactive prevention and control of complex risk factors in confined spaces. Attached Figure Description

[0012] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0013] Figure 1 This is a block diagram of the knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines provided in Embodiment 1 of the present invention;

[0014] Figure 2 This is a schematic diagram of the knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines provided in Embodiment 1 of the present invention.

[0015] Figure 3 This is a block diagram of the spatial entity association embedding subsystem provided in Embodiment 2 of the present invention;

[0016] Figure 4 This is a block diagram of the associated anomaly pattern extraction subsystem provided in Embodiment 6 of the present invention;

[0017] Figure 5 This is a block diagram of the risk situation simulation subsystem provided in Embodiment 9 of the present invention;

[0018] Figure 6 A block diagram of the electronic device provided by the present invention;

[0019] Figure 7 A block diagram of a computer-readable storage medium provided for this invention.

[0020] Reference numerals: 1. Spatial entity association embedding subsystem; 2. Association anomaly pattern extraction subsystem; 3. Risk situation simulation subsystem; 4. Central processing unit / microprocessor / main control chip; 5. Storage medium; 6. Data bus; 7. Input / output bus / external bus / device bus; 8. Display; 9. Input / output device; 10. Computer-readable instructions; 11. Non-transitory computer-readable storage medium. Detailed Implementation

[0021] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Hereinafter, the terms "first," "second," etc., are used for descriptive convenience only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0023] In this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed mechanical connection, a detachable mechanical connection, or an integral part; or, "connection" can be a direct connection or an indirect connection through an intermediate medium. Furthermore, unless otherwise explicitly specified and limited, the term "coupling" should be interpreted broadly. For example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components; it can also be understood as an electrical connection between different components in a circuit structure through physical lines capable of transmitting electrical signals, such as copper foil or wires on a printed circuit board (PCB), to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in a non-contact manner, such as an electrical connection between two components using capacitive coupling to transmit electrical signals.

[0024] In this embodiment of the invention, directional terms such as "up," "down," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and can change accordingly depending on the orientation of the components in the accompanying drawings.

[0025] Example 1: As Figure 1 As shown, this embodiment of the invention provides a knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines, comprising:

[0026] The spatial entity association embedding subsystem 1 is used to receive real-time monitoring data streams collected by various types of sensors deployed within the mine's enclosed space. These real-time monitoring data streams include parameters such as gas component concentration, ambient temperature and humidity, and air pressure fluctuations. The subsystem aligns each monitoring data stream according to timestamps and spatial coordinates to form a unified data frame. Based on a pre-constructed mine enclosed space entity model containing spatial entities such as roadways, working faces, ventilation facilities, and electromechanical equipment, and their topological relationships, the subsystem binds each monitoring data point to its corresponding mine enclosed space entity through spatial coordinate matching within the data frame. The bound real-time monitoring data undergoes parameter type identification and numerical quantization, extracting feature vectors for each monitoring data point containing information such as parameter type, value, time, and the entity it belongs to. These feature vectors are used as new nodes and connected to existing mine enclosed space entity nodes via edges representing monitoring relationships and influence relationships, dynamically updating the mine enclosed space knowledge graph and generating a new graph structure containing real-time monitoring data.

[0027] The associated anomaly pattern extraction subsystem 2 is used to receive dynamically updated knowledge graphs of mine enclosed spaces; traverse all monitoring data points in the knowledge graph, extract their parameter values, and compare them with the normal fluctuation range of the parameters in the same historical period or adjacent spatial areas, marking abnormal data nodes that exceed the threshold; starting from the abnormal data nodes, perform a breadth-first search along the associated edges of spatial adjacency, ventilation impact, and equipment connection in the knowledge graph to trace the possible propagation paths of abnormal parameters, forming an anomaly propagation subgraph; perform cluster analysis on the abnormal data nodes in the anomaly propagation subgraph, and group nodes with similar anomaly characteristics and propagation relationships into the same anomaly pattern based on the spatiotemporal distance and association strength between the abnormal data nodes; each anomaly pattern contains a set of abnormal parameters, its spatial entity, and the association paths between them; output the set of anomaly patterns and its corresponding subgraph structure.

[0028] The Risk Situation Simulation Subsystem 3 receives the set of abnormal patterns and their subgraph structures, and calls up snapshots of the mine confined space knowledge graph from multiple historical time points. It compares the current abnormal pattern subgraph with the normal evolution trajectory of the same spatial region in the historical mine confined space knowledge graph, calculating the rate and direction of change of the attribute values ​​of the current abnormal nodes, as well as the expansion trend of the abnormal propagation path. Based on the evolutionary characteristics of the rate and direction of change and the expansion trend, it constructs a risk transmission network using the dependencies of ventilation and electromechanical systems between entities in the mine confined space knowledge graph. By simulating the diffusion process of abnormal parameters in the risk transmission network, it simulates the range of changes in adjacent spatial entities and their parameters that may be affected in future time periods. Finally, it synthesizes the simulation results of each abnormal pattern, generates a risk level assessment and early warning area identifier, and outputs it to the monitoring terminal.

[0029] In this context, monitoring data points within a mine's enclosed space refer to the specific physical locations of various sensing devices deployed within enclosed areas of the mine, such as goafs, blind tunnels, and chambers. Each monitoring data point corresponds to a unique spatial coordinate and is associated with one or more monitored parameter types, such as methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure. The real-time monitoring data stream consists of raw data sequences continuously collected and transmitted to the system by the sensing devices at each monitoring data point. This sequence includes parameter values ​​at each sampling time, the corresponding timestamp, and the spatial coordinate information of the monitoring data point. The mine's enclosed space entity model is a pre-constructed digital model describing various physical objects and their spatial topological relationships within the mine's enclosed space. It includes entities such as roadways, working faces, ventilation facilities (e.g., air doors, air walls), and electromechanical equipment (e.g., local ventilators, sensors), defining the connection relationships, relative positions, and ventilation network logic between entities. The "entire" entity within the mine's enclosed space refers to the concrete physical objects in the model, i.e., the various components within the mine's enclosed space, such as a specific roadway, a working face, or a local ventilator. Each entity corresponds to a node in the knowledge graph and possesses attributes such as entity type, spatial coordinates, and current state. Feature vectors are structured information extracted from monitoring data points, representing the multi-dimensional attributes of each data point. For each monitoring data point, its feature vector is composed of the parameter type, parameter value, collection timestamp, and the bound spatial entity identifier. This vector connects real-time data and the knowledge graph, generating new graph nodes. The mine confined space knowledge graph is a knowledge base organized in a graph structure, integrating static spatial entities and dynamic monitoring data. Nodes in the graph include mine confined space entities and feature vector nodes transformed from monitoring data points. Edges in the graph include topological relationships between entities, such as adjacency, containment, dependency, and monitoring and influence relationships between entities and data. It is dynamically updated in real-time to reflect the current mine environment status. Abnormal data nodes in the mine confined space knowledge graph are the feature vector nodes corresponding to monitoring data points whose parameter values ​​exceed the normal fluctuation range. These nodes are marked as abnormal and serve as the starting point for abnormal propagation analysis. An anomaly pattern is a set of multiple spatiotemporally correlated anomaly data nodes with similar anomaly characteristics and their propagation paths. An anomaly pattern describes the occurrence, correlation, and possible propagation direction of anomaly parameters within a local area, such as the combination of anomaly methane concentration in a tunnel and anomaly oxygen concentration in an adjacent area, connected by a ventilation influence edge. The normal evolution trajectory is extracted from historical knowledge graph snapshots at multiple time points, showing typical patterns of parameter changes over time within the same spatial area. This trajectory reflects the fluctuation range, trend, and synergistic relationship of each parameter value under anomaly-free conditions, used to compare and identify the degree of deviation from the current anomaly. Change rate, direction, and expansion trend: dynamic characteristics obtained by comparing the current anomaly pattern subgraph with the normal evolution trajectory.The rate of change refers to how quickly the value of the abnormal parameter changes per unit time; the direction of change refers to whether the parameter increases or decreases; the expansion trend refers to the direction and range of the abnormal propagation path in space, such as the abnormal impact spreading to an adjacent mining face.

[0030] The principles described in the above embodiments are referenced in the appendix. Figure 2 This embodiment of a knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines achieves comprehensive perception, accurate anomaly identification, and proactive risk warning through the collaborative operation of a spatial entity association embedding subsystem, an association anomaly pattern extraction subsystem, and a risk situation inference subsystem. Firstly, it achieves deep fusion and dynamic mapping of monitoring data and spatial entities. By aligning real-time data streams collected by various sensors with time and spatial coordinates and binding them to pre-constructed entity models such as roadways, working faces, and ventilation facilities, discrete monitoring point data is transformed into structured information with clear spatial attribution and topological relationships. The extraction of feature vectors and dynamic updating of the knowledge graph enable real-time monitoring data to be embedded into the existing knowledge network as nodes, constructing a unified graph model that integrates static entity relationships and dynamic monitoring parameters. This effectively solves the data silo problem in traditional monitoring systems, forming a computable digital twin covering the entire mine space with data and entity associations. Secondly, the system achieves correlation mining of abnormal patterns and visualized tracking of propagation paths. It identifies abnormal nodes by comparing real-time data with historical normal fluctuation ranges, and performs a breadth-first search based on spatial adjacency and ventilation influence edges in the knowledge graph to reveal the diffusion trajectory of abnormal parameters, forming an abnormal propagation subgraph. Through cluster analysis, it groups spatiotemporally similar and closely related abnormal nodes into unified abnormal patterns, not only identifying isolated abnormal points but also capturing the correlation characteristics and clustering behavior of anomalies in the spatial network. This enhances the contextual awareness capability of anomaly detection, helping to distinguish between local disturbances and the emergence of systemic risks. Finally, the system achieves quantitative extrapolation and hierarchical early warning of risk situations. By comparing the current abnormal pattern subgraph with the normal evolution trajectory in historical knowledge graph snapshots, the system calculates the rate of change, direction, and expansion trend of the propagation path of abnormal parameters. Based on a risk transmission network constructed from relationships such as ventilation dependence and electromechanical dependence between entities, it simulates the diffusion process of anomalies within it, extrapolating the spatial entities that may be affected and the range of parameter changes in future periods. By combining the results of various anomaly patterns, the system generates risk level assessments and early warning area identifiers, thus achieving a leap from anomaly identification to risk prediction. This provides a temporal and spatial basis for pre-intervention and emergency decision-making.

[0031] In summary, this embodiment uses knowledge graphs to perform entity association and dynamic fusion of multi-source heterogeneous monitoring data, realizing the correlation extraction and propagation tracking of abnormal patterns, and on this basis, completing the simulation and deduction of risk situation and graded early warning; it enhances the systematicness, foresight and intelligence level of mine safety monitoring, and provides integrated technical support for the early detection, accurate positioning and proactive prevention and control of complex risk factors in confined spaces.

[0032] Example 2: Figure 3 As shown, based on Embodiment 1, the spatial entity association embedding subsystem 1 provided in this embodiment of the invention specifically includes:

[0033] The spatial entity mapping component is used to perform timestamp parsing and alignment processing on real-time monitoring data streams. It interpolates and synchronizes asynchronously arriving data based on a unified time base, forming a temporally continuous and aligned multi-parameter data sequence. It extracts the spatial coordinate information attached to each monitoring data point in the multi-parameter data sequence and performs matching calculations with the spatial coordinate range occupied by each entity in the pre-built mine-use enclosed space entity model. Based on the coordinate inclusion relationship, it assigns each data point to the corresponding mine-use enclosed space entity. After the matching is completed, it outputs a set of bound data units consisting of entity identifier, timestamp, parameter type, and original value.

[0034] The structured representation component is used to identify and determine the parameter type in each bound data unit. Based on the identified parameter type, it calls the preset physical quantity conversion rules to convert the original value into a quantized value under the standard physical dimensions. It extracts the entity identifier, timestamp, parameter type, and converted quantized value from each bound data unit and combines them to generate a feature vector of the monitoring data point. Each feature vector contains four dimensions: entity identifier, timestamp, parameter type, and quantized value. The feature vectors of all monitoring data points are organized into a feature vector sequence in chronological order.

[0035] The association edge construction component is used to instantiate each feature vector in the feature vector sequence into a new node in the knowledge graph of mine confined space. The attributes of the new node include timestamp, parameter type, and quantization value. It iterates through all new nodes, locates the corresponding mine confined space entity node in the existing knowledge graph based on the entity identifier carried by each new node, and adds a monitoring relationship edge between the new node and the entity node. At the same time, it calls the parameter influence rule library that pre-configured defines the possible mutual influence types between different spatial entities of different parameter types. Based on the parameter type and the entity to which the new node belongs, it identifies other spatial entities or parameters that may be affected by it, and adds influence relationship edges between related entity nodes or between the new node and the affected entity node. After all new nodes and relationship edges are added, the updated mine confined space knowledge graph is output.

[0036] In the above embodiments, the spatial entity association embedding subsystem of this embodiment achieves the technical effect of dynamically updating knowledge graphs from raw monitoring data through the synergistic effect of the spatial entity mapping component, the structured expression component, and the association edge construction component. Specifically, the spatial entity mapping component performs timestamp parsing and alignment processing on the real-time monitoring data stream, synchronously generating continuously aligned multi-parameter data sequences based on a unified time reference through interpolation; by extracting the spatial coordinate information of data points and matching it with the coordinate range of a pre-built mine-use enclosed space entity model, each data point is assigned to a corresponding entity, outputting a set of bound data units consisting of entity identifier, timestamp, parameter type, and original value; achieving accurate positioning of asynchronous multi-source data in the spatiotemporal dimension, providing a structured data foundation for subsequent processing. The structured expression component identifies the parameter types in the bound data units, converting the original values ​​into quantized values ​​under standard physical dimensions according to preset physical quantity conversion rules; extracting entity identifier, timestamp, parameter type, and quantized value combinations to generate feature vectors, fixedly containing four dimensions and organized into a feature vector sequence in chronological order; unifying the expression specifications of multi-parameter data, eliminating the influence of dimensional differences on association analysis, and forming a standardized data sequence suitable for graph-based processing. The association edge construction component instantiates feature vector sequences into new nodes in the knowledge graph. Node attributes include timestamps, parameter types, and quantization values. It locates corresponding entity nodes in the existing graph using entity identifiers and adds monitoring relationship edges. Based on a parameter influence rule base, it identifies the types of mutual influences that parameter types may generate between spatial entities and adds influence relationship edges between related entity nodes or new nodes and affected entity nodes. The final output is an updated knowledge graph for the mine's enclosed space. This achieves automatic mapping of monitoring data to graph nodes and, through rule-driven association edge construction, dynamically reveals the propagation path and influence relationships of parameter changes between spatial entities.

[0037] In summary, this embodiment transforms discrete and heterogeneous monitoring data into a knowledge graph structure with clear spatiotemporal attribution, unified dimensional expression, and multidimensional relationships through three stages of processing: spatiotemporal alignment, standardization transformation, and graph association. This provides a traceable and reasonable data foundation for the analysis of the state of enclosed spaces.

[0038] Example 3: Based on Example 2, the associated edge construction component provided in this embodiment of the invention specifically includes:

[0039] The node consistency verification module extracts entity identifiers from each feature vector to generate an entity identifier list. Based on the entity identifier list, it searches the entity index table of the mine confined space knowledge graph in parallel to obtain the current attribute set of each entity node. The attribute set includes entity type, spatial coordinate range, and current operating status. It compares the current attributes of each entity node with the monitoring point attributes implicit in the corresponding feature vector to verify whether the monitoring data point is located within the spatial range of the entity and whether the entity type matches the monitoring parameter type. After verification, it outputs a set of matching pairs containing the new node identifier, the entity node identifier, and a verification pass flag.

[0040] The type dynamic assignment module is used to define standard monitoring relationship types corresponding to different combinations of entity types and parameter types; it iterates through the matching pair set, queries and assigns specific monitoring relationship types based on the entity type of the entity node in each matching pair and the parameter type in the feature vector of the corresponding new node; after completing the assignment, it outputs a set of matching pairs with type, which includes the new node identifier, the entity node identifier, and the assigned monitoring relationship type.

[0041] The graph index update module is used to create a directed edge from a new node to an entity node for each type matching pair in the mine-use enclosed space knowledge graph. The edge type is the assigned monitoring relationship type. A pointer to the new node is added to the associated edge index structure of the entity node to form a bidirectional index. After all monitoring relationship edges are added, the spatiotemporal index of the knowledge graph in the mine-use enclosed space knowledge graph is incrementally updated. The new node is included in the index structure with its spatial coordinates and timestamp as the key value, and the updated mine-use enclosed space knowledge graph is output.

[0042] In the above embodiments, this embodiment ensures the physical rationality of the association through consistency verification, standardizes the relationship type through dynamic assignment, and supports efficient querying through index updates, forming a monitoring association network with clear semantic types, bidirectional traceability, and spatiotemporal indexability.

[0043] Example 4: Based on Example 3, the map index update module provided in this embodiment of the invention specifically includes:

[0044] The bidirectional pointer addition task generation submodule receives a set of created directed edges, each containing a source node identifier, a target node identifier, and an edge type; it traverses the set of directed edges, extracts the source node identifier or new node, target node identifier or mine confined space entity node, and edge type for each edge, and generates a pointer addition task unit containing the three elements of source node identifier, target node identifier, and edge type; it summarizes all task units and outputs a pointer addition task list.

[0045] The entity node outgoing pointer update submodule is used to locate the corresponding entity node in the node index of the mine-use closed space knowledge graph based on the target node identifier in each task unit, and access the outgoing associated edge index storage area of ​​the entity node; it combines the source node identifier and edge type in the task unit into a new outgoing pointer record and appends it to the end of the outgoing associated edge index storage area of ​​the entity node. After completing the processing of all task units, it outputs a set of confirmation information containing the entity node identifiers with updated outgoing pointers and their corresponding newly added pointer records;

[0046] The bidirectional association submodule is used to locate newly created nodes in the mine-use enclosed space knowledge graph based on the source node identifier in each task unit; check whether the incoming association edge index storage area of ​​the newly created node exists, and if not, initialize an empty incoming association edge index storage area for it; combine the target node identifier and edge type in the task unit into a new incoming pointer record and write it to the incoming association edge index storage area of ​​the new node; after completing all task units, output the mine-use enclosed space knowledge graph with updated bidirectional index.

[0047] In the above embodiments, this embodiment realizes the pipeline transformation from edge creation to index update through task generation, supports forward traversal from entity to monitoring node through outgoing pointer update, and supports reverse tracing from monitoring node to entity through incoming pointer update; finally, a complete bidirectional index structure is built in the knowledge graph to ensure that the relationship can be efficiently retrieved and verified on any node side, thereby improving the performance of the graph in complex queries and path analysis.

[0048] Example 5: Based on Example 4, the bidirectional association submodule provided in this embodiment of the invention specifically includes:

[0049] The incoming pointer record assembly unit receives each task unit from the bidirectional pointer addition task list. Each task unit contains a source node identifier, a target node identifier, and an edge type. It extracts the target node identifier and edge type from each task unit. It then calls the node identifier encoding table of the mine-use enclosed space knowledge graph storage layer to convert the target node identifier into a binary encoding sequence of fixed byte length. Next, it calls the edge type enumeration value mapping table to convert the edge type into a predefined binary enumeration value. Finally, it concatenates the converted target node identifier binary sequence with the edge type binary enumeration value in the order of node identifier followed by edge type, generating a fixed-length incoming pointer record data block. After all task units are processed, it outputs a set of incoming pointer record data blocks to be written for each task unit.

[0050] The inbound index storage area write unit receives the set of inbound pointer record data blocks to be written and the source node identifier corresponding to each data block; based on each source node identifier, it locates the corresponding newly created node in the node storage area index of the mine-use enclosed space knowledge graph, and obtains the metadata of the inbound association edge index storage area of ​​the newly created node; the metadata includes the starting physical address of the storage area, the number of records currently written, the fixed length of each record, and the total capacity of the storage area; based on the fixed length of each record and the number of records currently written, it calculates the offset of the next writable record, which is equal to the number of records currently written multiplied by the fixed length of the record; it adds the starting physical address to the offset to obtain the target physical address for this write; at the same time, it compares whether the number of new records after writing exceeds the total capacity of the storage area, and if it does, it marks that the storage area needs to be expanded; after completing the calculation, it outputs the write target physical address and the updated record count corresponding to each task unit;

[0051] The metadata update unit receives the set of inbound pointer record data blocks to be written, the write target physical address corresponding to each data block, and the updated record count; it writes each data block to the persistent storage medium of the mine-use enclosed space knowledge graph according to its target physical address and performs a write confirmation operation; after all data blocks are successfully written, it updates the metadata of the inbound associated edge index storage area according to the updated record count corresponding to each new node, modifies the current number of written records in the metadata to the updated record count, and recalculates the current write offset; it writes the updated metadata back to the storage area header; after all operations are completed, it outputs a list of node identifiers whose inbound index has been updated as confirmation information for the completion of the bidirectional index update.

[0052] In the above embodiments, this embodiment achieves binary serialization of associated data through record assembly, realizes accurate calculation of storage addresses and capacity monitoring through writing units, and achieves persistent storage and index status synchronization through metadata updates. Finally, a structured inbound association index is established on the monitoring node side, supporting rapid backtracking of associated entity nodes from the monitoring node, and together with the outbound index, it forms a complete bidirectional reference system, improving the efficiency and reliability of the knowledge graph for mine-use enclosed spaces in reverse query and association verification scenarios.

[0053] Example 6: As Figure 4 As shown, based on Example 1, the associated anomaly pattern extraction subsystem provided in this embodiment of the invention specifically includes:

[0054] The correlation metric extraction component receives an anomaly propagation subgraph, which contains a set of anomalous data nodes and correlation edges between nodes, such as spatial adjacency, ventilation impact, and equipment connection. It traverses all anomalous data nodes in the subgraph, extracting parameter type, parameter value, timestamp, and spatial entity identifier from each node's attributes to generate a multi-dimensional feature vector for each node. It then traverses all correlation edges, extracting the identifiers of the two nodes connected by each edge and the edge type, and applying preset correlation strength weights for the three edge types: spatial adjacency, ventilation impact, and equipment connection. Based on the spatial coordinates and timestamp of each anomalous data node, it calculates the Euclidean spatial distance and absolute time difference between each pair of nodes. It then weights and fuses the spatial distance, absolute time difference, and the type weights of the correlation edges between the two nodes to calculate a comprehensive correlation metric value between each pair of nodes. Finally, it outputs a metric matrix containing all anomalous node identifiers and their feature vectors, as well as the comprehensive correlation metric values ​​between node pairs.

[0055] The density clustering component receives the metric matrix and the anomaly propagation subgraph, and sets the comprehensive association metric threshold and the spatiotemporal distance threshold. Using each anomaly node as an initial seed, it traverses the metric matrix, filtering out nodes whose comprehensive association metric value with the current seed node is greater than the association metric threshold and whose spatial distance and temporal difference are both less than the corresponding thresholds, forming a candidate neighborhood node set. For each node in the candidate neighborhood node set, it checks in the anomaly propagation subgraph whether there is at least one connected path between it and the seed node, consisting of spatial adjacency, ventilation influence, or equipment connection edges. If such a path exists, the seed node is marked as a node in the same cluster; otherwise, it is discarded. The newly marked nodes in the same cluster are used as new seeds, and the above process is repeated until no further expansion is possible, forming a connected cluster. After all connected clusters are constructed, the node list of each connected cluster and the set of retained association edges between nodes within the cluster are output.

[0056] An abnormal pattern structure generation component is used to: count the frequency of parameter types of all abnormal data nodes within each connected cluster, and determine the most frequent parameter type as the dominant parameter type of the abnormal pattern; extract the earliest and latest timestamps of nodes within the connected cluster to generate the start and end times of the abnormal pattern; summarize the spatial entity identifiers of the nodes within the cluster to form a list of spatial regions involved in the abnormal pattern; construct a propagation path topology structure composed of node identifiers and edge types based on the set of associated edges between nodes within the cluster, and mark the starting node and propagation direction of the path according to the order of the timestamps of each node; integrate the dominant parameter type, start time, end time, list of spatial regions, and propagation path topology structure corresponding to each connected cluster into an abnormal pattern; and summarize the abnormal patterns of all connected clusters to output the abnormal pattern set and its corresponding subgraph structure.

[0057] In the above embodiments, the correlation anomaly pattern extraction subsystem of this embodiment achieves the identification and structured description of anomaly patterns with spatiotemporal and topological correlations from the anomaly propagation subgraph through the collaborative processing of the correlation metric extraction component, density clustering component, and anomaly pattern structuring generation component. Specifically, the technical effects are as follows: the correlation metric extraction component receives the anomaly propagation subgraph, traverses the anomaly data nodes to extract attributes such as parameter type, value, timestamp, and spatial entity identifier, and generates a multi-dimensional feature vector; it traverses the correlation edges to extract node identifiers and edge types, assigning preset weights to edge types such as spatial adjacency, ventilation influence, and equipment connection; based on the node spatial coordinates and timestamps, it calculates the Euclidean distance and absolute value of the time difference between node pairs, and performs weighted fusion based on the correlation edge type weights to calculate the comprehensive correlation metric value for each pair of nodes. This process quantifies the attribute differences, spatiotemporal proximity, and topological correlation strength of nodes into unified values, forming a metric matrix containing node features and node pair correlation, providing multi-dimensional quantitative basis for clustering. The density clustering component receives a metric matrix and anomaly propagation subgraph, and sets association metric thresholds and spatiotemporal distance thresholds. Using each anomaly node as a seed, it filters candidate nodes whose comprehensive association metric value is greater than the threshold and whose spatiotemporal distance is less than the threshold. In the anomaly propagation subgraph, it checks whether there are connected paths between the seed and candidate nodes formed by spatial adjacency, ventilation influence, or equipment connection edges. Nodes with existing paths are retained as nodes in the same cluster and iteratively expanded to form connected clusters. Combining quantitative metrics and graph structure connectivity ensures that nodes within the same cluster not only satisfy spatiotemporal and metric similarity but are also mutually reachable on the actual association network, avoiding the mis-clustering of nodes with no direct or indirect connections into one class and improving the topological rationality of the clustering results. The anomaly pattern structure generation component counts the frequency of node parameter types for each connected cluster, identifying the most frequent type as the dominant parameter type; extracts the earliest and latest timestamps of nodes within the cluster to generate the start and end times of the anomaly pattern; summarizes the spatial entity identifiers of nodes to form a list of spatial regions; constructs a propagation path topology based on the set of associated edges within the cluster, marking the starting node and propagation direction in timestamp order; and integrates the dominant parameter type, start and end times, list of spatial regions, and propagation path topology into a structured anomaly pattern. This transforms connected clusters into anomaly pattern descriptions with clear semantic features, including core anomaly parameters, spatiotemporal range, affected area, and propagation path, resulting in an interpretable and directly usable structured output for analysis and decision-making.

[0058] In summary, this embodiment achieves quantitative characterization of multidimensional relationships between anomalous nodes through correlation metric extraction, identifies anomalous groups based on metrics and connectivity through density clustering, and performs feature induction and topological reconstruction of anomalous patterns through structured generation. Finally, it extracts a set of anomalous patterns with clear spatiotemporal boundaries, dominant parameter features, and propagation paths from complex anomalous propagation subgraphs, providing structured data support for root cause analysis, impact assessment, and decision-making regarding anomalous events in confined spaces.

[0059] Example 7: Based on Example 6, the abnormal pattern structure generation component provided in this embodiment of the invention specifically includes:

[0060] The data structuring assembly subcomponent is used to assign an anomaly pattern identifier code to each connected cluster, write the dominant parameter type into the pattern attribute field corresponding to the anomaly pattern identifier code, combine the start time and end time into a time interval and write it into the pattern attribute field, sort the entity identifiers in the spatial region list according to spatial adjacency and write them into the spatial range field, and write the node sequence and edge type sequence in the propagation path topology into the propagation path field in chronological order. After all fields are filled, the anomaly pattern structured record containing the anomaly pattern identifier code and complete attribute fields is output.

[0061] The spatial entity association and binding sub-component is used to traverse the node list and extract the identifier of the spatial entity to which each node belongs; locate the corresponding entity node in the knowledge graph of mine confined space based on the entity identifier, write the identifier code of the abnormal mode and the start time of the abnormal mode into the associated event index area of ​​each entity node; at the same time, write the entity node's own identifier and entity type into the affected entity field of the abnormal mode structured record; after the binding is completed, output the complete record of the abnormal modes of the bound spatial entities.

[0062] The temporal index construction subcomponent is used to extract the start and end times of each record, using the start time as the first key and the end time as the second key to generate a temporal index entry for the anomaly pattern; it also extracts the entity identifiers contained in the spatial region list of each record, using each entity identifier as the key and the anomaly pattern identifier code as the value to generate a spatial index entry for the anomaly pattern; it writes all temporal and spatial index entries into the anomaly pattern index table of the mine-use enclosed space knowledge graph; after completing the index construction, it outputs a complete result containing the anomaly pattern set and its corresponding subgraph structure.

[0063] In the above embodiments, this embodiment achieves standardized encapsulation of abnormal patterns through data structured assembly, realizes bidirectional linking between patterns and graph entities through spatial entity association binding, and realizes efficient indexing of patterns in time and space dimensions through temporal index construction. Finally, it outputs a set of abnormal patterns with complete attributes, entity associations and multi-dimensional indexes, which can be integrated into the knowledge graph system as a structured knowledge unit, supporting fast query, association analysis and decision-making applications based on time, space, entity and topological features.

[0064] Example 8: Based on Example 7, the time-series index construction sub-component provided in this embodiment of the invention specifically includes:

[0065] The index entry receiving module is used to receive time index entries and spatial index entries, access the metadata of the abnormal mode index table, read the number of currently written entries and the remaining capacity, calculate the total number of entries and compare it with the remaining capacity. If the capacity is insufficient, it requests an expansion storage block from the storage management unit, obtains the address of the new storage block, and outputs a list of write tasks containing the set of entries and the starting address of the write.

[0066] The Write Task List module is used to map time index entries to corresponding time partitions by year and month according to the partitioning rules of the exception mode index table; map spatial index entries to corresponding spatial partitions by entity identifier prefix; sort each partition by start time; generate continuous data blocks and their write offsets for each partition; and output a set of data to be written containing partition identifier, offset, and data blocks.

[0067] The data set receiving module is used to call the persistent storage interface to write data blocks to the specified storage address in parallel, in units of partitions; collect write confirmation information for each partition; after all partitions are successfully written, update the metadata of the abnormal mode index table, increase the entry count of each partition, update the write offset, merge the global time range, and output the updated knowledge graph of the mine's enclosed space.

[0068] In the above embodiments, this embodiment ensures the availability of index storage through capacity monitoring and expansion, optimizes the organization efficiency of index data through spatiotemporal partitioning and sorting, and achieves efficient and reliable persistence operations through parallel writing and metadata updates. Ultimately, it constructs an anomaly pattern index structure in the knowledge graph that supports fast retrieval by time interval and spatial entity, improving the response performance and scalability of anomaly patterns in complex spatiotemporal query scenarios.

[0069] Example 9: As Figure 5 As shown, based on Embodiment 1, the risk situation simulation subsystem 3 provided in this embodiment of the invention specifically includes:

[0070] The network topology construction component is used to traverse all abnormal data nodes in the abnormal pattern subgraph and obtain the spatial entity identifier of each abnormal data node. Starting from the entity identifier, it extracts all upstream and downstream entity nodes and connecting edges that have ventilation dependencies with the entity in the knowledge graph of the mine confined space. At the same time, it extracts all associated entity nodes and connecting edges that have electromechanical dependencies. The extracted entity nodes are connected according to the dependency direction to construct a directed network topology structure containing the starting entity, dependency path, and ending entity, and outputs the risk transmission network.

[0071] The attenuation calculation component is used to set an attenuation coefficient based on the dependency type for each dependency edge in the risk propagation network. Starting from the entity where the abnormal data node is located, the abnormal parameters are propagated level by level along the dependency direction. After passing through each dependency edge, the abnormal parameter value of the upper-level entity is multiplied by the attenuation coefficient of the dependency edge to obtain the affected parameter value of the lower-level entity. This process is repeated until all reachable paths have been traversed, and the identifier, the path to the affected entity node, and the set of calculated affected parameter values ​​are output.

[0072] The parameter range generation component applies the rate of change of the current abnormal data node to the affected parameter values, calculates the parameter values ​​of each affected entity at a future specified time point in time, summarizes all affected entities, merges the affected parameter values ​​from multiple sources for the same entity into the parameter change range of the entity, and outputs a list of adjacent spatial entity identifiers that may be affected in the future time period and the corresponding parameter change range of each entity.

[0073] In the above embodiments, the risk situation simulation subsystem 3 of this embodiment achieves risk transmission network construction, impact propagation calculation, and future parameter range prediction based on anomaly pattern subgraphs through the collaborative operation of network topology construction components, attenuation calculation components, and parameter range generation components. Specifically, the technical effects are as follows: the network topology construction component traverses the abnormal data nodes in the anomaly pattern subgraph to obtain their spatial entity identifiers; starting from the entity identifiers, it extracts all upstream and downstream related entity nodes and connecting edges of ventilation and electromechanical dependencies in the knowledge graph; it connects the extracted entities according to the dependency direction to construct a directed network topology structure containing the starting entity, dependency path, and ending entity; based on the predefined entity dependency relationships in the knowledge graph, it expands isolated abnormal nodes into a risk transmission network with a clear direction, revealing potential paths for anomalies to propagate along the ventilation and electromechanical systems, providing a structural basis for risk impact range analysis. The attenuation calculation component sets an attenuation coefficient based on the dependency type for each dependency edge in the risk propagation network. Starting from the entity where the abnormal data node is located, the abnormal parameter value is propagated level by level along the dependency direction. After passing through each dependency edge, the abnormal parameter of the previous level is multiplied by the attenuation coefficient of that edge to obtain the affected parameter value of the next level entity. After traversing all reachable paths, the component outputs the identifier, the path to the destination, and the set of affected parameter values ​​for each affected entity. By introducing the attenuation coefficient, the component simulates the intensity attenuation of risk on the dependency path, realizes the quantitative propagation calculation of the abnormal impact in the dependency network, and obtains the specific degree of impact of each entity at the end of the propagation path, providing a quantitative basis for risk assessment. The parameter range generation component applies the rate of change of the current abnormal data nodes to the affected parameter values, calculates the parameter values ​​of each affected entity at a future specified time point in time, and aggregates all affected entities, merging multiple source parameter values ​​of the same entity into the parameter change range of that entity. It outputs a list of adjacent spatial entities that may be affected in the future time period and the corresponding parameter change range of each entity. By introducing the time dimension and rate of change, it expands the static affected parameter values ​​into a parameter range prediction that evolves over time, supporting the dynamic extrapolation of the risk situation in the future time period. It merges the results of multi-path impacts to form a comprehensive parameter change range at the entity level, reflecting the multi-source superposition impact that the entity may be subject to.

[0074] In summary, this embodiment constructs a dependency network to clarify the risk propagation structure, quantifies the path of impact intensity through attenuation calculation, and generates a parameter range to extrapolate the risk situation for future periods. Finally, it outputs the potential scope of entities affected by the abnormal pattern in the future and its parameter variation range, providing situation prediction support for risk warning and prevention decisions based on dependency networks, quantified attenuation, and time extrapolation.

[0075] Example 10: Based on Example 9, the network topology construction component provided in this embodiment of the invention specifically includes:

[0076] The dependency direction identification subcomponent is used to traverse each connection edge, identify the direction attribute corresponding to its dependency type, mark edges with positive direction attribute as pointing from the start point to the end point, and mark edges with negative direction attribute as pointing from the end point to the start point; based on the direction mark of each edge, perform topological sorting on the entity nodes on the same dependency path to generate a node sequence arranged according to the dependency flow direction.

[0077] The directed path generation sub-component connects the nodes in each node sequence in order to form a directed path from the starting entity through intermediate entities to the ending entity; it detects whether there are shared nodes between different paths; if there are shared nodes, it merges multiple paths at the shared nodes to form a branch structure; and it outputs the intermediate topology structure containing all nodes and the merged path branches.

[0078] The risk transmission network output sub-component is used to extract all entity nodes and directed connections between nodes in the intermediate topology, instantiate the nodes and connections into network nodes and directed edges, combine all the instantiated nodes and edges into a complete risk transmission network, and output the risk transmission network structure.

[0079] In the above embodiments, the network topology construction component transforms entity dependencies in a knowledge graph into a risk propagation network with a clear flow direction and branching structure by identifying dependency directions, generating directed paths, merging shared nodes, and instantiating the network. The final output is a structured directed network model, providing an accurate, clear, and structurally complete topological foundation for analyzing the propagation of risk along dependency paths.

[0080] Example 11: Based on Example 10, the directed path generation sub-component provided in this embodiment of the invention specifically includes:

[0081] The path association module is used to traverse all node sequences, extract node identifiers from each node sequence, and generate a list of node identifiers for each path; perform an intersection operation on the node identifier lists of all paths to detect node identifiers that appear in two or more paths, and mark the nodes as shared nodes; at the same time, it records the original path identifier of each shared node and its position number in the path; and outputs a set of shared node information containing shared node identifiers, a list of associated path identifiers, and position numbers.

[0082] The path reorganization module is used to, for each shared node, truncate all original paths passing through the shared node at the node according to its associated path identifier list, and generate subsequent path segments starting from the shared node; connect multiple subsequent path segments with the same shared node in parallel to the shared node to form a structure in which multiple sub-paths branch from the shared node; after all shared nodes have been processed, the module outputs an intermediate topology structure with branches, which is composed of original path segments and branch sub-paths.

[0083] The node deduplication module is used to traverse all path segments in the intermediate topology and detect whether the same node appears repeatedly in different path segments. If duplicates exist, an instance of the shared node is retained, and all upstream connections pointing to the shared node are merged. At the same time, all downstream branches leading from the shared node are connected to the retained instance. After deduplication and connection merging are completed, the final intermediate topology containing all nodes and merged path branches is output.

[0084] In the above embodiments, the directed path generation sub-component of this embodiment integrates a linear node sequence into an intermediate topology with branching structure and no node redundancy by identifying shared nodes, reorganizing branch paths, and deduplicating and merging nodes. The final output is a structured branching topology that accurately reflects the intersection and distribution relationships of paths in the dependency network, providing a clear and compact intermediate representation for the construction of risk transmission networks.

[0085] Example 12: Based on Example 11, the path reorganization module provided in this embodiment of the invention specifically includes:

[0086] The shared node branch path extraction submodule is used to filter all subsequent path segments starting from the shared node from the original path set based on the shared node identifier; each subsequent path segment contains a continuous sequence of nodes from the next node of the shared node to the end of the path and the connection edge type between the nodes; the extracted subsequent path segments are grouped according to the shared node identifier to generate a subsequent path segment group corresponding to each shared node; the output is a set of data to be connected containing the shared node identifier and its corresponding subsequent path segment group.

[0087] The spatial topology construction submodule is used to locate the spatial entity corresponding to each shared node in the mine enclosed space knowledge graph, obtain the exit direction and connectable entity types of the spatial entity; match the first node of each segment in the subsequent path segment group with the corresponding exit of the shared node, and connect the first node of the segment to the designated exit of the shared node according to the matching result, forming multiple branch paths starting from the shared node; at the same time, it records the sequence of entities such as roadways, ventilation doors, and mining faces passed through each branch path; and outputs a preliminary branch topology structure containing the shared node identifier and the attached branch paths.

[0088] The conflict resolution submodule is used to traverse each branch path, check whether the connection edge type between adjacent nodes in the branch path is consistent with the ventilation dependency or electromechanical dependency relationship defined in the mine enclosed space entity model; check whether different branch paths intersect or overlap in the same spatial area. If there is an intersection and the path direction conflicts, one path is retained according to the dependency priority, and the remaining paths are adjusted to connect to other exits of the shared node; after completing the verification and resolution, the output is the final branch topology structure containing all shared nodes and multiple sub-paths branching from each node.

[0089] In the above embodiments, this embodiment extracts subsequent path segments, attaches branches based on spatial constraints, and resolves dependencies and spatial conflicts to reorganize the linear subsequent segments at shared nodes into a branch topology that conforms to the layout and dependency relationships of a mine's enclosed space. The final output is a branch topology with spatial rationality and logical consistency, providing an accurate and feasible structured representation for the multipath propagation analysis of risks in enclosed space dependency networks.

[0090] Example 13: Based on Example 12, the spatial topology construction submodule provided in this embodiment of the invention specifically includes:

[0091] The first node feature extraction unit is used to traverse each subsequent path segment group and extract the node identifier of the first node from each subsequent path segment; based on the node identifier, it queries the attribute of the spatial entity corresponding to the node in the mine enclosed space knowledge graph. The attribute includes entity type, spatial coordinates, and orientation information relative to the shared node; it combines the orientation information of each first node with the entity type to generate the connection feature identifier of the subsequent path segment; and it outputs a matching feature set containing the shared node identifier, the identifier of each subsequent path segment, and its corresponding connection feature identifier.

[0092] The exit feature parsing unit is used to receive the matching feature set and locate the spatial entity corresponding to the shared node in the knowledge graph of the mine confined space based on the shared node identifier contained therein; access the exit attribute list of the spatial entity, where each record contains the exit identifier, exit direction, and range of entity types that the exit can connect to; combine the exit direction with the range of entity types that can connect to generate the exit feature identifier set of the shared node; and output the exit feature information containing the shared node identifier and its corresponding exit feature identifier set.

[0093] The exit matching and binding unit is used to compare the connection feature identifier of each subsequent path segment with each exit feature identifier in the set of exit feature identifiers of the shared node one by one. When the orientation information and entity type in the connection feature identifier completely match the exit direction and the range of connectable entity types in a certain exit feature identifier, the first node of the subsequent path segment is bound to the exit identifier. After all the matching is completed, the matching result set containing the shared node identifier, the matched exit identifier, and the corresponding subsequent path segment identifier is output.

[0094] In the above embodiments, this embodiment extracts the features of the first node, parses the features of the exit node, and performs feature matching to accurately attach subsequent path segments to the corresponding exits of the shared node based on spatial orientation and entity type constraints. The final output is a feature-matched attachment result, ensuring that the branch paths conform to the actual spatial layout and connection rules in the mine's enclosed space knowledge graph, providing a spatially accurate and type-compliant foundation for the construction of subsequent branch topologies.

[0095] Example 14: Based on Example 13, the export matching and mounting unit provided in this embodiment of the invention specifically includes:

[0096] The connection feature standardization encoding subunit is used to call the direction encoding table in the mine confined space knowledge graph to map the orientation information to standard direction codes; call the entity type encoding table to map the entity type to standard type codes; concatenate the standard direction codes and standard type codes to generate a standardized connection feature string; and output a standardized feature set containing shared node identifiers, subsequent path segment identifiers, and standardized connection feature strings.

[0097] The export feature matching domain construction subunit is used to convert the export direction into a standard direction code and decompose the range of connectable entity types into a list of standard type codes that are allowed to be connected; using the standard direction code as the key and the list of allowed connection type codes as the value, an export matching domain dictionary is constructed; the output contains matching domain information including the shared node identifier and its export matching domain dictionary;

[0098] The feature matching and exit binding subunit is used to traverse the standardized feature set, extract the direction code and type code of each entry, retrieve the corresponding list of allowed connection type codes in the exit matching domain dictionary with the direction code as the key, check whether the type code exists in the type code list, if it exists, bind the subsequent path segment identifier with the exit identifier corresponding to the direction code, and output the matching result set.

[0099] In the above embodiments, this embodiment uses standardized coding, matching domain construction, and feature comparison to standardize the matching of the first node features of subsequent path segments with the exit constraints of shared nodes, thereby achieving accurate exit binding of branch paths in the confined space of a mine. The final output is a binding result based on coding matching, ensuring that branch connections conform to standardized direction and type constraints, improving the efficiency and accuracy of the matching process, and providing reliable and consistent exit connection relationships for spatial topology construction.

[0100] Figure 6 A block diagram of an exemplary electronic device suitable for implementing embodiments of the present invention is shown.

[0101] The electronic device may include a central processing unit / microprocessor / main control chip 4; and a storage medium 5 coupled to the central processing unit / microprocessor / main control chip 4 and storing computer-executable instructions therein for performing the steps of various methods of embodiments of the present invention when executed by the processor.

[0102] The central processing unit / microprocessor / main control chip 4 may include, but is not limited to, one or more processors or microprocessors.

[0103] Storage medium 5 may include, but is not limited to, random access memory (RAM), read-only memory (ROM), flash memory, EPROM memory, EEPROM memory, registers, computer storage media (e.g., hard disk, floppy disk, solid-state drive, removable disk, CD-ROM, DVD-ROM, Blu-ray disc, etc.).

[0104] In addition, the electronic device may include (but is not limited to) a data bus 6, an input / output bus / external bus / device bus 7, a display 8, and input / output devices 9 (e.g., keyboard, mouse, speaker, etc.).

[0105] The central processing unit / microprocessor / main control chip 4 can communicate with external devices (8, 9, etc.) via wired or wireless networks (not shown) through the input / output bus / external bus / device bus 7.

[0106] The storage medium 5 may also store at least one computer-executable instruction for performing the steps of various functions and / or methods in the embodiments described herein when the central processing unit / microprocessor / main control chip 4 is running.

[0107] In one embodiment, the at least one computer-executable instruction may also be compiled into or comprise a software product, wherein one or more computer-executable instructions are executed by a processor to perform the steps of the various functions and / or methods in the embodiments described herein.

[0108] Figure 7A schematic diagram of a computer-readable storage medium according to an embodiment of the present invention is shown.

[0109] like Figure 7 As shown, the non-transitory computer-readable storage medium 11 stores instructions, such as computer-readable instructions 10. When the computer-readable instructions 10 are executed by a processor, the various methods described above can be performed. The non-transitory computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-transitory non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. For example, the non-transitory computer-readable storage medium 11 can be connected to a computing device such as a computer, and then, when the computing device executes the computer-readable instructions 10 stored on the non-transitory computer-readable storage medium 11, the various methods described above can be performed.

[0110] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or units may be electrical, mechanical, or other forms.

[0111] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0112] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0113] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for executing all or part of the steps of the methods of the various embodiments of this invention through a computer device (which may be a personal computer, server, or network device, etc.). The aforementioned storage medium includes: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, optical disks, and other media capable of storing program code.

[0114] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines, characterized in that, Include: The spatial entity association embedding subsystem is used to extract the feature vector of each monitoring data point; the feature vector is used as a new node and connected to the original mine-use enclosed space entity nodes with edges, and the mine-use enclosed space knowledge graph is dynamically updated. The associated anomaly pattern extraction subsystem is used to group nodes with similar anomaly characteristics and propagation relationships in the anomaly propagation subgraph into the same anomaly pattern; it outputs the anomaly pattern set and its corresponding subgraph structure. The risk situation simulation subsystem receives a set of abnormal patterns and their subgraph structures, and calls up snapshots of the mine confined space knowledge graph from multiple historical time points. It compares the current abnormal pattern subgraph with the normal evolution trajectory of the same spatial region in the historical mine confined space knowledge graph, calculating the rate and direction of change of the attribute values ​​of the current abnormal nodes, as well as the expansion trend of the abnormal propagation path. Based on the evolutionary characteristics of the rate and direction of change and the expansion trend, it constructs a risk transmission network using the ventilation and electromechanical dependencies between entities in the mine confined space knowledge graph. By simulating the diffusion process of abnormal parameters in the risk transmission network, it predicts the range of changes in adjacent spatial entities and their parameters that may be affected in future time periods. Finally, it synthesizes the simulation results of each abnormal pattern, generates a risk level assessment and early warning area identifier, and outputs it to the monitoring terminal. The spatial entity association embedding subsystem receives real-time monitoring data streams collected by various types of sensors deployed within the mine's enclosed space. These data streams include parameters such as gas component concentration, ambient temperature and humidity, and air pressure fluctuations. The subsystem aligns these data streams according to timestamps and spatial coordinates to form a unified data frame. Based on a pre-constructed mine enclosed space entity model containing spatial entities and their topological relationships, including roadways, working faces, ventilation facilities, and electromechanical equipment, the subsystem binds each monitoring data point to its corresponding mine enclosed space entity through spatial coordinate matching within the data frame. Finally, the subsystem identifies the parameter type and quantifies the values ​​of the bound real-time monitoring data, extracting a feature vector for each monitoring data point containing parameter type, value, time, and entity information. The feature vector is used as a new node, and connected to the original mine-use enclosed space entity node by monitoring relationship and influence relationship edges. The mine-use enclosed space knowledge graph is dynamically updated to generate a new graph structure containing real-time monitoring data. The associated anomaly pattern extraction subsystem receives dynamically updated knowledge graphs of mine enclosed spaces; it traverses all monitoring data points in the knowledge graph, extracts parameter values, and compares them with the normal fluctuation range of parameters in the same historical period or adjacent spatial regions, marking abnormal data nodes that exceed the threshold; starting from the abnormal data nodes, it performs a breadth-first search along the spatial adjacency, ventilation impact, and equipment connection association edges in the mine enclosed space knowledge graph to trace the possible propagation paths of abnormal parameters, forming an anomaly propagation subgraph; it performs cluster analysis on the abnormal data nodes in the anomaly propagation subgraph, grouping nodes with similar anomaly characteristics and propagation relationships into the same anomaly pattern based on the spatiotemporal distance and association strength between the abnormal data nodes; each anomaly pattern contains a set of abnormal parameters, its spatial entity, and the association paths between them; and outputs the set of anomaly patterns and its corresponding subgraph structure.

2. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 1, characterized in that, The spatial entity association embedding subsystem includes: The spatial entity mapping component is used to perform timestamp parsing and alignment processing on real-time monitoring data streams. It interpolates and synchronizes asynchronously arriving data based on a unified time base, forming a temporally continuous and aligned multi-parameter data sequence. It extracts the spatial coordinate information attached to each monitoring data point in the multi-parameter data sequence and performs matching calculations with the spatial coordinate range occupied by each entity in the pre-built mine-use enclosed space entity model. Based on the coordinate inclusion relationship, it assigns each data point to the corresponding mine-use enclosed space entity. After the matching is completed, it outputs a set of bound data units consisting of entity identifier, timestamp, parameter type, and original value. The structured expression component is used to identify and determine the parameter type in each bound data unit, and call the preset physical quantity conversion rules according to the identified parameter type to convert the original value into a quantized value under the standard physical dimension; Extract the entity identifier, timestamp, parameter type, and converted quantized value from each bound data unit, and combine them to generate a feature vector of the monitoring data point; Each feature vector contains four dimensions: entity identifier, timestamp, parameter type, and quantization value. The feature vectors of all monitored data points are organized into a feature vector sequence in chronological order. The association edge construction component is used to instantiate each feature vector in the feature vector sequence into a new node in the knowledge graph of mine confined space. The attributes of the new node include timestamp, parameter type, and quantization value. It iterates through all new nodes, locates the corresponding mine confined space entity node in the existing knowledge graph based on the entity identifier carried by each new node, and adds a monitoring relationship edge between the new node and the entity node. At the same time, it calls the parameter influence rule library that pre-configured defines the possible mutual influence types between different spatial entities of different parameter types. Based on the parameter type and the entity to which the new node belongs, it identifies other spatial entities or parameters that may be affected by it, and adds influence relationship edges between related entity nodes or between the new node and the affected entity node. After all new nodes and relationship edges are added, the updated mine confined space knowledge graph is output.

3. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 1, characterized in that, The risk situation simulation subsystem includes: The network topology construction component is used to traverse all abnormal data nodes in the abnormal pattern subgraph and obtain the spatial entity identifier of each abnormal data node. Starting from the entity identifier, it extracts all upstream and downstream entity nodes and connecting edges that have ventilation dependencies with the entity in the mine confined space knowledge graph. At the same time, it extracts all associated entity nodes and connecting edges that have electromechanical dependencies. The extracted entity nodes are connected according to the dependency direction to construct a directed network topology and output the risk transmission network. The attenuation calculation component is used to set an attenuation coefficient based on the dependency type for each dependency edge in the risk propagation network. Starting from the entity where the abnormal data node is located, the abnormal parameters are propagated level by level along the dependency direction. After passing through each dependency edge, the abnormal parameter value of the upper-level entity is multiplied by the attenuation coefficient of the dependency edge to obtain the affected parameter value of the lower-level entity. This process is repeated until all reachable paths have been traversed, and the identifier, the path to the affected entity node, and the set of calculated affected parameter values ​​are output. The parameter range generation component applies the rate of change of the current abnormal data node to the affected parameter values, calculates the parameter values ​​of each affected entity at a future specified time point in time, summarizes all affected entities, merges the affected parameter values ​​from multiple sources for the same entity into the parameter change range of the entity, and outputs a list of adjacent spatial entity identifiers that may be affected in the future time period and the corresponding parameter change range of each entity.

4. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 3, characterized in that, Network topology building components, including: The dependency direction identification subcomponent is used to traverse each connection edge, identify the direction attribute corresponding to its dependency type, mark edges with positive direction attribute as pointing from the start point to the end point, and mark edges with negative direction attribute as pointing from the end point to the start point; based on the direction mark of each edge, perform topological sorting on the entity nodes on the same dependency path to generate a node sequence arranged according to the dependency flow direction. The directed path generation sub-component connects the nodes in each node sequence in order to form a directed path from the starting entity through intermediate entities to the ending entity; it detects whether there are shared nodes between different paths; if there are shared nodes, it merges multiple paths at the shared nodes to form a branch structure; and it outputs the intermediate topology structure containing all nodes and the merged path branches. The risk transmission network output sub-component is used to extract all entity nodes and directed connections between nodes in the intermediate topology, instantiate the nodes and connections into network nodes and directed edges, combine all the instantiated nodes and edges into a complete risk transmission network, and output the risk transmission network structure.

5. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 4, characterized in that, Directed path generates child components, including: The path association module is used to traverse all node sequences, extract the node identifiers in each node sequence, and generate a list of node identifiers for each path; perform an intersection operation on the node identifier lists of all paths, detect the node identifiers that appear in two or more paths, and mark the nodes as shared nodes; at the same time, it records the original path identifier of each shared node and its position number in the path. Output a set of shared node information; The path reorganization module is used to, for each shared node, truncate all original paths passing through the shared node at the node according to its associated path identifier list, and generate subsequent path segments starting from the shared node; connect multiple subsequent path segments with the same shared node in parallel to the shared node to form a structure in which multiple sub-paths branch from the shared node; after all shared nodes have been processed, the module outputs an intermediate topology structure with branches, which is composed of original path segments and branch sub-paths. The node deduplication module is used to traverse all path segments in the intermediate topology and detect whether the same node appears repeatedly in different path segments. If duplicates exist, an instance of the shared node is retained, and all upstream connections pointing to the shared node are merged. At the same time, all downstream branches leading from the shared node are connected to the retained instance. After deduplication and connection merging are completed, the final intermediate topology containing all nodes and merged path branches is output.

6. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 5, characterized in that, The path reorganization module includes: The shared node branch path extraction submodule is used to filter all subsequent path segments starting from the shared node from the original path set based on the shared node identifier; each subsequent path segment contains a continuous sequence of nodes from the next node of the shared node to the end of the path and the connection edge type between the nodes; the extracted subsequent path segments are grouped according to the shared node identifier to generate a subsequent path segment group corresponding to each shared node; the output is a set of data to be connected containing the shared node identifier and its corresponding subsequent path segment group. The spatial topology construction submodule is used to locate the spatial entity corresponding to each shared node in the knowledge graph of the confined space in mining, obtain the exit direction and connectable entity types of the spatial entity; match the first node of each segment in the subsequent path segment group with the corresponding exit of the shared node, and connect the first node of the segment to the specified exit of the shared node according to the matching result, forming multiple branch paths starting from the shared node; at the same time, it records the entity sequence traversed by each branch path; and outputs a preliminary branch topology structure containing the shared node identifier and the attached branch paths. The conflict resolution submodule is used to traverse each branch path and check whether the connection edge type between adjacent nodes in the branch path is consistent with the ventilation dependency or electromechanical dependency relationship defined in the mine enclosed space entity model. The system detects whether different branch paths intersect or overlap in the same spatial region. If there is an intersection and the path directions conflict, one path is retained according to the priority of the dependency relationship, and the remaining paths are adjusted to connect to other exits of the shared node. After the verification and resolution are completed, the final branch topology structure containing all shared nodes and multiple sub-paths branching from each node is output.

7. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 6, characterized in that, The spatial topology construction submodule includes: The first node feature extraction unit is used to traverse each subsequent path segment group, extract the node identifier of the first node from each subsequent path segment; query the attribute of the spatial entity corresponding to the node in the knowledge graph of mine confined space based on the node identifier; combine the orientation information of each first node with the entity type to generate the connection feature identifier of the subsequent path segment; and output the matching feature set. The exit feature parsing unit is used to receive the matching feature set and locate the spatial entity corresponding to the shared node in the mine enclosed space knowledge graph based on the shared node identifier contained therein; access the exit attribute list of the spatial entity; and combine the exit direction with the range of connectable entity types to generate the exit feature identifier set of the shared node. Output export characteristic information; The exit matching and binding unit is used to compare the connection feature identifier of each subsequent path segment with each exit feature identifier in the set of exit feature identifiers of the shared node one by one; when the orientation information and entity type in the connection feature identifier completely match the exit direction and the range of connectable entity types in a certain exit feature identifier, the first node of the subsequent path segment is bound to the exit identifier; after all matching is completed, the matching result set is output.

8. The knowledge graph-based multi-parameter big data monitoring system for confined spaces in mines as described in claim 7, characterized in that, Export matching and mounting unit, including: The connection feature standardization encoding subunit is used to call the direction encoding table in the mine confined space knowledge graph to map the orientation information to standard direction codes; call the entity type encoding table to map the entity type to standard type codes; and concatenate the standard direction codes and standard type codes to generate a standardized connection feature string. Output a standardized feature set; The export feature matching domain construction subunit is used to convert the export direction into a standard direction code and decompose the range of connectable entity types into a list of standard type codes that allow connection. Construct an exit matching domain dictionary using the standard direction code as the key and the list of allowed connection type codes as the value; Output matching field information; The feature matching and exit binding subunit is used to traverse the standardized feature set, extract the direction code and type code of each entry, retrieve the corresponding list of allowed connection type codes in the exit matching domain dictionary with the direction code as the key, check whether the type code exists in the type code list, and if it exists, bind the subsequent path segment identifier with the exit identifier corresponding to the direction code, and output the matching result set.

Images