Coal mine safety risk intelligent management and control method, device, equipment and medium

Abstract

The invention relates to a coal mine safety risk intelligent management and control method, device and equipment and a medium. The method comprises the following steps: constructing a multi-source heterogeneous coal mine safety data system according to received data of a target coal mine park; the multi-source heterogeneous coal mine safety data system comprises static structure data, dynamic environment data, personnel behavior data and management data; constructing a coal mine three-dimensional space model according to the static structure data and the dynamic environment data; risk indexes in the dynamic environment data are extracted based on multi-algorithm fusion for evaluation, and a risk level is obtained; and if the risk level reaches a preset threshold value, triggering a corresponding linkage response mechanism, and forming a visual result in the coal mine three-dimensional space model. By the adoption of the method, closed-loop logic from sensing, evaluation to linkage treatment can be achieved, and the real-time performance, predictability and controllability of coal mine safety management are effectively improved through algorithm support of each stage and fine design of implementation details.

Description

Technical Field

[0001] The present invention belongs to the field of risk protection, and in particular relates to a method, device, equipment and medium for intelligent management and control of coal mine safety risks. Background Art

[0002] With the development of information and automation technology, coal mine safety management technology has emerged. Based on sensor data collection and big data analysis, it has initially realized the digital presentation of mine operating status and the expression of some risk information.

[0003] Traditional safety risk assessments are typically based on manual inspection records, empirical scoring methods, or historical models based on accident level statistics. Hidden danger reporting and handling processes rely on manual submission and departmental transfers, lacking oversight of closed-loop rectification processes. Personnel behavior management is primarily conducted through on-site supervision and post-event video review, resulting in delayed responses to abnormal behavior. Equipment status assessments rely on scheduled manual inspections, making it difficult to promptly identify operational anomalies or potential failures.

[0004] The above method also has defects such as data fragmentation and inconsistent interfaces between multiple systems, which leads to serious information island phenomenon and difficulty in forming coordinated linkage. The risk perception method is mainly based on manual experience, and there is a lack of modeled and dynamic early warning mechanisms, making it difficult to detect potential risks in a timely manner. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, device, equipment and medium for intelligent management of coal mine safety risks that can integrate multi-source data for intelligent analysis and visual management to address the above technical problems.

[0006] In the first aspect, the present application provides a method for intelligent management and control of coal mine safety risks, including:

[0007] Construct a multi-source heterogeneous coal mine safety data system based on the received data from the target coal mine park; the multi-source heterogeneous coal mine safety data system includes static structure data, dynamic environment data, personnel behavior data, and management data;

[0008] Construct a three-dimensional spatial model of a coal mine based on static structural data and dynamic environmental data;

[0009] Based on the fusion of multiple algorithms, risk indicators in dynamic environmental data are extracted and evaluated to obtain the risk level;

[0010] If the risk level reaches the preset threshold, the corresponding linkage response mechanism will be triggered, and the visual results will be formed in the three-dimensional spatial model of the coal mine.

[0011] In one embodiment, a multi-source heterogeneous coal mine safety data system is constructed based on the received data of the target coal mine park, including:

[0012] In response to the received measurement data, architectural drawing data, and equipment data of the target coal mine park, the structural data corresponding to the basic space is constructed to obtain static structural data; the basic space includes mine tunnels, working faces, equipment areas, and refuge chambers;

[0013] In response to the real-time monitoring data collected by the sensor network, the real-time monitoring data in different formats are standardized to obtain dynamic environmental data. The standardization process corresponds to format unification, storage redundancy, noise reduction and filtering, and missing data completion. The dynamic environmental data includes specific gas concentrations, electromechanical data, water data, roof delamination, and dust concentration.

[0014] In response to the data collected from the work order system, positioning system and individual equipment, personnel behavior data is obtained; personnel behavior data includes personnel operation records, trajectory information, operation process execution data and pre-shift inspection feedback;

[0015] In response to the received job standard procedures, historical accident files, safety training records, safety credit scores and emergency plan texts, a management knowledge graph is formed through text parsing and structured processing to obtain management data;

[0016] A multi-source heterogeneous coal mine safety data system is constructed based on static structural data, dynamic environmental data, personnel behavior data and management data.

[0017] In one embodiment, constructing a three-dimensional spatial model of a coal mine based on static structural data and dynamic environmental data includes:

[0018] Construct a 3D model of the coal mine using static structural data based on a 3D modeling engine;

[0019] The dynamic environment data mapping is bound to the spatial coordinates corresponding to the coal mine three-dimensional model to obtain the coal mine three-dimensional spatial model; the coal mine three-dimensional spatial model includes multiple monitoring points, road nodes, and areas.

[0020] In one embodiment, risk indicators are extracted from dynamic environmental data based on multi-algorithm fusion and evaluated to obtain risk levels, including:

[0021] Normalize the dynamic environmental data corresponding to the same monitoring point, extract risk indicators, and generate multiple dimensionless risk factor values;

[0022] Construct a multi-source risk indicator vector based on multiple risk factor values;

[0023] Use the time series prediction model to predict the trend of multi-source risk indicator vectors within a preset time period to obtain future risk change values;

[0024] Combining the future risk change value and the current risk factor value to form a risk feature set;

[0025] The classification model is used to perform level mapping on the risk feature set to obtain the risk level; the risk level includes low risk, medium risk, high risk and emergency.

[0026] In one embodiment, if the risk level reaches a preset threshold, a corresponding linkage response mechanism is triggered, and a visualization result is formed in the three-dimensional spatial model of the coal mine, including:

[0027] Activate broadcast warnings and video surveillance through a coordinated response mechanism;

[0028] Based on the preset risk level color display strategy, the corresponding areas of the coal mine 3D spatial model are highlighted according to the risk levels of different monitoring points;

[0029] According to the current risk type, the coal mine three-dimensional spatial model is used to simulate the evolution of emergency scenarios and obtain the disaster evolution results;

[0030] Risks are marked on road nodes based on the results of disaster evolution, and evacuation routes are marked on the three-dimensional spatial model of the coal mine based on the shortest path and low-risk priority principles;

[0031] Generate evacuation prompt push instructions and emergency work order push instructions based on the disaster evacuation route; the evacuation prompt push instructions are used to instruct the push of personnel evacuation information and disaster evacuation routes to the personnel terminals in the coal mine based on management data; the emergency work order push instructions are used to instruct the push of emergency work orders to relevant positions and on-duty personnel.

[0032] In one embodiment, the method further comprises:

[0033] Based on the multi-source heterogeneous coal mine safety data system, the behavior recognition model is used to identify abnormal behavior of personnel in the monitoring video frames, and the abnormal behavior recognition results are obtained; the abnormal behavior recognition results include personnel and abnormal points;

[0034] Build a relationship network between abnormal behavior recognition results and events based on the event-responsible person-deduction mapping rule table, and adjust the security integrity score according to the preset weight;

[0035] Build a personnel profile based on abnormal behavior identification results and security integrity scores;

[0036] Based on the content recommendation algorithm, the standard job process is pushed to the personnel terminal corresponding to the personnel portrait; the standard job process corresponds to the abnormal points.

[0037] In one embodiment, the method further comprises:

[0038] Record the incidents corresponding to the risk level that reaches the preset threshold into the historical incident archive;

[0039] Overlay risk levels, personnel behavior data, and historical accident records onto the three-dimensional spatial model of the coal mine in the form of layers;

[0040] In response to the obtained interactive operation query instruction, visually present the attributes and historical trajectory of any object in the three-dimensional spatial model of the coal mine;

[0041] Generate regional analysis reports based on risk levels and dynamic environmental data at preset time intervals; regional analysis reports include trend charts and risk level distribution tables.

[0042] Secondly, the present application also provides a coal mine safety risk intelligent management and control device, including:

[0043] A perception module is used to build a multi-source heterogeneous coal mine safety data system based on the data received from the target coal mine park;

[0044] Model building module, used to build a three-dimensional spatial model of the coal mine based on static structural data and dynamic environmental data;

[0045] The prediction module is used to extract risk indicators from dynamic environmental data based on multi-algorithm fusion and evaluate them to obtain risk levels;

[0046] The response module is used to trigger the corresponding linkage response mechanism if the risk level reaches the preset threshold, and form a visual result in the three-dimensional spatial model of the coal mine.

[0047] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above-mentioned methods for intelligent management and control of coal mine safety risks when executing the computer program.

[0048] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of any of the above-mentioned coal mine safety risk intelligent management and control methods are implemented.

[0049] The above-mentioned intelligent coal mine safety risk management and control method, device, equipment and medium are based on data fusion, three-dimensional modeling and intelligent algorithms. Through an end-to-end risk modeling and response mechanism, it realizes a closed-loop logic from perception, assessment to coordinated disposal. Through the careful design of algorithm support and implementation details at each stage, it can effectively improve the real-time, predictive and controllable nature of coal mine safety management. The fusion of multi-algorithm risk assessment mechanism can effectively improve the accuracy and timeliness of risk identification, and cooperate with the coordinated response mechanism to form a complete closed loop from identification to intervention. Using the three-dimensional model as the visualization carrier, it can realize the accurate presentation of risk information in the real structural space, enhancing the user operability and information interpretability of the system. Through intelligent risk identification and automatic response mechanism, the pressure of emergency response is reduced, human resource allocation is optimized, and the overall inherent safety management level of coal mines is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0051] Figure 1 Schematic diagram of the process of the intelligent management and control method of coal mine safety risks of the present invention;

[0052] Figure 2 Schematic diagram of the step-by-step process of step S103;

[0053] Figure 3 Schematic diagram of the step-by-step process of step S104;

[0054] Figure 4 This is a structural diagram of the coal mine safety risk intelligent management and control device of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0056] In one embodiment, Figure 1 As shown, a method for intelligent management and control of coal mine safety risks is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and implemented through the interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0057] S101. Construct a multi-source heterogeneous coal mine safety data system based on the received data of the target coal mine park; the multi-source heterogeneous coal mine safety data system includes static structure data, dynamic environment data, personnel behavior data and management data.

[0058] Multi-source heterogeneity refers to data with diverse sources and types, including but not limited to structural, environmental, behavioral and management dimensions. Specifically, static structural data mainly includes relatively stable engineering drawings and GIS information such as mining area geographic information, underground tunnel topology, equipment layout diagrams, and work area divisions; dynamic environmental data covers real-time indicators such as gas concentration, carbon monoxide content, ventilation flow rate, temperature and humidity, and motor current that are continuously collected through sensor networks; personnel behavior data comes from personnel positioning systems, camera behavior recognition modules, access control and clock-in records, etc., reflecting the spatial activity trajectory and safety behavior norms of operators; management data includes hidden danger inspection records, safety training history, accident records, rectification work orders and other event-related information from the management system. Furthermore, cross-system semantic fusion is achieved through data standardization, structured modeling, and timestamp alignment.

[0059] S102. Construct a three-dimensional spatial model of the coal mine based on the static structural data and the dynamic environmental data.

[0060] The three-dimensional spatial model of the coal mine assumes the spatial presentation function of the physical scene, as well as the subsequent visual expression of risks and positioning of the linkage mechanism. Schematically, three-dimensional modeling platforms such as Unity3D or Cesium can be used, combined with BIM modeling rules and GIS coordinate systems, to construct tunnels, shafts, mining faces, cable layouts, etc. as digital spatial objects, and bind them with real-time data to form a perceptible model. To achieve dynamic perception, the sensor collection values can be mapped to the three-dimensional model object properties through real-time communication protocols such as WebSocket or MQTT (message queue telemetry transmission), realizing functions such as local discoloration of the model when the gas concentration is abnormal, and real-time animation of ventilation flow. The construction of the spatial model not only improves the intuitiveness of data display, but also provides a scenario basis for subsequent intelligent decision-making such as risk positioning and path deduction.

[0061] S103. Risk indicators in dynamic environmental data are extracted based on multi-algorithm fusion and evaluated to obtain a risk level.

[0062] Risk indicators refer to a collection of variables highly correlated with accident precursors, such as excessive gas levels, ventilation anomalies, abnormal equipment vibration, sudden current changes, and decreased oxygen concentration. Schematically, a multi-algorithm fusion strategy is employed to jointly model risk indicators, improving the accuracy and robustness of risk perception. Specifically, this multi-algorithm fusion approach includes two processing paths: prediction using a time series prediction model and grading using a classification algorithm. A deep neural network based on LSTM (Long Short-Term Memory) networks can be used to predict short-term trends in environmental variables, identifying potential hazardous developments in advance. The classification algorithm can be a rule-based risk threshold determination, anomaly detection, or classification model to qualitatively assess whether the current state poses a safety hazard. For example, an LSTM model models historical gas concentration series, outputting a predicted value for the next 10 minutes and comparing it to an upper limit. If the value exceeds the 95% confidence interval, a primary risk is triggered. Furthermore, a multi-factor normalization analysis model is introduced to convert indicators of different dimensions into comparable scales, constructing a multidimensional risk vector, and calculating the overall risk level using a weighted scoring method.

[0063] S104. If the risk level reaches a preset threshold, a corresponding linkage response mechanism is triggered, and a visualization result is formed in the three-dimensional spatial model of the coal mine.

[0064] The linkage response calls a series of preset response strategy libraries based on the mapping relationship between the risk level and the trigger position in the three-dimensional model. Indicatively, when the gas concentration exceeds the standard and the predicted value shows a continuous upward trend, an evacuation route instruction can be automatically generated, and a voice warning can be issued through the underground broadcast module. At the same time, the linkage dispatch monitoring system switches the corresponding camera image to the dispatch center, and generates an electronic work order for the closed mining area and sends it to the emergency repair personnel. For medium-level risks, it may only trigger the on-duty confirmation mechanism, or push the inspection task to the mobile terminal. In all response processes, the spatial location of the risk event, high-risk indicator curve, response measures and other information are superimposed in real time on the three-dimensional space model to form an intuitive and operational dynamic visualization interface, thereby assisting managers to make intervention decisions quickly.

[0065] For example, a gas concentration sensor in a tunneling face area within a mine exhibited a fluctuating upward trend for 30 consecutive minutes. The LSTM model predicted that it would exceed the upper threshold of 1.0% in 5 minutes, and the carbon monoxide concentration in the local area also showed an abnormal increase. In this scenario, the data fusion model identified the area as medium-to-high risk, triggering a spatial warning annotation within the 3D model. A multi-indicator evolution chart for the area was automatically pushed to the monitoring and dispatching terminal. Simultaneously, the mine broadcast system was activated to announce that operations in the area needed to be suspended, and an inspection task was generated and sent to the nearest team. If the personnel positioning system also detected that personnel were still working, the risk level was upgraded to high, triggering a forced offline instruction and a mobile terminal prompt, enabling a rapid response.

[0066] The aforementioned intelligent coal mine safety risk management method incorporates static structural data, dynamic environmental data, human behavior data, and management data to form a data system covering multiple dimensions, including spatial structure, environmental evolution, human operations, and management systems. This enables risk perception to transcend the limitations of single sensor data and instead incorporates three-dimensional perception and cross-domain integration, significantly improving the integrity and accuracy of risk factor identification. The three-dimensional model constructed using static structural data not only recreates the actual structural layout of the coal mine complex but also, by mapping dynamic environmental data to model spatial coordinates, enables spatial localization and dynamic visualization of risk factors, enabling managers to understand the spatial distribution and evolution of risks. The collaborative use of multiple algorithms to extract and predict the evolutionary trends of risk factors from dynamic environmental data effectively mitigates data fluctuations and sensor noise, improving the robustness of risk assessment and trend identification, and enabling proactive early warning. By setting response thresholds for risk levels, once the warning threshold is reached, subsystems such as the broadcast system, video surveillance, and emergency response modules are automatically activated, enabling proactive intervention, improving the efficiency of accident prevention and initial response, and reducing the risk of human misjudgment and delayed response. The risk level is visualized in the three-dimensional spatial model of the coal mine. Combined with disaster evolution simulation and risk avoidance path planning, managers can grasp the current mining situation more intuitively and clearly, thereby formulating response measures or conducting personnel dispatch more scientifically.

[0067] In one embodiment, a multi-source heterogeneous coal mine safety data system is constructed based on the received data of the target coal mine park, including:

[0068] S11. In response to the received measurement data, architectural drawing data, and equipment data of the target coal mine park, construct structural data corresponding to the basic space to obtain static structural data; the basic space includes mine tunnels, working faces, equipment areas, and refuge chambers.

[0069] Schematically, architectural drawing data can come from CAD (Computer Aided Design) files or BIM (Building Information Modeling) models, measurement data can come from laser scanning or RTK (Real-Time Kinematic) mapping results, and equipment data can come from installation archives or construction records. Using algorithms such as primitive recognition, spatial registration, and topological modeling, the drawings and measurement data are spatially reconstructed to generate an interactive digital model of the underground spatial structure. This static structural data not only carries physical layout information but also provides a positioning reference and object association basis for subsequent dynamic data binding.

[0070] S12. In response to the real-time monitoring data collected by the sensor network, the real-time monitoring data in different formats are standardized to obtain dynamic environmental data; the standardization processing corresponds to format uniform processing, storage redundancy processing, noise reduction filtering processing and missing completion processing; the dynamic environmental data includes specific gas concentration, electromechanical data, water data, roof delamination and dust concentration.

[0071] The real-time monitoring data collected by the sensor network deployed at each monitoring node underground covers high-frequency data channels such as gas concentration, carbon monoxide concentration, wind speed, wind pressure, motor current, water level, roof separation and dust concentration. Since each sensor may come from different manufacturers and use different data protocols and sampling periods, the raw data needs to be standardized to unify the semantics and structure to ensure computability and fusion. The standardization process includes unifying the format of multi-source data, compressing and removing duplicates in units of time windows, that is, storing redundant data, using sliding average, wavelet denoising or Kalman filtering to suppress signal noise interference and perform noise reduction filtering, and combining historical trends with spatial neighboring node predictions to fill in missing values for missing completion processing. After multiple processing, a data structure with high timeliness and high availability is formed, namely dynamic environmental data, which provides a reliable foundation for subsequent risk calculations.

[0072] S13. In response to the data collected from the work order system, positioning system, and individual equipment, personnel behavior data is obtained; the personnel behavior data includes personnel operation records, trajectory information, operation process execution data, and pre-shift inspection feedback.

[0073] Personnel behavior data mainly comes from the work records of the work order system, the trajectory data generated by the underground personnel positioning system, the operation process records sent back by the individual operation terminal, and the safety inspection feedback before the shift. Schematically, through semantic annotation and timestamp alignment, the scattered data are uniformly modeled as a sequence of behavioral events. Each record includes fields such as the behavior subject, behavior type, occurrence area, associated facilities, and execution status. For example, a team member performs a gas inspection process at the excavation working face. This behavior is recorded as a process item in the work order system, a spatial trajectory is generated in the positioning system, and photo upload information is included in the individual terminal. Furthermore, the above content is unified and aggregated into a complete behavior record to ensure that the behavior data has temporal integrity, spatial clarity, and clarity of operation objectives, providing a basis for subsequent behavior compliance analysis and safety behavior identification.

[0074] S14. In response to the received job standard procedures, historical accident files, safety training records, safety integrity scores and emergency plan texts, a management knowledge graph is formed through text parsing and structured processing to obtain management data.

[0075] For the analysis and conversion of unstructured text information, in response to received documents such as job standard procedures, historical accident archives, safety training records, safety credit scores, and emergency plan texts, natural language processing technologies such as BERT (General Semantic Representation Model) encoders and text dependency parsers are used to complete the semantic parsing and structural modeling of the text. Specifically, through mechanisms such as keyword extraction, entity relationship recognition, and event abstraction, the original text is converted into a management knowledge graph. The nodes in the graph represent safety entities, including positions, types of work, risk sources, and response measures. The edges represent relationships such as position-needed execution-operation process, hidden danger-cause-accident type, etc., making management data computable.

[0076] S15. Construct a multi-source heterogeneous coal mine safety data system based on static structural data, dynamic environmental data, personnel behavior data and management data.

[0077] Integrate static structural data, dynamic environmental data, personnel behavior data and management data to build a multi-source heterogeneous coal mine safety data system covering the four dimensions of time and space, human factors, physics and management. In principle, integration is carried out through a unified data indexing mechanism and data warehouse modeling method to ensure that different types of data are semantically consistent, temporally aligned, and spatially projected. Specifically, when a gas sensor anomaly occurs in a certain working face equipment area, the structural information of the area, the corresponding personnel behavior trajectory, the most recent inspection record and the emergency plan can be quickly associated to support multi-angle risk reasoning and precise response, thereby breaking the traditional coal mine system data island pattern and providing full data support for subsequent intelligent perception, three-dimensional mapping and risk assessment.

[0078] In one embodiment, constructing a three-dimensional spatial model of a coal mine based on static structural data and dynamic environmental data includes:

[0079] S21. Construct a three-dimensional model of a coal mine using static structural data based on a three-dimensional modeling engine.

[0080] Schematically, the 3D modeling process uses industrial-grade modeling platforms such as Unity 3D, Unreal Engine or OpenSceneGraph as the underlying graphics rendering engine, and combines the attribute parameters such as tunnel direction, equipment layout, and spatial dimensions in the static structural data to initialize the model and reconstruct the geometry. Furthermore, in order to achieve the fineness of the model and the ability to connect with data, the BIM methodology will be introduced in the modeling process. Through object-oriented component modeling, different spatial structures will be refined into interactive modular units, including working faces, ventilation shafts, main transport tunnels, return air tunnels, etc., and unique spatial identifiers, topological relationships and logical attributes are given to make them searchable and data-carrying. In terms of spatial reconstruction algorithms, a voxel modeling method based on a spatial skeleton is adopted. Guided by the plane data of the mine map, the solid geometry is reconstructed through cross-section contour fitting and three-dimensional stretching operations, thereby realizing the spatial splicing of complex tunnel networks.

[0081] S22. Bind the dynamic environment data mapping to the spatial coordinates corresponding to the coal mine three-dimensional model to obtain the coal mine three-dimensional spatial model; the coal mine three-dimensional spatial model includes multiple monitoring points, road nodes, and areas.

[0082] Furthermore, the dynamic environmental data is mapped and bound to the corresponding spatial coordinates in the above-mentioned three-dimensional model to form a dynamically driven three-dimensional spatial model of the coal mine. Specifically, based on the monitoring equipment number, installation location and spatial label contained in each dynamic data, the spatial index of the corresponding structural unit in the three-dimensional model is retrieved, and a data-position mapping relationship is established. The binding process can be achieved through a triple mechanism of spatial coordinate alignment, semantic matching and topological verification: Among them, spatial coordinate alignment refers to the completion of spatial anchoring directly based on the three-dimensional positioning data such as the total station and UWB (ultra-wideband positioning system) ranging results recorded when the sensor is deployed; semantic matching is used to identify the consistency of the entity name and equipment code in the dynamic data and the three-dimensional model; topological verification is used to ensure that the mapping result is consistent with the structural topology logic to avoid binding errors or overlapping conflicts.

[0083] This creates a three-dimensional spatial data field containing multidimensional monitoring information, where each spatial node can host a set of time-series data or status information. The three-dimensional coal mine spatial model defines multiple key units, including monitoring points, road nodes, and regional units. Monitoring points can be the locations of real-time sensors for gas, carbon monoxide, water level, temperature, and other indicators. Road nodes correspond to the connection points and intersections of ventilation arteries, surface routes, and transportation routes. Regional units can be functional areas such as excavation areas, mining areas, equipment installation areas, and refuge chambers.

[0084] This 3D spatial model of a coal mine integrates structure, data, and events, supporting subsequent multidimensional analysis and interactive visualization. For example, when dust concentration in a certain area exceeds a threshold, the area is automatically highlighted in the 3D scene, and the associated personnel trajectories, equipment status, and historical risk records are displayed, enabling spatial visibility of risk tracing and coordinated early warning.

[0085] In one embodiment, Figure 2 As shown in the figure, risk indicators in dynamic environmental data are extracted based on multi-algorithm fusion for evaluation, and risk levels are obtained, including:

[0086] S201. Normalize the dynamic environmental data corresponding to the same monitoring point, extract risk indicators, and generate multiple dimensionless risk factor values.

[0087] Schematically, since different types of environmental monitoring data have different dimensions and value ranges, direct comparison or fusion can easily lead to bias imbalance. Therefore, normalization processing uses dimensionless standard conversion formulas such as Z-score normalization or min-max normalization to uniformly map the original indicators to [0,1] or the standard normal distribution space, so that each risk factor has an equal weight basis. After normalization, according to the predefined risk indicator extraction rules, representative and highly sensitive risk factors are extracted from each type of data. For example, the gas concentration change rate, the frequency of dust exceeding the standard, the sudden increase in water level, the duration of voltage anomaly, etc. can all be used as risk factor values.

[0088] S202. Construct a multi-source risk indicator vector based on multiple risk factor values.

[0089] Schematically, the multi-source risk indicator vector is constructed based on the synchronization alignment mechanism of the monitoring point spatial position and the monitoring time window, ensuring that the data from different sensors can be uniformly described at the same time, forming a multi-dimensional feature vector reflecting the environmental status of the time-space point. Specifically, at a certain moment, the gas, carbon monoxide, dust, wind speed, temperature, water level and other indicators deployed on a certain excavation working face are normalized to form a vector of length n, denoted as X(t) = [x1(t), x2(t), ..., x n(t)].

[0090] S203: Use a time series prediction model to perform trend prediction on the multi-source risk indicator vector within a preset time period to obtain a future risk change value.

[0091] Furthermore, multi-source risk indicator vectors are used as input samples, and a time series forecasting model is used for trend modeling. Time series forecasting models can take various forms, including traditional ARIMA (Autoregressive Integrated Moving Average Model) models and exponential smoothing models. Deep learning structures such as LSTM, GRU (Gated Recurrent Unit), or Transformer structures can also be used to achieve nonlinear multidimensional prediction. For example, an LSTM network is used as the training input, with the risk indicator sequence at time T in the past and the output being the predicted value of the risk factor at time Δt in the future, forming the prediction vector X^(t+Δt).

[0092] S204. Combining the future risk change value and the current risk factor value to form a risk feature set.

[0093] The current risk factor value X(t) and the risk change value X^(t+Δt) at the predicted time are combined to form the risk feature set F(t), i.e., F(t) = [X(t), X^(t+Δt)]. This risk feature set combines information about the current static state and future dynamic trends. Optionally, this feature set not only includes the main variable value but also includes additional temporal context features, such as night shift period and working face operation time, to enhance the model's context recognition capabilities.

[0094] S205. Use the classification model to perform level mapping on the risk feature set to obtain a risk level; the risk level includes low risk, medium risk, high risk and emergency.

[0095] The constructed risk feature set F(t) is input into the trained risk level classification model to output a discretized risk level label. In principle, the classification model can be implemented in a variety of forms, including traditional machine learning methods such as random forests, support vector machines, and gradient boosting trees. Deep neural networks can also be used to achieve higher-dimensional nonlinear mapping. The model training phase integrates historical accident data, manually labeled level data, and expert experience rules to construct training samples, realizing the automatic mapping of risk levels from feature vector space to discrete labels. Discrete labels can be low risk, medium risk, high risk, and emergency. Optionally, the risk level division can refer to the key indicator thresholds specified in the coal mine safety regulations or the enterprise's customized multi-level linkage standards.

[0096] Furthermore, to improve the robustness and interpretability of the model, an integrated learning framework can be introduced, that is, using voting or stacking strategies to fuse multiple model results to improve classification accuracy; at the same time, explanation tools such as SHAP values or LIME can be introduced to factor the classification results to assist experts in understanding the model output.

[0097] In one embodiment, Figure 3 As shown in the figure, if the risk level reaches the preset threshold, the corresponding linkage response mechanism is triggered and a visualization result is formed in the three-dimensional spatial model of the coal mine, including:

[0098] S301. Activate broadcast warning and video surveillance through the linkage response mechanism.

[0099] Once it is determined that the risk level of one or more monitoring points reaches a preset threshold, such as a high risk or emergency level, the linkage response mechanism is triggered, and active response measures including broadcast warnings and video surveillance are automatically initiated. The broadcast warning module connects to the underground broadcast system and accurately pushes warning voices by area, such as "The gas concentration at the working face exceeds the standard, please evacuate immediately", to ensure that the workers in the relevant areas are covered in the shortest time; while the video surveillance module automatically dispatches camera resources in the dispatch center or key areas of the mine, and links the monitoring screen in real time to assist the on-duty personnel in conducting on-site situation analysis. Optionally, multi-level linkage automated scheduling can be achieved by establishing a mapping rule table between risk levels and response actions. For example, high-risk linkage local broadcasts and central control notifications, and emergency levels linkage mine-wide broadcasts and video pan-tilt tracking.

[0100] S302: Based on a preset risk level color display strategy, highlight the corresponding areas in the three-dimensional spatial model of the coal mine according to the risk levels of different monitoring points.

[0101] Based on the preset risk level color display strategy, the areas where monitoring points with risk levels are located are highlighted in color in the three-dimensional spatial model of the coal mine. Schematically, the strategy follows a visual gradient mapping standard from green low risk to red emergency, allowing managers to identify high-risk areas at a glance in three-dimensional space. For example, if the dust concentration and gas concentration of a monitoring point both reach a high risk level, the corresponding working face area will be displayed as orange highlight in the three-dimensional model, accompanied by dynamic flashing to enhance the attention-awakening effect. Furthermore, in order to enhance the spatial perception effect, a risk factor change layer or risk level label text can be superimposed to achieve risk space explicitness.

[0102] S303. Call the three-dimensional spatial model of the coal mine according to the current risk type to simulate the evolution of the emergency scenario and obtain the disaster evolution result.

[0103] To assist management personnel in predicting the path of disaster spread and developing effective response measures, a three-dimensional spatial model of a coal mine is used to simulate the evolution of emergency scenarios based on the current risk type. This simulation process can be based on finite element calculations, physical modeling, or empirical models based on the evolution of historical accidents, combined with current environmental variables for parameter injection to achieve a dynamic demonstration of the risk field expansion trend. For example, in a scenario where gas exceeds the standard, the direction and speed of gas accumulation and diffusion can be deduced based on local wind speed, pressure distribution, and spatial structure; in the case of abnormal roof delamination, the subsidence of the rock structure and the scope of secondary collapse can be simulated. The simulation results will be superimposed on the three-dimensional spatial model as a dynamic layer, forming a visual evolution trajectory of the risk flow path, helping emergency decision makers grasp the rhythm of the situation development.

[0104] S304. Mark the road nodes for risk according to the disaster evolution results, and mark the disaster avoidance routes in the three-dimensional spatial model of the coal mine based on the shortest path and low-risk priority principles.

[0105] Schematically, based on the disaster evolution results obtained through simulation, each road node in the three-dimensional spatial model is risk-labeled to identify high-risk areas and safe traversable paths. During the path planning phase, a dual-objective optimization strategy of shortest path and low-risk priority is introduced. Starting from the current occupant's location and aiming at the refuge chamber or escape from the mine, a weighted graph search algorithm, such as a modified Dijkstra or A* algorithm, is used to prioritize risk avoidance by assigning negative weights to nodes or edges. The labeled results are displayed in real time on the three-dimensional spatial model as guide arrows, and can be combined with acoustic and optical navigation nodes to form a guidance path.

[0106] S305. Generate an evacuation prompt push instruction and an emergency work order push instruction based on the disaster avoidance route; the evacuation prompt push instruction is used to instruct the push of personnel evacuation information and disaster avoidance routes to the personnel terminals in the coal mine based on management data; the emergency work order push instruction is used to instruct the push of emergency work orders to relevant positions and on-duty personnel.

[0107] In principle, the evacuation prompt push command, through linkage with the personnel positioning system and the management knowledge graph, can accurately identify workers in dangerous areas. Based on their job roles, operating permissions, and accessibility, the evacuation route is pushed to their individual terminals, mining lamp screens, or underground broadcast systems in text, graphic, or voice format. For example, a prompt message can be pushed to a tunneling worker's terminal: "Please evacuate immediately, proceed along Channel 3, and turn left at the first fork to reach Refuge Chamber B." Simultaneously, an emergency work order push command sends task orders and emergency response suggestions to on-duty managers, inspection posts, or on-site rescue team terminals. For example, instructions include checking the startup status of Fan 5 and transmitting fault codes, starting Main Water Pump B, and preparing to seal the gates, ensuring that each emergency post initiates the response process according to its responsibility system.

[0108] In one embodiment, the method further comprises:

[0109] S41. Based on the multi-source heterogeneous coal mine safety data system, the behavior recognition model is used to identify abnormal behaviors of personnel in the monitoring video frames to obtain abnormal behavior recognition results; the abnormal behavior recognition results include personnel and abnormal points.

[0110] Based on surveillance video streams deployed at key coal mine locations, such as tunnel intersections, tunneling faces, and entrances and exits, an integrated behavior recognition model performs frame-by-frame identification and dynamic analysis of human behavior within video sequences. This behavior recognition model combines object detection algorithms based on YOLO and Faster-RCNN with action recognition algorithms based on I3D and SlowFast to accurately determine human posture, movement trajectory, and interaction objects. It can identify typical abnormal behaviors, including not wearing a helmet, illegal boundary crossing, abnormal pausing, running away, and carrying prohibited items. At the data input layer, the model integrates location information, work period tables, and job authority records from a multi-source, heterogeneous coal mine safety data system to enhance contextual semantic understanding and false detection prevention capabilities. The model outputs abnormal behavior recognition results in a structured format, including the person's identity, abnormal behavior category, abnormal occurrence time, and video frame location coordinates, thus achieving dual anchoring of both the person and the abnormal point.

[0111] S42. Build a relationship network between abnormal behavior recognition results and events based on the event-responsible person-deduction point mapping rule table, and adjust the security integrity points according to the preset weights.

[0112] Furthermore, based on the results of the identified abnormal behavior, the pre-built event-responsible person-deduction mapping rule table is called to bind the specific behavior event with the management responsibility node, forming a responsibility tracing mechanism for the behavior event. For example, if an employee is identified as not wearing protective glasses, it will be automatically classified as a minor violation - failure to comply with safety regulations, and the corresponding safety integrity points will be deducted based on the rule table, and the time and space location and responsible person information corresponding to the behavior will be recorded. Optionally, the rule table is usually stored in the form of a multi-dimensional matrix or knowledge graph, and supports dynamic adjustment of weights based on factors such as behavior type, job type, historical behavior frequency, etc., to achieve real-time maintenance and quantitative feedback of the integrity points system, and provide an objective basis for behavioral incentives and safety reward and punishment mechanisms.

[0113] S43. Build a personnel profile based on abnormal behavior identification results and security integrity scores.

[0114] The results of identifying a person's abnormal behavior are combined with their corresponding safety integrity score to construct a multi-dimensional person profile reflecting their behavioral preferences, safety habits, and risk propensity. This person profile includes not only static information such as job type, shift, and work area, but also dynamic behavioral characteristics such as the frequency of abnormal behavior, high-risk work periods, and operational compliance rates, thereby forming a set of behavioral labels that can be used for reasoning and decision support. For example, if a person has recently frequently crossed the line into a high-voltage substation and temporarily left their post, negative labels such as negligent operation and weak awareness of rules will be automatically added to their profile, and their recommended tasks or work authority level will be adjusted accordingly.

[0115] S44. Push the standard job process to the personnel terminal corresponding to the personnel portrait based on the content recommendation algorithm; the standard job process corresponds to the abnormal point.

[0116] Based on the constructed personnel portrait, the content recommendation algorithm is called to push the standard operating procedures of the positions corresponding to their abnormal points to the corresponding personnel terminals. Indicatively, the recommendation mechanism can adopt algorithm models such as collaborative filtering, content matching or reinforcement learning, and intelligently select operating instructions or training materials that are closely related to the current behavioral problems based on the matching degree between the personnel portrait label and the job operation standard library. For example, for an electrician who frequently misoperates equipment, his portrait is matched to the high-voltage equipment maintenance process standard, and the relevant operating procedures, precautions and video tutorials are automatically pushed to his portable terminal or miner's lamp screen to achieve closed-loop educational intervention and habit correction. Furthermore, the recommended content can also be combined with the pre-shift meeting or rotation training system, and regularly refreshed and pushed as part of daily management to improve the safety level and execution consistency of the overall operation team.

[0117] In one embodiment, the method further comprises:

[0118] S51. Record the accident corresponding to the risk level that reaches the preset threshold into the historical accident file.

[0119] In principle, events corresponding to risk levels that have reached a preset threshold are recorded and written to a historical accident archive database as structured accident data, enabling long-term preservation and traceability of major risk scenarios and their response processes. Accident records typically include risk level determination results, risk factor composition, monitoring point locations, risk response actions, event duration, response effectiveness assessment, and information on relevant responsible individuals. Archiving these records in the historical accident archive not only provides a basis for subsequent review but also provides high-value sample data for subsequent accident database-based model training and knowledge graph completion.

[0120] S52. Overlay the risk level, personnel behavior data, and historical accident files onto the three-dimensional spatial model of the coal mine in the form of layers.

[0121] The current risk level data, personnel behavior data, and historical accident archives are mapped to the three-dimensional spatial model of the coal mine in the form of layers, forming a spatial expression of multi-information fusion. Schematically, the layer overlay mechanism adopts a controllable layered rendering strategy to support the simultaneous presentation of information in multiple dimensions. For example, the current gas concentration level layer, the personnel behavior event thermal layer, and the historical accident-frequent area layer can be superimposed in the tunnel area of the three-dimensional spatial model of the coal mine. Through spatial overlay, not only can the risk data be expressed intuitively, but it also facilitates managers to conduct spatiotemporal correlation analysis of the historical behavior and current status of a certain area, thereby supporting regional precise management and control, key review, and resource priority allocation strategies.

[0122] S53: In response to the obtained interactive operation query instruction, visually present the attributes and historical trajectory of any object in the three-dimensional spatial model of the coal mine.

[0123] In response to interactive operation instructions initiated by the user terminal or console, it supports attribute query and historical trajectory tracing for any object such as a monitoring point, a section of tunnel, a personnel identification or equipment node in the three-dimensional space model. Through spatial positioning and data binding mechanism, it can accurately identify the unique spatiotemporal identification of the selected object, and extract its current status, risk level, affiliated team and other attribute information and its concentration curve changes, behavior frequency trajectory, movement path and other evolutionary trajectories within a specified time period from the corresponding data table. Visualization methods include attribute information floating windows, historical curve graphs, thermal path animations, etc., allowing managers to deeply understand the dynamic behavior characteristics and risk change trends of objects through natural interaction.

[0124] S54. Generate a regional analysis report based on the risk level and dynamic environmental data at preset time intervals; the regional analysis report includes a trend chart and a risk level distribution table.

[0125] At preset time intervals, i.e. daily, per shift or monthly, the risk levels and dynamic environmental data of each area are automatically summarized and analyzed to generate a structured regional analysis report. This report is usually presented in the form of a chart, including a risk level time trend chart, a risk level spatial distribution chart, and an indicator comparison table. Among them, the regional analysis report can be used as a daily operation safety assessment tool, as well as a reporting material for pre-shift meetings, a basis for monthly inspections, or a basis for reporting to supervision, to achieve multi-scenario reuse of analysis results. In addition, the report generation mechanism can be further expanded to adaptive analysis scheduling, that is, dynamically adjusting the report generation frequency and content dimensions according to historical risk change trends, thereby improving the practicality and timeliness of data analysis.

[0126] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0127] Based on the same inventive concept, the embodiments of the present application also provide a device for intelligently controlling coal mine safety risks for implementing the aforementioned method for intelligently controlling coal mine safety risks. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations in the embodiments of one or more devices for intelligently controlling coal mine safety risks provided below can be found in the limitations of the method for intelligently controlling coal mine safety risks described above and will not be repeated here.

[0128] In an exemplary embodiment, Figure 4 As shown, a coal mine safety risk intelligent management and control device is provided, comprising:

[0129] A perception module 401 is used to construct a multi-source heterogeneous coal mine safety data system based on the received data of the target coal mine park;

[0130] Model building module 402, for building a three-dimensional spatial model of the coal mine based on static structural data and dynamic environmental data;

[0131] Prediction module 403, for extracting risk indicators from dynamic environmental data based on multi-algorithm fusion and evaluating them to obtain risk levels;

[0132] The response module 404 is used to trigger the corresponding linkage response mechanism if the risk level reaches a preset threshold, and form a visual result in the three-dimensional spatial model of the coal mine.

[0133] In one embodiment, the perception module 401 is further configured to construct structural data corresponding to the basic space in response to the received measurement data, architectural drawing data, and equipment data of the target coal mine park, thereby obtaining static structural data;

[0134] The perception module 401 is further configured to, in response to the received real-time monitoring data collected by the sensor network, perform standardization processing on the real-time monitoring data in different formats to obtain dynamic environmental data;

[0135] The perception module 401 is also used to obtain personnel behavior data in response to data collected from the work order system, positioning system, and individual equipment;

[0136] The perception module 401 is also used to respond to the received job standard procedures, historical accident files, safety training records, safety credit scores and emergency plan texts, and form a management knowledge graph through text parsing and structured processing to obtain management data;

[0137] The perception module 401 is also used to construct a multi-source heterogeneous coal mine safety data system based on static structure data, dynamic environment data, personnel behavior data and management data.

[0138] In one embodiment, the model building module 402 is further configured to build a three-dimensional model of the coal mine using static structural data based on a three-dimensional modeling engine;

[0139] The model building module 402 is further used to map and bind the dynamic environment data to the spatial coordinates corresponding to the three-dimensional model of the coal mine to obtain the three-dimensional spatial model of the coal mine.

[0140] In one embodiment, it further includes:

[0141] The indicator extraction module is used to normalize the dynamic environmental data corresponding to the same monitoring point, extract risk indicators, and generate multiple dimensionless risk factor values;

[0142] A vector construction module is used to construct a multi-source risk indicator vector based on multiple risk factor values;

[0143] The prediction algorithm module is used to use the time series prediction model to predict the trend of multi-source risk indicator vectors within a preset time period and obtain the future risk change value;

[0144] Feature extraction module, used to combine future risk change values and current risk factor values to form a risk feature set;

[0145] The level mapping module is used to use the classification model to perform level mapping on the risk feature set to obtain the risk level.

[0146] In one embodiment, the response module 404 is further configured to activate broadcast warning and video surveillance through a linkage response mechanism;

[0147] The visualization module is used to highlight the corresponding areas in the three-dimensional spatial model of the coal mine according to the risk levels of different monitoring points based on the preset risk level color display strategy;

[0148] The disaster simulation module is used to call the coal mine three-dimensional spatial model to simulate the evolution of emergency scenarios according to the current risk type and obtain the disaster evolution results;

[0149] The path planning module is used to mark the risk of road nodes according to the results of disaster evolution, and mark the disaster avoidance routes in the three-dimensional spatial model of the coal mine based on the shortest path and low-risk priority principles;

[0150] The push module is used to generate evacuation prompt push instructions and emergency work order push instructions based on the disaster avoidance route.

[0151] In one embodiment, it further includes:

[0152] Anomaly detection module, which is used to identify abnormal human behavior in surveillance video frames using a behavior recognition model based on a multi-source heterogeneous coal mine safety data system;

[0153] The scoring module is used to build a relationship network between abnormal behavior recognition results and events based on the event-responsible person-deduction mapping rule table, and adjust the security integrity score according to the preset weight;

[0154] Personnel profiling module, used to construct personnel profiles based on abnormal behavior identification results and security integrity scores;

[0155] The training module is used to push standard job procedures to the personnel terminals corresponding to the personnel portraits based on the content recommendation algorithm; the standard job procedures correspond to the abnormal points.

[0156] In one embodiment, it further includes:

[0157] A historical storage module is used to record accidents corresponding to risk levels that reach a preset threshold into historical accident archives;

[0158] The visualization module is also used to overlay risk levels, personnel behavior data, and historical accident files onto the three-dimensional spatial model of the coal mine in the form of layers;

[0159] The visualization module is further used to respond to the obtained interactive operation query instruction and visualize the attributes and historical trajectory of any object in the three-dimensional spatial model of the coal mine;

[0160] The visualization module is also used to generate regional analysis reports based on risk levels and dynamic environmental data at preset time intervals; the regional analysis reports include trend charts and risk level distribution tables.

[0161] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0162] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0163] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0164] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for intelligent management and control of coal mine safety risks, characterized in that: The method comprises: Constructing a multi-source heterogeneous coal mine safety data system based on the received data of the target coal mine park; the multi-source heterogeneous coal mine safety data system includes static structure data, dynamic environment data, personnel behavior data and management data; Constructing a three-dimensional spatial model of a coal mine according to the static structural data and the dynamic environmental data; Extract risk indicators from the dynamic environmental data based on multi-algorithm fusion and evaluate them to obtain a risk level; If the risk level reaches a preset threshold, a corresponding linkage response mechanism is triggered, and a visualization result is formed in the three-dimensional spatial model of the coal mine.

2. The method according to claim 1, characterized in that The multi-source heterogeneous coal mine safety data system is constructed based on the received data of the target coal mine park, including: In response to the received measurement data, architectural drawing data, and equipment data of the target coal mine park, structural data corresponding to the basic space is constructed to obtain the static structural data; the basic space includes mine tunnels, working faces, equipment areas, and refuge chambers; In response to the received real-time monitoring data collected by the sensor network, the real-time monitoring data in different formats are standardized to obtain the dynamic environmental data; the standardization processing corresponds to format unification processing, storage redundancy processing, noise reduction filtering processing and missing completion processing; the dynamic environmental data includes specific gas concentration, electromechanical data, water data, roof delamination and dust concentration; In response to the data collected from the work order system, positioning system and individual equipment, the personnel behavior data is obtained; the personnel behavior data includes personnel operation records, trajectory information, operation process execution data and pre-shift inspection feedback; In response to the received job standard procedures, historical accident files, safety training records, safety credit scores and emergency plan texts, a management knowledge graph is formed through text parsing and structured processing to obtain the management data; The multi-source heterogeneous coal mine safety data system is constructed based on the static structure data, the dynamic environment data, the personnel behavior data and the management data.

3. The method according to claim 2, characterized in that The step of constructing a three-dimensional spatial model of a coal mine according to the static structural data and the dynamic environmental data includes: Constructing a three-dimensional model of the coal mine using the static structure data based on a three-dimensional modeling engine; The dynamic environment data mapping is bound to the spatial coordinates corresponding to the three-dimensional model of the coal mine to obtain the three-dimensional spatial model of the coal mine; the three-dimensional spatial model of the coal mine includes multiple monitoring points, road nodes, and areas.

4. The method according to claim 3, characterized in that The risk indicators extracted from the dynamic environment data based on multi-algorithm fusion are evaluated to obtain a risk level, including: Normalizing the dynamic environmental data corresponding to the same monitoring point, extracting risk indicators, and generating multiple dimensionless risk factor values; Constructing a multi-source risk indicator vector based on a plurality of risk factor values; Using a time series prediction model to perform trend prediction on the multi-source risk indicator vector within a preset time period to obtain a future risk change value; Combining the future risk change value and the current risk factor value to form a risk feature set; The risk feature set is graded and mapped using a classification model to obtain the risk grade; the risk grade includes low risk, medium risk, high risk and emergency.

5. The method according to claim 4, characterized in that If the risk level reaches a preset threshold, a corresponding linkage response mechanism is triggered, and a visualization result is formed in the three-dimensional spatial model of the coal mine, including: Activate broadcast warnings and video surveillance through a coordinated response mechanism; Based on a preset risk level color display strategy, highlight the area corresponding to the three-dimensional spatial model of the coal mine according to the risk level of different monitoring points; Calling the three-dimensional spatial model of the coal mine to simulate the evolution of sudden scenarios according to the current risk type to obtain disaster evolution results; Marking the road nodes for risk according to the disaster evolution results, and marking the disaster avoidance routes in the three-dimensional space model of the coal mine based on the shortest path and low-risk priority principles; An evacuation prompt push instruction and an emergency work order push instruction are generated according to the disaster avoidance route; the evacuation prompt push instruction is used to instruct the push of personnel evacuation information and disaster avoidance routes to the personnel terminals in the coal mine according to the management data; the emergency work order push instruction is used to instruct the push of emergency work orders to relevant posts and on-duty personnel.

6. The method according to any one of claims 2 to 5, characterized in that The method further comprises: Based on the multi-source heterogeneous coal mine safety data system, the behavior recognition model is used to identify abnormal behavior of personnel in the monitoring video frame, and abnormal behavior recognition results are obtained; the abnormal behavior recognition results include personnel and abnormal points; Constructing a relationship network between the abnormal behavior identification results and events based on the event-responsible person-deduction point mapping rule table, and adjusting the security integrity score according to the preset weight; Constructing a personnel profile based on the abnormal behavior identification results and the security integrity score; The standard job process is pushed to the personnel terminal corresponding to the personnel portrait according to the content recommendation algorithm; the standard job process corresponds to the abnormal point.

7. The method according to claim 6, characterized in that The method further comprises: Recording accidents corresponding to risk levels that reach a preset threshold into the historical accident file; Overlaying the risk level, the personnel behavior data, and the historical accident files onto the three-dimensional spatial model of the coal mine in the form of layers; In response to the obtained interactive operation query instruction, visually presenting the attributes and historical trajectory of any object in the three-dimensional spatial model of the coal mine; Generate a regional analysis report based on the risk level and the dynamic environmental data at preset time intervals; the regional analysis report includes a trend chart and a risk level distribution table.

8. An intelligent coal mine safety risk management and control device, characterized in that: The device comprises: A perception module is used to build a multi-source heterogeneous coal mine safety data system based on the data received from the target coal mine park; Model building module, used to build a three-dimensional spatial model of the coal mine based on static structural data and dynamic environmental data; A prediction module is used to extract risk indicators from the dynamic environmental data based on multi-algorithm fusion and evaluate them to obtain a risk level; The response module is used to trigger the corresponding linkage response mechanism if the risk level reaches a preset threshold, and form a visual result in the three-dimensional spatial model of the coal mine.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images