Mine intelligent monitoring and early warning decision system based on multi-source monitoring data fusion

Abstract

The present application relates to the technical field of mine safety, and in particular to a mine intelligent monitoring and early warning decision system based on multi-source monitoring data fusion, which comprises a multi-mine data management module, a cross-domain causal knowledge graph module and an intelligent prediction and decision module. By standardizing and aligning cross-mine multi-source heterogeneous data, a cross-mine disaster causal correlation network is constructed, the disaster causal chain of a source mine is adapted to a target mine, disaster probability prediction and disaster causal chain stage division are completed in combination with real-time data, four-level early warning is supported, and a matching disposal scheme is generated, thereby providing accurate and personalized monitoring and early warning and decision support for mine safety.

Description

Technical Field

[0001] This invention relates to the field of mine safety technology, and more specifically, to a mine intelligent monitoring, early warning and decision-making system based on the fusion of multi-source monitoring data. Background Technology

[0002] Mine safety monitoring is a core technology for ensuring the safety of mine production operations. It requires the integration of multi-source monitoring data to achieve disaster early warning and decision-making, thereby reducing the risk of accidents such as gas explosions and slope instability. Currently, mine monitoring largely relies on data collection and analysis from single mines. While this can obtain local parameters, it faces the challenge of integrating and utilizing data across multiple mines.

[0003] In existing technologies, multi-source monitoring data suffers from differences in sensor types, formats, and parameter definitions. Furthermore, the spatiotemporal coordinates and data acquisition frequencies vary across different mines, making it difficult to interoperate data across mines and establish a unified analytical foundation. Simultaneously, disaster causal correlation analysis is limited to a single mine, lacking the ability to adapt and transfer causal chains across mines. This makes it difficult to apply the established causal logic of disasters to other mines, and it fails to consider the impact of mine characteristics on the causal chain, resulting in a lack of personalized predictions. Moreover, most systems rely on single parameter thresholds for early warning, failing to consider the development stage of the causal chain and the probability of disaster occurrence. This leads to coarse early warning level classifications and a disconnect between early warnings and response plans, hindering accurate and efficient risk management.

[0004] To address the aforementioned shortcomings, this invention proposes a mine intelligent monitoring and early warning decision-making system based on multi-source monitoring data fusion, aiming to solve the problems of cross-mine data utilization, personalized causal analysis, and adaptability of early warning and response. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a mine intelligent monitoring, early warning, and decision-making system that integrates multi-source monitoring data.

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

[0007] The intelligent monitoring, early warning and decision-making system for mines that integrates multi-source monitoring data includes a multi-mine data governance module, a cross-domain causal knowledge graph module, and an intelligent prediction and decision-making module;

[0008] The multi-mine data governance module includes a heterogeneous data standardization unit, a cross-mine data alignment unit, and a three-dimensional structured database unit. It is used to standardize, align, and structure multi-source heterogeneous data across mines, and provide a unified data foundation for subsequent analysis.

[0009] The cross-domain causal knowledge graph module includes a knowledge graph construction unit and a causal chain migration algorithm unit, which are used to construct a disaster causal association network across mines, realize the migration and adaptation of the source mine disaster causal chain to the target mine, and output the personalized disaster causal chain of the target mine.

[0010] The intelligent prediction and decision-making module includes three units: a multi-mine disaster prediction model, a four-level early warning determination, and a decision tree generation. Based on the personalized disaster causal chain of the target mine and real-time monitoring data, it completes the prediction of the probability of disaster occurrence, the determination of the four-level early warning level, and generates a standardized response plan that matches the early warning level and the stage of the causal chain.

[0011] Furthermore, the specific implementation of the cross-mine data alignment unit includes:

[0012] Spatiotemporal alignment: Linear interpolation is used for stable data, and cubic spline interpolation is used for drastically fluctuating data to unify the acquisition frequency; Gauss-Kruger projection is used in conjunction with a geographic information system to convert the spatial coordinates of each mine into a unified plane rectangular coordinate system to achieve spatial matching;

[0013] Feature alignment: Construct a tag library containing core features and quantify various features; use the cosine similarity algorithm to calculate the feature similarity between the source mine and the target mine. When the feature similarity is greater than or equal to the similarity threshold, the source mine and the target mine are determined to be highly similar mines.

[0014] Furthermore, the specific implementation of the knowledge graph construction unit includes:

[0015] Entity hierarchy definition: encompasses mine characteristic entities, monitoring parameter entities, and disaster event entities;

[0016] Parameter anomaly degree calculation: Introduce the hazard characteristic coefficient K, and use the formula Calculate the degree of anomaly of the parameters, D-value; where, Range of values ; This refers to three types of standardized data collected by monitoring equipment and obtained through standardized processing during the mine monitoring process; This represents the security threshold for the three types of standardized data; This represents the critical threshold for the three types of standardized data.

[0017] Association weight calculation: Based on the statistical probability of disaster cases in multiple scenarios, determine the scenario-based weights of bidirectional associations between entities. The correlation types include bidirectional correlations between disaster causes and equipment status, and between equipment status and environmental parameters, with different weight values ​​for underground mines and open-pit mines.

[0018] Furthermore, the implementation steps of the causal chain migration algorithm unit include:

[0019] S21. Source Mine Causal Chain Extraction: Extract the complete source mine causal chain from the disaster event table of the knowledge graph, and clarify the hazard characteristic coefficient, original safety threshold, original hazard critical threshold and original association weight of each node in the chain as the adaptation benchmark.

[0020] S22. Mine similarity matching: Using the cosine similarity algorithm, if the feature similarity between the source mine and the target mine is greater than or equal to the similarity threshold, the source mine and the target mine are determined to be highly similar mines.

[0021] S23. Mining Feature Difference Analysis: Compare the core feature quantification values ​​of the two to identify key difference dimensions and their impact on parameter thresholds and weights;

[0022] S24. Target mine causal chain adaptation: A personalized causal chain for the target mine is obtained through threshold adaptation, parameter anomaly degree adaptation, and weight adaptation.

[0023] Furthermore, the threshold adaptation, parameter anomaly degree adaptation, and weight adaptation processes are as follows:

[0024] Threshold adaptation: Based on the established threshold adjustment rules, the threshold adjustment range is determined by calculating the feature difference quantification value;

[0025] Parameter anomaly degree adaptation: Calculate the parameter anomaly degree of each node in the target mine based on the adapted threshold and the real-time monitoring parameters of the target mine;

[0026] Weight adaptation: Based on the established weight adjustment rules and feature difference quantification values, calculate the weight of each node in the target mine.

[0027] Furthermore, the specific implementation of the multi-mine disaster prediction model unit includes:

[0028] Causal chain stage division: Based on the number of triggering nodes, continuity, and degree of parameter anomaly, the causal chain of the target mine disaster is divided into the initial stage, discrete risk stage, early stage, middle stage, and late stage.

[0029] Disaster Occurrence Probability Calculation: The probability P of disaster occurrence is calculated and predicted using a two-way correlation superposition formula. The specific calculation formula is as follows:

[0030] ;

[0031] Where m and n are the number of forward and reverse association trigger nodes, respectively, and the degree of parameter anomaly of the forward association i is: The corresponding scenario-based weight is The anomaly degree of the parameters of the reverse correlation j is The corresponding scenario-based weight is ; As a boundary constraint, ensure that the risk coefficient is at most 1.0.

[0032] Furthermore, the four-level early warning determination unit divides the early warning levels based on the probability of disaster occurrence and the causal chain stage, including Level 1 Blue Warning, Level 2 Yellow Warning, Level 3 Orange Warning, and Level 4 Red Warning.

[0033] Furthermore, the specific implementation of the decision tree generation unit is as follows:

[0034] After receiving real-time monitoring data from the target mine, the system first calculates the degree of parameter anomaly, node trigger status, and probability of disaster occurrence. Then, it determines the disaster causal chain stage in order of priority: initial stage, discrete risk stage, early stage, mid-term stage, and late stage. For each stage, it issues an early warning and outputs a corresponding response plan.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] 1. Solving the problem of cross-mine causal chain adaptation: A cross-mine disaster causal association network is constructed through a cross-domain causal knowledge graph module. Combined with a cosine similarity algorithm, the similarity between the source mine and the target mine is matched. Based on the analysis of mine feature differences, thresholds, parameter anomalies, and weights are adapted to complete the migration of the disaster causal chain. This method overcomes the limitations of single-mine causal analysis, enabling the target mine to obtain a personalized disaster causal chain. Combined with a bidirectional correlation superposition formula, the probability of disaster occurrence is calculated, improving the pertinence and accuracy of prediction.

[0037] 2. Achieving Precise Early Warning and Efficient Response: Based on the number, continuity, and degree of parameter anomalies of trigger nodes, the causal chain of target mine disasters is subdivided into initial stage, discrete risk stage, early stage, mid-stage, and late stage, accurately depicting the disaster risk transmission process and development trend; considering the superposition of two-way correlations, the probability of disaster occurrence is calculated to avoid the one-sidedness of unidirectional correlations; a multi-level early warning system is constructed, which differs from existing technologies that classify early warning levels based on a single threshold. Each early warning level in this invention is based on the aforementioned disaster causal chain stage and the probability of disaster occurrence, with clearly defined triggering conditions; finally, a linkage mechanism of early warning level - causal chain stage - response plan is established through a decision tree, realizing an immediate response plan upon early warning, significantly reducing risk response time, and avoiding safety hazards caused by excessive or insufficient early warnings. Attached Figure Description

[0038] Figure 1 A block diagram of a mine intelligent monitoring, early warning, and decision-making system that integrates multi-source monitoring data;

[0039] Figure 2 This is a flowchart of the causal chain migration algorithm unit of the present invention;

[0040] Figure 3This is a flowchart illustrating the implementation of the intelligent prediction and decision-making module of this invention. Detailed Implementation

[0041] Example, refer to Figure 1 The multi-source monitoring data fusion intelligent monitoring and early warning decision-making system for mines in this embodiment includes a multi-mine data governance module, a cross-domain causal knowledge graph module, and an intelligent prediction and decision-making module.

[0042] Multi-mine data governance module: The core function of this module is to realize the standardization, alignment and structured storage of multi-source data across mines, providing a data foundation for subsequent causal chain analysis and prediction and early warning. It includes heterogeneous data standardization unit, cross-mine data alignment unit and three-dimensional structured database unit.

[0043] Heterogeneous Data Standardization Unit: This unit develops standardized specifications for mine monitoring data, addressing differences in sensor types, data formats, and parameter definitions across various mines. It integrates multi-source data acquisition and preprocessing technologies to ensure data consistency.

[0044] Parameter classification and standardization definition: Covering three core data categories: disaster-causing data, equipment operation data, and environmental status data. Disaster-causing data includes gas concentration, rock stress, slope displacement, and roof subsidence. Equipment operation data includes load, speed, motor temperature, and insulation resistance of ventilation fans; cutting current, propulsion speed, and hydraulic system pressure of tunneling machines; brake oil pressure, wire rope tension, and operating speed of hoists; flow rate, outlet pressure, and motor power of water pumps; exhaust pressure, exhaust temperature, and oil level of air compressors; operating speed, deviation, and tension of conveyor belts; and other specific parameters of mining equipment. Environmental status data includes humidity, temperature, wind speed, atmospheric pressure, precipitation, and noise.

[0045] All parameters are categorized into two types based on their hazardous characteristics: higher values ​​indicate greater danger (e.g., gas concentration, equipment load) and lower values ​​indicate greater danger (e.g., insulation resistance), laying the foundation for subsequent calculations of the D-value, which indicates the degree of parameter anomaly. At the same time, parameter names, units, and sampling frequencies are standardized.

[0046] Data preprocessing and transmission: Deploy edge computing nodes (using low-power MCUs or GPU accelerator cards, with models adapted to the mining computing power requirements), eliminate random noise from sensors using Kalman filtering, process non-stationary signals using wavelet transform, and combine with the Laida criterion ( (Rules) Remove outliers to ensure data reliability ≥ 95%, and trigger sensor resampling when it falls below the threshold.

[0047] Cross-mine data transmission is achieved based on 5G / F5G slicing technology or industrial Ethernet. Underground mines prioritize 5G slicing to ensure real-time performance, while open-pit mines use F5G slicing to cover wide-area scenarios. The heterogeneous data (JSON / XML / binary) from each mine is converted into JSON-LD format using the Python PySpark framework, taking into account both machine readability and semantic relevance, which facilitates cross-mine data comparison and fusion.

[0048] Cross-mine data alignment unit: solves the inconsistency problem of multi-mine data in the spatiotemporal and feature dimensions, and ensures the accuracy of parameter anomaly calculation and causal chain migration;

[0049] 1. Spatiotemporal Alignment: Using millisecond-level timestamps as a benchmark, linear interpolation and cubic spline interpolation methods are combined to unify the data acquisition frequency of different mines. For scenarios with relatively stable data fluctuations (such as ventilation fan load data), linear interpolation is first used for initial frequency reduction, such as interpolating data from 10 seconds / time to 5 seconds / time. For monitoring items with drastic data changes and the need to retain high-order features (such as rockburst sensor data), cubic spline interpolation algorithm is used to ensure data continuity while more accurately fitting the curve change trend, ultimately achieving consistency of the time benchmark for all mine data.

[0050] Spatial Alignment: When establishing a universal coordinate system for the mine, the Gauss-Kruger projection method is adopted. The Earth's ellipsoid is divided into multiple projection zones according to a certain longitude difference using either 6-degree or 3-degree zoning. Each projection zone is independently calculated to convert spherical coordinates into plane rectangular coordinates. Specifically, geodetic control point data (including latitude, longitude, and elevation information) for each mining area are first collected. Then, using surveying equipment such as total stations, on-site measurements are conducted on underground working faces, roadways, open-pit mines, slopes, and other areas to obtain the three-dimensional spatial coordinates of key feature points.

[0051] Using professional GIS (Geographic Information System) software, such as ArcGIS and SuperMap, the collected coordinate data is imported into the system, Gauss-Kruger projection parameters (central meridian longitude, projection datum, etc.) are set, and the data is processed for coordinate transformation so that the core areas of each mine, such as underground working faces, roadways, open-pit mines, and slopes, are uniformly mapped to a plane coordinate system.

[0052] When comparing parameters in the same scenario across mines, a mine data integration platform is built, a unified data interface standard is formulated, and the monitoring data such as gas concentration collected by sensors in various mines are spatially matched and correlated according to the converted plane coordinates.

[0053] 2. Feature Alignment: Construct mine feature labels, including 12 core feature categories, and adopt appropriate quantization methods for different features:

[0054] Mine type: Using enumerated values, it is divided into open-pit mine 1 and underground mine 0;

[0055] Gas concentration: Set a numerical range for low-gas mines (gas emission ≤ 10m³). 3 / t) is assigned a value of 1-3, for high-gas mines (gas emission rate > 10m³). 3 / t) is assigned the value 4-5;

[0056] Mining depth: Numerical quantification is used, and the actual mining depth is recorded in meters;

[0057] Equipment configuration: Percentage quantification, proportion of intelligent integrated data acquisition equipment (0-100%).

[0058] Geological conditions: Set a complexity level, assign 1 to simple geology, 2 to medium geology, and 3 to complex geology (such as faults and fracture zones).

[0059] Hydrological conditions: Set a risk index, assign a value of 1 to areas with stable water temperature, and assign a value of 2-5 to areas with sudden water inrush or high water volume, depending on the degree of risk.

[0060] Mining scale: Using a production value range, small and medium-sized mines (annual production < 1 million tons) are assigned a value of 1-3, and large mines (annual production ≥ 1 million tons) are assigned a value of 4-5;

[0061] Mining duration: The actual mining time is recorded in years using time-based numerical values.

[0062] Ventilation system: The level of intelligence is set in stages, with traditional ventilation system assigned a value of 1, basic intelligent control assigned a value of 2, and intelligent ventilation control system assigned a value of 3.

[0063] Personnel density: Quantified by the average working area per person. The smaller the average working area per person, the larger the value (a reasonable range of 1-10 is set according to the actual situation).

[0064] Disaster history: Time-weighted records are used. Major accidents occurring in the past year are assigned a value of 5, those occurring in the past 2-3 years are assigned a value of 4, those occurring in the past 3-5 years are assigned a value of 3, and those with no major accident records are assigned a value of 1.

[0065] Safety investment: Record the percentage of annual safety investment (0-100%).

[0066] The cosine similarity algorithm is used to calculate the feature similarity between different mines after normalization for different data types. The calculation formula is as follows:

[0067] ;

[0068] Where A represents the source mine, which has a clear causal chain for a certain disaster and can be used as a reference sample; B represents the target mine, which is the mine for which disaster prediction needs to be carried out. This represents the total number of feature dimensions used to calculate mine similarity, in this embodiment. ; Indicates the first of the source mines A The values ​​of each feature index after normalization; Indicates the first [unclear] of target mine B The values ​​of each feature index are normalized; Sim(A,B) is the feature similarity between the source mine A and the target mine B, with a value range of [0,1]. When the similarity is ≥S0, A and B are determined to be highly similar mines, and subsequent disaster causal chain migration analysis can be performed. S0 is the similarity threshold. In this embodiment, S0=0.7, which is an empirical value.

[0069] The three-dimensional structured database unit is built on PostgreSQL and TimescaleDB (a time-series database). It uses Mine ID, monitoring dimension, and time axis as its core indexes to store three types of standardized data, supporting efficient querying and calculation. It adopts a distributed deployment, compatible with existing mine sensor networks and control systems. The data storage structure is as follows:

[0070] Basic Information Table: Records the quantified values ​​of characteristic tags for each mine;

[0071] Real-time monitoring table: stores preprocessed parameter values;

[0072] Disaster Event Table: Records the complete causal chain of historical disasters, such as Mine ID: K001, Disaster Type: Gas Explosion, Causal Chain Nodes: Humidity = 85% (D1 = 0.75) → Ventilation Fan Insulation Resistance = 0.8MΩ (D2 = 0.8) → Gas Concentration = 1.8% (D4 = 0.8) → Gas Explosion, Correlation Weights: W1 = 0.75, W2 = 0.75, W4 = 0.9;

[0073] Cross-domain causal knowledge graph module: This module constructs a cross-mine disaster causal association network to realize the migration of the causal chain from the source mine to the target mine, providing core logical support for multi-mine collaborative early warning. It includes two units: knowledge graph construction and causal chain migration algorithm.

[0074] 1. Knowledge Graph Construction Unit: Construct a knowledge graph of mine characteristics, monitoring parameters, and disaster events, and clarify the entity definition, the calculation method of parameter anomaly degree (D value), and the association weight (W value).

[0075] The entity hierarchy and definition are shown in Table 1:

[0076] Table 1

[0077]

[0078] Calculation of parameter anomaly degree D-value: Introducing the hazard characteristic coefficient K (K=1 corresponds to a higher value indicating greater danger, K=-1 corresponds to a lower value indicating greater danger), the calculation formula is as follows:

[0079] ;

[0080] in, This indicates the degree of abnormality of the parameter, and its range is [range missing]. The larger the value, the higher the degree of abnormality and the higher the level of danger. This represents the actual monitored parameter value, that is, the specific data collected by the equipment during the current mine monitoring process; The safety threshold is a critical value of a parameter that ensures the safe operation of mine production. When the actual parameter is within the safety threshold range, the production status is considered safe. This represents the critical threshold for danger. When the actual parameters reach or exceed this value (in conjunction with the K value to determine the direction), the mine is considered to be in an extremely dangerous state.

[0081] Association and Weight Calculation: Establishing bidirectional associations between entities and contextualized weights. The probability was determined based on statistical analysis of disaster cases across multiple scenarios over the past 10 years; details are as follows:

[0082] Multi-scenario mine disaster cases were integrated from official databases, enterprise ledgers, and core literature. After cleaning, they were categorized by mine type and disaster type, and the abnormal parameter sequences before the disaster were extracted from each case. The co-occurrence frequency of parameter B after parameter A anomaly was first counted (e.g., in 500 gas explosion cases, 400 cases showed increased humidity → decreased insulation resistance, with a co-occurrence frequency of 80%) to screen for potential associations. The chi-square test was used to verify the significance of the association (excluding accidental co-occurrences), and significant associations were included in subsequent calculations. Based on the probability of the association occurring in similar cases, for example, if decreased insulation resistance → increased load occurred in 450 / 500 gas explosion cases, the basic weight = 450 / 500 = 0.9. Adjustments were made according to mine type (underground / open-pit). For example, if the effect of humidity on insulation resistance is weak in open-pit mines, a scenario coefficient of 0.6 was used for correction, and the final scenario-based weight = 0.9 * 0.6 = 0.54 (retaining one decimal place as 0.5). The weights were updated every 6 months with new cases, combined with system early warning feedback, to ensure that the weights closely match the actual scenarios.

[0083] The correlation types include bidirectional correlations between disaster causes and equipment status. Positive correlations include, for example, abnormal gas concentration leading to increased ventilation fan load, with a positive scenario weight of 0.6 in underground mines and 0.1 in open-pit mines. Negative correlations include, for example, increased ventilation fan load leading to local gas flow obstruction, thus exacerbating abnormal concentration, with a negative scenario weight of 0.8 in underground mines and 0.1 in open-pit mines. Bidirectional correlations between equipment status and environmental parameters include, for example, positive correlations include, for example, ventilation fan failure leading to underground gas accumulation, with a positive scenario weight of 0.7 in underground mines and 0.3 in open-pit mines. Negative correlations include, for example, high humidity in underground mines leading to aging of ventilation fan motor insulation, further increasing the risk of failure, with a negative scenario weight of 0.75 in underground mines and 0.6 in open-pit mines.

[0084] 2. Causal Chain Migration Algorithm Unit: Adapts the association logic of the source mine to the target mine, outputting a personalized causal chain, such as... Figure 2 As shown, the steps are as follows:

[0085] S21. Source Mine Causal Chain Extraction: Extract the complete causal chain of a certain disaster from the disaster event table of the knowledge graph source mine, and clarify the core attributes of each node in the chain: parameter type (disaster trigger / equipment status / environmental status), hazard characteristic coefficient (…). Value), original security threshold ( Original), original hazard threshold ( Original), original association weight ( This forms a basic attribute table of the causal chain of the source mine, which serves as the adaptation benchmark.

[0086] S22. Mine similarity matching: The source mine and the target mine feature labels are compared using the cosine similarity algorithm. If the similarity is ≥0.7, it is determined to be a highly similar mine.

[0087] S23. Mine Feature Difference Analysis: Based on a mine feature tag library (containing 12 categories of features such as mine type, mining depth, equipment configuration, geological conditions, and environmental conditions), the feature quantification values ​​of the source mine and the target mine are compared to identify key difference dimensions. Common difference dimensions and their impact types are as follows:

[0088] Differences in mining depth affect the threshold values ​​of gas concentration and rock stress parameters. If the target mine's mining depth is less than that of the source mine, the gas pressure is lower, and the critical threshold values ​​for gas parameters can be increased. Differences in equipment configuration affect the threshold values ​​and weights of equipment status parameters. If the target mine's equipment is more intelligent than that of the source mine, the equipment has higher fault tolerance, and the critical threshold values ​​for equipment parameters can be decreased. Differences in geological conditions affect the threshold values ​​and weights of rock mass and slope parameters. If the target mine's rock mass stability is greater than that of the source mine, the slope displacement tolerance is higher, and the safety threshold values ​​for displacement parameters can be increased.

[0089] S24, Target Mine Causal Chain Adaptation:

[0090] S241, Threshold adaptation ( , For three types of parameters—disaster causes, equipment status, and environmental status—threshold adjustment rules are established based on the impact types of characteristic difference dimensions.

[0091] If the difference in characteristics leads to an increase in the safety redundancy of a certain parameter of the target mine (e.g., more advanced equipment, more stable geology), then this parameter... Adjust towards a safer direction; if the differences in characteristics lead to a decrease in the safety redundancy of a certain parameter of the target mine (such as deeper mining depth or harsher environment), then adjust this parameter accordingly. Adjustments are being made towards stricter standards; among them, and These represent the safety threshold and the critical danger threshold of the parameters corresponding to the target mine, respectively.

[0092] The threshold adjustment range is referenced to the feature difference quantization value: Let the feature difference quantization value be... , value range ,but (The parameters are dynamically set according to their importance. Core parameters such as gas concentration are set at 30%, and secondary parameters such as environmental noise are set at 10%. The core and secondary parameters are obtained with reference to industry standards.)

[0093] S242, D-value adaptation ( ): Based on the threshold after target ore adaptation ( , Following the D-value calculation formula, input the real-time monitoring parameter values ​​of the target mine. The degree of parameter anomaly at each node of the target mine was calculated. ;

[0094] S243, Weight Adaptation ( For the association weights between parameters in a causal chain, a weight adjustment rule is established based on the impact of feature differences on the association strength:

[0095] If the difference in characteristics enhances the effect of a certain association (e.g., higher humidity in the target ore → stronger association of humidity leading to equipment insulation), then ( (Quantification of the difference in this feature). The baseline adjustment factor is used, and adjustments are made based on parameter importance and mine type. In this embodiment... ;

[0096] If the difference in characteristics weakens the impact of a certain association (e.g., higher protection level of equipment in the target mine → weaker association of environmental factors causing equipment failure), then ; The baseline adjustment factor is used, and adjustments are made based on parameter importance and mine type. In this embodiment... ;

[0097] After weight adjustment, the following conditions must be met. (If the correlation is below 0.1, it is determined that the correlation has no significant impact on the target mine and can be removed from the causal chain), ensuring the rationality and effectiveness of the correlation weight;

[0098] Finally, the personalized causal chain of the target mine is obtained through threshold adaptation, D-value adaptation, and weight adaptation;

[0099] Intelligent prediction and decision-making module: such as Figure 3 As shown, this module, based on the personalized causal chain and real-time data of the target mine, realizes disaster probability prediction, four-level early warning triggering and standardized response plan output, including a multi-mine disaster prediction model unit, a four-level early warning judgment unit and a decision tree generation unit.

[0100] Multi-mine disaster prediction model unit: This unit predicts the likelihood of a disaster by matching stages and calculating the probability of occurrence, incorporating a two-way correlation risk superposition logic. First, based on the number of nodes triggered by real-time parameters of the target mine and the D value, the disaster causal chain is divided into five stages: The initial stage is characterized by a single minor anomaly with no risk correlation and an extremely low probability of disaster when only one node is triggered; the early stage is characterized by the continuous transmission of risk, the initial formation of the disaster causal chain, and the need for intervention to stop the transmission; the mid-stage is characterized by the formation of the core risk chain, approaching the disaster node, and requiring urgent intervention; the final stage is characterized by all four consecutive nodes triggered or any node having a D value of 1.0. The stage is classified as the late stage, characterized by a completely closed risk chain and an extremely high probability of disaster occurrence, requiring the activation of the highest level of emergency response. The stage is defined as the discrete risk stage when two or more discontinuous nodes trigger the event, indicating that the risk has not formed a complete transmission chain. The probability of disaster occurrence is lower than that of the same number of continuous nodes, but higher than that of a single node (the initial stage), requiring separate stage division to avoid over-warning or under-warning. The triggering condition is determined based on the D-value threshold. When the D-value is greater than the threshold, it is determined to be a trigger node. In this embodiment, the D-value thresholds for the initial stage, early stage, middle stage, late stage, and discrete risk stage are 0.3, 0.4, 0.5, 0.6, and 0.3, respectively, determined through regression analysis of historical monitoring data.

[0101] Disaster occurrence probability (P) calculation: A two-way correlation superposition formula is used, combining the D value of positive (cause → equipment → environment) and negative (environment → equipment → cause) correlations with scenario-based weights. The specific calculation formula is as follows:

[0102] ;

[0103] Where m and n are the number of forward and reverse association trigger nodes, respectively, and the degree of parameter anomaly of the forward association i is: The corresponding scenario-based weight is The anomaly degree of the parameters of the reverse correlation j is The corresponding scenario-based weight is ; To constrain the boundaries, ensure that the risk coefficient is at most 1.0;

[0104] Level 4 Early Warning Judgment Unit: The early warning level is determined by combining the probability of disaster occurrence P with the stage of the causal chain; the probability threshold for classifying the early warning level is determined based on the statistics of historical disaster cases;

[0105] Level 1 Blue Alert Triggering Conditions: 1. The causal chain stage is in the initial stage and P < 10%; 2. The causal chain stage is in the discrete risk stage and 10% ≤ P < 20%;

[0106] Level II Yellow Alert Triggering Conditions: 1. The causal chain stage is in the discrete risk stage and 20% ≤ P < 30%; 2. The causal chain stage is in the initial stage and 30% ≤ P < 60%.

[0107] The triggering conditions for a Level 3 Orange Alert are: the causal chain is in the middle stage and 60% ≤ P < 80%;

[0108] Level IV Red Alert Trigger Conditions: The causal chain is in a late stage and P≥80%;

[0109] The system calculates the P-value and stage at a fixed frequency (once every 5 seconds) and pushes them synchronously through the monitoring screen (displaying the causal chain stage, D-value, and P-value), mobile APP (SMS, handling suggestions), and underground AR glasses (pop-up warning information). Discrete risk stages are additionally marked with discontinuous nodes and potential related risks.

[0110] Decision tree generation unit: Based on the warning level and the causal chain stage, generate a decision tree that links warning, stage and response.

[0111] After receiving real-time monitoring data from the target mine, the system first calculates the D value, node trigger status, and probability P. It then calculates the anomaly quantification value for each monitoring node using the parameter anomaly degree (D value), identifies the continuity (continuous / discontinuous) of the current trigger node, and derives the disaster occurrence probability (P) based on a bidirectional correlation superposition formula, providing data support for subsequent stage judgments. Based on the above calculation results, the system enters the core branch of the causal chain stage judgment, judging according to the priority of the initial stage, discrete risk stage, early stage, mid-term stage, and late stage, with each branch corresponding to a specific handling plan.

[0112] Initial phase: Blue alert requires mild handling, with the monitors as the responsible party; measures include monitoring key parameters every 2 minutes and conducting equipment inspections once a day; time limit: 48 hours.

[0113] Discrete Risk Stage: Blue or yellow alerts require discrete risk handling, with the safety team as the responsible party; the measure is to investigate potential connections between discontinuous stages; the time limit is 24 hours.

[0114] Initial stage: Yellow or orange alerts require severe handling, with the safety team or safety management department as the responsible party; measures include adjusting equipment parameters to block risk transmission, and collecting data at all nodes every 30 seconds; time limit: 2 hours;

[0115] Mid-term stage: Orange or red alerts require moderate to severe response, with the safety management department or emergency command center as the responsible party; measures include stopping operations in the affected area, operating key equipment in dual-machine mode, and reducing risk parameters to the safety threshold through mandatory measures; the time limit is 30 minutes for intervention and 1 hour for evacuation;

[0116] Later stage: A red alert requires severe response. The responsible entity is the emergency command center, and the measures include shutting down the entire mine system, evacuating all personnel, coordinating with local rescue efforts, and deploying medical resources; the time limit is 10 minutes for evacuation and 30 minutes for rescue coordination.

[0117] Through the detailed description of the above embodiments, the intelligent mine monitoring and early warning decision-making system based on multi-source monitoring data fusion of the present invention constructs a cross-mine disaster causal association network through a cross-domain causal knowledge graph module, achieves similarity matching between source mines and target mines by combining an improved cosine similarity algorithm, performs threshold, parameter anomaly degree and weight adaptation based on mine feature difference analysis, completes disaster causal chain migration, and enables target mines to obtain personalized causal chains. Combined with a bidirectional correlation superposition formula, the probability of disaster occurrence is calculated, improving the pertinence and accuracy of prediction. Based on the disaster causal chain stage (initial, discrete risk, early stage, middle stage, late stage) and the probability of disaster occurrence, four levels of early warning (blue, yellow, orange, red) are divided, and the decision tree generation unit outputs disposal plans that match the early warning level and disaster causal chain stage, providing accurate and personalized monitoring, early warning and decision support for mine safety.

[0118] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.

[0119] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0120] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0121] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0122] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus 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 apparatuses or units may be electrical, mechanical, or other forms.

[0124] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion 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 to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0125] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A mine intelligent monitoring, early warning, and decision-making system based on multi-source monitoring data fusion, characterized in that: It includes a multi-mine data governance module, a cross-domain causal knowledge graph module, and an intelligent prediction and decision-making module; The multi-mine data governance module is used to standardize, align, and structure the multi-source heterogeneous data across mines. The cross-domain causal knowledge graph module includes a knowledge graph construction unit and a causal chain transfer algorithm unit, used to construct a disaster causal association network across mines, realize the transfer and adaptation of the source mine disaster causal chain to the target mine, and output a personalized disaster causal chain for the target mine; wherein, the implementation steps of the causal chain transfer algorithm unit include: S21. Source Mine Disaster Causal Chain Extraction: Extract the complete source mine disaster causal chain from the disaster event table of the knowledge graph, and clarify the hazard characteristic coefficient, original safety threshold, original hazard critical threshold and original association weight of each node in the chain; S22. Mine similarity matching: The cosine similarity algorithm is used to calculate the feature similarity between the source mine and the target mine. If the feature similarity is greater than or equal to the similarity threshold, the source mine and the target mine are determined to be highly similar mines. S23. Mining Feature Difference Analysis: Compare the core feature quantification values ​​of the two to identify key difference dimensions and their impact on parameter thresholds and weights; S24. Target Mine Causal Chain Adaptation: A personalized disaster causal chain for the target mine is obtained through threshold adaptation, parameter anomaly degree adaptation, and weight adaptation. The intelligent prediction and decision-making module includes a multi-mine disaster prediction model unit, a four-level early warning determination unit, and a decision tree generation unit. Based on the personalized disaster causal chain of the target mine and real-time monitoring data, it completes the prediction of the probability of disaster occurrence, the determination of the four-level early warning level, and generates a standardized response plan that matches the early warning level and the stage of the disaster causal chain.

2. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 1, characterized in that, In the multi-mine data governance module, the standardization processing step classifies and standardizes heterogeneous data to form three types of standardized data; the spatiotemporal and feature dimension alignment step first performs spatiotemporal alignment on the data collected across mines, then constructs mine feature labels, and uses differentiated quantification methods to quantify different features; The structured storage step uses a database to store the three types of standardized data.

3. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 2, characterized in that, The three types of standardized data include disaster cause data, equipment operation data, and environmental status data.

4. The mine intelligent monitoring, early warning, and decision-making system based on multi-source monitoring data fusion according to claim 2, characterized in that, The specific implementation of the knowledge graph construction unit includes: Entity hierarchy definition: encompasses mine characteristic entities, monitoring parameter entities, and disaster event entities; Parameter anomaly degree calculation: Introduce the hazard characteristic coefficient K, and use the formula Calculate the degree of anomaly of the parameters, D-value; where, Range of values ; This refers to three types of standardized data collected by monitoring equipment and obtained through standardized processing during the mine monitoring process; This represents the security threshold for the three types of standardized data; This represents the critical threshold for danger of the three types of standardized data; Association weight calculation: Based on the statistical probability of disaster cases in multiple scenarios, determine the scenario-based weights of bidirectional associations between entities. The correlation types include bidirectional correlations between disaster causes and equipment status, and between equipment status and environmental parameters, with different weight values ​​for underground mines and open-pit mines.

5. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 1, characterized in that, The threshold adaptation, parameter anomaly degree adaptation, and weight adaptation processes are as follows: Threshold adaptation: Based on the established threshold adjustment rules, the threshold adjustment range is determined by calculating the feature difference quantification value; Parameter anomaly degree adaptation: Calculate the parameter anomaly degree of each node in the target mine based on the adapted threshold and the real-time monitoring parameters of the target mine; Weight adaptation: Based on the established weight adjustment rules and feature difference quantification values, calculate the weight of each node in the target mine.

6. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 1, characterized in that, The specific implementation of the multi-mine disaster prediction model unit includes: Causal chain stage division: Based on the number of triggering nodes, the continuity of triggering nodes, and the degree of parameter anomaly, the causal chain of the target mine disaster is divided into the initial stage, discrete risk stage, early stage, middle stage, and late stage. Disaster Occurrence Probability Calculation: The probability P of disaster occurrence is calculated and predicted using a two-way correlation superposition formula. The specific calculation formula is as follows: ； Where m and n are the number of forward and reverse association trigger nodes, respectively, and the degree of parameter anomaly of the forward association i is: The corresponding scenario-based weight is The anomaly degree of the parameters of the reverse correlation j is The corresponding scenario-based weight is ; For boundary constraints.

7. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 1, characterized in that, The four-level early warning determination unit divides the early warning levels based on the probability of disaster occurrence and the causal chain stage, including Level 1 Blue Warning, Level 2 Yellow Warning, Level 3 Orange Warning and Level 4 Red Warning.

8. The intelligent monitoring, early warning, and decision-making system for mines based on multi-source monitoring data fusion according to claim 1, characterized in that, The specific implementation of the decision tree generation unit is as follows: After receiving real-time monitoring data from the target mine, the system first calculates the degree of parameter anomaly, node trigger status, and probability of disaster occurrence. Then, it determines the disaster causal chain stage in order of priority: initial stage, discrete risk stage, early stage, mid-term stage, and late stage. For each stage, it issues an early warning and outputs a corresponding response plan.

Images