Mine safety monitoring data fusion analysis and hierarchical early warning system

Abstract

The present application relates to the field of digital analysis technology, in particular to a mine safety monitoring data fusion analysis and hierarchical early warning system, comprising a monitoring data processing unit, a safety dynamic correlation unit, a risk propagation resonance calculation unit and a disaster adaptive fusion unit. The present application replaces the single-point maximum with the risk situation field through the disaster adaptive fusion unit, redefines the disaster severity from the spatial extension dimension, and through the three-dimensional dynamic gating mechanism of historical benchmark, spatial attention and trigger attention, makes the system rely on historical law to run stably in steady state, and focuses attention on the disaster type with the fastest spread and the highest chain risk in an instant in non-steady state, and finally generates a comprehensive risk index with sensitivity to sudden peaks and robustness to isolated noise, significantly improving the advance and accuracy of early warning.

Description

Technical Field

[0001] This invention relates to the field of digital analysis technology, specifically to a mine safety monitoring data fusion analysis and hierarchical early warning system. Background Technology

[0002] A mine safety monitoring data fusion analysis and hierarchical early warning system refers to a system that integrates multi-source heterogeneous monitoring data from underground mines, such as gas, roof, hydrology, and ventilation, and uses multi-sensor data fusion and correlation analysis technology to comprehensively analyze and extract features from multi-dimensional safety information. This eliminates the limitations and uncertainties of single monitoring methods, and then dynamically classifies the mine safety situation into different levels based on risk thresholds and early warning models, thereby achieving precise and hierarchical safety early warning and risk control.

[0003] However, existing technologies have several shortcomings. First, the indicator correlation models are rigid, often using statically preset indicator correlation networks or relying solely on data-driven correlation analysis. This makes it difficult to adapt to the dynamic changes in surrounding rock caused by mining operations and easily generates false correlation results without physical support. Second, the anomaly assessment methods are one-sided, generally treating multiple anomalies as independent events and using only simple linear weighted summation calculations. This ignores the differences in frequency characteristics of different disaster precursors and fails to consider the nonlinear risk amplification effect caused by the spatiotemporal synchronous coupling of multiple disasters. Third, the risk assessment mechanism is rigid, mainly relying on the maximum value of a single point or fixed weights to determine the overall risk, and cannot effectively identify the spatial distribution characteristics of the risk.

[0004] Based on this, the present invention provides a mine safety monitoring data fusion analysis and hierarchical early warning system to solve the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this invention is to provide a mine safety monitoring data fusion analysis and hierarchical early warning system. This invention uses a disaster adaptive fusion unit to replace the single-point maximum value with a risk situation field, redefining the severity of disasters from the dimension of spatial extensibility. Through a three-dimensional dynamic gating mechanism of historical benchmark, spatial attention and triggered attention, the system can operate robustly based on historical patterns in steady state, and instantly focus attention on the disaster type with the fastest spread and highest chain risk in non-steady state. The resulting comprehensive risk index has both sensitivity to sudden spikes and robustness to isolated noise, significantly improving the lead time and accuracy of early warning.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a mine safety monitoring data fusion analysis and hierarchical early warning system, comprising a monitoring data processing unit, a safety dynamic correlation unit, a risk propagation resonance calculation unit, and a disaster adaptive fusion unit, wherein: The monitoring data processing unit: collects data from each sensor in real time, cleans and aligns the data, and then performs standard transformation on the data. The safety dynamic association unit: Based on the engineering layout and geological conditions of the mine, the mine is converted into multiple types of nodes. On the basis of multiple types of nodes, physical edges and causal edges between nodes are established. The weight components of physical edges and causal edges are evaluated and weighted and fused to form a causal connection graph. The risk propagation resonance calculation unit: scores and weights each sensor, updates the causal connection graph, and uses the updated causal connection graph to calculate and map the risk energy value of all nodes; The disaster adaptive fusion unit: constructs a disaster potential energy field, calculates the coupling strength of the disaster trigger matrix and various disasters, generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warning based on the final risk index.

[0007] The monitoring data processing unit includes a data acquisition module and a data preprocessing module, wherein: The data acquisition module is used to collect sensor data collected in real time from multiple types of sensors, package the collected sensor data into a sensor data package, and upload it. The data preprocessing module receives sensor data packets uploaded by the data acquisition module, cleans the sensor data packets, performs time alignment and standardization after cleaning, and obtains the first dataset.

[0008] The secure dynamic association unit includes: a node definition module, a bilateral establishment module, and a weight component calculation module, wherein: The node definition module: Based on the engineering layout and geological conditions of the mine, each sensor is recorded as the first node, the underground engineering space of the mine is divided into several regions, a second node is established for each region, and a geological structure node is established for the geological anomaly. The bilateral establishment module: constructs physical edges by corresponding the first node and the second node one-to-one, and establishes causal edges between the second node and the third node; The weight component calculation module is used to evaluate the weight components of physical edges and causal edges and perform linear superposition to form a causal connection graph.

[0009] The weight component calculation module is used to evaluate the weight components of physical edges and causal edges and perform linear superposition to form a causal connection graph. The specific steps are as follows: Calculate the spatial distance between the centers of the nodes at both ends of the physical edge and the rock quality index, and determine the weight components of the physical edge based on the spatial distance between the centers of the nodes at both ends and the rock quality index. Take a sliding time window of fixed length T, obtain the monitoring data of the nodes at both ends of the causal edge within the sliding time window from the first dataset, and calculate the absolute value of the Pearson correlation coefficient of the nodes at both ends of the causal edge based on the monitoring data. Perform Granger causality tests within a sliding time window to obtain statistics. Sort the statistics corresponding to all causal edges and calculate the causal strength based on the maximum and minimum values ​​of the statistics. The absolute value of the Pearson correlation coefficient and the causal strength are fused to obtain the weight components of the causal edges. Then, the weight components of the physical edges and the causal edges are linearly superimposed to obtain the edge weights and the causal connection graph.

[0010] The risk propagation resonance calculation unit includes a signal diagnosis module and a risk energy mapping module, wherein: The signal diagnosis module: decomposes the first dataset, calculates the risk score of the first node at the current moment, generates a transmission weight, and updates the causal connection graph through the transmission weight; The risk energy mapping module calculates the external energy value based on the updated causal connection graph and performs coupling resonance, updates the risk energy value of the previous moment, and then performs mapping.

[0011] The signal diagnosis module decomposes the first dataset, calculates the risk score of the first node at the current moment, generates a transmission weight, and updates the causal connection graph using the transmission weight. The specific steps are as follows: The monitoring data of the first node is obtained from the first dataset. The monitoring data of the first node is smoothed to obtain the first component. The monitoring data of the first node is subtracted from the first component to obtain the second component and the instantaneous main frequency is calculated. The trend score and transient score are calculated based on the first and second components. The trend score and transient score are then weighted and summed to obtain the risk score. For each causal edge, the propagation weight at the current moment is calculated one by one. The calculation process is as follows: ;; In the formula: Let be the instantaneous dominant frequency of the source node of the causal edge at the current moment. The optimal resonant propagation frequency for this causal edge is... This represents the frequency response bandwidth parameter of the causal edge; Connect the edge weight with Multiply them to obtain the transmission weight.

[0012] The risk energy mapping module: calculates the external energy value based on the updated causal connection graph and performs coupling resonance, updates the risk energy value of the previous time step, and then performs mapping, including: By combining the risk score with the transmission weight, the external energy value is obtained. Resonant nodes are selected from all nodes for coupled resonance calculation to obtain the resonance intensity value. Based on the risk energy value of the previous moment, the risk energy value is updated by combining the external energy value and the resonance intensity value. The updated risk energy value is then mapped using the hyperbolic tangent function to obtain the comprehensive risk vector.

[0013] The disaster adaptive fusion unit includes a disaster probability calculation module and a graded early warning module, wherein: The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, and calculates the disaster triggering matrix and the coupling strength of various types of disasters; The graded early warning module generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warnings based on the final risk index.

[0014] The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, and calculates the disaster triggering matrix and the coupling strength of various disasters, including: Determine the set of nodes associated with each type of disaster, and calculate the disaster field strength at the center coordinates of the second node: ;; In the formula: This represents the risk value of the node at the current moment. The center coordinates of the current second node. Spatial coordinates of the field source node For spatial influence kernel function, It is a very small constant; The proportion of dangerous areas is selected based on the intensity of the disaster field.

[0015] The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, calculates the disaster triggering matrix and the coupling strength of various disasters, and also includes: For each type of disaster, the transfer intensity of each combination is calculated. The transfer intensity is then normalized by the sum of squares to obtain the disaster triggering matrix. The coupling strength of various disasters is determined based on the disaster triggering matrix.

[0016] Compared with the prior art, the beneficial effects of the present invention are: The safety dynamic correlation unit of this invention separates and linearly superimposes the weighted components of physical and causal edges. This not only reflects the inherent spatial attenuation law of rock mechanics but also adaptively captures the real-time causal emergence between monitoring data, solving the problem of integrating structural correlation and data-driven approaches. Secondly, the risk propagation resonance calculation unit achieves profound innovation. It adjusts the transmission efficiency of risk along the causal edge based on the instantaneous dominant frequency difference of abnormal signals, allowing high-frequency micro-seismic impacts and low-frequency stress accumulation to propagate along their respective optimal paths. Simultaneously, it specifically identifies dangerous patterns that surge synchronously from multiple sensors within the same spatiotemporal window, treating traditional methods as... The system transforms concurrent anomalies of isolated noise into nonlinear risk increments, possessing a unique ability to perceive precursors of cascading disasters. Finally, the disaster adaptive fusion unit replaces the single-point maximum value with a risk situation field, redefining the severity of disasters from the dimension of spatial extensibility. Through a three-dimensional dynamic gating mechanism of historical benchmarks, spatial attention, and triggered attention, the system operates robustly based on historical patterns in steady state, and instantly focuses attention on the disaster type with the fastest spread and highest cascading risk in non-steady state. The resulting comprehensive risk index combines sensitivity to sudden spikes with robustness to isolated noise, significantly improving the lead time and accuracy of early warnings. Attached Figure Description

[0017] Figure 1 This is a system diagram of a mine safety monitoring data fusion analysis and hierarchical early warning system according to the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example: like Figure 1 As shown, this embodiment provides a mine safety monitoring data fusion analysis and hierarchical early warning system. This invention provides a mine safety monitoring data fusion analysis and hierarchical early warning system, including a monitoring data processing unit, a safety dynamic correlation unit, a risk propagation resonance calculation unit, and a disaster adaptive fusion unit, wherein: The monitoring data processing unit: collects data from each sensor in real time, cleans and aligns the data, and then performs standard transformation on the data. The safety dynamic association unit: Based on the engineering layout and geological conditions of the mine, the mine is converted into multiple types of nodes. On the basis of multiple types of nodes, physical edges and causal edges between nodes are established. The weight components of physical edges and causal edges are evaluated and weighted and fused to form a causal connection graph. The risk propagation resonance calculation unit: scores and weights each sensor, updates the causal connection graph, and uses the updated causal connection graph to calculate and map the risk energy value of all nodes; The disaster adaptive fusion unit: constructs a disaster potential energy field, calculates the coupling strength of the disaster trigger matrix and various disasters, generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warning based on the final risk index.

[0020] The monitoring data processing unit includes a data acquisition module and a data preprocessing module, wherein: The data acquisition module is used to collect sensor data collected in real time from multiple types of sensors, package the collected sensor data into a sensor data package, and upload it. It should be noted that the sensor data includes: real-time concentration values ​​(in volume percentage) uploaded by each gas sensor at a fixed frequency. Stress-strain sensors read time-series data from anchor stress gauges, borehole stress gauges, and surrounding rock strain gauges; each record includes the measuring point number, acquisition time, and stress value (in megapascals) or strain value (in microstrain). Measurements also include delamination and tunnel convergence deformation, displacement values ​​(in millimeters), and temperature, wind speed (meters per second), and water inflow (cubic meters per hour) acquired by temperature sensors. The data preprocessing module receives sensor data packets uploaded by the data acquisition module, cleans the sensor data packets, performs time alignment and standardization after cleaning, and obtains the first dataset.

[0021] Data cleaning, time alignment, and standardization of sensor data packets are all existing technologies, and will not be elaborated upon in this application.

[0022] The secure dynamic association unit includes: a node definition module, a bilateral establishment module, and a weight component calculation module, wherein: The node definition module: Based on the engineering layout and geological conditions of the mine, each sensor is recorded as the first node, the underground engineering space of the mine is divided into several regions, a second node is established for each region, and a geological structure node is established for the geological anomaly. Each installed and normally functioning sensor corresponds to a first node. The attributes of a first node include: the sensor's unique number as a node representation, the data it monitors, the three-dimensional spatial coordinates of its installation location, and the disaster type it is associated with.

[0023] The underground engineering space in the mine is divided into several areas, and the basis for this division is as follows: Each independent mining unit of the coal mining face is a region, each independent tunneling head of the tunneling face is a region, the main transport roadway is divided into several regions according to the intersection and ventilation zone boundary, special functional spaces such as electromechanical chambers, pump rooms, and substations are each a separate region, and the boundary of the closed goaf is divided into one or more regions according to the isolation range of the sealing wall. The attributes of the second node include: region number, geometric coordinates of the region center, rock mass parameters (uniaxial compressive strength and rock mass integrity coefficient) within the region, and hazard type labels that the region may face.

[0024] The third node, based on the mine geological report and the actual geological conditions revealed during the mining process, establishes a geological structural node for each known geological anomaly that affects mining safety. These anomalies include, but are not limited to: faults (fault attitude, drop and fracture zone width need to be marked), rock fracture zones, karst development areas or caves, coal seam bifurcation lines or scour zones, old goaf boundaries, etc.

[0025] The bilateral establishment module: constructs physical edges by corresponding the first node and the second node one-to-one, and establishes causal edges between the second node and the third node; It should be noted that the physical edge is a pair of directed edges with opposite directions established between the first node and the second node based on the spatial coordinates of each first node. This pair of directed edges represents a two-way information relationship between the first node and its corresponding second node. An anomaly in the first node means an increase in risk within the second node, and the overall risk status of the second node also affects the interpretation of the anomaly level of the first node within the second node.

[0026] Causal edges are established between two second nodes and between a second node and a third node, using a 12-hour sliding time window, moving forward one hour at a time. Within each window, a bivariate Granger causality test is performed on the associated monitoring data of the two nodes. If the test results indicate that the past values ​​of the monitoring data of node j statistically significantly contribute to explaining the current values ​​of the monitoring data of node i (i.e., node j is a Granger cause of node i), then a causal edge is established in the graph from j to i. For the second node, its monitoring data is a composite sequence of standardized data from all first nodes within the region, weighted by area; for geological structure nodes, its monitoring data combination is the mean of the composite sequences of its adjacent second nodes.

[0027] The bivariate Granger causality test is a statistical hypothesis testing method based on time series forecasting, used to determine whether a statistically significant causal relationship exists between two variables. Its core idea is: if predicting the future value of Y using the past values ​​of both variables X and Y is significantly more effective than predicting the future value of Y using only its own past values, then X is said to be a Granger cause of Y. In this application, this test is used to analyze the combined sequences of monitoring indicators for any two second nodes or the second and third nodes. If the probability corresponding to the test result's statistic is less than 0.01, a directed causal edge is established between the two, thereby automatically constructing a time-series causal relationship network between the monitoring variables through a data-driven approach.

[0028] The weight component calculation module is used to evaluate the weight components of physical edges and causal edges and perform linear superposition to form a causal connection graph. The specific steps are as follows: Calculate the spatial distance between the centers of the nodes at both ends of the physical edge and the rock quality index. Based on the spatial distance between the centers of the nodes at both ends and the rock quality index (taking the value of the worse end between the two ends), determine the weight components of the physical edge. When the spatial distance is less than 20 meters and the rock quality index is greater than 75%, the weight component of the physical edge is 0.8. When the spatial distance is between 20 and 50 meters, or the rock quality index is between 50% and 75%, the weight component of the physical edge is 0.5. When the spatial distance is between 50 meters and 100 meters, or the rock quality index is between 25% and 50%, the weight component of the physical edge is taken as 0.3. When the spatial distance is greater than 100 meters, or the rock quality index is less than 25%, the weight component of the physical edge is set to 0.1.

[0029] Take a sliding time window of fixed length T, obtain the monitoring data of the nodes at both ends of the causal edge within the sliding time window from the first dataset, and calculate the absolute value of the Pearson correlation coefficient of the nodes at both ends of the causal edge based on the monitoring data. Perform Granger causality tests within a sliding time window to obtain statistics. Sort the statistics corresponding to all causal edges and calculate the causal strength based on the maximum and minimum values ​​of the statistics. The absolute value of the Pearson correlation coefficient and the causal strength are combined:

[0030] In the formula: The absolute value of the Pearson correlation coefficient. For Granger statistic, This is the balance coefficient, with a value of 0.5.

[0031] The weight components of the causal edges are obtained, and then the weight components of the physical edges and the causal edges are linearly superimposed to obtain the edge weights and the causal connection graph.

[0032] Then, the weights of the physical edges and causal edges are linearly superimposed, specifically as follows: ;; In the formula: , The contribution coefficients are set to 0.3 and 0.7. The risk propagation resonance calculation unit includes a signal diagnosis module and a risk energy mapping module, wherein: The signal diagnosis module decomposes the first dataset, calculates the risk score of the first node at the current moment, generates a transmission weight, and updates the causal connection graph using the transmission weight. The specific steps are as follows: The monitoring data of the first node is obtained from the first dataset. The monitoring data of the first node is smoothed to obtain the first component. The monitoring data of the first node is subtracted from the first component to obtain the second component and the instantaneous main frequency is calculated. It should be noted that the monitoring data of the first node is smoothed by using an exponentially weighted moving average as a low-pass filter. The smoothing coefficient of the exponentially weighted moving average is set to 0.1, which makes the trend component more sensitive to recent data and can track the slow drift of the parameters.

[0033] The trend score and transient score are calculated based on the first and second components. The specific steps are as follows: Calculate the mean of the first component, then calculate the absolute value of the deviation of the first component from the mean, and compare it with the historical standard deviation of the first component. If the deviation exceeds three times the historical standard deviation of the first component, the trend score is the value of the deviation divided by three times the historical standard deviation. If it does not exceed the historical standard deviation, the trend score is 0. The transient score is the absolute value of the second component. If the absolute value of the second component exceeds the standard deviation of the second component, then the absolute value is divided by three times the standard deviation of the second component. If it does not exceed the standard deviation, the transient score is 0. The risk score is obtained by weighted summation of the trend score and the transient score. The weights for trend score and transient score are 0.4 and 0.6, respectively. For each causal edge, the propagation weight at the current moment is calculated one by one. The calculation process is as follows: ;; In the formula: Let be the instantaneous dominant frequency of the source node of the causal edge at the current moment. The optimal resonant propagation frequency for this causal edge is... This represents the frequency response bandwidth parameter of the causal edge; Connect the edge weight with Multiply to obtain the propagation weight. .

[0034] The risk energy mapping module: calculates the external energy value based on the updated causal connection graph and performs coupling resonance, updates the risk energy value of the previous time step, and then performs mapping, including: By combining the risk score with the transmission weight, the external energy value is obtained: ;; In the formula: The time delay of the transmission path is calculated using the following formula: , V represents the spatial distance between the centers of the two nodes, and v represents the average longitudinal wave velocity of the mine rock mass, typically taken as 3,500 meters per second.

[0035] Resonant nodes are selected from all nodes for coupled resonance calculation to obtain resonance intensity values. As a specific implementation method, the above specific steps are as follows: for any node i, whether there are two or more first nodes whose risk scores have reached a local peak within the same sampling interval among all other nodes pointing to i; if so, they are considered as resonant nodes. Apply a positive lower bound to each resonant node: ; In the formula: Risk score for resonant nodes. It is a positive lower limit constant, with a value of 10. -6 .

[0036] The formula for calculating coupled resonance is: ;; In the formula: The number of resonant nodes. This is the resonance amplification factor, with a value of 1.5; The update is based on the risk energy value from the previous moment, combined with the external energy value and the resonance intensity value: ;; In the formula: The risk energy value at the previous moment. It is an exponential decay factor. The attenuation coefficient is... The time step is 60 seconds. Wherein, the attenuation coefficient is , The value is 2h, which means that in the absence of continuous abnormal stimuli in the mine, the previously accumulated risk energy decays to half of its peak value within two hours and to one-quarter after four hours. This decay rate is consistent with the engineering experience in the field of mine safety regarding the persistence of abnormalities, and is neither too sluggish nor too sensitive.

[0037] The updated risk energy value is mapped using the hyperbolic tangent function, and the maximum value of the local anomaly mapping value is used to obtain the comprehensive risk vector.

[0038] ;; In the formula: It is a hyperbolic tangent function with an output range of 0 to 1. This function grows approximately linearly in the low-energy region and gradually saturates in the high-energy region, which conforms to the marginal diminishing law of risk perception.

[0039] This is a local anomaly mapping value. , It is the 95th percentile value of all pulse intensities of the first node throughout the entire historical operation.

[0040] The disaster adaptive fusion unit includes a disaster probability calculation module and a graded early warning module, wherein: The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, and calculates the disaster triggering matrix and the coupling strength of various disasters, including: Determine the set of nodes associated with each type of disaster. Specifically, for the indicators monitored in the first node that are directly related to the disaster (e.g., a gas sensor corresponds to a gas disaster), calculate the disaster field strength at the center coordinates of the second node. ;; In the formula: This represents the risk value of the node at the current moment. The center coordinates of the current second node. Spatial coordinates of the field source node For spatial influence kernel function, It is a very small constant, with a value of 10. -6 ; ;; In the formula: d is the Euclidean distance between two points in space. This parameter represents the characteristic radius of the disaster's impact, and its determination is based on the physical impact range and engineering experience of each disaster type: 25 meters for roof collapse disasters, corresponding to the main impact radius of roof fall; 40 meters for gas disasters, corresponding to the diffusion impact range after abnormal gas outburst; 50 meters for water hazards, corresponding to the rapid expansion range of water flow after a water inrush; and 30 meters for rockburst disasters, corresponding to the effective damage radius of the shock stress wave. This parameter can be calibrated once by a safety engineer based on actual geological conditions during the initial stage of mine operation.

[0041] The types of disasters include: roof collapse disasters, gas disasters, water hazards, rock bursts, and fires.

[0042] The proportion of dangerous areas is selected based on the disaster field strength. Specifically, a field strength threshold of 0.3 is set, all second nodes with a field strength greater than the field strength threshold are counted, and the proportion of dangerous areas is obtained by dividing the second node by the total number of second nodes.

[0043] The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, calculates the disaster triggering matrix and the coupling strength of various disasters, and also includes: For each type of disaster, combinations are performed, and the transfer intensity of each combination is calculated: ;; In the formula: , They are disaster k and disaster l, respectively. Let be the comprehensive risk vector of source node i. It is a very small constant, with a value of 10. -6 .

[0044] The disaster triggering matrix is ​​obtained by normalizing the sum of squares of the transfer intensity, as follows: ;; In the formula: This is a regularization term with a value of 0.01.

[0045] For all K types of disasters, a disaster triggering matrix with K rows and K columns is formed; Determining the coupling strength of various disasters based on the disaster triggering matrix: .

[0046] The graded early warning module generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warnings based on the final risk index.

[0047] The graded early warning module generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warnings based on the final risk index.

[0048] The historical baseline weights, spatial attention factor, and triggering attention factor were calculated separately: Historical benchmark weights: ;; In the formula: The information entropy of a disaster is calculated by determining the risk score of the disaster over the last thousand time steps and determining the probability of it falling within each numerical range. This represents the sum of all disaster types.

[0049] Spatial attention factor: ;; Attention triggering factors: ;; In the formula: The spatial attention temperature parameter has a value of 2.0. Indicates the percentage of dangerous areas. This is the parameter to trigger attention. Its value is 3.0. A non-steady-state index is defined. The risk energy values ​​of all nodes are sorted in descending order, and the top 20% of nodes are selected as the standard node set. For each node in the standard node set, the absolute value of the rate of change of its risk energy value between two consecutive time steps is calculated, and then the average is calculated. ;; In the formula: The number of standard nodes. , Let J represent the risk energy values ​​of node j at the current time and the previous time. The non-steady-state threshold is selected during initialization by calculating the non-steady-state index at each time step within a continuous period of normal conditions, taking its arithmetic mean, multiplying it by three, and then updating it once a day during operation to maintain adaptability to changes in normal operating conditions.

[0050] Multi-dimensional adaptive weights: ;; In the formula: As a balance factor, For allocation coefficients, The historical benchmark weight.

[0051] When the unsteady-state index is less than the unsteady-state threshold, it is judged to be in stable operation, the balance factor is set to 0.7, and the distribution coefficient is set to 0.6. When the unsteady-state index is greater than or equal to the unsteady-state threshold, it is judged as unsteady operation, the balance factor is set to 0.3, and the allocation coefficient is set to 0.4. The formula for calculating the final risk index is: First, calculate the local peak intensity: ;; In the formula: Let k be the set of nodes for disaster k. Let the disaster field strength of disaster k be located at the center of the second node. This is a weighting factor, with a value of 0.4; ;; The final risk index is selected from the 30 most recent normal days. The 60th, 85th, and 95th percentiles were used as thresholds. , , .

[0052] Final Risk Index This is considered normal; Final Risk Index If so, continued monitoring is needed, and if the final risk index... If so, an early warning will be issued.

[0053] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0054] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A mine safety monitoring data fusion analysis and hierarchical early warning system, characterized in that, It includes a monitoring data processing unit, a safety dynamic correlation unit, a risk propagation resonance calculation unit, and a disaster adaptive fusion unit, among which: The monitoring data processing unit: collects data from each sensor in real time, cleans and aligns the data, and then performs standard transformation on the data. The safety dynamic association unit: Based on the engineering layout and geological conditions of the mine, the mine is converted into multiple types of nodes. On the basis of multiple types of nodes, physical edges and causal edges between nodes are established. The weight components of physical edges and causal edges are evaluated and weighted and fused to form a causal connection graph. The risk propagation resonance calculation unit: scores and weights each sensor, updates the causal connection graph, and uses the updated causal connection graph to calculate and map the risk energy value of all nodes; The disaster adaptive fusion unit: constructs a disaster potential energy field, calculates the coupling strength of the disaster trigger matrix and various disasters, generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warning based on the final risk index.

2. The mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 1, characterized in that, The monitoring data processing unit includes a data acquisition module and a data preprocessing module, wherein: The data acquisition module is used to collect sensor data collected in real time from multiple types of sensors, package the collected sensor data into a sensor data package, and upload it. The data preprocessing module receives sensor data packets uploaded by the data acquisition module, cleans the sensor data packets, performs time alignment and standardization after cleaning, and obtains the first dataset.

3. The mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 1, characterized in that, The secure dynamic association unit includes: a node definition module, a bilateral establishment module, and a weight component calculation module, wherein: The node definition module: Based on the engineering layout and geological conditions of the mine, each sensor is recorded as the first node, the underground engineering space of the mine is divided into several regions, a second node is established for each region, and a geological structure node is established for the geological anomaly. The bilateral establishment module: constructs physical edges by corresponding the first node and the second node one-to-one, and establishes causal edges between the second node and the third node; The weight component calculation module is used to evaluate the weight components of physical edges and causal edges and perform linear superposition to form a causal connection graph.

4. The mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 3, characterized in that, The weight component calculation module is used to evaluate the weight components of physical edges and causal edges and perform linear superposition to form a causal connection graph. The specific steps are as follows: Calculate the spatial distance between the centers of the nodes at both ends of the physical edge and the rock quality index, and determine the weight components of the physical edge based on the spatial distance between the centers of the nodes at both ends and the rock quality index. Take a sliding time window of fixed length T, obtain the monitoring data of the nodes at both ends of the causal edge within the sliding time window from the first dataset, and calculate the absolute value of the Pearson correlation coefficient of the nodes at both ends of the causal edge based on the monitoring data. Perform Granger causality tests within a sliding time window to obtain statistics. Sort the statistics corresponding to all causal edges and calculate the causal strength based on the maximum and minimum values ​​of the statistics. The absolute value of the Pearson correlation coefficient and the causal strength are fused to obtain the weight components of the causal edges. Then, the weight components of the physical edges and the causal edges are linearly superimposed to obtain the edge weights and the causal connection graph.

5. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 4, characterized in that, The risk propagation resonance calculation unit includes a signal diagnosis module and a risk energy mapping module, wherein: The signal diagnosis module: decomposes the first dataset, calculates the risk score of the first node at the current moment, generates a transmission weight, and updates the causal connection graph through the transmission weight; The risk energy mapping module calculates the external energy value based on the updated causal connection graph and performs coupling resonance, updates the risk energy value of the previous moment, and then performs mapping.

6. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 5, characterized in that, The signal diagnosis module decomposes the first dataset, calculates the risk score of the first node at the current moment, generates a transmission weight, and updates the causal connection graph using the transmission weight. The specific steps are as follows: The monitoring data of the first node is obtained from the first dataset. The monitoring data of the first node is smoothed to obtain the first component. The monitoring data of the first node is subtracted from the first component to obtain the second component and the instantaneous main frequency is calculated. The trend score and transient score are calculated based on the first and second components. The trend score and transient score are then weighted and summed to obtain the risk score. For each causal edge, the propagation weight at the current moment is calculated one by one. The calculation process is as follows: ； In the formula: Let be the instantaneous dominant frequency of the source node of the causal edge at the current moment. The optimal resonant propagation frequency for this causal edge is... This represents the frequency response bandwidth parameter of the causal edge; Connect the edge weight with Multiply them to obtain the transmission weight.

7. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 6, characterized in that, The risk energy mapping module: calculates the external energy value based on the updated causal connection graph and performs coupling resonance, updates the risk energy value of the previous time step, and then performs mapping, including: By combining the risk score with the transmission weight, the external energy value is obtained. Resonant nodes are selected from all nodes for coupled resonance calculation to obtain the resonance intensity value. Based on the risk energy value of the previous moment, the risk energy value is updated by combining the external energy value and the resonance intensity value. The updated risk energy value is then mapped using the hyperbolic tangent function to obtain the comprehensive risk vector.

8. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 1, characterized in that, The disaster adaptive fusion unit includes a disaster probability calculation module and a graded early warning module, wherein: The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, and calculates the disaster triggering matrix and the coupling strength of various types of disasters; The graded early warning module generates multi-dimensional adaptive weights and sets trigger thresholds, calculates the final risk index, and performs graded early warnings based on the final risk index.

9. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 8, characterized in that, The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, and calculates the disaster triggering matrix and the coupling strength of various disasters, including: Determine the set of nodes associated with each type of disaster, and calculate the disaster field strength at the center coordinates of the second node: ；； In the formula: This represents the risk value of the node at the current moment. The center coordinates of the current second node. Spatial coordinates of the field source node For spatial influence kernel function, It is a very small constant; The proportion of dangerous areas is selected based on the intensity of the disaster field.

10. A mine safety monitoring data fusion analysis and hierarchical early warning system according to claim 9, characterized in that, The disaster probability calculation module: determines the set of nodes associated with each type of disaster, calculates the disaster field strength at the center coordinates of the second node and constructs the disaster potential energy field, calculates the disaster triggering matrix and the coupling strength of various disasters, and also includes: For each type of disaster, the transfer intensity of each combination is calculated. The transfer intensity is then normalized by the sum of squares to obtain the disaster triggering matrix. The coupling strength of various disasters is determined based on the disaster triggering matrix.

Images