Multi-source information real-time identification and positioning method for coal mine rock burst risk area

Abstract

The invention discloses a multi-source information real-time identification and positioning method for a coal mine rock burst risk area. The method comprises the following steps: integrating a micro-seismic system, a rock noise system, a stress system and a geological system to construct an acquisition network; extracting multi-source characteristic indexes and carrying out space-time unification; establishing a dynamic weighted risk fusion model to calculate a comprehensive risk index RI of a three-dimensional grid unit, wherein the weight of the comprehensive risk index RI is dynamically and adaptively adjusted along with mining; based on the RI value, a high-risk area is identified and positioned through three-dimensional visualization, and graded early warning information is generated; according to the invention, a time-space synchronous multi-source information acquisition network is constructed, and uniform preprocessing and feature extraction are carried out on four types of data of microseism, rock noise, stress and geologic structure, so that deep fusion and collaborative analysis of multi-source information are realized, and an information island phenomenon existing in traditional monitoring is overcome; and the judgment on the rock burst risk is more comprehensive.

Description

Technical Field

[0001] This invention relates to the field of rockburst disaster prevention and control technology, and in particular to a method for real-time identification and location of multi-source information in coal mine rockburst risk areas. Background Technology

[0002] Coal mine rock bursts are a complex dynamic disaster in coal mining, characterized by their suddenness and destructive power, which seriously threaten mine safety and the lives of personnel. In order to achieve effective early warning of rock bursts, the industry has widely adopted a variety of technical means, including microseismic monitoring, ground sound monitoring, and stress monitoring, to identify potential risk areas by real-time monitoring of the physical state of coal and rock masses.

[0003] However, existing monitoring technologies often operate independently, forming "information silos." Data such as microseismic activity, ground acoustics, and stress lack effective synchronization and fusion mechanisms in time and space, making comprehensive risk assessment difficult. Existing analysis methods are mostly static or single-weighted superpositions, unable to adaptively adjust the weights of different monitoring parameters according to dynamic changes in the mining process (such as the working face approaching geological structures), resulting in insufficient sensitivity and accuracy in risk assessment, often leading to false alarms or missed alarms. Furthermore, traditional risk visualization is mostly limited to two-dimensional planes or single-parameter visualization, making it difficult to intuitively and three-dimensionally present the specific location, spatial form, and evolution trend of risks underground, which is not conducive to guiding precise prevention and control measures.

[0004] Therefore, in response to the problems mentioned above, this invention proposes a method for real-time identification and location of multi-source information in coal mine rockburst risk areas. Summary of the Invention

[0005] To overcome the problems of insufficient multi-source information fusion, rigid risk assessment models, and unintuitive risk location display in existing technologies, this invention proposes a real-time multi-source information identification and location method for coal mine rockburst risk areas. By constructing a spatiotemporally synchronized acquisition network, establishing a dynamically weighted risk fusion model, and visually locating and displaying the risk in three-dimensional space, the method achieves accurate and efficient real-time identification and location of rockburst risk areas.

[0006] The technical solution of this invention is: a method for real-time identification and location of multi-source information in coal mine rockburst risk areas, comprising the following steps: S1, in the monitoring area underground in the coal mine, deploys and integrates a microseismic monitoring system, a ground sound monitoring system, a coal mine stress online monitoring system and a geological structure exploration system to form a spatiotemporally synchronized multi-source information acquisition network; The microseismic monitoring system is used to collect high-energy and low-frequency vibration event signals. Its sensor array covers the mining face and its surrounding area within a range of at least 200 meters. The sensor spacing is set to 100-200 meters. A three-component detector is used, and the lowest frequency response is not less than 1Hz. The ground sound monitoring system is used to collect low-energy and high-frequency rock fracture signals. Its sensors are deployed in a high density in the roadway and tunnel of the mining face. The monitoring range focuses on the mining stress concentration area. The spacing between the ground sound sensors is set to 50-100 meters, the upper limit of the effective frequency band is not less than 1500Hz, and the anchor bolts are connected to the roof or side rock mass using anchoring agent. The online stress monitoring system for coal mines is used to collect dynamic stress change data inside the coal and rock mass. It includes a borehole stress gauge and a hydraulic support pressure sensor. The borehole stress gauge is installed in a deep borehole in the coal body in front of the working face and the roadway side. Its installation depth is 12-25 meters. It is arranged at a spacing of 20-40 meters in the advance support range in front of the working face. The real-time sampling interval is no more than 5 minutes. The geological structure exploration system is based on existing geological exploration data, downhole channel seismic data and radio wave perspective data, and constructs a geological model that includes faults, folds, coal seam thickness variations and roof and floor lithological boundaries. S2 preprocesses and extracts features from real-time multi-source data to obtain a spatiotemporally unified set of feature indicators. For microseismic and ground sound signals, filtering, noise reduction, and source location processing are performed. Event energy, event frequency, energy rate, apparent volume, b-value, and event spatial clustering are extracted as feature indicators. The event spatial clustering is estimated using a kernel density algorithm with a search radius of 50 meters, and the number of events per cubic kilometer per 24 hours is calculated as the density value. Simultaneously, for coal mine stress data, stress increment, stress gradient, and stress concentration factor are calculated as feature indicators. The feature index set is obtained by calculating the rate of change of the stress gauge monitoring values ​​of two adjacent boreholes over a unit distance, and then mapping all the extracted feature indicators to the same three-dimensional coordinate system and the same timestamp based on the mining engineering plan. S3. Based on the obtained feature index set, a dynamically weighted multi-source information risk fusion model is established. This model calculates the comprehensive risk index RI of each three-dimensional grid cell within a specific time window in the following way: the size of the three-dimensional grid cell is set to 10 meters × 10 meters × 10 meters, and the time window is 2-4 hours. The expression is as follows: ; in, , , , These are risk sub-functions based on microseismic, geophonic, stress, and geological structure information; EMS and LMS represent the energy and location characteristics of microseismic events, respectively; EAE and LAE represent the energy and location characteristics of geophonic events, respectively; , These represent stress increment and stress gradient, respectively. Representative geological structural influencing factors; , , It is a dynamic weighting coefficient that changes with time t, used to reflect the changes in the dominance of different monitoring methods during the mining process; It is the static weighting coefficient of geological structure; The dynamic weighting coefficient , , The system is adaptively adjusted based on the relative relationship between the current mining location and the geological structure, and the correlation between recent microseismic and geosonic activity. The initial weights are set to... =0.3, =0.3, =0.3, =0.1, and satisfies + + + =1; S4. The calculated three-dimensional comprehensive risk index RI of the entire monitoring area is spatially visualized. By setting a risk threshold, areas where the RI exceeds the threshold are identified as high-risk areas for rockburst. The risk threshold is set to 0.6 and can be adjusted within the range of 0.5-0.7 according to the specific conditions of the mine. The data is then rendered and displayed in real time on the three-dimensional geological model and the mining engineering plan. S5 generates graded early warning information based on the spatial location, size, and risk level of the identified high-risk areas, combined with the mining operation plan, and pushes it to the underground workers and the ground dispatch center to guide the implementation of pressure relief or area avoidance measures.

[0007] Preferably, in step S1, the geological model constructed by the geological structure exploration system quantifies and assigns values ​​to structural anomaly areas to generate geological structure influence factors. Among them, faults The value is determined based on its elevation drop, attitude, and distance from the mining face. The formula for calculating the value is: ; Where k is the fault property coefficient (1.2 for reverse faults and 0.8 for normal faults), H is the fault displacement (meters), and L is the normal distance between the monitoring point and the fault (meters). This item is included in the calculation when L is less than 50 meters.

[0008] Preferably, in step S2, the event spatial clustering degree is obtained by calculating the number of earthquake or ground sound events occurring per unit volume and per unit time, and using the kernel density estimation method for spatial smoothing. The kernel function is a Gaussian kernel, and the bandwidth parameter is adaptively selected according to the event distribution density of the monitoring area, ranging from 30 to 80 meters.

[0009] Preferably, in step S3, the dynamic weighting coefficient , , The adaptive adjustment rule is as follows: when the mining face is close to the main geological structure, the weight of the microseismic risk sub-function is automatically increased. and geological structure weight For example, when the working face advances to within 100 meters of the fault, Increase by 0.1-0.2 from the initial value; When a sharp increase in the frequency of ground sound events is detected while microseismic events are relatively calm, the weight of the ground sound risk subfunction is automatically increased. For example, when the frequency of local acoustic events increases by more than 300% within 2 hours while the microseismic energy does not change significantly, Increase the initial value by 0.15; When the borehole stress gauge detects the stress change gradient As the stress risk sub-function continues to increase, its weight is automatically increased. For example, when When the pressure exceeds 1 MPa / m for three consecutive sampling cycles, Increase the initial value by 0.1.

[0010] Preferably, in step S3, the risk function , , A machine learning model trained on historical rockburst case data is used. This machine learning model can be a random forest, gradient boosting tree, or neural network model. Its training features include at least the total energy, maximum energy, event frequency, b-value, stress increment and its rate of change of the past 24 hours. The output is a risk probability value between 0 and 1.

[0011] Preferably, the input features of the machine learning model also include mining process parameters, including but not limited to working face advance speed, mining height and roadway excavation progress, wherein the working face advance speed is in meters per day, and is input into the model as a time series feature and monitoring data.

[0012] Preferably, in step S4, the risk threshold is not a fixed value, but a floating threshold that is adjusted according to the mining stage, roof management method, and the effect of the pressure relief measures already taken. In the initial pressure stage, the threshold is lowered by 0.1 to improve sensitivity. In areas where large-scale pressure relief blasting has been implemented, the threshold can be temporarily raised by 0.05-0.1 to reduce interfering alarms.

[0013] Preferably, in step S4, the three-dimensional dynamic positioning display uses different colors and transparency to represent different risk levels. The risk level is divided according to the RI value, where 0.6-0.75 is yellow (medium risk), 0.75-0.9 is orange (high risk), and greater than 0.9 is red (danger level). It can also display the spatiotemporal evolution of the risk area in the form of time-series animation.

[0014] Preferably, in step S5, the graded early warning information is divided into at least three levels: "Attention", "Early Warning", and "Alarm". Specific response procedures and handling measures are specified for each level. The "Attention" level (RI=0.5-0.6) requires the team leader to strengthen patrols; the "Early Warning" level (RI=0.6-0.75) requires limiting personnel and developing a pressure relief plan; and the "Alarm" level (RI>0.75) requires immediately stopping operations, evacuating personnel, and implementing directional pressure relief blasting. These handling measures include strengthening monitoring, limiting personnel, stopping operations, and implementing directional pressure relief blasting.

[0015] The beneficial effects of this invention are: 1. This invention constructs a spatiotemporally synchronized multi-source information acquisition network and performs unified preprocessing and feature extraction on four types of data: microseismic, ground sound, stress, and geological structure. This achieves deep fusion and collaborative analysis of multi-source information, overcomes the information silo phenomenon in traditional monitoring, and makes the assessment of rockburst risk more comprehensive.

[0016] 2. This invention introduces a dynamically weighted risk fusion model and incorporates a weight coefficient rule that adaptively adjusts based on the correlation between mining progress, geological conditions, and monitoring parameters. This enables the risk assessment model to flexibly respond to the dynamic changes of the dominant risk factors during the mining process, overcoming the rigidity of traditional static or fixed-weight models. As a result, it significantly improves the sensitivity and accuracy of risk warnings and reduces false alarms and missed alarms.

[0017] 3. This invention calculates and visualizes the comprehensive risk index in a three-dimensional spatial grid, and uses color and transparency for multi-level rendering and dynamic display. This enables the precise mapping of rockburst risk areas from abstract data to three-dimensional and intuitive spatial locations, making the shape, range and evolution trend of high-risk areas very clear. It provides clear and efficient decision support for pressure relief and personnel evacuation in the well, and greatly improves the intuitiveness and operability of risk management. Attached Figure Description

[0018] Figure 1 The diagram shown is a schematic representation of the method flow of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 This invention provides an embodiment of a method for real-time identification and location of multi-source information in coal mine rockburst risk areas: S1. Three-component geophones are deployed at intervals of 100-200 meters in the main roadways, rock gates and chambers of the underground mining area. The three-component geophones should be installed in the borehole and tightly coupled to the surrounding rock using a special coupling agent to ensure good reception of low-frequency vibration signals of 1-200Hz. The entire array can cover the mining face and its surrounding area of ​​at least 200 meters to ensure effective capture and location of high-energy vibration events.

[0021] Acoustic emission sensors are deployed at high-density intervals of 50-100 meters in the roadways (transport roadways and return airway) of the mining face. These sensors are typically installed directly on the roof or sides of the roadway using anchor bolts and anchoring agents to ensure integrated vibration with the coal and rock mass, effectively receiving high-frequency rock fracture signals of 100-1500Hz. Their monitoring range focuses on the mining-induced stress concentration zone within approximately 150 meters in front of the working face.

[0022] Hydraulic borehole stress gauges are installed in the two sides and middle of the roadway ahead of the working face, with a drilling depth of 12-25 meters. These stress gauges are arranged at intervals of 20-40 meters within the support area to monitor the dynamic changes of stress inside the coal body in real time, with a sampling interval of 2-5 minutes.

[0023] In the longwall mining face, one pressure sensor is installed on every 10 hydraulic supports to monitor changes in roof pressure and indirectly reflect the distribution of support pressure.

[0024] This system primarily relies on preliminary exploration results. It collects and integrates mine geological exploration reports, 3D seismic interpretation results, and underground channel seismic and radio wave perspective data. Using professional geological modeling software, it constructs a refined 3D geological model containing precise coordinates of faults, folds, coal seam floor contour lines, coal seam thickness contour lines, and roof and floor lithology information.

[0025] All monitoring substations must be connected to a unified precision clock signal to ensure that the synchronization error of the acquisition timestamps for microseismic, ground sound, and stress data is less than 1 millisecond.

[0026] S2. In the data preprocessing stage, for microseismic / ground sound signals, the original waveform data is first formatted and bad channels are removed. Then, bandpass filtering (microseismic: 1-200Hz, ground sound: 100-1500Hz) is performed to remove power frequency interference and random noise. The polarization analysis algorithm is used to automatically identify and pick the arrival times of valid P-waves and S-waves. Finally, the source location algorithm of the double-difference positioning method is used to calculate the spatiotemporal source parameters (occurrence time, X / Y / Z coordinates) of each event. Outlier removal and moving average filtering are performed on the original stress readings to smooth random fluctuations. The stress values ​​are converted to megapascals (MPa). Finally, the three-dimensional geological model is unified to the same coordinate system as the mining engineering plan.

[0027] In the feature extraction stage, event energy, event frequency, energy rate, apparent volume, b-value, and event spatial clustering degree calculated using the kernel density estimation method (with a search radius set to 50 meters and the number of events per unit cubic kilometer per 24 hours as the density value) are extracted from microseismic and geosound events as feature indicators. Stress increment, stress change gradient (obtained by the rate of change of stress gauge monitoring values ​​in adjacent boreholes over a unit distance) and stress concentration factor are calculated from stress data as feature indicators. Finally, all extracted feature indicators are uniformly mapped to the same three-dimensional coordinate system and the same timestamp based on the mining engineering plan, ensuring that the time synchronization error of all data is no more than 1 second, thus forming a spatiotemporally aligned feature indicator set for subsequent model use.

[0028] The kernel density estimation method used to extract spatial clustering specifically includes: defining a three-dimensional Gaussian kernel function centered on the location point of each microseismic or geosound event, setting the search radius (bandwidth) to 50 meters, calculating the sum of the kernel function values ​​of all events falling within each 10m×10m×10m grid cell, and then dividing by the time window. (24 hours) and unit volume (1 km³), finally obtaining the event space clustering degree of this grid cell. The formula can be simplified to: where K is the kernel function; in It is the event space clustering degree calculated at the three-dimensional spatial coordinate point (x, y, z); For time windows; per unit volume; The three-dimensional straight-line distance from the currently being calculated point (x, y, z) to the location of the source of the i-th microseismic / geothermal event; For bandwidth.

[0029] S3. To establish a dynamically weighted risk fusion model that can adapt to changes in operating conditions, it is first necessary to construct four risk functions. , , and These functions are implemented using machine learning models trained on historical rockburst case data, with the inputs being the energy and location characteristics of microseismic events, respectively. ), Earth sound event energy and location characteristics ( ), stress increment and stress gradient ( ) and geological structural influencing factors ( The output is a normalized risk probability value between 0 and 1, which transforms characteristic indicators of different physical dimensions into fusionable risk contribution values. Then, dynamic weighting coefficients need to be set and managed, where the geological structure weight δ is fixed at 0.1 as a static coefficient, while the microseismic weight... Earth tone weight and stress weight The initial values ​​are all set to 0.3, and satisfy the following condition: + + + The constraint condition is 1. These dynamic weights are adaptively adjusted according to predefined rules. For example, when the working face advances to within 100 meters of the main fault (elevation > 5 meters), Automatically increase by 0.15 when the frequency of local acoustic events increases sharply by more than 300% within 2 hours while microseismic activity is calm. Automatically increases by 0.15 when the stress gradient monitored by the borehole stress gauge... When the pressure exceeds 1.0 MPa / m for three consecutive sampling periods The index is automatically increased by 0.1 to reflect the dynamic changes of the dominant risk factors during the mining process. Finally, based on the above function and weights, the comprehensive risk index RI of each three-dimensional grid cell (10m × 10m × 10m) is calculated within a specific time window (2-4 hours). The calculation formula is as follows: The calculation is performed iteratively across all grids in the entire monitoring area, ultimately generating a three-dimensional risk field that can be accurately quantified and dynamically evolves with space and time.

[0030] S4 imports the calculated three-dimensional comprehensive risk index (RI) data of the entire monitoring area into the visualization engine. The system automatically scans and identifies the entire three-dimensional space using a preset risk threshold (the base value is set at 0.6, and can be adjusted within the range of 0.5-0.7 according to the specific conditions of the mine). The system will identify all three-dimensional grid cells with RI values ​​greater than or equal to the threshold, and automatically identify and delineate these spatially connected high-risk grid clusters as "high-risk areas for rockburst", thus completing the transformation from data to risk objects.

[0031] The system then overlays and renders these identified high-risk areas onto the 3D geological model and mining engineering plan. To convey the risk level intuitively, a set of color and transparency coding rules is used. Areas with an RI value between 0.6 and 0.75 are displayed as yellow (medium risk), areas with an RI value between 0.75 and 0.9 are displayed as orange (high risk), and areas with an RI value exceeding 0.9 are displayed as red (danger). The higher the risk level, the higher the opacity of the display, thus achieving a prominent visual warning effect in the 3D scene.

[0032] The system also provides powerful spatiotemporal evolution retrospective and preview functions, which can dynamically display the complete spatiotemporal evolution process of these high-risk areas in the form of time series animations, from their generation, development, migration and even disappearance in the past few hours or days. This transforms the abstract numerical calculation results into a three-dimensional, visual and traceable "risk cloud map", ultimately achieving intuitive positioning of rockburst risk areas in three-dimensional space.

[0033] S5. Based on the spatial location, three-dimensional range, and comprehensive risk index (RI) level of the identified high-risk areas, a graded early warning response mechanism is automatically generated and executed. When the system determines that an area is at the "Attention" level (RI value between 0.5 and 0.6), the system will prominently mark the area on the monitoring interface and automatically send a prompt message to the inspection personnel and team leaders in that area, requiring them to increase the frequency and intensity of inspections for macroscopic phenomena such as sound, sidewall spalling, or floor heave in the area, without changing the normal work plan. When the risk level rises to the "Warning" level (RI value between 0.6 and 0.75), the system will automatically generate a detailed early warning report. The report will clearly state the spatial coordinates, impact range, and main risk factors (such as active microseismic activity, stress exceeding limits, or geophone emissions) of the high-risk area. Through linkage with the mine dispatch system and personnel positioning system, mandatory management of limited personnel operations will be implemented in the area. At the same time, the anti-rockfall department will be notified to immediately formulate a targeted pressure relief plan based on the report information and prepare for its implementation. Once the risk level reaches the "alarm" level (RI value greater than 0.75), the system will immediately trigger the highest level of audio-visual alarm and automatically send emergency evacuation instructions to the personnel positioning cards of all personnel located in the risk area and potentially threatened areas. At the same time, the underground broadcast will be activated to continuously broadcast the alarm in the area and related roadways. The dispatch center must immediately order the cessation of all work activities in the area and affected areas and authorize the anti-impact team to carry out precise directional pressure relief blasting measures in the pre-identified risk core area.

[0034] Comparative Example 1 provided by the present invention: This embodiment designs a computer simulation experiment. The simulation environment is based on typical deep coal mine geology and mining conditions. The software environment used in this simulation experiment is a combination of COMSOL Multiphysics simulation software and a self-developed Python algorithm set. COMSOL is mainly used to build a detailed mine geomechanical model and simulate key physical processes such as stress field, displacement field changes and fault activation during mining. Python is used to simulate the occurrence of microseismic / earthquake events, generate stress data, and implement the fusion algorithm of this invention.

[0035] This simulation established a virtual mine model with a length of 2000 meters, a width of 1500 meters, and a vertical depth of 800 meters. The coal seam thickness is 5 meters, the dip angle is 15°, and two main faults are set: a reverse fault F1 with a drop of 8 meters and a normal fault F2 with a drop of 4 meters.

[0036] The simulated mining activity is set as a longwall longwall mining face with an advance length of 200 meters, advancing at a uniform rate of 5 meters per day, thereby simulating a continuous evolution process of mining stress field. In this embodiment, 20 microseismic sensors, 40 ground acoustic sensors, and 15 borehole stress gauges are virtually arranged in the model space.

[0037] This experiment simulates the entire process of the working face moving from away from the fault to crossing the fault. Two comparative examples are used. Comparative example 1 uses the traditional single microseismic monitoring method, which issues warnings based solely on the energy and frequency of microseismic events. A warning is issued when an event with energy greater than 1e5J or an hourly frequency greater than 10 occurs. Comparative example 2 uses the static weighted fusion method, employing the same multi-source data as this invention, but the weights α, β, γ, and δ in the risk fusion model are fixed values ​​(0.3, 0.3, 0.3, 0.1) and do not change with the working conditions.

[0038] This experiment simulated a scenario where the working face advanced to 150 meters from the F1 fault, and localized stress concentration and microfracture development occurred in the coal seam ahead. Ground acoustic monitoring showed that the frequency of these events increased from 5 times / hour to 25 times / hour within 4 hours. Stress gauges indicated localized... When the pressure reaches 0.8 MPa / m, the microseismic activity is calm. Please refer to the table below for details.

[0039]

[0040] As shown in the table above, the early warning rule of Comparative Example 1 relies on high-energy microseismic events or high frequencies. Since the microseismic events were calm and did not reach the single frequency threshold, no early warning was triggered. The static weight model of Comparative Example 2 could not make a specific response to the sudden activity of ground sounds, and the comprehensive risk index did not reach the early warning threshold of 0.6. The dynamic weight model of this invention automatically increases the ground sound weight β(t) from 0.3 to about 0.45 according to the ground sound frequency surge rule, amplifying the contribution of ground sound information, thereby significantly improving the comprehensive risk index and realizing early and clear early warning of risks.

[0041] Comparative Example 2 is provided in this invention: This experiment simulates the working face advancing to 50 meters from the F1 fault. Fault activation induces a series of high-energy microseismic scenarios. Three microseismic events with energies greater than 1e5J occur consecutively near the fault. Geosonous activity remains high, and stress continues to rise. For details, please refer to the table below.

[0042]

[0043] As shown in the table above, Comparative Example 1 successfully detected a high-energy microseismic event and triggered an early warning, but the warning area was limited to the vicinity of discrete event points, failing to clearly present the overall risk area. Comparative Example 2 fused multi-source information, and the risk index exceeded the threshold, but due to the fixed weights, it failed to particularly highlight the dominant role of tectonic and microseismic activity at this stage, resulting in insufficient sensitivity and level of risk assessment. This invention automatically adjusts the microseismic weights... The RI value increased from 0.3 to approximately 0.45. After fusion calculation, the RI value reached as high as 0.85 in the area surrounding the fault, triggering an "alarm" level. The high-risk area centered on the fault was clearly delineated on the 3D map, with a clear scope and a higher level.

[0044] Comparative Example 3 is provided in this invention: This experiment simulates a scenario where, after the working face pushes past the F1 fault, a risk scenario re-emerges in the footwall due to stress redistribution. After passing the fault, microseismic activity weakens, but sonic activity becomes active again, and stress gauges show rapid stress accumulation in the footwall coal seam. It reaches 1.2 MPa / m. Please refer to the table below for details.

[0045]

[0046] As shown in the table above, the early warning mechanism of Comparative Example 1 is highly dependent on high-energy microseismic events. Because microseismic activity weakens after passing through the fault, the system cancels the early warning and fails to identify new static load-type risks. In Comparative Example 2, the static weight model suffers from a decrease in the contribution of microseismic events, leading to a decline in the comprehensive risk index, which is only on the edge of "early warning" and fails to fully reflect the ongoing risks brought about by stress concentration and active ground sounds. This invention automatically increases the risk level based on the high stress gradient rule. Meanwhile, the active geophony also contributed a high weight, and the final RI value in the footwall region of the F1 fault was calculated to be 0.78, continuously issuing "early warnings" and accurately capturing the secondary risks after crossing the fault.

[0047] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for real-time identification and location of multi-source information in coal mine rockburst risk areas, characterized in that, It includes the following steps: S1. In the underground coal mine monitoring area, a microseismic monitoring system, a ground acoustic monitoring system, a coal mine stress online monitoring system, and a geological structure exploration system are deployed and integrated to form a spatiotemporally synchronized multi-source information acquisition network. The microseismic monitoring system is used to collect high-energy and low-frequency vibration event signals, and its sensor array covers the mining face and its surrounding area of ​​at least 200 meters. The ground acoustic monitoring system is used to collect low-energy and high-frequency rock fracture signals, and its sensors are deployed at high density in the mining face roadway and roadway, with the monitoring range focused on the mining stress concentration area. The coal mine stress online monitoring system is used to collect dynamic stress change data inside the coal and rock mass, and it includes borehole stress gauges and hydraulic support pressure sensors. The borehole stress gauges are installed in deep boreholes in the coal body in front of the roadway sidewall and the working face. The geological structure exploration system is based on existing geological exploration data, underground channel wave seismic and radio wave perspective data to construct a geological model including faults, folds, coal seam thickness changes, and roof and floor lithological boundaries. S2 preprocesses and extracts features from real-time multi-source data to obtain a spatiotemporally unified set of feature indicators. For microseismic and ground sound signals, filtering, noise reduction, and source location processing are performed to extract event energy, event frequency, energy rate, apparent volume, b-value, and event spatial clustering as feature indicators. At the same time, for coal mine stress data, stress increment, stress change gradient, and stress concentration coefficient are calculated as feature indicators. All extracted feature indicators are uniformly mapped to the same three-dimensional coordinate system and the same timestamp based on the mining engineering plan, forming a set of feature indicators. S3. Based on the obtained feature index set, a dynamically weighted multi-source information risk fusion model is established. This model calculates the comprehensive risk index RI of each three-dimensional grid cell within a specific time window in the following way: in, , , , These are risk sub-functions based on microseismic, geophonic, stress, and geological structure information; EMS and LMS represent the energy and location characteristics of microseismic events, respectively; EAE and LAE represent the energy and location characteristics of geophonic events, respectively; , These represent stress increment and stress gradient, respectively. Representative geological structural influencing factors; , , It is a dynamic weighting coefficient that changes with time t, used to reflect the changes in the dominance of different monitoring methods during the mining process; It is the static weighting coefficient of geological structure; The dynamic weighting coefficient , , Adaptive adjustments are made based on the relative relationship between the current mining location and the geological structure, as well as the correlation between recent microseismic events and geosound activity. S4. The calculated three-dimensional comprehensive risk index RI of the entire monitoring area is spatially visualized. By setting a risk threshold, areas where the RI exceeds the threshold are identified as high-risk areas for rockburst. The data is then rendered and displayed in real time on the three-dimensional geological model and the mining engineering plan. S5 generates graded early warning information based on the spatial location, size, and risk level of the identified high-risk areas, combined with the mining operation plan, and pushes it to the underground workers and the ground dispatch center to guide the implementation of pressure relief or area avoidance measures.

2. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S1, the geological model constructed by the geological structure exploration system quantifies and assigns values ​​to structural anomaly areas, generating geological structure influence factors. Among them, faults The value is determined based on its elevation drop, attitude, and distance from the mining face.

3. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S2, the event spatial clustering degree is obtained by calculating the number of earthquakes or ground sound events occurring per unit volume and per unit time, and then using a kernel density estimation method for spatial smoothing.

4. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that, In step S3, the dynamic weighting coefficient , , The adaptive adjustment rule is as follows: when the mining face is close to the main geological structure, the weight of the microseismic risk sub-function is automatically increased. and geological structure weight When a sharp increase in the frequency of ground sound events is detected while microseismic events are relatively calm, the weight of the ground sound risk sub-function is automatically increased. When the borehole stress gauge detects the stress gradient As the stress risk sub-function continues to increase, its weight is automatically increased. .

5. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S3, the risk function , , A machine learning model trained based on historical rockburst case data was used.

6. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 5, characterized in that: The machine learning model is a random forest, gradient boosting tree, or neural network model.

7. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 6, characterized in that: The input features of the machine learning model also include mining process parameters, which include, but are not limited to, working face advance speed, mining height and roadway excavation progress.

8. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S4, the risk threshold is not a fixed value, but a floating threshold that is adjusted according to the mining stage, roof management method, and the effect of the pressure relief measures already taken.

9. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S4, the three-dimensional dynamic positioning display uses different colors and transparency to represent different risk levels. High-risk areas are highlighted with warm colors and high opacity, and the spatiotemporal evolution of the risk areas can be displayed in the form of time-series animation.

10. The method for real-time identification and location of multi-source information in coal mine rockburst risk areas according to claim 1, characterized in that: In step S5, the graded early warning information is divided into at least three levels: "attention", "early warning" and "alarm". Specific response procedures and handling measures are specified for each level. These handling measures include strengthening monitoring, limiting personnel operations, stopping operations, and performing directional pressure relief blasting.

Images