Deep rock mass dynamic disaster multi-source monitoring intelligent early warning method

Abstract

This invention relates to the field of safety monitoring and disaster early warning technology for deep rock mass engineering, and in particular to an intelligent early warning method for multi-source monitoring of dynamic disasters in deep rock masses. The method includes: S1 collecting multi-source heterogeneous monitoring data of the target area by deploying a multi-source sensor network and performing spatiotemporal fusion to obtain a unified spatiotemporal characterization; S2 combining stress, microseismic, and energy parameters, and introducing damage variables to construct a three-field coupled dynamic evolution model of stress field, seismic field, and energy field; S3 inputting the unified spatiotemporal characterization into the three-field coupled dynamic evolution model, and using a finite element-discrete element coupled numerical method to extrapolate the disaster process and predict the spatiotemporal probability distribution of future disasters; S4 using an adaptive dynamic threshold mechanism to perform graded early warning of the spatiotemporal probability distribution and output early warning information. This invention can realize real-time monitoring, intelligent identification, and advanced early warning of dynamic disasters such as rockburst, rock burst, water inrush, and mine tremors in deep coal mines, metal mines, tunnels, and other engineering projects.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring and disaster early warning technology for deep rock mass engineering, and in particular to an intelligent early warning method for multi-source monitoring of dynamic disasters in deep rock masses, which is applicable to real-time monitoring, intelligent identification and early warning of dynamic disasters such as rock bursts, rockbursts, water inrushes and mine tremors in deep coal mines, metal mines and tunnels. Background Technology

[0002] Deep rock dynamic hazards (rock bursts, rock bursts, water inrushes, mine tremors, etc.) are major technical challenges restricting the safe mining of deep resources. As mining depth increases, the rock mass is in a complex disaster-prone environment characterized by "three highs and one disturbance" (high ground stress, high rock temperature, high well depth, and mining disturbance). The mechanisms of disaster occurrence are complex, and traditional single monitoring methods are insufficient to achieve accurate early warning.

[0003] Existing technologies mainly include: (1) Microseismic monitoring technology: This technology involves collecting microseismic signals generated by rock fractures to locate the seismic source and calculate energy. However, it has limitations such as signal recognition being susceptible to interference from downhole mechanical vibrations, mismatch between the static sampling frequency and the dominant frequency of the dynamic fracture signal, and spatial uncertainty in seismic source location.

[0004] (2) Multi-source data fusion technology: Some technologies simply weight and fuse data such as microseismic, stress, and displacement, but lack deep coupling between physical mechanisms and data-driven approaches. The warning threshold is fixed and it is difficult to adapt to the time-varying characteristics under dynamic disturbances in mining.

[0005] (3) Intelligent early warning model: Existing early warning models based on neural networks or support vector machines are mostly "black box" predictions, lacking explanation of the physical mechanism of the entire disaster gestation process, with poor interpretability, and do not consider the temporal asynchrony and spatial heterogeneity of multi-source data.

[0006] (4) TBM development monitoring technology: The micro-vibration monitoring technology for tunnel boring machines has realized automatic sensor disassembly and assembly and frequency conversion acquisition, but it is mainly for linear projects and is difficult to adapt to the three-dimensional monitoring needs of complex mining spaces in deep mines.

[0007] In summary, existing technologies lack effective spatiotemporal correlation mechanisms among multi-source monitoring data, fail to establish a dynamic evolution model coupling the stress, vibration, and energy fields, and the spatial resolution and temporal lead of early warning results are insufficient to meet engineering requirements. Therefore, this invention proposes an intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters. Summary of the Invention

[0008] The purpose of this invention is to provide a multi-source intelligent early warning method for monitoring dynamic disasters in deep rock masses. By constructing a dual-engine early warning system of "physical mechanism constraints + data-driven optimization", it enables real-time monitoring, intelligent identification, and early warning of dynamic disasters such as rock bursts, rockbursts, water inrushes, and mine tremors in deep coal mines, metal mines, tunnels, and other engineering projects.

[0009] To achieve the above objectives, the present invention provides the following solution: A multi-source intelligent early warning method for monitoring deep rock mass dynamic hazards includes: S1. Multi-source heterogeneous monitoring data of the target area is collected by deploying a multi-source sensor network. A local coordinate system is established with the location of the mining face as a dynamic reference. The multi-source heterogeneous monitoring data is aligned in time at multiple scales. The spatial field is reconstructed based on spatial interpolation and physical constraints to obtain a unified spatiotemporal representation. S2, combining the stress, microseismic, and energy parameters in the spatiotemporal unified characterization, and introducing rock mass damage variables, a three-field coupled dynamic evolution model of stress field-vibration field-energy field is constructed, which includes stress balance, energy conservation, and damage evolution process. The actual monitored microseismic events are used as the source term input of the energy field. S3, input the spatiotemporal unified representation into the three-field coupled dynamic evolution model, and use the established twin to extrapolate the rock mass response under mining disturbance using the finite element-discrete element coupled numerical method, and combine stochastic simulation and time series prediction network to predict the spatiotemporal probability distribution of disasters occurring in future time windows; S4. Based on the spatiotemporal probability distribution, a graded early warning is performed, and early warning information is output. When the output early warning information reaches the set early warning level, rock mass decompression control measures matching the early warning level are triggered and executed. The graded early warning adopts an adaptive dynamic threshold mechanism. The adaptive dynamic threshold mechanism dynamically updates the early warning threshold parameters over time by constructing a weighted objective function that includes the number of missed reports, the number of false reports, and the early warning lead time.

[0010] Preferably, the multi-source heterogeneous monitoring data in S1 includes: microseismic monitoring data, geostress monitoring data, electromagnetic radiation monitoring data, and rock mass displacement monitoring data.

[0011] Preferably, in S1, the spatial field reconstruction based on spatial interpolation and physical constraints is based on the coupling method of Kriging interpolation and finite element to reconstruct discrete measurement point data into continuous field quantities.

[0012] Preferably, the finite element-discrete element coupled numerical method used in S3 includes: using finite element method to calculate stress wave propagation and overall deformation in the far-field continuous region, using discrete element method to simulate crack initiation and block motion in the near-field fracture region, and realizing information transmission and energy conservation in the transition region through the coupling interface.

[0013] Preferably, the finite element-discrete element coupled numerical method in S3 further includes: by monitoring the Gaussian point damage variable of the finite element unit, when the damage variable exceeds a set threshold, generating a set of discrete element blocks inside the unit and activating discrete element particles.

[0014] Preferably, the adaptive dynamic threshold mechanism in S4 is also constrained by the mining disturbance coefficient and the periodic adjustment coefficient reflecting the impact of the mining cycle. The weighted objective function is learned through feedback from historical disaster occurrences and forecast records to minimize the comprehensive loss and optimize the early warning threshold parameters.

[0015] Preferably, the early warning information output in S4 is a three-level progressive early warning information of mine-region-local, and the early warning level is classified into four levels: blue, yellow, orange and red.

[0016] Preferably, the rock mass decompression control measures in S4 include: hydraulic fracturing and deep-hole blasting decompression. During the execution of the decompression control measures, multi-source heterogeneous monitoring data are continuously collected to evaluate the decompression effect. If the expected results are not achieved, the decompression parameters are dynamically adjusted until the probability of disaster is reduced to a safe range.

[0017] The beneficial effects of this invention are as follows: This invention establishes a unified spatiotemporal characterization framework for multi-source heterogeneous monitoring data, solving the challenges of asynchronous data acquisition and spatial registration; it constructs a three-field coupled dynamic evolution model of stress field, vibration field, and energy field to reveal the physical mechanisms of disaster formation; it develops disaster process extrapolation and early warning technologies based on digital twins to achieve accurate prediction of disaster location, time, and severity; and it establishes an adaptive threshold dynamic adjustment mechanism to adapt to the time-varying environmental characteristics under mining disturbances. This embodiment, through the construction of a dual-engine early warning system of "physical mechanism constraints + data-driven optimization," enables real-time monitoring, intelligent identification, and early warning of dynamic disasters such as rock bursts, rockbursts, water inrushes, and mine tremors in deep coal mines, metal mines, and tunnels. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a multi-source intelligent early warning method for monitoring deep rock dynamic disasters, according to an embodiment of the present invention. Detailed Implementation

[0020] 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 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 skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0022] This embodiment provides a multi-source intelligent early warning method for monitoring deep rock mass dynamic disasters, including: S1. By deploying a multi-source sensor network to collect multi-source heterogeneous monitoring data of the target area, a local coordinate system is established with the location of the mining face as a dynamic reference. The multi-source heterogeneous monitoring data is aligned over time at multiple scales, and the spatial field is reconstructed based on spatial interpolation and physical constraints to obtain a unified spatiotemporal representation. S2, combining stress, microseismic and energy parameters in the unified spatiotemporal characterization, and introducing rock mass damage variables, constructs a three-field coupled dynamic evolution model of stress field-vibration field-energy field, which includes stress balance, energy conservation and damage evolution process, wherein the actual monitored microseismic events are used as the source term input of the energy field; S3 inputs a unified spatiotemporal representation into a three-field coupled dynamic evolution model. Through the established twin, the finite element-discrete element coupled numerical method is used to extrapolate the rock mass response under mining disturbance. Combined with stochastic simulation and time series prediction network, the spatiotemporal probability distribution of disasters occurring within future time windows is predicted. S4 performs graded early warning based on spatiotemporal probability distribution and outputs early warning information. When the output early warning information reaches the set early warning level, rock mass decompression control measures matching the early warning level are triggered and executed. Among them, the graded early warning adopts an adaptive dynamic threshold mechanism. The adaptive dynamic threshold mechanism dynamically updates the early warning threshold parameters over time by constructing a weighted objective function that includes the number of missed reports, the number of false reports, and the early warning lead time.

[0023] Specifically, this embodiment establishes a unified spatiotemporal characterization framework for multi-source heterogeneous monitoring data to solve the challenges of asynchronous data acquisition and spatial registration; it constructs a dynamic evolution model of three coupled fields—stress field, vibration field, and energy field—to reveal the physical mechanisms of disaster formation; it develops disaster process extrapolation and early warning technologies based on digital twins to achieve accurate prediction of the location, time, and severity of disasters; and it establishes an adaptive threshold dynamic adjustment mechanism to adapt to the time-varying environmental characteristics under mining disturbances. This embodiment, by constructing a dual-engine early warning system of "physical mechanism constraints + data-driven optimization," can achieve real-time monitoring, intelligent identification, and early warning of dynamic disasters such as rock bursts, rockbursts, water inrushes, and mine tremors in deep coal mines, metal mines, and tunnels.

[0024] Furthermore, the multi-source heterogeneous monitoring data in S1 includes: microseismic monitoring data, geostress monitoring data, electromagnetic radiation monitoring data, and rock mass displacement monitoring data.

[0025] Specifically, in this embodiment, a multi-source sensor network is deployed in the surrounding rock of deep mining faces and roadways, including: Microseismic / acoustic emission monitoring: A combined "surface + downhole" deployment method is adopted, deploying low-frequency microseismic detectors (for far-field large-scale fracture location) and high-frequency acoustic emission sensors (for near-field small-scale micro-fracture capture).

[0026] Stress monitoring: Borehole stress gauges are installed in coal pillars and on both sides of roadways to monitor changes in vertical and horizontal stress in real time.

[0027] Electromagnetic radiation induction: A ring-shaped charge induction sensor is installed in the anchor bolts, anchor cables, and deep within the borehole. This sensor utilizes the electric field fluctuations caused by free charges generated when the rock mass undergoes stress deformation and fracture to capture microscopic damage precursor information that conventional micro-seismic methods cannot detect.

[0028] Surface and rock strata movement monitoring: Integrating BeiDou GNSS, InSAR and borehole inclinometers to monitor the three-dimensional continuous deformation of the surface and deep rock strata.

[0029] Furthermore, in S1, the spatial field is reconstructed based on spatial interpolation and physical constraints by using the coupling method of Kriging interpolation and finite element method to reconstruct discrete measurement point data into continuous field quantities.

[0030] Specifically, in this embodiment, a local coordinate system is established that dynamically updates with the advance position of the mining face, using the position of the mining face as the spatial reference. The current position of the mining face is set as the origin. The direction of advancement is Axis, perpendicular to the bedding direction Establish a right-handed coordinate system based on the axes.

[0031] An adaptive resampling technique is employed, based on Shannon's sampling theorem, to perform multi-scale alignment of data at different frequencies. ; in, For the first Class of monitoring data, For the corresponding resampling kernel function, The time delay compensation amount is determined through cross-correlation analysis.

[0032] Based on the coupling method of Kriging interpolation and finite element method, discrete measurement point data are reconstructed into continuous field quantities: ; in, For spatial location Field value at that location, These are the Kriging weighting coefficients. These are finite element physical constraints to ensure that the reconstructed field satisfies the rock mechanics equilibrium equations.

[0033] Furthermore, the three-field coupled dynamic evolution model in S2 includes: stress balance equation, energy conservation equation and damage evolution equation, in which microseismic events serve as the source term input of the energy field.

[0034] Specifically, this embodiment establishes a stress field. Vibration field (Particle vibration velocity), energy field The system of coupled equations: ; ; in, This is the feedback force of the vibration field on the stress field. For energy dissipation, This refers to the energy source term for microseismic events.

[0035] The microseismic monitoring data is converted into a source term input for the energy field. For the first... The energy release of a microseismic event is as follows: ; in, This refers to the radiated energy of microseismic events. To account for the energy diffusion coefficient due to the heterogeneity of the rock mass, it is updated using real-time wave velocity tomography.

[0036] Introducing rock mass damage variables Establish a damage evolution equation coupled with three fields: ; in, For equivalent stress, The damage threshold stress, These are material parameters.

[0037] For spatial influence kernel function, Characterizing the first Micro-seismic events at location The damage contribution weight caused by the microseismic energy release reflects the spatial influence range and attenuation law of the microseismic energy release on the damage to the surrounding rock mass.

[0038] Basic form: ; ; in, For spatial points To the location of the microseismic event The Euclidean distance; For the first The radius of influence of a microseismic event is related to the energy of the event. The scaling factor is determined through on-site calibration. It represents the energy index.

[0039] Determination Method: This embodiment adopts the inversion calibration method based on monitoring data. It uses historical microseismic data and subsequent damage observation to invert kernel function parameters, and uses Bayesian optimization or genetic algorithm to solve for the optimal parameters. It also incorporates anisotropic kernel functions based on the rock mass structural surface attitude for anisotropic correction, so that the kernel function parameters are dynamically updated with the rock mass damage evolution.

[0040] Furthermore, the finite element-discrete element coupled numerical method used in S3 includes: using finite element method to calculate stress wave propagation and overall deformation in the far-field continuous region, using discrete element method to simulate crack initiation and block motion in the near-field fracture region, and realizing information transfer and energy conservation in the transition region through the coupling interface.

[0041] Specifically, by monitoring the Gaussian point damage variable of the finite element unit, when the damage variable exceeds a set threshold, a discrete element block set is generated inside the unit and the discrete element particles are activated.

[0042] Specifically, this embodiment establishes a three-dimensional digital twin model including geological structure, mining engineering, and monitoring system to realize real-time mapping between physical space and digital space. The twin includes: (1) geological twin: three-dimensional visualization of the distribution of strata, faults, and joints; (2) engineering twin: dynamic updates of mining working face, roadway, and goaf; (3) monitoring twin: real-time presentation of sensor network, data stream, and early warning status.

[0043] Based on the established three-field coupled model, the finite element-discrete element coupled numerical method (FEM-DEM) is used to extrapolate the rock mass response under future mining disturbances: ; in, It is a state vector (stress, displacement, damage, etc.). These are the parameters for mining disturbance (advance speed, support strength, etc.). This is a coupling operator.

[0044] Predicting the future using a fusion of Monte Carlo simulation and machine learning methods. Spatiotemporal probability distribution of disaster occurrence within a given time period: ; in, For indicator functions, The critical damage threshold, The spatial transition probability kernel is learned from historical data through a Long Short-Term Memory (LSTM) network.

[0045] This embodiment achieves "taking advantage of each and coupling in different areas" through FEM-DEM: far-field continuous region → finite element (FEM): efficient calculation of stress wave propagation and overall deformation; near-field fracture region → discrete element (DEM): fine simulation of crack initiation, propagation and block motion; transition region → coupling interface: realizing information transmission and energy conservation.

[0046] The coupling interface adopts the interface element method: FEM boundary node → interface element → DEM virtual particle → contact force aggregation → reaction force mapping → stress mapping → shape function interpolation → FEM boundary node.

[0047] Specifically, the algorithm flow in this embodiment includes: (1) monitoring the damage variable D of the Gaussian point of the FEM unit; (2) when (e.g., 0.8), mark the element as “to be broken”; (3) generate a discrete element block set (Voronoi or polyhedral subdivision) inside the element; (4) map the element stress state to the block initial stress; (5) delete the element from the FEM mesh and activate the DEM particles; (6) update the coupling interface topology.

[0048] Furthermore, the adaptive dynamic threshold mechanism in S4 is also constrained by the mining disturbance coefficient and the periodic adjustment coefficient reflecting the impact of the mining cycle. The weighted objective function is learned through feedback from historical disaster occurrences and forecast records to minimize the comprehensive loss and optimize the early warning threshold parameters.

[0049] Specifically, this embodiment establishes a dynamic threshold optimization model: Warning threshold No longer a constant, but a variable that adapts to time: ; in, As the baseline threshold, This is the mining disturbance coefficient. This is a periodic adjustment factor (considering the impact of the mining cycle). This refers to the mining cycle.

[0050] Threshold correction based on feedback learning: Establish an early warning effectiveness feedback mechanism. Based on actual disaster occurrences or false / missed reports, use reinforcement learning to optimize threshold parameters. The loss function is: ; in, The number of times the report was missed. Number of false alarms To provide advance warning time, The weights are used to minimize the loss function using the Soft Actor-Critic algorithm. Optimize parameters .

[0051] Furthermore, the early warning information output in S4 is a three-level progressive early warning information of mine-region-local, and the early warning level is classified into four levels: blue, yellow, orange and red.

[0052] Specifically, in this embodiment, the following settings are made: (1) Mine level: Based on the overall energy release rate and macro stress level, output long-term early warning for 24-72 hours; (2) Regional level: Based on the field evolution within a 300m radius around the mining face, output a mid-term early warning for 4-24 hours; (3) Local level: Based on real-time microseismic clustering and stress mutation, output short-term warning of 0-4 hours.

[0053] The warning levels are classified into four levels: blue (caution), yellow (warning), orange (alert), and red (emergency).

[0054] Specifically, when the mining face enters the medium stress zone, the microseismic activity increases compared to the background value, but does not form a significant accumulation. When the stress monitoring value reaches 40-60% of the design intensity, a blue warning (Level IV - Caution) is triggered.

[0055] When microseismic events accumulate in a specific area in front of the work site, stress monitoring shows a significant upward trend (growth rate > 10% / day), and multi-field coupling analysis indicates that the risk may escalate within the next 24-72 hours, a yellow alert (Level III - Warning) is triggered.

[0056] When digital twin simulations show that the probability of disaster P > 50% within the next 4-24 hours, microseismic activity enters a "critical state" (the event frequency-energy relationship deviates from the power law distribution), and the stress field and damage field exhibit spatial coupling (the high stress area overlaps with the high damage area), an orange alert (Level II - Warning) is triggered.

[0057] A red alert (Level I - Emergency) is triggered when any of the following conditions are met: the probability of a disaster within the next 4 hours is greater than 75% according to the digital twin projection; the microseismic event forms a "migration trajectory" pointing towards the mining space; the stress monitoring shows a "sudden drop" precursor (stress transfer); or macroscopic precursors such as rock spalling and bursting sounds are observed on site.

[0058] Furthermore, the rock mass decompression control measures in S4 include hydraulic fracturing and deep-hole blasting decompression. During the execution of the decompression control measures, multi-source heterogeneous monitoring data are continuously collected to evaluate the decompression effect. If the expected results are not achieved, the decompression parameters are dynamically adjusted until the probability of disaster is reduced to a safe range.

[0059] Specifically, the rock mass decompression control measures in this embodiment include: Contingency plan matching: When the warning level exceeds the set threshold (e.g., orange), the system automatically matches the corresponding pressure relief control measures (e.g., hydraulic fracturing parameters, deep hole blasting layout scheme, borehole pressure relief density) from the contingency plan library.

[0060] Execution and Feedback: The pre-planned procedure is executed by remotely controlling equipment such as pressure relief drilling rigs or high-pressure water pumps. During the pressure relief process, data is continuously collected and the pressure relief effect is evaluated in real time (such as stress transfer and energy release rate).

[0061] Dynamic control: If the pressure relief effect does not meet expectations, the pressure relief parameters are automatically adjusted until the disaster risk is reduced to a safe range, realizing fully automatic closed-loop management of "monitoring-early warning-control-assessment".

[0062] As a specific application example of the method proposed in this embodiment, it was carried out in a deep gold mine (mining depth > 1000m), and the following layout was carried out: Microseismic sensors: A 32-channel array is deployed within a 200m radius around the mining face, with a sensor spacing of 30-50m; Ground stress sensors: Six sets of hollow-encapsulated stress gauges are installed on the roadway sidewalls to monitor three-dimensional stress changes in real time; Displacement sensors: Fiber optic grating displacement gauges are used to monitor the displacement of key points with an accuracy of 0.01mm; Electromagnetic radiation sensors: Deployed in the return airway of the working face to monitor electromagnetic radiation signals from coal and rock fractures.

[0063] Data from each sensor is transmitted to a ground data center via an industrial Ethernet network. The sampling rates are as follows: microseismic data 10kHz, stress data 10Hz, electromagnetic data 500Hz, and displacement data 1Hz.

[0064] On a certain day in a certain month of a certain year, microseismic events were detected 80 meters in front of the mining face, with localized stress concentration and an abnormally high energy release rate. Digital twin simulation predicted a 70% probability of a moderate rockburst in the area within 24 hours. An orange alert was issued, and the mine implemented pressure relief blasting and reinforced support measures. A magnitude 2.1 rockburst actually occurred 16 hours later. Due to the early prevention and control measures, there were no casualties or equipment damage.

[0065] Compared with the prior art, this embodiment has the following significant advantages: (1) Significantly improved spatiotemporal resolution: Through unified spatiotemporal representation of multi-source data, the spatial resolution of monitoring data is better than 2m, and the time synchronization accuracy reaches the millisecond level, which solves the problem of data "fighting alone" in traditional methods.

[0066] (2) Deep integration of physical mechanism and data-driven approach: Unlike the "black box" model driven by pure data, the "stress-vibration-energy" three-field coupling model established in this embodiment has a clear physical meaning. At the same time, by optimizing the model parameters through machine learning, the unity of mechanism interpretability and prediction accuracy is achieved.

[0067] (3) Extended early warning time: Through digital twin simulation technology, the early warning time is extended from the traditional several hours to 24-72 hours, leaving sufficient time for disaster prevention and control.

[0068] (4) Adaptive to environmental changes: The dynamic threshold mechanism enables the early warning system to automatically adapt to environmental factors such as changes in mining speed and sudden changes in geological conditions, reducing the false alarm rate by more than 60% and the missed alarm rate by more than 80%.

[0069] (5) Improved spatial positioning accuracy: Through the probability field characterization method, the disaster positioning accuracy has been improved from the traditional 10-20m to within 2m, meeting the needs of precise prevention and control.

[0070] In summary, this embodiment breaks through the limitations of traditional simple fusion of multi-source data, establishes a dual-engine architecture of physical mechanism constraints and data-driven optimization, and realizes the spatiotemporal simulation of the disaster incubation process through digital twin technology, which significantly improves the accuracy, foresight and spatial resolution of early warning, and has significant technological progress and industrial applicability.

[0071] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-source intelligent early warning method for monitoring dynamic disasters in deep rock masses, characterized in that, include: S1. Multi-source heterogeneous monitoring data of the target area is collected by deploying a multi-source sensor network. A local coordinate system is established with the location of the mining face as a dynamic reference. The multi-source heterogeneous monitoring data is aligned in time at multiple scales. The spatial field is reconstructed based on spatial interpolation and physical constraints to obtain a unified spatiotemporal representation. S2, combining the stress, microseismic, and energy parameters in the spatiotemporal unified characterization, and introducing rock mass damage variables, a three-field coupled dynamic evolution model of stress field-vibration field-energy field is constructed, which includes stress balance, energy conservation, and damage evolution process. The actual monitored microseismic events are used as the source term input of the energy field. S3, input the spatiotemporal unified representation into the three-field coupled dynamic evolution model, and use the established twin to extrapolate the rock mass response under mining disturbance using the finite element-discrete element coupled numerical method, and combine stochastic simulation and time series prediction network to predict the spatiotemporal probability distribution of disasters occurring in future time windows; S4. Based on the spatiotemporal probability distribution, a graded early warning is performed, and early warning information is output. When the output early warning information reaches the set early warning level, rock mass decompression control measures matching the early warning level are triggered and executed. The graded early warning adopts an adaptive dynamic threshold mechanism. The adaptive dynamic threshold mechanism dynamically updates the early warning threshold parameters over time by constructing a weighted objective function that includes the number of missed reports, the number of false reports, and the early warning lead time.

2. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The multi-source heterogeneous monitoring data in S1 includes: microseismic monitoring data, geostress monitoring data, electromagnetic radiation monitoring data, and rock mass displacement monitoring data.

3. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The spatial field reconstruction based on spatial interpolation and physical constraints in S1 is a method that uses Kriging interpolation and finite element coupling to reconstruct discrete measurement point data into continuous field quantities.

4. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The finite element-discrete element coupled numerical method used in S3 includes: using finite element method to calculate stress wave propagation and overall deformation in the far-field continuous region, using discrete element method to simulate crack initiation and block motion in the near-field fracture region, and realizing information transmission and energy conservation in the transition region through the coupling interface.

5. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The finite element-discrete element coupled numerical method in S3 also includes: by monitoring the Gaussian point damage variable of the finite element unit, when the damage variable exceeds a set threshold, generating a set of discrete element blocks inside the unit and activating discrete element particles.

6. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The adaptive dynamic threshold mechanism in S4 is also constrained by the mining disturbance coefficient and the periodic adjustment coefficient reflecting the impact of the mining cycle. The weighted objective function learns from historical disaster occurrences and forecast records to minimize the comprehensive loss and optimize the early warning threshold parameters.

7. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The early warning information output in S4 is a three-level progressive early warning information of mine-region-local, and the early warning level is classified into four levels: blue, yellow, orange and red.

8. The intelligent early warning method for multi-source monitoring of deep rock mass dynamic disasters according to claim 1, characterized in that, The rock mass decompression control measures in S4 include hydraulic fracturing and deep-hole blasting decompression. During the execution of the decompression control measures, multi-source heterogeneous monitoring data are continuously collected to evaluate the decompression effect. If the expected results are not achieved, the decompression parameters are dynamically adjusted until the probability of disaster is reduced to a safe range.

Images