A large-range rock burst disaster classification method and system based on microseismic monitoring

Abstract

The application discloses a large-range rock burst disaster classification method and system based on microseismic monitoring, relates to the technical field of coal mine safety, and comprises the following steps: collecting multi-source microseismic monitoring data of a coal mine working face, including source energy, source position, event frequency and waveform characteristics, and pre-processing the multi-source microseismic monitoring data to construct a space-time data cube; based on the space-time data cube, the improved clustering algorithm is used to analyze the spatial distribution characteristics of microseismic events, identify the microseismic event concentration area, and extract the vibration energy gradient and event density of each area. Through multi-source data fusion and an improved algorithm, a scientific classification matrix is constructed, the quantitative determination standards of three types of disasters, namely, fault activation type, stress concentration type and composite type, are determined, the disadvantages of traditional single parameters and subjective experience judgment are overcome, and the accuracy and scientificity of disaster classification are greatly improved, thereby laying a foundation for precise prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of coal mine safety, more particularly, the present application relates to a large-range rock burst disaster classification method and system based on microseismic monitoring. BACKGROUND

[0002] With the gradual extension of coal resource mining to the deep part, the geological conditions of the mine become more and more complex, and the rock burst, as a typical coal and rock dynamic disaster, has significantly increased in frequency, influence range and damage degree, and has become the core bottleneck restricting the safe and efficient mining of deep coal mines. The causes of rock burst involve multiple factors such as geological structure (such as fault, fold), mining disturbance, stress concentration, and the disaster types are complex and diverse, and accurate identification of the disaster causes and types is the prerequisite for effective prevention and control.

[0003] At present, the existing rock burst disaster prevention and control technology has the following outstanding problems: the traditional disaster classification mostly relies on a single monitoring parameter (such as microseismic energy or coal stress) or artificial experience judgment, and does not fully integrate multi-source information such as geological structure, mine pressure and support state, which makes it difficult to accurately identify the core cause of the disaster, resulting in a fuzzy boundary in the determination of different types of disasters such as "fault activation" and "stress concentration", and the classification results are highly subjective and have low reliability.

[0004] Therefore, it has become an urgent need for safe mining in the coal industry to develop an integrated technology that can accurately classify rock burst disasters. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a large-range rock burst disaster classification method and system based on microseismic monitoring, which constructs a scientific classification matrix through multi-source data fusion and improved algorithm, and clearly defines the quantitative determination standards for three types of disasters, namely fault activation type, stress concentration type and composite type, thereby overcoming the disadvantages of traditional single parameter and subjective experience judgment, greatly improving the accuracy and scientificity of disaster classification, and laying a foundation for accurate prevention and control.

[0006] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0007] A large-range rock burst disaster classification method based on microseismic monitoring, comprising the following steps: collecting multi-source microseismic monitoring data of a coal mine working face, including source energy, source location, event frequency and waveform characteristics, and preprocessing the multi-source microseismic monitoring data to construct a space-time data cube; based on the space-time data cube, using an improved clustering algorithm to analyze the spatial distribution characteristics of microseismic events, identifying microseismic event concentration areas, and extracting the vibration energy gradient and event density of each area; constructing a fault activation risk evaluation model, calculating the fault activation tendency index according to the space-time migration law and energy release characteristics of microseismic events, and dividing different levels of fault activation danger zones in combination with geological structure data; using multi-parameter fusion technology, fusing microseismic characteristic parameters with mine pressure data and support resistance data to establish a rock burst disaster classification matrix, and classifying rock burst disasters into three categories: fault activation type, stress concentration type and composite type; based on the rock burst disaster classification results, generating a disaster prevention and control decision diagram, and recommending corresponding prevention and control measures for different disaster types.

[0008] In a preferred embodiment, the multi-source microseismic monitoring data of the coal mine working face, including source energy, source location, event frequency and waveform characteristics, and preprocessing the multi-source microseismic monitoring data to construct a space-time data cube, specifically: deploying an on- and off-shaft combined microseismic monitoring network to collect microseismic event waveform data of the working face; denoising and filtering the waveform data to extract source parameters including magnitude, location, energy and frequency; associating the processed source parameters with the mining advance sequence in space and time to construct a three-dimensional space-time data cube.

[0009] In a preferred embodiment, based on the space-time data cube, using an improved clustering algorithm to analyze the spatial distribution characteristics of microseismic events, identifying microseismic event concentration areas, and extracting the vibration energy gradient and event density of each area, specifically: using a DBSCAN clustering algorithm based on density to identify the spatial aggregation area of microseismic events; calculating the energy gradient value and event density value of each aggregation area; in combination with geological structure information, spatially correlating and analyzing the aggregation area with known faults, folds and other geological structures.

[0010] In a preferred embodiment, the construction of a fault activation risk evaluation model, the calculation of the fault activation tendency index according to the space-time migration law and energy release characteristics of microseismic events, and the division of different levels of fault activation danger zones in combination with geological structure data, specifically: analyzing the space-time migration law of microseismic events along the fault plane, calculating the event migration speed and direction; evaluating the energy accumulation rate and release characteristics of the fault area; using a fuzzy comprehensive evaluation method, quantifying the fault activation tendency index into three levels of high risk, medium risk and low risk.

[0011] In a preferred embodiment, the multi-parameter fusion technology is used to fuse the microseismic characteristic parameters with mine pressure data and support resistance data, a rock burst disaster classification matrix is established, and the rock burst disaster is divided into three categories of fault activation type, stress concentration type and composite type, specifically: the characteristic parameters of microseismic monitoring are extracted, including event concentration, energy release rate and b value; mine pressure data are collected, including support working resistance, roof subsidence and coal stress; the main characteristic vectors are extracted by dimension reduction processing of multi-source parameters through principal component analysis; and the classification matrix is constructed, and the rock burst disaster is divided into three categories of fault activation type, stress concentration type and composite type according to the combination mode of the characteristic vectors.

[0012] In a preferred embodiment, the classification matrix is constructed, and the rock burst disaster is divided into three categories of fault activation type, stress concentration type and composite type according to the combination mode of the characteristic vectors, specifically: the fault activation type: the microseismic events are distributed along the fault zone, the energy release is concentrated, and the fault activation tendency index is high; the stress concentration type: the microseismic events are concentrated near the mining working face, the energy gradient changes sharply, and the mine pressure appears obviously; and the composite type: the microseismic events are spatially distributed and multiple precursor characteristics coexist.

[0013] In a preferred embodiment, the disaster prevention and control decision diagram is generated, specifically: the classification results are fused with the three-dimensional geological model, and the spatial distribution of different disaster types is visualized; for the fault activation type disaster, pressure relief blasting and fault grouting measures are recommended; for the stress concentration type disaster, roof pre-splitting and coal pressure relief measures are recommended; and for the composite type disaster, comprehensive prevention and control measures are recommended, including optimization of mining parameters and strengthening of support resistance.

[0014] In a preferred embodiment, it further includes: based on the multi-source microseismic monitoring data and the classification results, a dynamic early warning model is established by using an LSTM neural network to predict the development trend of the rock burst disaster, and the classification matrix parameters are updated regularly.

[0015] The technical effects and advantages of the large-range rock burst disaster classification method and system based on microseismic monitoring are as follows:

[0016] The present application overcomes the disadvantages of traditional single parameter and subjective experience judgment by multi-source data fusion and improved algorithm, constructs a scientific classification matrix, clearly defines the quantitative judgment standards of three types of disasters of fault activation type, stress concentration type and composite type, greatly improves the accuracy and scientificity of disaster classification, and lays a foundation for precise prevention and control;

[0017] Based on the fault activation risk evaluation model and LSTM neural network, the spatio-temporal migration law, energy release characteristics and other data of microseismic are integrated to realize the quantitative classification of fault activation tendency and the dynamic prediction of disaster development trend. The system has self-adaptive learning ability, can update and optimize the model parameters through real-time data, and can lock the high-risk area and evolution direction in advance to gain sufficient time for prevention and control.

[0018] Based on the accurate classification results, a three-dimensional visual prevention and control decision graph is generated, and different measures are pushed for different disaster types: focus on stability + pressure relief for fault activation type, focus on stress release for stress concentration type, and focus on comprehensive prevention and control + parameter optimization for composite type. It not only ensures that the prevention and control measures directly hit the disaster causes and improve the success rate of prevention and control, but also avoids blind governance and excessive investment, saving the cost of disaster prevention.

[0019] An integrated and modular system of data acquisition, spatial analysis, risk evaluation, disaster classification and decision support is constructed to realize full-process automatic processing. The system adapts to the deployment requirements of mine site, and the modules can be individually debugged and optimized. The three-dimensional visual design reduces the professional threshold, solves the problems of scattered traditional technology monitoring, analysis and decision-making modules, complex operation and difficult popularization, and improves the engineering applicability and landing efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 The overall flowchart of the large-range rock burst disaster classification method based on microseismic monitoring according to the present application.

[0021] Figure 2 The data acquisition and spatio-temporal data cube construction flowchart of the large-range rock burst disaster classification method based on microseismic monitoring according to the present application.

[0022] Figure 3 The spatial analysis and fault evaluation flowchart of the large-range rock burst disaster classification method based on microseismic monitoring according to the present application.

[0023] Figure 4 The disaster classification and decision support flowchart of the large-range rock burst disaster classification method based on microseismic monitoring according to the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] Example 1, Figures 1-4The application discloses a large-range rock burst disaster classification method based on microseismic monitoring, and comprises the following steps.

[0026] Collecting multi-source microseismic monitoring data of a coal mine working face, including source energy, source position, event frequency and waveform characteristics, and pre-processing the multi-source microseismic monitoring data to construct a time-space data cube.

[0027] Collecting multi-source microseismic monitoring data of a coal mine working face, including source energy, source position, event frequency and waveform characteristics, and pre-processing the multi-source microseismic monitoring data to construct a time-space data cube, specifically, a joint microseismic monitoring network of an underground mine and a ground surface is arranged, and microseismic event waveform data of the working face is collected; the waveform data is subjected to denoising and filtering processing, and source parameters including a magnitude, a position, energy and a frequency are extracted; the processed source parameters are associated with a mining advance sequence in time and space, and a three-dimensional time-space data cube is constructed.

[0028] It should be noted that the monitoring network comprises underground sensors and ground surface sensors, and forms a three-dimensional coverage.

[0029] The collected object is microseismic event waveform data, that is, an earthquake wave signal generated when rock is broken or stress is released, which is the original basis for extracting source parameters.

[0030] The limitation of a single monitoring method is solved: the underground sensor has high collection accuracy but is easily disturbed by mining; the ground surface sensor has a wide coverage and can make up for the blind area of underground monitoring, and the joint network can realize dual protection of a large range and high accuracy, and ensure that no key microseismic event is missed.

[0031] Denoising processing: invalid signals generated by mining operations (such as blasting and mechanical operation) and environmental interference (such as cable interference and natural vibration) are removed, and real microseismic waveforms are reserved; filtering processing: specific frequency waveforms related to rock breaking are selected through technical means, high-frequency noise and low-frequency interference are filtered out, and the quality of the waveform signal is optimized; parameter extraction: based on the processed pure waveforms, core parameters are calculated through a seismological algorithm, including a magnitude (reflecting vibration intensity), a position (three-dimensional space coordinates, accurately positioning the source), energy (total energy released by the source) and frequency (number of microseismic events per unit time).

[0032] The data authenticity and effectiveness are ensured: the original waveform data contains a large amount of interference signals, and if directly used, the subsequent analysis will be distorted; the four types of source parameters extracted after processing are core indexes for quantifying microseismic characteristics, and provide standardized data for subsequent spatial analysis and energy gradient calculation.

[0033] Temporal Correlation: Each seismic source parameter is timestamped and mapped one-to-one with the coal mine mining progress (such as working face advance distance, advance time, and mining procedures), clarifying the microseismic event occurring at a specific moment and corresponding to the mining location and the specific procedure performed. Spatial Correlation: The three-dimensional location coordinates of the seismic source are matched with the geological spatial model of the mine (such as coal seam thickness, rock strata distribution, and fault location) to clarify the geological region where the microseismic event occurred and whether it is near a fault / fold. Cube Construction: A structured spatiotemporal data cube is formed using the time axis (T) + three-dimensional spatial axes (X / Y / Z) + seismic source parameter axes (energy / frequency, etc.) as the core dimensions—essentially organizing the scattered microseismic parameters into a queryable and analyzable three-dimensional database according to the logic of time and space.

[0034] To achieve data traceability and correlation: so that microseismic events are no longer isolated data points, but spatiotemporal events that can correspond to specific mining stages and geological locations.

[0035] It provides a platform for subsequent analysis: Subsequent DBSCAN clustering algorithms and spatiotemporal migration pattern analysis all need to be carried out in this cube. For example, the cube can be used to quickly query the microseismic energy changes in a certain spatial region within a certain time period, directly supporting the identification of clustered areas and the calculation of energy gradients.

[0036] Based on spatiotemporal data cubes, an improved clustering algorithm is used to analyze the spatial distribution characteristics of microseismic events, identify concentrated areas of microseismic events, and extract the vibration energy gradient and event density of each area.

[0037] Based on spatiotemporal data cubes, an improved clustering algorithm is used to analyze the spatial distribution characteristics of microseismic events, identify concentrated areas of microseismic events, and extract the seismic energy gradient and event density of each area. Specifically, the density-based DBSCAN clustering algorithm is used to identify spatial clusters of microseismic events; the energy gradient value and event density value of each cluster are calculated; and combined with geological structural information, spatial correlation analysis is performed between the clusters and known geological structures such as faults and folds.

[0038] It should be noted that, considering the uneven density and geological interference points in mine microseismic data, the traditional DBSCAN algorithm has been optimized (e.g., dynamically adjusting the neighborhood radius and minimum point threshold) to adapt to the microseismic distribution characteristics in complex geological environments.

[0039] Based on the three-dimensional spatial coordinates (X / Y / Z) in the spatiotemporal data cube, microseismic events with similar distances and sufficient density are automatically clustered to form independent spatial clusters, while scattered isolated microseismic points (interference or risk-free events) are removed.

[0040] This approach overcomes the limitations of isolated analysis of individual microseismic events, precisely locating concentrated microseismic event zones. These zones often directly reflect stress concentration or fault activity and are high-risk candidate areas for rockbursts. Energy gradient values: Calculate the energy difference rate between different sub-regions within a single cluster, reflecting the degree of energy concentration and the intensity of energy changes within the region (e.g., a large gradient value indicates a rapid increase in energy from the edge to the center within a region). Event density values: Statistically count the number of microseismic events per unit volume (e.g., 10m × 10m × 10m), quantifying the event density of the cluster.

[0041] Replacing qualitative descriptions with quantitative indicators visualizes the risk characteristics of clustered areas. Areas with large energy gradients and high event density indicate intense rock fracturing activity, rapid energy accumulation, and a higher probability of rockbursts.

[0042] This method utilizes existing geological exploration data from the mine (such as fault distribution maps and fold location coordinates) and spatial coordinate matching to determine whether each microseismic cluster overlaps with, is adjacent to, or is distributed along known geological structures. The output correlation results are as follows: cluster A is distributed along fault F1; cluster B is located near the fold axis; and cluster C has no obvious geological structural correlation. The core objective is to establish a direct correlation between microseismic spatial characteristics and geological genesis, providing crucial evidence for subsequent disaster type determination (e.g., fault-activated types require cluster-fault correlation), and avoiding ignoring geological origins based solely on data characteristics.

[0043] A fault activation risk assessment model was constructed. Based on the spatiotemporal migration patterns and energy release characteristics of microseismic events, fault activation tendency indicators were calculated, and different levels of fault activation hazard zones were delineated in conjunction with geological structural data.

[0044] A fault activation risk assessment model was constructed. Based on the spatiotemporal migration patterns and energy release characteristics of microseismic events, fault activation tendency indices were calculated. Combined with geological structural data, fault activation hazard zones of different levels were delineated. Specifically, the spatiotemporal migration patterns of microseismic events along the fault plane were analyzed, and the migration speed and direction of the events were calculated. The energy accumulation rate and release characteristics of the fault area were assessed. The fuzzy comprehensive evaluation method was used to quantify the fault activation tendency index into three levels: high risk, medium risk, and low risk.

[0045] Based on the established spatial relationship between clustered areas and faults, this study focuses on microseismic events distributed along and around the fault plane. The three-dimensional coordinates and timestamps of these events are extracted to construct a correlation dataset of fault plane-time-microseismic location. Using trajectory analysis algorithms, the movement paths of microseismic events along the fault plane are fitted, and the migration distance (migration velocity) and dominant movement direction (e.g., along the fault strike or dip) of the event clusters per unit time are calculated. The directional and accelerated migration of microseismic events along the fault is a direct precursor to fault slippage or activation due to mining disturbances. Migration parameters can provide a preliminary assessment of whether the fault is in an active state.

[0046] Identify the microseismic accumulation area in and around the fault plane, statistically analyze the total microseismic energy (energy accumulation rate) in this area per unit time, and simultaneously analyze the energy release pattern: is it a sudden, concentrated release (a dense cluster of high-energy events in a short period) or a slow, continuous release (a uniform distribution of low-energy events)? Compare the energy characteristics of non-fault areas to clarify the energy anomalies in the fault area (e.g., a significantly higher accumulation rate than other areas, or a pulsed release pattern). The essence of fault activation is the energy release after stress accumulates to a critical state. The energy accumulation rate reflects the rate of risk escalation, while the release characteristics determine the urgency of activation; together, they constitute the energy dimension indicators of fault risk.

[0047] Migration speed, migration direction concentration, energy accumulation rate, and energy release suddenness coefficient were selected as core evaluation indicators. Combined with mine geological structure data (such as fault size, dip angle, and historical activation records), weights were assigned to each indicator. A fuzzy comprehensive evaluation model was used to transform the quantitative data of each indicator into a fuzzy matrix. After calculation, a comprehensive score was output, and then classified into three levels according to preset thresholds: high risk (fault highly likely to activate), medium risk (expected activation), and low risk (stable with no signs of activation). This addresses the problems of difficulty in quantifying fault activation risk and subjective judgment, integrating multi-dimensional qualitative and quantitative information into a clear risk level, providing an intuitive basis for subsequent disaster classification and prevention and control measures.

[0048] By employing multi-parameter fusion technology, microseismic characteristic parameters are integrated with mine pressure data and support resistance data to establish a rockburst disaster classification matrix, classifying rockburst disasters into three major categories: fault activation type, stress concentration type, and composite type.

[0049] Multi-parameter fusion technology was employed to integrate microseismic characteristic parameters with mine pressure data and support resistance data to establish a rockburst disaster classification matrix. Rockburst disasters were categorized into three main types: fault-activated, stress-concentrated, and composite. Specifically, the following steps were taken: extracting characteristic parameters from microseismic monitoring, including event concentration, energy release rate, and b-value; collecting mine pressure data, including support working resistance, roof subsidence, and coal seam stress; performing dimensionality reduction on multi-source parameters using principal component analysis to extract key feature vectors; and constructing a classification matrix to classify rockburst disasters into three categories—fault-activated, stress-concentrated, and composite—based on the combination patterns of the feature vectors.

[0050] Event Concentration: The proportion of high-energy events within a specific area (e.g., a microseismic cluster) reflects the degree of energy concentration. Energy Release Rate: The total energy released by microseismic events per unit time, reflecting the pace of energy release. b-value: Calculated based on the Gutenberg-Richard law, reflecting the magnitude distribution characteristics of microseismic events; a smaller b-value indicates a higher proportion of high-magnitude events and more intense stress accumulation. From the three dimensions of distribution, pace, and intensity, the core characteristics of microseismic data are extracted to form a set of microseismic dimensional indicators for disaster classification.

[0051] By utilizing existing monitoring equipment in the mine, key mechanical data from the working face are collected in real time. Support working resistance reflects the stress state of the support structure, roof subsidence reflects roof stability, and coal stress directly reflects the degree of stress accumulation in the coal body. Supplementing this mechanical data with microseismic parameters complements the existing data. Microseismic data reflects the results of rock fracturing, while mine pressure data reflects the process of stress application. The combination of these two data provides a more comprehensive characterization of the causes of disasters.

[0052] Microseismic characteristic parameters (3 types) and mine pressure data (3 types) are integrated into a 6-dimensional original parameter set. Principal component analysis (PCA) is used to remove redundant information and correlation interference between parameters, and 2-3 principal components with the largest variance contributions are selected to form the main feature vectors. This solves the problems of multi-parameter dimensional redundancy and information overlap by replacing complex original parameters with a small number of core feature vectors, which simplifies calculations, preserves key classification information, and avoids overfitting.

[0053] Using the main feature vectors as the rows / columns of a matrix, threshold ranges are set for different feature vectors (e.g., a principal component above X is a strong feature, and below Y is a weak feature), forming a standardized classification matrix. Classification rules are matched based on the combination patterns of feature vectors: for example, a strong fault activation-related feature vector + a weak stress concentration feature vector corresponds to a fault activation type, and vice versa for a stress concentration type; if both are strong, it is a composite type. This transforms multi-dimensional, abstract parameter data into clear-cut, binary classification standards, enabling automated and standardized determination of disaster types, replacing traditional subjective experience-based judgments.

[0054] A classification matrix was constructed, and rockburst disasters were divided into three categories based on the combination pattern of feature vectors: fault activation type, stress concentration type, and composite type. Specifically: Fault activation type: microseismic events are distributed along fault zones, energy release is concentrated, and the fault activation tendency index is high; Stress concentration type: microseismic events are concentrated near the mining face, energy gradient changes drastically, and mine pressure is obvious; Composite type: it has the characteristics of both fault activation and stress concentration, microseismic events are spatially dispersed, and multiple precursor features coexist.

[0055] The spatial distribution of microseismic events extends along fault zones, directly matching the clustered region-fault correlation results; energy release is characterized by concentrated bursts; and the calculated fault activation tendency index reaches a high / medium risk level.

[0056] The key basis for judgment is the strong correlation between microseismic distribution and faults. Energy characteristics and fault activation indicators serve as auxiliary verification. All three point to the core cause of fault slippage or activation due to disturbance.

[0057] Microseismic events are concentrated around the mining face (without obvious fault association), exhibiting strong spatial clustering; the energy gradient changes drastically, reflecting rapid local stress accumulation; and mine pressure data shows significant anomalies. Key identification points: The core cause is localized stress concentration due to mining activities, without fault activation. When determining the cause, it is crucial to match the spatial correspondence between the microseismic cluster area and the mining face, as well as the anomalous responses of mechanical parameters.

[0058] It simultaneously exhibits the key characteristics of the first two types of disasters: both a microseismic cluster zone distributed along the fault (signs of fault activation) and an independent high-energy microseismic concentration zone in the mining area (signs of stress concentration); the spatial distribution of microseismic activity is generally dispersed (no single dominant cluster area); and multiple parameters, such as fault activation tendency indicators, energy gradient, and mine pressure, all show anomalies (superposition of precursor features). The cause is the superposition of fault activation and stress concentration. The key to judgment is that both core characteristics must be met simultaneously, rather than a single characteristic dominating. Cross-validation of multiple parameters is necessary to rule out misjudgments based on a single cause.

[0059] Based on the classification results of rockburst disasters, a disaster prevention and control decision map is generated, and corresponding prevention and control measures are recommended for different disaster types.

[0060] The disaster prevention and control decision map is generated by integrating the classification results with a three-dimensional geological model to visualize the spatial distribution of different disaster types. For fault activation type disasters, pressure relief blasting and fault grouting are recommended. For stress concentration type disasters, roof pre-fracturing and coal body pressure relief are recommended. For compound disasters, comprehensive prevention and control measures are recommended, including optimizing mining parameters and strengthening support resistance.

[0061] The existing 3D geological model of the mine (including spatial information on coal seams, rock strata, faults, and mining roadways) is used to overlay the identified three types of hazards (fault activation type, stress concentration type, and combined type) onto the model according to their actual spatial locations (such as fault zones and the periphery of mining faces). Different colors (e.g., red = high risk, yellow = medium risk, blue = low risk) are used to label the hazard type and risk level, and key information such as microseismic accumulation areas and high-energy areas are simultaneously labeled to form a 3D hazard prevention and control decision map. This transforms abstract classification results into an intuitive spatial map, allowing technicians to quickly locate high-risk areas and identify hazard types without analyzing complex data, thus lowering the professional threshold for on-site decision-making.

[0062] Based on the core causes of each type of disaster, targeted prevention and control measures are matched to form a precise correspondence between type and measure: Fault activation type: Recommended pressure relief blasting (releasing the elastic potential energy accumulated in the fault area and alleviating the slippage trend) + fault grouting (strengthening the fault fracture zone through grouting, improving fault stability, and preventing further activation). Stress concentration type: Recommended roof pre-fracturing (breaking the integrity of the hard roof strata through blasting or hydraulic fracturing, releasing the pressure transmission from the roof to the coal body) + coal body pressure relief (such as borehole pressure relief, hydraulic fracturing, directly releasing the concentrated stress inside the coal body). Composite type: Recommended comprehensive prevention and control measures (integrating the core logic of the first two types of measures), while focusing on optimizing mining parameters (such as adjusting the working face advance speed, optimizing roadway layout, and reducing disturbance to faults and stress concentration areas) + strengthening support resistance (selecting high-strength supports and increasing support density to resist the impact loads caused by the dual causes). To avoid the traditional one-size-fits-all approach to prevention and control, measures should be directly matched to the causes of disasters—fault activation type focuses on stabilizing faults and releasing slip energy, stress concentration type focuses on releasing local stress, and composite type focuses on taking into account both causes and strengthening the ability to withstand shocks, thus ensuring the effectiveness of the measures.

[0063] Based on multi-source microseismic monitoring data and classification results, an LSTM neural network is used to establish a dynamic early warning model to predict the development trend of rockburst disasters and to update the classification matrix parameters regularly.

[0064] It should be noted that multi-source microseismic monitoring data (source energy, location, etc.), multi-parameter fusion data (microseismic characteristics + mine pressure + support resistance), and historical disaster classification results are all used as input datasets for model training and prediction. This ensures that the early warning model is not only based on raw monitoring data but also incorporates validated classification conclusions, guaranteeing a strong correlation between prediction results and disaster causes and types, thus improving prediction accuracy. Leveraging the strength of LSTM neural networks in processing time-series data, a dynamic early warning model is constructed by training the input multi-source time-series data. The prediction objective is to output the development trend of rockburst disasters, including changes in disaster risk level (increasing / decreasing / stabilizing) over a future period (e.g., 1 week, 1 month), the migration direction of high-risk areas, and the predicted rate of energy accumulation, rather than simply determining whether an event will occur. This overcomes the limitations of traditional static post-event analysis and early warning, allowing for advance prediction of disaster evolution patterns and providing more preparation time for on-site prevention and control.

[0065] Based on the prediction results of the LSTM model and feedback from real-time monitoring data (such as whether a disaster actually occurred and whether the disaster type is consistent with the prediction), the key parameters of the classification matrix (such as feature vector thresholds and indicator weights) are adjusted periodically (e.g., monthly, quarterly, or according to mining progress milestones). This allows the classification matrix to adapt to changes in geological conditions during mining (e.g., stress redistribution due to mining progress) and changes in microseismic activity patterns, avoiding the decrease in accuracy caused by unchanging classification standards and achieving continuous improvement in classification and early warning accuracy.

[0066] Example 2 presents a large-scale rockburst disaster classification system based on microseismic monitoring according to the present invention, including a data acquisition module, a spatial analysis module, a fault evaluation module, and a decision support module: The data acquisition module collects multi-source microseismic monitoring data from coal mine working faces, performs preprocessing, and constructs a spatiotemporal data cube; the spatial analysis module uses an improved clustering algorithm to analyze the spatial distribution characteristics of microseismic events and identify concentrated areas of microseismic events; the fault evaluation module constructs a fault activation risk assessment model, calculates fault activation tendency indices, and delineates fault activation hazard zones; the disaster classification module uses multi-parameter fusion technology to establish a rockburst disaster classification matrix, classifying rockburst disasters into different types; and the decision support module generates a disaster prevention and control decision map and recommends corresponding prevention and control measures for different disaster types.

[0067] The data acquisition module undertakes the acquisition and preprocessing of multi-source microseismic data. It is responsible for obtaining raw data from the monitoring network and constructing a spatiotemporal data cube after processing. This serves as the system's data entry point, providing standardized, high-quality foundational data for all subsequent analysis stages, ensuring data integrity and reliability.

[0068] The spatial analysis module performs spatial distribution characteristic analysis of microseismic events. It identifies microseismic clustering areas using an improved DBSCAN clustering algorithm, extracts key indicators such as energy gradient and event density, and correlates them with geological structural information. The system's spatial feature mining tool transforms raw data into spatialized indicators reflecting risk targets, providing a basis for fault assessment and hazard classification.

[0069] The fault assessment module runs a fault activation risk assessment model. Based on spatial analysis results, it analyzes the spatiotemporal migration patterns and energy release characteristics of microseisms, quantifies fault activation tendency indicators, and classifies high, medium, and low risk zones. The system's dedicated fault risk assessment unit specifically targets fault activation as a core cause of disasters, outputting accurate risk classification results.

[0070] The disaster classification module implements multi-parameter fusion and classification, integrating quantitative results from spatial analysis and fault evaluation, and incorporating mine pressure and support resistance data. After dimensionality reduction, it outputs three types of disaster classification results—fault activation type, stress concentration type, and composite type—through a classification matrix. The system's core classification engine realizes the transformation from multi-dimensional data to clear disaster types, serving as a key hub connecting analysis and decision-making.

[0071] The decision support module executes prevention and control decision outputs, generating a 3D visualized disaster prevention and control decision map, and pushes differentiated prevention and control measures based on disaster classification results. It also connects to dynamic early warning systems, providing a data feedback channel for the LSTM model. The system's output transforms the analysis and classification results into decision products that can be directly applied in the field, realizing the implementation of technical methods into engineering practice.

[0072] Furthermore, the data acquisition module includes: a microseismic monitoring unit: a sensor network deployed above and below ground for acquiring waveform data of microseismic events; a data preprocessing unit: performing noise reduction and filtering on the waveform data to extract source parameters; and a spatiotemporal construction unit: associating the source parameters with the mining progress sequence to construct a three-dimensional spatiotemporal data cube.

[0073] It should be noted that the microseismic monitoring unit is essentially a sensor network deployed both above and below ground. Specifically, it includes short-range sensors deployed in key areas such as underground mining faces and fault zones, and a wide-area coverage sensor network deployed on the surface, forming a three-dimensional monitoring network. These sensors capture microseismic waveform signals generated by rock fracturing in real time, directly acquiring unprocessed raw data. This addresses the issue of data source, ensuring wide-area coverage of raw waveform data and accurate capture of key areas through the three-dimensional network, avoiding the omission of crucial microseismic events and providing complete raw material for subsequent processing.

[0074] The data preprocessing unit specifically receives the raw waveform data acquired by the microseismic monitoring unit and performs two core operations: denoising (removing interference signals such as blasting and mechanical operation) and filtering (selecting specific frequency waveforms related to rock fracture). Based on the purified waveforms, standardized source parameters (magnitude, location, energy, and frequency) are extracted using seismic algorithms. This addresses the problems of disorganized raw data and abundant invalid information by ensuring data authenticity through purification and transforming waveform signals into analyzable quantitative indicators through parameter refinement, providing standardized data for subsequent spatiotemporal correlation.

[0075] The spatiotemporal construction unit receives the source parameters output from the data preprocessing unit and precisely spatiotemporally binds them to the coal mine mining sequence (such as working face advancement time, advancement distance, and mining procedures). Each source parameter is labeled with a timestamp, spatial coordinates, and mining stage. Based on the bound data, a three-dimensional spatiotemporal data cube with a time axis, a three-dimensional spatial axis, and a parameter axis is constructed. This solves the problem of scattered and isolated data by making each microseismic parameter traceable and associative structured data through spatiotemporal correlation. The final output is the core data carrier of the entire system, the spatiotemporal data cube, which directly connects to the subsequent spatial analysis module.

[0076] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0077] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0078] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0079] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0080] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0081] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A classification method for large-scale rockburst hazards based on microseismic monitoring, characterized in that, Includes the following steps: Multi-source microseismic monitoring data of coal mine working face are collected, including source energy, source location, event frequency and waveform characteristics, and the multi-source microseismic monitoring data are preprocessed to construct a spatiotemporal data cube; Based on the spatiotemporal data cube, an improved clustering algorithm is used to analyze the spatial distribution characteristics of microseismic events, identify concentrated areas of microseismic events, and extract the vibration energy gradient and event density of each area. Specifically: The density-based DBSCAN clustering algorithm was used to identify spatial clusters of microseismic events. Calculate the energy gradient and event density values ​​for each cluster region; By combining geological structural information, spatial correlation analysis is performed between the clustered areas and known faults and folds; A fault activation risk assessment model was constructed. Based on the spatiotemporal migration patterns and energy release characteristics of microseismic events, fault activation tendency indicators were calculated, and different levels of fault activation hazard zones were delineated in conjunction with geological structural data. Based on the spatial correlation analysis and the identified fault activation hazard zones, a multi-parameter fusion technique is used to integrate microseismic characteristic parameters with mine pressure data and support resistance data to establish a rockburst disaster classification matrix, classifying rockburst disasters into three major categories: fault activation type, stress concentration type, and composite type. Based on the classification results of rockburst disasters, a disaster prevention and control decision map is generated, and corresponding prevention and control measures are recommended for different disaster types.

2. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 1, characterized in that, The multi-source microseismic monitoring data collected from the coal mine working face includes source energy, source location, event frequency, and waveform characteristics. The multi-source microseismic monitoring data is then preprocessed to construct a spatiotemporal data cube, specifically as follows: A combined microseismic monitoring network was deployed both above and below ground to collect waveform data of microseismic events at the working face; The waveform data is denoised and filtered to extract source parameters including magnitude, location, energy, and frequency. The processed seismic source parameters are spatiotemporally correlated with the mining progress sequence to construct a three-dimensional spatiotemporal data cube.

3. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 2, characterized in that, The aforementioned fault activation risk assessment model calculates fault activation tendency indices based on the spatiotemporal migration patterns and energy release characteristics of microseismic events, and delineates fault activation hazard zones of different levels in conjunction with geological structural data. Specifically: Analyze the spatiotemporal migration patterns of microseismic events along the fault plane, and calculate the event migration velocity and direction; Assess the energy accumulation rate and release characteristics in the fault region; The fuzzy comprehensive evaluation method was used to quantify the fault activation tendency index into three levels: high risk, medium risk, and low risk.

4. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 3, characterized in that, The method employs multi-parameter fusion technology to integrate microseismic characteristic parameters with mine pressure data and support resistance data, establishing a rockburst disaster classification matrix. Rockburst disasters are categorized into three main types: fault-activated, stress-concentrated, and composite. Extract characteristic parameters from microseismic monitoring, including event concentration, energy release rate, and b-value; Collect mine pressure data, including support working resistance, roof subsidence, and coal stress; Principal component analysis is used to reduce the dimensionality of multi-source parameters and extract the main feature vectors. A classification matrix was constructed, and rockburst disasters were classified into three categories based on the combination patterns of feature vectors: fault activation type, stress concentration type, and composite type.

5. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 4, characterized in that, The construction of the classification matrix, based on the combination pattern of feature vectors, classifies rockburst disasters into three main categories: fault activation type, stress concentration type, and composite type. Fault-activated type: Microseismic events are distributed along the fault zone, with concentrated energy release and a high fault activation tendency index; Stress concentration type: Microseismic events are concentrated near the mining face, the energy gradient changes drastically, and the mine pressure is obvious; Composite type: It has the characteristics of fault activation and stress concentration at the same time, the microseismic events are spatially dispersed, and multiple precursor features coexist.

6. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 5, characterized in that, The generation of the disaster prevention and control decision map specifically involves: The classification results are integrated with a three-dimensional geological model to visualize the spatial distribution of different disaster types; For fault-activated disasters, pressure relief blasting and fault grouting are recommended. For stress concentration disasters, it is recommended to take measures such as roof pre-fracturing and coal body decompression. For complex disasters, comprehensive prevention and control measures are recommended, including optimizing mining parameters and strengthening support resistance.

7. The method for classifying large-scale rockburst hazards based on microseismic monitoring according to claim 6, characterized in that, Also includes: Based on the multi-source microseismic monitoring data and classification results, an LSTM neural network is used to establish a dynamic early warning model to predict the development trend of rockburst disasters and to update the classification matrix parameters regularly.

8. A large-scale rockburst hazard classification system based on microseismic monitoring, applied to the large-scale rockburst hazard classification method based on microseismic monitoring described in any one of claims 1-7, characterized in that, It includes a data acquisition module, a spatial analysis module, a fault assessment module, and a decision support module. Data acquisition module: used to collect multi-source microseismic monitoring data from coal mine working faces, perform preprocessing, and construct a spatiotemporal data cube; Spatial Analysis Module: Used to analyze the spatial distribution characteristics of microseismic events using an improved clustering algorithm, and to identify areas where microseismic events are concentrated; Fault assessment module: used to construct a fault activation risk assessment model, calculate fault activation tendency index, and delineate fault activation hazard zones; Disaster classification module: Used to establish a rockburst disaster classification matrix using multi-parameter fusion technology, classifying rockburst disasters into different types; Decision support module: Used to generate disaster prevention and control decision maps and recommend corresponding prevention and control measures for different disaster types.

9. A large-scale rockburst disaster classification system based on microseismic monitoring according to claim 8, characterized in that, The data acquisition module includes: Microseismic monitoring unit: A sensor network deployed above and below ground level to collect waveform data of microseismic events; Data preprocessing unit: performs noise reduction and filtering on waveform data, and extracts source parameters; Spatiotemporal construction unit: Correlate seismic source parameters with the mining progress sequence to construct a three-dimensional spatiotemporal data cube.

Images