Rock burst early warning method based on causal mechanism

Abstract

This invention relates to the field of rockburst early warning technology, specifically to a method for sensing early warning precursors of rockbursts based on causal mechanisms. The method includes: constructing a multi-dimensional feature matrix based on multi-source heterogeneous data; identifying the true causal relationships between day-level, hour-level, and grade-level parameters using a conditional independence test; determining causal directions through structural analysis to construct day-level, hour-level, and grade-level causal transmission networks; quantifying the causal intensity of each causal transmission path; extracting the corresponding key causal transmission paths at the day-level, hour-level, and grade-level; obtaining day-level rockburst early warning signals; issuing hour-level rockburst early warning signals after obtaining day-level rockburst early warning signals; and issuing grade-level emergency rockburst early warning signals after obtaining hour-level rockburst early warning signals. This invention constructs a complete time-gradient rockburst early warning system from day-level early warning to graded imminent warning, forming a complete early warning time chain.

Description

Technical Field

[0001] This invention relates to the field of rockburst early warning technology, specifically to a method for sensing early warning signs of rockbursts based on causal mechanisms. Background Technology

[0002] Currently, there are three main technical solutions in the field of rockburst precursor sensing: First, the single-parameter threshold monitoring method, which uses fixed thresholds based on single parameters such as microseismic energy, stress changes, and displacement rates for precursor identification; second, the multi-parameter statistical analysis method, which uses statistical methods such as principal component analysis and cluster analysis to comprehensively analyze multi-parameter monitoring data; and third, the machine learning recognition method, which uses algorithms such as neural networks and support vector machines to establish a recognition model between monitoring parameters and rockburst precursors. These methods constitute the main technical routes of current rockburst precursor sensing technology, but they have the following drawbacks:

[0003] (1) The precursor recognition time is seriously delayed. Existing technologies can generally only identify precursor signals when rockburst is about to occur (5-30 minutes in advance). The precursor recognition time is seriously delayed, and it is impossible to achieve true advance perception. It can only provide emergency warnings and lacks sufficient time window to take effective preventive measures.

[0004] (2) Lack of deep integration capability of multi-source information. Existing technologies usually identify precursors based on a single type of monitoring data, or only perform simple multi-parameter overlay analysis, which cannot effectively integrate heterogeneous information from multiple sources such as geology, mining, and monitoring, resulting in insufficient comprehensiveness and accuracy of precursor identification.

[0005] (3) Lack of scientific support for causal mechanisms. Existing technologies rely on statistical correlation for precursor identification, but lack a deep understanding of the formation mechanism of rockbursts and cannot reveal the complete causal chain from the root cause to the manifestation of precursors, resulting in insufficient reliability and interpretability of precursor identification.

[0006] In summary, existing methods for detecting precursors of rockbursts generally suffer from key shortcomings such as delayed identification, weak fusion capabilities, and a lack of mechanistic support. These technical bottlenecks severely restrict the effectiveness of early warning systems for rockbursts. There is an urgent need to develop a new generation of precursor sensing technologies with advanced sensing capabilities, deep multi-source fusion, and a causal mechanism-based approach. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention proposes a method for sensing the precursors of rockbursts based on causal mechanisms, comprising the following steps:

[0008] S1. Construct a multi-dimensional feature matrix based on multi-source heterogeneous data;

[0009] S2. Identify the true causal relationships among parameters at the daily, time, and grade levels using the conditional independence test; construct causal transmission networks at the daily, time, and grade levels by determining causal directions through structural analysis; quantify the causal strength of each causal transmission path; and extract the corresponding key causal transmission paths at the daily, time, and grade levels.

[0010] S3. Obtain a day-level early warning for rockbursts; after obtaining a day-level early warning signal for rockbursts, issue a time-level early warning for rockbursts; after obtaining a time-level early warning signal for rockbursts, issue a graded emergency early warning for rockbursts.

[0011] Preferably, step S1 includes: acquiring multi-source heterogeneous data; the multi-source heterogeneous data includes geological exploration data, mining planning and support data, and safety monitoring data; data quality control and spatiotemporal alignment to ensure complete consistency in spatiotemporal dimensions.

[0012] Preferably, step S1 further includes: constructing a multidimensional feature matrix: extracting a comprehensive feature vector from the spatiotemporally aligned data using multi-domain feature engineering methods, and performing dimensionality reduction using principal component analysis and independent component analysis methods to obtain a multidimensional feature matrix. Each row of the multidimensional feature matrix represents a sample, and the number of proxy feature dimensions is the number of parameters.

[0013] Preferably, in step S2, identifying the true causal relationship between day-level, time-level, and grade-level parameters using the conditional independence test method includes: obtaining standardized multidimensional feature matrices at different time scales through a sliding window, where the day-level time span covers the number of samples at the day level, the time-level time span covers the number of samples at the time level, and the grade-level time span covers the number of samples at the grade level; and introducing a condition set using the conditional independence test method to control confounding factors in order to accurately identify the direct causal relationship between the corresponding parameters at the day-level, time-level, and grade-level.

[0014] Preferably, in step S2, the step of determining the causal direction through structural analysis and constructing the day-level, time-level, and hierarchical causal transmission network includes: determining the causal direction by analyzing the conditional dependency patterns between parameters, and constructing the day-level, time-level, and hierarchical causal transmission network based on the causal direction; firstly, finding a V-shaped structure composed of three parameters, and then calculating a Bayesian information criterion score for each possible combination of directed edges by constructing all possible combinations, and selecting the direction with the highest score as the final causal direction.

[0015] Preferably, in step S2, the quantification of the causal strength of each causal transmission path includes: for parameters with causal relationships, setting a reasonable range of intervention amount, calculating the expected values ​​of the target parameters before and after intervention, and taking the difference between the two as the direct causal strength; calculating the indirect causal effect, that is, the influence of a parameter on the target parameter through an intermediary parameter; and calculating the total causal effect between all parameter pairs through chain transmission rules and matrix operation methods to form a complete causal strength matrix.

[0016] Preferably, in step S2, the extraction of key causal transmission paths corresponding to the daily, time-based, and grade-based levels includes: establishing a directed acyclic graph to represent the topology of the causal transmission network; using a depth-first search algorithm to traverse all possible causal transmission paths from the front parameter of the causal transmission path to the precursor target parameter; calculating the transmission strength and transmission efficiency of each causal transmission path; and sorting all causal transmission paths using a weighted scoring mechanism, selecting the highest-scoring paths as key causal transmission paths at the daily, time-based, and grade-based levels, respectively.

[0017] Preferably, in step S3, obtaining the early warning of a day-level rockburst includes: continuously acquiring data of the frontmost parameters of the key causal transmission path at the day level; calculating the degree of deviation when these parameters deviate from their historical baseline level; performing weighted calculations based on the corresponding causal intensity; introducing a time decay function to highlight the importance of recent signals; forming a comprehensive early activation index; and triggering an early warning of a day-level rockburst when the activation exceeds an adaptive threshold based on historical data statistics.

[0018] Preferably, in step S3, the early activation index is determined using the following formula:

[0019] A_early(t)=∑(i=1 to m)wi×sigmoid(C_total(Xi→Y)×Δ(Xi(t),Xi_baseline))×T_weight(t)

[0020] In the formula: A_early(t): the overall early activation level at time t;

[0021] m: The total number of geological root parameters (usually m=5-10), located at the forefront of the critical causal transmission path at the celestial level;

[0022] wi: The weighting coefficient of the i-th geological root parameter;

[0023] sigmoid(): Activation function used for normalization.

[0024] C_total(Xi→Y): The intensity of the total causal effect of the geological root parameter Xi on the rockburst event Y;

[0025] Δ(Xi(t),Xi_baseline): The degree of deviation of the geological root parameter Xi from the baseline at time t, expressed as absolute difference divided by variance;

[0026] T_weight(t): Time weighting function used to calculate time decay parameters; average decay is generally used.

[0027] Preferably, in step S3, the time-level rockburst early warning includes: calculating in real time the changing trend and stability index of the parameters at the forefront of the time-level key causal transmission path using the moving average method and trend analysis method; when the trend change exceeds the set threshold or the stability index deteriorates significantly, it is judged as a sign of a turning point in the stability of the geological body.

[0028] Preferably, in step S3, the step of conducting a graded rockburst emergency warning includes: using a high-frequency sampling and mutation detection algorithm to monitor the rate of change of the parameters at the forefront of the graded key causal transmission path in real time; when the rate of change is detected to exceed three times the historical average rate of change, or when the parameters at the forefront of multiple graded key causal transmission paths simultaneously show abnormalities, a graded rockburst emergency warning is immediately triggered.

[0029] Beneficial technical effects: This invention constructs a complete time-gradient rockburst early warning system, from day-level advanced sensing to tiered near-term warning; it amplifies weak geological root cause signals by utilizing causal transmission intensity, achieving day-level advanced warning capability; time-level identification can capture critical moments of geological body stability transitions, providing mid-term warning information; and tiered identification continues the traditional near-term warning function, ensuring final safety assurance. These three levels cooperate and progress step by step, forming a complete early warning time chain. Attached Figure Description

[0030] Figure 1 This is a structural diagram of the rockburst precursor sensing method based on causal mechanism according to the present invention. Detailed Implementation

[0031] The specific embodiments of the present invention will now be described in conjunction with the accompanying drawings.

[0032] like Figure 1 As shown, the rockburst precursor sensing method based on causal mechanism proposed in this invention includes the following steps:

[0033] S1. Constructing a multidimensional feature matrix based on multi-source heterogeneous data.

[0034] S11. Acquire multi-source heterogeneous data

[0035] The multi-source heterogeneous data includes geological exploration data, mining planning and support data, and safety monitoring data. These multi-source heterogeneous data have significant differences in terms of time scale, spatial benchmark, and sampling accuracy (frequency). Direct fusion analysis will lead to information distortion and analysis bias.

[0036] The geological exploration data includes coal seam geological information and strata geological and structural information. This data is static background information, represented in a geographic coordinate system, with a low update frequency but high information density. The coal seam geological information includes geological parameters related to coal seam characteristics, such as coal seam thickness, coal seam strength, and main components. The strata geological and structural information includes parameters related to strata geology and geological structure, such as strata type, stacking pattern, rock mass strength, fault structure, and hydrological information.

[0037] The mining planning and support data includes mining planning data and support scheme data. The mining planning data includes parameters related to working face mining and roadway excavation, such as working face layout, roadway arrangement, advance plan, mining speed, and mining parameters. The support scheme data includes parameters related to roadway support, such as support method and anchor bolt support layout. The mining planning and support data is dynamic disturbance information, represented in an engineering coordinate system, and updated according to the planned cycle.

[0038] The safety monitoring data is data obtained through monitoring after taking corresponding monitoring measures to maintain mine safety. The safety monitoring data includes parameters such as microseismic signals, stress state, displacement deformation, and rock bursts. The safety monitoring data is represented in the coordinate system of the monitoring equipment, with a high sampling frequency but a relatively low signal-to-noise ratio.

[0039] S12. Data quality control and spatiotemporal alignment

[0040] The process involves quality checks on multi-source heterogeneous data, reasonable estimation and repair of substandard data, and removal of obviously erroneous data points. Then, the timestamps of the multi-source heterogeneous data are unified to a standard time reference through a time synchronization function. Interpolation is used to handle different sampling frequencies, and the position information of different coordinate systems is uniformly mapped to a standard spatial reference through a spatial coordinate transformation function to ensure complete consistency in spatiotemporal dimensions.

[0041] S13. Constructing a multidimensional feature matrix

[0042] Multi-domain feature engineering methods are used to extract comprehensive feature vectors from spatiotemporally aligned data. These vectors include a multi-dimensional feature set, such as time-domain features reflecting the basic statistical laws of the data, frequency-domain features reflecting the frequency components of the data, and wavelet transform features reflecting the multi-scale characteristics of the data. To avoid computational complexity caused by excessive dimensionality, dimensionality reduction methods such as principal component analysis and independent component analysis are employed to optimize the feature dimensions while maintaining maximum data content. The final result is a standardized multi-dimensional feature matrix X∈R^(n×p) suitable for causal relationship discovery, where n is the number of samples and p is the feature dimension, i.e., the number of parameters.

[0043] S2. Construct causal transmission networks at the celestial, temporal, and hierarchical levels, and quantify the strength of each transmission path accordingly.

[0044] S21. Identify the true causal relationship between parameters using the conditional independence test.

[0045] Because multi-source heterogeneous data contains a large number of parameters, these parameters exhibit both direct causal relationships and spurious associations caused by common causes, which must be distinguished using scientific methods. A sliding window is used to obtain standardized multidimensional feature matrices at different time scales, including days (e.g., 3-7 days), hours (e.g., 6-24 hours), and minutes (e.g., 5-60 minutes). The day-scale time span covers the number of samples at the day level, the hour-scale time span covers the number of samples at the hour level, and the minute-scale time span covers the number of samples at the minute level. A conditional independence test is used to introduce a condition set to control for confounding factors, accurately identifying the direct causal relationships between corresponding parameters at the day, hour, and minute levels. When two parameters remain dependent given all other relevant parameters, a direct causal relationship exists between them. Conversely, if they become independent under conditional control, the previously observed association is an indirect effect caused by other parameters, and no direct causal relationship exists.

[0046] S22. Determine causal directions and construct causal transmission networks through structural analysis.

[0047] After establishing a causal relationship between parameters, it is necessary to determine which parameter is the cause and which is the effect. A V-structure identification algorithm, based on the structural features of a causal graph, is employed to determine the causal direction by analyzing the conditional dependency patterns between parameters. Based on this causal direction, day-level, time-level, and hierarchical causal transmission networks are constructed. Specifically, determining the causal direction involves: first, identifying a V-shaped structure composed of three parameters (two parameters pointing to the same parameter). This structure possesses special statistical properties in causal graphs, clearly indicating the causal direction; then, by constructing all possible combinations of directed edges, calculating a Bayesian information criterion score for each combination, and selecting the direction with the highest score as the final causal direction.

[0048] S23. Quantify the causal strength of each causal transmission path.

[0049] A causal transmission path is a route in a causal transmission network. Using Pearl's do-calculus framework, a simulated intervention experiment is employed to measure the causal influence of one parameter on another, thereby quantifying the causal strength of each causal transmission path at the daily, hourly, and graded levels. This allows for the determination of corresponding monitoring priorities and early warning priorities at the daily, hourly, and graded levels based on the intensity. Specifically, this involves: First, for parameters with causal relationships, a reasonable intervention range is set (generally ±10% of the statistical average of the relationship), and the expected values ​​of the target parameter before and after the intervention are calculated. The difference between the two values ​​represents the direct causal strength. Then, the indirect causal effect is calculated, i.e., the influence of one parameter on the target parameter through an intermediary parameter. Finally, the total causal effect between all parameter pairs is calculated using chain transmission rules and matrix operations, forming a complete causal strength matrix. This causal strength matrix not only records the direct influence strength between parameters but also includes the indirect influence strength generated through multi-step transmission.

[0050] S24. Extracting key causal transmission pathways

[0051] Complex causal transmission networks contain numerous causal transmission paths, but not all of them are equally important for sensing rockburst precursors. It is necessary to identify the key causal transmission paths. Assessing the importance of causal transmission paths requires considering both transmission strength and efficiency: transmission strength reflects the amplification capability of a signal propagating along the causal transmission path; a stronger causal transmission path can amplify weak parameter changes into more significant rockburst precursor signals. Transmission efficiency considers the impact of causal transmission path length on signal attenuation; excessively long causal transmission paths cause the signal to gradually attenuate during propagation. A directed acyclic graph (DAG) is constructed to represent the topology of the causal transmission network. A depth-first search algorithm is then used to traverse all possible causal transmission paths from the leading parameter to the precursor target parameter. For each causal transmission path, its transmission strength (the product of the strengths of all causal transmissions) and transmission efficiency (considering the correction factor for path length attenuation) are calculated. All causal transmission paths are then ranked using a weighted scoring mechanism, and the top K paths with the highest scores are selected as the key causal transmission paths at the daily, hourly, and graded levels, respectively.

[0052] S3. Earthquake Burst Precursor Sensing

[0053] S31. Obtain advanced early warning of sky-level rockburst.

[0054] Rockbursts are a gradual process of energy accumulation and release. Several days before the final eruption, subtle but persistent anomalous changes occur in the geological parameters at the forefront of the critical causal transmission path at the time-scale level. By continuously acquiring data on the parameters at the forefront of this critical causal transmission path, calculating the degree of deviation when these parameters deviate from their historical baseline levels, and weighting the deviations according to the corresponding causal strength, while introducing a time decay function to highlight the importance of recent signals, a comprehensive early activation index is formed. When the activation index exceeds an adaptive threshold based on historical data statistics, the system triggers a time-scale rockburst early warning, gaining valuable time for the formulation and implementation of prevention and control measures. Preferably, the early activation index is determined using the following formula:

[0055] A_early(t)=∑(i=1 to m)wi×sigmoid(C_total(Xi→Y)×Δ(Xi(t),Xi_baseline))×T_weight(t)

[0056] In the formula: A_early(t): the overall early activation level at time t;

[0057] m: The total number of geological root parameters (usually m=5-10), located at the forefront of the critical causal transmission path at the celestial level;

[0058] wi: The weighting coefficient of the i-th geological root parameter;

[0059] sigmoid(): Activation function used for normalization.

[0060] C_total(Xi→Y): The intensity of the total causal effect of the geological root parameter Xi on the rockburst event Y;

[0061] Δ(Xi(t),Xi_baseline): The degree of deviation of the geological root parameter Xi from the baseline at time t, expressed as absolute difference divided by variance;

[0062] T_weight(t): Time weighting function used to calculate time decay parameters; average decay is generally used.

[0063] S32. Obtain time-level rockburst early warning.

[0064] After obtaining the terrestrial-level rockburst early warning signal, a time-level rockburst early warning is issued. This stage corresponds to the critical transition of the geological body from a relatively stable state to a critically unstable state, serving as a crucial bridge connecting the terrestrial-level early warning and the graded proximity identification. Within the time window, phenomena of medium time scales, such as stress field redistribution, surrounding rock state evolution, and changes in microseismic activity, gradually emerge. Although these changes are not as weak as the terrestrial-level signal or as dramatic as the graded signal, they have significant indicative significance of turning points. The changing trends and stability indicators of the parameters at the forefront of the time-level key causal transmission path are calculated in real time using the moving average method and trend analysis. When the trend change exceeds a set threshold or the stability indicator deteriorates significantly, it is judged as a sign of a turning point in the stability of the geological body.

[0065] S33. Obtain emergency early warning for graded rockbursts.

[0066] After acquiring the hourly rockburst early warning signal, a graded rockburst emergency warning is issued. This stage focuses on capturing rapidly changing signals before the critical failure of the geological body, which is the main scope of traditional precursor identification methods. However, in this invention, it plays a more significant role in final confirmation and emergency response. When a geological body approaches a critical failure state, phenomena such as microseismic energy release, abrupt changes in stress state, and accelerated deformation of surrounding rock will erupt in a short period of time, with large signal amplitudes and extremely rapid change rates. High-frequency sampling and abrupt change detection algorithms are used to monitor the rate of change of parameters at the forefront of the graded key causal transmission paths in real time. When the detected rate of change exceeds three times the historical average rate of change, or when parameters at the forefront of multiple graded key causal transmission paths simultaneously show anomalies, a graded rockburst emergency warning is immediately triggered.

[0067] This invention is not limited to the preferred embodiments described above. Anyone can derive other methods in various forms under the guidance of this invention. Any technical solution that is the same as or similar to this application falls within the protection scope of this invention.

Claims

1. A method for sensing the precursors of rockbursts based on causal mechanisms, characterized in that, Includes the following steps: S1. Construct a multi-dimensional feature matrix based on multi-source heterogeneous data; S2. Identify the true causal relationships among parameters at the daily, time, and grade levels using the conditional independence test; construct causal transmission networks at the daily, time, and grade levels by determining causal directions through structural analysis; quantify the causal strength of each causal transmission path; and extract the corresponding key causal transmission paths at the daily, time, and grade levels. S3. Obtain a day-level rockburst early warning; after obtaining a day-level rockburst early warning signal, issue a time-level rockburst warning; after obtaining a time-level rockburst warning signal, issue a graded rockburst emergency warning. The acquisition of the day-level rockburst early warning includes: continuously acquiring data of the front-end parameters of the key causal transmission path at the day level; calculating the degree of deviation when these parameters deviate from their historical baseline level; performing weighted calculations based on the corresponding causal strength; introducing a time decay function to highlight the importance of recent signals; forming a comprehensive early activation index; and triggering the day-level rockburst early warning when the activation exceeds an adaptive threshold based on historical data statistics.

2. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 1, characterized in that, Step S1 includes: acquiring multi-source heterogeneous data; the multi-source heterogeneous data includes geological exploration data, mining planning and support data, and safety monitoring data; data quality control and spatiotemporal alignment to ensure complete consistency in spatiotemporal dimensions.

3. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 2, characterized in that, Step S1 also includes: constructing a multidimensional feature matrix: extracting comprehensive feature vectors from the spatiotemporally aligned data using multi-domain feature engineering methods, and performing dimensionality reduction using principal component analysis and independent component analysis methods to obtain a multidimensional feature matrix. Each row of the multidimensional feature matrix represents a sample, and the number of proxy feature dimensions is the number of parameters.

4. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 3, characterized in that, In step S2, identifying the true causal relationships between day-level, time-level, and grade-level parameters using the conditional independence test method includes: obtaining standardized multidimensional feature matrices at different time scales through a sliding window, where the day-level time span covers the number of samples at the day level, the time-level time span covers the number of samples at the time level, and the grade-level time span covers the number of samples at the grade level; and introducing a condition set using the conditional independence test method to control confounding factors in order to accurately identify the direct causal relationships between the corresponding parameters at the day-level, time-level, and grade-level.

5. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 4, characterized in that, In step S2, the step of determining the causal direction through structural analysis and constructing the day-level, time-level, and hierarchical causal transmission network includes: determining the causal direction by analyzing the conditional dependency patterns between parameters, and constructing the day-level, time-level, and hierarchical causal transmission network based on the causal direction; firstly, finding a V-shaped structure composed of three parameters, and then constructing all possible combinations of directed edges, calculating the Bayesian information criterion score for each combination, and selecting the direction with the highest score as the final causal direction.

6. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 5, characterized in that, In step S2, the quantification of the causal strength of each causal transmission path includes: for parameters with causal relationships, setting a reasonable range of intervention amount, calculating the expected value of the target parameter before and after intervention, and taking the difference between the two as the direct causal strength; calculating the indirect causal effect, that is, the influence of a parameter on the target parameter through the intermediary parameter; and calculating the total causal effect between all parameter pairs through the chain transmission rule and matrix operation method to form a complete causal strength matrix.

7. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 6, characterized in that, In step S2, the extraction of key causal transmission paths corresponding to the daily, time-based, and grade-based levels includes: establishing a directed acyclic graph to represent the topology of the causal transmission network; using a depth-first search algorithm to traverse all possible causal transmission paths from the front parameter of the causal transmission path to the precursor target parameter; calculating the transmission strength and transmission efficiency of each causal transmission path; and sorting all causal transmission paths through a weighted scoring mechanism, selecting the highest-scoring paths as key causal transmission paths at the daily, time-based, and grade-based levels respectively.

8. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 1 or 7, characterized in that, In step S3, the early activation index is determined using the following formula: A_early(t)=∑(i=1 to m)wi×sigmoid(C_total(Xi→Y)×Δ(Xi(t),Xi_baseline))×T_weight(t) In the formula: A_early(t): the overall early activation level at time t; m: The total number of geological root parameters (usually m=5-10); wi: The weighting coefficient of the i-th geological root parameter; sigmoid(): Activation function used for normalization. C_total(Xi→Y): The intensity of the total causal effect of the geological root parameter Xi on the rockburst event Y; Δ(Xi(t),Xi_baseline): The degree of deviation of the geological root parameter Xi from the baseline at time t, expressed as absolute difference divided by variance; T_weight(t): Time weighting function used to calculate time decay parameters.

9. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 1 or 7, characterized in that, In step S3, the time-level rockburst early warning includes: calculating the change trend and stability index of the parameters at the forefront of the time-level key causal transmission path in real time using the moving average method and trend analysis method. When the trend change exceeds the set threshold or the stability index deteriorates significantly, it is judged as a sign of the geological body's stability turning point.

10. The method for sensing the precursors of rockbursts based on causal mechanisms according to claim 9, characterized in that, In step S3, the step of conducting graded rockburst emergency early warning includes: using high-frequency sampling and mutation detection algorithms to monitor the rate of change of parameters at the forefront of the graded key causal transmission path in real time; when the rate of change is detected to exceed three times the historical average rate of change, or when the parameters at the forefront of multiple graded key causal transmission paths are abnormal at the same time, the graded rockburst emergency early warning is immediately triggered.

Images