Earth rock burst advanced early warning method based on cumulative damage space-time distribution characteristics

Abstract

The invention discloses a rock burst advanced early warning method based on cumulative damage spatial-temporal distribution characteristics, and the method comprises the steps: obtaining all micro-seismic data of a to-be-predicted region, the micro-seismic data being composed of the three-dimensional coordinates and corresponding energy of each point in the predicted region; dividing the to-be-predicted area into a plurality of grids, and obtaining all microseismic data in each grid; based on all the micro-seismic data in each grid, calculating the fractal dimension of the roadway surrounding rock damage of each grid; based on all microseismic data in each grid, the number of crack planes, the crack area and the crack shape in each grid are obtained; and on the basis of the parting dimension, the number of the crack planes, the crack area and the crack shape in each grid, the rock burst danger degree of each grid is obtained, and early warning is performed according to the danger degree. According to the method, the parting dimension is obtained based on the microseismic data, meanwhile, various indexes of cracks are introduced, the danger degree of rock burst is obtained, early warning is conducted, and the timeliness and accuracy of early warning are improved.

Description

Technical Field

[0001] The invention relates to the technical field of mining engineering, and in particular to an advance warning method for rock burst based on temporal and spatial distribution characteristics of cumulative damage. Background Art

[0002] In recent years, the mining depth of coal mines has been increasing, and mining work has gradually entered more complex geological areas. Rock burst is the main dynamic disaster in my country's mining. Rock burst often leads to roof accidents, damages shafts and tunnels, causes casualties, destroys equipment, pollutes the working environment, and affects production. Rock burst sometimes leads to strong gas gushing, and sometimes causes explosions due to large amounts of dust, causing greater damage.

[0003] The existing rock burst early warning methods mainly rely on the post-perception of microseismic monitoring data. The sensitivity of traditional stress indicators in rock burst prediction is low, and monitoring and analysis have lags. The interpretability of rock burst early warning based on microseismic parameters is unclear. Mechanical mechanisms such as rock burst pressure mechanics have not been effectively combined into the early warning method, and the cumulative damage of surrounding rocks and the degree of crack expansion cannot be quantitatively and dynamically evaluated. Secondly, unlike the research on cracks by existing technologies such as hydraulic pressure, the existing technology only studies the area or shape of three-dimensional fracture networks based on microseismic events, but does not explore the impact of cracks on rock burst, and cannot use key parameters such as the degree of crack expansion and the spatiotemporal distribution of cracks to effectively evaluate the impact risk. Thirdly, in the current rock burst prediction and early warning methods, most of them use artificial judgments on the degree of danger of rock burst based on experience and make forecasts, which obviously requires a lot of manpower and material resources and has the characteristics of low efficiency.

[0004] Rock burst prediction and early warning methods are gradually developing in the direction of intelligence. The introduction of machine learning algorithms makes the extraction of disaster precursor features more accurate. The data-driven rock burst prediction model has gradually become the mainstream research direction, which can improve the accuracy and reliability of prediction in complex mining environments and better solve the complexity and nonlinearity of rock burst. Therefore, it is crucial to dynamically evaluate the impact hazard based on intelligent algorithms and realize advanced intelligent early warning of impact manifestation.

[0005] Traditional stress indicators have low sensitivity in rock burst prediction and cannot effectively capture key disaster precursors in a timely manner. Existing rock burst early warning methods mainly rely on post-analysis of microseismic monitoring data, which has a lag and is difficult to reflect the precursor characteristics of the impact. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide an early warning method for rock burst based on the temporal and spatial distribution characteristics of cumulative damage in response to the deficiencies in the above-mentioned prior art. The fractal dimension is obtained based on microseismic data, and various indicators of cracks are introduced at the same time to obtain the degree of danger of rock burst and issue an early warning, thereby improving the timeliness and accuracy of the early warning.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a rock burst advance warning method based on the temporal and spatial distribution characteristics of cumulative damage, comprising the following steps:

[0008] Step 1: Acquire all microseismic data of the area to be predicted, wherein the microseismic data consists of the three-dimensional coordinates and corresponding energy of each point in the prediction area;

[0009] Step 2: Divide the area to be predicted into multiple grids, and obtain all microseismic data in each grid;

[0010] Step 3: Based on all the microseismic data in each of the grids, calculate the fractal dimension of the tunnel surrounding rock damage in each of the grids;

[0011] Step 4: based on all microseismic data in each of the grids, obtain the number of fracture planes, fracture area and fracture shape in each of the grids;

[0012] Step 5: Based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each of the grids, the rock burst hazard level of each of the grids is obtained, and an early warning is issued according to the hazard level.

[0013] Furthermore, the area to be predicted is a mining face or a mining tunnel.

[0014] Further, in step 2, the obtaining of all microseismic data in each of the grids includes:

[0015] Determine the size of the window and the sliding step, and determine the number of the grids and the positions of the grids according to the size of the window;

[0016] Allocating the microseismic data to each of the grids according to the three-dimensional coordinates in the microseismic data;

[0017] The microseismic data in each of the grids is filtered.

[0018] Furthermore, filtering the microseismic data in each of the grids includes:

[0019] Determining the energy of each microseismic data in each of the grids, and removing the microseismic data with energy less than a set energy;

[0020] Counting the number of microseismic data in each of the grids;

[0021] Grids whose number of microseismic data is greater than a set number are screened out.

[0022] Further, in step 4, based on all microseismic data in each of the grids, the number of fracture planes, fracture area and fracture shape in each of the grids are obtained, including:

[0023] Clustering all microseismic data in the grid by using DBSCAN algorithm;

[0024] Performing spatial plane fitting on the clustered microseismic data through RANSAC to obtain a plane with cracks;

[0025] The number of cracks in the plane and the shape of each crack are extracted by using the Alpha-Shape algorithm;

[0026] The area of ​​each crack is obtained by vector cross multiplication method.

[0027] Further, in step 5, the rock burst danger level of each grid is obtained based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each grid, including:

[0028] The fractal dimension, the number of crack planes, the crack area and the crack shape in the grid are input into a pre-trained machine learning model to obtain the rock burst hazard level of the grid;

[0029] The machine learning model is used to receive the fractal dimension, the number of crack planes, the crack area and the crack shape in the grid, and output the rock burst hazard level of the grid after processing.

[0030] Further, in step 5, the early warning is issued according to the danger level, including:

[0031] comparing the risk level with a set risk level value;

[0032] If the danger level is not greater than the set danger level value, no operation is performed;

[0033] If the danger level is greater than the set danger level value, an early warning is initiated.

[0034] Further, after step 4 and before step 5, the method further includes:

[0035] The shapes of individual cracks in the mesh are visualized in three dimensions.

[0036] Compared with the prior art, the present invention has the following advantages:

[0037] The present invention provides an early warning method for rock burst based on the temporal and spatial distribution characteristics of cumulative damage. The prediction area is divided into multiple grids, and the surrounding rock damage classification dimension is obtained for each grid based on microseismic data. At the same time, the DBSCAN+RANSAC+Alpha-Shape algorithm is used to introduce various indicators of cracks, namely the number of crack planes, crack area and crack shape in each of the grids, and crack visualization operations are performed, thereby effectively moving the prediction threshold and the warning threshold forward, and constructing a Bayes-XGBoost intelligent hierarchical prediction model to improve the timeliness and accuracy of the warning.

[0038] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic block diagram of the overall process of a rock burst advance warning method based on the temporal and spatial distribution characteristics of cumulative damage provided by the present invention.

[0040] Figure 2 This is a schematic diagram of performing spatial scanning in step 2 of the present invention.

[0041] Figure 3 The figure is a flow chart of establishing the early warning model in the present invention. DETAILED DESCRIPTION

[0042] like Figure 1 As shown, the present invention provides an early warning method for rock burst based on the temporal and spatial distribution characteristics of cumulative damage, comprising the following steps:

[0043] Step 1: Acquire all microseismic data of the area to be predicted, wherein the microseismic data consists of the three-dimensional coordinates and corresponding energy of each point in the prediction area;

[0044] Step 2: Divide the area to be predicted into multiple grids, and obtain all microseismic data in each grid;

[0045] Step 3: Based on all the microseismic data in each of the grids, calculate the fractal dimension of the tunnel surrounding rock damage in each of the grids;

[0046] Step 4: based on all microseismic data in each of the grids, obtain the number of fracture planes, fracture area and fracture shape in each of the grids;

[0047] Step 5: Based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each of the grids, the rock burst hazard level of each of the grids is obtained, and an early warning is issued according to the hazard level.

[0048] Wherein, in step 1, the area to be predicted is a mining face or a mining tunnel. The acquired data is filtered and converted to obtain a point set of microseismic data (x, y, z, m) containing three-dimensional coordinates and energy. Wherein, x, y, z are three-dimensional coordinates, and m is energy.

[0049] In step 2, the spatial scanning algorithm is used to divide the grid window, and the microseismic events are assigned to the corresponding grids according to their coordinates and time, thereby forming microseismic data sets in different spatial regions.

[0050] Rock burst has significant spatial characteristics, but existing research rarely explores the identification of rock burst areas in depth. The present invention uses a spatial scanning algorithm to divide the grid window and assigns microseismic events to corresponding grids according to their coordinates and time, thereby forming microseismic data sets in different spatial regions. By analyzing the microseismic parameters in the grid window, the cumulative damage and crack propagation of the coal body in the spatial region are evaluated. The dynamic evolution of rock burst in the spatial region of the tunneling roadway can be tracked, thereby accurately realizing advance warning of impact danger. The schematic diagram of spatial scanning is shown in Figure 2 shown.

[0051] The spatial scanning method is as follows:

[0052] 1) First, determine the scope of the spatial scan and set the length and width of the working surface rectangle to accommodate all microseismic event points to form an enclosing grid.

[0053] 2) Determine the size and sliding step of the grid window. Determine the number of grids and the coordinate position of each grid according to the length and width of the enclosing grid.

[0054] 3) Allocate event points to corresponding grids. Screen data points one by one, determine whether they are within the enclosing grid based on their coordinate positions, and then transfer the point information (including date, coordinates, and energy) to the corresponding grid.

[0055] The magnitude of each event in each grid is determined. In order to eliminate the influence of small energy events on the sample, points with single event energy less than 100 J are deleted, and grids with more than 50 data points are screened to form a grid data point set.

[0056] That is, the step of obtaining all microseismic data in each of the grids in step 2 includes:

[0057] Determine the size of the window and the sliding step, and determine the number of the grids and the positions of the grids according to the size of the window;

[0058] Allocating the microseismic data to each of the grids according to the three-dimensional coordinates in the microseismic data;

[0059] The microseismic data in each of the grids is filtered.

[0060] The filtering of the microseismic data in each of the grids includes:

[0061] Determining the energy of each microseismic data in each of the grids, and removing the microseismic data with energy less than a set energy;

[0062] Counting the number of microseismic data in each of the grids;

[0063] Grids whose number of microseismic data is greater than a set number are screened out.

[0064] In step 3, the fractal dimension of the tunnel surrounding rock damage in different grid windows is calculated using the fractal dimension theory and algorithm. Fractal dimension is a dimensional indicator used to describe complex geometric forms, which is a generalization of the conventional Euclidean dimension. Fractal dimension is often used to quantify irregular or broken geometric forms in nature, such as terrain, cracks, stock market fluctuations, etc. It reflects the "fractal characteristics" of a form by measuring the complexity of a form at different scales. The present invention uses the total number method to calculate the fractal dimension in different spatial regions, that is, to divide a geometric shape into multiple boxes, calculate the filling conditions of these boxes at different scales, and obtain the fractal dimension through scale change analysis.

[0065] In step 4, based on all the microseismic data in each of the grids, the number of fracture planes, fracture area and fracture shape in each of the grids are obtained, including:

[0066] Clustering all microseismic data in the grid by using DBSCAN algorithm;

[0067] Performing spatial plane fitting on the clustered microseismic data through RANSAC to obtain a plane with cracks;

[0068] The number of cracks in the plane and the shape of each crack are extracted by using the Alpha-Shape algorithm;

[0069] The area of ​​each crack is obtained by vector cross multiplication method.

[0070] The present invention uses DBSCAN to cluster the spatial point set, uses the point set obtained by RANSAC clustering to perform spatial plane fitting (fitting each plane representing the crack), and then extracts the crack polygon boundary through the Alpha-Shape algorithm, and uses the vector cross multiplication method to calculate the polygon area to obtain the number of crack planes, crack area and crack shape in the current spatial area.

[0071] DBSCAN is a density-based clustering algorithm that forms clusters by gathering points in high-density areas and treating points in low-density areas as noise. It is suitable for detecting cluster structures in space, especially point cloud data with irregular shapes and noise. The specific steps are:

[0072] 1) Select a point as the core point. If the number of points in its neighborhood is greater than min_samples, a cluster is formed.

[0073] 2) Expand the points in the neighborhood until all reachable points are included in the cluster.

[0074] 3) If a point cannot reach the min_samples number, it is marked as a noise point.

[0075] RANSAC is an iterative method for fitting mathematical models and excluding outliers from data. Its goal is to find a model from a set of points containing a large number of outliers so that most points conform to the model. It is suitable for extracting data that conforms to a certain pattern (such as a plane, a straight line, etc.) from a large amount of noise. The specific steps are:

[0076] 1) Randomly sample a subset of the data and fit a model (such as a plane) to this subset.

[0077] 2) Calculate the residual of the model and mark the points whose residual is lower than the threshold as inliers.

[0078] 3) Repeat the above steps and select the model with the most inliers as the final result.

[0079] Alpha-Shape is a geometric algorithm for calculating the shape of a point set. It generates a shape containing a point set by adjusting a parameter alpha. It can be used to represent the convex hull or non-convex shape of a point set. It is used to describe the boundaries of irregular geometric shapes, especially for 3D point clouds and crack shape extraction. The specific steps are:

[0080] 1) Build a neighborhood based on the given alpha value and connect the points within the neighborhood.

[0081] 2) Connect these points to generate a triangulation to form the geometric shape of the point set.

[0082] 3) By adjusting the alpha value, the "compactness" of the shape can be controlled, that is, the generated shape can be closer to the boundary of the original point set from the convex hull.

[0083] In step 5, the rock burst hazard level of each grid is obtained based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each grid, including:

[0084] The fractal dimension, the number of crack planes, the crack area and the crack shape in the grid are input into a pre-trained machine learning model to obtain the rock burst hazard level of the grid;

[0085] During the development of rock burst, as the stress state of coal-rock mass changes, stress and energy accumulate inside it, the coal mass continues to break, and various microseismic indicators change. The location and energy of large-energy microseismic events are related to the surrounding rock fracture damage. Therefore, from the perspective of seismology, the surrounding rock failure mechanism is regarded as an important cause of microseismic events. The numerical results of the evolution of microseismic indicators in the region are used as a data set for machine learning algorithms to be applied to regional rock burst prediction. It is reliable and can be used as a reasonable representation of the degree of impact danger, thereby capturing impact precursor information.

[0086] Therefore, in order to solve this problem, starting from the surrounding rock damage mechanism, the fractal dimension theory is introduced, and the various parameters of the cracks are inverted based on the algorithm of large-scale data grid processing. Through the intelligent machine learning algorithm, the interpretability of microseismic parameters for rock burst warning is improved, and the rock burst hazard is accurately and effectively evaluated, so as to successfully warn. Specifically:

[0087] Rocks break after being damaged, and the strain energy dissipated by the rupture is released in the form of elastic waves. The microseismic monitoring system can monitor the spatial source coordinates of these elastic waves. The evolution of these rupture damages has fractal characteristics, so the fractal dimension can be used to estimate the degree of surrounding rock damage, thereby further characterizing the degree of rock burst incubation. At the same time, through these spatial source coordinates, the three-dimensional fracture network of microseismic events can be inverted through large-scale data gridding algorithms, and parallel and staggered complex fractures can be accurately identified to obtain key parameters such as the area and shape of the fractures. Combined with the fractal dimension, the impact hazard characteristics can be effectively evaluated through intelligent machine learning algorithms.

[0088] The machine learning model is used to receive the fractal dimension, the number of crack planes, the crack area and the crack shape in the grid, and output the rock burst hazard level of the grid after processing.

[0089] The present invention constructs a data set for evaluating the level of the current spatial danger zone based on the various parameters that have been obtained, divides the impact danger of the current spatial area into levels, and constructs an intelligent grading prediction model using the XGBoost algorithm optimized by the Bayesian algorithm. That is, the fractal dimension, number of crack planes, crack area and crack shape of the current window are used as feature sets, and the danger level of the current spatial area is used as the target set for model training. Finally, the impact rock pressure danger level of the current spatial area is output for early warning at any time.

[0090] In the present invention, the machine learning model adopts the Bayes-XGBoost early warning model, that is, the Bayesian optimized XGBoost model, such as Figure 3 As shown, the steps for building the XGBoost model based on Bayesian optimization are as follows:

[0091] 1) The fractal dimension, number of crack planes, crack area and crack shape in the previously obtained historical window are used as feature variables, and the hazard level of the current spatial area is used as the target variable to construct a data set. Specifically, the maximum energy occurring in each area is used as the impact hazard level. The specific level division is shown in Table 1, and finally a data set for training is obtained.

[0092] Table 1 Microseismic energy and impact hazard level

[0093]

[0094] 2) Divide the training data in 1) into training samples and test samples in a ratio of 8:2 to prepare for model training and prediction.

[0095] 3) Establish a Bayes-XGBoost rock burst prediction model. The training samples are used for training, and the multiple hyperparameters of XGBoost are comprehensively optimized by the Bayesian optimization method, so as to find a better model configuration in a shorter time, thereby improving the model performance, reducing overfitting, and enhancing the generalization ability of the model. The optimization range of each hyperparameter setting is shown in Table 2. After determining the optimal hyperparameter combination of the model, an intelligent identification model of rock burst danger area and danger level based on the spatiotemporal distribution characteristics of cumulative damage is obtained.

[0096] Table 2 XGBoost hyperparameters and their optimization range

[0097]

[0098]

[0099] 4) Iterate and output the optimal hyperparameter values ​​to set the model parameters, use the test set to test the warning level effect of the model, and output the model test results.

[0100] The following is a brief introduction to the concepts of the algorithms involved:

[0101] (1) Bayesian optimization is a search strategy based on probability distribution. It uses a prior function to represent the distribution assumption and uses the acquisition function to determine the sampling location. For the parameter selection of the model, the prior function uses a Gaussian process to reflect the distribution assumption of the optimization function. The acquisition function uses the boost probability function to select the next set of parameters from the posterior distribution and evaluate them. The process of Bayesian optimization of XGBoost parameters mainly consists of the following steps:

[0102] 1. Initialize parameters and Gaussian process model. Use the training and validation results to initialize the Gaussian process model, which will be used as a proxy model for Bayesian optimization to approximate the objective function.

[0103] 2. Prediction and model modification: Use the Gaussian process model to predict the parameter space and use the Bayesian rule to update the Gaussian process model, that is, modify the model to better reflect the relationship between parameters and performance.

[0104] 3. Parameter selection and evaluation: Use a collection function to select the next set of parameters based on the predictions of the current Gaussian process model, and use the selected set of parameters to train and validate the XGBoost model.

[0105] 4. Loop iteration and termination conditions. Repeat steps 2 and 3, iteratively select new parameters, train and verify the model, and update the Gaussian process model.

[0106] When the error caused by the newly added parameter combination meets the preset threshold or reaches the iteration limit, the optimization process is terminated. The parameters at this time should be the XGBoost model parameters that Bayesian optimization considers to be optimal in the current iteration.

[0107] (2) Extreme Gradient Boosting Tree (XGBoost) is an efficient machine learning algorithm for regression and classification problems. Its core principle is to gradually reduce the model error by increasing the tree model. By continuously adding new trees and letting the new trees learn the residuals of all previous trees, the model is gradually optimized during the training process. It can be effectively used for nonlinear problems in rock burst classification warning.

[0108] In step 5, the early warning is issued according to the danger level, including:

[0109] comparing the risk level with a set risk level value;

[0110] If the danger level is not greater than the set danger level value, no operation is performed;

[0111] If the danger level is greater than the set danger level value, an early warning is initiated.

[0112] Meanwhile, the present invention further comprises, after step 4 and before step 5:

[0113] The shapes of individual cracks in the mesh are visualized in three dimensions.

[0114] The present invention performs three-dimensional visualization of the fitted crack plane and Alpha-Shape boundary to display the geometric shape of the crack. It also combines the fractal dimensions in different spatial regions to comprehensively evaluate the danger level of rock burst in the current spatial region and issue early warnings at any time.

[0115] The present invention combines intelligent algorithms to perform real-time intelligent analysis of monitoring data, minimizes the delay of microseismic monitoring and analysis, and improves the timeliness and reliability of predicting rock burst risk. Through on-site microseismic monitoring, recording and transmission, the monitoring and recording results of the real-time database are sent to the computer, the collected signals are subjected to noise reduction processing, the errors of the original information are mastered or cleared, and then the required useful information is identified and extracted from the information based on the characteristics, and input into the program of the method, the damage and crack expansion of the tunnel surrounding rock are observed in real time, and early warning information is analyzed and given in time, which can effectively prevent the losses caused by rock burst disasters.

[0116] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A rock burst advance warning method based on the temporal and spatial distribution characteristics of cumulative damage, characterized in that: The following steps are involved: Step 1: Acquire all microseismic data of the area to be predicted, wherein the microseismic data consists of the three-dimensional coordinates and corresponding energy of each point in the prediction area; Step 2: Divide the area to be predicted into multiple grids, and obtain all microseismic data in each grid; Step 3: Based on all the microseismic data in each of the grids, calculate the fractal dimension of the tunnel surrounding rock damage in each of the grids; Step 4: based on all microseismic data in each of the grids, obtain the number of fracture planes, fracture area and fracture shape in each of the grids; Step 5: Based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each of the grids, the rock burst hazard level of each of the grids is obtained, and an early warning is issued according to the hazard level.

2. A method for early warning of rock burst based on the temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: The area to be predicted is a mining working face or a mining tunnel.

3. A method for early warning of rock burst based on temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: In step 2, the obtaining of all microseismic data in each of the grids includes: Determine the size of the window and the sliding step, and determine the number of the grids and the positions of the grids according to the size of the window; Allocating the microseismic data to each of the grids according to the three-dimensional coordinates in the microseismic data; The microseismic data in each of the grids is filtered.

4. A method for early warning of rock burst based on temporal and spatial distribution characteristics of cumulative damage according to claim 2, characterized in that: The filtering of the microseismic data in each of the grids includes: Determining the energy of each microseismic data in each of the grids, and removing the microseismic data with energy less than a set energy; Counting the number of microseismic data in each of the grids; Grids whose number of microseismic data is greater than a set number are screened out.

5. The method for early warning of rock burst based on the temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: In step 4, based on all the microseismic data in each of the grids, the number of fracture planes, fracture area and fracture shape in each of the grids are obtained, including: Clustering all microseismic data in the grid by using DBSCAN algorithm; Performing spatial plane fitting on the clustered microseismic data through RANSAC to obtain a plane with cracks; The number of cracks in the plane and the shape of each crack are extracted by using the Alpha-Shape algorithm; The area of ​​each crack is obtained by vector cross multiplication method.

6. The method for early warning of rock burst based on the temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: In step 5, the rock burst hazard level of each grid is obtained based on the fractal dimension, the number of crack planes, the crack area and the crack shape in each grid, including: The fractal dimension, the number of crack planes, the crack area and the crack shape in the grid are input into a pre-trained machine learning model to obtain the rock burst hazard level of the grid; The machine learning model is used to receive the fractal dimension, the number of crack planes, the crack area and the crack shape in the grid, and output the rock burst hazard level of the grid after processing.

7. The method for early warning of rock burst based on the temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: In step 5, the early warning is issued according to the danger level, including: comparing the risk level with a set risk level value; If the danger level is not greater than the set danger level value, no operation is performed; If the danger level is greater than the set danger level value, an early warning is initiated.

8. The method for early warning of rock burst based on the temporal and spatial distribution characteristics of cumulative damage according to claim 1, characterized in that: After step 4 and before step 5, the method further includes: The shapes of individual cracks in the mesh are visualized in three dimensions.

Images