Method for comprehensively evaluating stability of surrounding rock structure under deep engineering dynamic disaster

Abstract

The invention provides a method for comprehensively evaluating the stability of a surrounding rock structure under a deep engineering dynamic disaster, and the method comprises the steps: carrying out the geological environment analysis of a target region, and obtaining a rockburst influence factor; determining rockburst evaluation indexes according to the rockburst influence factors, and integrating the rockburst evaluation indexes into a data set; constructing a machine learning model and performing hyper-parameter optimization on the machine learning model to obtain an evaluation model; performing data mining on the data set through the evaluation model, and predicting a rockburst risk grade; and executing different prevention and control schemes according to the rockburst risk levels. According to the method, through multi-dimensional data acquisition, core index screening, machine learning prediction, risk grade division and targeted prevention and control scheme design, accurate assessment and efficient prevention and control of the rockburst risk are realized, and the safety of deep vertical shaft engineering is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering and geological disaster prevention and control, and in particular to a comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic disasters. Background Art

[0002] With the continuous advancement of engineering construction, rockbursts in my country's underground engineering projects have become a serious engineering geological hazard. Rockbursts are a unique dynamic hazard and ground pressure phenomenon that manifests in extremely high stress environments, and are particularly prone to occurring in deep tunnels. Under high stress conditions, when the rock mass is disturbed by mining, the elastic energy stored in the rock mass exceeds its strength limit, resulting in a sudden and violent release of the energy, causing severe damage to the rock mass, posing a serious threat to personnel, equipment, and project stability.

[0003] Currently, traditional rockburst assessment methods rely on a single mechanical parameter, ignoring the coupling effects of multiple factors, such as geological structure, hydrological conditions, and rock thermophysical properties. This results in insufficient risk assessment accuracy. Traditional criteria often rely on static thresholds, making it difficult to dynamically reflect nonlinear characteristics such as stress redistribution and surrounding rock damage evolution in deep engineering projects. Furthermore, they rely on empirical formulas or simple regression models, which cannot effectively handle the complex relationships between multidimensional data. Therefore, it is crucial to design a comprehensive assessment method for the stability of surrounding rock structures under dynamic hazards in deep engineering projects. Summary of the Invention

[0004] The purpose of this invention is to provide a comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic hazards. Through multi-dimensional data collection, core indicator screening, machine learning prediction, risk level classification and targeted prevention and control plan design, it can achieve accurate assessment and efficient prevention and control of rockburst risks and improve the safety of deep shaft projects.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic disasters includes the following steps:

[0007] Conduct geological environment analysis of the target area to identify factors affecting rock burst;

[0008] Determine rockburst evaluation indicators based on rockburst influencing factors and integrate the rockburst evaluation indicators into a data set;

[0009] Build a machine learning model and optimize its hyperparameters to obtain an evaluation model;

[0010] The evaluation model is used to mine the data set and predict the rockburst risk level; the rockburst risk levels include: no rockburst, slight rockburst, moderate rockburst, and severe rockburst;

[0011] Different prevention and control plans are implemented according to the rock burst risk level.

[0012] Optionally, rockburst influencing factors include: geological conditions, ground stress, hydrological conditions, formation temperature, rock mechanical properties, rock physical indicators and mining data;

[0013] Rock mechanical properties include: elastic modulus, Poisson's ratio, rock uniaxial tensile strength, rock uniaxial compressive strength, rock shear strength and internal friction angle;

[0014] Rock physical indicators include: density, bulk density, porosity, permeability, mineral composition and thermophysical properties;

[0015] Mining data include: rock strength coefficient, drilling parameters, explosive parameters and detonation parameters.

[0016] Optionally, the ground stress is measured by hydraulic fracturing or casing core stress relief method.

[0017] Optionally, the rockburst evaluation indicators include: stress criterion, rock brittleness index, maximum tangential stress of surrounding rock and elastic energy index.

[0018] Optionally, the stress criterion is the ratio of the maximum tangential stress of the surrounding rock to the uniaxial compressive strength of the rock. The calculation formula of the stress criterion is: R = σ θ / σ c ; Among them, R is the stress criterion, σ θ is the maximum tangential stress of surrounding rock, σ c is the uniaxial compressive strength of rock;

[0019] The rock brittleness index is the ratio of the rock uniaxial compressive strength to the rock uniaxial tensile strength. The calculation formula of the rock brittleness index is: B = σ c / σ t ; Among them, B is the rock brittleness index, σ t Uniaxial tensile strength of rock;

[0020] The calculation formula of elastic energy index is: Where, is the elastic energy index, is the elastic strain energy stored in the rock, is the plastic strain energy.

[0021] Optionally, determining the rock burst evaluation index based on rock burst influencing factors further includes: quantitatively evaluating the stability of the surrounding rock and the difficulty of disaster control through a difficulty coefficient and a hazard coefficient;

[0022] The calculation formula for the difficulty coefficient is: Among them, D f is the difficulty coefficient, H is the actual depth of the underground project, and Hcr is the critical depth of the deep project;

[0023] The formula for calculating the risk factor is: Among them, D c is the risk factor, σ1 is the maximum principal stress. m It is the strength of engineering rock mass.

[0024] Optionally, the machine learning model includes: an integrated model and a single model; the integrated model includes: a bagging model, a voting model and a boosting model; the single model includes: a support vector machine model, a decision tree model, a logistic regression model, a K-nearest neighbor model and a neural network model.

[0025] Optionally, the process of performing hyperparameter optimization on the machine learning model is specifically as follows: performing hyperparameter optimization on the machine learning model through the TPE algorithm, and performing 5-fold cross-validation on the optimized machine learning model through grid search, genetic algorithm and particle swarm optimization algorithm.

[0026] Optionally, the rockburst risk level classification standard is:

[0027] When 0≤R<0.2、0≤W et <2, 0≤B<10 and 0≤σ θ <24 hours, the rock burst risk level is no rock burst;

[0028] When 0.2≤R<0.3, 2≤W et <3.5, 10≤B<14, 24≤σ θ <60, the rockburst risk level is minor rockburst;

[0029] When 0.3≤R<0.5, 3.5≤W et <5, 14≤B<17, 60≤σ θ <126, the rockburst risk level is moderate rockburst;

[0030] When 0.5≤R<2, 5≤W et <20, 17≤B<60, 126≤σ θ ≤200, the rock burst risk level is severe rock burst.

[0031] Optionally, the prevention and control scheme includes: passive prevention and control and active prevention and control; passive prevention and control includes: anchor spraying support, steel support and flexible wire mesh; active prevention and control includes: advance anchor rods, stress relief holes and high-pressure water injection.

[0032] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the present invention provides a comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic disasters, the method comprising: conducting a geological environment analysis of the target area to obtain rockburst influencing factors; determining rockburst evaluation indicators based on the rockburst influencing factors, and integrating the rockburst evaluation indicators into a data set; constructing a machine learning model and performing hyperparameter optimization on the machine learning model to obtain an evaluation model; performing data mining on the data set through the evaluation model and predicting the rockburst risk level; and executing different prevention and control plans based on the rockburst risk level. This method achieves accurate assessment and efficient prevention and control of rockburst risks through multi-dimensional data collection, core indicator screening, machine learning prediction, risk level classification, and targeted prevention and control plan design, thereby improving the safety of deep vertical shaft projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0034] Figure 1 This is a flow chart of the comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters of the present invention. DETAILED DESCRIPTION

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0036] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0037] like Figure 1 As shown, the present invention provides a comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic disasters, comprising the following steps:

[0038] Step 100: Analyze the geological environment of the target area to obtain rock burst influencing factors;

[0039] Specifically, factors affecting rock burst include: geological conditions, ground stress, hydrological conditions, formation temperature, rock mechanical properties, rock physical indicators and mining data.

[0040] Geological conditions are composed of geological structures and stratigraphic structures. Geological structures are the various deformation and displacement traces left in rock layers and rock masses due to tectonic changes and are the product of crustal movement. Strata are layered rocks formed at a certain time in geological history, a layer or group of rocks with a specific position in the crust.

[0041] In-situ stress is the naturally occurring stress state within underground rock masses or formations in the absence of external disturbances. It includes gravitational stress and tectonic stress. Gravity stress manifests as vertical stress, while tectonic stress manifests as horizontal stress. In this embodiment, in-situ stress is measured using hydraulic fracturing or core stress relief methods.

[0042] Hydrological conditions refer to the general conditions related to the formation, distribution, and variation of groundwater. Formation temperature refers to the temperature of rocks at different depths underground. It increases with depth, and its temperature distribution is affected by many factors, including heat flow inside the earth, thermal conductivity of rocks, and groundwater activity. Rock mechanical properties include: elastic modulus E, Poisson's ratio v, uniaxial tensile strength σ of rocks t , rock uniaxial compressive strength σ c , rock mass shear strength τ and internal friction angle The calculation formula is as follows:

[0043]

[0044] Where σ is stress, ε is strain, and ε x is the transverse strain, ε y is the axial strain, P t is the maximum axial tensile load, A is the cross-sectional area of the specimen, P is the maximum axial pressure, c is the bond force on the structural surface, σ n is the normal stress acting on the structural surface. Rock physical indicators include: density ρ, bulk density γ, porosity n, permeability Q x , mineral composition and thermal properties, the calculation formula is as follows:

[0045]

[0046] Where G is the mass of the rock being tested, V is the volume of the rock being tested, W is the weight of the rock being tested, and V P is the total volume of rock pores. Thermophysical properties include thermal conductivity, specific heat capacity and thermal diffusivity.

[0047] Mining data includes rock strength coefficient, drilling parameters, explosive parameters, and detonation parameters. Drilling parameters include hole spacing, row spacing, and blasthole depth. Explosive parameters include charge structure, charge quantity, and unit explosive consumption.

[0048] Step 200: determining rockburst evaluation indicators based on rockburst influencing factors, and integrating the rockburst evaluation indicators into a data set;

[0049] Specifically, rockbursts typically occur in rock masses with high stress concentrations. From a topographical perspective, rockbursts typically occur in mountainous areas or deep underground projects, or in rock masses with high tectonic stresses. From a geological perspective, rockbursts tend to occur near hard structural surfaces. From a structural layout perspective, the more irregular the excavation section, the greater the likelihood of a rockburst. All of these factors can be reflected by the maximum tangential stress in the surrounding rock. Rockburst sections primarily exhibit tensile failure, accompanied by shear failure, while the uniaxial tensile strength of rock represents both its tensile and shear mechanical properties. Rockbursts primarily occur in structurally intact hard rock, while the uniaxial compressive strength of rock measures the rock's hardness. The elastic energy index reflects the rock's ability to store and release energy. Under the same stress conditions, the larger the elastic energy index, the better the rock's ability to store and release energy.

[0050] Furthermore, rockburst evaluation indicators include: stress criterion, rock brittleness index, maximum tangential stress of surrounding rock, and elastic energy index. The stress criterion is the ratio of the maximum tangential stress of surrounding rock to the uniaxial compressive strength of rock. It takes into account the mechanical properties of rock and the stress redistribution after cavern excavation. It is used to determine the possibility of rockburst. The calculation formula is:

[0051] R=σ θ / σ c ;

[0052] Among them, R is the stress criterion, σ θ is the maximum tangential stress of surrounding rock, σ c is the uniaxial compressive strength of rock;

[0053] The rock brittleness index is the ratio of the rock's uniaxial compressive strength to its uniaxial tensile strength. The higher the rock brittleness index, the more likely the rock will fail brittlely when subjected to stress, thereby increasing the risk of rock burst. The calculation formula is:

[0054] B=σ c / σ t ;

[0055] Among them, B is the rock brittleness index, and σ t Uniaxial tensile strength of rock;

[0056] The elastic energy index is an indicator of the elastic energy stored and released by rock during stress. It is the ratio of the elastic strain energy stored in the rock during loading to the plastic strain energy. The larger the elastic energy index, the greater the energy released during rock failure and the greater the risk of rock burst. The calculation formula is:

[0057]

[0058] Where, is the elastic energy index, is the elastic strain energy stored in the rock, is the plastic strain energy.

[0059] Specifically, determining rock burst evaluation indicators based on rock burst influencing factors also includes: quantitatively evaluating the stability of surrounding rock and the difficulty of disaster control through difficulty coefficient and hazard coefficient;

[0060] Difficulty level D f It refers to the ratio of the depth of the underground project to its critical depth. The calculation formula is:

[0061]

[0062] Among them, H is the actual depth of the underground project, Hcr is the critical depth of the deep project; when D f When D is less than 1, it indicates that the engineering rock mass is in a linear working state and its stability can be controlled by conventional methods. f When ≥1, it indicates that the soft rock engineering rock group is in a nonlinear large deformation working state, and the hard rock engineering rock group will experience nonlinear dynamic phenomena such as rock burst.

[0063] Risk Factor D c It refers to the ratio of the maximum principal stress borne by the surrounding rock of the shaft to the strength of its engineering rock mass. The calculation formula is:

[0064]

[0065] Among them, σ1 is the maximum principal stress. σ m is the engineering rock mass strength. c When D is less than 1, it indicates that the engineering rock mass is in a linear working state and will not produce nonlinear dynamic disaster phenomena; when D c When D > 1, it indicates that most of the engineering rock mass is in a nonlinear working state, which will produce nonlinear dynamic disaster phenomena, and D c The larger it is, the more serious the dynamic disasters of deep rock mechanics nonlinear problems faced during underground engineering construction are.

[0066] Furthermore, the degree of surrounding rock hazard is quantitatively assessed by the failure proximity index (FAI). The calculation formula for failure proximity is:

[0067]

[0068] Where w is the stress state parameter of the surrounding rock, and FD is the damage factor. When FAI < 0.8, the rock is in the elastic development stage, the stress-strain curve develops linearly, and the surrounding rock mainly stores energy, so no corresponding support is required to maintain the surrounding rock stability. When 0.8 ≤ FAI < 1, the rock is in the damage development stage, and the stress exceeds the rock damage stress (the volume strain turning point). Under the action of higher stress, even if the stress does not increase, it may cause the development of cracks in the surrounding rock. Appropriate support measures are needed to improve the ultimate strength of the rock. When 1 ≤ FAI < 2, the stress exceeds the ultimate strength of the rock, and obvious macroscopic cracks are generated in the rock. Support measures should be taken in a timely manner to ensure the stability of the surrounding rock. When FAI ≥ 2, the rock is severely broken, and the larger the value, the more severe the rock damage.

[0069] Step 300: Build a machine learning model and optimize hyperparameters of the machine learning model to obtain an evaluation model;

[0070] Specifically, principal component analysis (PCA) was used to reduce the number of features of the dataset to 5, and the feature values were ensured to be orthogonal to each other. Then, 70% of the dataset after the number of features was reduced was used as the training set, and the remaining 30% was used as the test set.

[0071] The machine learning models constructed in this embodiment include: integrated models and single models; the integrated models include: Bagging model, Voting model and Boosting model; the single models include: support vector machine model, decision tree model, logistic regression model, K-nearest neighbor model and neural network model.

[0072] The Bagging model is a comprehensive prediction model formed by integrating a large number of basic single learners of the same type. Each basic single learner is trained based on a random subset of the training set, and the learners are independent of each other. The Bagging model averages the prediction results of all basic single learners as the final output.

[0073] Compared to the Bagging model, the Voting model has far fewer basic single learners and combines the prediction results of each basic single learner through a majority voting strategy or an averaging strategy as the final output. This example selects five machine learning classifiers: the support vector machine model (SVM), the decision tree model (DT), the logistic regression model (LR), the K-nearest neighbor model (KNN), and the neural network model (NN) as the basic single learners of the Voting model.

[0074] SVM uses only a portion of the training set as support vectors to determine the hyperplane, rather than relying on all data, thus achieving both accuracy and computational efficiency. Its objective function and constraints are both convex, transforming the classification problem into a convex optimization problem to find the global optimal solution. Furthermore, the built-in kernel function enables the model to handle nonlinear problems.

[0075] DT processes datasets in the form of a tree, consisting of three types of nodes: root nodes, internal nodes, and leaf nodes. The root node is at the top of the tree. Internal nodes are segmented according to the eigenvalues, and data is split or classified according to the eigenvalues as it passes through the internal nodes, ensuring that the entropy of the entire dataset decreases after each split. The tree grows in this way until it reaches a specified depth or the number of samples at each internal point reaches a minimum. Leaf nodes are the bottom nodes of the tree. After multiple splits, the data passes through the leaf nodes to output the final prediction results. This embodiment uses the Gini index to measure the entropy of the dataset after passing through the internal nodes.

[0076] LR uses a logistic expression to describe the probability that the training set belongs to the "1" class. At the same time, through its binary classification solution, the identification of each category can be regarded as a separate binary classification problem, and multiple classifiers can be constructed, thereby improving computational efficiency.

[0077] KNN estimates the category by referencing the categories of several nearest neighbors and combining it with a majority voting strategy. The number of neighbors is a hyperparameter of the classifier, and the distance between two sample points is represented by the Euclidean distance, ensuring the efficiency of the training and prediction process.

[0078] NN establishes a mapping relationship between the input layer (features) and the output layer (target) by introducing one or more hidden layers. The transfer coefficients between the input layer and the hidden layer and between the hidden layer and the output layer are determined by the gradient descent method in the back propagation process.

[0079] The Boosting model is also a comprehensive prediction model formed by integrating a large number of basic single learners of the same type. However, the training of its i+1th basic single learner depends on the comprehensive prediction results of the previous i basic single learners.

[0080] The single model of this embodiment is consistent with the basic single learner of the Voting model.

[0081] The TPE algorithm fits the relationship between the machine learning model's variables X(x1, x2, ...) and the target y, and identifies extreme points. First, TPE assumes a prior distribution P(y) for y and randomly samples several data points. Second, it uses kernel density estimation to estimate a surrogate function for the target function based on P(y) and the known data points, and simultaneously derives the posterior distribution P(y|X) of y. Expected lift is then used to derive the acquisition function of y from the surrogate function and P(y|X). The points where the acquisition function reaches its maximum are added to the sampled data points to improve the surrogate function's simulation capabilities in the next iteration. Finally, a new iteration is started based on the expanded data point set until the specified number of iterations is reached.

[0082] Furthermore, grid search (GS), genetic algorithm (GA), particle swarm optimization algorithm (PSO) and TPE algorithm are used together with 5-fold cross validation, that is, the training set is divided into 5 equal parts, and 5 tests are performed under a certain sampled parameter combination. In each test, 4 parts are used for training, and the remaining 1 part is used as a validation set for verification until each data set has been verified once. Finally, the parameter combination with the highest average prediction accuracy in the training set is selected as the hyperparameter optimization result.

[0083] Step 400: performing data mining on the data set using the evaluation model and predicting the rockburst risk level; the rockburst risk levels include: no rockburst, slight rockburst, moderate rockburst, and severe rockburst;

[0084] Specifically, the evaluation model is trained using the training set. The trained model is then used to estimate the sample categories of the test set. The Shapley value method is then used to calculate the contribution of each characteristic value to the rockburst prediction accuracy. This example compares the prediction accuracy of the trained model on an incomplete dataset with characteristic factors removed with the prediction accuracy of the complete dataset. The difference between the two is considered the marginal contribution of the removed characteristic factor. A larger difference indicates a greater impact of the characteristic factor on prediction accuracy.

[0085] Specifically, the rockburst risk level classification standards are as follows:

[0086] When 0≤R<0.2、0≤W et <2, 0≤B<10 and 0≤σ θ <24:00, the rock burst risk level is no rock burst; it means that the surrounding rock has not been torn, peeled off or collapsed, the tunnel remains relatively intact, and no protective and monitoring measures are required;

[0087] When 0.2≤R<0.3, 2≤W et <3.5, 10≤B<14, 24≤σ θ<60, the rockburst risk level is minor rockburst; the surrounding rock begins to loosen and peel off, with occasional slight bursting sounds, and appropriate protective measures need to be taken;

[0088] When 0.3≤R<0.5, 3.5≤W et <5, 14≤B<17, 60≤σ θ <126, moderate rockburst risk level; manifested by massive spalling of surrounding rock, often accompanied by sharp ejection sounds, occasional rock ejection, severe spalling and bottom heave, which can easily cause personal injury and equipment damage, requiring on-site real-time monitoring measures;

[0089] When 0.5≤R<2, 5≤W et <20, 17≤B<60, 126≤σ θ ≤200, the rock burst risk level is severe rock burst; it is manifested by large-scale peeling of the surrounding rock, continuous sharp ejection sounds, accompanied by rock ejection, rapid deformation of the surrounding rock, and the appearance of a large number of blast pits, which seriously threaten the safety of personnel and equipment. High-level protection measures must be taken in a timely manner.

[0090] Step 500: Execute different prevention and control plans according to the rockburst risk level.

[0091] Specifically, the prevention and control plans include: passive prevention and control and active prevention and control; passive prevention and control include: anchor spraying support, steel support and flexible wire mesh; active prevention and control include: advance anchor rods, stress relief holes and high-pressure water injection.

[0092] More specifically, when the rock burst risk level is no rock burst, anchor spraying support is adopted; when the rock burst risk level is mild rock burst, advance anchor rods, stress relief holes or flexible wire mesh are adopted; when the rock burst risk level is moderate rock burst, advance anchor rods, anchor spraying support or steel support are adopted; when the rock burst risk level is severe rock burst, a multi-layer support system is adopted, with the innermost layer being high-strength sprayed concrete support, the middle layer being anchor spraying support, and the outermost layer being steel support.

[0093] The beneficial effects of the present invention are as follows:

[0094] 1) By integrating multi-dimensional data such as geological structure, geostress, hydrological conditions, rock mechanical properties, and mining disturbance, combined with machine learning models, the limitations of traditional static criteria are overcome and dynamic prediction of rockburst risk is achieved;

[0095] 2) A quantitative rockburst classification standard was established based on key parameters such as the elastic energy index, maximum tangential stress of the surrounding rock, and rock brittleness index. Targeted prevention and control solutions were provided for different risk levels, achieving an integrated "precise classification and active prevention and control" approach, reducing project costs and improving safety.

[0096] 3) By optimizing machine learning hyperparameters through the TPE algorithm and combining it with 5-fold cross-validation, we solved the problems of low efficiency and overfitting in traditional grid search and improved the generalization ability of the model.

[0097] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0098] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A comprehensive evaluation method for the stability of surrounding rock structures under deep engineering dynamic disasters, characterized by: The steps include: Conduct geological environment analysis of the target area to identify factors affecting rock burst; Determining rockburst evaluation indicators based on the rockburst influencing factors, and integrating the rockburst evaluation indicators into a data set; Constructing a machine learning model and optimizing hyperparameters of the machine learning model to obtain an evaluation model; performing data mining on the data set through the evaluation model and predicting the rockburst risk level; The rockburst risk levels include: no rockburst, slight rockburst, moderate rockburst and severe rockburst; Different prevention and control plans are implemented according to the rockburst risk level.

2. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: The rockburst influencing factors include: geological conditions, ground stress, hydrological conditions, formation temperature, rock mechanical properties, rock physical indicators and mining data; The rock mechanical properties include: elastic modulus, Poisson's ratio, rock uniaxial tensile strength, rock uniaxial compressive strength, rock shear strength and internal friction angle; The rock physical indicators include: density, bulk density, porosity, permeability, mineral composition and thermophysical properties; The mining data includes: rock strength coefficient, drilling parameters, explosive parameters and detonation parameters.

3. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 2 is characterized in that: The ground stress is measured by a hydraulic fracturing method or a casing core stress relief method.

4. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: The rockburst evaluation indexes include: stress criterion, rock brittleness index, maximum tangential stress of surrounding rock and elastic energy index.

5. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 4 is characterized in that: The stress criterion is the ratio of the maximum tangential stress of the surrounding rock to the uniaxial compressive strength of the rock. The calculation formula of the stress criterion is: R = σ θ / σ c ; Among them, R is the stress criterion, σ θ is the maximum tangential stress of the surrounding rock, σ c is the uniaxial compressive strength of rock; The rock brittleness index is the ratio of the rock uniaxial compressive strength to the rock uniaxial tensile strength. The calculation formula of the rock brittleness index is: B = σ c / σ t ; Among them, B is the rock brittleness index, σ t Uniaxial tensile strength of rock; The calculation formula of the elastic energy index is: Where, is the elastic energy index, is the elastic strain energy stored in the rock, is the plastic strain energy.

6. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: Determining rock burst evaluation indicators based on the rock burst influencing factors also includes: quantitatively evaluating the stability of surrounding rock and the difficulty of disaster control through difficulty coefficient and hazard coefficient; The calculation formula of the difficulty coefficient is: Among them, D f is the difficulty coefficient, H is the actual depth of the underground project, and Hcr is the critical depth of the deep project; The calculation formula of the risk factor is: Among them, D c is the risk factor, σ1 is the maximum principal stress. m It is the strength of engineering rock mass.

7. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: The machine learning models include: integrated models and single models; the integrated models include: Bagging models, Voting models and Boosting models; the single models include: support vector machine models, decision tree models, logistic regression models, K-nearest neighbor models and neural network models.

8. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: The process of performing hyperparameter optimization on the machine learning model is specifically as follows: performing hyperparameter optimization on the machine learning model through the TPE algorithm, and performing 5-fold cross-validation on the optimized machine learning model through grid search, genetic algorithm and particle swarm optimization algorithm.

9. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 5 is characterized in that: The rockburst risk level classification standards are as follows: When 0≤R<0.2、0≤W et <2, 0≤B<10 and 0≤σ θ <24:00, the rock burst risk level is no rock burst; When 0.2≤R<0.3, 2≤W et <3.5, 10≤B<14, 24≤σ θ <60, the rockburst risk level is the minor rockburst; When 0.3≤R<0.5, 3.5≤W et <5, 14≤B<17, 60≤σ θ <126, the rockburst risk level is the medium rockburst; When 0.5≤R<2, 5≤W et <20, 17≤B<60, 126≤σ θ ≤200, the rock burst risk level is the severe rock burst.

10. The comprehensive evaluation method for surrounding rock structure stability under deep engineering dynamic disasters according to claim 1 is characterized in that: The prevention and control plan includes: passive prevention and control and active prevention and control; the passive prevention and control includes: anchor spraying support, steel support and flexible wire mesh; the active prevention and control includes: advance anchor rods, stress relief holes and high-pressure water injection.

Images