A stress-energy-structure plane integrated deep surrounding rock stability evaluation method

Abstract

The application provides a stress-energy-structure surface fusion deep surrounding rock stability evaluation method, relates to the technical field of deep underground engineering surrounding rock stability evaluation, and comprises the following steps: S1. Collecting geological and mechanical data of an engineering area, and constructing a stress and energy model of the engineering area; according to the stress and energy model, extracting characteristic parameters of an evaluation unit; S2. According to the characteristic parameters of the evaluation unit and a failure threshold value, calculating an overrun coefficient under the current surrounding rock state, and forming a failure criterion; S3. Inputting the overrun coefficient into a multi-source intelligent fusion model of a stacked self-encoder, and outputting a surrounding rock stability index S; S4. According to the judgment result of the surrounding rock stability index S and the failure criterion, performing a stability classification evaluation on the deep surrounding rock. The application breaks through the limitation of a single stress or deformation criterion, realizes quantitative evaluation of surrounding rock stability, and is suitable for deep high-energy geological environments.

Description

Technical Field

[0001] This invention relates to the field of deep underground engineering surrounding rock stability evaluation technology, and more specifically, to a deep surrounding rock stability evaluation method that integrates stress, energy, and structural surfaces. Background Technology

[0002] Deep underground engineering projects (typically referring to underground caverns and tunnels with a vertical burial depth exceeding 600m) are situated in high-energy geological environments characterized by "three highs and one disturbance" (high ground stress, high osmotic pressure, high ground temperature, and strong excavation disturbance). Construction in such environments faces severe geological hazard risks, including rock bursts, large deformations and fractures of the surrounding rock, and high-temperature thermal hazards. These hazards directly threaten construction safety and the long-term operational stability of the cavern. Therefore, surrounding rock stability assessment, as a core issue throughout the entire lifecycle of deep engineering projects, is the fundamental basis for determining engineering design, construction plans, and economic safety.

[0003] Currently, the main methods for evaluating the stability of surrounding rock in underground engineering include theoretical analysis, numerical simulation, engineering analogy, monitoring back analysis, and intelligent comprehensive evaluation. However, as deep engineering projects become deeper and more complex, the aforementioned existing technologies are gradually revealing significant limitations and shortcomings in addressing the instability of surrounding rock in deep, high-energy environments.

[0004] Theoretical analytical methods (such as Fenner's formula) are based on elastoplastic mechanics and strength criteria such as Mohr-Coulomb and Hawke-Brown, deriving stress and plastic zone distribution through simplified models. Although the physical meaning is clear, their assumptions are too idealized and cannot reflect the dynamic internal process of surrounding rock from cracking and damage to instability and failure in deep and complex geological environments.

[0005] Numerical simulation methods (finite element method, discrete element method, etc.) can visualize stress fields, displacement fields, and plastic zones, and are currently the most widely used quantitative methods. However, their core theoretical foundation remains limited to classical elastoplastic mechanics. Their evaluation criteria still primarily rely on a single stress or strain index (i.e., determining whether the stress exceeds the rock's limit). In shallow engineering, this evaluation based on apparent mechanical behavior is still applicable; however, in deep, high-energy environments, rocks exhibit significant nonlinearity, rheological properties, and energy accumulation and dissipation characteristics, rendering traditional theoretical systems no longer entirely applicable. Relying solely on a single stress or deformation index cannot accurately capture the essence of deep surrounding rock disasters, easily leading to inaccurate predictions and even missed reports of major risks.

[0006] The engineering analogy method relies on hierarchical systems such as BQ, RMR, and Q systems to conduct qualitative or semi-quantitative evaluations by comparing with the experience of existing projects. This method lacks rigorous quantitative standards, and the evaluation results are highly dependent on the subjective experience of engineers. More importantly, there are relatively few cases of deep underground engineering, geological conditions vary greatly, and there is a lack of comparable objects for reference, making it difficult to guarantee the accuracy of the analogy and failing to meet the safety evaluation needs of special deep engineering projects.

[0007] The monitoring-based inverse analysis method collects data such as displacement and stress by deploying sensors and combines this with inverse algorithms to dynamically evaluate the stability of the surrounding rock. Although it is based on measured data, existing monitoring methods often rely on a single indicator (such as only looking at deformation or only looking at stress values). In actual engineering projects, phenomena such as "large deformation but no failure" or "significant cracking and spalling but not obvious deformation" often occur, and a single indicator cannot truly reflect the safety state of the surrounding rock. In addition, deep, large underground cavern groups have complex structures, a limited number of monitoring points, and insufficient data representativeness, making it easy for monitoring to go unreported or false alarms, and difficult to comprehensively cover the precursory information of surrounding rock instability.

[0008] While the intelligent comprehensive evaluation method integrates machine learning and other technologies to process multi-source data, its effectiveness is highly dependent on high-quality, large-sample training datasets. Currently, the number of deep-sea engineering projects is limited, and complete unstable sample data is scarce, resulting in poor generalization ability of the model. Furthermore, a unified engineering application standard has not yet been established, and it can currently only be used as an auxiliary tool.

[0009] In summary, existing methods for evaluating the stability of surrounding rock generally suffer from a lack of simplistic criteria. The instability and failure of surrounding rock is essentially a dynamic process of energy accumulation, release, and evolution; both deformation and cracking are external manifestations of energy release. Especially in deep, high-energy environments, energy transfer and release dominate the mechanical response. Most existing methods are based on external, apparent indicators such as stress or deformation, neglecting the intrinsic physical processes of crack propagation and energy release during the instability and failure of surrounding rock. This results in evaluations that only characterize the apparent mechanical behavior of the surrounding rock, failing to address the core mechanisms underlying disaster development. Consequently, scientific and accurate assessments are difficult to achieve, leading to situations where deformation does not lead to failure, or failure occurs but deformation is minimal.

[0010] Therefore, there is a need for a new method for evaluating the stability of deep surrounding rock that can break through the limitations of single stress or deformation criteria, integrate multiple factors such as energy, stress and structural surfaces, and conform to the essential law of "gestation-evolution-development" of disasters in deep surrounding rock. Summary of the Invention

[0011] The purpose of this invention is to provide a method for evaluating the stability of deep surrounding rock by integrating stress, energy, and structural planes. Based on the nature of stress concentration, energy release from crack propagation, and increased number of cracks during the incubation process of surrounding rock failure, this invention establishes a criterion for surrounding rock stability that runs through the entire process of crack initiation, energy release, and crack propagation, thereby making a scientific and reasonable evaluation of whether the surrounding rock is unstable.

[0012] To achieve the above objectives, this invention provides a method for evaluating the stability of deep surrounding rock by integrating stress, energy, and structural planes. The technical solution of this invention is implemented as follows:

[0013] A method for evaluating the stability of deep surrounding rock by integrating stress, energy, and structural planes includes the following steps:

[0014] S1. Collect geological and mechanical data of the engineering area and construct a stress and energy model for the engineering area; the geological and mechanical data includes in-situ stress, rock mass mechanical parameters, and structural surface geometric parameters; based on the stress and energy model, extract characteristic parameters of the evaluation unit, including: Type I stress intensity factor K. I The total energy released by the propagation of new cracks is G. I The number of newly formed macroscopic cracks per unit volume, N;

[0015] S2. Based on the characteristic parameters and instability threshold value of the evaluation unit, calculate the exceedance coefficients under the current surrounding rock condition and form an instability criterion; the exceedance coefficients include: stress exceedance coefficient α, energy exceedance coefficient β, and structural plane exceedance coefficient γ; the instability threshold value includes fracture toughness K. ICM Energy threshold value G IC Threshold value N for the number of new cracks C ;

[0016] S3. Input the overlimit coefficient into the fusion model of the stacked autoencoder and output the surrounding rock stability index S;

[0017] S4. Based on the judgment results of the surrounding rock stability index S and the instability criterion, the stability of the deep surrounding rock is evaluated and classified.

[0018] In step S2, the fracture toughness K ICM The calculation method is as follows: the plane strain fracture toughness K of intact rock is determined by indoor rock mechanics test. IC Combined with the rock mass integrity coefficient K v By applying size effect and crack correction, the fracture toughness K is obtained. ICM .

[0019] In step S2, the energy threshold value G IC The calculation method is as follows: first calculate the critical energy release rate per unit area. Where ν is Poisson's ratio and E is the elastic modulus; then calculate the energy threshold value. Where A is the total critical area of ​​crack propagation before instability and penetration within the evaluation unit.

[0020] In step S2, the threshold value N for the number of newly formed cracks c The calculation formula is: Where N0 is the density of internal macroscopic cracks in the rock when it is in a critical failure state, as determined by indoor rock tests, and B is the reduction factor.

[0021] In step S3, the formula for calculating the stress exceedance factor α is: .

[0022] In step S3, the formula for calculating the energy exceedance coefficient β is: .

[0023] In step S3, the formula for calculating the structural surface exceedance coefficient γ is as follows: .

[0024] In step S3, the construction process of the fusion model of the stacked autoencoder includes: using unsupervised layer-by-layer pre-training to extract deep implicit information of multi-source features; the network structure includes an input layer, two sparse autoencoder hidden layers, and an output layer; the input layer sets neurons to correspond to stress over-limit coefficient, energy over-limit coefficient, and structural surface over-limit coefficient respectively; in the input layer initialization stage, the feature weights are set: the energy feature weight is the highest, the structural surface feature weight is the second highest, and the stress feature weight is the lowest; the output layer uses the Sigmoid activation function to map the output to the [0,1] interval to obtain the surrounding rock stability index S.

[0025] In step S4, the instability criterion is: when When the stress-driven criterion is satisfied, where ;when At that time, the energy release criterion is satisfied, where ;when When the structural surface expansion criterion is satisfied, where .

[0026] In step S4, the surrounding rock stability index ,by The threshold for distinguishing between stable and unstable conditions; the stability classification evaluation includes:

[0027] when When the surrounding rock is in a stable or basically stable state, it is determined that the surrounding rock is in a stable state.

[0028] when If the value approaches 1, the surrounding rock is determined to be in a state of complete failure.

[0029] For other values, the judgment is made in conjunction with the satisfaction of the aforementioned instability criterion.

[0030] Compared with existing technologies, the stress-energy-structural plane fusion method for evaluating the stability of deep surrounding rock described in this invention has the following advantages:

[0031] 1. Breaking through the traditional single-index paradigm of stress / deformation, this approach deeply integrates the essence of fracture mechanics with energy principles, constructing a three-stage coupled full-chain physical criterion system of "stress-driven - energy release - structural surface evolution," truly achieving a leap from "phenomenon description" to "mechanism analysis." It establishes a hierarchical and logically progressive triple criterion: the stress criterion corresponds to crack initiation; the energy criterion corresponds to crack propagation; and the structural surface criterion corresponds to macroscopic instability precursors. These three criteria mutually corroborate each other and are indispensable, effectively avoiding misjudgments such as "deformation without destruction" or "destruction without deformation" caused by a single index.

[0032] 2. A multi-source heterogeneous data fusion model is constructed using a stacked autoencoder (SAE), which supports three types of nonlinear and small-sample inputs: stress exceedance coefficient, energy exceedance coefficient, and structural surface exceedance coefficient. Through a training strategy of prior weight initialization, unsupervised pre-training, and supervised fine-tuning, the model's generalization ability and real-time computation efficiency are significantly improved while ensuring physical interpretability.

[0033] 3. All threshold values ​​are strictly derived from the basic formulas of fracture mechanics, rather than empirically determined: the stress threshold is corrected by inversion from indoor Brazilian splitting / triaxial tests; the energy threshold is calculated based on the quantitative relationship between energy release rate and fracture toughness under plane strain conditions; and the threshold value for the number of new cracks is obtained by correcting the baseline value from indoor tests. All parameters can be measured or inverted, combining theoretical rigor with engineering operability.

[0034] 4. A five-level quantitative grading standard with the surrounding rock stability index S as the core is proposed. To establish a stable / unstable boundary threshold, and to dynamically determine the excess combination of three types of criteria, a continuous, interpretable, and traceable expression of risk level is achieved, supporting dynamic design and intelligent support decision-making.

[0035] 5. Applicable to deep, high-energy geological environments. Traditional methods are based on the theory and practical experience of shallow underground engineering, which are difficult to apply to deep, high-energy geological environments. They cannot reflect the energy-dominated, stress-driven, and multi-factor-influenced characteristics of deep surrounding rock disasters. This method uses energy release and crack propagation as the core criteria to accurately match the internal mechanism of deep surrounding rock instability, avoiding the shortcomings of one-sided and distorted evaluation of the apparent behavior of surrounding rock, and filling the technical gap in the multi-factor integrated stability evaluation of deep surrounding rock. Attached Figure Description

[0036] Figure 1This is a schematic diagram of the mechanical mechanism of crack propagation in the surrounding rock of deep tunnels as described in this invention.

[0037] Explanation of reference numerals in the attached figures:

[0038] 1. Surrounding rock; 2. Tunnel; 3. Unit; 4. Crack or structural surface; 5. Newly formed crack. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the described embodiments are only some, not all, of the embodiments of this invention. The specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0040] This invention provides a method for evaluating the stability of deep surrounding rock by integrating stress, energy, and structural planes, the specific steps of which include:

[0041] S1. Multi-source data acquisition and model building:

[0042] Based on actual deep engineering projects, geological and mechanical data of the engineering area were collected, including in-situ stress, rock mass mechanical parameters, and structural surface geometric parameters. Indoor uniaxial and triaxial compression tests were conducted on the rocks to obtain rock mechanical parameters.

[0043] A stress field and energy field distribution model for the engineering area is constructed using methods such as numerical simulation and inversion analysis.

[0044] From the established model or field monitoring system, extract the characteristic parameters of the evaluation unit. The characteristic parameters include: Type I stress intensity factor K. I The total energy released by the propagation of new cracks is G. I The number of newly formed macroscopic cracks per unit volume, N.

[0045] S2. Based on the characteristic parameters and instability threshold of the evaluation unit, calculate the exceedance coefficient under the current surrounding rock condition and form the instability criterion:

[0046] S21. Stress, energy, and structural threshold values:

[0047] Based on relevant theories such as fracture mechanics and elastoplastic mechanics, threshold values ​​for stress, energy, and structural surface criteria are obtained through rock experiments and tests.

[0048] Stress threshold value: The fracture toughness K of intact rock is determined by indoor rock mechanics tests such as the Brazilian splitting test and the three-point bending test. IC Combined with the rock mass integrity coefficient K v By applying size effect and fracture correction, the fracture toughness of the rock mass is obtained. Using KICM The stress threshold value σ of the rock mass was calculated. c .

[0049] Energy threshold value: based on energy release rate and fracture toughness K under plane strain conditions. ICM The quantitative relationship is used to calculate the critical energy release rate per unit area. Where ν is Poisson's ratio and E is the elastic modulus; the total energy threshold value is calculated by combining the critical propagation area A before crack instability and penetration within the evaluation unit obtained from numerical simulation or field statistics. .

[0050] Structural threshold value: This is determined through laboratory rock tests, including but not limited to triaxial compression and Brazilian splitting, to measure the internal crack density N0 when the rock is in a critical failure state. CT scans can be used for measurement, and acoustic emission testing can be used for calibration. In addition to laboratory tests, in-situ rock tests can also be conducted, combined with microseismic monitoring, and adjustments can be made based on rock mass integrity. A reduction factor B is introduced to correct the baseline value from the laboratory tests, thus determining the threshold value for the number of newly formed cracks. .

[0051] S22. Calculate the excess coefficient:

[0052] Define the excess coefficients for the three types of features respectively:

[0053] The formula for calculating the stress exceedance factor α is: ;

[0054] The formula for calculating the energy exceedance factor β is: ;

[0055] The formula for calculating the surface exceedance factor γ is: .

[0056] Type I stress intensity factor K I The total energy released by the propagation of new cracks is G. I The number of newly formed macroscopic cracks per unit volume, N, is obtained from the feature extraction in step S1.

[0057] S23. Three Instability Criteria:

[0058] In the evaluation of surrounding rock stability, to achieve accurate perception of the disaster incubation process, it is necessary to establish criteria based on key stages of the process. Therefore, criteria are established for the three stages of crack initiation, propagation, and penetration, based on three factors: stress, energy, and structural planes. The underlying principles are as follows:

[0059] S231. Stress Criterion:

[0060] Stress criteria are used to determine whether a crack has initiated. The essence of crack initiation is that the stress field intensity at the crack tip reaches the material's resistance limit, and the crack enters an unstable propagation state from a static state (a precursor to rock failure).

[0061] The stress threshold for crack initiation (critical fracture stress) is: ;

[0062] Where Y is the geometric shape factor, which is dimensionless and reflects the influence of crack morphology, component geometry, and loading method on stress intensity factor; a is the characteristic size of crack, which is half the length of the crack for a center-penetrating crack, and the crack depth for a single-sided crack or surface crack.

[0063] Considering the weakening effect of primary structural planes in the rock mass, the rock mass integrity factor K is adopted. v The fracture toughness of intact rock is corrected using the following formula: .

[0064] Substituting into the stress formula, the final crack initiation criterion is: when the Type I stress intensity factor reaches the fracture toughness K of the material's plane strain... ICM At this time, the crack becomes unstable and begins to crack, that is... .

[0065] S232. Energy Criterion:

[0066] Energy criteria are used to determine whether a crack propagates or penetrates. The energy theory for crack instability initiation is as follows: the driving force for crack propagation is the release of the system's elastic strain energy, and the resistance is the surface energy required to form a new fracture surface. When the crack propagates per unit area, the energy released by the system (energy release rate G) reaches the material's inherent resistance value (critical energy release rate, i.e., energy threshold value G). IC When the crack becomes unstable and initiates cracking, it will cause the crack to become unstable.

[0067] Total energy released by the propagation of new cracks (G) I The elastic potential energy released by the system when a crack propagates per unit area, i.e., the propagation force per unit fracture area, is expressed as follows:

[0068] ;

[0069] Where σ is the applied tensile stress in the far field, and E is the elastic modulus of the material.

[0070] Energy threshold value G IC The energy release threshold for crack instability initiation is the inherent resistance of a material to crack propagation.

[0071] ;

[0072] Where δ is the surface energy per unit area of ​​the material, that is, the energy required to form a new fracture surface per unit area.

[0073] Therefore, the energy criterion for crack instability initiation can be simply stated as: That is, when the newly formed crack expands, the total energy G released cumulatively... I Greater than or equal to the energy threshold value G IC At that time, it was determined that the crack had started to propagate unstably.

[0074] S233. Structural Surface Criterion:

[0075] The structural plane criterion is used to determine whether cracks propagate or penetrate. Numerous rock mechanics tests show that when cracks propagate unstably, the number of internal cracks and structural planes in the rock increases dramatically, indicating impending rock failure. Therefore, a rapid increase in internal cracks in rock can be considered a precursor to rock failure.

[0076] Figure 1 The diagram illustrates the damage evolution process of the surrounding rock after tunnel excavation, used to explain the aforementioned mechanical processes and principles. The main body is the surrounding rock (1), in which a circular tunnel (2) is formed by excavation; a unit cell (3) is taken near the tunnel wall for local mechanical analysis. A main crack or structural plane (4) already exists in the unit cell (3), which expands under the influence of excavation unloading and stress redistribution, and induces new cracks at the tip (5), reflecting the crack initiation-expansion-branching process.

[0077] To establish a structural criterion for rock failure, a full-process rock failure test can be conducted to determine the threshold value N for the number of new cracks when the rock is in a critical failure state. C This refers to the number of structural surfaces per unit volume. Therefore, the structural surface criterion is: when the number of newly generated structural surfaces per unit volume reaches or exceeds the threshold value N for the number of new cracks. c At that time, the crack was in an unstable propagation state, indicating that the rock was about to undergo overall failure.

[0078] In summary, three hierarchical instability criteria are constructed, with criteria 2 and 3 being the core criteria, of higher importance than criterion 1. Criterion 2 is a direct manifestation of the first principles of thermodynamics and serves as a sufficient thermodynamic threshold for instability; criterion 3 further introduces evolution and scale effects; criterion 2 determines "whether instability can occur," while criterion 3 determines "whether instability has actually occurred," together forming the core framework of the integrity criterion chain. They respectively anchor the necessary thermodynamic conditions and sufficient kinetic conditions for surrounding rock instability, jointly constituting the modern rock mass fracture mechanics paradigm of "energy-driven damage response." Criterion 1 serves as an auxiliary indicator.

[0079] Criterion 1 (Stress-Driven Criterion): ,Right now , ;

[0080] Criterion 2 (Energy Release Criterion): ,Right now , ;

[0081] Criterion 3 (Structural Surface Extension Criterion): ,Right now , .

[0082] S3. Construction of a multi-source intelligent fusion model based on stacked autoencoders:

[0083] Stacked Autoencoder (SAE) is used as the core algorithm for multi-source fusion. The excess coefficients of three types of feature parameters are used as input, and the output is the surrounding rock stability index S with a value range of 0 to 1.

[0084] S31. Algorithm selection criteria:

[0085] Heterogeneous feature adaptability: Stress, energy, and crack quantity are heterogeneous data with different physical meanings and dimensions. SAE can automatically extract deep implicit information through unsupervised layer-by-layer pre-training, realize nonlinear fusion of multi-source features, and avoid the subjectivity of manual feature engineering.

[0086] Prior knowledge integration: Feature weights can be initialized at the input layer based on the differences in the importance of the criteria, and then adaptively optimized through supervised fine-tuning. This integrates the basic understanding of principles, mechanisms and engineering experience while retaining the objectivity of data-driven approaches.

[0087] Small sample generalization ability: The number of measured samples of deep surrounding rock instability is limited. SAE's unsupervised pre-training mechanism can make full use of unlabeled monitoring data to improve the model's generalization ability and meet the needs of small sample evaluation.

[0088] Computational efficiency advantages: SAE has fast training speed and low inference latency, enabling real-time dynamic evaluation of the stability of surrounding rock and meeting the timeliness requirements of on-site monitoring and early warning.

[0089] S32. Model building process:

[0090] S321. Input Feature Construction: Calculate the exceedance coefficients of three types of features: stress exceedance coefficient Energy exceedance coefficient Structural surface exceedance coefficient α, β, and γ are used as the model input vectors.

[0091] S322. Prior Weight Initialization: Initialize the input feature weights based on the differences in the importance of the criteria. The core essence of energy release and surrounding rock cracking failure has the highest weight; the increase in structural surfaces is a macroscopic manifestation and has the next highest weight; the initiation stress, as a threshold value, has a lower weight than structural surfaces. Therefore, the stress feature weights... Energy feature weights Structural feature weights The total weight is 1.

[0092] S323. Network Structure: The model consists of an input layer, two sparse autoencoder hidden layers, and an output layer. The input layer has 3 neurons, the first hidden layer has 8 neurons, and the second hidden layer has 4 neurons. Unsupervised, layer-by-layer pre-training is used to extract deep fusion information from multi-source features. The output layer has 1 neuron, and the Sigmoid activation function maps the output to... The surrounding rock stability index S is obtained from the interval.

[0093] S324. Model Training Optimization: The model is fine-tuned in a supervised manner using a dataset consisting of engineering measured data and numerical simulation samples. The network weights and biases are optimized through the backpropagation algorithm to achieve adaptive nonlinear fusion of multi-source features.

[0094] S4. Conduct stability classification evaluation of deep surrounding rock:

[0095] Surrounding rock stability index ,by The threshold for distinguishing between stable and unstable conditions; the stability classification evaluation includes:

[0096] when When the surrounding rock is in a stable or basically stable state, it is determined that the surrounding rock is in a stable state.

[0097] when If the value approaches 1, the surrounding rock is determined to be in a state of complete failure.

[0098] For other values, the judgment is made in conjunction with the satisfaction of the aforementioned instability criterion.

[0099] Furthermore, the grading rules are subdivided as follows:

[0100] ① Or it may approach 1, corresponding to the simultaneous satisfaction of all three criteria, indicating that the surrounding rock is in a state of complete failure;

[0101] ② When both criteria 2 and criterion 3 are met, the surrounding rock is in an unstable state and there is a high risk of instability.

[0102] ③ When criteria 1 and criterion 2 are satisfied at the same time, but criterion 3 is not satisfied, the surrounding rock is in a basically stable state, but there is a low to medium risk of instability. The risk increases with the increase of the over-limit coefficient.

[0103] ④ When criteria 1 and criterion 3 are satisfied at the same time, but criterion 2 is not satisfied, the surrounding rock is in a critical stable state and there is a medium to high risk of instability. The risk increases with the increase of the over-limit coefficient.

[0104] ⑤ When criterion 2 is satisfied (regardless of whether criteria 1 and 3 are satisfied), the risk increases with the increase of the excess coefficient;

[0105] ⑥ When criterion 3 is satisfied (regardless of whether criteria 1 and 2 are satisfied), the risk increases with the increase of the excess coefficient;

[0106] ⑦ When only criterion 1 is satisfied, the surrounding rock is in a stable state;

[0107] ⑧ When all criteria are not met, the surrounding rock is in a stable state.

[0108] The grading evaluation rules are listed below (√ indicates compliance, × indicates non-compliance): Table 1. Grading Evaluation Rules

[0109] When the above methods are applied to deep engineering sites, stress, energy, and structural surfaces are obtained through on-site monitoring or testing. Stress can be measured by sensors, energy by microseismic monitoring or optical fiber grating, and structural surfaces can be obtained through borehole testing or 3D laser scanning. Threshold values ​​can be obtained through indoor or on-site testing.

[0110] When the above method is used in numerical simulation software, such as numerical simulation of deep engineering excavation process, the stress, energy, and structural surface data of each unit in the surrounding rock can be determined in the simulation program.

[0111] Example 1

[0112] This embodiment uses a deep-buried tunnel project as an example. The tunnel is 600m deep, and the surrounding rock is medium-coarse-grained granite, belonging to a high-stress region dominated by tectonic stress. The measured in-situ stress data are: maximum principal stress 35MPa, intermediate principal stress 18MPa, and minimum principal stress 13MPa. A 1m×1m×1m cube surrounding rock unit at a typical location on the tunnel sidewall is selected as the evaluation object, and the stress-energy-structural surface fusion method proposed in this invention is used for stability evaluation. The specific implementation steps are as follows:

[0113] S1. Multi-source data acquisition and model building:

[0114] S11. Geological and Mechanical Data Acquisition: Based on the actual deep engineering site, collect data on the distribution of geostress, basic rock mechanics parameters (such as elastic modulus, cohesion, and internal friction angle), and geometric parameters of structural surfaces in the engineering area. For the medium-coarse-grained granite in this embodiment, through indoor uniaxial and triaxial compression tests, the elastic modulus was measured to be 31.5 GPa, the cohesion to be 21 MPa, the internal friction angle to be 48°, and the plane strain fracture toughness K... IC 1.8 MPa·m 0.5 .

[0115] S12. Numerical Simulation and Inversion Analysis: Using FLAC3D-PFC coupled numerical simulation software, the aforementioned geostress and rock mechanics parameters were input to simulate the stress redistribution, energy evolution, and crack propagation patterns of the surrounding rock during tunnel excavation and unloading. Simultaneously, the simulation results were calibrated using on-site microseismic monitoring data to ensure the model's consistency with the actual physical field, thereby obtaining the dynamic response data of the evaluation unit during the excavation process.

[0116] S13. Real-time feature extraction:

[0117] The feature parameters of the evaluation unit are extracted from the model established in step S12 as follows:

[0118] Maximum Type I stress intensity factor at the tip of the structural surface: ;

[0119] Total energy released by the propagation of new cracks under excavation disturbance: (1m) 3 (within the evaluation unit)

[0120] Number of newly formed macroscopic cracks per unit volume: .

[0121] Through steps S11 and S12, a high-precision mechanical response model of the surrounding rock was established, which solved the problem that traditional theoretical analytical methods could not take into account the heterogeneity of deep and complex geological environments, and provided reliable physical field basis data for subsequent criterion calculations.

[0122] S2. Calculate the excess coefficient:

[0123] Based on the characteristic parameters and instability threshold value of the evaluation unit, the exceedance coefficient under the current surrounding rock condition is calculated. The instability threshold value includes the fracture toughness K. ICM Energy threshold value G IC Threshold value N for the number of new cracks C Characteristic parameters include: Type I stress intensity factor K I The total energy released by the propagation of new cracks is G. IThe number of newly formed macroscopic cracks per unit volume, N.

[0124] S21. Calculate fracture toughness:

[0125] Based on the linear elastic fracture mechanics' type I crack instability propagation criterion, the stress threshold value is determined to be the plane strain fracture toughness of the rock mass structural plane. Considering the weakening effect of the primary structural plane in the rock mass, a rock mass integrity coefficient K is introduced. v (In this embodiment) The fracture toughness of intact rock is corrected. The calculation formula is: .

[0126] Substitute the numerical values ​​into the calculation: This value represents the stress intensity factor at the tip of the structural surface. When this value is reached, the crack will become unstable and initiate.

[0127] S22. Calculate the energy threshold value:

[0128] Based on the energy release rate theory of fracture mechanics, the critical energy release rate G per unit area under plane strain state is utilized. ICM With fracture toughness K ICM Quantitative relationship: Substitute the numerical values ​​into the calculation: .

[0129] Based on the statistical results of discrete element simulation, the critical total area of ​​crack propagation before instability and penetration within the evaluation unit is determined. Then the energy threshold value is: .

[0130] This value represents the energy released when the crack propagation energy within the unit reaches this value, which will trigger an unstable energy release.

[0131] S23. Calculate the structural surface threshold value:

[0132] The method adopted is a three-step approach: "indoor test benchmark + rock mass reduction correction + field monitoring calibration".

[0133] S231. Triaxial compression tests were conducted on granite, combined with acoustic emission and CT scans. The number of macroscopic cracks N0 in the post-peak failure area was statistically analyzed, which was approximately 14 cracks / m³. 3 .

[0134] S232. Considering that the rock mass contains primary structural planes, a reduction factor of 0.75 is taken, resulting in a preliminary threshold value of 10.5 planes / m. 3 .

[0135] S233. Establish a mapping relationship by combining microseismic monitoring data from similar projects in the same region, and finally calibrate and determine the threshold value N for the number of new cracks in this project. c =10 strips / m 3This value represents the number of new cracks per unit volume that, when reached, will gradually penetrate to form a failure surface.

[0136] Through steps S21 to S23, three core criterion thresholds based on physical mechanisms rather than empirical statistics were constructed. These thresholds directly correspond to the critical physical conditions for surrounding rock instability, overcoming the shortcomings of existing engineering analogy methods that rely on subjective experience and have low quantification levels, thus significantly improving the scientific rigor and accuracy of the evaluation.

[0137] In this embodiment, the key parameters involved in the threshold value are obtained through indoor tests or on-site testing and monitoring. An alternative solution is to obtain them through on-site monitoring inversion, that is, based on the monitoring data of on-site stress, micro-seismic activity, deformation, etc., the stress level and released energy of the surrounding rock when the rock instability occurs are inverted, and the corresponding threshold value is obtained through statistical regression.

[0138] S24. Calculate the excess coefficient:

[0139] Calculate the exceedance coefficients for the three types of features respectively, and use them as the model input vector. Type I stress intensity factor K I The total energy released by the propagation of new cracks is G. I The number of newly formed macroscopic cracks per unit volume, N, is obtained from the feature extraction in step S1.

[0140] Stress exceedance factor:

[0141]

[0142] Energy Exceedance Factor:

[0143]

[0144] Crack quantity exceeding limit factor:

[0145]

[0146] The calculation of the overlimit coefficient normalizes the characteristic parameters of different physical dimensions (stress, energy, quantity) into dimensionless overlimit coefficients, eliminating the influence of dimensional differences and providing standardized input data for subsequent multi-source intelligent fusion.

[0147] S25. Instability Criterion Satisfaction Judgment:

[0148] Criterion 1 (Stress-Driven Criterion): Not satisfied.

[0149] Criterion 2 (Energy Release Criterion): ,satisfy.

[0150] Criterion 3 (Structural Surface Extension Criterion): ,satisfy.

[0151] S3. Construction and Inference of a Multi-Source Intelligent Fusion Model Based on Stacked Autoencoders (SAE):

[0152] S31. Model Initialization and Weight Setting: Construct an SAE model consisting of an input layer (3 neurons, corresponding to α, β, and γ), two sparse autoencoder hidden layers (8 neurons in the first hidden layer and 4 neurons in the second hidden layer), and an output layer (1 neuron). Based on the difference in criterion importance, initialize the input feature weights: energy feature weight β = 0.4 (highest), and structural surface feature weight... Stress characteristic weights .

[0153] S32. Model Training and Optimization: Using a dataset containing 120 sets of numerical simulation samples and 32 sets of engineering measurement samples, the SAE model was unsupervised pre-trained to extract deep hidden information, followed by supervised fine-tuning. The network weights and biases were optimized using the backpropagation algorithm to achieve nonlinear adaptive fusion of multi-source heterogeneous features.

[0154] S33. Stability index calculation: The calculated excess coefficient vector ( , , The input is the trained SAE model. After nonlinear fusion computation during network forward propagation, the output layer uses the Sigmoid activation function to map the output to... The range is used to obtain the surrounding rock stability index. .

[0155] Through step S3, the SAE algorithm was used to organically integrate the three elements of stress, energy, and structural planes. This method incorporates prior knowledge of fracture mechanics principles (through weight initialization) and possesses data-driven objectivity (through fine-tuning), effectively avoiding the subjective bias of manual weighting. It also maintains excellent generalization ability under small sample conditions, achieving accurate quantitative evaluation of the stability state of surrounding rock.

[0156] S4. Surrounding Rock Stability Determination and Comprehensive Evaluation:

[0157] Further judgment is made based on the satisfaction of instability criteria: According to the list of graded evaluation rules, when criteria 2 and 3 are satisfied at the same time, the surrounding rock of the evaluation unit is determined to be in an unstable state and has a high risk of instability.

[0158] In this embodiment, although the stress did not reach the crack initiation threshold, the energy accumulation had exceeded the critical threshold and the number of cracks increased sharply, which is consistent with the instability characteristics of "energy-crack dual drive".

[0159] Overall conclusion: The surrounding rock of this evaluation unit is in an unstable state with a high risk of instability, which is consistent with the typical pattern of "stress not reaching the strength threshold, but energy accumulation and crack propagation have exceeded the critical point, which can easily induce energy-driven spalling and rockburst" in the surrounding rock of deeply buried tunnels.

[0160] Engineering recommendations: The excavation and blasting of this tunnel section should be adjusted, the blasting parameters optimized, and the unloading weakened; and timely measures should be taken to strengthen the support (such as lengthening anchor bolts, prestressed anchor cables, etc.) and reduce the degree of energy accumulation in the surrounding rock in conjunction with stress relief holes; at the same time, the frequency of monitoring such as microseismic monitoring should be increased to continuously observe the evolution of the stability state of the surrounding rock.

[0161] The S4 step not only provides a quantitative risk level, but also explains the cause of instability by combining physical mechanisms, providing engineering designers with a decision-making basis with clear guidance, and effectively making up for the problems of high misjudgment rate and delayed early warning in the traditional single index evaluation in deep high-energy environment.

[0162] Using the same evaluation unit in this embodiment, stability was evaluated using three mainstream traditional methods:

[0163] Theoretical analysis (stress criterion): Based on the Mohr-Coulomb criterion, the safety factor of the surrounding rock was calculated to be 1.28, which was judged as "stable"; Numerical simulation (displacement criterion): Using the maximum horizontal displacement of 8.2 mm as the threshold (recommended value in the code), the measured displacement was 7.5 mm, which was judged as "basically stable"; Microseismic monitoring back analysis (b-value method): The calculated b-value of the microseismic event was 0.93, which was judged as "low risk". None of the three methods indicated a high risk of instability in this area.

[0164] On-site response verification: During subsequent construction, local spalling occurred at the evaluation unit location within 48 hours of excavation, with a total mass of approximately 1.2 tons of detached rock. Dense new cracks were observed, and the energy level of microseismic events significantly increased (maximum moment magnitude reached...). This closely matches the "unstable state" predicted by this invention. The aforementioned three traditional methods failed to identify this risk in advance.

[0165] Regarding parameter portability and adaptability, the threshold value determination process and SAE model architecture used in this embodiment are universally applicable. When applied to other lithologies (such as marble and sandstone) or projects at different burial depths (800m, 1200m), it is only necessary to re-conduct indoor fracture tests on the corresponding rock mass, adjust the reduction factor, and use a limited sample under the new working conditions to perform lightweight fine-tuning of the SAE model (without full retraining) to achieve rapid adaptation.

[0166] Regarding real-time dynamic early warning capabilities, this method supports online calculation throughout the entire process: the on-site microseismic system outputs the energy release G every 5 minutes. I The fiber optic strain sensor updates the stress intensity factor K in real time.I The borehole panoramic scanner updates the structural surface density N weekly. The SAE model takes less than 0.2 seconds for a single inference, enabling minute-level dynamic updates of the surrounding rock stability index S and automatic threshold alarms.

[0167] In summary, this embodiment fully demonstrates the entire process of the method of the present invention, from data acquisition, mechanism modeling, parameter calculation, intelligent fusion to hierarchical early warning. It realizes the quantitative coupling of energy evolution, the core driving force of catastrophe, with measurable engineering parameters, and elevates the evaluation of surrounding rock stability from "experience-driven and phenomenon-description" to a new stage of "mechanism-driven and process-quantitative". It provides scientific, practical and forward-looking technical support for ensuring the safety of deep major underground engineering projects.

[0168] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various modifications and alterations without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A method for evaluating the stability of deep surrounding rock by integrating stress, energy, and structural planes, characterized in that, Includes the following steps: S1. Collect geological and mechanical data of the engineering area and construct a stress and energy model of the engineering area; the geological and mechanical data includes in-situ stress, rock mass mechanical parameters, and structural surface geometric parameters; Based on the stress and energy model, characteristic parameters of the evaluation unit are extracted, including: Type I stress intensity factor K. I The total energy released by the propagation of new cracks is G. I The number of newly formed macroscopic cracks per unit volume, N; S2. Based on the characteristic parameters and instability threshold value of the evaluation unit, calculate the exceedance coefficients under the current surrounding rock condition and form an instability criterion; the exceedance coefficients include: stress exceedance coefficient α, energy exceedance coefficient β, and structural plane exceedance coefficient γ; the instability threshold value includes fracture toughness K. ICM Energy threshold value G IC Threshold value N for the number of new cracks C ; S3. Input the overlimit coefficient into the fusion model of the stacked autoencoder and output the surrounding rock stability index S; S4. Based on the judgment results of the surrounding rock stability index S and the instability criterion, the stability of the deep surrounding rock is evaluated and classified.

2. The method according to claim 1, characterized in that, In step S2, the fracture toughness K ICM The calculation method is as follows: the plane strain fracture toughness K of intact rock is determined by indoor rock mechanics test. IC Combined with the rock mass integrity coefficient K v By applying size effect and crack correction, the fracture toughness K is obtained. ICM .

3. The method according to claim 1, characterized in that, In step S2, the energy threshold value G IC The calculation method is as follows: first calculate the critical energy release rate per unit area. Where ν is Poisson's ratio and E is the elastic modulus; then calculate the energy threshold value. Where A is the total critical area of ​​crack propagation before instability and penetration within the evaluation unit.

4. The method according to claim 1, characterized in that, In step S2, the threshold value N for the number of newly formed cracks c The calculation formula is: Where N0 is the density of internal macroscopic cracks in the rock when it is in a critical failure state, as determined by indoor rock tests, and B is the reduction factor.

5. The method according to claim 1, characterized in that, In step S3, the formula for calculating the stress exceedance factor α is: .

6. The method according to claim 1, characterized in that, In step S3, the formula for calculating the energy exceedance coefficient β is: .

7. The method according to claim 1, characterized in that, In step S3, the formula for calculating the structural surface exceedance coefficient γ is as follows: .

8. The method according to claim 1, characterized in that, In step S3, the construction process of the fusion model of the stacked autoencoder includes: using unsupervised layer-by-layer pre-training to extract deep implicit information of multi-source features; the network structure includes an input layer, two sparse autoencoder hidden layers, and an output layer; the input layer sets neurons to correspond to stress over-limit coefficient, energy over-limit coefficient, and structural surface over-limit coefficient respectively; in the input layer initialization stage, the feature weights are set: the energy feature weight is the highest, the structural surface feature weight is the second highest, and the stress feature weight is the lowest; the output layer uses the Sigmoid activation function to map the output to the [0,1] interval to obtain the surrounding rock stability index S.

9. The method according to claim 1, characterized in that, In step S4, the instability criterion is: when When the stress-driven criterion is satisfied, where ;when At that time, the energy release criterion is satisfied, where ;when When the structural surface expansion criterion is satisfied, where .

10. The method according to claim 1, characterized in that, In step S4, the surrounding rock stability index ,by This serves as the threshold separating stability from instability. The stability grading evaluation includes: when When the surrounding rock is in a stable or basically stable state, it is determined that the surrounding rock is in a stable state. when If the value approaches 1, the surrounding rock is determined to be in a state of complete failure. For other values, the judgment is made in conjunction with the satisfaction of the aforementioned instability criterion.

Images