Intelligent hierarchical method for deep rock mass multi-field coupling test data driving
By constructing a multi-field coupling mechanism database and using intelligent feature indicators to train machine learning models, the problem of low accuracy in deep rock mass classification was solved, achieving accurate rock mass classification and future behavior prediction in deep environments, supporting engineering design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA HYDROELECTRIC ENGINEERING CONSULTING GROUP CHENGDU RESEARCH HYDROELECTRIC INVESTIGATION DESIGN AND INSTITUTE
- Filing Date
- 2026-02-27
- Publication Date
- 2026-05-15
AI Technical Summary
Existing methods for classifying deep rock masses have low accuracy in deep environments with high stress, high seepage, and high temperature, and cannot effectively reflect the dynamic coupling effect of stress field, seepage field, and temperature field. Furthermore, their reliance on scarce field data leads to insufficient model generalization ability and a lack of forward-looking prediction of the future behavior evolution of the surrounding rock.
By constructing a multi-field coupling mechanism database, intelligent feature indicators are refined, including stress-seepage coupling effect, stress-temperature coupling effect, seepage-temperature coupling effect, and stress-seepage-temperature three-field coupling effect indicators. Based on machine learning models, intelligent classification is performed to obtain the rock mass grade at the engineering site.
It enables precise classification of rock mass in deep environments, improves the accuracy of classification and the generalization ability of the model, and has the ability to predict the future behavior evolution of surrounding rock, supporting dynamic decision-making in engineering design.
Smart Images
Figure CN121743811B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of deep rock mass classification technology, and in particular to an intelligent classification method driven by multi-field coupled test data of deep rock masses. Background Technology
[0002] Rock mass classification is a core basis for engineering investigation, design, and construction, and its accuracy is crucial. For a long time, the engineering community has widely adopted semi-quantitative methods based on experience and statistics, such as RMR (Rock Mass Rating) and the Q system. These methods mainly rely on engineers to qualitatively describe and quantitatively score several key parameters (such as rock mass strength, RQD, joint spacing and characteristics, groundwater conditions, etc.) on-site, and finally determine the rock mass grade through weighted summation / product.
[0003] However, in the deep environment of "three highs" (high temperature, high humidity, and high temperature), these traditional methods reveal inherent and insurmountable limitations.
[0004] Static nature and reliance on experience: Traditional methods are based on statistical induction from a large number of completed engineering cases. They are essentially "static" empirical formulas, which are difficult to accurately reflect the dynamic and nonlinear morphological evolution process caused by the dynamic adjustment of stress field and seepage field under excavation disturbance of deep surrounding rock. They rely too much on the personal experience of engineers and are highly subjective.
[0005] Lack of Mechanism: The parameter system of traditional methods fails to fully consider and quantify the controlling influence of multi-physics field (stress-seepage-temperature) coupling on the mechanical behavior of rock masses. For example, high ground stress can compact fractures and thus change the seepage field, while high water pressure can change the effective stress and thus change the stress field. This complex feedback mechanism cannot be reflected in a simple weighted scoring.
[0006] Scale effect and lack of foresight: Sparse field survey data points make it difficult to fully reflect the heterogeneity of the macroscopic rock mass. Traditional methods classify based on a "snapshot" of the current state, lacking the ability to predict the potential deterioration path of the surrounding rock in future engineering activities (such as excavation and support), leading to conservative or risky designs.
[0007] Traditional experience-based grading methods (such as RMR and Q systems) and recently emerging field data-driven machine learning methods suffer from the following three fundamental shortcomings:
[0008] 1. Lack of mechanism and model distortion: Static empirical formulas cannot characterize dynamic coupling effects.
[0009] The linear weighted formulas relied upon by traditional classification methods are essentially static models. They cannot describe or quantify the dynamic coupling and feedback effects between stress, seepage, and temperature fields under conditions of high ground stress, high osmotic pressure, and high temperature in deep geological environments. For example, they cannot reflect the complete, nonlinear dynamic process by which excavation unloading causes stress redistribution, which in turn alters fracture aperture and seepage path, ultimately leading to the deterioration of surrounding rock strength. Therefore, under complex deep conditions, the calculation results of traditional methods deviate significantly from the actual state of the surrounding rock, and the risk of model distortion is extremely high.
[0010] 2. Data-driven dilemma: Reliance on “scarce field data” leads to weak model generalization ability.
[0011] Recent intelligent grading methods, while incorporating machine learning algorithms, rely entirely on sparse and expensive in-situ data collected directly from the field for model training. This leads to typical "data bottleneck" problems: ① Obtaining sufficient, high-quality training data is costly and time-consuming; ② Data is particularly scarce in the early stages of engineering exploration, resulting in insufficient model training; ③ Models trained based on specific engineering data are difficult to generalize to other projects with vastly different geological conditions, exhibiting severely insufficient generalization ability.
[0012] 3. Insufficient predictive ability: Lack of forward-looking judgment on the future evolution of the surrounding rock's behavior.
[0013] Both traditional and existing data-driven methods primarily rely on the "current state" of the surrounding rock for their judgments. They are more like "intelligent identification" of the current situation, lacking the ability to proactively predict the potential deterioration paths and damage risks of the surrounding rock under the disturbance of subsequent engineering activities (such as blasting, tunneling, and support). This results in insufficient foresight in engineering design, failing to provide accurate basis for dynamic design and proactive prevention. Summary of the Invention
[0014] The technical problem solved by this invention: This invention provides an intelligent classification method driven by multi-field coupling test data of deep rock masses, which solves the problem of low accuracy in existing deep rock mass classification methods.
[0015] The technical solution adopted by this invention to solve the above-mentioned technical problems is an intelligent classification method driven by multi-field coupled test data of deep rock masses, comprising the following steps:
[0016] S1. Obtain multi-field coupling test data of deep rock mass and construct a multi-field coupling mechanism database;
[0017] S2. Intelligent feature indicators characterizing multi-field coupling effects are extracted from the multi-field coupling mechanism database; the intelligent feature indicators include traditional indicators and composite indicators, and the composite indicators include: stress and seepage coupling effect indicators, stress and temperature coupling effect indicators, seepage and temperature coupling effect indicators, and stress, seepage and temperature three-field coupling effect indicators.
[0018] S3. Determine composite indices based on geological environmental parameters of deep rock masses;
[0019] S4. Using intelligent feature indicators as input and the corresponding deep rock mass grade as output, train the machine learning model to obtain an intelligent grading model of deep rock mass under different geological environment parameters.
[0020] S5. Obtain the intelligent characteristic indicators of the deep rock mass to be classified at the engineering site and the local geological environment parameters. Classify the deep rock mass using an intelligent classification model that matches the local geological environment parameters to obtain the deep rock mass grade.
[0021] Furthermore, the stress-seepage coupling effect index includes the stress-seepage coupling fracture propagation coefficient, the stress sensitivity-seepage capacity correlation coefficient, and the stress unloading-seepage abrupt change lag time; the stress-temperature coupling effect index includes the stress-temperature synergistic strength reduction coefficient, the high-temperature high-stress rock mass elastic modulus decay rate, and the stress-temperature coupling brittleness index; the seepage-temperature coupling effect index includes the temperature-seepage coupling rock mass dissolution rate and the seepage-temperature synergistic heat exchange coefficient; and the stress-seepage-temperature three-field coupling effect index includes the three-field coupling rock mass integrity deterioration index.
[0022] Furthermore, the formula for the stress-seepage coupled fracture propagation coefficient is as follows: ,in, This represents the stress-seepage coupling fracture propagation coefficient. This represents the change in the normal aperture of the fracture. This represents the effective stress change. This indicates the change in pore water pressure.
[0023] Furthermore, the formula for the correlation coefficient between stress sensitivity and seepage capacity is as follows: ,in, This represents the correlation coefficient between stress sensitivity and seepage capacity. This represents the absolute value of the permeability coefficient with respect to the rate of change of stress. This represents the permeability coefficient under the current stress.
[0024] Furthermore, the formulas for stress unloading and seepage abrupt change lag time are as follows: ,in, This indicates the hysteresis time between stress unloading and seepage abrupt change. This indicates the time corresponding to the point of abrupt change in seepage flow. This indicates the start time of stress unloading.
[0025] Furthermore, the formula for the stress-temperature synergistic strength reduction factor is: ,in, This represents the stress-temperature combined strength reduction factor. This represents the reduction in peak rock mass strength under the combined effects of temperature and stress. This represents the reduction in peak strength of the rock mass under a single stress. This represents the reduction in peak rock mass strength under a single temperature effect.
[0026] Furthermore, the formula for the stress-temperature coupled brittleness index is: ,in, Indicates the stress-temperature coupled brittleness index. This represents the absolute value of the tangent modulus of the post-peak stress-strain curve. Indicates the elastic modulus. It represents Poisson's ratio.
[0027] Furthermore, the formula for the synergistic heat exchange coefficient between seepage and temperature is: ,in, This represents the heat exchange coefficient resulting from the interaction between seepage and temperature. This represents the heat exchange power caused by seepage. Indicates the heat exchange area. It represents the average temperature difference between the fluid and the solid.
[0028] Furthermore, the formula for the three-field coupled rock mass integrity deterioration index is as follows: ,in, This represents the degradation index of the integrity of the rock mass coupled with three fields. This represents the initial integrity coefficient. This represents the current integrity coefficient. The weight representing the effect of stress on integrity. This represents the effective stress change. This indicates the weight of the impact of seepage on integrity. This represents the change in pore water pressure. This indicates the weight of the effect of temperature on integrity. It represents the amount of temperature change.
[0029] Furthermore, the composite index determined based on the geological environment parameters of the deep rock mass includes: when the maximum principal stress is ≥20MPa and groundwater is present, the composite index is the stress-seepage coupling effect index; when the maximum principal stress is ≥20MPa and the ground temperature is >28℃, the composite index is the stress-temperature coupling effect index; when the ground temperature is >28℃ and groundwater is present, the composite index is the seepage-temperature coupling effect index; when the maximum principal stress is ≥20MPa, the ground temperature is >28℃ and groundwater is present, the composite index is the stress-seepage-temperature three-field coupling effect index.
[0030] The beneficial effects of this invention are as follows: This invention provides an intelligent grading method driven by multi-field coupling test data of deep rock masses. A multi-field coupling mechanism database is constructed using multi-field coupling test data of deep rock masses, from which intelligent feature indicators characterizing the multi-field coupling effect are extracted. These intelligent feature indicators include traditional and composite indicators. Composite indicators are determined based on the geological environmental parameters of the deep rock mass location, obtaining intelligent feature indicators corresponding to deep rock masses under different geological environmental parameters. Using the intelligent feature indicators as input and the corresponding deep rock mass grade as output, a machine learning model is trained to obtain intelligent grading models of deep rock masses under different geological environmental parameters. The intelligent feature indicators of the deep rock mass to be graded at the engineering site and the local geological environmental parameters are obtained. Grading is performed using the intelligent grading model of the deep rock mass matched with the local geological environmental parameters to obtain the deep rock mass grade, thus solving the problem of low accuracy in existing deep rock mass grading methods. Attached Figure Description
[0031] Figure 1 This is a flowchart illustrating an intelligent classification method driven by multi-field coupling test data for deep rock masses, provided by the present invention. Detailed Implementation
[0032] This invention addresses the problem of low accuracy in existing deep rock mass classification methods by providing an intelligent classification method driven by multi-field coupled experimental data for deep rock masses, such as... Figure 1 As shown, it includes the following steps:
[0033] S1. Obtain multi-field coupling test data of deep rock mass and construct a multi-field coupling mechanism database.
[0034] Specifically, the acquisition of multi-field coupling test data of deep rock mass includes indoor physical tests, in-situ tests and numerical simulation tests. The indoor physical tests include high temperature and high pressure triaxial servo tests and rock creep tests. The in-situ tests include large pressure plate tests and hydraulic fracturing tests. The numerical simulation tests include multi-physics coupling simulation based on FLAC3D or COMSOL.
[0035] Multi-field coupling test data can accurately reflect the response information of rock mass from initial loading to peak failure under the coupling effect of at least two of the stress field, seepage field, and temperature field. Specifically, it includes multi-path mechanical test, temperature-mechanical coupling test, stress-seepage coupling test, and in-situ monitoring of damage evolution.
[0036] The multi-path mechanical tests include conventional triaxial compression tests, true triaxial stress loading tests, simulated excavation unloading path tests, and cyclic loading and unloading tests, which are used to obtain the strength and deformation characteristics of rock masses under different stress states and load histories.
[0037] The temperature-mechanical coupling test is used to test the mechanical properties of rock masses under different constant temperatures and thermal cycling paths, so as to quantify the deterioration effect of temperature and thermal fatigue on the mechanical parameters of rock masses.
[0038] The stress-seepage coupling test includes mechanical tests at different pore water pressure levels, permeability evolution tests under different stress paths, and long-term seepage damage tests under different hydraulic gradients, which are used to obtain response data of the rock mass from micro-damage initiation to macro-fracture under stress-seepage coupling.
[0039] The in-situ damage evolution monitoring includes the simultaneous integration of acoustic emission, microseismic monitoring systems, active ultrasonic detection systems, real-time CT scanning, or digital image correlation (DIC) technology during multi-path mechanical tests, thermo-mechanical coupling tests, and stress-flow coupling tests. This enables qualitative and quantitative characterization of the damage field of the rock mass throughout the entire process from microcrack initiation to macroscopic rupture.
[0040] This allows us to obtain multi-dimensional, synchronous response data, including full-process stress-strain data, acoustic emission event sequences and location data, wave velocity evolution data, fracture CT three-dimensional model data, permeability dynamic change data, simulated rock quality indicators and fracture density, thereby constructing a complete multi-field coupling mechanism database.
[0041] S2. Extract intelligent feature indicators characterizing multi-field coupling effects from the multi-field coupling mechanism database; the intelligent feature indicators include traditional indicators and composite indicators, and the composite indicators include: stress and seepage coupling effect indicators, stress and temperature coupling effect indicators, seepage and temperature coupling effect indicators, and stress, seepage and temperature three-field coupling effect indicators.
[0042] Specifically, the stress-seepage coupling effect index is adapted to the risks of water inrush and fracture instability; the stress-temperature coupling effect index is adapted to the risks of rock burst and strength degradation; the seepage-temperature coupling effect index is adapted to the risks of long-term degradation and support corrosion; and the stress-seepage-temperature three-field coupling effect index is adapted to the risks of multiple disasters superimposed.
[0043] The stress-seepage coupling effect index includes the stress-seepage coupling fracture propagation coefficient, the stress sensitivity-seepage capacity correlation coefficient, and the stress unloading-seepage abrupt change lag time; the stress-temperature coupling effect index includes the stress-temperature synergistic strength reduction coefficient, the high-temperature high-stress rock mass elastic modulus decay rate, and the stress-temperature coupling brittleness index; the seepage-temperature coupling effect index includes the temperature-seepage coupling rock mass dissolution rate and the seepage-temperature synergistic heat exchange coefficient; the stress-seepage-temperature three-field coupling effect index includes the three-field coupling rock mass integrity deterioration index.
[0044] The formula for the stress-seepage coupled fracture propagation coefficient is: ,in, This represents the stress-seepage coupling fracture propagation coefficient. This represents the change in the normal aperture of the fracture. This represents the effective stress change. This represents the change in pore water pressure. The stress-seepage coupling fracture propagation coefficient is used to reflect the degree of accelerated fracture propagation under the combined effects of stress and seepage.
[0045] The formula for the correlation coefficient between stress sensitivity and seepage capacity is: ,in, This represents the correlation coefficient between stress sensitivity and seepage capacity. This represents the absolute value of the rate of change of the permeability coefficient with respect to stress, also known as stress sensitivity. This represents the permeability coefficient under the current stress. The correlation coefficient between stress sensitivity and seepage capacity is used to reflect the stability of high-stress rock masses under seepage. For example, stress-sensitive rock masses are prone to fracture instability when exposed to seepage.
[0046] The formulas for stress unloading and seepage abrupt change lag time are as follows: ,in, This indicates the hysteresis time between stress unloading and seepage abrupt change. This indicates the time corresponding to the point of abrupt change in seepage flow. This indicates the start time of stress unloading, also known as the time corresponding to the stress unloading abrupt change point. The stress unloading and seepage abrupt change lag time is used to reflect the delayed response characteristics of the seepage field after stress unloading induces crack propagation; the shorter the lag time, the more sudden the risk.
[0047] The formula for the stress-temperature synergistic strength reduction factor is: ,in, This represents the stress-temperature combined strength reduction factor. This represents the reduction in peak rock mass strength under the combined effects of temperature and stress. This represents the reduction in peak strength of the rock mass under a single stress. This represents the reduction in peak rock mass strength under a single temperature effect. A stress-temperature synergistic strength reduction factor greater than 1 indicates a detrimental synergistic effect of temperature and stress.
[0048] The rate of decrease in elastic modulus of high-temperature and high-stress rock mass is used to reflect the deterioration trend of rock mass stiffness under long-term thermal stress. The lower the stiffness, the more likely it is to undergo plastic deformation, leading to support failure.
[0049] The formula for the stress-temperature coupled brittleness index is: ,in, Indicates the stress-temperature coupled brittleness index. This represents the absolute value of the tangent modulus of the post-peak stress-strain curve. Indicates the elastic modulus. This represents Poisson's ratio. The larger the stress-temperature coupled brittleness index, the steeper the stress drop after the peak, and the stronger the brittleness.
[0050] The temperature-seepage coupling of rock dissolution rate reflects the degree of chemical erosion of rock fissure walls by deep, highly mineralized groundwater (such as sulfate-containing groundwater) under the influence of temperature. The faster the dissolution, the easier it is for the fissures to expand.
[0051] The formula for the heat exchange coefficient synergistic between seepage and temperature is: ,in, This represents the heat exchange coefficient resulting from the interaction between seepage and temperature. This represents the heat exchange power caused by seepage. Indicates the heat exchange area. This represents the average temperature difference between the fluid and the solid. The seepage-temperature synergistic heat exchange coefficient is used to reflect the indirect coupling effect of seepage carrying away or bringing in heat, which causes changes in the rock mass temperature and thus alters the rock mass strength. For example, low-temperature seepage increases the brittleness of the rock mass, while high-temperature seepage softens it.
[0052] The formula for the three-field coupled rock mass integrity degradation index is: ,in, This represents the degradation index of the integrity of the rock mass coupled with three fields. This represents the initial integrity coefficient. This represents the current integrity coefficient. The weight representing the effect of stress on integrity. This represents the effective stress change. This indicates the weight of the impact of seepage on integrity. This represents the change in pore water pressure. This indicates the weight of the effect of temperature on integrity. It represents the amount of temperature change. The three-field coupled rock mass integrity degradation index comprehensively quantifies the dynamic decay degree of the rock mass integrity index under the combined effects of stress, seepage, and temperature. It reflects the ultimate destructive effect of the three fields superimposed on the rock mass structure, which is different from the local effects of single or two-field coupling.
[0053] Traditional indicators include: acoustic emission event rate, cumulative number of events, acoustic emission energy rate, cumulative energy, b-value (magnitude-frequency relationship), RA value (rise time / amplitude), and average frequency extracted from acoustic emission data. Among them, the decrease in b-value is a key precursor to the development of microcracks in rocks from random distribution to localization and macroscopic fracturing; the RA value is used to differentiate fracturing patterns (tensional cracks are usually high-frequency and low-RA, while shear cracks are usually low-frequency and high-RA).
[0054] Extract the following from ultrasonic data: longitudinal wave velocity, transverse wave velocity, wave velocity damage factor, and wave velocity anisotropy ratio. The wave velocity damage factor quantifies the stiffness degradation caused by damage; the wave velocity anisotropy ratio reflects the material anisotropy caused by damage.
[0055] Extract the following from the stress-strain curve: peak strength, elastic modulus and Poisson's ratio, stress and strain at the yield point, and stress drop rate after the peak.
[0056] Traditional indicators also include the RQD value of borehole cores measured in field tests, joint spacing and orientation, field point load strength, and wave velocity obtained from acoustic or seismic wave tests.
[0057] S3. Based on the geological environment parameters of the deep rock mass, determine composite indices to obtain intelligent characteristic indices corresponding to deep rock masses under different geological environment parameters.
[0058] Specifically, the geological environmental parameters of the deep rock mass are divided into four cases: maximum principal stress ≥ 20 MPa and groundwater presence, maximum principal stress ≥ 20 MPa and ground temperature > 28℃, ground temperature > 28℃ and groundwater presence, and maximum principal stress ≥ 20 MPa, ground temperature > 28℃ and groundwater presence. The composite index determined based on the geological environmental parameters of the deep rock mass includes: when the maximum principal stress ≥ 20 MPa and groundwater presence, the composite index is the stress-seepage coupling effect index; when the maximum principal stress ≥ 20 MPa and ground temperature > 28℃, the composite index is the stress-temperature coupling effect index; when the ground temperature > 28℃ and groundwater presence, the composite index is the seepage-temperature coupling effect index; and when the maximum principal stress ≥ 20 MPa, ground temperature > 28℃ and groundwater presence, the composite index is the stress-seepage-temperature three-field coupling effect index.
[0059] S4. Using intelligent feature indicators as input and the corresponding deep rock mass grade as output, train a machine learning model to obtain an intelligent grading model of deep rock mass under different geological environment parameters.
[0060] Specifically, for four different geological environment parameters, intelligent grading models for deep rock masses were obtained for each parameter. During training, intelligent feature indicators were used as inputs, and the corresponding deep rock mass grades were used as outputs. A sample set was constructed and divided into training, validation, and test sets. The gradient boosting decision tree algorithm was used to train the model on the training set, and the model hyperparameters were automatically optimized using cross-validation and Bayesian optimization methods, guided by the performance on the validation set, to obtain the optimal model. Accuracy, F1 score, and confusion matrix were used as evaluation indicators. The model performance was finally evaluated unbiasedly on the test set, which had not participated in any training or hyperparameter tuning, to ensure that the model had good generalization ability. Model interpretability techniques such as SHAP were used to analyze the contribution of each intelligent feature indicator to the grading results, confirming that the composite feature indicators played a dominant role in the model's decision-making, thus indicating that the model reasoned based on physical mechanisms rather than simple data fitting.
[0061] S5. Obtain the intelligent characteristic indicators of the deep rock mass to be classified at the engineering site and the local geological environment parameters. Classify the deep rock mass using an intelligent classification model that matches the local geological environment parameters to obtain the deep rock mass grade.
[0062] Specifically, by analyzing the geological environmental parameters of the deep rock mass to be classified at the engineering site, a matching intelligent classification model for the deep rock mass is determined. Intelligent characteristic indicators are input, and the matching intelligent classification model outputs the deep rock mass classification. This provides support for subsequent engineering decisions, such as support design and risk warning.
Claims
1. A smart classification method driven by multi-field coupled test data of deep rock masses, characterized in that, Includes the following steps: S1. Obtain multi-field coupling test data of deep rock mass and construct a multi-field coupling mechanism database; S2. Intelligent feature indicators characterizing multi-field coupling effects are extracted from the multi-field coupling mechanism database; the intelligent feature indicators include traditional indicators and composite indicators, and the composite indicators include: stress and seepage coupling effect indicators, stress and temperature coupling effect indicators, seepage and temperature coupling effect indicators, and stress, seepage and temperature three-field coupling effect indicators. The stress-seepage coupling effect indices include the stress-seepage coupling fracture propagation coefficient, the stress sensitivity-seepage capacity correlation coefficient, and the stress unloading-seepage abrupt change lag time; the stress-temperature coupling effect indices include the stress-temperature synergistic strength reduction coefficient, the high-temperature high-stress rock mass elastic modulus decay rate, and the stress-temperature coupling brittleness index; the seepage-temperature coupling effect indices include the temperature-seepage coupling rock mass dissolution rate and the seepage-temperature synergistic heat exchange coefficient; and the stress-seepage-temperature three-field coupling effect indices include the three-field coupling rock mass integrity deterioration index. The formula for the stress-seepage coupled fracture propagation coefficient is: ,in, This represents the stress-seepage coupling fracture propagation coefficient. This represents the change in the normal aperture of the fracture. This represents the effective stress change. This indicates the change in pore water pressure; The formula for the correlation coefficient between stress sensitivity and seepage capacity is: ,in, This represents the correlation coefficient between stress sensitivity and seepage capacity. This represents the absolute value of the permeability coefficient with respect to the rate of change of stress. This represents the permeability coefficient under the current stress. The formulas for stress unloading and seepage abrupt change lag time are as follows: ,in, This indicates the hysteresis time between stress unloading and seepage abrupt change. This indicates the time corresponding to the point of abrupt change in seepage flow. Indicates the stress unloading start time; The formula for the stress-temperature synergistic strength reduction factor is: ,in, This represents the stress-temperature combined strength reduction factor. This represents the reduction in peak rock mass strength under the combined effects of temperature and stress. This represents the reduction in peak strength of the rock mass under a single stress. This represents the reduction in peak rock mass strength under a single temperature effect. The formula for the stress-temperature coupled brittleness index is: ,in, Indicates the stress-temperature coupled brittleness index. This represents the absolute value of the tangent modulus of the post-peak stress-strain curve. Indicates the elastic modulus. Indicates Poisson's ratio; The formula for the heat exchange coefficient synergistic between seepage and temperature is: ,in, This represents the heat exchange coefficient resulting from the combined effects of seepage and temperature. This represents the heat exchange power caused by seepage. Indicates the heat exchange area. This represents the average temperature difference between the fluid and the solid. The formula for the three-field coupled rock mass integrity degradation index is: ,in, This represents the degradation index of the integrity of the rock mass under three-field coupling. This represents the initial integrity coefficient. This represents the current integrity coefficient. The weight representing the effect of stress on integrity. This represents the effective stress change. This indicates the weight of the impact of seepage on integrity. This represents the change in pore water pressure. This indicates the weight of the effect of temperature on integrity. Indicates the amount of temperature change; S3. Determine composite indices based on the geological environment parameters of the deep rock mass to obtain intelligent characteristic indices corresponding to the deep rock mass under different geological environment parameters; S4. Using intelligent feature indicators as input and the corresponding deep rock mass grade as output, train the machine learning model to obtain an intelligent grading model of deep rock mass under different geological environment parameters. S5. Obtain the intelligent characteristic indicators of the deep rock mass to be classified at the engineering site and the local geological environment parameters. Classify the deep rock mass using an intelligent classification model that matches the local geological environment parameters to obtain the deep rock mass grade.
2. The intelligent classification method driven by multi-field coupled test data of deep rock mass according to claim 1, characterized in that, The composite index determined based on the geological environment parameters of the deep rock mass includes: when the maximum principal stress is ≥20MPa and groundwater is present, the composite index is the stress-seepage coupling effect index; when the maximum principal stress is ≥20MPa and the ground temperature is >28℃, the composite index is the stress-temperature coupling effect index; when the ground temperature is >28℃ and groundwater is present, the composite index is the seepage-temperature coupling effect index; when the maximum principal stress is ≥20MPa, the ground temperature is >28℃ and groundwater is present, the composite index is the stress-seepage-temperature three-field coupling effect index.