A refined intelligent modeling and deformation prediction method for tunnel surrounding rock during the construction process

Abstract

This invention provides a refined intelligent modeling and deformation prediction method for tunnel surrounding rock during the construction process, belonging to the field of intelligent technology in tunnel engineering and geotechnical engineering. The method includes: S1: acquiring original drilling parameters; S2: analyzing original drilling parameters; S3: obtaining equivalent Mohr-Coulomb mechanical parameters; S4: generating a three-dimensional spatial distribution field of rock mass mechanical parameters; S5: constructing a refined heterogeneous numerical model; S6: deformation prediction. This invention overcomes the technical defects of existing technologies, such as insufficient accuracy in tunnel deformation prediction and difficulty in revealing the mechanical response of non-uniform surrounding rock, due to the lack of physical mechanisms in inversion, insufficient fidelity in spatial modeling, and lack of dynamic evolution of the model.

Description

Technical Field

[0001] This invention relates to the field of intelligent technology in tunnel engineering and geotechnical engineering, and in particular to a method for refined intelligent modeling and deformation prediction of tunnel surrounding rock for the construction process. Background Technology

[0002] With the continuous advancement of infrastructure construction, tunnel engineering is developing towards longer, deeper, and larger cross-sections. Tunnels traversing complex geological structures, such as the deep-buried extra-long tunnels in Southwest China, exhibit highly heterogeneous, anisotropic, and uncertain surrounding rock conditions. Traditional geological exploration methods, such as core drilling and laboratory testing, while providing precise point-based mechanical parameters, are costly, time-consuming, and produce sparse data, making it difficult to comprehensively characterize the continuous distribution of surrounding rock mechanical properties in three-dimensional space. In numerical simulations, engineers typically use homogeneous models or simple layered models based on the average values ​​of a few test points. This simplification masks the inherent spatial variability of the rock mass, leading to significant deviations between numerical prediction results (such as deformation and stress) and field monitoring data. It fails to accurately reveal key mechanical behaviors such as asymmetric deformation and stress concentration induced by local weak zones, severely restricting the optimization of support design and the proactive control of construction risks.

[0003] The maturity of Measurement While Drilling (MWD) technology has provided a new data source for solving the aforementioned problems. This technology, by recording response parameters such as drilling speed, torque, pressure, and impact energy in real time during the drilling process, provides a new means for continuous, in-situ characterization of the surrounding rock ahead of the drilling face. In recent years, it has been widely used for surrounding rock quality classification, lithology identification, and estimation of uniaxial compressive strength, and has shown promising application prospects in tunnel engineering. However, existing research mostly focuses on mapping MWD data to lithology categories or a few empirical indicators, lacking a systematic inversion framework for constitutive model parameters (such as elastic modulus, cohesion, and internal friction angle) directly coupled with numerical simulation. In engineering applications, MWD data is more often used as "auxiliary interpretive information" and has not yet been fully transformed into a high-value parameter set that can directly drive three-dimensional mechanical analysis. The existing technological system still has fundamental shortcomings in achieving the leap from "drilling signals" to "mechanical parameter fields that can be used for high-fidelity numerical simulation."

[0004] First, at the parameter inversion level, there is a lack of rigorous physical mechanisms to bridge the gap. Although most data-driven models (such as neural networks and random forests) can establish a black-box mapping between MWD parameters and target variables (such as UCS), the models lack clear theoretical support for rock fracturing energy dissipation and rock mass strength. They are essentially "black-box" operations with poor physical interpretability. Furthermore, the inversion output parameters are of a single type, making it difficult to systematically obtain the complete set of Mohr-Coulomb parameters (elastic modulus E, cohesion c, internal friction angle φ) with inherent consistency required for numerical simulation.

[0005] Secondly, achieving a balance between "high fidelity" and "non-uniformity" in spatial modeling is challenging. Even if mechanical parameters can be derived from the borehole location, extrapolating the sparse one-dimensional borehole "hard data" to continuous three-dimensional space remains a challenge. While traditional Kriging interpolation methods can perform spatial estimation, their inherent smoothing effect significantly weakens the spatial variability of parameters, smoothing out local weak areas crucial to engineering safety. Pure random field simulation methods (such as sequential Gaussian simulations) can characterize spatial variability, but the generated results are randomized and may not strictly adhere to known borehole data points, leading to discrepancies between the model and measured values ​​at critical locations (i.e., the borehole itself), thus losing the constraint of data accuracy.

[0006] Third, at the closed-loop application level, the model lacks dynamic evolution capabilities. Existing modeling methods are mostly "open-loop" models, meaning that after building a model based on historical data, it is directly used for prediction of newly excavated sections. The model cannot use the actual deformation monitoring data obtained after the new excavation section is exposed to perform feedback correction and optimization of the inversion model or parameter field. As a result, the model's predictive ability cannot evolve with the progress of the project and is difficult to adapt to the dynamic changes in the surrounding rock conditions along the tunnel axis.

[0007] The aforementioned bottlenecks have resulted in current MWD data applications mostly remaining at the level of surrounding rock classification, failing to deeply empower the refined design, dynamic feedback, and precise risk management of tunnel engineering. Therefore, there is an urgent need to develop a new method for refined modeling of tunnel surrounding rock that integrates physical mechanism-driven and data science-driven approaches. This method should construct a complete technical closed loop from "real-time drilling signals" to "high-fidelity non-uniform numerical models" and then to "precise deformation prediction and dynamic optimization," providing a core digital analysis engine for intelligent tunnel construction and safe operation and maintenance. Summary of the Invention

[0008] This invention provides a refined intelligent modeling and deformation prediction method for tunnel surrounding rock in the construction process. It can overcome the technical defects of existing technologies, such as insufficient accuracy of tunnel deformation prediction and difficulty in revealing the mechanical response of non-uniform surrounding rock, due to the lack of physical mechanism in inversion, insufficient fidelity of spatial modeling, and lack of dynamic evolution of the model.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A refined intelligent modeling and deformation prediction method for tunnel surrounding rock oriented towards the construction process includes:

[0011] S1: During the drilling process at the tunnel face, raw drilling parameters are collected in real time, including drilling speed, propulsion pressure, impact pressure, and rotation pressure; the raw drilling parameters are systematically cleaned, outlier removed, and standardized preprocessed to form a drilling time series dataset;

[0012] S2: Based on the borehole time series dataset, combined with the mechanical specific energy theory and the Hoek-Brown rock strength criterion, a multi-stage physical inversion model is constructed to resolve the original drilling parameters into equivalent Mohr-Coulomb mechanical parameters with clear physical meaning.

[0013] S3: Obtain the equivalent Mohr-Coulomb mechanical parameters corresponding to each borehole spatial location, including elastic modulus, cohesion, and internal friction angle;

[0014] S4: A hybrid spatial interpolation algorithm combining sequential Gaussian simulation and Hoffman correction is adopted. With equivalent Mohr-Coulomb mechanical parameters as spatial constraints, conditional simulation and iterative correction are performed in the three-dimensional tunnel model domain to generate a three-dimensional spatial distribution field of rock mass mechanical parameters.

[0015] S5: Establish a tunnel numerical model, assign the three-dimensional rock mass mechanical parameter spatial distribution field to the tunnel numerical model, and construct a refined heterogeneous numerical model that can reflect the spatial variability of the surrounding rock.

[0016] S6: On a refined heterogeneous numerical model, simulate the entire tunnel excavation process, calculate and output the deformation prediction results of the surrounding rock under excavation disturbance, including crown settlement, perimeter convergence and asymmetric deformation mode.

[0017] This specification also includes the following methods for refined intelligent modeling and deformation prediction of tunnel surrounding rock for the construction process:

[0018] S7: Compare the deformation prediction results with the actual monitoring data obtained at the corresponding mileage of the tunnel site to quantitatively evaluate the prediction accuracy of the refined heterogeneous numerical model and obtain the quantitative evaluation results.

[0019] S8: Based on the quantitative evaluation results, the parameters of the multi-stage physical inversion model are corrected and optimized, and the optimized multi-stage physical inversion model is applied to the next prediction, forming a closed-loop iterative process of data acquisition-inversion-modeling-prediction-verification-optimization.

[0020] In this specification, the specific methods of systematic cleaning, outlier removal, and standardized preprocessing described in S1 are as follows: A multi-algorithm consensus mechanism composed of Z-score, IQR, and isolated forest methods is used to independently detect outliers in the original drilling parameters. A voting mechanism is implemented, and if the majority of methods determine that an outlier is an outlier, it is confirmed as an outlier and removed. Linear interpolation is performed on the missing positions. At the same time, non-representative data segments within 0.5 meters at the beginning and end of the borehole are removed. The retained valid data is standardized by Z-score, and finally, a borehole time series dataset is formed.

[0021] In this specification, the specific execution process of the multi-stage physical inversion model described in S2 is as follows: First, the original drilling parameters in the borehole time series dataset are converted into force and torque, and the mechanical specific energy is calculated; then, the uniaxial compressive strength of the rock is inverted from the mechanical specific energy through empirical relationships calibrated by field rock sample experiments; the geological strength index (GSI) is determined by combining the geological sketch of the working face or the statistical characteristics of the original drilling parameters, and the uniaxial compressive strength of the rock obtained by inversion is substituted into the generalized Hoek-Brown criterion to calculate the rock mass strength parameters; finally, under the stress level of engineering interest, the Hoek-Brown strength envelope is linearly fitted and equivalently transformed into cohesion and internal friction angle, while the elastic modulus is estimated by combining the empirical relationship between the rock mass deformation modulus and the uniaxial compressive strength.

[0022] In this specification, the specific process of the hybrid spatial interpolation algorithm described in S4 is as follows: First, the elastic modulus, cohesion, and internal friction angle in the equivalent Mohr-Coulomb mechanical parameters are subjected to Box-Cox transformation to obtain Gaussian spatial variables that follow or are close to a standard normal distribution; then, based on the data of the Gaussian spatial variables at the borehole location, the experimental variogram is calculated and fitted to obtain a variogram model describing the spatial autocorrelation of the parameters; next, based on the variogram model, a conditional random field is generated on the unsampled grid nodes in the three-dimensional tunnel model domain; Hoffman correction is applied to the generated initial simulation field to force the corrected field to be strictly equal to the observed values ​​of the equivalent Mohr-Coulomb mechanical parameters at all known borehole points, while maintaining the spatial structure of the initial field; finally, the corrected Gaussian field is inversely transformed to restore the three-dimensional rock mass mechanical parameter spatial distribution field.

[0023] In this specification, the assignment of the three-dimensional rock mass mechanical parameter spatial distribution field in S5 adopts the center point mapping method: calculate the center coordinates of each unit of the tunnel numerical model, map them to the three-dimensional rock mass mechanical parameter spatial distribution field, obtain the elastic modulus, cohesion and internal friction angle values ​​at the center point through the hybrid spatial interpolation algorithm, and assign them to the corresponding tunnel numerical model unit.

[0024] In this specification, when constructing a refined heterogeneous numerical model in S5, the spatial extrapolation range of the three-dimensional rock mass mechanical parameter spatial distribution field is determined based on sensitivity analysis, and the optimal range is twice the tunnel diameter (D).

[0025] In this specification, the actual monitoring data mentioned in S7 is multi-source fusion monitoring data, including surrounding rock displacement data obtained through displacement convergence meters and surrounding rock stress data obtained through stress sensors. After time-series alignment processing of the two types of data, a weight allocation algorithm based on the reliability of monitoring data is used to assign corresponding weights to different data sources, and the data is fused to form a unified monitoring dataset. At the same time, combined with the real-time progress of tunnel excavation, the deformation prediction results of the corresponding mileage are dynamically matched with the unified monitoring dataset to achieve real-time synchronization of "excavation-monitoring-assessment", thereby improving the comprehensiveness and accuracy of quantitative assessment.

[0026] In this specification, the specific method for correcting and optimizing the parameters of the multi-stage physical inversion model described in S8 is as follows: calculate the prediction error between the actual monitoring data and the deformation prediction results, and use the parameter sensitivity back analysis method based on error analysis to fine-tune the regression coefficients and geological strength index (GSI) values ​​of the rock uniaxial compressive strength inversion model, thereby reducing the error of the next prediction.

[0027] In this specification, when estimating the elastic modulus, the relevant parameters are taken within the empirical range based on the integrity of the rock mass and verified by the confining pressure correction formula. At the same time, constraints are applied based on rock mechanics theory to ensure that the values ​​of elastic modulus, cohesion and internal friction angle are consistent with the global rock mechanics database.

[0028] In summary, the present invention has at least the following beneficial effects:

[0029] 1) This invention pioneers a multi-stage physical inversion path of "MSE→UCS→HB→equivalent MC," endowing high-dimensional and noisy drilling signals with clear rock mechanics connotations, and outputting... E , c and φ The meaning of the parametric physical fields is clear and their internal consistency is consistent, fundamentally eliminating the uninterpretability of the "black box" model;

[0030] 2) The proposed SGS+Hoffman hybrid spatial interpolation algorithm creatively integrates the spatial variability characterization capability of geostatistics with the ability to strictly adhere to borehole "hard data", achieving high-fidelity, non-smooth three-dimensional parameter field reconstruction, effectively overcoming the smoothing effect of traditional interpolation methods and the uncertainty of pure random simulation.

[0031] 3) The refined numerical model based on the non-uniform parameter field can more realistically simulate the mechanical response of tunnel excavation under complex geological conditions. Engineering verification shows that, compared with the traditional homogeneous model, this method significantly reduces the average relative error of the surrounding rock deformation prediction and can accurately capture and interpret key phenomena such as asymmetric deformation and local stress concentration observed in the field.

[0032] 4) By introducing a closed-loop optimization mechanism of “prediction-verification-feedback”, the method has the characteristics of self-learning and self-optimization, and can dynamically track and adapt to the changes in surrounding rock conditions along the tunnel axis, which significantly improves the long-term applicability and reliability of the model in long-distance tunnel engineering.

[0033] 5) It provides a complete digital solution from real-time on-site perception to intelligent mechanism analysis, and then to high-fidelity numerical simulation and dynamic decision support, providing a powerful core analysis engine for the intelligent design, construction and safety management of tunnel engineering. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a schematic diagram of the refined intelligent modeling and deformation prediction method for tunnel surrounding rock oriented towards the construction process involved in this invention.

[0036] Figure 2 This is a schematic diagram of the modules included in an embodiment of the present invention.

[0037] Figure 3 This is a schematic diagram of the multi-stage physics-energy inversion process in an embodiment of the present invention.

[0038] Figure 4 This is a schematic diagram of the SGS+Hoffman hybrid spatial interpolation algorithm in an embodiment of the present invention.

[0039] Figure 5 This is a schematic diagram of the fine reconstruction of the three-dimensional model based on the elastic modulus in an embodiment of the present invention.

[0040] Figure 6 This is a schematic diagram of the cohesive force-based fine reconstruction of the three-dimensional model in an embodiment of the present invention.

[0041] Figure 7 This is a schematic diagram of the fine reconstruction of the three-dimensional model of the internal friction angle in an embodiment of the present invention.

[0042] Figure 8 This is a schematic diagram of on-site displacement monitoring in an embodiment of the present invention.

[0043] Figure 9 This is a schematic diagram of the 3C cyclic cross-section model in an embodiment of the present invention.

[0044] Figure 10 This is a schematic diagram of the 8C cyclic cross-section model in an embodiment of the present invention.

[0045] Figure 11 This is a schematic diagram of the 13C cyclic cross-section model in an embodiment of the present invention. Detailed Implementation

[0046] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0047] The following disclosure provides many different implementations or examples for carrying out different structures of the embodiments of the present invention. To simplify the disclosure of the embodiments of the present invention, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. Furthermore, reference numerals and / or reference letters may be repeated in different examples of the embodiments of the present invention; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various implementations and / or arrangements discussed.

[0048] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0049] like Figure 1 As shown, this embodiment provides a refined intelligent modeling and deformation prediction method for tunnel surrounding rock oriented towards the construction process, including:

[0050] S1: During the drilling process at the tunnel face, raw drilling parameters are collected in real time, including drilling speed, propulsion pressure, impact pressure, and rotation pressure; the raw drilling parameters are systematically cleaned, outlier removed, and standardized preprocessed to form a drilling time series dataset;

[0051] S2: Based on the borehole time series dataset, combined with the mechanical specific energy theory and the Hoek-Brown rock strength criterion, a multi-stage physical inversion model is constructed to resolve the original drilling parameters into equivalent Mohr-Coulomb mechanical parameters with clear physical meaning.

[0052] S3: Obtain the equivalent Mohr-Coulomb mechanical parameters corresponding to each borehole spatial location, including elastic modulus, cohesion, and internal friction angle;

[0053] S4: A hybrid spatial interpolation algorithm combining sequential Gaussian simulation and Hoffman correction is adopted. With equivalent Mohr-Coulomb mechanical parameters as spatial constraints, conditional simulation and iterative correction are performed in the three-dimensional tunnel model domain to generate a three-dimensional spatial distribution field of rock mass mechanical parameters.

[0054] S5: Establish a tunnel numerical model, assign the three-dimensional rock mass mechanical parameter spatial distribution field to the tunnel numerical model, and construct a refined heterogeneous numerical model that can reflect the spatial variability of the surrounding rock.

[0055] S6: On a refined heterogeneous numerical model, simulate the entire tunnel excavation process, calculate and output the deformation prediction results of the surrounding rock under excavation disturbance, including crown settlement, perimeter convergence and asymmetric deformation mode.

[0056] In some embodiments, the method for refined intelligent modeling and deformation prediction of tunnel surrounding rock for the construction process further includes:

[0057] S7: Compare the deformation prediction results with the actual monitoring data obtained at the corresponding mileage of the tunnel site to quantitatively evaluate the prediction accuracy of the refined heterogeneous numerical model and obtain the quantitative evaluation results.

[0058] S8: Based on the quantitative evaluation results, the parameters of the multi-stage physical inversion model are corrected and optimized, and the optimized multi-stage physical inversion model is applied to the next prediction, forming a closed-loop iterative process of data acquisition-inversion-modeling-prediction-verification-optimization.

[0059] In some embodiments, the systematic cleaning, outlier removal, and standardization preprocessing described in S1 are performed as follows: a multi-algorithm consensus mechanism consisting of Z-score, IQR, and isolated forest methods is used to independently detect outliers in the original drilling parameters. A voting mechanism is implemented, and if the majority of methods determine that an outlier is an outlier, it is confirmed as an outlier and removed. Linear interpolation is performed on the missing positions. At the same time, non-representative data segments within 0.5 meters at the beginning and end of the borehole are removed. The retained valid data is standardized by Z-score, and finally a borehole time series dataset is formed.

[0060] In some embodiments, the specific execution process of the multi-stage physical inversion model described in S2 is as follows: First, the original drilling parameters in the borehole time series dataset are converted into force and torque, and the mechanical specific energy is calculated; then, the uniaxial compressive strength of the rock is inverted from the mechanical specific energy through the empirical relationship calibrated by the field rock sample test; the geological strength index (GSI) is determined by combining the geological sketch of the working face or the statistical characteristics of the original drilling parameters, and the uniaxial compressive strength of the rock obtained by inversion is substituted into the generalized Hoek-Brown criterion to calculate the rock mass strength parameters; finally, under the stress level of engineering interest, the Hoek-Brown strength envelope is linearly fitted and equivalently transformed into cohesion and internal friction angle, while the elastic modulus is estimated by combining the empirical relationship between the rock mass deformation modulus and the uniaxial compressive strength.

[0061] In some embodiments, the specific process of the hybrid spatial interpolation algorithm in S4 is as follows: First, the elastic modulus, cohesion, and internal friction angle in the equivalent Mohr-Coulomb mechanical parameters are subjected to Box-Cox transformation to obtain Gaussian spatial variables that follow or are close to a standard normal distribution; then, based on the data of the Gaussian spatial variables at the borehole location, the experimental variogram is calculated and fitted to obtain a variogram model describing the spatial autocorrelation of the parameters; next, based on the variogram model, a conditional random field is generated on the unsampled grid nodes in the three-dimensional tunnel model domain; Hoffman correction is applied to the generated initial simulation field to force the corrected field to be strictly equal to the observed values ​​of the equivalent Mohr-Coulomb mechanical parameters at all known borehole points, while maintaining the spatial structure of the initial field; finally, the corrected Gaussian field is inversely transformed to restore the three-dimensional rock mass mechanical parameter spatial distribution field.

[0062] In some embodiments, the assignment of the three-dimensional rock mass mechanical parameter spatial distribution field in S5 adopts the center point mapping method: calculate the center coordinates of each unit of the tunnel numerical model, map them to the three-dimensional rock mass mechanical parameter spatial distribution field, obtain the elastic modulus, cohesion and internal friction angle values ​​at the center point through the hybrid spatial interpolation algorithm, and assign them to the corresponding tunnel numerical model unit.

[0063] In some embodiments, when constructing a refined heterogeneous numerical model in S5, the spatial extrapolation range of the three-dimensional rock mass mechanical parameter spatial distribution field is determined based on sensitivity analysis, and the optimal range is twice the tunnel diameter (D).

[0064] In some embodiments, the actual monitoring data mentioned in S7 is multi-source fusion monitoring data, including surrounding rock displacement data obtained by displacement convergence meter and surrounding rock stress data obtained by stress sensor. After time-series alignment processing of the two types of data, a weight allocation algorithm based on the reliability of monitoring data is used to assign corresponding weights to different data sources, and the data is fused to form a unified monitoring dataset. At the same time, combined with the real-time progress of tunnel excavation, the deformation prediction results of the corresponding mileage are dynamically matched with the unified monitoring dataset to achieve real-time synchronization of "excavation-monitoring-assessment" and improve the comprehensiveness and accuracy of quantitative assessment. Specifically, it includes: (1) Data source: surrounding rock displacement data (arch settlement, perimeter convergence value) obtained through displacement convergence meter, and surrounding rock stress data (radial stress, circumferential stress value) obtained through stress sensor; (2) Time alignment processing: taking the tunnel excavation mileage as the synchronization benchmark, the displacement data (collection frequency 1 time / hour) and stress data (collection frequency 1 time / 30 minutes) are uniformly interpolated to "one set of data corresponding to every 0.5 meters of excavation mileage", to ensure the time consistency of the two types of data under the same mileage; (3) Reliability assessment: calculate the reliability value of the two types of data respectively, displacement data reliability = 1 - data fluctuation coefficient (fluctuation coefficient = standard deviation / mean), stress data reliability = sensor calibration validity Coefficient × (1 - environmental interference coefficient), where the effective coefficient of sensor calibration is 0.8~1.0 based on the most recent calibration result, and the environmental interference coefficient is converted to 0~0.2 based on the humidity and dust concentration of the tunnel face according to the empirical formula; (4) Weight allocation: adopt the linear weighting algorithm, data weight = self-confidence value / (displacement data confidence value + stress data confidence value), and based on this weight, the two types of data are integrated to form a unified monitoring dataset; (5) Dynamic matching: through the real-time mileage positioning module of the tunnel boring machine, the unified monitoring dataset is bound to the deformation prediction results of the corresponding mileage one by one, so as to realize the real-time synchronization of "tunneling progress - monitoring data - evaluation process" and improve the comprehensiveness and accuracy of quantitative evaluation.

[0065] In some embodiments, the specific method for correcting and optimizing the parameters of the multi-stage physical inversion model described in S8 is as follows: calculate the prediction error between the actual monitoring data and the deformation prediction results, and use the parameter sensitivity back analysis method based on error analysis to fine-tune the regression coefficients and geological strength index (GSI) values ​​of the rock uniaxial compressive strength inversion model, thereby reducing the error of the next prediction.

[0066] In some embodiments, when estimating the elastic modulus, the relevant parameters are taken within an empirical range based on the integrity of the rock mass and verified by the confining pressure correction formula. At the same time, constraints are applied based on rock mechanics theory to ensure that the values ​​of elastic modulus, cohesion and internal friction angle are consistent with the global rock mechanics database.

[0067] The technical concept of this invention is as follows:

[0068] This paper describes in detail a long highway tunnel project in a complex mountainous area of ​​southwestern China. The tunnel's surrounding rock is mainly composed of tuff and granite, exhibiting significant spatial variation in rock mass integrity and traversing multiple local fault zones. This provides an ideal scenario for verifying the effectiveness of the method described in this invention under complex geological conditions. Figure 1 As shown, the method for refined intelligent modeling and deformation prediction of tunnel surrounding rock for the construction process includes the following steps:

[0069] S1: Drilling Parameter Acquisition and Preprocessing

[0070] A fully computerized three-arm drilling rig was used for full-face drilling of the tunnel. High-precision sensors were installed on the pipelines of the drill arm's propulsion, impact, and rotation mechanisms, with sampling points at 2cm depth intervals, to automatically collect and record the following four core drilling while-drilling parameters (MWD) for each borehole in real time:

[0071] Drilling rate (PR): Unit: m / min, reflects the ease with which the drill bit penetrates the rock mass;

[0072] Feed pressure (FP): Unit: bar, characterizing axial thrust;

[0073] Hammer Pressure (HP): Unit: bar, representing the energy required for impact fracturing;

[0074] Rotation pressure (RP): Unit: bar, characterizing cutting torque.

[0075] Each excavation cycle (approximately 3-5m advance) can generate tens of thousands to hundreds of thousands of raw data points. The construction log system synchronously records the tunnel mileage, station number, cycle number, and corresponding geological sketch information, achieving strict binding between data and spatial location.

[0076] The raw MWD data contains noise and outliers introduced by factors such as sensor malfunctions, hydraulic fluctuations, and rock disturbances during drilling start-up / stop phases. To ensure the accuracy of subsequent inversion, rigorous standardization preprocessing is necessary. This invention employs an automated cleaning process based on multi-algorithm consensus, and examples of the code implementation for its core steps are shown in Table 1.

[0077] Table 1. Code Examples for Automated Cleaning Implementation

[0078] ;

[0079] The technical implementation process of this automated cleaning process based on multi-algorithm consensus is described in detail below:

[0080] Parallel execution of anomaly detection algorithms: The system executes three independent anomaly detection algorithms in parallel to scan the original dataset.

[0081] Z-score normalization detection: Calculates the standard deviation multiple of each data point from the mean of its parameter sequence. Data points with an absolute value greater than a set threshold (usually 3) are marked as suspected outliers. This method is good at identifying global outliers far from the center of the overall distribution.

[0082] IQR (Interquartile Range) detection: Calculate the upper quartile (Q3) and lower quartile (Q1) of each parameter sequence, and define a reasonable upper bound (Q3 + 1.5 * IQR) and lower bound (Q1 - 1.5 * IQR). Data points outside this range are marked as suspected anomalies. This method is sensitive to the dispersion of the data itself and can effectively identify anomalies at the tail of the distribution.

[0083] Isolation Forest detection isolates sample points by randomly partitioning the feature space and calculating the path length required for each data point to be isolated. Points with significantly shorter path lengths are considered "easily isolated" outliers. This unsupervised learning method can capture complex, multivariate coupled local anomaly patterns.

[0084] Voting Consensus and Final Anomaly Determination: To avoid misjudgments by a single algorithm, a consensus mechanism based on majority voting is adopted. Each data point will receive three independent "anomaly" or "normal" labels from the three algorithms mentioned above. Only when a data point is consistently determined to be an anomaly by more than half (i.e., at least two) of the algorithms is it finally identified as an outlier requiring cleaning. This mechanism greatly enhances the robustness of anomaly detection and reduces false cleansing caused by the limitations of the algorithms themselves.

[0085] Outlier Handling and Data Reconstruction: All data points identified as outliers by the consensus mechanism are deleted. Subsequently, missing data points resulting from deletion are filled using linear interpolation to ensure the continuity of the data sequence in time or depth, forming a complete, clean, and high-quality borehole time-series dataset.

[0086] Through the above process, this invention achieves the processing of two typical anomalies in drilling parameters: Type I (non-representative data of low pressure and low velocity at the beginning / end of the borehole) and Type II (instantaneous extremely high / low values ​​caused by sensor errors). The data dispersion after processing is reduced by an average of about 30%-40%, forming a reliable and high-quality borehole time series dataset, laying a reliable data foundation for high-precision inversion.

[0087] S2: Multi-stage physics-energy inversion

[0088] like Figure 3As shown, this step aims to establish the constitutive parameters (elastic modulus under the Mohr-Coulomb criterion) required for numerical simulation from the operating parameters (MWD). E Cohesion c internal friction angle φ The rigorous physical channel is divided into four sub-stages.

[0089] S2.1: Calculation of mechanical specific energy

[0090] The drilling parameters after cleaning are converted into force and energy indicators. Axial thrust. Combating pressure and torque T The calculation formula is as follows:

[0091] ;

[0092] ;

[0093] ;

[0094] in, , , These are respectively the pressures of advancement, attack, and reversal; , This corresponds to the effective area of ​​the piston; Hydraulic motor displacement; , For impact and rotational mechanical efficiency ， Based on the model of the rock drilling rig and the rock type at the construction site, the values ​​are taken as 40% and 60%, respectively. Then, the mechanical specific energy (MSE), which is the total energy consumed to break a unit volume of rock, is calculated.

[0095] ;

[0096] in, This refers to the drilling speed; N This refers to the drill bit rotation speed; f The impact frequency; Energy for a single strike; This represents the cross-sectional area of ​​the drill bit.

[0097] S2.2: Uniaxial compressive strength inversion: Based on Teale specific energy principle and a large amount of field calibration data, a nonlinear regression model was established for UCS, MSE, and MWD parameters.

[0098] ;

[0099] In the formula, The uniaxial compressive strength of intact rock; α ,a , b , c , d The regression coefficients obtained through field rock sample experiments are constrained by the basic quality index BQ of the rock mass, thus achieving adaptive calibration for rock masses of different qualities.

[0100] S2.3: Geological Intensity Index and Hoek-Brown Parameter Determination. The geological intensity index (GSI) is determined based on the geological sketch of the working face or the statistical characteristics of drilling parameters. Then, it is combined with the UCS obtained from step S2.2 (…). ), calculate rock mass strength parameters according to the generalized Hoek-Brown criterion , s , β :

[0101] ;

[0102] ;

[0103] ;

[0104] In the formula, Rock mass material constants are parameters that reflect the shape of the failure envelope of a rock mass. s Rock mass material constant is a parameter that reflects the ratio of the strength of the rock mass to the strength of intact rock. β Rock mass material constants are parameters that control the curvature of the failure envelope. The Hoek-Brown constant for intact rock. D This represents the rock mass disturbance factor.

[0105] S2.4: Finally, the stress level within the influence range of tunnel excavation ( Under these conditions, the Hoek-Brown intensity envelope is linearly fitted and equivalently transformed into Mohr-Coulomb intensity parameters. c and φ Simultaneously determine the deformation modulus of the rock mass. E The conversion formula is:

[0106] ;

[0107] ;

[0108] ;

[0109] In the formula This represents the minimum principal stress under the Hoek-Brown strength criterion; meanwhile, the elastic modulus E is estimated using the empirical relationship between the rock mass deformation modulus and the UCS.

[0110] ;

[0111] Among them, parameters γ and n Values ​​are taken within an empirical range based on the rock mass integrity (intact rock or jointed rock mass), and verified using the confining pressure correction formula to ensure... E The reasonableness of the value.

[0112] This completes the process of evolving from MWD to a comprehensive set of surrounding rock mechanical parameters with clear physical meaning. E , c , φ Intelligent parsing of ).

[0113] S3: Output of mechanical parameters at the drilling point

[0114] Step S2 performs analytical calculations at each sampling point of the borehole trajectory. This step binds the location of each borehole point (spatial coordinates recorded by the acquisition system) to its corresponding inversion result, outputting a set of discrete "hard data" with clearly defined spatial locations. Each data point contains a complete triplet mechanical parameter: elastic modulus. E Cohesion c and internal friction angle φ These data points accurately characterize the mechanical properties of the local surrounding rock traversed by the borehole, providing the most critical constraints for the next step of spatial interpolation.

[0115] S4: Hybrid spatial interpolation algorithm for constructing heterogeneous parameter fields

[0116] This step is the core innovation that enables the leap from "discrete points" to "continuous fields." For example... Figure 4 As shown, in order to construct a high-fidelity, non-uniform three-dimensional parametric field, this invention adopts a hybrid algorithm that combines sequential Gaussian simulation (SGS) with Hoffman correction.

[0117] S4.1: Mechanical parameters of the drilling point output in step S3 ( E , c , φ Perform Box-Cox transformations on each variable to make them conform to or approximate a standard normal distribution, thus obtaining the Gaussian spatial variable Z( x To meet the prerequisites of SGS.

[0118] S4.2: Based on Gaussian space variable Z ( x ) Calculate and fit the experimental variation function based on the data at the borehole location. γ ( h ), thus obtaining the variogram model describing the autocorrelation of the parameter space, whose expression is:

[0119] ;

[0120] in, Gold nugget value (characterizing microscale variation and measurement error), sill value +C (representing total spatial variability), range a (Characterizing the spatially related range); spatial spacing h (Characterizing spatial correlation as it varies with distance). This process generates a spatial random field that conforms to a preset variogram function (e.g., exponential type, correlation distance 25m), reflecting the spatial correlation and uncertainty of the parameters.

[0121] S4.3: Based on the variogram model established in step S4.2, a conditional random field is generated on all unsampled mesh nodes within the three-dimensional tunnel model domain using a sequential Gaussian simulation method. This process ensures that the simulated field conforms to the preset spatial statistical structure, and code examples are shown in Table 2.

[0122] Table 2. Code examples for implementing Conditional Random Fields

[0123] ;

[0124] Within the transformed Gaussian space, SGS is performed to construct the initial three-dimensional parameter field. This process is a sequential, conditional stochastic simulation, with the following specific steps:

[0125] A completely random access order is generated for all grid nodes to be simulated in the 3D model. This randomness ensures the statistical stability of the simulation results and avoids systematic biases introduced by artificial paths. Each grid node is visited sequentially according to this random order. For the current node: its K nearest neighbors are searched and determined within the spatial range (usually based on the correlation distance defined by the variogram model). These neighbors include "hard data" points (drill points) that have already been assigned values, as well as grid nodes that have been visited and assigned values ​​in this round of simulation. Based on the known Gaussian values ​​of these neighbors, spatial interpolation is performed using the Ordinary Kriging method. The Kriging method not only provides the conditional expectation (mean) of the current node but also its Kriging variance. This variance quantifies the uncertainty of the current point's value given the known neighborhood data. A value is randomly drawn from the conditional Gaussian distribution defined with the Kriging mean as the expectation and the Kriging variance as the standard deviation, and assigned to the current grid node. This step organically combines spatial correlation (reflected by Kriging) with inherent randomness (reflected by random sampling). Repeat step 2 until all grid nodes have been visited and assigned values. Finally, a Gaussian random field is generated that statistically conforms to the preset spatial structure (variance function) and passes through all "hard data" points.

[0126] S4.4: Hoffman iterative hard data correction for the initial simulation field generated by SGS. Hoffman correction is applied. While the SGS simulated field is statistically reasonable, it may deviate from the actual measurements at the borehole. Therefore, Hoffman correction is introduced for iterative forced matching. By solving an optimization problem with hard data constraints (minimizing the L2 norm of the overall field correction), the initial field is iteratively adjusted, resulting in a field with forced correction. At all known borehole points The value is strictly equal to the observed value. At the same time, to preserve the spatial structure of the initial field as much as possible, the optimization model for the correction field is as follows:

[0127] ;

[0128] Constraints:

[0129] ;

[0130] After correction, the final Gaussian field is obtained, which is consistent with both the statistical structure and the hard data points.

[0131] The specific implementation code for this part is shown in Table 3.

[0132] Table 3. Code Examples for Hoffman Correction Implementation

[0133] ;

[0134] While the initial field generated by SGS is statistically reasonable, its simulated values ​​at "hard data" points may slightly deviate from the measured values. To address this, Hoffman correction is introduced for iterative optimization: the initial simulated field is iterated multiple times (typically 3 to 5 times). In each iteration, all simulated grid nodes (excluding "hard data" points) are traversed, and the value of each node is updated to the arithmetic mean of the values ​​of all its neighboring nodes (including other simulated points and "hard data" points) within a certain spatial range. This operation acts as a diffusion process, effectively enhancing the spatial continuity of the parametric field and suppressing unreasonable and drastic fluctuations in local areas. At the smooth beginning and end of each iteration, a core constraint is enforced: the field values ​​of all "hard data" points (i.e., borehole locations) are forcibly reset to their original true measured values ​​(i.e., values ​​after Gaussian transformation). This step ensures that the entire optimization process is anchored to the measured data, which is crucial for the algorithm to strictly match the borehole data.

[0135] After multiple iterations, the simulation field maintains good spatial continuity while strictly matching all borehole "hard data," effectively overcoming the oversmoothing problem of the traditional Kriging method, and can accurately characterize local heterogeneous bodies such as weak interlayers.

[0136] S4.5: Inverse Transformation: For the corrected Gaussian field Perform the inverse transformation of step S4.1 to restore it to the mechanical parameter field in physical space, thus obtaining a high-fidelity three-dimensional non-uniform elastic modulus field. E Field), cohesion field ( c field) and internal friction angle field ( φ (Field). Examples of results for constructing heterogeneous three-dimensional mechanical parameter fields using a typical tunnel surrounding rock mixed space interpolation algorithm are shown in Table 4.

[0137] Table 4. Examples of heterogeneous three-dimensional mechanical parameter fields constructed using a typical tunnel surrounding rock hybrid spatial interpolation algorithm.

[0138] ;

[0139] S5: Refined Numerical Model Construction

[0140] This step transforms the aforementioned parameter field into a computable numerical model. The three-dimensional mechanical parameter field generated in step S4 is imported into the tunnel numerical model built using the "center point method" (assigning parameter values ​​to the material group corresponding to the center point of the mesh element). For example... Figure 5 , Figure 6 and Figure 7 As shown, the constructed model strictly adheres to the actual tunnel design dimensions (horseshoe cross-section, net width 11.75m, net height 5.70m) and includes detailed initial support (shotcrete, steel arch frame, anchor bolts) and secondary lining structure. The surrounding rock constitutive model adopts the Mohr-Coulomb criterion. In this way, a refined heterogeneous numerical model that can realistically reflect the spatial variability of the mechanical properties of the surrounding rock is constructed, replacing the homogeneous parameter assumption model used in traditional methods.

[0141] S6: Simulation and Deformation Prediction of the Entire Tunnel Excavation Process

[0142] On the refined heterogeneous numerical model constructed in step S5, the tunnel excavation process is dynamically simulated using the "element birth and death method," and the corresponding support structure is activated. Running numerical calculations, the model outputs the detailed mechanical response of the tunnel surrounding rock under excavation disturbance. The deformation prediction results mainly include: the quantified values ​​of crown settlement and peripheral convergence (horizontal displacement), as well as the crucial asymmetric deformation modes (such as differential settlement between left and right arch shoulders, local bulging, etc.). This embodiment compares the mechanical responses of the model under different extrapolation ranges (1.0D, 1.5D, 2.0D, 2.5D, where D is the tunnel diameter). It finds that when the extrapolation range is 2.0D, the model can completely cover the main excavation disturbance area, and its predicted surrounding rock displacement field and lining axial force distribution exhibit continuous and gradual characteristics, avoiding insufficient load transfer in smaller areas or non-physical stress / displacement abrupt changes (local axial force abrupt changes exceeding 15%) in larger areas (2.5D). Therefore, 2.0 times the tunnel diameter is determined as the recommended extrapolation range for this method under similar geological conditions. These results are difficult to accurately predict using traditional homogeneous models, while this model, thanks to the heterogeneous parameter field, can more realistically reflect the mechanical behavior caused by geological inhomogeneity.

[0143] S7: Engineering Verification

[0144] To quantitatively evaluate the prediction accuracy of the method of this invention, the deformation prediction results obtained in step S6 (such as the displacement values ​​of key points like the arch crown, arch waist, and arch foot of a specific cross-section) are compared point by point with the prediction results of a traditional homogeneous model using the average value of surrounding rock parameters, as well as the actual monitoring data (convergence meter monitoring results) deployed at the corresponding mileage of the tunnel site. Figure 8 , Figure 9 , Figure 10 and Figure 11 As shown, quantitative evaluation is performed by calculating indicators such as average relative error and root mean square error. In this embodiment, verification based on three typical cross-sections of the tunnel (3C, 8C, and 13C cyclic cross-sections) shows that the average prediction error of the refined model of this invention is significantly lower than that of the traditional homogeneous model used for comparison. The model also accurately captures the local axial force and bending moment concentration areas of the secondary lining under heterogeneous surrounding rock, providing precise guidance for targeted support reinforcement. Furthermore, it highly matches the field monitoring results, further demonstrating the effectiveness of this invention in improving deformation prediction accuracy.

[0145] S8: Model Optimization and Closed-Loop Feedback

[0146] This invention is not merely a one-time prediction tool, but a continuously learning system. Based on the verification results of step S7, if a systematic bias is found in the prediction of a specific section (such as a certain type of surrounding rock), the optimization process is initiated. Specifically, returning to step S2, the newly accumulated MWD data with verification results is used to correct and optimize the regression coefficients in the UCS inversion model or the empirical parameters in the HB-MC conversion. The optimized inversion model will be used for prediction in subsequent new tunneling cycles. Thus, a closed-loop iterative process of "data acquisition (S1) - inversion (S2 / S3) - modeling (S4 / S5) - prediction (S6) - verification (S7) - optimization (S8)" is formed. This process enables the system to adapt to the constantly changing geological conditions ahead of the tunnel, achieving continuous evolution of prediction capabilities and providing long-term, reliable technical support for tunnel dynamic design and construction.

[0147] This embodiment also provides a system for refined modeling and deformation prediction of tunnel surrounding rock that implements the above method, such as... Figure 2 As shown, it includes functional modules corresponding to steps S1-S8: drilling parameter acquisition and processing module 110, drilling parameter multi-stage physical-energy inversion module 120, borehole point mechanical parameter output module 130, multi-parameter heterogeneous field construction module 140, Flac3D refined numerical modeling module 150, tunnel excavation simulation and deformation prediction module 160, engineering verification module 170, and model optimization and closed-loop feedback module 180.

[0148] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-described methods for refined modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process, as described in S1 to S8.

[0149] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described methods for refined modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process, as described in S1 to S8.

[0150] The embodiments described above are for illustrative purposes only and are not intended to limit the invention. Therefore, any changes in numerical values ​​or substitutions of equivalent elements should still fall within the scope of this invention.

[0151] The above detailed description will enable those skilled in the art to understand that the present invention can indeed achieve the aforementioned objectives and has complied with the provisions of the Patent Law.

[0152] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the invention. The above descriptions are merely preferred embodiments of the invention and are not intended to limit the invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

[0153] It should be noted that the above description of the process is for illustrative purposes only and does not limit the scope of this specification. Those skilled in the art can make various modifications and changes to the process under the guidance of this specification. However, these modifications and changes remain within the scope of this specification.

[0154] The basic concepts have been described above. Obviously, for those skilled in the art who have read this application, the above disclosure is merely illustrative and does not constitute a limitation of this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and corrections to this application. Such modifications, improvements, and corrections are suggested in this application, and therefore, such modifications, improvements, and corrections still fall within the spirit and scope of the exemplary embodiments of this application.

[0155] Furthermore, this application uses specific terms to describe its embodiments. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "an embodiment," "one embodiment," or "an alternative embodiment" mentioned twice or more in different positions in this specification do not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application can be appropriately combined.

[0156] Furthermore, those skilled in the art will understand that aspects of this application can be described and illustrated through several patentable types or situations, including any new and useful combination of processes, machines, products, or substances, or any new and useful improvements thereof. Therefore, aspects of this application can be implemented entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or a combination of hardware and software. All of the above hardware or software can be referred to as a “unit,” “module,” or “system.” Furthermore, aspects of this application can take the form of a computer program product embodied in one or more computer-readable media, wherein computer-readable program code is contained therein.

[0157] The computer program code required for the operation of each part of this application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, and Python; general programming languages ​​such as C; Visual Basic, Fortran2103, Perl, COBOL2102, PHP, and ABAP; dynamic programming languages ​​such as Python, Ruby, and Groovy; or other programming languages. This program code can run entirely on the user's computer, or as a standalone software package on the user's computer, or partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer via any network, such as a local area network (LAN) or wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as Software as a Service (SaaS).

[0158] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this application are not intended to limit the order of the processes and methods of this application. Although some currently considered useful embodiments of the invention have been discussed in the foregoing disclosure by way of various examples, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the substance and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a purely software solution, such as an installation on an existing server or mobile device.

[0159] Similarly, it should be noted that, in order to simplify the description of the present application and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of the embodiments of the present application sometimes combines multiple features into a single embodiment, drawing, or description thereof. However, this approach of the present application should not be construed as reflecting an intention that the claimed subject matter requires more features than expressly recited in each claim. Rather, the subject of the invention should possess fewer features than in any single embodiment described above.

Claims

1. A refined intelligent modeling and deformation prediction method for tunnel surrounding rock oriented towards the construction process, characterized in that, include: S1: During the drilling process at the tunnel face, raw drilling parameters are collected in real time, including drilling speed, propulsion pressure, impact pressure, and rotation pressure; the raw drilling parameters are systematically cleaned, outlier removed, and standardized preprocessed to form a drilling time series dataset; S2: Based on the borehole time series dataset, combined with the mechanical specific energy theory and the Hoek-Brown rock strength criterion, a multi-stage physical inversion model is constructed to resolve the original drilling parameters into equivalent Mohr-Coulomb mechanical parameters with clear physical meaning. The specific execution process of the multi-stage physical inversion model is as follows: First, the original drilling parameters in the borehole time series dataset are converted into force and torque, and the mechanical specific energy is calculated; then, the uniaxial compressive strength of the rock is inverted from the mechanical specific energy through the empirical relationship calibrated by the field rock sample test; the geological strength index is determined by combining the geological sketch of the working face or the statistical characteristics of the original drilling parameters, and the uniaxial compressive strength of the rock obtained by inversion is substituted into the generalized Hoek-Brown criterion to calculate the rock mass strength parameters; finally, under the stress level of engineering concern, the Hoek-Brown strength envelope is linearly fitted and equivalently transformed into cohesion and internal friction angle, and the elastic modulus is estimated by combining the empirical relationship between the rock mass deformation modulus and the uniaxial compressive strength. S3: Obtain the equivalent Mohr-Coulomb mechanical parameters corresponding to each borehole spatial location, including elastic modulus, cohesion, and internal friction angle; S4: A hybrid spatial interpolation algorithm combining sequential Gaussian simulation and Hoffman correction is adopted. With equivalent Mohr-Coulomb mechanical parameters as spatial constraints, conditional simulation and iterative correction are performed in the three-dimensional tunnel model domain to generate a three-dimensional spatial distribution field of rock mass mechanical parameters. S5: Establish a tunnel numerical model, assign the three-dimensional rock mass mechanical parameter spatial distribution field to the tunnel numerical model, and construct a refined heterogeneous numerical model that can reflect the spatial variability of the surrounding rock. S6: On a refined heterogeneous numerical model, simulate the entire tunnel excavation process, calculate and output the deformation prediction results of the surrounding rock under excavation disturbance, including crown settlement, perimeter convergence and asymmetric deformation mode.

2. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, Also includes: S7: Compare the deformation prediction results with the actual monitoring data obtained at the corresponding mileage of the tunnel site to quantitatively evaluate the prediction accuracy of the refined heterogeneous numerical model and obtain the quantitative evaluation results. S8: Based on the quantitative evaluation results, the parameters of the multi-stage physical inversion model are corrected and optimized, and the optimized multi-stage physical inversion model is applied to the next prediction, forming a closed-loop iterative process of data acquisition-inversion-modeling-prediction-verification-optimization.

3. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, The specific methods for systematic cleaning, outlier removal, and standardized preprocessing described in S1 are as follows: A multi-algorithm consensus mechanism composed of Z-score, IQR, and isolated forest methods is used to independently detect outliers in the original drilling parameters. A voting mechanism is implemented, and if the majority of methods determine that an outlier is an outlier, it is confirmed as an outlier and removed. Linear interpolation is performed on the missing positions. At the same time, non-representative data segments within 0.5 meters at the beginning and end of the borehole are removed. The retained valid data is standardized by Z-score, and finally, a borehole time series dataset is formed.

4. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, The specific process of the hybrid spatial interpolation algorithm described in S4 is as follows: First, the elastic modulus, cohesion, and internal friction angle in the equivalent Mohr-Coulomb mechanical parameters are subjected to Box-Cox transformation to obtain Gaussian spatial variables that follow or are close to a standard normal distribution; then, based on the data of the Gaussian spatial variables at the borehole location, the experimental variogram is calculated and fitted to obtain a variogram model describing the spatial autocorrelation of the parameters; next, based on the variogram model, a conditional random field is generated on the unsampled grid nodes in the three-dimensional tunnel model domain; Hoffman correction is applied to the generated initial simulation field to force the corrected field to be strictly equal to the observed values ​​of the equivalent Mohr-Coulomb mechanical parameters at all known borehole points, while maintaining the spatial structure of the initial field; finally, the corrected Gaussian field is inversely transformed to restore the three-dimensional rock mass mechanical parameter spatial distribution field.

5. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, The assignment of the three-dimensional rock mass mechanical parameter spatial distribution field in S5 adopts the center point mapping method: calculate the center coordinates of each unit of the tunnel numerical model, map them to the three-dimensional rock mass mechanical parameter spatial distribution field, obtain the elastic modulus, cohesion and internal friction angle values ​​at the center point through the hybrid spatial interpolation algorithm, and assign them to the corresponding tunnel numerical model unit.

6. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, When constructing a refined heterogeneous numerical model in S5, the spatial extrapolation range of the three-dimensional rock mass mechanical parameter spatial distribution field is determined based on sensitivity analysis, and the optimal range is twice the tunnel diameter.

7. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 2, characterized in that, The specific method for correcting and optimizing the parameters of the multi-stage physical inversion model described in S8 is as follows: calculate the prediction error between the actual monitoring data and the deformation prediction results, and use the parameter sensitivity back analysis method based on error analysis to fine-tune the regression coefficients and geological strength index values ​​of the rock uniaxial compressive strength inversion model, thereby reducing the error of the next prediction.

8. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 1, characterized in that, When estimating the elastic modulus, the relevant parameters are taken within the empirical range based on the integrity of the rock mass and verified by the confining pressure correction formula. At the same time, constraints are applied based on rock mechanics theory to ensure that the values ​​of elastic modulus, cohesion and internal friction angle are consistent with the global rock mechanics database.

9. The method for refined intelligent modeling and deformation prediction of tunnel surrounding rock oriented towards the construction process as described in claim 2, characterized in that, The actual monitoring data described in S7 is multi-source fusion monitoring data, including surrounding rock displacement data obtained through displacement convergence meters and surrounding rock stress data obtained through stress sensors. After time-series alignment processing of the two types of data, a weight allocation algorithm based on the reliability of monitoring data is used to assign corresponding weights to different data sources, and the data is fused to form a unified monitoring dataset. At the same time, combined with the real-time progress of tunnel excavation, the deformation prediction results of the corresponding mileage are dynamically matched with the unified monitoring dataset to achieve real-time synchronization of excavation-monitoring-assessment, thereby improving the comprehensiveness and accuracy of quantitative assessment.

Images