Method and system for determining inherent frequency of tunnel surrounding rock based on numerical samples

Abstract

The invention provides a tunnel surrounding rock inherent frequency determination method and system based on a numerical sample, and relates to the technical field of tunnel engineering, and the method comprises the steps: obtaining geological survey data of target deep-buried jointed rock tunnel engineering; performing parameter model establishment and parameter setting by combining a three-dimensional numerical inversion method according to the geological survey data to obtain parameterized model configuration; performing numerical model construction processing according to the parameterized model configuration, and generating an inherent frequency numerical sample data set; sensitivity analysis is carried out according to the inherent frequency numerical value sample data set, and a main control factor screening result is obtained; establishing a prediction model according to a main control factor screening result to obtain an inherent frequency prediction model; and inputting real-time main control factor data obtained by field monitoring into the inherent frequency prediction model for prediction processing to obtain inherent frequency time-frequency characteristics and main frequency information of the target deeply-buried jointed rock tunnel. The construction safety and the engineering decision-making efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering technology, and more specifically, to a method and system for determining the natural frequency of tunnel surrounding rock based on numerical samples. Background Technology

[0002] As tunnel construction in my country continues to extend deeper, the stress environment of deeply buried tunnels is highly complex and dynamically changing due to multiple factors such as geological structure and excavation unloading, posing a severe challenge to the safety and stability of engineering construction. Among these challenges, unfavorable combinations of structural planes, represented by jointed fracture zones, have become key control factors for the stability of hard rock surroundings. Under the combined effects of high ground stress and construction disturbances, deeply buried jointed rock tunnels face numerous engineering geological hazard risks. Various external dynamic disturbance sources during construction (such as full-face tunnel boring machine (TBM) vibration, blasting vibration, seismic waves, and train vibration) are transmitted into the surrounding rock. If their disturbance frequencies are close to the natural frequencies of the surrounding rock, a significant resonance effect will be triggered. This resonance phenomenon accelerates the initiation and propagation of cracks in the jointed rock mass, thereby inducing rockburst disasters, causing casualties and equipment damage. Therefore, it is urgent to establish a rapid method for obtaining the natural frequencies of deeply buried jointed rock tunnels to lay the foundation for the prevention and control of rockbursts triggered by dynamic disturbances and ensure the safety of on-site personnel. Existing methods for obtaining natural frequencies mainly rely on field tests or numerical simulation calculations. While field tests are direct and reliable, they are time-consuming, costly, and greatly affected by environmental factors; while numerical simulations are more accurate, they require the construction of detailed geometric models and the setting of material parameters, making the calculation process complex and time-consuming.

[0003] Based on the shortcomings of the existing technologies, there is an urgent need for a method and system for determining the natural frequency of tunnel surrounding rock based on numerical samples. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for determining the natural frequencies of tunnel surrounding rock based on numerical samples, aiming to overcome the shortcomings of the prior art and provide a deployment method and system with high modeling efficiency, accurate positioning, logical consistency, and strong adaptability. To achieve the above objective, the technical solution adopted by this invention is as follows: Firstly, this application provides a method for determining the natural frequencies of tunnel surrounding rock based on numerical samples, including: Obtain geological survey data for the target deep-buried jointed rock mass tunnel project; Based on the geological survey data, a parameter model was established and parameters were set using a three-dimensional numerical inversion method. The excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade were set as core influencing factors to obtain the parameterized model configuration. Numerical model construction is performed based on the parameterized model configuration. By simulating the joint distribution and mechanical response under different surrounding rock qualities, and combining the cross-combination of working conditions, dynamic mechanical analysis is performed to generate a numerical sample dataset of natural frequencies. Sensitivity analysis is performed on the numerical sample dataset of the inherent frequency. Based on the preset grey relational theory model, the degree of correlation between each influencing factor and the inherent frequency is quantified to obtain the screening results of the main control factors. Based on the screening results of the main control factors, a prediction model is established. Based on the preset machine learning model, features are extracted from numerical samples and the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel is constructed to obtain the natural frequency prediction model. The real-time control factor data obtained from on-site monitoring are input into the natural frequency prediction model for prediction processing to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep-buried jointed rock tunnel.

[0005] Secondly, this application also provides a system for determining the natural frequencies of tunnel surrounding rock based on numerical samples, comprising: The acquisition module is used to acquire geological survey data for the target deep-buried jointed rock mass tunnel project; The configuration module is used to establish and set parameters based on the geological survey data and the three-dimensional numerical inversion method. By setting the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as core influencing factors, the parameterized model configuration is obtained. The construction module is used to construct a numerical model based on the parameterized model configuration. It simulates the joint distribution and mechanical response under different surrounding rock qualities, and performs dynamic mechanical analysis by combining working conditions to generate a numerical sample dataset of natural frequencies. The analysis module is used to perform sensitivity analysis based on the inherent frequency numerical sample dataset, quantify the correlation between each influencing factor and the inherent frequency based on a preset grey relational theory model, and obtain the screening results of the main control factors. The modeling module is used to build a prediction model based on the screening results of the main control factors. Based on the preset machine learning model, it extracts features from numerical samples and constructs the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel, thus obtaining the natural frequency prediction model. The prediction module is used to input the real-time main control factor data obtained from on-site monitoring into the natural frequency prediction model for prediction processing, so as to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep buried jointed rock tunnel.

[0006] The beneficial effects of this invention are as follows: This invention achieves rapid and high-precision acquisition of the natural frequencies of deeply buried jointed rock tunnels by integrating parameterized model configuration driven by geological exploration data, natural frequency sample dataset generated by numerical simulation, selection of main control factors dominated by grey relational theory, and natural frequency prediction model constructed by machine learning. This provides a scientific basis for the proactive prevention and control of rockbursts triggered by dynamic disturbances under high ground stress environment, and significantly improves construction safety and engineering decision-making efficiency. Attached Figure Description

[0007] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 This is a schematic diagram of a method for determining the natural frequency of tunnel surrounding rock based on numerical samples, as described in an embodiment of the present invention. Figure 2 This is a schematic diagram of a system for determining the natural frequency of tunnel surrounding rock based on numerical samples, as described in an embodiment of the present invention. Figure 3 Schematic diagrams of deep-buried jointed rock mass tunnels with different surrounding rock grades; Figure 4 This is a schematic diagram illustrating the effect of GSI variation on the natural frequency of deeply buried jointed tunnels.

[0009] The diagram is labeled as follows: 901, Acquisition Module; 902, Configuration Module; 903, Construction Module; 904, Analysis Module; 905, Modeling Module; 906, Prediction Module. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0011] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0012] Example 1: This embodiment provides a method for determining the natural frequency of tunnel surrounding rock based on numerical samples.

[0013] See Figure 1 The figure shows that the method includes steps S100 to S600.

[0014] Step S100: Obtain geological survey data for the target deep-buried jointed rock mass tunnel project; To address the complex dynamic response mechanism of deeply buried jointed rock tunnels, this step focuses on obtaining the complete engineering geological background characteristics. By extracting basic parameters such as tunnel geometry, surrounding rock mechanical strength, and geostress distribution, it provides a physical basis for constructing a numerical model under real high geostress conditions, thereby avoiding the limitations of idealized geological parameters in traditional empirical formulas.

[0015] Step S200: Based on geological survey data, a parameter model is established and parameters are set using a three-dimensional numerical inversion method. The core influencing factors are set as excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade to obtain the parameterized model configuration. Understandably, in order to address the complexity of quantitative characterization of multi-factor coupling effects, this step is based on combining geological survey data with three-dimensional numerical inversion methods to set three core variables: excavation disturbance (rate), ground pressure environment (burial depth, lateral pressure coefficient), and surrounding rock integrity (level). A physical model of dynamic correlation between construction process and geological conditions is established, and its parameter configuration directly determines the degree of reproduction of the nonlinear response of jointed rock mass in subsequent numerical simulations.

[0016] Step S300: Based on the parameterized model configuration, a numerical model is constructed and processed. By simulating the joint distribution and mechanical response under different surrounding rock qualities, and combining the cross-combination of working conditions, a dynamic mechanical analysis is performed to generate a natural frequency numerical sample dataset. It should be noted that, in order to meet the training data requirements of machine learning, the rock mass structure characteristics of different surrounding rock qualities (GSI) are restored by using discrete joint networks. Combined with dynamic mechanical calculations of high ground stress conditions, an inherent frequency sample library covering the variable surrounding rock conditions in actual engineering is generated, thus eliminating the constraints of scarce field measured data and sample bias on model training.

[0017] Step S400: Perform sensitivity analysis based on the numerical sample dataset of inherent frequencies, quantify the correlation between each influencing factor and the inherent frequency based on the preset grey relational theory model, and obtain the screening results of the main control factors. To address the issue of unclear weights for multiple factors, this step utilizes grey relational analysis to analyze the impact paths of excavation parameters, geostress, and surrounding rock characteristics on the natural frequency in the numerical samples. It quantifies the contribution of each factor to the dynamic response, eliminates secondary variables to focus on key rockburst triggers, and improves the effectiveness and interpretability of subsequent model inputs.

[0018] Step S500: Based on the screening results of the main control factors, a prediction model is established. Based on the preset machine learning model, features are extracted from numerical samples and the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel is constructed to obtain the natural frequency prediction model. This step, based on a structured dataset of controlling factors, uses machine learning to extract the stiffness attenuation law and vibration mode characteristics of jointed rock masses unique to high geostress environments, and constructs a nonlinear mapping relationship between geostress boundary, surrounding rock integrity, and natural frequency.

[0019] Step S600: Input the real-time main control factor data obtained from on-site monitoring into the natural frequency prediction model for prediction processing to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep-buried jointed rock tunnel.

[0020] Finally, by combining the vibration monitoring data stream during construction, the main control parameters of the surrounding rock monitored in real time are input into the prediction model, and the natural frequency time-frequency characteristics and main frequency value based on the current geological conditions are directly output. This enables online prediction of dynamic characteristics from numerical simulation to engineering scenarios, providing real-time decision parameters for proactive prevention and control of dynamic disturbance risks in deep-buried tunnels.

[0021] Further, step S200 includes steps S210 to S230.

[0022] Step S210: Combining geological survey data with three-dimensional numerical inversion methods, basic parameter extraction processing is performed. By identifying tunnel geometric dimensions, surrounding rock mechanical parameters and geostress characteristics, a set of basic parameters is obtained. Step S220: Based on the basic parameter set, the core influencing factors are set and processed. By selecting the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as dynamic variables, the surrounding rock grade is quantitatively classified based on the geological strength index to obtain the definition of the core factors. Step S230: Based on the definition of core factors, perform parameter integration processing, construct the coupling relationship of influencing factors by associating geological strength and geostress characteristics, and obtain the parameterized model configuration.

[0023] Preferably, in the scenario of a deeply buried jointed rock tunnel, step S210 first combines geological survey data with three-dimensional numerical inversion methods to identify tunnel geometric dimensions (such as tunnel diameter), surrounding rock mechanical parameters (strength indices such as cohesion and internal friction angle), and geostress characteristics (vertical stress to lateral pressure ratio), forming a set of basic parameters characterizing the intrinsic properties of a high geostress environment; step S220 selects excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade as dynamic variables based on this parameter set, wherein the surrounding rock grade is quantitatively classified through geological strength indices, which will... The structural features such as the number of joint groups and spacing are converted into scale values ​​in the range of 50-80 (for example, low values ​​correspond to multiple dense joint groups, and high values ​​correspond to intact rock masses), realizing the quantifiable classification of complex jointed rock masses; step S230 constructs a coupled relationship model of construction disturbance-rock mass integrity-geostress environment by associating geological strength indicators with geostress characteristics (such as high ground pressure inhibiting joint opening and deep burial conditions strengthening surrounding rock constraints), generating a parameterized configuration that takes into account geological heterogeneity and construction dynamics, providing a physical basis for the subsequent accurate simulation of the dynamic response of jointed rock masses.

[0024] Further, step S300 includes steps S310 to S330.

[0025] Step S310: Based on the parameterized model configuration, the jointed rock mass model is constructed. A three-dimensional discrete joint network is established by mapping the number of joint groups, spacing and spatial distribution characteristics through geological strength index, and a parameterized surrounding rock quality model is obtained. Step S320: Perform multi-condition dynamic response analysis based on the parametric surrounding rock quality model. Simulate the vibration characteristics of surrounding rock under high ground stress environment by cross-combining tunnel burial depth, lateral pressure coefficient and excavation rate to obtain dynamic response data. Step S330: Generate natural frequency samples based on dynamic response data. By extracting the fundamental frequency vibration modes under different surrounding rock qualities and quantifying the frequency values, a natural frequency numerical sample dataset is obtained.

[0026] Specifically, considering that rockburst hazards generally occur in Class II to III surrounding rock, to accurately reflect the gradual process of the surrounding rock environment from harsh to favorable, the quality of the surrounding rock is quantitatively classified as GSI=50-80 based on the GSI (Geological Strength Index) scale. This range is primarily based on the characteristics of the surrounding rock in deeply buried tunnels: deeply buried tunnels are mostly located in high-stress environments, and the rock mass classification corresponding to GSI=50-80 is Class II to IV, covering typical surrounding rock types from "relatively stable" (Class II to III, GSI=60-80) to "prone to instability" (Class IV, GSI=50-55). Simultaneously, the GSI value is closely related to the joint environment: typically, the more joint groups there are, the lower the GSI value. For example, GSI=80 corresponds to 2 sets of joints (relatively intact rock mass), GSI=60-75 corresponds to 3 sets of joints (moderately developed), and GSI=50-55 corresponds to 4 sets of joints (relatively developed). This indicates that the more developed the joints, the worse the rock mass integrity, and the lower the GSI. Regarding joint spacing: the larger the joint spacing (from 2.0m to 5.5m) and the shorter the length (from 60m to 30m), the higher the GSI value, reflecting a weakening of the cutting effect of joints on the rock mass and an improvement in integrity. Based on these characteristics, discontinuities with different spacing, dip, and friction coefficients were introduced into the study area to simulate the rock mass quality of Class II-IV surrounding rocks, establishing deep-buried jointed rock mass tunnel models for different surrounding rock classes, such as... Figure 3 As shown in the table below, the correspondence between engineering rock mass classification and GSI is as follows: Table 1 Correspondence between Engineering Rock Mass Classification and GSI

[0027] Step S310 maps the structural characteristics of jointed rock masses in deep-buried tunnels using the Geological Strength Index (GSI). The index values ​​in the 50-80 range are transformed into the specific number of joint groups (2-4 groups), spacing (2.0-5.5m), and spatial distribution (different dip / angle combinations). A three-dimensional discrete joint network model is established. This process accurately restores the gradual characteristics of joint development under high geostress conditions (such as rock mass fragmentation caused by joint density corresponding to a decrease in index value). It overcomes the shortcomings of traditional methods in simplifying the modeling of complex rock mass structures and outputs a parameterized surrounding rock quality model. Step S320, based on this quality model, uses a cross-combination of three typical variables for deep-buried tunnels: tunnel depth (500-2500m), lateral pressure coefficient, and excavation rate (0.1-20m / day). It simulates the unique surrounding rock vibration propagation constraints (such as deep confining pressure inhibiting joint opening and rapid excavation triggering unloading dynamic waves) in a discrete element platform. Its core innovation lies in quantifying the dynamic coupling effect of construction disturbance, ground stress boundary, and joint network to generate dynamic response data that reflects the variability of actual working conditions. Step S330 extracts the fundamental frequency vibration mode (resonance harmonic components dominated by the rock mass) from the dynamic response data. A sample set is constructed by quantifying the fundamental frequency values ​​under different surrounding rock qualities, revealing the core law that increased geological strength leads to enhanced rock mass integrity, increased surrounding rock stiffness, and a monotonic increase in natural frequency. Comparing with previous research, in high-stress, deeply buried jointed hard rock tunnels, as the GSI scale increases from 50 to 80, the natural frequency of the deeply buried tunnel increases from 10Hz to 26.15Hz. That is, the more intact the surrounding rock (Level IV → Level II), the higher the natural frequency of the deeply buried jointed rock tunnel. If the volume of internal cracks in the rock mass structure is larger and the overall stiffness of the surrounding rock is smaller, its natural frequency increases. The conclusions obtained are consistent with the results of laboratory tests. Figure 4 As shown in the figure, this processing method enables the digital characterization of the stiffness attenuation characteristics of deeply buried jointed rock masses, providing a high-fidelity training sample library covering Class II-IV surrounding rocks for machine learning.

[0028] Further, step S400 includes steps S410 to S430.

[0029] Step S410: Based on the numerical sample dataset of natural frequencies, construct the evaluation sequence. Define the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as the evaluation index sequence, and the natural frequency as the reference sequence to obtain the initial evaluation matrix. Step S420: Perform data standardization processing based on the initial evaluation matrix, eliminate the differences in the dimensions of each factor by mean value, and generate a dimensionless data sequence matrix; Step S430: Perform correlation quantification processing on the dimensionless data sequence matrix, calculate the correlation coefficient and mean correlation degree, output the correlation ranking of each influencing factor and the inherent frequency, and obtain the screening results of the main control factors.

[0030] This process, based on numerical samples, conducts a sensitivity analysis of the natural frequencies of deeply buried jointed rock tunnels. A grey relational analysis model is used to calculate the correlation coefficients of each influencing factor. The specific process is as follows: A matrix of evaluation indexes is established to determine the influencing factors of the natural frequency of the surrounding rock in deeply buried jointed rock tunnels. (The matrix is ​​provided in the original text.) The evaluation sequence matrix is ​​composed of data from the evaluation indicators and reference indicators: ; In the formula, This represents the total number of evaluation index sequences and reference index sequences; This indicates the total number of evaluation indicators and reference indicators; Represents the evaluation sequence matrix; to This indicates specific evaluation indicators and reference indicators.

[0031] Next, determine the evaluation indicator list. Compared with the reference indicator series : Evaluation index sequence : ; Reference indicator sequence: : ; Then, the indicator data is made dimensionless.

[0032] Considering the different physical meanings of the factors, the dimensions of the data may not be the same. This paper uses mean normalization to make the reference index series dimensionless: ; ; In the formula, Indicates the first The first evaluation indicator Dimensionless data after mean normalization; Indicates the first The first of the evaluation indicators (such as excavation rate, tunnel depth, etc.) The raw data, i.e., the raw observations that have not undergone standardization; Indicates the sequence number of the sample or data point.

[0033] The dimensionless data sequence matrix is ​​as follows: ; In the formula, This represents the dimensionless data sequence matrix. to This represents the specific element values ​​contained in the dimensionless data sequence matrix.

[0034] Then calculate the absolute value of the difference between each evaluation index sequence and the reference index sequence:

[0035] In the formula, Indicates the first In the sample, the reference sequence (intrinsic frequency) and the... The absolute value of the numerical difference of a series of evaluation indicators (excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade) after dimensionless processing; This represents the reference sequence (natural frequency) after being averaged and dimensionless. One sample data; Indicates the first The evaluation index sequence (influencing factors) after being processed by meanization and dimensionlessness is the th... Sample data.

[0036] The maximum and minimum values ​​of the differences between the evaluation and reference indicator sequences are denoted as follows: and : ; ; Then, the correlation coefficients between each evaluation and the reference indicator series are calculated: ; In the formula, Indicates the first The evaluation index sequence (influencing factors) and the reference sequence (inherent frequency) at the 1st The correlation coefficient of each sample point; The resolution coefficient is a parameter used to adjust the sensitivity of the correlation coefficient, and its value range is typically [range missing]. .

[0037] For each evaluation indicator, the mean of the correlation coefficient between the individual indicator and the corresponding elements of the reference sequence is calculated to reflect the correlation between each evaluation indicator and the reference indicator sequence, i.e., the correlation degree. ; In the formula, Indicates the first The correlation between a series of evaluation indicators and a reference series (inherent frequency).

[0038] Based on the correlation between the evaluation indicators of the natural frequency of the deep-buried jointed rock mass tunnel determined in the above steps and the reference indicators, the evaluation indicators are ranked.

[0039] Input optimization: Key control factors with high correlation are selected as input parameters. Data preprocessing is performed on the input parameters, and the Pearson correlation coefficient is used to assess the correlation between the key control factors. Highly correlated redundant parameters are eliminated to improve the generalization ability of the subsequent prediction model. The Pearson correlation coefficient reflects the degree of linear correlation between two variables: when the linear relationship between two variables strengthens, the correlation coefficient tends to 1 or -1; when one variable increases, the other also increases, indicating a positive correlation with a correlation coefficient greater than 0; conversely, a negative correlation indicates a negative correlation with a correlation coefficient less than 0; if the correlation coefficient equals 0, it indicates that there is no linear correlation between them. The formula is: ; ; In the formula, Representing variables and The Pearson correlation coefficient between two variables is used to quantify the degree of linear correlation between the two variables. Representing variables and covariance; and Representing variables respectively and Standard deviation; Representing variables The One observation value; Representing variables The average of all observations; Representing variables The One observation value; Representing variables The average of all observations.

[0040] Further, step S500 includes steps S510 to S530.

[0041] Step S510: Based on the screening results of the main control factors, construct and process the training dataset. By associating the surrounding rock grade, tunnel burial depth and lateral pressure coefficient parameters with the frequency labels in the inherent frequency numerical samples, the training set of the main control factors is obtained. Step S520: Perform feature space mapping processing based on the main control factor training set, and extract the stiffness attenuation and vibration mode characteristics of jointed rock mass under high ground stress environment through nonlinear transformation to obtain high-dimensional feature vectors; Step S530: Train the prediction model based on the high-dimensional feature vector, optimize the mapping weights between the main control factor parameters and the inherent frequency through error backpropagation, and obtain the inherent frequency prediction model.

[0042] Specifically, the above process uses random sampling to divide the selected dataset of main control factors into training and testing sets at an 8:2 ratio, establishing a natural frequency prediction model for deeply buried jointed rock mass tunnels based on the SVM algorithm. The model is trained using the training dataset, with the input data being the parameters of each main control factor, and the output data being the time-frequency characteristics and main frequency information of the natural frequency of deeply buried jointed rock mass tunnels. The penalty factor in the SVM model during the natural frequency prediction process is analyzed based on control variables and cross-validation. With kernel function coefficients The changes in parameters are analyzed to determine the optimal values ​​for the hyperparameters. After model construction, performance evaluation is also conducted: the prediction accuracy of the SVM model is validated using a test set, employing root mean square error (RMSE), mean absolute percentage error (MAPE), and goodness-of-fit R-squared. 2 Evaluation metrics for the left and right prediction models. Under the same dataset and parameter settings, the prediction performance of the SVM model is established and compared with that of other machine learning models (random forest, neural network, etc.) to verify the superiority of the SVM model.

[0043] Further, step S600 includes steps S610 to S630.

[0044] Step S610: Based on the on-site construction environment, the vibration signal is collected and processed in real time. The vibration waveform time domain data of the surrounding rock in each direction is obtained synchronously through the vibration monitoring instrument to obtain the original vibration signal set. Step S620: Extract the main control factor features based on the original vibration signal set, separate the fundamental harmonic components of jointed rock mass under high ground stress environment through filtering and noise reduction and time-frequency transformation, and obtain the standardized main control factor input vector. Step S630: Perform natural frequency prediction processing based on the standardized main control factor input vector, and calculate the fundamental frequency response value dominated by the stiffness attenuation of the surrounding rock through forward propagation to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep buried jointed rock tunnel.

[0045] Step S610 synchronously acquires multi-directional surrounding rock vibration time-domain data (X / Y / Z three-axis waveforms) through a vibration monitoring instrument to obtain a mixed signal set containing construction disturbance and the actual response of the rock mass. This multi-directional synchronous acquisition strategy is specifically designed to address the spatial anisotropy problem of the dynamic response of jointed rock mass in deep-buried tunnels. It can separate key components such as axial excavation disturbance, radial stress wave and vertical ground pressure fluctuation, providing undistorted original vibration signals for subsequent feature extraction. It should be noted that step S620 extracts the main control factor features based on the signal set, eliminates high-frequency noise interference from construction such as TBM tunneling through filtering and noise reduction, and uses time-frequency transformation technology to separate the fundamental harmonic components unique to jointed rock masses under high ground stress. The core of this processing lies in quantifying the joint stiffness attenuation characteristics and ground stress boundary response, and finally outputs a dimensionally standardized input vector of the main control factors, whose parameter structure is strictly consistent with the numerical sample. Step S630 inputs the standardized input vector into a pre-trained machine learning model, calculates the output fundamental frequency response value dominated by the stiffness attenuation of the surrounding rock through forward propagation, and the internal learning mechanism of the model strictly follows the mapping law between geological strength index and natural frequency, and automatically corrects the boundary effect of high ground stress environment, and finally generates natural frequency time-frequency features and main frequency information that can be directly used for rockburst resonance risk early warning.

[0046] Example 2: like Figure 2 As shown, this embodiment provides a system for determining the natural frequencies of tunnel surrounding rock based on numerical samples. The system includes: The acquisition module 901 is used to acquire geological survey data for the target deep-buried jointed rock mass tunnel project; The configuration module 902 is used to establish and set parameters based on geological survey data and three-dimensional numerical inversion methods. By setting the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as core influencing factors, the parameterized model configuration is obtained. Module 903 is used to construct numerical models based on parameterized model configurations. It simulates joint distribution and mechanical response under different surrounding rock qualities, and performs dynamic mechanical analysis by combining working conditions to generate a dataset of natural frequency numerical samples. Analysis module 904 is used to perform sensitivity analysis based on the numerical sample dataset of inherent frequencies. It quantifies the degree of correlation between each influencing factor and the inherent frequency based on a preset grey relational theory model, and obtains the screening results of the main control factors. Modeling module 905 is used to establish a prediction model based on the screening results of the main control factors. Based on the preset machine learning model, it extracts features from numerical samples and constructs the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel to obtain the natural frequency prediction model. The prediction module 906 is used to input the real-time main control factor data obtained from on-site monitoring into the natural frequency prediction model for prediction processing, so as to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep buried jointed rock tunnel.

[0047] In one specific embodiment of this application, the configuration module 902 includes: The first configuration unit is used to combine geological exploration data with three-dimensional numerical inversion methods to extract basic parameters. By identifying tunnel geometry, surrounding rock mechanical parameters and geostress characteristics, a set of basic parameters is obtained. The second configuration unit is used to set the core influencing factors based on the basic parameter set. By selecting the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as dynamic variables, the surrounding rock grade is quantitatively classified based on geological strength index to obtain the definition of the core factors. The third configuration unit is used to integrate parameters based on the definition of core factors. By associating geological strength and geostress characteristics, it constructs the coupling relationship of influencing factors and obtains the parameterized model configuration.

[0048] In one specific embodiment of this application, the construction module 903 includes: The first building unit is used to construct the jointed rock mass model according to the parameterized model configuration. It establishes a three-dimensional discrete joint network by mapping the number of joint groups, spacing and spatial distribution characteristics through geological strength index to obtain the parameterized surrounding rock quality model. The second building unit is used to perform multi-condition dynamic response analysis based on the parametric surrounding rock quality model. By cross-combining tunnel burial depth, lateral pressure coefficient and excavation rate, the vibration characteristics of surrounding rock under high ground stress environment are simulated to obtain dynamic response data. The third building unit is used to generate natural frequency samples based on dynamic response data. By extracting the fundamental frequency vibration modes under different surrounding rock qualities and quantifying the frequency values, a natural frequency numerical sample dataset is obtained.

[0049] In one specific embodiment of this application, the analysis module 904 includes: The first analysis unit is used to construct and process the evaluation sequence based on the numerical sample dataset of natural frequencies. By defining the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as the evaluation index sequence and the natural frequency as the reference sequence, the initial evaluation matrix is ​​obtained. The second analysis unit is used to perform data standardization processing based on the initial evaluation matrix, eliminate the differences in the dimensions of each factor through mean value, and generate a dimensionless data sequence matrix. The third analysis unit is used to perform correlation quantification processing on the dimensionless data sequence matrix. By calculating the correlation coefficient and mean correlation degree, it outputs the correlation ranking of each influencing factor with the inherent frequency and obtains the screening results of the main control factors.

[0050] In one specific embodiment of this application, the modeling module 905 includes: The first modeling unit is used to construct and process the training dataset based on the screening results of the main control factors. It obtains the training set of the main control factors by associating the surrounding rock grade, tunnel burial depth and lateral pressure coefficient parameters with frequency labels in the numerical samples of inherent frequency. The second modeling unit is used to perform feature space mapping processing based on the training set of main control factors. It extracts the stiffness attenuation and vibration mode characteristics of jointed rock mass under high ground stress environment through nonlinear transformation to obtain high-dimensional feature vectors. The third modeling unit is used to train the prediction model based on the high-dimensional feature vector. It optimizes the mapping weights between the main control factor parameters and the inherent frequency through error backpropagation to obtain the inherent frequency prediction model.

[0051] In one specific embodiment of this application, the prediction module 906 includes: The first prediction unit is used to collect and process vibration signals in real time according to the on-site construction environment. It synchronously acquires the time-domain data of the surrounding rock vibration waveform in each direction through the vibration monitoring instrument to obtain the original vibration signal set. The second prediction unit is used to extract the main control factor features based on the original vibration signal set, and separate the fundamental harmonic components of jointed rock mass under high ground stress environment through filtering and noise reduction and time-frequency transformation to obtain the standardized main control factor input vector. The third prediction unit is used to perform natural frequency prediction processing based on the standardized main control factor input vector. It calculates and outputs the fundamental frequency response value dominated by the stiffness attenuation of the surrounding rock through forward propagation, thereby obtaining the natural frequency time-frequency characteristics and main frequency information of the target deep-buried jointed rock tunnel.

[0052] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for determining the natural frequencies of tunnel surrounding rock based on numerical samples, characterized in that, include: Obtain geological survey data for the target deep-buried jointed rock mass tunnel project; Based on the geological survey data, a parameter model was established and parameters were set using a three-dimensional numerical inversion method. The excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade were set as core influencing factors to obtain the parameterized model configuration. Numerical model construction is performed based on the parameterized model configuration. By simulating the joint distribution and mechanical response under different surrounding rock qualities, and combining the cross-combination of working conditions, dynamic mechanical analysis is performed to generate a numerical sample dataset of natural frequencies. Sensitivity analysis is performed on the numerical sample dataset of the inherent frequency. Based on the preset grey relational theory model, the degree of correlation between each influencing factor and the inherent frequency is quantified to obtain the screening results of the main control factors. Based on the screening results of the main control factors, a prediction model is established. Based on the preset machine learning model, features are extracted from numerical samples and the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel is constructed to obtain the natural frequency prediction model. The real-time control factor data obtained from on-site monitoring is input into the natural frequency prediction model for prediction processing to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep-buried jointed rock tunnel. The process involves constructing a numerical model based on the parameterized model configuration, simulating joint distribution and mechanical response under different surrounding rock qualities, and performing dynamic mechanical analysis by combining various working conditions to generate a natural frequency numerical sample dataset, including: The jointed rock mass model is constructed based on the parameterized model configuration. A three-dimensional discrete joint network is established by mapping the number of joint groups, spacing and spatial distribution characteristics through geological strength indexes, and a parameterized surrounding rock quality model is obtained. Based on the parametric surrounding rock quality model, multi-condition dynamic response analysis was performed. By cross-combining tunnel burial depth, lateral pressure coefficient and excavation rate, the vibration characteristics of surrounding rock under high ground stress environment were simulated to obtain dynamic response data. The natural frequency sample generation process is performed based on the dynamic response data. By extracting the fundamental frequency vibration modes under different surrounding rock masses and quantifying the frequency values, a natural frequency numerical sample dataset is obtained.

2. The method for determining the natural frequencies of tunnel surrounding rock based on numerical samples according to claim 1, characterized in that, Based on the geological survey data, a parameter model was established and parameters were set using a three-dimensional numerical inversion method. By setting excavation rate, tunnel depth, lateral pressure coefficient, and surrounding rock grade as core influencing factors, a parameterized model configuration was obtained, including: Based on the geological survey data, the basic parameters are extracted and processed using a three-dimensional numerical inversion method. By identifying the tunnel's geometric dimensions, surrounding rock mechanical parameters, and geostress characteristics, a set of basic parameters is obtained. Based on the aforementioned set of basic parameters, the core influencing factors are set and processed. By selecting excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as dynamic variables, the surrounding rock grade is quantitatively classified based on geological strength index, and the core factor definition is obtained. Based on the definition of the core factors, parameter integration processing is performed, and the coupling relationship of influencing factors is constructed by associating geological intensity and geostress characteristics to obtain the parameterized model configuration.

3. The method for determining the natural frequencies of tunnel surrounding rock based on numerical samples according to claim 1, characterized in that, Sensitivity analysis was performed on the numerical sample dataset of the inherent frequencies. Based on a pre-defined grey relational theory model, the correlation between each influencing factor and the inherent frequency was quantified to obtain the screening results of the main controlling factors, including: The evaluation sequence is constructed based on the numerical sample dataset of the natural frequency. The excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade are defined as evaluation index sequences, and the natural frequency is used as the reference sequence to obtain the initial evaluation matrix. Based on the initial evaluation matrix, data standardization is performed, and the differences in the dimensions of each factor are eliminated by mean value to generate a dimensionless data sequence matrix. Based on the dimensionless data sequence matrix, correlation quantification is performed. By calculating the correlation coefficient and mean correlation degree, the correlation ranking of each influencing factor with its inherent frequency is output, and the screening results of the main control factors are obtained.

4. The method for determining the natural frequencies of tunnel surrounding rock based on numerical samples according to claim 1, characterized in that, Based on the screening results of the main controlling factors, a prediction model is established. Features are extracted from numerical samples using a pre-set machine learning model, and a mapping relationship is constructed between the parameters of each main controlling factor and the natural frequency of the target deep-buried jointed rock tunnel. This yields a natural frequency prediction model, including: Based on the screening results of the main control factors, a training dataset is constructed and processed. By associating the surrounding rock grade, tunnel burial depth and lateral pressure coefficient parameters with frequency labels in the inherent frequency numerical samples, a training set of main control factors is obtained. Based on the training set of the main controlling factors, feature space mapping processing is performed, and the stiffness attenuation and vibration mode characteristics of jointed rock mass under high ground stress environment are extracted by nonlinear transformation to obtain high-dimensional feature vectors; The prediction model is trained based on the high-dimensional feature vector, and the mapping weights between the main control factor parameters and the inherent frequency are optimized through error backpropagation to obtain the inherent frequency prediction model.

5. A system for determining the natural frequencies of tunnel surrounding rock based on numerical samples, characterized in that, include: The acquisition module is used to acquire geological survey data for the target deep-buried jointed rock mass tunnel project; The configuration module is used to establish and set parameters based on the geological survey data and the three-dimensional numerical inversion method. By setting the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as core influencing factors, the parameterized model configuration is obtained. The construction module is used to construct a numerical model based on the parameterized model configuration. It simulates the joint distribution and mechanical response under different surrounding rock qualities, and performs dynamic mechanical analysis by combining working conditions to generate a numerical sample dataset of natural frequencies. The analysis module is used to perform sensitivity analysis based on the inherent frequency numerical sample dataset, quantify the correlation between each influencing factor and the inherent frequency based on a preset grey relational theory model, and obtain the screening results of the main control factors. The modeling module is used to build a prediction model based on the screening results of the main control factors. Based on the preset machine learning model, it extracts features from numerical samples and constructs the mapping relationship between the parameters of each main control factor and the natural frequency of the target deep-buried jointed rock tunnel, thus obtaining the natural frequency prediction model. The prediction module is used to input the real-time main control factor data obtained from on-site monitoring into the natural frequency prediction model for prediction processing, so as to obtain the natural frequency time-frequency characteristics and main frequency information of the target deep-buried jointed rock tunnel; The building module includes: The first construction unit is used to construct a jointed rock mass model according to the parameterized model configuration. It establishes a three-dimensional discrete joint network by mapping the number of joint groups, spacing and spatial distribution characteristics through geological strength index to obtain a parameterized surrounding rock quality model. The second construction unit is used to perform multi-condition dynamic response analysis based on the parameterized surrounding rock quality model. By cross-combining tunnel burial depth, lateral pressure coefficient and excavation rate, the vibration characteristics of surrounding rock under high ground stress environment are simulated to obtain dynamic response data. The third construction unit is used to generate natural frequency samples based on the dynamic response data. By extracting the fundamental frequency vibration modes under different surrounding rock masses and quantifying the frequency values, a natural frequency numerical sample dataset is obtained.

6. The system for determining the natural frequency of tunnel surrounding rock based on numerical samples according to claim 5, characterized in that, The configuration module includes: The first configuration unit is used to extract basic parameters based on the geological survey data and in combination with the three-dimensional numerical inversion method, and obtain the basic parameter set by identifying the tunnel geometry, surrounding rock mechanical parameters and geostress characteristics. The second configuration unit is used to set core influencing factors based on the basic parameter set. By selecting excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as dynamic variables, the surrounding rock grade is quantitatively classified based on geological strength index to obtain the definition of core factors. The third configuration unit is used to perform parameter integration processing based on the core factor definition, and to construct the coupling relationship of influencing factors by associating geological strength and geostress characteristics to obtain the parameterized model configuration.

7. The system for determining the natural frequency of tunnel surrounding rock based on numerical samples according to claim 5, characterized in that, The analysis module includes: The first analysis unit is used to construct and process the evaluation sequence based on the numerical sample dataset of the natural frequency. By defining the excavation rate, tunnel depth, lateral pressure coefficient and surrounding rock grade as the evaluation index sequence and the natural frequency as the reference sequence, an initial evaluation matrix is ​​obtained. The second analysis unit is used to perform data standardization processing based on the initial evaluation matrix, and eliminate the dimensional differences of each factor by mean value to generate a dimensionless data sequence matrix. The third analysis unit is used to perform correlation quantification processing on the dimensionless data sequence matrix, and output the correlation ranking of each influencing factor and its inherent frequency by calculating the correlation coefficient and mean correlation degree, thereby obtaining the screening results of the main control factors.

8. The system for determining the natural frequency of tunnel surrounding rock based on numerical samples according to claim 5, characterized in that, The modeling module includes: The first modeling unit is used to construct and process the training dataset based on the screening results of the main control factors. By associating the surrounding rock grade, tunnel burial depth and lateral pressure coefficient parameters with frequency labels in the inherent frequency numerical samples, the main control factor training set is obtained. The second modeling unit is used to perform feature space mapping processing based on the training set of the main control factors, and extract the stiffness attenuation and vibration mode features of jointed rock mass under high ground stress environment through nonlinear transformation to obtain a high-dimensional feature vector. The third modeling unit is used to train the prediction model based on the high-dimensional feature vector, and optimize the mapping weights between the main control factor parameters and the inherent frequency through error backpropagation to obtain the inherent frequency prediction model.

Images