Intelligent judgment method and system for tunnel anti-dislocation fortification length based on XGBoost algorithm

Abstract

The invention discloses a tunnel anti-dislocation fortification length intelligent judgment method and system based on an XGBoost algorithm, relates to the technical field of crossing of tunnel engineering and artificial intelligence, and solves the technical problem that intelligent fortification length judgment cannot be realized by quickly and accurately predicting tunnel deformation length in the prior art. The method comprises the following steps: S1, constructing a tunnel three-dimensional beam-spring model, performing a simulation test on the model to obtain simulation data to construct a database, and processing the data in the database to obtain a data set; s2, using an XGBoost algorithm to construct a multi-target regression model, and using the data set to train and optimize the multi-target regression model to obtain a prediction model; s3, tunnel and surrounding rock parameters are input into the prediction model to predict tunnel deformation length data, and then the tunnel anti-dislocation fortification length is calculated through a safety coefficient method; the tunnel deformation length can be quickly and efficiently predicted by inputting the tunnel geometric parameters, the stratum parameters and the like, and the prediction error is small.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel engineering and artificial intelligence, and in particular to a method and system for intelligently determining the length of a tunnel anti-fault defense system based on an XGBoost algorithm. Background Art

[0002] Earthquake fault movement is one of the main causes of tunnel structural damage. Fault activity can cause tunnel longitudinal stretching, shearing or bending deformation, and even cause structural fracture in severe cases.

[0003] Traditional seismic design relies on empirical formulas or simplified mechanical models, which make it difficult to accurately predict tunnel responses under complex geological conditions, resulting in conservative design of defense lengths.

[0004] The existing tunnel anti-fault length prediction method has the following main shortcomings: (1) Empirical formula method: This method is based on linear or nonlinear equations fitted with a small amount of test data (such as regression formulas based on relative stiffness ratio). This method ignores the nonlinear behavior of soil and the dynamic interaction between structure and soil, and has large prediction errors.

[0005] (2) Numerical simulation method, using finite element or discrete element method to simulate fault-tunnel interaction. Although the accuracy is relatively high, the modeling is complex, the calculation is time-consuming (a single simulation takes hours to days), and it relies on expert experience to adjust parameters, which makes it difficult to quickly apply to actual projects.

[0006] (3) Physical testing method: obtaining data through centrifuge testing or model testing, but it is costly, time-consuming, and difficult to cover all working conditions.

[0007] For example, the Chinese patent "A method for calculating the length of a tunnel crossing an active fault and its fortified extension section" (patent application number: CN202111023070.5, publication number: CN113685199A). This patent solves the problem of errors and limitations in determining the length of the fortified section by numerical simulation when the width of the fault fracture zone is small through quantitative calculation.

[0008] However, this patent uses a formula for segmented calculations, which involves many intermediate calculation processes, and the calculation accuracy and work efficiency of related engineering personnel cannot be guaranteed.

[0009] For example, the Chinese patent "Prediction method of tunnel surrounding rock extrusion deformation based on GA-XGBoost model" (patent application number: CN202110673385.8, publication number: CN113326660B). This patent predicts the extrusion deformation of tunnel surrounding rock by constructing and training the GA-XGBoost model.

[0010] However, this patent mainly focuses on the tunnel cross-section deformation caused by surrounding rock compression and is not suitable for the prediction and calculation of parameters related to the fortification length. Summary of the invention

[0011] In order to solve the problems existing in the above-mentioned prior art, the present invention provides a method and system for intelligently determining the anti-fracture defense length of a tunnel based on the XGBoost algorithm, which solves the technical problem that the prior art cannot achieve intelligent determination of the defense length by quickly and accurately predicting the deformation length of the tunnel.

[0012] The intelligent determination method of tunnel anti-fault defense length based on XGBoost algorithm includes: S1: Construct a three-dimensional beam-spring model of the tunnel, conduct simulation tests on the model to obtain simulation data to build a database, and then process the data in the database to obtain a data set; S2: Use the XGBoost algorithm to build a multi-objective regression model, and use the data set to train and optimize the multi-objective regression model to obtain a prediction model; S3: Input tunnel and surrounding rock parameters into the prediction model to predict tunnel deformation length data, and then calculate the tunnel anti-fault defense length through the safety factor method. The tunnel and surrounding rock parameters include diameter, burial depth, thickness, and rock-soil friction angle.

[0013] Furthermore, the step of constructing the three-dimensional beam-spring model of the tunnel in S1 includes: ANSYS software was used to establish a coupling model of tunnel and soil spring unit; boundary conditions were set: the footplate of the fault was fixed and the upper plate was free; the tunnel material constitutive model was defined as a bilinear strain hardening model, the tunnel elastic modulus value was determined according to the tunnel material, and then the characteristics including rock-soil friction angle, tunnel burial depth, burial depth and diameter ratio, and diameter-thickness ratio were determined to cover the typical values ​​of actual engineering; the soil spring resistance was calculated to obtain simulation data; the simulation data was compared with the test data to verify the model accuracy and ensure that the numerical simulation results were consistent with the measured data, and the test data was obtained through centrifuge test.

[0014] The simulation data and test data specifically include: tunnel longitudinal bending strain, tunnel axis displacement, tunnel deformation length; The calculation formula of soil spring resistance is: Axial spring force

[0015] Where γ is the effective weight of soil, K0 is the static earth pressure coefficient, and k is the reduction factor. In concrete tunnels and steel-lined tunnels, k is taken as 1.0 and 0.6 respectively. is the internal friction angle, D is the tunnel diameter, and H is the tunnel depth.

[0016] Lateral spring force

[0017] Where N qh is the horizontal bearing capacity coefficient.

[0018] Vertical spring force

[0019] Where N qv is the vertical tensile coefficient, N q is the bearing capacity coefficient related to the foundation depth, N y is the bearing capacity factor related to the foundation width and soil weight.

[0020] Furthermore, the database described in S1 is composed of input features and labels for each data, wherein the input features include: tunnel geometric parameters (diameter, burial depth, thickness), rock-soil friction angle; the labels (numerical simulation results) include: tunnel deformation field, strain distribution and deformation length generated by a three-dimensional beam-spring model; and data preprocessing is performed to remove abnormal values.

[0021] Further, the S2 includes: S2.1, determine the number of XGBoost learning trees and the initial range of tree depth; S2.2. Optimize the number of learning trees and the maximum depth of the trees through grid search: perform an exhaustive search within the preset range, calculate the model performance under each set of parameters, and select the best performing hyperparameter set as the candidate value; Then, the Bayesian optimization method is used to adjust other hyperparameters including learning rate, minimum number of sub-node samples and sub-sample ratio to reduce unnecessary calculations, improve computing efficiency, and further find the optimal hyperparameter set.

[0022] Further, the S2 includes: S2.3: On the basis of hyperparameter optimization, combined with the complexity-performance balance optimization strategy, find the optimal balance between model complexity and performance, ensure that the model can effectively capture the data rules, avoid overfitting or underfitting, and improve the robustness and prediction accuracy of the model in practical applications. Specifically, use 5-fold cross-validation to evaluate the model: divide the training data into 5 parts, use 4 of them as training sets each time, and the remaining 1 as a validation set; repeat 5 times to ensure that each part of the data is used as a validation set; calculate the mean square error M MSE and standard error SE MSE , to ensure the stability of the model; with the mean square error M MSE Not higher than the minimum mean square error M MSE Add a standard error SE MSEThe model is used as a reference to determine the hyperparameter values, which ensures that a model with lower computational complexity is selected without significantly losing performance; finally, the final model is retrained on the entire training set to obtain the prediction model.

[0023] The calculation formula for the five-fold cross validation is:

[0024] in is the mean square error of the k-th training.

[0025] Further, the S3 includes: S3.1. Analyze the influence of various characteristics including geotechnical friction angle, tunnel burial depth, burial depth to diameter ratio, and diameter-thickness ratio on the anti-dislocation defense length through SHAP value, generate characteristic contribution graph, and analyze the influence of various characteristics on the prediction results according to the characteristic contribution graph; S3.2. By comparing the predicted values ​​with the actual values, the mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are calculated to verify the accuracy and applicability of the model. Finally, the verified prediction model is used to predict the tunnel deformation length data.

[0026] Furthermore, the tunnel is a concrete tunnel or a steel-lined tunnel.

[0027] Furthermore, the calculation formula for the tunnel anti-fault protection length calculated by the safety factor method described in S3 is:

[0028] Where, L is the tunnel anti-fault protection length, L a is the total deformation length of the tunnel, L ah is the deformation length of the upper plate, L af is the deformation length of the lower plate and α is the safety factor.

[0029] Furthermore, the All are predicted by the prediction model. The numerical relationship between them further determines the fault location to determine the fortification location and range.

[0030] The intelligent determination system of tunnel anti-fracture defense length based on XGBoost algorithm includes a data input module, a prediction module and a defense length output module. The data input module is used to input tunnel and surrounding rock parameters, including diameter, burial depth, thickness, and rock-soil friction angle. The prediction module includes a prediction model in the intelligent determination method of tunnel anti-fracture defense length based on XGBoost algorithm. The prediction model predicts the deformation length of the tunnel according to the input tunnel and surrounding rock parameters. The defense length output module calculates the tunnel anti-fracture defense length using a safety factor method and outputs it.

[0031] The beneficial effects of the present invention include: 1. Simple data acquisition: A large amount of data for training and testing can be generated through a three-dimensional beam-spring numerical model that is easy to operate and can reflect the deformation characteristics of tunnels across active faults.

[0032] 2. The model prediction efficiency and accuracy are high. Compared with the traditional empirical formula method and physical test method, the present invention can quickly and efficiently predict the tunnel deformation length by inputting tunnel geometric parameters and formation parameters, and the prediction error is small.

[0033] 3. Analytical feature influence. The present invention clarifies the influence of soil friction angle, tunnel burial depth, tunnel diameter-thickness ratio and other characteristics on the tunnel anti-fault defense length, provides intuitive and interpretable results, and is conducive to engineering decision-making.

[0034] 4. Wide application scope. The present invention can be applied to the anti-fault protection of concrete tunnels and steel-lined tunnels, and has a wide application scope. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flow chart of the method for intelligently determining the tunnel anti-fault defense length based on the XGBoost algorithm involved in an embodiment of the present application.

[0036] Figure 2 It is a schematic diagram of a three-dimensional beam-spring model involved in an embodiment of the present application.

[0037] Figure 3 This is a flowchart of hyperparameter optimization and model optimization involved in the embodiments of the present application.

[0038] Figure 4 Schematic diagram of input feature contribution involved in the embodiments of the present application, wherein (a) is a schematic diagram of input feature contribution in a concrete tunnel, and (b) is a schematic diagram of input feature contribution in a steel-lined tunnel.

[0039] Figure 5These are model verification and prediction diagrams in a concrete tunnel involved in an embodiment of the present application, wherein (a) is a model verification and prediction diagram of a concrete tunnel in an upper plate, (b) is a model verification and prediction diagram of a concrete tunnel in a lower plate, and (c) is a model verification and prediction diagram in the entire concrete tunnel.

[0040] Figure 6 1 is a comparison diagram of the errors between the model prediction value and the actual value in the concrete tunnel involved in the embodiment of the present application, wherein (a) is a comparison diagram of the mean square error, (b) is a comparison diagram of the mean absolute error, and (c) is a comparison diagram of the mean absolute percentage error.

[0041] Figure 7 These are model verification and prediction diagrams in a steel-lined tunnel involved in an embodiment of the present application, wherein (a) is a model verification and prediction diagram for a steel-lined tunnel in an upper plate, (b) is a model verification and prediction diagram for a steel-lined tunnel in a lower plate, and (c) is a model verification and prediction diagram for the entire steel-lined tunnel.

[0042] Figure 8 : This is a comparison diagram of the errors between the model prediction value and the actual value in the steel-lined tunnel involved in the embodiment of the present application, wherein (a) is a mean square error comparison diagram, (b) is a mean absolute error comparison diagram, and (c) is a mean absolute percentage error comparison diagram. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0044] An intelligent determination method for tunnel anti-fault protection length based on XGBoost algorithm, such as Figure 1 As shown, including: S1: Construct a three-dimensional beam-spring model of the tunnel, conduct simulation tests on the model to obtain simulation data to build a database, and then process the data in the database to obtain a data set; S2: Use the XGBoost algorithm to build a multi-objective regression model, and use the data set to train and optimize the multi-objective regression model to obtain a prediction model; S3: Input tunnel and surrounding rock parameters into the prediction model to predict tunnel deformation length data, and then calculate the tunnel anti-fault defense length through the safety factor method. The tunnel and surrounding rock parameters include diameter D, burial depth H, thickness t, and rock-soil friction angle φ.

[0045] The calculation formula of the tunnel anti-fault protection length calculated by the safety factor method is:

[0046] Where, L is the tunnel anti-fault protection length, L a is the total deformation length of the tunnel, L ah is the deformation length of the upper plate, L af is the deformation length of the lower plate and α is the safety factor.

[0047] Specifically, the safety factor is summarized based on actual engineering and numerical simulation, and is the maximum value of the ratio of deformation length under the most unfavorable conditions to that under general conditions, which is 1.6 in concrete tunnels.

[0048] In another embodiment, the All are predicted by the prediction model. The numerical relationship between them further determines the fault location to determine the fortification location and range.

[0049] In another embodiment, the database construction in S1 is mainly based on numerical simulation and Python language, including: Construct a three-dimensional beam-spring model: use ANSYS software to establish a coupling model of tunnel and soil spring unit; set boundary conditions: the lower plate of the fault is fixed and the upper plate is free; define the tunnel material constitutive model as a bilinear strain hardening model, determine the value of the tunnel elastic modulus according to the tunnel type, and then determine the characteristics including the rock-soil friction angle, tunnel burial depth, burial depth and diameter ratio, and diameter-thickness ratio to cover the typical values ​​of actual engineering; calculate the soil spring resistance and obtain simulation data; compare the simulation data with the test data to verify the model accuracy and ensure that the numerical simulation results are consistent with the measured data. The test data is obtained through centrifuge testing.

[0050] The simulation data and test data specifically include: tunnel longitudinal bending strain, tunnel axis displacement, tunnel deformation length; Centrifuge test process: 1. Test preparation: Determine the test acceleration (centrifuge speed) based on the proposed similarity ratio. Paste strain gauges on the tunnel model in advance (test tunnel strain).

[0051] 2. Burying of stratum and tunnel models: Burying of tunnel models and stratum models: Filling the model box with model soil and burying the tunnel at a preset depth in the stratum.

[0052] 3. Arrangement of other sensors: Set a displacement sensor above the stratum model (to test the surface displacement) and set a camera at the front of the model box (to test the displacement of strata at different burial depths).

[0053] 4. Place and fix the model box on the centrifuge, turn on the centrifuge and accelerate the speed of the centrifuge to the designed speed.

[0054] 5. Record the initial readings of sensors (strain gauges, displacement sensors, cameras).

[0055] 6. Use the model box to carry out staggered loading, which is divided into 20 steps for distributed loading. After each step of loading, record the sensor reading.

[0056] 7. After the displacement loading is completed, turn off the centrifuge. Take out the tunnel model from the stratum and measure the displacement of the tunnel axis with a vernier caliper.

[0057] The constructed three-dimensional beam-spring model is as follows Figure 2 As shown in the figure, it is composed of beam units and spring units. The axial spring is used to simulate the axial tensile and compressive deformation of the tunnel, the transverse spring is used to simulate the transverse shear deformation of the tunnel, and the vertical spring is used to simulate the vertical deformation of the tunnel. Δx represents the axial displacement of the tunnel, Δy represents the transverse displacement of the tunnel, and Δz represents the vertical displacement of the tunnel. The lower plate of the model is set as a fixed boundary, and the upper plate is set as a free boundary that can be displaced, so as to simulate the actual project.

[0058] Specifically, the tunnel is a concrete tunnel, and the elastic modulus of the concrete tunnel is 4.4-42.68MPa; the characteristic range is determined: the friction angle (φ) is 29.7°-33.4°, the tunnel burial depth (H) is 2-56m, the ratio of burial depth to diameter (H / D) is 0.5-4, and the ratio of diameter to thickness (D / t) is 14-20, covering the typical values ​​of actual engineering; the soil spring resistance is calculated by formula, and 480 sets of simulation data are obtained; The calculation formula of soil spring resistance is: Axial spring force

[0059] Where γ is the effective weight of the soil, K0 is the static earth pressure coefficient, k is the reduction factor, which is taken as 1.0 in concrete tunnels, D is the tunnel diameter, and H is the tunnel depth.

[0060] Lateral spring force

[0061] Where N qh is the horizontal bearing capacity coefficient.

[0062] Vertical spring force

[0063] Where N qv is the vertical tensile coefficient, N q is the bearing capacity coefficient related to the foundation depth, N y is the bearing capacity factor related to the foundation width and soil weight.

[0064] The database consists of input features and labels for each piece of data, wherein the input features include: tunnel geometric parameters (diameter D, burial depth H, thickness t), rock-soil friction angle φ; the labels (numerical simulation results) include: tunnel deformation field, strain distribution and deformation length generated by a three-dimensional beam-spring model; The database data was processed and grouped, specifically by identifying and removing outliers through the quantile method, and then divided into training set and test set in a ratio of 8:2.

[0065] In another embodiment, the S2 includes: S2.1, determine the number of XGBoost learning trees and the initial range of tree depth; S2.2. Optimize the number of learning trees and the maximum depth of the tree through grid search: perform exhaustive search within the preset range (the number of learning trees n_estimators ∈ [10, 100], the maximum depth of the tree max_depth ∈ [3, 8]), calculate the model performance under each set of parameters, and select the best performing hyperparameter set as the candidate value; Then, the Bayesian optimization method is used to adjust other hyperparameters including learning rate, minimum number of sub-node samples, and sub-sample ratio to reduce unnecessary calculations, improve calculation efficiency, and further find the optimal hyperparameter set; In another embodiment, the S2 includes: S2.3: Based on hyperparameter optimization, combined with the complexity-performance balance optimization strategy, find the optimal balance between model complexity and performance, ensure that the model can effectively capture data patterns while avoiding overfitting or underfitting, and improve the model in practice. The robustness and prediction accuracy of the application are evaluated by using a five-fold cross-validation model: the training data is divided into five parts, four of which are used as training sets each time, and the remaining one is used as a validation set; repeat five times to ensure that each part of the data is used as a validation set; calculate the mean square error M MSE and standard error SE MSE , to ensure the stability of the model; with the mean square error M MSE Not higher than the minimum mean square error M MSE Add a standard error SE MSEThe model is used as a reference to determine the hyperparameter values, which ensures that a model with lower computational complexity is selected without significant performance loss; finally, the final model is retrained on the entire training set to obtain the prediction model. The specific process is shown in Figure 3 .

[0066] The calculation formula for the five-fold cross validation is:

[0067] in is the mean square error of the k-th training.

[0068] In another embodiment, the S3 includes: S3.1. Analyze the influence of various characteristics including geotechnical friction angle (φ), tunnel depth (H), depth to diameter ratio (H / D), and diameter-thickness ratio (D / t) on the anti-fault fortification length through SHAP value, and generate characteristic contribution diagram. For details, please refer to Figure 4 (a) Schematic diagram of input feature contribution in a concrete tunnel. It can be seen from the figure that the burial depth (H) is particularly important for the anti-fault of concrete tunnels. S3.2. By comparing the predicted values ​​with the actual values, the mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are calculated to verify the accuracy and applicability of the model. Finally, the verified prediction model is used to predict the tunnel deformation length data.

[0069] In this embodiment, the tunnel is a concrete tunnel, and the specific verification effect is as follows: Figure 5 , 6 shown.

[0070] Specifically, Figure 5 These are model verification and prediction diagrams in a concrete tunnel involved in an embodiment of the present application, wherein (a) is a model verification and prediction diagram of a concrete tunnel in an upper plate, (b) is a model verification and prediction diagram of a concrete tunnel in a lower plate, and (c) is a model verification and prediction diagram in the entire concrete tunnel. Figure 6 1 is a comparison diagram of the errors between the model prediction value and the actual value in the concrete tunnel involved in the embodiment of the present application, wherein (a) is a comparison diagram of the mean square error, (b) is a comparison diagram of the mean absolute error, and (c) is a comparison diagram of the mean absolute percentage error.

[0071] exist Figure 5 In, L ah / D refers to the ratio of the deformation length of the concrete tunnel in the upper wall to the diameter of the concrete tunnel. af / D refers to the ratio of the deformation length of the concrete tunnel in the footwall to the diameter of the concrete tunnel, L a / D refers to the ratio of the total deformation length of the concrete tunnel to the diameter of the concrete tunnel; The formula refers to the result obtained based on the existing empirical formula method, XGBoost training refers to the result obtained by using the XGBoost algorithm in the training set data, and XGBoost testing refers to the result obtained by using the XGBoost algorithm in the test set data.

[0072] exist Figure 6 In, L ah Refers to the deformation length of the concrete tunnel in the upper wall, L af Refers to the deformation length of the concrete tunnel in the footwall, L a Refers to the total deformation length of the concrete tunnel; The training set XGBoost refers to the error of the result obtained by using the XGBoost algorithm in the training set data, the training set formula refers to the error of the result obtained by using the empirical formula method in the training set data, the test set XGBoost refers to the error of the result obtained by using the XGBoost algorithm in the test set data, and the test set formula refers to the error of the result obtained by using the empirical formula method in the test set data.

[0073] Figure 5 It shows that most of the training and test results of the concrete tunnel prediction model based on the XGBoost algorithm are highly consistent with the ideal prediction line, but the results calculated by the formula are scattered around the ideal prediction line, which proves that the model has high accuracy; Figure 6 It shows that the error of the concrete tunnel prediction model results based on the XGBoost algorithm is at least one order of magnitude lower than that calculated by the formula. The mean square error of the total deformation length of the concrete tunnel calculated by the XGBoost model and the formula on the training set are 0.016 and 1.896 respectively, the mean absolute error of the total deformation length calculated by the XGBoost model and the formula on the training set are 0.09 and 1.109 respectively, and the mean absolute percentage error of the total deformation length calculated by the XGBoost model and the formula on the training set are 0.5% and 6.1% respectively, which proves that the model has a small error.

[0074] In another embodiment, the tunnel is a steel-lined tunnel, and the elastic modulus of the steel-lined tunnel is 0.96-31.45 MPa; the characteristic range is determined: the rock and soil friction angle (φ) is 29.7°-33.4°, the burial depth (H) of the steel-lined tunnel is 0.1-32m, the ratio of burial depth to diameter (H / D) is 0.5-8, and the ratio of diameter to thickness (D / t) is 20-50.

[0075] In the calculation formula of the axial spring of the soil spring resistance, the reduction coefficient K is taken as 0.6.

[0076] The safety factor α in the calculation formula of the defense length is taken as 2.0.

[0077] The SHAP value is used to analyze the influence of various characteristics including geotechnical friction angle (φ), tunnel depth (H), depth to diameter ratio (H / D), and diameter-thickness ratio (D / t) on the anti-fault fortification length, and generate a characteristic contribution diagram. For details, please refer to Figure 4 (b) Schematic diagram of input feature contribution in steel-lined tunnels. It can be seen from the figure that tunnel depth (H) is the most important factor for the anti-fault of steel-lined tunnels, followed by depth and diameter ratio (H / D), then diameter-thickness ratio (D / t), and the influence of rock-soil friction angle (φ) is the smallest.

[0078] The predicted values ​​of the prediction model are compared with the actual values, and the mean square error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) are calculated to verify the accuracy and applicability of the model. The comparison results are shown in Figure 7 and Figure 8 .

[0079] Specifically, Figure 7 These are model verification and prediction diagrams in a steel-lined tunnel involved in an embodiment of the present application, wherein (a) is a model verification and prediction diagram for a steel-lined tunnel in an upper plate, (b) is a model verification and prediction diagram for a steel-lined tunnel in a lower plate, and (c) is a model verification and prediction diagram for the entire steel-lined tunnel. Figure 8 : This is a comparison diagram of the errors between the model prediction value and the actual value in the steel-lined tunnel involved in the embodiment of the present application, wherein (a) is a mean square error comparison diagram, (b) is a mean absolute error comparison diagram, and (c) is a mean absolute percentage error comparison diagram.

[0080] exist Figure 7 In, L ah / D refers to the ratio of the deformation length of the steel-lined tunnel in the upper wall to the diameter of the steel-lined tunnel. af / D refers to the ratio of the deformation length of the steel-lined tunnel in the footwall to the diameter of the steel-lined tunnel. a / D refers to the ratio of the total deformation length of the steel-lined tunnel to the diameter of the steel-lined tunnel; The formula refers to the result obtained based on the existing empirical formula method, XGBoost training refers to the result obtained by using the XGBoost algorithm in the training set data, and XGBoost testing refers to the result obtained by using the XGBoost algorithm in the test set data.

[0081] exist Figure 8 In, L ah Refers to the deformation length of the steel-lined tunnel in the upper wall, L af Refers to the deformation length of the steel-lined tunnel in the footwall, L a Refers to the total deformation length of the steel-lined tunnel; The training set XGBoost refers to the error of the result obtained by using the XGBoost algorithm in the training set data, the training set formula refers to the error of the result obtained by using the empirical formula method in the training set data, the test set XGBoost refers to the error of the result obtained by using the XGBoost algorithm in the test set data, and the test set formula refers to the error of the result obtained by using the empirical formula method in the test set data.

[0082] Figure 7 It shows that most of the training and testing results of the steel-lined tunnel prediction model based on the XGBoost algorithm are highly consistent with the ideal prediction line, but the results calculated by the formula are scattered around the ideal prediction line, proving that the model has high accuracy; Figure 8 It shows that the error of the prediction model results of steel-lined tunnel based on XGBoost algorithm is at least one order of magnitude lower than that calculated by formula. The mean square error of the total deformation length of steel-lined tunnel calculated by XGBoost model and formula on the training set is 2.37 and 591.9 respectively. The mean absolute error of the total deformation length calculated by XGBoost model and formula on the training set is 1.134 and 18.67 respectively. The mean absolute percentage error of the total deformation length calculated by XGBoost model and formula on the training set is 1.9% and 29% respectively, which proves that the error of the model is small.

[0083] In another embodiment, a system for intelligently determining the tunnel anti-fracture defense length based on an XGBoost algorithm is provided, comprising a data input module, a prediction module and a defense length output module, wherein the data input module is used to input tunnel and surrounding rock parameters including diameter D, burial depth H, thickness t, and rock-soil friction angle φ; the prediction module comprises a prediction model in an intelligent method for determining the tunnel anti-fracture defense length based on an XGBoost algorithm, wherein the prediction model predicts the deformation length of the tunnel according to the input tunnel and surrounding rock parameters; and the defense length output module calculates and outputs the tunnel anti-fracture defense length using a safety factor method.

[0084] The above-mentioned embodiments only express the specific implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the protection scope of the present application. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the technical solution concept of the present application, and these all belong to the protection scope of the present application.

Claims

1. An intelligent determination method for tunnel fault-proof defense length based on XGBoost algorithm, characterized in that: The following steps are involved: S1: Construct a three-dimensional beam-spring model of the tunnel, conduct simulation tests on the model to obtain simulation data to build a database, and then process the data in the database to obtain a data set; S2: Use the XGBoost algorithm to build a multi-objective regression model, and use the data set to train and optimize the multi-objective regression model to obtain a prediction model; S3: Input tunnel and surrounding rock parameters into the prediction model to predict tunnel deformation length data, and then calculate the tunnel anti-fault defense length through the safety factor method. The tunnel and surrounding rock parameters include diameter, burial depth, thickness, and rock-soil friction angle.

2. According to the XGBoost algorithm-based intelligent determination method for tunnel fault-proof defense length according to claim 1, it is characterized in that: The construction of the tunnel three-dimensional beam-spring model described in S1 includes: ANSYS software was used to establish a coupling model of tunnel and soil spring unit; boundary conditions were set: the footplate of the fault was fixed and the upper plate was free; the tunnel material constitutive model was defined as a bilinear strain hardening model, the tunnel elastic modulus value was determined according to the tunnel material, and then the characteristics including rock-soil friction angle, tunnel burial depth, burial depth and diameter ratio, and diameter-thickness ratio were determined to cover the typical values ​​of actual engineering; the soil spring resistance was calculated to obtain simulation data; the simulation data was compared with the test data to verify the model accuracy and ensure that the numerical simulation results were consistent with the measured data, and the test data was obtained through centrifuge test.

3. According to the XGBoost algorithm-based intelligent determination method for tunnel fault-proof defense length according to claim 1, it is characterized in that: The database described in S1 is composed of input features and labels for each piece of data. The input features include: diameter, burial depth, thickness, rock-soil friction angle; the labels, i.e., numerical simulation results, include: tunnel deformation field, strain distribution, and deformation length generated by a three-dimensional beam-spring model.

4. According to the XGBoost algorithm-based intelligent determination method for tunnel anti-fault defense length according to claim 1, it is characterized in that: The S2 includes: S2.1, determine the number of XGBoost learning trees and the initial range of tree depth; S2.

2. Optimize the number of learning trees and the maximum depth of the trees through grid search: perform an exhaustive search within the preset range, calculate the model performance under each set of parameters, and select the best performing hyperparameter set as the candidate value; The Bayesian optimization method is then used to adjust other hyperparameters including the learning rate, the minimum number of sub-node samples, and the sub-sample ratio to further find the optimal hyperparameter set.

5. According to the XGBoost algorithm-based intelligent determination method for tunnel fault-proof defense length according to claim 4, it is characterized in that: The S2 includes: S2.3: Use K-fold cross validation to evaluate the model: Divide the training data into K parts, use K-1 parts as the training set each time, and use the remaining 1 part as the validation set; repeat K times to ensure that each part of the data is used as the validation set; calculate the mean square error M MSE and standard error SE MSE ; Take the mean square error M MSE Not higher than the minimum mean square error M MSE Add a standard error SE MSE The model is used as a reference to determine the hyperparameter values, and finally the final model is retrained on the entire training set to obtain the prediction model.

6. According to the XGBoost algorithm-based intelligent determination method for tunnel fault-proof defense length according to claim 1, it is characterized in that: The S3 includes: S3.

1. Analyze the influence of various characteristics including geotechnical friction angle, tunnel burial depth, burial depth to diameter ratio, and diameter-thickness ratio on the anti-dislocation defense length through SHAP value, generate characteristic contribution graph, and analyze the influence of various characteristics on the prediction results according to the characteristic contribution graph; S3.

2. By comparing the predicted value of the prediction model with the actual value, and calculating the mean square error, mean absolute error and mean absolute percentage error, the accuracy and applicability of the model are verified. Finally, the verified prediction model is used to predict the tunnel deformation length data.

7. The method for intelligently determining the length of tunnel fault-proof defense based on XGBoost algorithm according to claim 1 is characterized in that: The tunnel is a concrete tunnel or a steel-lined tunnel.

8. The method for intelligently determining the length of a tunnel fault-proof defense system based on the XGBoost algorithm according to any one of claims 1 to 7, characterized in that: The calculation formula for the tunnel anti-fault protection length calculated by the safety factor method described in S3 is: ； Where, L is the tunnel anti-fault protection length, L a is the total deformation length of the tunnel, L ah is the deformation length of the upper plate, L af is the deformation length of the lower plate and α is the safety factor.

9. The method for intelligently determining the length of tunnel fault-proof defense based on XGBoost algorithm according to claim 8 is characterized in that: Said All are predicted by the prediction model. The numerical relationship between them further determines the fault location to determine the fortification location and range.

10. The intelligent determination system of tunnel anti-fault defense length based on XGBoost algorithm is characterized by: It includes a data input module, a prediction module and a defense length output module. The data input module is used to input tunnel and surrounding rock parameters, including diameter, burial depth, thickness, and rock-soil friction angle. The prediction module includes the prediction model in the tunnel anti-fracture defense length intelligent determination method based on the XGBoost algorithm as described in claim 9. The prediction model predicts the tunnel deformation length according to the input tunnel and surrounding rock parameters. The defense length output module calculates the tunnel anti-fracture defense length using the safety factor method and outputs it.

Images