Rock-fill dam material parameter single-target inversion method and system based on CPO-XGBoost

Abstract

The invention provides a rock-fill dam material parameter single-target inversion method and system based on CPO-XGBoost, and the method comprises the steps: carrying out the sensitivity analysis of Duncan-Zhang E-B model parameters in a finite element rock-fill dam model, and obtaining to-be-inverted parameters; segmenting the to-be-inverted parameters, and training and verifying the constructed XGBoost model based on a segmentation result to obtain a rock-fill dam material parameter inversion model; performing iterative optimization in a parameter space of the rock-fill dam material parameter inversion model through a crown porcupine optimization algorithm based on a fitness function defining a mean square error as a target to obtain an optimal parameter combination and an optimal fitness value, and retaining the optimal parameter combination to obtain an inversion result; and inputting the inversion parameters into a finite element model, and verifying the consistency of calculated settlement and actually measured data. According to the method, the to-be-inverted parameters which have obvious influence on the deformation of the rock-fill dam can be accurately identified, and the calculation time can be obviously shortened on the premise of keeping relatively high precision.

Description

Technical Field

[0001] This invention belongs to the field of water conservancy engineering technology, specifically to a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost. Background Technology

[0002] As a common hydraulic engineering structure, the safety and stability of rockfill dams directly depend on the accuracy of their material parameters. The material parameters of rockfill dams (such as elastic modulus, Poisson's ratio, internal friction angle, and cohesion) are key indicators reflecting the mechanical properties of the dam body and have a significant impact on the stress, deformation, and stability analysis of the dam. However, due to the non-homogeneity and complexity of rockfill dam materials and the uncertainties during construction, it is difficult to directly obtain these parameters through experiments or theoretical calculations. Therefore, a method for inverting the material parameters of rockfill dams has been proposed.

[0003] Traditional methods for inverting material parameters in rockfill dams primarily rely on finite element numerical simulation (FEM) and optimization algorithms. The process typically involves constructing a finite element model based on measured data and adjusting material parameters using optimization algorithms to ensure the numerical simulation results closely match the measured data. While these methods can yield inversion results, the computational burden of finite element numerical simulation is substantial, especially during parameter inversion, requiring multiple model runs and resulting in excessively long computation times, making it difficult to meet practical engineering needs. Furthermore, due to the complex nonlinear relationship between rockfill dam material parameters and dam response, traditional optimization algorithms are prone to getting trapped in local optima, compromising inversion accuracy and leading to significant nonlinear effects and processing difficulties. Additionally, the strong coupling effect among rockfill dam material parameters makes it difficult for traditional methods to effectively handle simultaneous inversion of multiple parameters, resulting in insufficient reliability of the inversion results. Summary of the Invention

[0004] To address the problems of low computational efficiency and significant impact of nonlinearity on the inverted structure, leading to low inversion accuracy, traditional methods for inverting material parameters of rockfill dams provide a single-objective inversion method for rockfill dam material parameters based on CPO-XGBoost.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention proposes a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost, comprising the following steps:

[0007] Based on the sensitivity analysis of the Duncan-Chang EB model parameters in the finite element rockfill dam model, the parameters to be inverted are obtained.

[0008] The parameters to be inverted are segmented, and the constructed XGBoost model is trained and validated based on the segmentation results to obtain the inversion model of material parameters of rockfill dam.

[0009] Based on the defined fitness function with mean square error as the objective, the optimal parameter combination and the optimal fitness value are obtained by iterative optimization in the parameter space of the rockfill dam material parameter inversion model through the hog optimization algorithm. The optimal parameter combination is retained to obtain the inversion result.

[0010] The inversion parameters were input into the finite element model to verify the consistency between the calculated settlement and the measured data.

[0011] Preferably, the sensitivity analysis based on the Duncan-Chang EB model parameters in the finite element rockfill dam model includes:

[0012] Step 1.1: Construct a finite element model of the rockfill dam based on the acquired engineering data;

[0013] Step 1.2: Determine the actual measuring points on the rockfill dam and obtain the material parameters of the rockfill dam at the actual measuring points, such as the tangent modulus coefficient K, modulus exponent n, and failure ratio R at the actual measuring points. f Initial internal friction angle Changes in bulk modulus exponent m and friction angle and bulk modulus parameter K b The EB model parameters at the actual measuring points on the rockfill dam were obtained. All parameters in the EB model parameters were set with increments of ±10% to 30% to obtain the parameter combination set.

[0014] Step 1.3: Perform sensitivity analysis on the parameter combination set using orthogonal experiments to obtain the rockfill dam response values ​​for each parameter; and set up three experimental groups using the parameter combination set, including negative increment experimental group, normal experimental group and positive increment experimental group. Input the parameter combinations in the parameter combination set of the negative increment experimental group, normal experimental group and positive increment experimental group into the corresponding nodes of the finite element rockfill dam calculation model to perform finite element analysis calculation to obtain the rockfill dam response values ​​for each parameter;

[0015] Step 1.4: Analyze the response values ​​of the rockfill dam using the range method and calculate the range values ​​of each parameter group;

[0016] Step 1.5: Sort the calculated range values ​​in descending order according to the size of the data to obtain a descending sequence of range value data. Extract the data with the largest range value from the descending sequence of range value data, and extract the first 40% of the data in the descending sequence of range value data. This will give us the parameters of the Duncan-Zhang EB model that are highly sensitive to the response of rockfill dams, which are the parameters to be inverted.

[0017] Preferably, in step 1.4, the calculation process of the range method is as follows:

[0018]

[0019] R j =A j,max -A j,min

[0020] Among them, A ij Let K1 be the average value of the rockfill dam response value j at the i-th level, and H be the number of tests conducted on the rockfill dam response value j at the i-th level. n The response value j of the rockfill dam is the result of the nth test at the i-th level. R is the average of all test results for the rockfill dam response value j. j Let A be the range of the experimental results for factor j. j,max A is the maximum value of the rockfill dam response value j obtained through multiple experiments for the same parameter. j,min The minimum value of the rockfill dam response value j obtained through multiple experiments for the same parameter.

[0021] Preferably, the step of segmenting the parameters to be inverted and training and validating the XGBoost model based on the segmentation results includes:

[0022] Step 2.1: Based on the parameters to be inverted, data parameters are sampled and combined using Latin hypercube sampling to generate 83 sets of parameter samples. Each parameter sample is input into the finite element model to calculate the corresponding settlement response value. The parameter samples are combined with the corresponding settlement response values ​​to obtain a sample set containing parameter combinations and settlement responses. The sample set is divided into a training set and a test set in a ratio of 8:2.

[0023] Step 2.2: Perform normalization preprocessing on the training set and the test set to obtain a normalized training set and a normalized test set;

[0024] Step 2.3: Use the Crowned Porcupine optimization algorithm to traverse the parameter space, and combine it with ten-fold cross-validation to evaluate the performance of each parameter combination in each normalized training set, and train to obtain the optimal parameters of XGBoost;

[0025] Step 2.4: Input the training data subset into the XGBoost model for training to obtain the XGBoost training model, and calculate the prediction error by inputting the validation data subset into the XGBoost training model. Based on the global search optimization mechanism, traverse the hyperparameter space of the XGBoost training model to dynamically generate candidate hyperparameter combinations.

[0026] Step 2.5, repeat step 2.4 until each data subset in the multiple data subsets is used as a validation data subset and input into the corresponding trained XGBoost training model for calculation, and calculate the mean squared error or mean absolute error of the prediction error.

[0027] Compare the mean squared error or the mean absolute error with the prediction error respectively, select the hyperparameter combination corresponding to the prediction error that is closest to the mean squared error or the mean absolute error among all prediction errors as the optimal hyperparameter combination, and replace the corresponding parameters in the XGBoost training model with the optimal hyperparameter combination to obtain the optimal XGBoost model.

[0028] Step 2.6: Input the normalized test set into the XGBoost optimal model, calculate the predicted value, and then perform denormalization on the predicted value to restore it to the actual physical settlement value.

[0029] Step 2.7: Calculate the verification error based on the predicted value, compare the verification error with the error threshold, and determine the inversion model of the material parameters of the rockfill dam.

[0030] Preferably, in step 2.2, the normalization preprocessing is as follows:

[0031]

[0032] Among them, X norm Here, x represents the normalized value, and x is the parameter to be inverted in the EB model parameter combination. max X is the maximum value of the parameter variable to be inverted in the parameter combination of the EB model. max This represents the maximum value of the parameter variable to be inverted in the parameter combination of the EB model.

[0033] Preferably, in step 3.1, the denormalization process is as follows:

[0034] y = y norm ×(y max -y min )+y min

[0035] Where y is the settlement value after inverse normalization, y norm The normalized predicted value is y. min and y max These are the minimum and maximum values ​​of the original settlement, respectively.

[0036] Preferably, the verification error is calculated based on the predicted value, and the verification error is compared with the error threshold to determine the inversion model of the rockfill dam material parameters, including:

[0037] The verification error is calculated based on the predicted value and the settlement value, and then compared with a preset error threshold. The verification error includes mean absolute error, mean absolute percentage error, root mean square error, and goodness of fit. The error threshold includes the coefficient error threshold and the root mean square error threshold.

[0038] Specifically, the determination coefficient is compared with the coefficient error threshold, and the root mean square error is compared with the root mean square error threshold.

[0039] If the determination coefficient is less than the coefficient error threshold or the root mean square error is greater than the root mean square error threshold, then the XGBoost optimal model is determined not to be an effective surrogate model. Steps 2.3 to 2.6 are repeated to re-optimize the hyperparameter combination of the XGBoost model and verify its accuracy until the verification error of the rockfill dam material parameter inversion training model reaches the preset error threshold.

[0040] If the determination coefficient is greater than or equal to the coefficient error threshold, and the root mean square error is less than or equal to the root mean square error threshold, then the XGBoost optimal model is determined to be an effective surrogate model, and the inversion model of the material parameters of the rockfill dam is obtained.

[0041] Preferably, the step of iteratively optimizing the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model using the Crowned Porcupine optimization algorithm based on a defined fitness function with mean square error as the objective, and retaining the optimal parameter combination to obtain the inversion result, includes:

[0042] Step 3.1: The global search optimization mechanism generates the combination of primary and secondary EB model parameters based on the data in the parameters to be inverted, according to the preset constraints.

[0043] Step 3.2: Input the combined EB model parameters of the primary and secondary rockfill areas into the rockfill dam material parameter inversion model to calculate the corresponding predicted settlement displacement;

[0044] Step 3.3: Obtain the measured settlement displacement, calculate the error between the measured settlement displacement and the predicted settlement displacement using the fitness function of the global search optimization mechanism, and obtain the fitness value;

[0045] Step 3.4: Obtain the measured settlement displacement, define a fitness function with mean square error as the objective, and calculate the error between the measured settlement displacement and the predicted settlement displacement using the Crown Porcupine optimization algorithm to obtain the fitness value.

[0046] Step 3.5: Compare the current fitness values ​​of the EB model parameter combinations for the primary and secondary rockfill areas with the historical best fitness values:

[0047] If the current fitness value is better than the historical best fitness, then the historical best model parameter combination is replaced with the primary and secondary rockfill EB model parameter combination corresponding to the current fitness value, and the historical best fitness value is replaced with the current fitness value.

[0048] If the current fitness value is not better than the historical best fitness, then retain the historical best fitness value and the model parameter combination corresponding to the historical best fitness value to obtain the best parameter combination and the best fitness value;

[0049] Step 3.6: Obtain the optimal parameter combination and the optimal fitness value. Input the parameters to be inverted, the optimal parameter combination, and the optimal fitness value into the rockfill dam material parameter inversion model to obtain the final settlement displacement and the inversion result.

[0050] Preferably, in step 3.1, the constraint condition is:

[0051]

[0052] Where: x m This is the m-th parameter to be inverted; The lower limit of the parameters to be inverted. M represents the upper limit of the parameters to be inverted, and M represents the total number of parameters to be inverted.

[0053] In step 4.3, the fitness function is:

[0054]

[0055] In the formula: u is the measured settlement displacement at the i-th point; i (x) represents the predicted settlement displacement at the i-th point; x = (x1, x2, ..., xk)T represents the parameters to be inverted for the rockfill dam material, and n represents the number of measurement points.

[0056] This invention proposes a single-objective inversion system for material parameters of rockfill dams based on CPO-XGBoost. The system applies the aforementioned single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost, including:

[0057] The sensitivity analysis module is configured to perform sensitivity analysis on the parameters of the Duncan-Chang EB model in the finite element rockfill dam model to obtain the parameters to be inverted.

[0058] The model training and validation module is configured to segment the parameters to be inverted, and based on the segmentation results, train and validate the constructed XGBoost model to obtain the inversion model of the material parameters of the rockfill dam.

[0059] The inversion module is configured to iteratively optimize the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model based on a defined fitness function with mean square error as the objective, and retain the optimal parameter combination to obtain the inversion result.

[0060] The verification module is configured to input the inversion parameters into the finite element model to verify the consistency between the calculated settlement and the measured data.

[0061] Compared with the prior art, the present invention has the following beneficial technical effects:

[0062] This invention proposes a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost. This method constructs a finite element model of the rockfill dam and performs sensitivity analysis on parameter samples to accurately screen key parameters that significantly contribute to the dam's response, significantly reducing the dimensionality of the inversion problem. It can accurately identify parameters that significantly affect the deformation of the rockfill dam, ensuring that the inversion process optimizes the key parameters that truly affect dam deformation, thereby improving the accuracy and efficiency of the inversion. Then, by training an XGBoost surrogate model, the computation time can be significantly reduced while maintaining high accuracy, replacing the traditional time-consuming finite element calculation, making the inversion process more efficient. Simultaneously, based on CPO, the optimal solution can be quickly found during the parameter search process, ensuring the accuracy and reliability of the inversion results. This allows for accurate prediction and evaluation of the deformation behavior of rockfill dams, thereby reducing engineering risks and improving the efficiency and reliability of the design.

[0063] Furthermore, this method constructs a finite element model of a rockfill dam based on measured data, ensuring a high degree of consistency between the model and the actual situation. This allows the model to more accurately reflect the mechanical behavior and deformation characteristics of the rockfill dam. Sensitivity analysis is performed on the mechanical material parameters of the rockfill dam, and orthogonal experimental design and range method are used to screen out key parameters that have a significant impact on dam deformation, improving the efficiency and accuracy of the inversion. The range method is used to quantify the degree of influence of parameters on rockfill dam deformation, screening out key parameters that have a significant impact on deformation, ensuring the accuracy of the parameters to be inverted. At the same time, a reasonable range of values ​​is determined based on the maximum and minimum values ​​of the parameters, providing a scientific search space for parameter inversion and avoiding inversion failures or low efficiency caused by excessively large or small value ranges.

[0064] Furthermore, this method acquires the geometric data, material distribution data, and measured deformation data of the rockfill dam, and inputs this data into finite element software for simulation. This ensures that the constructed finite element rockfill dam model is highly consistent with the actual situation, improving the model's accuracy. By performing finite element analysis on the 3D model of the rockfill dam, the stress, strain, and deformation of the dam under various working conditions are accurately calculated. Simultaneously, the locations of the finite element analyses are marked on the 3D model, forming a finite element mesh, further improving the accuracy and reliability of the analysis. By comparing the actual deformation data with the simulation results, engineers can identify shortcomings in the design and make corresponding optimizations, reducing engineering risks and improving the rationality and economy of the design.

[0065] Furthermore, this method replaces traditional finite element method (FEM) calculations with an XGBoost surrogate model, significantly improving computational efficiency. The XGBoost model, optimized based on the training dataset, accurately reflects the complex nonlinear relationship between rockfill dam deformation and material parameters. Simultaneously, CPO is used to optimize the hyperparameters of the XGBoost model, and ten-fold cross-validation is used to evaluate model performance, ensuring the stability and reliability of the surrogate model under different parameter combinations. Error validation of the surrogate model (e.g., MAE, MAPE, RMSE, R-squared) is then performed. 2 (etc.) to ensure that its prediction accuracy meets engineering requirements and to provide reliable technical support for parameter inversion.

[0066] Furthermore, the CPO algorithm in this method, by traversing the parameter space and combining it with a fitness function to evaluate the merits of parameter combinations, can effectively search for the optimal or near-optimal solution within a large parameter space, ensuring the accuracy and reliability of the inverted parameters. The iterative optimization strategy and dynamic fitness update mechanism in the algorithm enable it to adaptively adjust its search direction and strategy according to the dynamic changes of the optimization problem, helping to improve the robustness and effectiveness of the algorithm in complex parameter inversion problems. By comparing the current fitness value with the existing best fitness value, the optimal parameter combination is dynamically updated, ensuring that the final output inversion parameters have high accuracy and reliability. Attached Figure Description

[0067] Figure 1 A flowchart of a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost provided by the present invention;

[0068] Figure 2 A schematic diagram of a three-dimensional finite element model of a rockfill dam in a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost provided by the present invention;

[0069] Figure 3A schematic diagram of the target monitoring location of a rockfill dam in a single-objective inversion method for material parameters of a rockfill dam based on CPO-XGBoost provided by this invention;

[0070] Figure 4 A comparison chart of ES5 survey line inversion results in a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost provided by this invention;

[0071] Figure 5 A comparison chart of ES4 survey line inversion results in a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost provided by this invention;

[0072] Figure 6 A comparison of ES1 survey line inversion results in a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost provided by this invention. Detailed Implementation

[0073] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0074] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0075] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0076] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a communication connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0077] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0078] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0079] This invention proposes a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost, such as... Figure 1 As shown, it includes the following steps:

[0080] Step 1: Based on the sensitivity analysis of the Duncan-Chang EB model parameters in the finite element rockfill dam model, the parameters to be inverted are obtained.

[0081] Specifically, it includes:

[0082] Step 1.1: Construct a finite element model of the rockfill dam based on the acquired engineering data;

[0083] Specifically, the engineering data of the rockfill dam is obtained, including the drawings of the rockfill dam and the deformation monitoring layout diagram, as well as the material zoning, filling construction and settlement monitoring data of the rockfill dam. Considering the material zoning and the graded procedure of dam construction, the above data is input into the finite element software ABAQUS for simulation, simulating the construction loading process, and the simulation obtains the three-dimensional model of the rockfill dam, thus constructing the rockfill dam model.

[0084] Corresponding to the locations where settlement meters are installed in the rockfill dam, finite element nodes are arranged at the corresponding positions in the rockfill dam model. A finite element mesh is formed through these nodes, resulting in the finite element calculation model of the rockfill dam, as shown below. Figure 2 As shown.

[0085] Step 1.2, determine the actual measuring points on the rockfill dam, such as... Figure 3 As shown, the material parameters of the rockfill dam at the actual measuring points are obtained, such as the tangent modulus coefficient K, modulus exponent n, and failure ratio R at the actual measuring points. f Initial internal friction angle Changes in bulk modulus exponent m and friction angle and bulk modulus parameter K b We obtained the parameters of the Duncan-Chang EB model at the actual measuring points on the rockfill dam. We set the same positive and negative increments for all parameters in the Duncan-Chang EB model, that is, increments of positive and negative 10% to 30%, to obtain the parameter combination set.

[0086] Among them, the tangent modulus coefficient K reflects the material stiffness; the modulus exponent n controls the nonlinearity of the initial tangent modulus as a function of confining pressure; and the failure ratio R... f (R f <1), representing the ratio of the deviatoric stress at failure to the ultimate deviatoric stress; initial internal friction angle. The friction angle of the soil before shearing reflects the frictional characteristics between particles; the bulk modulus exponent m controls the nonlinearity of the bulk modulus change with confining pressure; the change in friction angle... The internal friction angle is expressed as an increment of plastic strain or stress state; bulk modulus parameter K. b Used to calculate bulk modulus.

[0087] Step 1.3: Perform sensitivity analysis on the parameter combination set using orthogonal experiments to obtain the rockfill dam response values ​​for each parameter; and set up three experimental groups using the parameter combination set, including negative increment experimental group, normal experimental group and positive increment experimental group. Input the parameter combinations in the parameter combination set of the negative increment experimental group, normal experimental group and positive increment experimental group into the corresponding nodes of the finite element rockfill dam calculation model to perform finite element analysis calculation to obtain the rockfill dam response values ​​for each parameter;

[0088] Step 1.4: Analyze the response values ​​of the rockfill dam using the range method to calculate the range values ​​of each parameter group; the larger the range value, the higher the sensitivity of the parameter to the response of the rockfill dam.

[0089] The calculation process of the range method is as follows:

[0090]

[0091] R j =A j,max -A j,min

[0092] Among them, A ijLet K1 be the average value of the rockfill dam response value j at the i-th level, and H be the number of tests conducted on the rockfill dam response value j at the i-th level. n The response value j of the rockfill dam is the result of the nth test at the i-th level. R is the average of all test results for the rockfill dam response value j. j Let A be the range of the experimental results for factor j. j,max A is the maximum value of the rockfill dam response value j obtained through multiple experiments for the same parameter. j,min The minimum value of the rockfill dam response value j obtained through multiple experiments for the same parameter.

[0093] Step 1.5: Sort the calculated range values ​​in descending order of data size to obtain a descending sequence of range values. Extract the data with the largest range from the descending sequence, and extract the top 40% of the data from the descending sequence. By analyzing and summarizing the data types, extract the three data points with the highest proportion, including the bulk modulus parameter K. b The tangent modulus coefficient K and the bulk modulus exponent m account for a large proportion of the data, so the bulk modulus parameter K is extracted. b The tangent modulus coefficient K and the bulk modulus exponent m are used as key parameters for the main rockfill area and the secondary rockfill area to obtain the Duncan-Chang EB model parameters that are highly sensitive to the response of rockfill dams, that is, to obtain the parameters to be inverted.

[0094] Step 2: Divide the parameters to be inverted into segments. Based on the segmentation results, train and validate the XGBoost model to obtain the inversion model of the material parameters of the rockfill dam.

[0095] Specifically, in step 2.1, based on the parameters to be inverted obtained in step 1.5, data parameters are sampled and combined using Latin hypercube sampling (LHS) to generate 83 sets of parameter samples. Each parameter sample is input into the finite element model to calculate the corresponding settlement response value. The parameter samples are combined with the corresponding settlement response values ​​to obtain a sample set containing parameter combinations and settlement responses. The sample set is divided into a training set and a test set in an 8:2 ratio to obtain the training set and the test set.

[0096] Step 2.2: Perform normalization preprocessing on the training set and test set, converting the EB model parameter combinations in the training set and test set into a normalized training set and a normalized test set with a unified format and a unified data range.

[0097] The unified data range is [0, 1].

[0098] The normalization preprocessing process is as follows:

[0099]

[0100] Among them, X norm Here, x represents the normalized value, and x is the parameter to be inverted in the EB model parameter combination. max X is the maximum value of the parameter variable to be inverted in the parameter combination of the EB model. max This represents the maximum value of the parameter variable to be inverted in the parameter combination of the EB model.

[0101] Step 2.3: Use the Crowned Porcupine optimization algorithm to traverse the parameter space, and combine it with ten-fold cross-validation to evaluate the performance of each parameter combination in each normalized training set, and train to obtain the optimal parameters of XGBoost;

[0102] Specifically, initialize the relevant parameters of the Crowned Hog optimization algorithm, including population size, maximum number of iterations, etc., divide the normalized training set into ten data subsets, select one data subset as the validation data subset in turn, and use the rest as the training data subsets;

[0103] Step 2.4: Input a subset of training data into the XGBoost model for training to obtain the XGBoost training model. Calculate the prediction error by inputting a subset of validation data into the XGBoost training model. Based on the Crowned Porcupine Optimization Algorithm (CPO), traverse the hyperparameter space of the XGBoost training model to dynamically generate candidate hyperparameter combinations.

[0104] In each round of training with a subset of training data and a subset of validation data, the Crowned Porcupine Optimization Algorithm (CPO) comes into play. The CPO Optimization Algorithm explores and develops the hyperparameter space by simulating four defensive behaviors of the crested porcupine (visual, auditory, olfactory, and physical attacks). In the exploration phase, using visual and auditory strategies, the CPO Optimization Algorithm investigates different regions, performs a global search, and generates a series of different combinations of XGBoost hyperparameters (such as learning rate, maximum tree depth, regularization term, etc.). These hyperparameter combinations are applied to the training of the XGBoost model based on the subset of training data.

[0105] Step 2.5: Repeat step 2.4 until each data subset in the multiple data subsets is used as a validation data subset and input into the corresponding trained XGBoost training model for calculation. Calculate the mean squared error (MSE) or mean absolute error (MAE) of the prediction error. Compare the MSE or MAE with the prediction error. Select the hyperparameter combination corresponding to the prediction error with the closest MSE or MAE among all prediction errors as the optimal hyperparameter combination. Replace the corresponding parameters in the XGBoost training model with the optimal hyperparameter combination to obtain the optimal XGBoost model.

[0106] Step 2.6: Input the normalized test set into the XGBoost optimal model to calculate the predicted values. Then, denormalize the predicted values ​​to restore them to the actual physical settlement values, such as... Figures 4-6 As shown;

[0107] The inverse normalization process is as follows:

[0108] y = y norm ×(y max -y min )+y min

[0109] In the formula, y is the settlement value after inverse normalization. norm The normalized predicted value is y. min and y max These are the minimum and maximum values ​​of the original settlement, respectively.

[0110] Step 2.7: Calculate the verification error based on the predicted and settlement values. Compare the verification error with the error threshold to determine the inversion model for the material parameters of the rockfill dam. The verification error includes mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and goodness of fit (R²). 2 );

[0111] Specifically, the verification error is calculated based on the predicted values, including:

[0112] The mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), and goodness of fit (R²) were calculated from the predicted values ​​and settlement values, respectively. 2 ):

[0113] The calculation process for Mean Absolute Error (MAE) is as follows:

[0114]

[0115] Among them, y i For the i-th settlement value, Let be the i-th predicted value, and n be the number of samples. The smaller the MAE, the smaller the prediction error of the material parameter inversion model for rockfill dams.

[0116] The calculation process for Mean Absolute Percentage Error (MAPE) is as follows:

[0117]

[0118] Among them, y i For the i-th settlement value, Let n be the i-th predicted value and n be the number of samples. The smaller the MAPE, the smaller the relative error of the inversion model of the material parameters of the rockfill dam.

[0119] The calculation process for the root mean square error (RMSE) is as follows:

[0120]

[0121] Among them, y i For the i-th settlement value, Let be the i-th predicted value, and n be the number of samples. The smaller the RMSE, the higher the prediction accuracy of the rockfill dam material parameter inversion model.

[0122] Goodness of fit (R) 2 The calculation process for ) is as follows:

[0123]

[0124] Among them, y i For the i-th settlement value, Let be the i-th predicted value, n be the number of samples, and Ri be the predicted value. 2 The closer the value is to 1, the stronger the explanatory power of the rockfill dam material parameter inversion model for data variability.

[0125] Comparing the verification error with a preset error threshold involves using the coefficient of determination (R²) to determine the error. 2 The coefficient of determination (R) is compared with the coefficient error threshold of 0.98, and the root mean square error (RMSE) is compared with the root mean square error threshold of 2.0 cm. 2 If the coefficient error is less than the threshold of 0.98 or the root mean square error (RMSE) is greater than the threshold of 2.0 cm, then the optimal XGBoost model is determined not to be an effective surrogate model. Steps 2.3 to 2.7 are repeated to re-optimize the hyperparameter combination of the XGBoost model and verify its accuracy until the verification error of the rockfill dam material parameter inversion training model reaches the preset error threshold. If the coefficient of determination (R²) is less than the coefficient error threshold of 0.98 or greater than the root mean square error (RMSE) threshold of 2.0 cm, then the optimal XGBoost model is determined not to be an effective surrogate model. Steps 2.3 to 2.7 are repeated to re-optimize the hyperparameter combination of the XGBoost model and verify its accuracy until the verification error of the rockfill dam material parameter inversion training model reaches the preset error threshold. 2 If the coefficient error is greater than or equal to the threshold of 0.98, and the root mean square error (RMSE) is less than or equal to the threshold of 2.0 cm, then the XGBoost optimal model is determined as the effective surrogate model, and the inversion model of the material parameters of the rockfill dam is obtained.

[0126] Step 3: Based on the defined fitness function with mean squared error (MSE) as the objective, the Caucasian porcupine optimization algorithm (CPO) is used to iteratively optimize the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model. The optimal parameter combination is retained to obtain the inversion result.

[0127] Specifically, in step 3.1, the global search optimization mechanism (CPO) generates the combination of primary and secondary rockfill area EB model parameters based on the data in the parameters to be inverted, according to the preset constraints.

[0128] The constraints are as follows:

[0129]

[0130] Where: x m This is the m-th parameter to be inverted; The lower limit of the parameters to be inverted. is the upper limit of the parameters to be inverted, and M is the total number of parameters to be inverted.

[0131] Step 3.2: Input the combined EB model parameters of the primary and secondary rockfill areas into the rockfill dam material parameter inversion model to calculate the corresponding predicted settlement displacement;

[0132] Step 3.3: Obtain the measured settlement displacement, define a fitness function with mean squared error (MSE) as the objective, and calculate the error between the measured settlement displacement and the predicted settlement displacement using the Caucasian Pig Optimization Algorithm (CPO) to obtain the fitness value;

[0133] The fitness function is:

[0134]

[0135] In the formula: u is the measured settlement displacement at the i-th point; i (x) represents the predicted settlement displacement at the i-th point; x = (x1, x2, ..., xk)T represents the parameters to be inverted for the rockfill dam material, and n represents the number of measurement points.

[0136] Step 3.4: Latin hypercube sampling is used to generate the initial population to ensure uniform coverage of the parameter space. The initial population is then subjected to adaptive screening, and individuals with fitness below the threshold are eliminated. The initial optimal parameter combination and its corresponding initial fitness are recorded. A population size decay coefficient (α = 0.95) is set, and population reduction is performed every 10 generations. Based on fitness ranking, the top k% of high-quality individuals are retained (k decreases linearly from 80 to 50). An elite replacement strategy is implemented for eliminated individuals, and the population is supplemented with variants of the historical optimal solution.

[0137] Individuals are randomly selected from the current population. A visual search mechanism is used to generate random perturbations in the parameter space. Based on the upper bound vector of the current parameter space, the lower bound vector of the current parameter space, and a random number uniformly distributed in the interval [-1,1], a new first candidate solution is generated. The specific process is as follows:

[0138] x 1new =x old +β·(x1max -x 1min rand(-1,1)

[0139] Where β is the exploration step size factor, which decreases linearly from 0.5 to 0.1 with the number of iterations, x old This is the parameter vector for the current individual (or a randomly selected individual). During the iteration process, x... old It is an individual randomly selected from the current population to generate the perturbation, x. 1new For the newly generated first candidate solution (parameter vector), by perturbating x old We get x 1max Let x be the upper bound vector of the current parameter space (the maximum value in each dimension). 1min The lower bound vector of the current parameter space (the minimum value of each dimension) is rand(-1,1), which is a random number (scalar) uniformly distributed in the interval [-1,1]. Each dimension generates a random number independently, so the perturbation is anisotropic.

[0140] Extract the parameter vector of the current best individual in the current population from the first candidate solution. Use a sound propagation mechanism to guide pheromone use based on the best individual in the population. Randomly select two individuals x from the current population. i1 and x j1 A second candidate solution is generated based on a uniformly distributed random number in the interval [0, 1] and the parameter vector of the current best individual. The specific process is as follows:

[0141] x 2new =x 1best +γ·(x i1 -x j1 rand(0,1)

[0142] Where γ is the exploration step size factor, which decreases linearly from 0.5 to 0.1 with the number of iterations, x 1best Let x be the parameter vector of the current best individual in the population (i.e., the parameter combination corresponding to the historical best fitness). 2new For the newly generated second candidate solution (parameter vector), x i1 and x j1 For each of the two individuals randomly selected from the population, rand(0,1) is a random number (scalar) uniformly distributed within the interval [0,1]. Each dimension is generated independently, controlling the weights of the difference vector.

[0143] Extract the parameter vector of the current best individual in the current population from the second candidate solution. Based on the random numbers in the current population under the standard normal distribution after obtaining the second candidate solution, generate the upper bound vector and the lower bound vector of the parameter space after generating the second candidate solution. Use the odor tracking mechanism to perform Gaussian perturbation near the current best solution to generate a new third candidate solution. The specific process is as follows:

[0144]

[0145] Where σ is the variation intensity, which decays exponentially from 0.3 to 0.05 with the number of iterations, x 2best Let x be the parameter vector of the current best individual in the population. 3new For the newly generated third candidate solution, N(0,1) is a random number (scalar) of a standard normal distribution (mean 0, standard deviation 1), generated independently for each dimension, producing Gaussian noise. 2max To generate the upper bound vector of the parameter space after the second candidate solution is generated, x 2min This is the lower bound vector of the parameter space after generating the second candidate solution.

[0146] Obtain the parameter vector of the third candidate solution, and the parameter vectors of two different individuals randomly selected from the population under the third candidate solution. Use a physical attack mechanism and a differential evolution strategy to generate new parameter vectors. The specific process is as follows:

[0147] x 4new =x i2 +F·(x j2 -x k )

[0148] Where F is the scaling factor, with a value range of [0.5, 1.0], x 4new For the newly generated parameter vector, x i2 Let x be the parameter vector of the current target vector. k x j2 The parameter vectors (j≠k≠i) of two randomly selected individuals in the population are used to construct the difference vector.

[0149] Calculate the fitness improvement rate Δf for 10 consecutive iterations. When Δf < ∈ (∈ is 0.01), trigger the transition from exploration to development and obtain the current fitness and historical best fitness values ​​of the EB model parameter combination for the primary and secondary rockfill areas.

[0150] The calculation process for the fitness improvement rate Δf is as follows:

[0151]

[0152] Among them, f t-10f is the historical best fitness value at iteration step t-10. t The historical best fitness value at iteration step t.

[0153] Step 3.5: Compare the current fitness values ​​of the EB model parameter combinations for the primary and secondary rockfill areas with the historical best fitness values:

[0154] If the current fitness value is better than the historical best fitness, then the historical best model parameter combination is replaced with the primary and secondary rockfill EB model parameter combination corresponding to the current fitness value, and the historical best fitness value is replaced with the current fitness value.

[0155] If the current fitness value is not better than the historical best fitness, then retain the historical best fitness value and the model parameter combination corresponding to the historical best fitness value to obtain the best parameter combination and the best fitness value;

[0156] Step 3.6: Obtain the optimal parameter combination and optimal fitness value. Input the parameters to be inverted, the optimal parameter combination, and the optimal fitness value into the rockfill dam material parameter inversion model to obtain the final settlement displacement, and obtain the inversion result, such as... Figures 4-6 As shown, the inversion parameters are input into the finite element model to verify the consistency between the calculated settlement and the measured data.

[0157] This invention also proposes a single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost. The method includes a sensitivity analysis module, a model training and verification module, an inversion module, and a verification module.

[0158] The sensitivity analysis module is configured to perform sensitivity analysis on the parameters of the Duncan-Chang EB model in the finite element rockfill dam model to obtain the parameters to be inverted.

[0159] The model training and validation module is configured to segment the parameters to be inverted, and based on the segmentation results, train and validate the constructed XGBoost model to obtain the inversion model of the material parameters of the rockfill dam.

[0160] The inversion module is configured to iteratively optimize the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model based on a defined fitness function with mean square error as the objective, and retain the optimal parameter combination to obtain the inversion result.

[0161] The verification module is configured to input the inversion parameters into the finite element model to verify the consistency between the calculated settlement and the measured data.

[0162] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0163] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost, characterized in that, Includes the following steps: Based on the sensitivity analysis of the Duncan-Chang EB model parameters in the finite element rockfill dam model, the parameters to be inverted are obtained. The parameters to be inverted are segmented, and the constructed XGBoost model is trained and validated based on the segmentation results to obtain the inversion model of material parameters of rockfill dam. Based on the defined fitness function with mean square error as the objective, the optimal parameter combination and the optimal fitness value are obtained by iterative optimization in the parameter space of the rockfill dam material parameter inversion model through the hog optimization algorithm. The optimal parameter combination is retained to obtain the inversion result. The inversion parameters were input into the finite element model to verify the consistency between the calculated settlement and the measured data.

2. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 1, characterized in that, The sensitivity analysis based on the Duncan-Chang EB model parameters in the finite element rockfill dam model includes: Step 1.1: Construct a finite element model of the rockfill dam based on the acquired engineering data; Step 1.2: Determine the actual measuring points on the rockfill dam and obtain the material parameters of the rockfill dam at the actual measuring points, such as the tangent modulus coefficient K, modulus exponent n, and failure ratio R at the actual measuring points. f Initial internal friction angle Changes in bulk modulus exponent m and friction angle and bulk modulus parameter K b The EB model parameters at the actual measuring points on the rockfill dam were obtained. All parameters in the EB model parameters were set with increments of ±10% to 30% to obtain the parameter combination set. Step 1.3: Perform sensitivity analysis on the parameter combination set using orthogonal experiments to obtain the rockfill dam response values ​​for each parameter; and set up three experimental groups using the parameter combination set, including negative increment experimental group, normal experimental group and positive increment experimental group. Input the parameter combinations in the parameter combination set of the negative increment experimental group, normal experimental group and positive increment experimental group into the corresponding nodes of the finite element rockfill dam calculation model to perform finite element analysis calculation to obtain the rockfill dam response values ​​for each parameter; Step 1.4: Analyze the response values ​​of the rockfill dam using the range method and calculate the range values ​​of each parameter group; Step 1.5: Sort the calculated range values ​​in descending order according to the size of the data to obtain a descending sequence of range value data. Extract the data with the largest range value from the descending sequence of range value data, and extract the first 40% of the data in the descending sequence of range value data. This will give us the parameters of the Duncan-Zhang EB model that are highly sensitive to the response of rockfill dams, which are the parameters to be inverted.

3. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 2, characterized in that, In step 1.4, the calculation process of the range method is as follows: R j =A j,max -A j,min Among them, A ij Let K1 be the average value of the rockfill dam response value j at the i-th level, and H be the number of tests conducted on the rockfill dam response value j at the i-th level. n The response value j of the rockfill dam is the result of the nth test at the i-th level. R is the average of all test results for the rockfill dam response value j. j Let A be the range of the experimental results for factor j. j,max A represents the maximum value of the rockfill dam response value j obtained through multiple experiments for the same parameter. j,min The minimum value of the rockfill dam response value j obtained through multiple experiments for the same parameter.

4. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 1, characterized in that, The step of segmenting the parameters to be inverted and training and validating the XGBoost model based on the segmentation results includes: Step 2.1: Based on the parameters to be inverted, data parameters are sampled and combined using Latin hypercube sampling to generate 83 sets of parameter samples. Each parameter sample is input into the finite element model to calculate the corresponding settlement response value. The parameter samples are combined with the corresponding settlement response values ​​to obtain a sample set containing parameter combinations and settlement responses. The sample set is divided into a training set and a test set in a ratio of 8:

2. Step 2.2: Perform normalization preprocessing on the training set and the test set to obtain a normalized training set and a normalized test set; Step 2.3: Use the Crowned Porcupine optimization algorithm to traverse the parameter space, and combine it with ten-fold cross-validation to evaluate the performance of each parameter combination in each normalized training set, and train to obtain the optimal parameters of XGBoost; Step 2.4: Input the training data subset into the XGBoost model for training to obtain the XGBoost training model, and calculate the prediction error by inputting the validation data subset into the XGBoost training model. Based on the global search optimization mechanism, traverse the hyperparameter space of the XGBoost training model to dynamically generate candidate hyperparameter combinations. Step 2.5, repeat step 2.4 until each data subset in the multiple data subsets is used as a validation data subset and input into the corresponding trained XGBoost training model for calculation, and calculate the mean squared error or mean absolute error of the prediction error. Compare the mean squared error or the mean absolute error with the prediction error respectively, select the hyperparameter combination corresponding to the prediction error that is closest to the mean squared error or the mean absolute error among all prediction errors as the optimal hyperparameter combination, and replace the corresponding parameters in the XGBoost training model with the optimal hyperparameter combination to obtain the optimal XGBoost model. Step 2.6: Input the normalized test set into the XGBoost optimal model, calculate the predicted value, and then perform denormalization on the predicted value to restore it to the actual physical settlement value. Step 2.7: Calculate the verification error based on the predicted value, compare the verification error with the error threshold, and determine the inversion model of the material parameters of the rockfill dam.

5. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 4, characterized in that, In step 2.2, the normalization preprocessing is as follows: Among them, X norm Here, x represents the normalized value, and x is the parameter to be inverted in the EB model parameter combination. max X is the maximum value of the parameter variable to be inverted in the parameter combination of the EB model. max This represents the maximum value of the parameter variable to be inverted in the parameter combination of the EB model.

6. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 4, characterized in that, In step 3.1, the inverse normalization process is as follows: y=y norm ×(and max -and min )+and min Where y is the settlement value after inverse normalization, y norm The normalized predicted value is y. min and y max These are the minimum and maximum values ​​of the original settlement, respectively.

7. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 4, characterized in that, The verification error is calculated based on the predicted values. The verification error is then compared with an error threshold to determine the inversion model for the material parameters of the rockfill dam, including: The verification error is calculated based on the predicted value and the settlement value, and then compared with a preset error threshold. The verification error includes mean absolute error, mean absolute percentage error, root mean square error, and goodness of fit. The error threshold includes the coefficient error threshold and the root mean square error threshold. Specifically, the determination coefficient is compared with the coefficient error threshold, and the root mean square error is compared with the root mean square error threshold. If the determination coefficient is less than the coefficient error threshold or the root mean square error is greater than the root mean square error threshold, then the XGBoost optimal model is determined not to be an effective surrogate model. Steps 2.3 to 2.6 are repeated to re-optimize the hyperparameter combination of the XGBoost model and verify its accuracy until the verification error of the rockfill dam material parameter inversion training model reaches the preset error threshold. If the determination coefficient is greater than or equal to the coefficient error threshold, and the root mean square error is less than or equal to the root mean square error threshold, then the XGBoost optimal model is determined to be an effective surrogate model, and the inversion model of the material parameters of the rockfill dam is obtained.

8. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 1, characterized in that, The fitness function, defined with mean squared error as the objective, is used to iteratively optimize the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model using the Crowned Porcupine optimization algorithm. The optimal parameter combination is then retained to obtain the inversion result, including: Step 3.1: The global search optimization mechanism generates the combination of primary and secondary EB model parameters based on the data in the parameters to be inverted, according to the preset constraints. Step 3.2: Input the combined EB model parameters of the primary and secondary rockfill areas into the rockfill dam material parameter inversion model to calculate the corresponding predicted settlement displacement; Step 3.3: Obtain the measured settlement displacement, calculate the error between the measured settlement displacement and the predicted settlement displacement using the fitness function of the global search optimization mechanism, and obtain the fitness value; Step 3.4: Obtain the measured settlement displacement, define a fitness function with mean square error as the objective, and calculate the error between the measured settlement displacement and the predicted settlement displacement using the Crown Porcupine optimization algorithm to obtain the fitness value. Step 3.5: Compare the current fitness values ​​of the EB model parameter combinations for the primary and secondary rockfill areas with the historical best fitness values: If the current fitness value is better than the historical best fitness, then the historical best model parameter combination is replaced with the primary and secondary rockfill EB model parameter combination corresponding to the current fitness value, and the historical best fitness value is replaced with the current fitness value. If the current fitness value is not better than the historical best fitness, then retain the historical best fitness value and the model parameter combination corresponding to the historical best fitness value to obtain the best parameter combination and the best fitness value; Step 3.6: Obtain the optimal parameter combination and the optimal fitness value. Input the parameters to be inverted, the optimal parameter combination, and the optimal fitness value into the rockfill dam material parameter inversion model to obtain the final settlement displacement and the inversion result.

9. The single-objective inversion method for material parameters of rockfill dams based on CPO-XGBoost according to claim 8, characterized in that, In step 3.1, the constraint condition is: Where: x m This is the m-th parameter to be inverted; The lower limit of the parameters to be inverted. M represents the upper limit of the parameters to be inverted, and M represents the total number of parameters to be inverted. In step 4.3, the fitness function is: In the formula: u is the measured settlement displacement at the i-th point; i (x) represents the predicted settlement displacement at the i-th point; x = (x1, x2, ..., xk)T represents the parameters to be inverted for the rockfill dam material, and n represents the number of measurement points.

10. A single-objective inversion system for material parameters of rockfill dams based on CPO-XGBoost, employing the single-objective inversion method for material parameters of rockfill dams based on any one of claims 1 to 9, characterized in that, include: The sensitivity analysis module is configured to perform sensitivity analysis on the parameters of the Duncan-Chang EB model in the finite element rockfill dam model to obtain the parameters to be inverted. The model training and validation module is configured to segment the parameters to be inverted, and based on the segmentation results, train and validate the constructed XGBoost model to obtain the inversion model of the material parameters of the rockfill dam. The inversion module is configured to iteratively optimize the parameter combination and the optimal fitness value in the parameter space of the rockfill dam material parameter inversion model based on a defined fitness function with mean square error as the objective, and retain the optimal parameter combination to obtain the inversion result. The verification module is configured to input the inversion parameters into the finite element model to verify the consistency between the calculated settlement and the measured data.

Images