Dam body bedrock operation parameter inversion method combining three-dimensional finite element and optimization algorithm

Abstract

The invention relates to the technical field of dam safety monitoring, and discloses a dam body bedrock operation parameter inversion method combining a three-dimensional finite element and an optimization algorithm, and the method comprises the following steps: S1, collecting and sorting related data such as geometric parameters of an area where a reservoir is located, carrying out the data preprocessing, building a historical database, and constructing a three-dimensional finite element model; s2, on the basis of the historical monitoring database, constructing a dam deformation monitoring statistical model; s3, separating a hydrostatic pressure deformation component from the prototype monitoring data by using a multiple linear regression model, performing hyper-parameter optimization on the proxy model by using an improved particle swarm optimization algorithm, and performing fitting prediction on a separation result; s4, evaluating the performance of the agent model; and S5, the downstream deformation of the dam body is analyzed, and safety evaluation is carried out. According to the method, multi-dimensional data are collected through combination of automatic observation equipment and manual observation, original deformation monitoring data are processed through soft threshold wavelet transform noise reduction, and the integrity and accuracy of the monitoring data are effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dam safety monitoring, in particular to a dam body bedrock operation parameter inversion method combining three-dimensional finite elements and optimization algorithms. BACKGROUND

[0002] In the field of water conservancy projects, concrete gravity dams, as important water retaining structures, their safety and stability are directly related to people's life and property safety and regional economic development. With the continuous development of water conservancy construction, a large number of concrete gravity dams have been in operation for many years. The performance of the dam structure will gradually change under the action of long-term water pressure, temperature changes, foundation deformation and other internal and external loads, and may have safety hazards such as cracks, leakage and deformation.

[0003] Currently, the safety monitoring and performance analysis of concrete gravity dams mainly face the following challenges: on the one hand, the traditional safety monitoring method often cannot comprehensively and accurately obtain the response data of the dam under the action of complex environment, the completeness and accuracy of the monitoring data are insufficient, which leads to the inability to accurately evaluate the actual performance of the dam; on the other hand, the existing analysis model is not perfect in describing the coupling effect of water pressure, temperature, time effect and other factors when dealing with the deformation mechanism of the dam, and there are problems such as high data quality requirements and complex calculation in the parameter inversion process, which makes it difficult to accurately identify the material constitutive model parameters of the dam, and further affects the reliability and effectiveness of the dam safety evaluation.

[0004] In addition, with the development of artificial intelligence technology, how to effectively combine it with traditional finite element analysis methods to build an efficient and accurate mapping proxy model to improve the utilization efficiency and analysis accuracy of dam safety monitoring data has become a technical problem to be solved in this field. SUMMARY

[0005] (I) Technical problems solved In view of the deficiencies of the prior art, the present application provides a dam body bedrock operation parameter inversion method combining three-dimensional finite elements and optimization algorithms, which solves the problem of "low work efficiency" in the above background art.

[0006] (II) Technical solutions In order to achieve the above purpose, the present application is implemented by the following technical solutions: a dam body bedrock operation parameter inversion method combining three-dimensional finite elements and optimization algorithms, comprising the following steps: S1, collect and arrange the hydrogeological, meteorological data, monitoring and dam body geometric parameters and other related data of the reservoir area, perform data preprocessing, establish a historical database, and construct a three-dimensional finite element model; S2, based on the historical monitoring database, construct a three-dimensional finite element model containing water pressure deformation components (εx, εy, εz), temperature deformation components (εx, εy, εz), time effect components (εx, εy, εz) and other factors; S3, input the monitoring data into the three-dimensional finite element model, and perform parameter inversion on the dam body bedrock operation parameters by using the optimization algorithm. ) and the time-dependent deformation component (TDC) of the classical dam deformation monitoring statistical model; S3, separating the hydrostatic deformation component from the prototype monitoring data by using a multiple linear regression model (MLR), optimizing the hyperparameters of the surrogate model by using an IPSO algorithm, and fitting and predicting the results by using a random forest algorithm (RFA) established surrogate model; S4, using R2, MAE and MSE as main indicators to evaluate the performance of the model; S5, inputting the inversion parameters into the three-dimensional finite element model to analyze the deformation of the dam along the water flow direction and evaluate the safety.

[0007] Preferably, S1 is specifically collecting horizontal displacement data perpendicular to the water flow direction by automatic observation equipment and manual observation means, selecting a soft threshold to denoise the original deformation monitoring data by wavelet transform, monitoring the upstream and downstream water levels by using water level gauges on the upstream and downstream of the dam, and monitoring the air temperature and rainfall by using thermometers and tipping bucket rain gauges arranged on the dam top. The monitoring items and layout standards of the concrete gravity dam refer to the Technical Standard for Concrete Dam Safety Monitoring (GBT51416-2020), a soft threshold is selected to denoise the original deformation monitoring data by wavelet transform, and a three-dimensional finite element model is constructed by using second-order tetrahedral isoparametric elements according to the geometric parameters of the dam.

[0008] Preferably, in S2, the water pressure deformation component, the temperature deformation component and the time-dependent deformation component are mainly selected as the dam deformation in the classical deformation monitoring model. Among them, linearly related to the 1-3 or 1-4 polynomial function of the water depth, i.e. , wherein, is the statistical coefficient; represents the upstream water depth of the dam, when the analysis object is a gravity dam, 3 or 4, In the HTT model, a plurality of average air temperature values in the early stage are used to calculate the temperature deformation component, i.e. , wherein is the statistical day, the average air temperature of the first i days, is the lag day, selecting a multi-term expression containing a logarithmic function, i.e. , wherein, , represents the total number of days from the initial monitoring day to the i-th day, and represent the statistical coefficients of the time-dependent deformation component.

[0009] Preferably, S3 is specifically: using a multiple linear regression model (MLR) to separate the accurate hydrostatic deformation component from the prototype monitoring data, so as to fit the deformation component caused by water pressure. The model is represented as: , wherein is the independent variable matrix, and the independent variable is a variable that can be independently and freely changed, is the dependent variable that is not independent and is affected by other variables, and here represents the observation value, is represented as a regression parameter matrix, and the parameter estimate is calculated according to the least squares method, and the multivariate function to be minimized for calculation is represented as: , the first-order partial derivative of each parameter is calculated, and it is equal to 0, and the estimated value is represented as , wherein X represents the sample observation value matrix, represents the transpose matrix of the sample observation value matrix.

[0010] Preferably, in S3, a random forest algorithm (RFA) is used to fit and predict the results, and in order to prevent large displacement measuring points from affecting the accuracy of small displacement measuring points, the data of the measuring points need to be uniformly normalized before fitting. The normalization calculation is as follows: , wherein is the normalized value, is the original observation value; and are the minimum value and the maximum value of the data sequence, respectively. The normalized data is divided into a training set, a validation set and a test set according to the ratio of 8:1:1. The hyperparameters of RFA are optimized by improving the particle swarm optimization algorithm (IPSO), and the formula is is the velocity of the i-th particle in the d-dimensional space in the k-th iteration, is the position of the i-th particle in the d-dimensional space in the k-th iteration, w is the inertia weight, c1 and c2 are learning factors, the particle population is set to 20, the inertia weight w is 1.0, and the learning factors c1 and c2 are both set to 2.0, and are random numbers between 0 and 1, is the individual optimal solution, is the global optimal solution.

[0011] Preferably, the FEM mapping agent model is constructed for prediction and fitting analysis with the hydrostatic pressure component separated from the prototype monitoring data, the point deformation feature point data are taken as input, the deformation is taken as output, the super parameter configuration is optimized based on the IPSO algorithm, the RFA is used to train multiple decision trees, the mapping relationship between the characteristic factor and the effect is constructed, and the agent finite element model is calculated. For the conventional elastic material, the response equation of the deformation and the material parameter is , and finally the prediction result is normalized to obtain the final prediction result, and the normalization calculation is .

[0012] Preferably, in the S4, the main indicators of model performance evaluation are as follows: wherein, is the model inversion value, is the actual value, and N is the total sample number Preferably, the S5 is specifically: inputting the parameters obtained by inversion into a three-dimensional finite element model, and focusing on analyzing the deformation of the dam body along the water flow direction. The dam deformation law is analyzed from the deformation change of the X direction of the water retaining dam section and the overflow dam section.

[0013] (Three) beneficial effects The dam body bedrock operation parameter inversion method combining three-dimensional finite element and optimization algorithm is provided. The following beneficial effects are provided: (1) The present application collects multi-dimensional data by combining automatic observation equipment with manual observation, adopts soft threshold wavelet transform to process the original deformation monitoring data, effectively guarantees the integrity and accuracy of the monitoring data, and at the same time, fuses a multiple linear regression model, an improved particle swarm optimization algorithm and a random forest algorithm to construct a model, realizes accurate separation of the dam deformation component, greatly improves the utilization efficiency and analysis accuracy of the monitoring data compared with the traditional monitoring method, and provides reliable data support for accurately evaluating the actual performance of the dam body.

[0014] (2) The present application constructs a classical dam deformation monitoring statistical model containing water pressure deformation component, temperature deformation component and aging deformation component, respectively uses an expression related to the water depth polynomial, an HTT model and a multiple expression containing a logarithmic function to describe each component, perfects the description of the dam deformation mechanism under the coupling action of multiple factors, optimizes the super parameters of the random forest algorithm through the improved particle swarm optimization algorithm, separates the static water pressure deformation component combined with the multiple linear regression model, reduces the high requirement of parameter inversion on data quality, simplifies the calculation process, successfully realizes accurate identification of the dam material constitutive model parameters, and improves the reliability and effectiveness of the dam safety evaluation.

[0015] (3) The improved particle swarm optimization algorithm is used to optimize the hyperparameters of the random forest algorithm to construct a mapping agent model, which is combined with a three-dimensional finite element model to input parameters into the three-dimensional finite element model to analyze the deformation of the dam body and perform safety evaluation; the combination mode fully gives play to the efficient data processing capacity of artificial intelligence technology and the structural analysis advantages of traditional finite element analysis, constructs an efficient and accurate mapping agent model, effectively improves the utilization efficiency and analysis accuracy of dam safety monitoring data, and solves the technical problems in the field of combination of artificial intelligence technology and traditional finite element analysis method. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 It is a flowchart of the inversion method of the present application; Figure 2 It is a schematic diagram of the measured horizontal deformation sequence of a certain dam measuring point of the present application; Figure 3 It is a schematic diagram of the separation result of the prototype monitoring data of a certain dam measuring point of the present application; Figure 4 It is a schematic diagram of the comparison between the model inversion and the measured results of the selected displacement measuring point of the present application; Figure 5 It is a schematic diagram of the X-direction deformation of the water retaining dam section (a) and the overflow dam section (b) of a certain dam of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0018] Please refer to Figure 1 - Figure 5 The present application provides an inversion method of dam bedrock operation parameters combined with three-dimensional finite element and optimization algorithm, which fuses multiple linear regression model, improved particle swarm optimization algorithm and random forest algorithm to construct a monitoring data separation and parameter inversion model, and realizes scientific evaluation of dam safety performance through three-dimensional finite element model, specifically: Taking a large concrete wide-slot gravity dam as an example, the dam crest is 209.12 meters long, the maximum dam height is 38.5 meters, and there are 16 dam sections, including the left bank non-overflow dam section (2#-8# dam section), the overflow dam section (9#-13# dam section), and the right bank non-overflow dam section (14#-17# dam section). The dam body is provided with a grouting and drainage gallery, and the lowest elevation of the gallery floor is 25 meters. First, data collection is carried out, and horizontal displacement data perpendicular to the water flow direction are obtained through automatic observation equipment and manual observation means, with a time range of February 2022 to May 2023. At the same time, water level gauges on the upstream and downstream of the dam are used to monitor water level changes, and thermometers and tipping bucket rain gauges are arranged on the crest of the 2# dam section to monitor air temperature and rainfall. In addition, 3 thermometers are arranged on the upstream side of the 7# dam section to monitor reservoir water temperature, and 3 vibrating wire thermometers are arranged on the downstream slope to monitor dam body temperature. In order to reduce the influence of noise, soft threshold is selected to carry out wavelet transform denoising on the original deformation monitoring data, db10 wavelet type is used, wavelet decomposition layer is set to 5, and threshold factor is 0.5. As shown in Figure 2 , after denoising, the horizontal deformation sequence of the measuring point presents a more stable change trend. According to the geometric parameters of the dam, a three-dimensional finite element model is constructed by using second-order tetrahedral isoparametric elements to provide a basis for subsequent analysis.

[0019] On the basis of establishing the historical database, a statistical model containing water pressure deformation component , temperature deformation component and time-dependent deformation component is constructed according to the classical deformation monitoring model. Among them, the water pressure deformation component is linearly related to the water depth with a 1-3 or 1-4 polynomial function, and the expression is , wherein is the statistical coefficient, represents the upstream water depth of the dam, and for a gravity dam, the highest degree is 4. The temperature deformation component is calculated by the HTT model, and the expression is , wherein is the statistical day, is the average temperature of the previous days, is the lag day, and the lag day is set to 30 days. The time-dependent deformation component selects a multi-term expression containing a logarithmic function, which is , wherein represents the total number of days from the initial monitoring day to the i-th day, and are statistical coefficients. Through these mathematical models, the complex deformation data is decomposed into multiple independent components, which is convenient for subsequent analysis.

[0020] In the process of separating the static water pressure deformation component, a multiple linear regression model (MLR) is used to fit the prototype monitoring data. The expression of the MLR model is where X is the independent variable matrix, Y is the dependent variable matrix, and β is the regression parameter matrix. The parameter estimation is based on the least square method, which aims to minimize the sum of squared errors By taking the first-order partial derivative of each parameter and setting it to zero, the regression parameter estimate is obtained as where X represents the sample observation value matrix, and XTrepresents the transpose matrix of the sample observation value matrix. By processing the prototype monitoring data of a certain measuring point through the model, the accurate hydrostatic pressure deformation component is separated, and the results are shown in Figure 3 The separated hydrostatic pressure component curve and the original reservoir water level curve show a high linear correlation, verifying the effectiveness of the model.

[0021] To further improve the prediction accuracy, the random forest algorithm (RFA) is used to fit and predict the separation results. Before fitting, to avoid the influence of large displacement measuring points on small displacement measuring points, the measuring point data needs to be normalized. The normalization formula is where is the normalized value, is the original observation value; and are the minimum and maximum values of the data sequence, respectively. The normalized data is divided into training set, validation set, and test set in the ratio of 8:1:1. The improved particle swarm optimization algorithm (IPSO) is used to optimize the hyperparameters of RFA, and the formula is is the velocity of the ithparticle in the dthdimension in the kthiteration, is the position of the ithparticle in the dthdimension in the kthiteration, w is the inertia weight, c1 and c2 are the learning factors, the particle population size is set to 20, the inertia weight w is taken as 1.0, and the learning factors c1 and c2 are both set to 2.0, and are random numbers between 0 and 1, is the individual optimal solution, is the global optimal solution. Based on the optimized hyperparameter configuration, multiple decision trees are trained through RFA to establish the mapping relationship between the characteristic factors and the effect size. Finally, the response equation of the proxy finite element model is where is the true elastic modulus, is the assumed material elastic modulus, and are the water pressure components calculated by the finite element method and actually measured, respectively. The prediction results are restored to the actual values through the inverse normalization formula .

[0022] Model performance evaluation adopts As a key indicator The calculation formula is , The calculation formula is , The calculation formula is ,in These are the values ​​retrieved from the model. The values ​​are actual values, and N is the total number of samples. Evaluation results show that the IPSO-RF model constructed in this invention achieves a maximum R² of 0.996, while the MSE and MAE are 0.0168 and 0.0128 respectively, both being the lowest values. This indicates that the finite element results calculated based on the optimized constitutive model parameters have high fitting accuracy with the MLR separation results. The optimal comprehensive elastic modulus of the dam body obtained from the inversion is 27.462 GPa, and the elastic modulus of the dam foundation is 19.779 GPa. Figure 4 The comparison between the reconstructed water pressure component corresponding to the optimal comprehensive elastic modulus obtained by inversion and the actual water pressure component is shown. In the inversion process of 30 working conditions, all inversion values ​​are within the predetermined range and no outliers appear, which further verifies the reliability and accuracy of the surrogate model.

[0023] The parameters obtained from the inversion are input into a three-dimensional finite element model to analyze the deformation characteristics of the dam body along the water flow direction. The study focuses on the deformation changes in the X-direction of the dam section and the spillway section, such as... Figure 5 As shown. From Figure 5 As can be seen from a, under the combined action of hydrostatic pressure and uplift pressure, the displacement in the X-direction of the dam section gradually increases with the dam height, exhibiting a layered distribution. The horizontal displacement of the dam body meets the basic requirement of being less than 1 / 10000 to 1 / 5000 of the dam height, with appropriate relaxation for medium and low dams. From Figure 5 As can be seen from b, the spillway section also exhibits similar deformation patterns under the same conditions. Furthermore, the bedrock of the dam body experienced a certain degree of settlement, inclined downstream, but the settlement was less than 1 / 20,000 of the dam height, meeting the basic requirements. The overall deformation pattern is consistent with the actual situation, verifying the scientific validity and practicality of the method of this invention in dam safety evaluation.

[0024] This invention, by integrating multiple advanced algorithms and three-dimensional finite element analysis technology, achieves precise separation of deformation components and efficient identification of material parameters in concrete gravity dams. A dam safety evaluation system is established, which can accurately reveal the deformation patterns of the dam and the operational performance of the bedrock. This provides a scientific basis for the safety status assessment of aging gravity dams, significantly enhances the dynamic monitoring capability of dam operation safety, and has important application value for ensuring the long-term stable operation of water conservancy projects.

[0025] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. A method for inverting the bedrock operating parameters of a dam body by combining three-dimensional finite element analysis and optimization algorithms, characterized in that, Includes the following steps: S1. Collect and organize hydrogeological and meteorological data, monitoring data and dam geometric parameters of the reservoir area, establish a historical database after data preprocessing, and construct a three-dimensional finite element model; S2. Based on the historical database, construct a statistical model for dam deformation monitoring that includes water pressure deformation components, temperature deformation components, and time-effect deformation components; S3. A multiple linear regression model is used to separate the hydrostatic deformation component from the prototype monitoring data. An improved particle swarm optimization algorithm is used to optimize the hyperparameters of the surrogate model. Then, the surrogate model constructed by the random forest algorithm is used to fit and predict the separation results. S4. The performance of the surrogate model is evaluated using the coefficient of determination, mean absolute error, and mean square error as indicators. S5. Input the inversion parameters that have passed the performance evaluation into the three-dimensional finite element model, analyze the deformation of the dam body in the direction of water flow, and conduct a safety evaluation.

2. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: The monitoring data in S1 includes horizontal displacement data perpendicular to the water flow direction, water level data upstream and downstream of the dam, temperature data, and rainfall data. The data preprocessing includes using soft thresholding to perform wavelet transform noise reduction on the original deformation monitoring data.

3. The method for inverting dam bedrock operating parameters combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: The water pressure deformation component in S2 is linearly related to the upstream water depth of the dam using a polynomial function of degree 1 to 4, expressed as follows: ,in, These are statistical coefficients; The upstream water depth of the dam is indicated. The temperature deformation component is calculated using the HTT model, and the expression is as follows: ,in To count the number of days, Average temperature of the previous i days The number of days is the lag.

4. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 3, characterized in that: The aging deformation component in S2 is described by a polynomial expression containing a logarithmic function, the expression being: ,in, , This represents the total number of days from the initial monitoring period to day i. and Statistical coefficients representing the age-related deformation components.

5. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: The expression for the multiple linear regression model in S3 is as follows: Where X is the independent variable matrix, Y is the dependent variable matrix, and β is the regression parameter matrix. The regression parameters are calculated using the least squares method, which requires minimizing the sum of squared errors and taking the first-order partial derivatives with respect to each parameter, then setting them to zero to obtain the estimated regression parameter values. Where X represents the sample observation matrix, This represents the transpose of the sample observation matrix.

6. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: Before using the random forest algorithm to fit and predict data in step S3, the measurement point data needs to be normalized. The normalization formula is as follows: ,in The value is the normalized value. These are the original observations; and These are the minimum and maximum values ​​of the data sequence, respectively. The normalized data is divided into training, validation, and test sets in an 8:1:1 ratio.

7. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: The speed update formula of the improved particle swarm optimization algorithm is: Let be the velocity of the i-th particle in the d-th dimension during the k-th iteration. It represents the position of the i-th particle in the d-dimensional space during the k-th iteration, where w is the inertia weight and c1 and c2 are learning factors. and A random number between [0,1] This is the optimal solution for the individual. This is the globally optimal solution.

8. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element method and optimization algorithm as described in claim 7, characterized in that: In the improved particle swarm optimization algorithm, the number of particles is set to 20, the inertia weight w is set to 1.0, and the learning factors c1 and c2 are both set to 2.

0. The random forest algorithm constructs the mapping relationship between feature factors and effect sizes by training multiple decision trees. The response equation of the surrogate finite element model is: ,in For the true elastic modulus, Assuming the elastic modulus of the material, and These are the water pressure components calculated using finite element analysis and those measured in practice.

9. The method for inverting dam bedrock operating parameters by combining three-dimensional finite element analysis and optimization algorithms according to claim 1, characterized in that: The deformation analysis of the dam body in the direction of water flow in S5 includes the deformation changes of the water-retaining dam section and the overflow dam section in the X direction; the safety evaluation needs to determine whether the horizontal displacement of the dam body meets the requirement of being less than 1 / 10000 to 1 / 5000 of the dam height, and whether the settlement of the bedrock of the dam body is less than 1 / 20000 of the dam height.

Images