Intelligent identification method for monitoring and simulation mutual feedback correction of concrete dam structure parameters

Abstract

The application discloses a method for monitoring and simulating mutual feedback correction of concrete dam structure parameter intelligent identification, belongs to the technical field of dam safety monitoring, and comprises the following steps: S1, a finite element model is constructed, and an LHS parameter set X is generated; S2, a water pressure component is calculated to obtain a data set Y, and a sample S=[X, Y] is normalized; S3, the training set and the test set are divided according to a ratio of 4:1, and a CBLA model is trained; S4, a water pressure component is separated by HST, and optimization is performed by combining ICPKO and CBLA; S5, iteration is terminated, and the optimal parameters are output, and finite element verification is performed. The method for monitoring and simulating mutual feedback correction of concrete dam structure parameter intelligent identification is adopted, the problem that a traditional algorithm is prone to local optimization is solved, the efficiency and precision of gravity dam structure mechanics parameter feedback are improved, the optimal combination of deformation modulus of each partition of a dam body and a dam foundation that is in line with actual conditions is successfully inverted, and a new technical means is provided for operation period gravity dam structure response analysis.

Description

Technical Field

[0001] This invention relates to the field of dam safety monitoring technology, and in particular to an intelligent identification method for concrete dam structural parameters based on the mutual feedback correction between monitoring and simulation. Background Technology

[0002] Accurately understanding the operational behavior of concrete dams is crucial for ensuring project safety. Obtaining the mechanical parameters of the dam body and bedrock and using them for structural analysis and the formulation of monitoring indicators are important means of assessing structural safety. However, under complex and variable environmental loads, the degradation of dam construction material properties leads to discrepancies between actual and design values ​​of its mechanical parameters. Currently, inversion analysis is performed using prototype dam observation data to obtain equivalent values ​​of mechanical parameters that are compatible with the service characteristics of the dam body and foundation, thereby enabling the tracking and monitoring of dam deformation behavior.

[0003] Currently, parameter inversion methods mainly fall into three categories: inverse method, direct method, and approximation method. The approximation method (also known as the surrogate model method) calculates the response output of the simulation model by pre-setting parameter samples, establishing an approximate mapping relationship between the parameter space and the response space to replace repetitive and time-consuming simulation model calculations. This method is widely used in engineering fields due to its high inversion efficiency. It is worth noting that when using the surrogate model method to invert the mechanical parameters of dam bodies and foundations, difficulties arise due to the large number of structural performance parameters to be inverted and the difficulty in constructing the surrogate model. Therefore, how to select the parameters to be inverted and the surrogate model suitable for the inversion problem is a research challenge in establishing a method for tracking and monitoring the deformation behavior of concrete arch dams. Mechanical parameter inversion is a spatial search problem involving multiple parameter combinations. Metaheuristic algorithms (such as genetic algorithms, particle swarm optimization, and gray wolf algorithms) are widely used in inversion problems due to their strong global search capabilities. Among these algorithms, genetic algorithms often encounter low computational efficiency and long processing times, while particle swarm optimization often suffers from premature convergence, low accuracy, and poor robustness when solving optimization problems of complex functions. In recent years, the Spotted Kingfisher (PKO) algorithm has been applied in many fields due to its advantages such as simple structure, ease of implementation, and strong robustness. However, the standard PKO algorithm lacks a global search mechanism and is prone to getting trapped in local optima in complex optimization problems.

[0004] Therefore, this invention provides an intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction, enabling efficient and accurate inversion of the deformation modulus of the dam body and foundation zones. This provides a new approach for the structural response analysis of gravity dams during operation and promotes the transformation of gravity dam mechanical parameter inversion from inefficient and coarse to efficient and accurate. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction, so as to solve the problems existing in the above-mentioned background technology.

[0006] To achieve the above objectives, this invention provides an intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction, comprising the following steps:

[0007] S1. Construct a finite element model based on engineering data and determine the range of values ​​for the parameters to be inverted. Construct the parameter set X using Latin hypercube sampling (LHS).

[0008] S2. Input the parameter set X into the finite element model, calculate the dam water pressure components corresponding to different water levels, and then obtain the water pressure deformation difference dataset Y under different parameter combinations, and obtain the calculation sample S=[X,Y]; in order to eliminate the influence of dimensions between data, the data is normalized.

[0009] S3. Construct an attention-enhanced hybrid neural network (CBLA) model that reflects the nonlinear relationship between the deformation modulus and structural response of each zone of the gravity dam and dam foundation. Divide the calculation sample S into a training set and a test set in a 4:1 ratio. Input the training set into the CBLA model for training.

[0010] S4. Construct the optimization objective function for deformation modulus inversion, use the ICPKO algorithm and the CBLA model to search for the mechanical parameters of the gravity dam to be determined, and update the parameter combination with the minimum fitness as the current optimal parameter.

[0011] S5. If the maximum number of iterations is reached, the optimization ends, the optimal parameters are output, the combined parameters obtained from the inversion are substituted into the finite element model for forward calculation, and the inversion results are evaluated.

[0012] Preferably, step S3 specifically includes:

[0013] S31. The samples obtained in step S2 are normalized using the Min-Max method to reduce the data to the range [0,1]. Then, the data samples are divided into training set and test set.

[0014] S32. Input the training set samples into the CNN model optimized based on the attention mechanism, extract local features through the convolutional kernel of the convolutional layer, and establish non-linear relationships using the ReLU activation function;

[0015] S33. Introduce a channel attention mechanism to enhance the feature extraction capability of CNNs. For the structural response feature F, calculate the channel attention feature. And refined features extracted through channel attention ;

[0016] S34, will By inputting into the BiLSTM layer, the state associations at any distance within the sequence are explicitly modeled through its gating mechanism, enabling the CBLA model to capture the globally relevant features in the structural response.

[0017] Preferably, in step S32, the convolutional layers of the CNN model extract local features using the following formula:

[0018] (1);

[0019] Where ReLU represents a nonlinear activation function. Indicates the current convolutional layer's th... The output of each convolutional kernel, This represents the input to the current convolutional layer. This represents the convolution operation. Indicates the corresponding scalar weight. This represents the corresponding scalar bias term.

[0020] Preferably, in step S33, the channel attention features The calculation formula is expressed as:

[0021] (2);

[0022] in, Indicates input features Perform global average pooling. Indicates input features Perform global max pooling. Indicates to The result of global average pooling. Indicates to The result of global max pooling. and The weights of the fully connected layers in a multilayer perceptron are represented by MLP, where MLP stands for Multilayer Perceptron. Indicates the activation function;

[0023] Refine features The calculation formula is expressed as:

[0024] (3).

[0025] Preferably, in step S34, the weighted features are input into the BiLSTM layer, and the state associations at arbitrary distances in the sequence are explicitly modeled through a gating mechanism, making the CBLA model more effective in capturing global features in the structural response. After the outputs from the forward LSTM and backward LSTM, the hidden layers of both are superimposed to jointly determine the output layer. The final output of BiLSTM can be expressed as:

[0026] (4);

[0027] (5);

[0028] (6);

[0029] in, , and Both represent activation functions. and These represent the LSTM hidden states during the forward and backward iterations, respectively. Indicates the weights of different levels. Representing refined features .

[0030] Preferably, in step S3, the accuracy of the CBLA model is measured by the relative error between the CBLA model and the finite element model. The calculated values ​​from the finite element model and the predicted values ​​from the CBLA model are selected to calculate the accuracy of the surrogate model. The calculation formula is expressed as follows:

[0031] (7);

[0032] in, and Let represent the learnable parameters of CNN and BiLSTM, respectively. This represents the total number of training samples. express The first result obtained by inputting the CBLA model Water pressure components at each measuring point; This represents the calculated value of the water pressure component in the corresponding finite element model.

[0033] Preferably, in step S4, when seeking the optimal deformation modulus inversion for the dam and foundation zones, the mechanical parameter inversion problem is transformed into an optimization problem by constructing an objective function. The objective function for optimization is expressed as:

[0034] (8);

[0035] in, This represents the elastic modulus of the dam body. This represents the deformation modulus of the dam foundation. Indicates the number of deformation measuring points. This indicates the number of times a monitoring point is monitored within the selected time period. Indicates the first under the action of reservoir water pressure Each measuring point is at The water pressure component in the statistical model at time. Represents the fitting coefficient. Indicates the height of the reservoir water level power term Indicates the first under the action of reservoir water pressure Each measuring point is at The CBLA model at time 1000 calculates the water pressure component value.

[0036] Preferably, in step S4, to address the problem that the PKO algorithm is prone to getting stuck in local optima, resulting in low inversion accuracy, the ICPKO algorithm uses Logistic-Tent mapping to make the population spatial distribution more uniform, and then introduces α-stable distribution to adjust the algorithm's global exploration and local optimization capabilities. Based on Levy flight and back learning to escape the local optimal solution of the algorithm, it is used to find the optimal solution and obtain the optimal combination of mechanical parameters of each zone of the dam and dam foundation.

[0037] Preferably, in step S4, the ICPKO algorithm expression is:

[0038] (9);

[0039] (10);

[0040] (11);

[0041] (12);

[0042] in, This represents the state value of an individual in the population during the current iteration. This represents the new individual state value in the next iteration. Indicates control parameters, For modulo operation; For random variables, To stabilize the index, To amplify parameters, For displacement parameters, The imaginary unit, For integration variables; This indicates that the i-th individual is at iteration step The position at that time This represents the random step size that follows a Lévy distribution; Indicates the iteration step The optimal position of the population at that time. This indicates the upper and lower bounds of the search for the parameters to be inverted. b Represents the learning factor. Indicates the iteration step as Time The position of each individual This represents the optimal position of the population during the historical iteration process.

[0043] Therefore, this invention employs the aforementioned intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction. It combines this with the Latin hypercube experimental method to generate deformation modulus samples, and calculates the water pressure components at measuring points under different combinations of deformation modulus samples based on a finite element model. Furthermore, by establishing a CBLA model reflecting the relationship between structural response and mechanical parameters, and using the ICPKO algorithm, it obtains the equivalent values ​​of the dam body and foundation deformation modulus. This invention improves the efficiency and accuracy of gravity dam structural mechanical parameter feedback by constructing a CBLA model and improving the PKO algorithm, providing a new technical means for the inversion of the deformation modulus of gravity dams and their foundation zones during operation.

[0044] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0045] Figure 1 This is a flowchart of the steps of the intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to the present invention.

[0046] Figure 2 This is an inversion flowchart of the intelligent identification method for concrete dam structural parameters based on the mutual feedback correction of monitoring and simulation in this invention;

[0047] Figure 3 This is a schematic diagram of the horizontal displacement values ​​of each measuring point in dam section #2 according to an embodiment of the present invention;

[0048] Figure 4 This is the finite element model of dam section #2 and the deformation modulus zoning diagram of the dam and dam foundation according to an embodiment of the present invention;

[0049] Figure 5 The graphs are curves showing the fitting process of the statistical model of horizontal displacement at each measuring point in dam section #2 of this invention. Among them, (a) is the fitting and residual analysis graph of horizontal displacement at measuring point PL5, (b) is the fitting and residual analysis graph of horizontal displacement at measuring point PL2, and (c) is the fitting and residual analysis graph of horizontal displacement at measuring point IP2.

[0050] Figure 6 The following are the component separation results of the measuring points of dam section #2 in this embodiment of the invention: (a) is the displacement component separation result of measuring point PL5, (b) is the displacement component separation result of measuring point PL2, and (c) is the displacement component separation result of measuring point IP2.

[0051] Figure 7 These are verification comparison diagrams of the water pressure component inversion results of dam section 2 in this embodiment of the invention. Among them, (a) is a comparison diagram of the relative water pressure components of PL5 measuring point, and (b) is a comparison diagram of the relative water pressure components of PL2 measuring point. Detailed Implementation

[0052] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0054] Example

[0055] This example uses a gravity dam located in southeastern China. The dam has a maximum height of 113 m, a crest elevation of 179 m, a length of 308.5 m, a normal reservoir water level of 173 m, and a check flood level of 177.8 m. The dam consists of six sections, numbered 1-6 from left to right bank. Sections 1, 2, 5, and 6 are non-overflow sections, while sections 3 and 4 are overflow sections. Considering the completeness and continuity of the monitoring data, and because the measuring points below one-third of the dam height are significantly constrained by the dam foundation, the left bank section 2 (the water-retaining dam section) is used for verification. Measuring points PL5, PL2, and IP2 on section 2 are selected. The horizontal displacement curves of these measuring points are shown below. Figure 3 As shown.

[0056] like Figures 1-2 As shown, the intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction mainly involves the following steps:

[0057] Step S1: Construct a finite element model based on engineering data, determine the value range of the parameters to be inverted, and construct a parameter set using Latin hypercube sampling (LHS). .

[0058] Based on the dam's geometric parameters and relevant design data, a three-dimensional finite element model was constructed, such as... Figure 4 As shown, the dam foundation extends 1.5 times the dam height along the upstream and downstream, left and right banks and deep into the riverbed. The model contains 9203 elements and 10355 nodes. Except for the parameters to be inverted, the mechanical parameters of the dam concrete and dam foundation are all taken from the design values. The specific parameters are shown in Table 1.

[0059] Table 1 Material parameters of different regions in the finite element model

[0060] ;

[0061] According to relevant test data, the range of values ​​for the elastic modulus of the dam body and the deformation modulus of the dam foundation are set as follows: E1∈[12.5,35.5] GPa, E2∈[15,35] GPa and E3∈[5,25] GPa.

[0062] Step S2: Input the parameter set X into the finite element model, calculate the dam water pressure components corresponding to different water levels, and then obtain the water pressure deformation difference dataset Y under different parameter combinations, and obtain the calculation sample S=[X,Y].

[0063] The parameter combination samples are generated by Latin hypercube sampling (LHS), and the water pressure components are calculated by inputting them into the finite element model to construct a sample set S=[X,Y]: the training set is set with 300 sets of parameter samples and the test set is set with 75 sets of parameter samples; the finite element models of the training set and the test set samples are input to calculate the water pressure components of each measuring point of the dam under different water levels, and then the water pressure deformation difference dataset Y under different parameter combinations is obtained.

[0064] A HST model was established using 2192 sets of horizontal displacement data from January 1, 2003 to December 31, 2008. Data before December 31, 2007 was used as the training set, and the remaining data was used as the test set. MLR was used to solve for the fitting coefficients of the HST model input factors. The statistical model fitting process curves for the horizontal displacement at each measuring point are shown below. Figure 5 As shown, the separation results of each component at the measuring point are as follows: Figure 6 As shown.

[0065] Step S3: Construct an attention-enhanced hybrid neural network (CBLA) model that reflects the nonlinear relationship between the deformation modulus and structural response of each zone of the gravity dam and its foundation. Divide the computational samples S into a training set and a test set in a 4:1 ratio. Input the training set into the CBLA model for training. The specific steps include:

[0066] Step S31: Use the Min-Max normalization method to reduce the data to the range [0, 1] of the samples obtained in step S2, and then divide the data samples into training set and test set.

[0067] Step S32: Input the training set samples into the CNN model optimized based on the attention mechanism, extract local features through the convolutional kernel of the convolutional layer, and use the ReLU activation function to establish a non-linear relationship between the input and output.

[0068] The convolutional layers of a CNN model extract local features using the following formula:

[0069] (1);

[0070] Where ReLU represents a nonlinear activation function. Indicates the current convolutional layer's th... The output of each convolutional kernel, This represents the input to the current convolutional layer. This represents the convolution operation. Indicates the corresponding scalar weight. This represents the corresponding scalar bias term.

[0071] Step S33: Introduce a channel attention mechanism. For the structural response feature F, calculate the channel attention feature. And refined features extracted through channel attention .

[0072] Channel attention features The calculation formula is expressed as follows:

[0073] (2);

[0074] in, Indicates input features Perform global average pooling. Indicates input features Perform global max pooling. Indicates to The result of global average pooling. Indicates to The result of global max pooling. and The weights of the fully connected layers in a multilayer perceptron are represented by MLP, where MLP stands for Multilayer Perceptron. Indicates the activation function;

[0075] Refine features The calculation formula is expressed as:

[0076] (3).

[0077] Step S34: Input the weighted features into the BiLSTM layer, explicitly model the state associations at arbitrary distances in the sequence through a gating mechanism, capture the global features of the structural response, and weight the features. After the outputs from the forward and backward LSTMs, the hidden layers of both layers are superimposed to jointly determine the output layer. The final output of BiLSTM is expressed as:

[0078] (4);

[0079] (5);

[0080] (6);

[0081] in, , and Both represent activation functions. and These represent the LSTM hidden states during the forward and backward iterations, respectively. Indicates the weights of different levels. That is, refine the features .

[0082] The accuracy of the CBLA model is measured by the relative error between the CBLA model and the finite element model. The accuracy of the surrogate model is calculated using both the calculated values ​​from the finite element model and the predicted values ​​from the CBLA model. The calculation formula is as follows:

[0083] (7);

[0084] in, and Let represent the learnable parameters of CNN and BiLSTM, respectively. This represents the total number of training samples. express The first result obtained by inputting the CBLA agent model Water pressure components at each measuring point; This represents the calculated value of the water pressure component in the corresponding finite element model.

[0085] Step S4: Construct the optimal objective function for deformation modulus inversion, use the ICPKO algorithm and the CBLA model to search for the mechanical parameters of the gravity dam to be determined, and update the parameter combination with the minimum fitness as the current optimal parameter.

[0086] In seeking the optimal deformation modulus inversion for the dam and foundation zones, the mechanical parameter inversion problem is transformed into an optimization problem by constructing an objective function. The objective function for optimization is expressed as:

[0087] (8);

[0088] in, This represents the elastic modulus of the dam body. This represents the deformation modulus of the dam foundation. Indicates the number of deformation measuring points. This indicates the number of times a monitoring point is monitored within the selected time period. Indicates the first under the action of reservoir water pressure Each measuring point is at The water pressure component in the statistical model at time. Represents the fitting coefficient. Indicates the height of the reservoir water level power term Indicates the first under the action of reservoir water pressure Each measuring point is at The CBLA model at time 1000 calculates the water pressure component value.

[0089] The ICPKO algorithm employs a Logistic-Tent mapping to homogenize the population spatial distribution, and then introduces an α-stable distribution to adjust the algorithm's global exploration and local optimization capabilities. Based on Lévy fly-through and back-learning to escape local optima, it is used to find the optimal combination of mechanical parameters for each zone of the dam and its foundation. The expression is:

[0090] (9);

[0091] (10);

[0092] (11);

[0093] (12);

[0094] in, This represents the state value of an individual in the population during the current iteration. This represents the new individual state value in the next iteration. Indicates control parameters, For modulo operation; For random variables, To stabilize the index, To amplify parameters, For displacement parameters, The imaginary unit, For integration variables; This indicates that the i-th individual is at iteration step The position at that time This represents the random step size that follows a Lévy distribution; Indicates the iteration step The optimal position of the population at that time. This indicates the upper and lower bounds of the search for the parameters to be inverted. b Represents the learning factor. Indicates the iteration step as Time The position of each individual This represents the optimal position of the population during the historical iteration process.

[0095] Step S5: If the maximum number of iterations is reached, the optimization ends, the optimal parameters are output, the combined parameters obtained from the inversion are substituted into the finite element model for forward calculation, and the inversion results are evaluated.

[0096] Through iterative optimization using the ICPKO algorithm, the inversion results of the deformation modulus of the dam section and foundation zoning are E1=25.01GPa, E2=25.01GPa, and E3=15.65GPa. The equivalent values ​​of the inverted parameters are substituted into the gravity dam finite element model to calculate the displacement-hydraulic pressure components. The results are compared and verified with the hydraulic pressure component data separated by the HST model. The verification calculation results are as follows: Figure 7 As shown.

[0097] Therefore, this invention employs the aforementioned intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction. It generates deformation modulus samples using the Latin hypercube experimental method and calculates the water pressure components at measuring points under different combinations of deformation modulus samples using a finite element model. Furthermore, it constructs a CBLA model between the mechanical parameters and structural response of gravity dams and their foundation zones using an attention-enhanced hybrid neural network method, replacing finite element calculations with the CBLA model. Based on the HST displacement water pressure components and the water pressure components calculated using the CBLA model, an optimization objective function for the mechanical parameters of gravity dams and their foundation zones is established. The improved Kingfisher algorithm is then used to optimize this function through Logistic-Tent mapping initialization of population position, α-stable distribution, Lévy flight, and back-learning, yielding the optimal combination of deformation moduli for each zone. This method provides a novel technique for inverting the deformation modulus of gravity dams and their foundation zones during operation.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for intelligent identification of concrete dam structural parameters based on monitoring and simulation feedback correction, characterized in that, Includes the following steps: S1. Construct a finite element model based on engineering data and determine the range of values ​​for the parameters to be inverted. Construct the parameter set X using Latin hypercube sampling (LHS). S2. Input the parameter set X into the finite element model, calculate the dam water pressure components corresponding to different water levels, and then obtain the water pressure deformation difference dataset Y under different parameter combinations, and obtain the calculation sample S=[X,Y]; S3. Construct an attention-enhanced hybrid neural network (CBLA) model that reflects the nonlinear relationship between the deformation modulus and structural response of each zone of the gravity dam and dam foundation. Divide the calculation sample S into a training set and a test set in a 4:1 ratio. Input the training set into the CBLA model for training. S4. Construct the optimization objective function for deformation modulus inversion, use the ICPKO algorithm and the CBLA model to search for the mechanical parameters of the gravity dam to be determined, and update the parameter combination with the minimum fitness as the current optimal parameter. S5. If the maximum number of iterations is reached, the optimization ends, the optimal parameters are output, the combination of parameters obtained from the inversion is substituted into the finite element model for forward calculation, and the inversion results are evaluated. Step S3 specifically includes: S31. The samples obtained in step S2 are normalized using the Min-Max method to reduce the data to the range [0, 1]. Then, the data samples are divided into training set and test set. S32. Input the training set samples into the CNN model optimized based on the attention mechanism, extract local features through the convolution kernels in the convolutional layer, and use the ReLU activation function to establish a non-linear relationship between the input and output; S33. Introduce a channel attention mechanism to calculate the channel attention features for the structural response feature F. And refined features extracted through channel attention ; S34, Refine the features By inputting into the BiLSTM layer, the state associations at any distance within the sequence are explicitly modeled through its gating mechanism, enabling the CBLA model to capture the globally relevant features in the structural response; In step S4, when seeking the optimal deformation modulus inversion for the dam and foundation zones, the mechanical parameter inversion problem is transformed into an optimization problem by constructing an objective function. The objective function for optimization is expressed as: （8） in, This represents the elastic modulus of the dam body. This represents the deformation modulus of the dam foundation. Indicates the number of deformation measuring points. This indicates the number of times a monitoring point is monitored within the selected time period. Indicates the first under the action of reservoir water pressure Each measuring point is at The water pressure component in the statistical model at time. Represents the fitting coefficient. Indicates the height of the reservoir water level power term Indicates the first under the action of reservoir water pressure Each measuring point is at The CBLA model at time 1000 calculates the water pressure component value; In step S4, the ICPKO algorithm uses Logistic-Tent mapping to homogenize the spatial distribution of the population, and then introduces α-stable distribution to adjust the algorithm's global exploration and local optimization capabilities. Based on Levy flight and back learning to escape the local optimal solution of the algorithm, it is used to find the optimal combination of mechanical parameters of each zone of the dam and dam foundation.

2. The intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to claim 1, characterized in that, In step S32, the convolutional layers of the CNN model extract local features using the following formula: （1） Where ReLU represents a nonlinear activation function. Indicates the current convolutional layer's th... The output of each convolutional kernel, This represents the input to the current convolutional layer. This represents the convolution operation. Indicates the corresponding scalar weight. This represents the corresponding scalar bias term.

3. The intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to claim 1, characterized in that, In step S33, channel attention features The calculation formula is: （2） in, Indicates input features Perform global average pooling. Indicates input features Perform global max pooling. Indicates to The result of global average pooling. Indicates to The result of global max pooling. and The weights of the fully connected layers in a multilayer perceptron are represented by MLP, where MLP stands for Multilayer Perceptron. Indicates the activation function; Refine features The calculation formula is expressed as: （3）。 4. The intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to claim 2, characterized in that: In step S34, the weighted features are input into the BiLSTM layer. A gating mechanism is used to explicitly model the state associations at arbitrary distances in the sequence, capturing global features of the structural response, and refining the weighted features. After the outputs from the forward and backward LSTMs, the hidden layers of both layers are superimposed to jointly determine the output layer. The final output of BiLSTM is expressed as: （4） （5） （6） in, , and Both represent activation functions. and These represent the LSTM hidden states during the forward and backward iterations, respectively. Indicates the weights of different levels. Representing refined features .

5. The intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to claim 4, characterized in that, In step S3, the accuracy of the CBLA model is measured by the relative error between the CBLA model and the finite element model. The calculated values ​​from the finite element model and the predicted values ​​from the CBLA model are selected to calculate the accuracy of the surrogate model. The calculation formula is as follows: （7） in, and Let represent the learnable parameters of CNN and BiLSTM, respectively. This represents the total number of training samples. express The first result obtained by inputting the CBLA model Water pressure components at each measuring point; This represents the calculated value of the water pressure component in the corresponding finite element model.

6. The intelligent identification method for concrete dam structural parameters based on monitoring and simulation feedback correction according to claim 1, characterized in that, In step S4, the expression for the ICPKO algorithm is: （9） （10） （11） （12） in, This represents the state value of an individual in the population during the current iteration. This represents the new individual state value in the next iteration. Indicates control parameters, For modulo operation; For random variables, To stabilize the index, To amplify parameters, For displacement parameters, The imaginary unit, For integration variables; Indicates the first Each individual in the iteration step The position at that time This represents the random step size that follows a Lévy distribution; Indicates the iteration step The optimal position of the population at that time. This indicates the upper and lower bounds of the search for the parameters to be inverted. Represents the learning factor. Indicates the iteration step as Time The position of each individual This represents the optimal position of the population during the historical iteration process.

Images