A running period high face plate dam deformation rapid prediction method based on intelligent agent model and gradient estimation

Abstract

The application utilizes easily accessible high face dam parameters, fully utilizes neural network, multivariate function gradient estimation and other technologies, and constructs a running period high face dam deformation rapid prediction method based on a neural network intelligent agent model and gradient estimation technology. The method is based on high face dam basic parameters and measured deformation data, utilizes the super nonlinear modeling capacity and factor sensitivity analysis technology of the neural network, obtains an optimized agent model through two times of training, realizes explicit of the deformation prediction model through the gradient estimation technology of the complex function, has clear physical meaning and convenient model use, provides a reliable and convenient method for rapid prediction of the running period deformation of the high face dam, and effectively makes up for the shortage of the traditional technology. The application result can be directly applied to preliminary estimation of the high face dam deformation, meanwhile, the method has universal generality, and can also provide a unified technical paradigm for dam operation state prediction.

Description

Technical Field

[0001] This disclosure relates to a rapid prediction method for the deformation of high panel dams during operation based on intelligent agent models and gradient estimation techniques. This method is applicable to the field of safety evaluation and monitoring of panel dams, and can also provide technical ideas and references for the rapid prediction and monitoring evaluation of the operational performance of other types of dams. Background Technology

[0002] Deformation of high panel dams is an important indicator for design control and operation monitoring and evaluation. Among them, the maximum settlement deformation of the dam crest is a key focus in the project and a control indicator for the design and operation safety of panel dams.

[0003] Excessive deformation will cause cracks or damage to the panels of high-face dams, endangering the dam's seepage prevention system and threatening the overall stability of the dam. Excessive deformation may also lead to insufficient dam height and failure to meet flood control requirements.

[0004] In recent years, with the construction of a number of high panel dams, especially ultra-high panel dams over 200m, how to quickly and conveniently predict and evaluate their deformation during operation is crucial for the safety of panel dams.

[0005] Currently, the deformation prediction of panel dams mainly adopts numerical calculation methods, that is, deformation parameters are obtained by inverting experimental or monitoring data, and then the deformation of the panel dam at a certain moment is calculated. In addition, engineering empirical methods are also commonly used for deformation prediction, such as using 1% of the dam height as the control standard for maximum settlement deformation.

[0006] Numerical calculation methods have many drawbacks. First, they are slow and time-consuming, making it difficult to quickly predict the deformation of panel dams. Second, due to the difficulty in obtaining accurate calculation parameters, the calculated values ​​often have large errors compared to the actual values. In addition, the calculation methods and assumptions also reduce the reliability of the calculation results. The biggest drawback of directly using engineering experience methods is that the deformation prediction is too rough and the accuracy is not high, making it difficult to meet the requirements of dam monitoring and evaluation during operation in most cases.

[0007] With the commissioning and accumulation of measured data of high-faced concrete dams at home and abroad, there is an urgent need for a method that can quickly predict the deformation of high-faced concrete dams during operation based on simple, objective and easily obtainable dam parameters, without the need for complex tools or calculations. This method will significantly improve the efficiency and level of safety monitoring and safety evaluation of high-faced concrete dams, thereby improving the quality of operation and management of high-faced concrete dams throughout their entire life cycle and ensuring dam safety. Summary of the Invention

[0008] The purpose of this disclosure is to provide a fast prediction method for the deformation of high panel dams during operation based on a neural network intelligent agent model and gradient estimation, so as to solve the problems of existing high panel dam deformation prediction methods, such as difficulty in obtaining input parameters, slow calculation speed, large calculation error and low reliability.

[0009] The rapid deformation prediction method for high-faced concrete dams established in this disclosure is characterized by a clear mechanism, easily obtainable parameters, explicit model expression, convenient deformation prediction, and strong versatility in its construction method. Its main steps are as follows:

[0010] S1: Construct an index system for characterizing the influencing factors of high-faced dam deformation;

[0011] S2: Collect characteristic deformation monitoring data and corresponding influencing factor indicators of existing high-face dam projects, and construct a case dataset consisting of characteristic deformation, influencing factor indicators and running time;

[0012] S3: Construct a neural network model and train the neural network model using the established dataset of characteristic deformation and influencing factor index system of high panel dams to obtain a primary surrogate model for estimating characteristic deformation of high panel dams.

[0013] S4: Quantitatively analyze the influence of each influencing factor index on the characteristic deformation of high panel dam using the primary surrogate model, and rank them according to the influence degree to obtain the significant influencing factors on the characteristic deformation of high panel dam;

[0014] S5: Based on the significant influencing factors obtained in step S4, construct a new dataset consisting of characteristic deformation and significant influencing factor indicators, and construct and train a new neural network model based on the new dataset to obtain an optimized surrogate model for high panel dam deformation estimation.

[0015] S6: For the significant influencing factors and running time of the characteristic deformation, according to the value range of each significant influencing factor in the case dataset, equal-interval interpolation is performed between the corresponding upper and lower limits to obtain the representative points of each significant influencing factor. Orthogonal combination design is then performed on the representative points of each significant influencing factor to obtain the representative combination used for the deformation calculation of high panel dam.

[0016] S7: Using the optimized surrogate model obtained in step S5, apply a preset small increment to each representative combination obtained in step S6, calculate the corresponding feature deformation estimate, and determine the gradient of feature deformation relative to each significant influencing factor index and running time based on the change of feature deformation estimate before and after the incremental perturbation.

[0017] S8: For any high panel dam, determine its combination of influencing factors, and select the representative combination with the highest similarity to the combination of influencing factors from the representative combinations obtained in step S6. Estimate the deformation of the panel dam at any time based on the characteristic deformation and gradient of the combination at different times.

[0018] Further, step S1 includes the following steps:

[0019] S1-1: Based on the mechanical mechanism and engineering examples of high-faced concrete dam deformation, determine the influencing factors used to characterize the deformation of high-faced concrete dams. The influencing factors include at least: dam height, dam axis length, upstream dam slope, downstream dam slope, porosity of the main rockfill area, strength of the main rockfill rock mass, porosity of the secondary rockfill area, strength of the secondary rockfill rock mass, thickness of the dam foundation overburden layer, and compressibility modulus of the overburden soil. The total number of the influencing factors is denoted as n.

[0020] S1-2: Based on the influencing factors determined in step S1-1, construct an index system for the influencing factors of deformation of high-faced concrete dams. For the i-th high-faced concrete dam, its influencing factor index system is denoted as v. i =(v i1 , v i2 , ……, v in ), where v in This represents the nth influencing factor index corresponding to the i-th high panel dam.

[0021] Further, step S2 includes the following steps:

[0022] S2-1: Collect detailed information on measured characteristic deformation data of high panel dams at home and abroad, as well as the corresponding influencing factor index system. The characteristic deformation includes at least: maximum settlement of the dam crest, maximum horizontal displacement, settlement of characteristic points of the dam body, and panel deflection. The number of high panel dams is denoted as N.

[0023] S2-2: The characteristic deformation data of the i-th high-face concrete dam mentioned in step S2-1 is denoted as d. i =[(t i1 , δ i1 ), (t i2 ,δ i2 ), ……, (t im , δ im )], where t im The operating time is calculated from the completion or commencement date, and the selected operating periods include at least: completion date, 3 months of operation, 6 months of operation, 1 year of operation, 3 years of operation, and 5 years of operation. im For t im The characteristic deformation value corresponding to the given time;

[0024] S2-3: For any high-faced concrete dam, construct a deformation sample dataset for the dam based on its influencing factor indicators and characteristic deformation data: D i =[(v i1 , v i2 , ……, v in , t i1 , δ i1 ), (v i1 , v i2 , ……, v in , t i2 ,δ i2 ), ……, (v i1 , v i2 , ……, v in , t im , δ im ]], where i represents the i-th high panel dam, and each high panel dam corresponds to m measured deformation sample data;

[0025] S2-4: Based on step S2-3, construct deformation sample datasets for each high panel dam, forming a measured high panel dam characteristic deformation dataset D = (D1, D2, ..., D...). N The total number of data samples is N×m.

[0026] Furthermore, step S3 includes the following steps:

[0027] S3-1: Construct a BP neural network for training a proxy model for estimating the deformation of a high panel dam. The BP neural network includes an input layer, at least one hidden layer, and an output layer, wherein the number of neurons in the input layer is n+1, the number of neurons in the output layer is 1, the number of hidden layers is greater than or equal to 1, and the number of neurons in the hidden layer is greater than 1.

[0028] S3-2: Train the BP neural network model established in step S3-1 using the high panel dam feature deformation dataset constructed in step S2, and randomly extract a validation dataset from the dataset, so that the ratio of training data to validation data is 4:1; taking the data of the i-th dam at the j-th time as an example, the model input is (v i1 , v i2 , ……, v in , t ij The output is δ. ij , where i∈[1,N], j∈[1,m], by training and validating the case dataset, a primary surrogate model for estimating the feature deformation of high panel dams is obtained.

[0029] Further, in step S3-2, the primary proxy model satisfies the following accuracy criterion: on the validation dataset, the standard deviation of the deviation between the high panel dam feature deformation calculated by the primary proxy model and the corresponding measured value is not greater than a preset control threshold, and the primary proxy model satisfies the preset overfitting criterion.

[0030] Further, step S4 includes the following steps:

[0031] S4-1: For all high-face dams, determine the value range of each influencing factor index in the case dataset, and denote the lower limit of each influencing factor index as v. min =(v 1_min , v 2_min , ……, v n_min The upper limit of the value is v. max =(v 1_max , v 2_max , ……, v n_max ), where v i_min =min(v i1, v i2 , ......, v iN ), v i_max =max(v i1, v i2 ,......, v iN );

[0032] S4-2: For any high panel dam, for each influencing factor indicator, under the condition that the other influencing factor indicators remain unchanged, take the lower limit and upper limit of the value of the influencing factor indicator respectively, use the primary surrogate model to calculate the corresponding feature deformation estimation result, and determine the degree of influence of the influencing factor indicator on the feature deformation according to the percentage change of the feature deformation estimation result. Sort the influencing factor indicators whose influence degree is not less than the preset threshold according to the degree of influence from large to small, and select the influencing factor indicators in the first r positions as the significant influencing factors of the high panel dam, where r∈[1, n];

[0033] S4-3: Repeat step S4-2 for all high panel dams, count the number of times each influencing factor index is selected as a significant influencing factor, sort the cumulative selection counts, and select the influencing factor index with the top r cumulative selection counts as the final significant influencing factor of the characteristic deformation of the high panel dam.

[0034] Further, step S5 includes the following steps:

[0035] S5-1: Based on the significant influencing factors obtained in step S4, remove the non-significant influencing factor indicators from the influencing factor index system of high-face panel dams, and sort the significant influencing factor indicators in descending order of their influence on characteristic deformation to obtain a new influencing factor index system for high-face panel dam deformation, denoted as u. i =(u i1 , u i2 , ……, u ir ), where i represents the number of the high-face dam, and the influence of the former factor is greater than that of the latter factor;

[0036] S5-2: Replace the original influencing factor index system in dataset D constructed in step S2 with the significant influencing factor index system u obtained in step S5-1, and repeat step S3 to train the neural network model to obtain an optimized surrogate model for estimating the feature deformation of high panel dams; the new neural network model includes an input layer, at least one hidden layer and an output layer, wherein the number of neurons in the input layer is r+1, the number of neurons in the output layer is 1, and the number of neurons in each hidden layer is greater than 1.

[0037] Further, step S6 includes the following steps:

[0038] S6-1: For all collected high-level panel dams, determine the value range of the significant influencing factor indicators, and obtain the lower limit u of the value of each significant influencing factor indicator. min =(u 1_min , u 2_min , ……, u r_min The upper limit of the value u max =(u 1_max , u 2_max ,……, u r_max And determine the lower limit of the running time t used for deformation estimation. min and runtime limit t max ;

[0039] S6-2: For any significant influencing factor indicator and its operating time, representative points are selected at equal intervals between the corresponding lower and upper limits of values; for the k-th significant influencing factor indicator, its representative points include the lower limit u of that significant influencing factor indicator. k_min and several equally spaced intermediate representative points u located between the lower limit and the upper limit of the value. k_Mi M i The constant value represents the representative point of the i-th high panel dam, and is determined based on the influence and distribution range of each factor.

[0040] S6-3: Orthogonal combination design is performed on representative points corresponding to each significant influencing factor index and operating time to form a representative sample set w=(w1, w2, ……, w) for high panel dam deformation calculation. H ), where H is the number of representative samples obtained through orthogonal combination design, and any representative sample w i It consists of representative points of each significant influencing factor indicator and its corresponding running time, denoted as w. i =(w i1 , w i2 , ……, w ir , t i ), where w ij This represents the point corresponding to the j-th significant influencing factor indicator.

[0041] Further, step S7 includes the following steps:

[0042] S7-1: For any representative combination, input the index values ​​of its significant influencing factors and the running time into the optimized surrogate model obtained in step S5, and calculate the corresponding feature deformation estimate.

[0043] S7-2: Apply a preset small increment to any significant influencing factor index or running time in the feature deformation estimation values ​​obtained in step S7-1, and determine the gradient of feature deformation at the representative combination relative to each significant influencing factor index and running time based on the ratio between the change in the feature deformation estimation values ​​before and after applying the increment and the increment.

[0044] S7-3: For each representative combination, construct a calculation table containing the estimated value of the characteristic deformation corresponding to the representative combination and its gradient information relative to each significant influencing factor index and running time.

[0045] Further, step S8 includes the following steps:

[0046] S8-1: For any high panel dam, determine its combination of significant influencing factor indicators, and in the representative combination obtained in step S6, select the representative combination with the smallest distance to the combination of significant influencing factor indicators as the matching combination based on the distance metric between the significant influencing factor indicators;

[0047] S8-2: Obtain the estimated characteristic deformation values ​​of the matching combination at different operating times and their gradients relative to each significant influencing factor index and operating time, and perform a linear approximation on the combination of influencing factor indices of the high panel dam based on the gradients to calculate the estimated characteristic deformation values ​​of the high panel dam at the corresponding operating times.

[0048] The advantages of this disclosure are as follows: The method for rapid prediction of deformation of high panel dams during operation based on neural network surrogate model and gradient estimation established in this disclosure can construct prediction models for deformations with arbitrary characteristics. The physical meaning is clear, and no computer or complex tools are required. It can achieve rapid prediction of deformation of high panel dams at any time based on easily obtainable parameters. It can provide a basis for monitoring and evaluating the deformation of panel dams during construction and operation, and also provide a unified paradigm for the construction of prediction models for the operational performance of panel dams and other dams. Attached Figure Description

[0049] Figure 1 This is a publicly available implementation flowchart.

[0050] Figure 2 This is a reference format for the calculation table of the deformation rapid prediction method constructed in this publication. Detailed Implementation

[0051] The preferred embodiments of this disclosure will be further described below with reference to the accompanying drawings. This disclosure provides a method for rapid prediction of deformation of high panel dams during operation based on a neural network surrogate model and gradient estimation. Figure 1 As shown, the specific steps are as follows:

[0052] S1-1: Based on the mechanical mechanism and engineering examples of high-faced concrete dam deformation, determine the influencing factors used to characterize the deformation of high-faced concrete dams. The influencing factors include at least: dam height, dam axis length, upstream dam slope, downstream dam slope, porosity of the main rockfill area, strength of the main rockfill rock mass, porosity of the secondary rockfill area, strength of the secondary rockfill rock mass, thickness of the dam foundation overburden layer, and compressibility modulus of the overburden soil. The total number of the influencing factors is denoted as n.

[0053] S1-2: Based on the influencing factors determined in step S1-1, construct an index system for the influencing factors of deformation of high-faced concrete dams. For the i-th high-faced concrete dam, its influencing factor index system is denoted as v. i =(v i1 , v i2 , ……, v in ), where v in This represents the nth influencing factor index corresponding to the i-th high panel dam.

[0054] S2-1: Collect measured characteristic deformation data and detailed information on the corresponding influencing factor index system of high panel dams with a height of 70m or more at home and abroad. The characteristic deformation includes at least: maximum settlement of the dam crest, maximum horizontal displacement, settlement of characteristic points of the dam body and panel deflection. The number of high panel dams is denoted as N.

[0055] S2-2: The characteristic deformation data of the i-th high-face concrete dam mentioned in step S2-1 is denoted as d. i =[(t i1 , δi1 ), (t i2 ,δ i2 ), ……, (t im , δ im )], where t im The operating time is calculated from the completion or commencement date, and the selected operating periods include at least: completion date, 3 months of operation, 6 months of operation, 1 year of operation, 3 years of operation, and 5 years of operation. im For t im The characteristic deformation value corresponding to the time.

[0056] S2-3: For any high-faced concrete dam, construct a deformation sample dataset for the dam based on its influencing factor indicators and characteristic deformation data: D i =[(v i1 , v i2 , ……, v in , t i1 , δ i1 ), (v i1 , v i2 , ……, v in , t i2 ,δ i2 ), ……, (v i1 , v i2 , ……, v in , t im , δ im )], where i represents the i-th high panel dam, and each high panel dam corresponds to m measured deformation sample data.

[0057] S2-4: Based on step S2-3, construct deformation sample datasets for each high panel dam, forming a measured high panel dam characteristic deformation dataset D = (D1, D2, ..., D...). N The total number of data samples is N×m.

[0058] S3-1: Construct a BP neural network for training a proxy model for estimating the deformation of high panel dams. The BP neural network includes an input layer, at least one hidden layer, and an output layer, wherein the number of neurons in the input layer is n+1, the number of neurons in the output layer is 1, the number of hidden layers is greater than or equal to 1, and the number of neurons in the hidden layer is greater than 1.

[0059] S3-2: Train the BP neural network model established in step S3-1 using the high panel dam feature deformation dataset constructed in step S2, and randomly extract a validation dataset from the dataset, so that the ratio of training data to validation data is 4:1; taking the data of the i-th dam at the j-th time as an example, the model input is (v i1 , v i2, ……, v in , t ij The output is δ. ij , where i∈[1,N], j∈[1,m], by training and validating the case dataset, a primary surrogate model for estimating the feature deformation of high panel dams is obtained.

[0060] In step S3-2, the primary surrogate model satisfies the following accuracy criteria: on the validation dataset, the standard deviation between the high panel dam feature deformation calculated by the primary surrogate model and the corresponding measured value is no greater than 10%, and the surrogate model does not show obvious overfitting.

[0061] S4-1: For all high-face dams, determine the value range of each influencing factor index in the case dataset, and denote the lower limit of each influencing factor index as v. min =(v 1_min , v 2_min , ……, v n_min The upper limit of the value is v. max =(v 1_max , v 2_max , ……, v n_max ), where v i_min =min(v i1, v i2 , ......, v iN ), v i_max =max(v i1, v i2 ,......, v iN ).

[0062] S4-2: For any high panel dam, for each influencing factor indicator, under the condition that the other influencing factor indicators remain unchanged, take the lower limit and upper limit of the value of the influencing factor indicator respectively, use the primary surrogate model to calculate the corresponding characteristic deformation estimation result, and determine the degree of influence of the influencing factor indicator on the characteristic deformation according to the percentage change of the characteristic deformation estimation result. Sort the influencing factor indicators with an influence degree of not less than 5% in descending order of influence degree, and select the influencing factor indicators in the first r positions as the significant influencing factors of the high panel dam, where r∈[1, n], and it is recommended to take 3~5.

[0063] S4-3: Repeat step S4-2 for all high panel dams, count the number of times each influencing factor index is selected as a significant influencing factor, sort the cumulative selection counts, and select the influencing factor index with the top r cumulative selection counts as the final significant influencing factor of the characteristic deformation of the high panel dam.

[0064] S5-1: Based on the significant influencing factors obtained in step S4, remove the non-significant influencing factor indicators from the influencing factor index system of high-face panel dams, and sort the significant influencing factor indicators in descending order of their influence on characteristic deformation to obtain a new influencing factor index system for high-face panel dam deformation, denoted as u. i =(u i1 , u i2 , ……, u ir ), where i represents the number of the high-face dam, and the influence of the former factor is greater than that of the latter factor.

[0065] S5-2: Replace the original influencing factor index system in dataset D constructed in step S2 with the significant influencing factor index system u obtained in step S5-1, and repeat step S3 to train the neural network model to obtain an optimized surrogate model for estimating the feature deformation of high panel dams; the new neural network model includes an input layer, at least one hidden layer and an output layer, wherein the number of neurons in the input layer is r+1, the number of neurons in the output layer is 1, and the number of neurons in each hidden layer is greater than 1.

[0066] S6-1: For all collected high-level panel dams, determine the value range of the significant influencing factor indicators, and obtain the lower limit u of the value of each significant influencing factor indicator. min =(u 1_min , u 2_min , ……, u r_min The upper limit of the value u max =(u 1_max , u 2_max ,……, u r_max And determine the lower limit of the running time t used for deformation estimation. min and runtime limit t max ;

[0067] S6-2: For any significant influencing factor indicator and its operating time, representative points are selected at equal intervals between the corresponding lower and upper limits of values; for the k-th significant influencing factor indicator, its representative points include the lower limit u of that significant influencing factor indicator. k_min and several equally spaced intermediate representative points u located between the lower limit and the upper limit of the value. k_Mi M i This represents the constant value for the representative point of the i-th high-face dam. The value is determined based on the influence and distribution range of various factors, and a suggested range is M. i ∈[1, 7].

[0068] S6-3: Orthogonal combination design is performed on representative points corresponding to each significant influencing factor index and operating time to form a representative sample set w=(w1, w2, ……, w) for high panel dam deformation calculation. H ), where H is the number of representative samples obtained through orthogonal combination design, and any representative sample w i It consists of representative points of each significant influencing factor indicator and its corresponding running time, denoted as w. i =(w i1 , w i2 , ……, w ir , t i ), where w ij This represents the point corresponding to the j-th significant influencing factor indicator.

[0069] S7-1: For any representative combination, input the index values ​​of its significant influencing factors and the running time into the optimized surrogate model obtained in step S5, and calculate the corresponding feature deformation estimate: (1)

[0070] S7-2: Apply a preset small increment to any significant influencing factor index or running time in the feature deformation estimation values ​​obtained in step S7-1. Determine the gradient of the feature deformation at the representative combination relative to each significant influencing factor index and running time based on the ratio between the change in the feature deformation estimation value before and after applying the increment and the increment itself. (2)

[0071] S7-3: For each representative combination, construct a calculation table containing the estimated characteristic deformation values ​​corresponding to the representative combination and their gradient information relative to each significant influencing factor index and running time. The calculation table format is as follows: Figure 2 As shown.

[0072] S8-1: For any high panel dam, determine the combination of significant influencing factors w. j In step S6, among the representative combinations obtained, the representative combination w with the smallest distance to the combination of significant influencing factor indicators is selected based on the distance metric between the significant influencing factor indicators. i As a matching combination;

[0073] S8-2: Obtain the matching combination w i The estimated characteristic deformation values ​​at different operating times and their gradients relative to various significant influencing factor indicators and operating time are calculated. Based on the gradients, a linear approximation is performed on the combination of influencing factor indicators of the high-faced concrete dam to obtain the estimated characteristic deformation values ​​of the high-faced concrete dam at the corresponding operating times. (3).

Claims

1. A method for rapid prediction of deformation of high panel dams during operation based on a neural network intelligent agent model and gradient estimation, characterized in that, Includes the following steps: S1: Construct an index system for characterizing the influencing factors of high-faced dam deformation; S2: Collect measured characteristic deformation data and corresponding influencing factor index data of existing high panel dams, and associate the influencing factor index, running time and characteristic deformation data to construct a case dataset for high panel dam deformation estimation. S3: Construct a neural network model and train the neural network model using the established dataset of characteristic deformation and influencing factor index system of high panel dams to obtain a primary surrogate model for estimating characteristic deformation of high panel dams. S4: Quantitatively analyze the influence of each influencing factor index on the characteristic deformation of high panel dam using the primary surrogate model, and rank them according to the influence degree to obtain the significant influencing factors on the characteristic deformation of high panel dam; S5: Based on the significant influencing factors obtained in step S4, construct a new dataset consisting of characteristic deformation and significant influencing factor indicators, and construct and train a new neural network model based on the new dataset to obtain an optimized surrogate model for high panel dam deformation estimation. S6: For the significant influencing factors and running time of the characteristic deformation, according to the value range of each significant influencing factor in the case dataset, equal-interval interpolation is performed between the corresponding upper and lower limits to obtain the representative points of each significant influencing factor. Orthogonal combination design is then performed on the representative points of each significant influencing factor to obtain the representative combination used for the deformation calculation of high panel dam. S7: Using the optimized surrogate model obtained in step S5, apply a preset small increment to each representative combination obtained in step S6, calculate the corresponding feature deformation estimate, and determine the gradient of feature deformation relative to each significant influencing factor index and running time based on the change of feature deformation estimate before and after the incremental perturbation. S8: For any high panel dam, determine its combination of influencing factors, and select the representative combination with the highest similarity to the combination of influencing factors from the representative combinations obtained in step S6. Estimate the deformation of the panel dam at any time based on the characteristic deformation and gradient of the combination at different times.

2. The method according to claim 1, characterized in that: Step S1 includes the following steps: S1-1: Based on the mechanical mechanism and engineering examples of high-faced concrete dam deformation, determine the influencing factors used to characterize the deformation of high-faced concrete dams. The influencing factors include at least the geometric parameters of the dam body, the structural parameters of the rockfill area, and the mechanical parameters of the dam foundation. S1-2: Based on the influencing factors determined in step S1-1, construct an index system for the influencing factors of deformation of high-faced concrete dams. For the i-th high-faced concrete dam, its influencing factor index system is denoted as v. i =(v i1 , v i2 , ……, v in ), where n is the number of influencing factors.

3. The method according to claim 1, characterized in that: Step S2 includes the following steps: S2-1: Collect measured characteristic deformation data of existing high panel dams and corresponding influencing factor index data. The characteristic deformation data shall include at least the characteristic deformation amount reflecting the overall or local deformation state of the dam body. The number of high panel dams shall be denoted as N. S2-2: The characteristic deformation data of the i-th high-face concrete dam mentioned in step S2-1 is denoted as d. i =[(t i1 , δ i1 ), (t i2 ,δ i2 ), ……, (t im , δ im )], where t im The runtime is calculated from the time of completion or commencement of construction, δ im For t im The characteristic deformation value corresponding to the given time; S2-3: For any high-faced concrete dam, combine the influencing factor indicators, operating time, and characteristic deformation values ​​to construct the deformation sample dataset D of that high-faced concrete dam. i , where i represents the number of the high-face dam; S2-4: Summarize the deformation sample datasets of each high panel dam constructed in step S2-3 to form a case dataset for deformation estimation of high panel dams.

4. The method according to claim 1, characterized in that: Step S3 includes the following steps: S3-1: Construct a neural network surrogate model for estimating the deformation of high panel dams, wherein the neural network surrogate model includes an input layer, at least one hidden layer, and an output layer; S3-2: Using the high panel dam feature deformation dataset constructed in step S2, train and validate the neural network model established in step S3-1 to obtain a primary surrogate model for estimating the feature deformation of high panel dams.

5. The method according to claim 4, characterized in that: In step S3-2, the primary surrogate model satisfies the following accuracy criterion: on the validation dataset, the standard deviation of the deviation between the high panel dam feature deformation calculated by the primary surrogate model and the corresponding measured value is not greater than a preset control threshold, and the primary surrogate model satisfies the preset overfitting criterion.

6. The method according to claim 1, characterized in that: Step S4 includes the following steps: S4-1: For each high-altitude dam included in the case dataset, determine the value range of each influencing factor index in the case dataset, and obtain the lower limit and upper limit of each influencing factor index respectively. S4-2: For any high panel dam, for each influencing factor indicator, under the condition that the other influencing factor indicators remain unchanged, take the lower limit and upper limit of the value of the influencing factor indicator respectively, use the primary surrogate model to calculate the corresponding feature deformation estimation result, and determine the degree of influence of the influencing factor indicator on the feature deformation according to the percentage change of the feature deformation estimation result. Sort the influencing factor indicators whose influence degree is not less than the preset threshold according to the degree of influence from large to small, and select the influencing factor indicators in the first r positions as the significant influencing factors of the high panel dam, where r∈[1, n]; S4-3: Repeat step S4-2 for all high panel dams, count the number of times each influencing factor index is selected as a significant influencing factor, sort the cumulative selection counts, and select the influencing factor index with the top r cumulative selection counts as the final significant influencing factor of the characteristic deformation of the high panel dam.

7. The method according to claim 1, characterized in that: Step S5 includes the following steps: S5-1: Based on the significant influencing factors obtained in step S4, remove the non-significant influencing factor indicators from the influencing factor index system of high panel dams, and sort them according to the degree of influence of the significant influencing factor indicators on the characteristic deformation to construct a new influencing factor index system for high panel dam deformation. S5-2: Replace the original influencing factor index system in the dataset constructed in step S2 with the significant influencing factor index system constructed in step S5-1, and retrain the neural network model based on the significant influencing factor index system to obtain an optimized surrogate model for estimating the characteristic deformation of high panel dams.

8. The method according to claim 1, characterized in that: Step S6 includes the following steps: S6-1: For the high panel dam, determine the value range of each significant influencing factor index and running time in the case dataset, and obtain the lower limit and upper limit of each significant influencing factor index and running time respectively; S6-2: For any significant influencing factor indicator and running time, select multiple representative points between its corresponding lower limit and upper limit of value in an equally spaced manner; S6-3: Orthogonally combine representative points corresponding to each significant influencing factor index and running time to form a representative sample set for high panel dam deformation calculation. Each representative sample consists of representative points of each significant influencing factor index and the corresponding running time.

9. The method according to claim 1, characterized in that: Step S7 includes the following steps: S7-1: For any representative combination, input the index values ​​of its significant influencing factors and the running time into the optimized surrogate model obtained in step S5, and calculate the corresponding feature deformation estimate. S7-2: Apply a preset small increment to any significant influencing factor index or running time in the feature deformation estimation values ​​obtained in step S7-1, and determine the gradient of feature deformation at the representative combination relative to each significant influencing factor index and running time based on the ratio between the change in the feature deformation estimation values ​​before and after applying the increment and the increment. S7-3: For each representative combination, construct a calculation table containing the estimated value of the characteristic deformation corresponding to the representative combination and its gradient information relative to each significant influencing factor index and running time.

10. The method according to claim 1, characterized in that: Step S8 includes the following steps: S8-1: For any high panel dam, determine its combination of significant influencing factor indicators, and in the representative combination obtained in step S6, select the representative combination with the smallest distance to the combination of significant influencing factor indicators as the matching combination based on the distance metric between the significant influencing factor indicators; S8-2: Obtain the estimated characteristic deformation values ​​of the matching combination at different operating times and their gradients relative to each significant influencing factor index and operating time, and perform a linear approximation on the combination of influencing factor indices of the high panel dam based on the gradients to calculate the estimated characteristic deformation values ​​of the high panel dam at the corresponding operating times.

Images