Earth and rockfill dam constitutive model parameter inversion method based on Diffusion MCMC

Abstract

The invention provides an earth and rockfill dam constitutive model parameter inversion method based on a Diffusion MCMC, and the method comprises the steps: building a finite element model according to the field survey data of a dam, selecting inversion parameters according to sensitivity analysis, determining the range of parameters to be inverted, constructing different parameter combinations, and carrying out the finite element calculation, thereby obtaining a proxy model training sample; then establishing nonlinear mapping of constitutive parameters and earth and rockfill dam settlement values based on an Ivy-RF model by using the training sample; training a Diffusion model by using the initial sample, and generating a plurality of groups of samples matched with the input data after training so as to prevent MCMC sampling from falling into local optimum; and finally, obtaining an optimal parameter combination by using a Diffusion-MCMC sampling method according to the settlement data of the earth and rockfill dam actually measured by engineering.

Description

Technical Field

[0001] The invention belongs to the technical field of water conservancy engineering, and in particular relates to a parameter inversion method of an earth-rock dam constitutive model based on Diffusion MCMC. Background Art

[0002] As a very common dam type in water conservancy projects, earth-rock dams have broad development prospects due to their good adaptability, small project scale, convenient construction and high safety. However, due to the complexity of design and construction, the actual parameters of the dam body materials of the built earth-rock dams are often significantly different from the design preset values. This difference is due to the nonlinear and elastoplastic properties of the dam body materials and their base covering layers, as well as the limitations of equipment and costs, and the uncertainty of the test environment and operation. These factors make it extremely challenging to accurately obtain material parameters through traditional tests and empirical comparison methods. In view of the variability and uncertainty of earth-rock materials, as well as the complex and changing engineering environment, it is particularly critical to study the uncertainty-based earth-rock dam parameter inversion method. Summary of the invention

[0003] In view of the actual needs and gaps in the prior art, the present invention proposes a random inversion method for parameters of the earth-rock dam constitutive model based on Diffusion-MCMC, which effectively solves the problems of traditional methods in terms of poor convergence, slow calculation speed and failure to fully consider the uncertainty of geotechnical materials.

[0004] The construction process of the scheme includes the following steps: 1) According to the field survey data of the dam, a finite element model is established, key inversion parameters are selected according to sensitivity analysis, the range of parameters to be inverted is determined, and different parameter combinations are constructed for finite element calculation to obtain proxy model training samples; 2) Using the training samples, a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​is established based on the Ivy-RF model; 3) The Diffusion model is trained using the initial samples, and after training, multiple groups of samples matching the input data can be generated to avoid MCMC sampling falling into local optimality; 4) The settlement data of the earth-rock dam measured in the project is used to obtain the optimal parameter combination using the Diffusion-MCMC sampling method. The present invention not only effectively solves the time-consuming problem of the current finite element, but also further improves the accuracy of parameter random inversion by optimizing the Markov chain Monte Carlo sampling technology.

[0005] The technical solution specifically adopted by the present invention to solve the technical problem is:

[0006] A parameter inversion method for a constitutive model of an earth-rock dam based on Diffusion MCMC is characterized by: establishing a finite element model according to field survey data of the dam, selecting inversion parameters according to sensitivity analysis, determining the range of parameters to be inverted, constructing different parameter combinations for finite element calculation to obtain proxy model training samples; then using the training samples to establish a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​based on an Ivy-RF model; and using initial samples to train a Diffusion model, generating multiple groups of samples matching the input data after training to avoid MCMC sampling falling into a local optimum; finally, using the settlement data of the earth-rock dam measured in the project, a Diffusion-MCMC sampling method is used to obtain the optimal parameter combination.

[0007] Furthermore, the method of establishing a finite element model based on the on-site survey data of the dam, selecting inversion parameters based on sensitivity analysis, determining the range of parameters to be inverted, and constructing different parameter combinations for finite element calculation to obtain proxy model training samples is as follows: construct a finite element model and perform meshing based on the design drawings and field measurement data of the earth-rock dam; perform finite element calculations in the finite element model according to the dam body material and construction period partitions; for inversion reference point selection: based on the distribution of monitoring equipment on the dam body cross section, select reference points for constitutive parameter inversion search from points on the dam body cross section where monitoring equipment is set and corresponding to the finite element grid; use the Duncan-Zhang EB model as the basis for finite element forward analysis; and determine the parameters K, Based on engineering experience, the impact factor K is set. The value range and number of levels of each influencing factor are selected; the appropriate orthogonal table is selected for the number of levels of each influencing factor, the table header is designed, and the test is performed; the test results are statistically analyzed to extract multiple groups of constitutive model parameter groups; these parameters are used to perform finite element analysis to simulate the displacement of the dam body; the settlement displacement of the monitoring equipment nodes is extracted from the simulation results as the training samples of the proxy model.

[0008] Furthermore, using the training samples, a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​is established based on the Ivy-RF model, including:

[0009] For the input data of the training samples (K, ) for preprocessing; the preprocessed data is input into the random forest regression RF model as a training set; and the number of trees and the minimum number of leaf nodes in the random forest are optimized using the Ivy optimization algorithm during the training process; the fitness function is the mean square error MSE of the RF model for the settlement values ​​of the training set and the settlement values ​​of the test set;

[0010] fitness=MSE[predict(train)]+MSE[predict(test)]

[0011] Finally, the optimal hyperparameter configuration is used to retrain the RF model to obtain the best prediction effect and model performance.

[0012] Furthermore, the initial samples are used to train the Diffusion model, and after training, multiple groups of samples matching the input data are generated, including:

[0013] Training the forward diffusion model: The diffusion model is trained using the initial samples. The diffusion model uses the following formula to perform the forward diffusion process and gradually transforms the initial samples into pure Gaussian noise:

[0014]

[0015] Among them, ε~N(0,1) is the noise weight added to each time step, β t is the noise width in the diffusion process, from the original vector x 0 First, generate a series of vectors x with increasing noise 1 ,...x t ,...x T , x T is pure Gaussian noise;

[0016] Train the inverse diffusion model: define a new data vector y extracted from the pure Gaussian noise T , length and x T Similarly, the reverse diffusion process is carried out using the following formula:

[0017] y t-1 =y t -(x t -x t-1 )φ t

[0018] Among them, φ t It is the parameter obtained by using the following loss function:

[0019]

[0020] Furthermore, the measured earth-rock dam settlement data is used to obtain the optimal parameter combination using the Diffusion-MCMC sampling method. In this process, the Metropolis-Hastings algorithm and the diffusion model, namely Diffusion-MH, are combined to improve the efficiency of MCMC sampling, including:

[0021] First, calculate the prior distribution: determine K, The prior distribution of π(X);

[0022] Then, the log-likelihood estimation function is obtained according to the individual differences of the data as the objective function;

[0023] The following diffusion sample acceptance criteria were used:

[0024] Generate a random number. When the random number is less than the diffusion probability, perform Diffusion-MH sampling based on the initial sample set to enhance sample diversity and improve sample quality. The following formula is used to calculate the histogram probability ratio Q of the candidate sample:

[0025]

[0026] Among them, θ represents the current sample, θ' represents the candidate sample;

[0027] By substituting the candidate sample into the surrogate model to generate D', the log-likelihood ratio L is calculated using the following formula: ratio :

[0028]

[0029] Where f(θ|D) represents the posterior probability density function;

[0030] The receiving probability a is calculated using the following formula:

[0031] a=min(1,L ratio ×Q ratio )

[0032] Decide whether to accept a candidate sample by random sampling:

[0033] The following MH sample acceptance criteria were used:

[0034] When the random number is greater than or equal to the diffusion probability, PureMH sampling, namely the Metropolis-Hastings algorithm, is performed based on the initial sample set. D' is generated by substituting the candidate sample into the proxy model, and the log-likelihood ratio L is calculated using the following formula: ratio :

[0035]

[0036] Where f(θ|D) represents the posterior probability density function;

[0037] The receiving probability a' is calculated using the following formula:

[0038] a'=min(1,L ratio )

[0039] Decide whether to accept candidate samples through random sampling;

[0040] Then optimize the diffusion model: use all generated samples to train the diffusion model again until the predetermined number of retrainings is reached; each retraining has a probability to update the diffusion model parameters, and iterates repeatedly to further optimize the model;

[0041] Finally, the valid Markov chain samples are used as samples of the posterior probability density function f(X|D) to calculate the estimated value of the posterior distribution f(X|D), including the mean and standard deviation.

[0042] And, a parameter inversion system of earth-rock dam constitutive model based on Diffusion MCMC, including:

[0043] The training sample generation module is used to establish a finite element model based on the on-site survey data of the dam, select inversion parameters based on sensitivity analysis, determine the range of parameters to be inverted, and construct different parameter combinations for finite element calculation to obtain proxy model training samples;

[0044] A random forest training module, used to establish a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​based on an Ivy-RF model using the training samples;

[0045] Diffusion model training module, which is used to train the Diffusion model using the initial samples. After training, multiple groups of samples matching the input data are generated to avoid MCMC sampling falling into the local optimum.

[0046] As well, the optimal parameter combination calculation module adopts the measured earth-rock dam settlement data from the engineering project and uses the Diffusion-MCMC sampling method to obtain the optimal parameter combination.

[0047] Furthermore, the training sample generation module uses the Duncan-Zhang EB model as the basis for finite element forward analysis; through displacement sensitivity analysis, the parameters K, Finite element analysis is performed to simulate the displacement of the dam body; the settlement displacement of the monitoring equipment nodes is extracted from the simulation results as training samples for the proxy model;

[0048] The Diffusion model training module includes training of a forward diffusion model and a reverse diffusion model.

[0049] Furthermore, the optimal parameter combination calculation module is based on the calculation of the prior distribution, and uses the log-likelihood estimation function obtained according to the individual differences of the data as the objective function, and judges whether to accept the candidate sample through the diffusion sample acceptance criterion and the MH sample acceptance criterion respectively; the diffusion model is trained again using all the generated samples until a predetermined number of retraining times is reached, and each retraining probabilistically updates the diffusion model parameters, and iterates repeatedly to further optimize the model, and finally outputs the static parameter group through the calculation of the posterior distribution.

[0050] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the parameter inversion method of the earth-rock dam constitutive model based on Diffusion MCMC are implemented as described above.

[0051] A non-transitory computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a parameter inversion method for a constitutive model of an earth-rock dam based on Diffusion MCMC as described above.

[0052] Compared with the prior art, the present invention and its preferred solution effectively solve the problems of poor convergence, slow calculation speed and insufficient consideration of geotechnical material uncertainty in the traditional method. It not only effectively solves the time-consuming problem of the current finite element method, but also further improves the accuracy of parameter random inversion by optimizing the Markov chain Monte Carlo sampling (MCMC) technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0054] Figure 1 The present invention is a flowchart of a random inversion method for parameters of an earth-rock dam constitutive model based on Diffusion-MCMC according to an embodiment of the present invention.

[0055] Figure 2 It is a schematic diagram of a three-dimensional finite element model of an earth-rock dam according to an embodiment of the present invention.

[0056] Figure 3 Schematic diagram of reference point setting and selection according to an embodiment of the present invention. DETAILED DESCRIPTION

[0057] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail as follows:

[0058] It should be noted that the following detailed descriptions are illustrative and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0059] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0060] like Figure 1 As shown, the embodiment of the present invention proposes a random inversion method for parameters of earth-rock dam constitutive model based on Diffusion-MCMC, comprising the following steps:

[0061] Step 1: Based on the field survey data of the dam, a finite element model is established, key inversion parameters are selected based on sensitivity analysis, the range of parameters to be inverted is determined, and different parameter combinations are constructed for finite element calculation to obtain proxy model training samples; its specific implementation includes the following sub-steps:

[0062] Step 1-1: Based on the design drawings of the earth-rock dam and the field measurement data, the professional finite element software ABAQUS is used to construct a finite element model and perform meshing.

[0063] like Figure 2 As shown, this embodiment takes a certain slab rockfill dam as an example, and uses solid185 units to create a three-dimensional finite element model in ABAQUS. In the model, the dam body material and construction period are divided into zones for finite element calculation.

[0064] Step 1-2: Inversion reference point selection

[0065] According to the distribution of monitoring equipment on the dam cross section, the reference points for constitutive parameter inversion search are selected from the points on the dam cross section where monitoring equipment is set and corresponding to the finite element mesh, such as Figure 3 shown.

[0066] Step 1-3: Constitutive model selection

[0067] In this embodiment, the Duncan Zhang EB model is selected as the basis for finite element forward analysis.

[0068] Step 1-4: Determine the parameters to be inverted and the fluctuation range

[0069] According to the selection criteria of test indicators in parameter sensitivity analysis, the maximum vertical displacement of the dam body is taken as the main test indicator. In view of the material characteristics of rockfill dam, R in the Duncan Zhang EB model is selected. f ,m,n,K,K b , A total of six parameters were analyzed for sensitivity. Based on the parameter values ​​of the generalized plasticity model determined in the article "Jia, Yufeng, and Shichun Chi. Back-analysis of soil parameters of the Malutang II concrete face rockfill dam using parallel mutation particle swarm optimization. Computers and Geotechnics 65 (2015) 87–96", each analysis parameter was adjusted by ±30% to form three different test levels. According to the number of test factors and levels, L was selected. 18 The orthogonal table was used for testing, and finite element calculation was performed to obtain the maximum settlement data of the dam body. By analyzing the calculation results, the factors with larger ranges have greater sensitivity, and finally the parameters K,

[0070] Step 1-5: Generate a valid model parameter set

[0071] Select L 9 The orthogonal table is used to perform experiments to obtain multiple constitutive model parameter groups required for subsequent training.

[0072] Step 1-5: Generate training samples

[0073] Substitute all constitutive model parameter groups into finite element calculations to obtain multiple groups of settlement results of measuring points. Arrange the input and output data of the training samples.

[0074] Step 2: Using the training samples, a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​is established based on the Ivy-RF model. Its specific implementation includes the following sub-steps:

[0075] Step 2-1: Data preprocessing

[0076] Considering the large difference in the values ​​of the input material parameters, the input data of the training samples (K, ) for normalization.

[0077] Step 2-2: RF model training

[0078] The preprocessed training set samples are input into the random forest regression algorithm; in order to improve the regression prediction accuracy of the random forest data, the number of trees and the minimum number of leaf points in the random forest are optimized using the Ivy optimization algorithm. The fitness function is the mean square error (MSE) of the RF model for the settlement values ​​of the training set and the settlement values ​​of the test set. The lower the MSE, the better. The fitness is calculated using the following formula.

[0079] fitness=MSE[predict(train)]+MSE[predict(test)] (1)

[0080] Retrain the RF model using the optimal hyperparameter configuration to obtain the best prediction effect and model performance.

[0081] Step 3: Use the initial samples to train the diffusion model. After training, multiple groups of samples matching the input data can be generated to avoid MCMC sampling falling into the local optimum. Its specific implementation includes the following sub-steps:

[0082] Step 3-1: Train the forward diffusion model

[0083] The diffusion model is trained using the initial samples. The diffusion model uses the following formula to perform the forward diffusion process, gradually converting the initial samples into pure Gaussian noise:

[0084]

[0085] Among them, ε~N(0,1) is the noise weight added to each time step, β t is the noise width in the diffusion process, from the original vector x 0 First, generate a series of vectors x with increasing noise 1 ,...x t ,...x T , x T is pure Gaussian noise.

[0086] Step 3-2: Training the reverse diffusion model

[0087] Define a new data vector y extracted from pure Gaussian noise T , whose length is the same as x T The reverse diffusion process is carried out using the following formula

[0088] y t-1 =y t -(x t -x t-1 )φ t (3)

[0089] Among them, φ t is the parameter obtained by using the following loss function

[0090]

[0091] Step 4: Using the measured earth-rock dam settlement data from the project, the Diffusion-MCMC sampling method is used to obtain the optimal parameter combination; its specific implementation includes the following sub-steps, such as Figure 1 As shown:

[0092] Step 4-1: Calculation of prior distribution

[0093] Determine K. The prior distribution of π(X);

[0094] K~N(1467,225 2 )

[0095]

[0096] Step 4-2: Construct the objective function

[0097] The log-likelihood estimation function is obtained based on the individual differences in the data;

[0098] Step 4-3: Diffusion Sample Acceptance Criteria

[0099] Generate a random number. When the random number is less than the diffusion probability, perform Diffusion-MH sampling based on the initial sample set to enhance sample diversity and improve sample quality. The following formula is used to calculate the histogram probability ratio Q of the candidate sample:

[0100]

[0101] Among them, θ represents the current sample and θ' represents the candidate sample.

[0102] By substituting the candidate sample into the surrogate model to generate D', the log-likelihood ratio L is calculated using the following formula: ratio

[0103]

[0104] Where f(θ|D) represents the posterior probability density function.

[0105] The receiving probability a is calculated using the following formula:

[0106] a=min(1,L ratio ×Q ratio ) (7)

[0107] Decide whether to accept candidate samples by random sampling

[0108] Step 4-4: MH Sample Acceptance Criteria

[0109] When the random number is greater than or equal to the diffusion probability, PureMH sampling is performed based on the initial sample set. D' is generated by substituting the candidate sample into the proxy model, and the log-likelihood ratio L is calculated using the following formula: ratio

[0110]

[0111] Where f(θ|D) represents the posterior probability density function.

[0112] The receiving probability a is calculated using the following formula:

[0113] a=min(1,L ratio ) (8)

[0114] Decide whether to accept candidate samples by random sampling

[0115] Step 4-5: Optimize the diffusion model

[0116] The diffusion model is trained again using all generated samples until a predetermined number of retrainings is reached. Each retraining may update the diffusion model parameters, and the model is iterated repeatedly to further optimize the model.

[0117] Step 4-6: Calculation of the posterior distribution

[0118] Take the valid Markov chain samples as samples of the posterior probability density function f(X|D) and calculate the estimated value (mean, standard deviation) of the posterior distribution f(X|D).

[0119] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0120] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1A device that provides the functions specified in a block or multiple blocks.

[0121] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0122] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0123] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0124] The above is only a preferred embodiment of the present invention, and does not limit the present invention in other forms. Any technician familiar with the profession may use the above disclosed technical content to change or modify it into an equivalent embodiment with equivalent changes. However, any simple modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention still belongs to the protection scope of the technical solution of the present invention.

[0125] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of a parameter inversion method of the earth-rock dam constitutive model based on Diffusion MCMC under the inspiration of this patent. All equal changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of this patent.

Claims

1. A parameter inversion method for earth-rock dam constitutive model based on Diffusion MCMC, characterized by: According to the field survey data of the dam, a finite element model is established, the inversion parameters are selected according to the sensitivity analysis, and the range of the parameters to be inverted is determined. Different parameter combinations are constructed for finite element calculation to obtain the proxy model training samples. Then, the training samples are used to establish a nonlinear mapping between the constitutive parameters and the settlement values ​​of the earth-rock dam based on the Ivy-RF model. The Diffusion model is trained with the initial samples, and multiple groups of samples matching the input data are generated after training to avoid MCMC sampling falling into the local optimum. Finally, the settlement data of the earth-rock dam measured in the project are used to obtain the optimal parameter combination using the Diffusion-MCMC sampling method.

2. The parameter inversion method of earth-rock dam constitutive model based on Diffusion MCMC according to claim 1 is characterized by: The method of establishing a finite element model based on the field survey data of the dam, selecting inversion parameters based on sensitivity analysis, determining the range of parameters to be inverted, and constructing different parameter combinations for finite element calculation to obtain agent model training samples is as follows: construct a finite element model and perform meshing based on the design drawings and field measurement data of the earth-rock dam; perform finite element calculation in the finite element model according to the dam body material and construction period partition; for inversion reference point selection: based on the distribution of monitoring equipment on the dam body cross section, select reference points for constitutive parameter inversion search from points on the dam body cross section where monitoring equipment is set and corresponding to the finite element grid; use the Duncan-Zhang EB model as the basis for finite element forward analysis; determine the parameters K, Based on engineering experience, the impact factor K is set. The value range and number of levels of each influencing factor; select a suitable orthogonal table for the number of levels of each influencing factor, design the table header, and perform the test; perform statistical analysis on the test results and extract multiple constitutive model parameter groups; Finite element analysis is performed using these parameters to simulate the displacement of the dam body; the settlement displacement of the monitoring equipment nodes is extracted from the simulation results as training samples for the proxy model.

3. The parameter inversion method of earth-rock dam constitutive model based on Diffusion MCMC according to claim 2 is characterized by: Using the training samples, a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​is established based on the Ivy-RF model, including: Input data for training samples Normalization is performed for preprocessing; the preprocessed data is input into the random forest regression RF model as a training set; and the number of trees and the minimum number of leaf nodes in the random forest are optimized using the Ivy optimization algorithm during the training process; the fitness function is the mean square error (MSE) of the RF model for the settlement values ​​of the training set and the test set; fitness=MSE[predict(train)]+MSE[predict(test)] Finally, the optimal hyperparameter configuration is used to retrain the RF model to obtain the best prediction effect and model performance.

4. The parameter inversion method of earth-rock dam constitutive model based on Diffusion MCMC according to claim 3 is characterized by: The initial samples are used to train the Diffusion model, and after training, multiple groups of samples matching the input data are generated, including: Training the forward diffusion model: The diffusion model is trained using the initial samples. The diffusion model uses the following formula to perform the forward diffusion process and gradually transforms the initial samples into pure Gaussian noise: Among them, ε~N(0,1) is the noise weight added to each time step, β t is the noise width in the diffusion process. Starting from the original vector x0, a series of vectors x1,...x with gradually increasing noise are generated. t ,...x T , x T is pure Gaussian noise; Train the inverse diffusion model: define a new data vector y extracted from the pure Gaussian noise T , length and x T Similarly, the reverse diffusion process is carried out using the following formula: and t-1 =and t -(x t -x t-1 )φ t Among them, φ t It is the parameter obtained by using the following loss function:

5. The parameter inversion method of earth-rock dam constitutive model based on Diffusion MCMC according to claim 4 is characterized by: The measured earth-rock dam settlement data is used to obtain the optimal parameter combination using the Diffusion-MCMC sampling method, including: First, calculate the prior distribution: determine K, The prior distribution of π(X); Then, the log-likelihood estimation function is obtained according to the individual differences of the data as the objective function; The following diffusion sample acceptance criteria were used: Generate a random number. When the random number is less than the diffusion probability, perform Diffusion-MH sampling based on the initial sample set to enhance sample diversity and improve sample quality. The following formula is used to calculate the histogram probability ratio Q of the candidate sample: Among them, θ represents the current sample, θ' represents the candidate sample; By substituting the candidate sample into the surrogate model to generate D', the log-likelihood ratio L is calculated using the following formula: ratio : Where f(θ|D) represents the posterior probability density function; The receiving probability a is calculated using the following formula: a=min(1,L ratio ×Q ratio ) Decide whether to accept a candidate sample by random sampling: The following MH sample acceptance criteria were used: When the random number is greater than or equal to the diffusion probability, PureMH sampling is performed based on the initial sample set. D' is generated by substituting the candidate sample into the proxy model, and the log-likelihood ratio L is calculated using the following formula: ratio : Where f(θ|D) represents the posterior probability density function; The receiving probability a' is calculated using the following formula: a'=min(1,L ratio ) Decide whether to accept candidate samples through random sampling; Then optimize the diffusion model: use all generated samples to train the diffusion model again until the predetermined number of retraining is reached; each retraining has the probability of updating the diffusion model parameters, and iterates repeatedly to further optimize the model; Finally, the valid Markov chain samples are used as samples of the posterior probability density function f(X|D) to calculate the estimated value of the posterior distribution f(X|D), including the mean and standard deviation.

6. A parameter inversion system for earth-rock dam constitutive model based on DiffusionMCMC, characterized in that: include: The training sample generation module is used to establish a finite element model based on the on-site survey data of the dam, select inversion parameters based on sensitivity analysis, determine the range of parameters to be inverted, and construct different parameter combinations for finite element calculation to obtain proxy model training samples; A random forest training module, used to establish a nonlinear mapping between constitutive parameters and earth-rock dam settlement values ​​based on an Ivy-RF model using the training samples; Diffusion model training module, which is used to train the Diffusion model using the initial samples. After training, multiple groups of samples matching the input data are generated to avoid MCMC sampling falling into the local optimum. As well, the optimal parameter combination calculation module adopts the measured earth-rock dam settlement data from the engineering project and uses the Diffusion-MCMC sampling method to obtain the optimal parameter combination.

7. The earth-rock dam constitutive model parameter inversion system based on DiffusionMCMC according to claim 6, characterized in that: The training sample generation module uses the Duncan-Zhang EB model as the basis for finite element forward analysis; through displacement sensitivity analysis, the parameters K, Finite element analysis is performed to simulate the displacement of the dam body; the settlement displacement of the monitoring equipment node is extracted from the simulation result as a sample for training the proxy model; the Diffusion model training module includes the training of the forward diffusion model and the reverse diffusion model.

8. The earth-rock dam constitutive model parameter inversion system based on Diffusion MCMC according to claim 7, characterized in that: The optimal parameter combination calculation module is based on the calculation of the prior distribution, and uses the log-likelihood estimation function obtained according to the individual differences of the data as the objective function, and judges whether to accept the candidate sample through the diffusion sample acceptance criterion and the MH sample acceptance criterion respectively; the diffusion model is trained again using all the generated samples until the predetermined number of retraining times is reached, and the diffusion model parameters are updated probabilistically each time the retraining is performed, and it is iterated repeatedly to further optimize the model, and finally the static parameter group is output through the calculation of the posterior distribution.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the parameter inversion method of the earth-rock dam constitutive model based on Diffusion MCMC as described in any one of claims 1 to 5 are implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the parameter inversion method of the earth-rock dam constitutive model based on Diffusion MCMC as described in any one of claims 1 to 5 are implemented.

Images