Data screening method for earth and rockfill dam parameter inversion

Abstract

The invention discloses a data screening method for earth and rockfill dam parameter inversion, belongs to the technical field of earth and rockfill dam parameter inversion, and can solve the problem that existing earth and rockfill dam inversion analysis is low in accuracy and efficiency. The method comprises the following steps: S1, establishing a finite element model of the earth and rockfill dam, and determining a plurality of to-be-screened material parameters according to a constitutive model adopted by the finite element model; the finite element model has a plurality of nodes; s2, generating a parameter group sample set of the plurality of material parameters, and establishing a displacement agent model of each node by using the parameter group sample set and the finite element model; the displacement agent model is used for simulating the mapping relation between the displacement of the node and the plurality of material parameters; and S3, determining parameters to be inverted from the plurality of material parameters according to the displacement agent models of all the nodes, and determining monitoring data for parameter inversion according to the parameters to be inverted. The method is used for screening the to-be-inverted data of the earth and rockfill dam.

Description

Technical Field

[0001] The invention relates to a data screening method for earth-rock dam parameter inversion, and belongs to the technical field of earth-rock dam parameter inversion. Background Art

[0002] With the continuous development of water conservancy projects in my country, the height of earth-rock dams under construction has been continuously breaking through, and the evolution of the stress state and deformation characteristics of the dam has become increasingly obvious in terms of the safety of the dam structure and the risk of engineering operation. In order to accurately describe the stress-deformation characteristics of earth-rock dams, scholars have proposed a variety of constitutive models to describe the stress-deformation relationship of earth-rock materials, such as the Duncan-Zhang EB model and the elastoplastic model. These constitutive models all show the characteristics of high nonlinearity and multiple parameters. In addition, the uncertainty difference between the experimental parameters and the actual parameters makes the analysis process time-consuming and the uncertainty error of the analysis results difficult to control. For this reason, it is necessary to perform inverse analysis on the material parameters of the earth-rock dam to determine the accurate values ​​of the material parameters.

[0003] Due to the complexity of the stress-deformation characteristics of earth-rock dams, there are many material parameters in their constitutive models. In addition, due to the structural characteristics of earth-rock dams, there are often multiple partitions with different materials inside the dam body. If all material parameters in each material partition are inverted, it will take a long time. Therefore, it is necessary to select material parameters that are more sensitive to dam body deformation as parameters to be inverted to improve the efficiency of inversion analysis.

[0004] In addition, in the parameter inversion analysis, the screening of the monitoring data of the dam displacement is equally important, which directly affects the rationality of the inversion analysis results. Studies have shown that there are obvious differences in the displacements of different parts of the dam body corresponding to the same material parameters. Therefore, the use of monitoring data from different parts of the dam body for inversion analysis often leads to large differences in the inversion analysis results. When the displacement of the part where the monitoring data is located is more sensitive to the material parameters, the inversion analysis results are more realistic; when the displacement of the part where the monitoring data is located is less sensitive to the material parameters, the rationality of the inversion analysis results is greatly reduced. At present, the screening of monitoring data in parameter inversion mostly relies on the personal experience of researchers, and when the error of the inversion analysis result is large, it is necessary to re-screen the monitoring data and re-perform the inversion analysis until the error meets the requirements, resulting in low accuracy and efficiency of the inversion analysis. Summary of the invention

[0005] The invention provides a data screening method for earth-rock dam parameter inversion, which can solve the problems of low accuracy and low efficiency in the existing earth-rock dam parameter inversion analysis.

[0006] The present invention provides a data screening method for earth-rock dam parameter inversion, the method comprising:

[0007] S1. Establishing a finite element model of an earth-rock dam, and determining a plurality of material parameters to be screened according to a constitutive model adopted by the finite element model; the finite element model has a plurality of nodes;

[0008] S2, generating a parameter group sample set of the plurality of material parameters, and using the parameter group sample set and the finite element model to establish a displacement proxy model of each node; the displacement proxy model is used to simulate the mapping relationship between the displacement of the node and the plurality of material parameters;

[0009] S3. Determine the parameters to be inverted from the multiple material parameters according to the displacement proxy models of all nodes, and determine the monitoring data for parameter inversion according to the parameters to be inverted.

[0010] Optionally, the earth-rock dam is a rockfill dam, and the rockfill dam includes a rockfill body;

[0011] The constitutive model adopted in the region corresponding to the rockfill body on the finite element model is a nonlinear elastic constitutive model; and the multiple material parameters are all material parameters of the nonlinear elastic constitutive model.

[0012] Optionally, the rockfill body includes a plurality of rockfill areas with different materials; the parameter group sample set of the plurality of material parameters generated in S2 is specifically:

[0013] A parameter group sample set of the plurality of material parameters is generated according to the material of any one of the plurality of rockfill areas and according to a preset number of samples.

[0014] Optionally, the displacement proxy model of each node is established by using the parameter group sample set and the finite element model in S2, specifically:

[0015] Using the parameter group sample set and the finite element model, obtaining a displacement sample set of a preset position on the finite element model;

[0016] Establishing a displacement proxy model of the preset position according to the parameter group sample set and the displacement sample set of the preset position;

[0017] A displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position.

[0018] Optionally, the displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position, specifically:

[0019] Determining a calculation error of the displacement proxy model at the preset position, increasing a preset number of samples when the calculation error does not satisfy a preset error range, and updating the parameter group sample set according to the increased preset number of samples;

[0020] Re-establishing the displacement proxy model of the preset position according to the updated parameter group sample set until the calculation error of the displacement proxy model of the preset position satisfies the preset error range;

[0021] According to the parameter group sample set corresponding to the displacement proxy model of the preset position finally established, and using the finite element model, a displacement proxy model of each node on the finite element model is established.

[0022] Optionally, in S3, the parameters to be inverted are determined from the multiple material parameters according to the displacement proxy models of all nodes, specifically:

[0023] According to the displacement proxy models of all nodes, the sensitivity index of the displacement of each node to each material parameter is determined;

[0024] The parameters to be inverted in each rockfill area are determined from the multiple material parameters according to the sensitivity index.

[0025] Optionally, the monitoring data for parameter inversion is determined according to the parameter to be inverted in S3, specifically:

[0026] The displacement proxy models of all nodes are updated according to the parameters to be inverted in each rockfill area, and the monitoring data for parameter inversion are determined according to the updated displacement proxy models of all nodes.

[0027] Optionally, the updating of the displacement proxy models of all nodes according to the parameters to be inverted in each rockfill area is specifically as follows:

[0028] Generate a parameter group sample set of all parameters to be inverted according to the materials in each rockfill area;

[0029] According to the parameter group sample set of all parameters to be inverted, the displacement proxy models of all nodes are updated by using the finite element model.

[0030] Optionally, the monitoring data for parameter inversion is determined according to the displacement proxy models of all nodes after the update, specifically:

[0031] According to the displacement proxy models of all nodes after the update, the sensitivity index of the displacement of each node to each parameter to be inverted is determined;

[0032] Monitoring data for parameter inversion is determined according to the sensitivity index.

[0033] Optionally, the determining of the monitoring data for parameter inversion according to the sensitivity index is specifically:

[0034] Determining the sensitivity distribution of each parameter to be inverted on the finite element model according to the sensitivity index;

[0035] Sensitive monitoring points on the rockfill body are determined according to the sensitivity distribution, and displacement monitoring data of the sensitive monitoring points are determined as monitoring data for parameter inversion.

[0036] The beneficial effects that the present invention can produce include:

[0037] The present invention uses a finite element model of an earth-rock dam and a parameter group sample set of multiple material parameters to establish a displacement proxy model of all nodes on the finite element model, and determines the parameters to be inverted according to the displacement proxy model, and then determines the monitoring data used for parameter inversion according to the parameters to be inverted, so that the parameters to be inverted that have a greater impact on the displacement of the dam body and the monitoring data that are more sensitive to the parameters to be inverted can be screened out, which is beneficial to improving the accuracy and efficiency of the inversion analysis.

[0038] The present invention determines the parameters to be inverted of each partition according to the partitions of different materials inside the earth-rock dam body, and then determines the monitoring data used for parameter inversion according to the parameters to be inverted of each partition, so that the screening process of the parameters to be inverted and the monitoring data can be combined with the structural characteristics of the earth-rock dam, and fully consider the mutual influence of stress and strain between multiple partitions and the influence of the deformation of each partition on the overall displacement of the earth-rock dam, thereby improving the accuracy of data screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A schematic flow chart of a data screening method for earth-rock dam parameter inversion provided by an embodiment of the present invention;

[0040] Figure 2 A specific flow chart of the data screening method provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of a finite element model of an earth-rock dam provided in an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of the arrangement of monitoring points on the largest cross section provided by an embodiment of the present invention;

[0043] Figure 5 A statistical graph of the ranking of the sensitivity index of each material parameter of the maximum cross-sectional displacement under the single material assumption provided in an embodiment of the present invention, wherein (a) is a statistical graph of the ranking of the sensitivity index of each material parameter of the horizontal downstream displacement, (b) is a statistical graph of the ranking of the sensitivity index of each material parameter of the vertical displacement, and (a) is a statistical graph of the ranking of the sensitivity index of each material parameter of the horizontal downstream displacement;

[0044] Figure 6 The parameters K to be inverted for different material partitions on the maximum cross section provided by the embodiment of the present invention are bThe vertical displacement sensitivity index distribution contour map of the upstream rockfill area is shown in Figure 2, where (a) is the parameter K to be inverted in the upstream rockfill area. b1 The vertical displacement sensitivity index distribution contour map, (b) is the parameter K to be inverted in the main rockfill area b2 (c) is the inversion parameter K in the downstream rockfill area. b3 The vertical displacement sensitivity index distribution contour map;

[0045] Figure 7 The vertical displacement sensitivity index distribution contour map of the parameter n to be inverted for different material zones on the maximum cross section provided by the embodiment of the present invention; wherein (a) is the vertical displacement sensitivity index distribution contour map of the parameter n1 to be inverted in the upstream rockfill area, (b) is the vertical displacement sensitivity index distribution contour map of the parameter n2 to be inverted in the main rockfill area, and (c) is the vertical displacement sensitivity index distribution contour map of the parameter n3 to be inverted in the downstream rockfill area;

[0046] Figure 8 The parameters K to be inverted for different material partitions on the maximum cross section provided by the embodiment of the present invention are b The distribution contour map of the horizontal downstream displacement sensitivity index, where (a) is the parameter K to be inverted in the upstream rockfill area b1 (a) The distribution contour map of the sensitivity index of horizontal downstream displacement, (b) the parameter K to be inverted in the main rockfill area b2 (c) is the inversion parameter K in the downstream rockfill area. b3 The distribution contour map of the horizontal downstream displacement sensitivity index;

[0047] Fig. 9 The contour map of the horizontal downstream displacement sensitivity index distribution of the parameter n to be inverted for different material zones on the maximum cross section provided by the embodiment of the present invention. (a) is the contour map of the horizontal downstream displacement sensitivity index distribution of the parameter n1 to be inverted in the upstream rockfill area, (b) is the contour map of the horizontal downstream displacement sensitivity index distribution of the parameter n2 to be inverted in the main rockfill area, and (c) is the contour map of the horizontal downstream displacement sensitivity index distribution of the parameter n3 to be inverted in the downstream rockfill area.

[0048] Reference numerals:

[0049] 1. Concrete face plate; 2. Concrete toe plate; 3. Extrusion wall; 4. Pad layer; 5. Transition layer; 6. Upstream rockfill area; 7. Main rockfill area; 8. Downstream rockfill area. DETAILED DESCRIPTION

[0050] The present invention is described in detail below in conjunction with embodiments, but the present invention is not limited to these embodiments.

[0051] The embodiment of the present invention provides a data screening method for earth-rock dam parameter inversion, see Figure 1 , the method comprising:

[0052] S1. Establish a finite element model of an earth-rock dam, and determine multiple material parameters to be screened according to a constitutive model adopted by the finite element model; the finite element model has multiple nodes;

[0053] S2. Generate a parameter group sample set of multiple material parameters, and use the parameter group sample set and the finite element model to establish a displacement proxy model for each node; the displacement proxy model is used to simulate the mapping relationship between the displacement of the node and the multiple material parameters;

[0054] S3. Utilizing the displacement proxy models of all nodes, the parameters to be inverted are determined from a plurality of material parameters, and monitoring data for parameter inversion is determined based on the parameters to be inverted.

[0055] Specifically, this embodiment establishes a finite element model of the earth-rock dam based on the engineering design and construction data of the earth-rock dam.

[0056] Specifically, the engineering design and construction data of earth-rock dams include: earth-rock dam plan layout, earth-rock dam cross-section design, dam deformation monitoring layout, earth-rock dam construction progress report, etc. Based on the engineering design data, the division of different materials in the earth-rock dam, the layered rolling construction, and the layout of the dam displacement monitoring points can be determined, and then the finite element model of the earth-rock dam can be established. Figure 3 .

[0057] Specifically, the mesh of the finite element model of the earth-rock dam can be quadrilateral units (two-dimensional model) or hexahedral units (three-dimensional model). The size of the unit mesh is 1 / 20 to 1 / 30 of the dam height and should not exceed 10m. A fixed boundary is set at the bottom of the finite element model, and a water pressure boundary is set on the upstream surface. At the same time, a layered filling method is set to simulate the construction steps, and a water pressure load loading method is set to simulate the staged water storage. To facilitate the screening of monitoring data, nodes are set at the positions corresponding to the positions of the dam displacement monitoring points on the earth-rock dam on the finite element model, refer to Figure 4 .

[0058] Specifically, earth-rock dams include earth-rock dams constructed of a single material and earth-rock dams constructed of multiple materials. Among them, earth-rock dams constructed of a single material include homogeneous dams, etc., and earth-rock dams constructed of multiple materials include rockfill dams, core wall earth dams, inclined wall earth-rock dams, etc.

[0059] Specifically, the constitutive models of earth-rock dams include linear elastic constitutive models, nonlinear elastic constitutive models, plastic constitutive models, viscoelastic constitutive models, hyperelastic constitutive models, elastoplastic constitutive models, damage constitutive models, etc. In practice, the corresponding constitutive model needs to be selected according to the material properties and analysis requirements of the earth-rock dam.

[0060] In this embodiment, the earth-rock dam is a concrete face rockfill dam (hereinafter referred to as a face rockfill dam). Figure 3 The structure of the face rockfill dam is, from upstream to downstream, a concrete face plate 1, a concrete toe plate 2, an extrusion wall 3, a cushion layer 4, a transition layer 5 and a rockfill body.

[0061] The rockfill body can usually be divided into a plurality of zones made of different materials. In this embodiment, the rockfill zone includes three zones, which are an upstream rockfill zone 6, a main rockfill zone 7 and a downstream rockfill zone 8 from upstream to downstream.

[0062] Specifically, the constitutive model used in the area corresponding to the soil and rock materials (cushion 4, transition layer 5 and rockfill body) on the finite element model is a nonlinear elastic constitutive model, and the constitutive model used in the area corresponding to the concrete materials (concrete panel 1, concrete toe plate 2, extrusion wall 3) is a linear elastic constitutive model.

[0063] Specifically, the nonlinear elastic constitutive model may include the Duncan-Zhang EB model, the Duncan-Zhang E-ν model, the Tsinghua nonlinear decoupling KG model, etc., and the elastoplastic model may include the Cambridge model, the Lade model, the Shen Zhujiang double yield surface model, etc.

[0064] The nonlinear elastic constitutive model used in this embodiment is the Duncan-Zhang EB model.

[0065] In practice, the displacement of the dam body is basically only affected by the deformation of the rockfill body, so the material parameter values ​​of the constitutive model of the concrete face plate 1 and the concrete toe plate 2, the extrusion wall 3, the cushion layer 4 and the transition layer 5 can all be regarded as constant values, and the design parameter values ​​can be used. In this embodiment, it is only necessary to screen the material parameters of the constitutive model of the rockfill body, that is, to screen the material parameters in the Duncan-Zhang EB model.

[0066] The material parameters in the Duncan-Zhang EB model include cohesion C, initial internal friction angle Internal friction angle increment Initial tangent modulus intercept K, initial tangent modulus slope n, tangent bulk modulus intercept K b , tangent bulk modulus slope m, unloading modulus intercept K ur and destruction ratio R f Among them, the cohesion C of the rockfill body is generally 0, so in this embodiment, only the numerical changes of the remaining 8 material parameters are considered, and the parameters to be inverted are screened from these 8 material parameters.

[0067] The rockfill body of this embodiment includes three rockfill areas made of different materials, namely, an upstream rockfill area 6, a main rockfill area 7 and a downstream rockfill area 8. Since the mechanical properties of the rockfill areas made of different materials are different, the values ​​of the eight material parameters of the Duncan-Zhang EB model corresponding to each rockfill area are different.

[0068] Therefore, there are 24 material parameters that affect the displacement of the concrete face rockfill dam, including the initial internal friction angle of the upstream rockfill area 6, Internal friction angle increment Initial tangent modulus intercept K1, initial tangent modulus slope n1, tangent bulk modulus intercept K b1 , tangent bulk modulus slope m1, unloading modulus intercept K ur1 and destruction ratio R f1 ; Initial internal friction angle of the main rockfill area 7 Internal friction angle increment Initial tangent modulus intercept K2, initial tangent modulus slope n2, tangent bulk modulus intercept K b2 , tangent bulk modulus slope m2, unloading modulus intercept K ur2 and destruction ratio R f2 ; and the initial internal friction angle of the downstream rockfill area 8 Internal friction angle increment Initial tangent modulus intercept K3, initial tangent modulus slope n3, tangent bulk modulus intercept K b3 , tangent bulk modulus slope m3, unloading modulus intercept K ur3 and destruction ratio R f3 .

[0069] Obviously, if the 24 material parameters are inverted at the same time, the calculation difficulty is large and the time consumption is long, resulting in low efficiency of the inversion analysis. To solve this problem, this embodiment selects the material parameters that have a greater impact on the dam body displacement from the 24 material parameters as the parameters to be inverted, which can greatly reduce the calculation difficulty, reduce the calculation amount and shorten the calculation time while ensuring the analysis accuracy, thereby improving the efficiency of the inversion analysis.

[0070] Specifically, this embodiment screens the parameters to be inverted by performing sensitivity analysis on various material parameters.

[0071] It is understandable that although sensitivity analysis can be performed on all 24 material parameters, the large number of material parameters makes the sensitivity analysis time-consuming, resulting in low parameter screening efficiency. Considering that the 8 material parameters of each rockfill area correspond to the 8 material parameters of the Duncan-Zhang EB model, in order to improve the screening efficiency, this implementation selects any rockfill area from multiple rockfill areas and assumes that the materials of the remaining rockfill areas are the same as those of the selected rockfill area, that is, assuming that the rockfill body is composed of a single material, so that the rockfill body has only 8 material parameters, namely K,n,K b , m, K ur and R f In this embodiment, only sensitivity analysis of the eight material parameters is required to screen out the parameters to be inverted corresponding to a single material, and then the screened parameters to be inverted are combined with the three rockfill areas to obtain the parameters to be inverted for each rockfill area. In this way, there is no need to perform sensitivity analysis on all material parameters, which effectively improves the efficiency of parameter screening.

[0072] In this embodiment, a parameter group sample set of multiple material parameters is generated in S2, which may be:

[0073] According to the material of any one of the multiple rockfill areas, a parameter group sample set of multiple material parameters is generated according to a preset number of samples.

[0074] In this embodiment, it is assumed that the numerical distributions of the eight material parameters in the Duncan-Zhang EB model all obey the normal distribution N(μ,σ 2 ), where the mean μ is the design parameter value of the corresponding material parameter, and the coefficient of variation C υ is 0.1, then the standard deviation is σ=μC υ Among them, the mean value μ is related to the characteristics of the material. Different materials have different values ​​of the mean value μ.

[0075] According to the probability distribution form of the material parameter values, a parameter group sample set of multiple material parameters is generated by sampling according to a preset sample quantity, and each parameter group sample in the parameter group sample set contains a sampling value of the above 8 material parameters.

[0076] Specifically, this embodiment assumes that the materials of the upstream rockfill area 6 and the downstream rockfill area 8 are the same as the material of the main rockfill area 7, that is, it is assumed that the rockfill body is composed of a single material of the main rockfill area 7. According to the normal distribution form N(μ,σ 2 ), Latin Hypercube Sampling (LHS) method is used to generate a material parameter group sample set according to the preset sample quantity t. The parameter group sample set generated by sampling is recorded as X = {X1, X2, ..., X t}, where each parameter group sample X i All contain the single sampling values ​​of the eight parameters of the Duncan-Zhang EB model, namely

[0077] In this embodiment, the displacement proxy model of each node is established by using the parameter group sample set and the finite element model in S2, which can be specifically:

[0078] Using the parameter group sample set and the finite element model, a displacement sample set of a preset position on the finite element model is obtained;

[0079] Establishing a displacement proxy model of the preset position according to the parameter group sample set and the displacement sample set of the preset position;

[0080] A displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position.

[0081] Furthermore, the displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position, specifically:

[0082] Determine the calculation error of the displacement proxy model at the preset position, increase the preset number of samples when the calculation error does not meet the preset error range, and update the parameter group sample set according to the increased preset number of samples;

[0083] Re-establishing the displacement proxy model of the preset position according to the updated parameter group sample set until the calculation error of the displacement proxy model of the preset position meets the preset error range;

[0084] According to the parameter group sample set corresponding to the displacement proxy model of the preset position finally established, and using the finite element model, a displacement proxy model of each node on the finite element model is established.

[0085] This embodiment verifies and evaluates the calculation accuracy of the displacement proxy model by first establishing a displacement proxy model at a preset position on the finite element model and determining the calculation error of the displacement proxy model, and then updates the parameter group sample set by increasing the preset number of samples to ensure that the calculation error of the displacement proxy model meets the preset error range. Then, based on the updated parameter group sample set, the displacement proxy models of all nodes on the finite element model are established. Compared with directly establishing the displacement proxy models of all nodes, this can effectively improve the construction efficiency of the displacement proxy models of all nodes. Specifically, due to the uneven overall deformation of the earth-rock dam, it generally presents the deformation characteristics of "large in the middle, small on both sides, large inside, and small outside". In order to more clearly show the influence of different material parameters on the displacement of the dam body, the position with the largest deformation on the dam body should be selected as the preset position. Specifically, refer to Figure 3 In the three-dimensional coordinate system, the X direction is the horizontal river direction, the Y direction is the vertical direction, and the Z direction is the horizontal cross-river direction. In this embodiment, the largest cross-section is selected from the cross-sections of the dam body perpendicular to the horizontal cross-river direction, and then the position with the largest vertical displacement is selected from the largest cross-section as the preset position.

[0086] In this embodiment, the parameter group sample set X={X1,X2,...,X t} is substituted into the finite element model, and the displacement sample set Y = {Y1, Y2, ..., Y t}; then according to the parameter group sample set X={X1,X2,...,X t} and displacement sample set Y = {Y1, Y2, ..., Y t}, establish a displacement proxy model at a preset position.

[0087] More specifically, the displacement of the preset position includes its X-direction displacement (horizontal displacement along the river), Y-direction displacement (vertical displacement) and Z-direction displacement (horizontal displacement across the river), that is, the displacement sample set Y = {Y1, Y2, ..., Y t Each displacement sample in} is Y i ={[Y Xi ,Y Yi ,Y Zi ]}, i=1,2,...,t.

[0088] Integration parameter set sample set X = {X1, X2, ..., X t} and the displacement sample set Y of the preset position i ={[Y Xi ,Y Yi ,Y Zi ]} forms a "parameter-displacement" sample set T = {X, Y} = {[X1, Y1], ..., [X t ,Y t ]}. Then the data in the "parameter-displacement" sample set of the preset position is divided according to the ratio of 2:1 to form the training set and test set of the preset position, that is:

[0089] The training set of the preset position is T = {X, Y} = {[X1, Y1], ..., [X 2t / 3 ,Y 2t / 3 ]};

[0090] The test set of the preset position is T′={X′,Y′}={[X 2t / 3+1 ,Y 2t / 3+1 ],...,[X t ,Y t ]}.

[0091] Then, this embodiment constructs a displacement proxy model of the preset position based on the probability distribution form of the material parameter value and using the training set and sample set of the preset position.

[0092] Specifically, this embodiment uses a sparse polynomial chaos expansion (sPCE) method to construct a displacement proxy model.

[0093] Specifically, in the probability space (Ω, F, P), if the probability density function of any random variable is square integrable, the random response model established based on the random variable can be expanded into an infinite series, that is:

[0094]

[0095] In formula (1), y is the model response; are the coefficients of the polynomial to be solved; are random variables that are independent of each other and obey a certain probability distribution, and are the selected input variables in uncertainty analysis; x0, x1(·), x2(·), x3(·),... are basis functions of the chaos expansion. The basis functions are a set of orthogonal function families, and their types are determined by the probability distribution form of the random variables; s is the number of random variables; i1, i2, i3, ... = 1, 2, ..., s.

[0096] Since the numerical distribution of material parameters is assumed to follow a normal distribution, the Hermite orthogonal polynomial is selected according to the Askey scheme As the basis function to establish the complete PCE proxy model. Convert equation (1) into an expression arranged by the number of expanded terms:

[0097]

[0098] In formula (2), y is the model response; a j for is the abbreviation of ; ξ is a random variable abbreviation of; Abbreviation of .

[0099] In order to meet the requirements of computational efficiency and accuracy (convergence), 2nd to 3rd order polynomial expansion is generally used for fitting analysis, and the relative error can be basically controlled within 5%.

[0100] Furthermore, a least squares minimization problem is established to solve the expansion term coefficients of the PCE proxy model. The infinite series in equation (2) can be written as the sum of its truncation equation and truncation error:

[0101]

[0102] In formula (3), y is the model response; P = cardA M,P ; ε P is the truncation error; a Τ ={a0,...,a P-1} Τ is the coefficient vector of the undetermined expansion term; Ψ(x)={Ψ0(x),...,Ψ P-1 (x)} Τ is the vector of all orthogonal polynomial values. Then the least squares minimization problem can be transformed into:

[0103]

[0104] In formula (4), is the estimated value of the coefficient vector of the expansion term; a Τ ={a0,...,a P-1} Τ is the coefficient vector of the undetermined expansion term; Ψ(x)={Ψ0(x),...,Ψ P-1 (x)} Τ is the vector of all orthogonal polynomial values.

[0105] Since different input variables (material parameters) have different degrees of influence on the output variables (dam displacement), when solving the expansion coefficients of the above PCE proxy model, the sparsity of the above PCE proxy model is used to select the basis function terms that have a greater impact on the output variables, and the coefficients of the remaining expansion terms that have a smaller impact on the output variables are set to 0, and the sPCE proxy model is obtained. The sPCE proxy model retains the main influence relationship between the input variables and the output variables, and reduces the coefficients of the expansion terms to be calculated to improve the analysis efficiency.

[0106] The least angle regression method (LAR) is used as the optimal item set screening strategy for the sPCE proxy model. This method directly modifies the least squares minimization problem in equation (4) by adding a penalty term in the form of λ||a||1. The solution of the expansion term coefficient of the sPCE proxy model can be expressed as:

[0107]

[0108] In formula (5), is the estimated value of the coefficient vector of the expansion term; λ is the penalty factor; the regularization term λ||a||1=∑|a| promotes minimization to favor low-order solutions. Thus, a sPCE proxy model with an optimal expansion term set can be established.

[0109] Based on the above method, a sPCE proxy model with optimal expanded itemset can be established.

[0110] Then, the parameter group sample X={X1,...,X 2t / 3} as input variables, and the displacement samples Y={Y1,...,Y 2t / 3} as the output variable, and substitute it into the above sPCE proxy model with the optimal expanded item set for fitting training, and use the LAR method to solve the non-zero term coefficients to obtain the displacement proxy model of the preset position; then use the parameter group sample X′={X 2t / 3+1 ,...,X t} as the input variable, and substitute it into the displacement proxy model of the preset position to obtain the calculated displacement sample y′={y2′ t / 3+1 ,...,y t′}(There are t / 3 displacement values ​​y in the calculated displacement sample y′ i ′,y i ′ only indicates the corresponding relationship with the input variable sample), the calculated displacement sample y′ is compared with the displacement sample Y′ in the test set at the preset position = {Y 2t / 3+1 ,...,Y t}A comparative analysis is performed to verify the calculation error of the displacement proxy model at the preset position.

[0111] Specifically, this embodiment uses the root mean square relative error RMSPE, the average relative error MAPE, and the correlation coefficient R as indicators to calculate and verify the calculation error of the displacement proxy model of the preset position.

[0112] In this embodiment, the preset error ranges of RMSPE and MAPE are less than 0.05, and the preset error range of R is that its absolute value is greater than 0.9.

[0113] If the calculation error of the displacement proxy model at the preset position does not meet the above preset error range, it indicates that the training sample size for the displacement proxy model at the preset position is insufficient. At this time, the preset sample size t should be increased and resampled to generate a new parameter group sample set, and then the displacement proxy model at the preset position should be re-established according to the above steps until the calculation error of the displacement proxy model at the preset position meets the above preset error range.

[0114] If the calculation error of the displacement proxy model at the preset position meets the above preset error range, it means that the displacement proxy model at the preset position established by sampling the parameter group sample set under the preset sampling number t can accurately simulate the complex nonlinear relationship between material parameters and dam displacement. The displacement proxy model can be used for the next step of screening the parameters to be inverted and the monitoring data.

[0115] In this embodiment, when the preset sample number t=300, the calculation error of the displacement proxy model at the preset position under the single material assumption is shown in Table 1.

[0116] Table 1 Calculation error of displacement proxy model at preset position under single material assumption

[0117]

[0118] It can be seen from Table 1 that the calculation errors of the displacement proxy model in the three directions corresponding to the RMSPE at the preset position are all less than 4.6×10 -2 , the corresponding MAPE calculation errors are less than 3.0×10 -2, the correlation coefficients R between the calculated displacement y′ and the test displacement Y′ in the three directions are all greater than 0.95, indicating that under 300 groups of samples, the sPCE proxy model has high simulation accuracy and adaptability to the Duncan-Zhang EB model, and the displacement proxy model of all nodes on the finite element model can be established based on this number of samples.

[0119] Specifically, since the maximum cross-section of the dam body can more clearly show the influence of different material parameters on the displacement of the dam body, this embodiment only needs to establish the displacement proxy model of all nodes on the maximum cross-section of the dam body.

[0120] Specifically, this embodiment establishes the displacement proxy model of all nodes on the maximum cross-section of the dam body according to the parameter group sample set corresponding to the displacement proxy model of the preset position finally established (the parameter group sample set generated by sampling at the preset sample number t=300) and the establishment method of the displacement proxy model of the preset position mentioned above.

[0121] In this embodiment, in S3, the parameters to be inverted are determined from multiple material parameters according to the displacement proxy models of all nodes, which may be:

[0122] According to the displacement proxy model of all nodes on the largest cross section, the sensitivity index of each node displacement to each material parameter is determined; and all sensitivity indices are statistically analyzed to obtain the average sensitivity index of the dam displacement to each material parameter;

[0123] The parameters to be inverted in each rockfill area are determined from multiple material parameters according to the average sensitivity index.

[0124] Specifically, this embodiment adopts the Sobol sensitivity analysis method to calculate the sensitivity index of the displacement in three directions of each node on the maximum cross section to each material parameter, so as to reflect the influence of each material parameter on the displacement in different directions of the dam body.

[0125] Specifically, the analysis process of the Sobol sensitivity analysis method is as follows:

[0126] Assume the input variable is N-dimensional independent random variables x=[x1,...,x N ], the truncated form of the sPCE proxy model at any point can be written as:

[0127]

[0128] In formula (6), y(x) is the model response; α0,α i ,α i,j ,...,α 1,...,N are expansion term coefficients; Ψ0,Ψ i ,Ψ i,j,...,Ψ 1,...,N is an orthogonal polynomial.

[0129] The response output of the sPCE proxy model is also a random variable, and its variance can be expressed as:

[0130]

[0131] In formula (7), y0=α0Ψ0 is a constant term; y(x) is the model response.

[0132] Due to the orthogonality of the basis functions, the output response variance can be decomposed into the sum of the partial variances of different combinations of input variables:

[0133]

[0134] In formula (8), is the partial variance, a t is the chaos expansion coefficient; The Sobol sensitivity index is therefore:

[0135]

[0136] In formula (9), It is the sensitivity index, which represents the contribution of the elements in the input variable set to the output response variance.

[0137] From formula (9), we can see that the indexes of each order satisfy the following relationship:

[0138]

[0139] In formula (10), is a high-order Sobol sensitivity index, such as S i,j Represents parameter x i 、x j The interaction effect sensitivity index of i is the parameter x i The first-order Sobol sensitivity index of i impact.

[0140] A single input variable x i The effect on the output response can be expressed as the sum of the exponentials of each order of the variable (including x i The first-order and all higher-order sensitivity indices of ), which is called the global sensitivity index:

[0141]

[0142] In formula (11), is the global sensitivity index; is the high-order Sobol sensitivity index, S ～i Does not contain x i The sum of all sensitivity indices.

[0143] After calculating the sensitivity index of the displacement in three directions of each node on the maximum cross section to each material parameter by the Sobol sensitivity analysis method, all sensitivity indices are statistically analyzed, and the average value of the sensitivity index of the displacement of all nodes to each material parameter is calculated to obtain the average sensitivity index of the dam body displacement to each material parameter, as shown in Table 2.

[0144] Table 2 Average sensitivity index of dam displacement to various material parameters under single material assumption

[0145]

[0146] It can be seen from Table 2 that among the material parameters, only n. K b The average sensitivity indexes of the three parameters corresponding to displacements in different directions are all greater than 0.1.

[0147] at the same time, Figure 5 The chart shows the ranking statistics of the sensitivity index of the displacement of the largest cross section to each material parameter under the assumption of a single material. Specifically, this embodiment first counts the ranking of the sensitivity index of the displacement of each node on the largest cross section to the eight material parameters, and then counts the proportion of the first, second, third and fourth rankings of each material parameter in the ranking of all nodes, and then obtains Figure 5 The ranking of sensitivity index is the ranking of sensitivity of material parameters. It can be seen that horizontal displacement along the river has a significant effect on material parameters n and K. b , All of them have strong sensitivity: n ranks first in sensitivity among 65.2% of the nodes, K b The sensitivity ranked in the top two in 60.3% of the nodes. The sensitivity ranks in the top four in 87.6% of the nodes. b 、n、 Both have strong sensitivity: K b It ranks first in sensitivity in 42.5% of the nodes, and ranks in the top two in sensitivity in 54.6% of the nodes. The sensitivity ranks in the top four in 99.4% of the nodes. Similar to the above two displacements, the horizontal transverse displacement also has a significant effect on the material parameters n, K b Shows strong sensitivity: n ranks first in sensitivity among 83.9% of nodes. The sensitivity ranks in the top two in 60.6% of the nodes, K bThe sensitivity ranks in the top four in more than 98% of the nodes. The above analysis shows that the material parameters n. K b is the main influencing parameter of the dam displacement in this embodiment. Therefore, the parameters in different material partitions of the rockfill body should be n. K b As the inversion parameter, the upstream rockfill area 6 n1, K b1 , the main rockfill area 7 n2, K b2 , and the downstream rockfill area 8 n3, K b3 The parameters to be inverted for the parameter inversion analysis of the concrete face rockfill dam are determined.

[0148] After obtaining the parameters to be inverted, the sensitivity distribution of the parameters to be inverted on the finite element model can be determined according to the sensitivity index of the dam displacement to each parameter to be inverted, and then the monitoring data for parameter inversion can be determined according to the sensitivity distribution.

[0149] It is understandable that since the sensitivity index of displacement of all nodes to various material parameters is based on the assumption that the rockfill body is composed of a single material, the sensitivity distribution of material parameters reflected by it also only conforms to this assumption. If the earth-rock dam is a homogeneous dam, since the actual internal structure of the dam body conforms to this assumption, it can be directly based on n. K b The sensitivity distribution of the parameters is analyzed to select the monitoring data for parameter inversion.

[0150] The earth-rock dam in this embodiment is a face rockfill dam, whose rockfill body is composed of multiple rockfill areas with different materials. The volume proportion, shape structure and filling position of different rockfill areas are quite different, which leads to great differences in the sensitivity distribution of material parameters in different rockfill areas. Therefore, it is necessary to select the 9 parameters to be inverted (the 6 upstream rockfill area) according to the actual zoning of the face rockfill dam and the 9 parameters to be inverted (the 6 upstream rockfill area n1, K b1 , the main rockfill area 7 n2, K b2 , and the downstream rockfill area 8 n3, K b3 ), and re-establish the displacement proxy models of all nodes on the maximum cross section according to the aforementioned displacement proxy model establishment method, and then determine the sensitivity distribution of the nine parameters to be inverted based on the re-established displacement proxy model, and then screen out the monitoring data for parameter inversion based on the sensitivity distribution.

[0151] In this embodiment, in S3, the monitoring data used for parameter inversion is determined according to the parameters to be inverted, which may specifically be:

[0152] The displacement proxy models of all nodes on the maximum cross section are updated according to the parameters to be inverted in each rockfill area, and the monitoring data used for parameter inversion are determined according to the updated displacement proxy models of all nodes.

[0153] Furthermore, the displacement proxy model of all nodes on the maximum cross section is updated according to the parameters to be inverted in each rockfill area, which can be specifically:

[0154] Generate a parameter group sample set of all parameters to be inverted according to the materials in each rockfill area;

[0155] According to the parameter group sample set of all parameters to be inverted, the displacement proxy models of all nodes on the maximum cross section are updated using the finite element model.

[0156] Furthermore, the monitoring data for parameter inversion determined according to the displacement proxy model of all nodes after the update may specifically be:

[0157] According to the displacement proxy model of all nodes after the update, the Sobol sensitivity analysis method is used to determine the sensitivity index of each node displacement to each parameter to be inverted.

[0158] Monitoring data for parameter inversion is determined according to the sensitivity index.

[0159] Furthermore, the monitoring data used for parameter inversion determined according to the sensitivity index may specifically be:

[0160] Determine the sensitivity distribution of each parameter to be inverted on the finite element model according to the sensitivity index;

[0161] Sensitive monitoring points on the rockfill are determined according to the sensitivity distribution, and the displacement monitoring data of the sensitive monitoring points are determined as monitoring data for parameter inversion.

[0162] Specifically, in this embodiment, the upstream rockfill area 6 n1, K b1 , the main rockfill area 7 n2, K b2 , and the downstream rockfill area 8 n3, K b3 , a total of 9 material parameters are used as parameters to be inverted, and the remaining material parameters of the three rockfill areas are regarded as constant values, and the design parameter values ​​can be used.

[0163] Assuming that the distribution of the parameter values ​​to be inverted obeys the normal distribution, the parameter group sample set X = {X1, X2, ..., X t′}, where each parameter group sample X iEach contains a sampling value of 9 parameters to be inverted, namely

[0164] The parameter group sample set X = {X1, X2, ..., X t′} is substituted into the finite element model, and the displacement sample set Y = {Y1, Y2, ..., Y t′}, where each displacement sample contains displacements in three directions, namely, Y i ={[Y Xi ,Y Yi ,Y Zi ]}, i=1,2,...,t′.

[0165] The parameter group sample set X = {X1, X2, ..., X t′} and the displacement sample set Y = {Y1, Y2, ..., Y t′ The data are divided into a training set and a test set in a ratio of 2:1, and then the displacement proxy model of the preset position can be re-established according to the aforementioned method for establishing the displacement proxy model.

[0166] Furthermore, the accuracy of the re-established displacement proxy model of the preset position is checked. When the calculation error of the displacement proxy model of the preset position does not meet the preset error range, the preset number of samples is increased and the displacement proxy model of the preset position is updated until its calculation error meets the preset error range.

[0167] In this embodiment, when the preset sample number t′=300, the calculation errors of the displacement proxy model at the preset positions under different material partitions are shown in Table 3.

[0168] Table 3 Calculation error of displacement proxy model at preset positions under different material partitions

[0169]

[0170] It can be seen from Table 3 that the calculation error of the displacement proxy model at the preset position under different material partitions has no significant change compared with the calculation error of the displacement proxy model at the preset position under the single material assumption in Table 1, and is still basically at the same order of magnitude; the calculation errors of the displacement proxy model at the preset position in the three directions corresponding to the RMSPE are all less than 9.7×10 -3 , the corresponding MAPE calculation errors are less than 7.7×10 -3 The correlation coefficients R between the calculated displacement y′ and the test displacement Y′ in the three directions are all greater than 0.99, showing a strong correlation. This shows that the sPCE proxy model still has a high simulation accuracy when considering different material partitions, and can be used for sensitivity analysis of the parameters to be inverted under different partition materials.

[0171] Furthermore, according to the parameter group sample set of the parameters to be inverted corresponding to the displacement proxy model of the preset position finally established (the parameter group sample set generated by sampling at the preset sample number t′=300), the displacement proxy model of all nodes on the maximum cross section is established using the finite element model.

[0172] Furthermore, based on the displacement proxy models of all nodes, the aforementioned Sobol sensitivity analysis method is used to calculate the sensitivity index of the displacements in three directions of all nodes on the maximum cross section to each parameter to be inverted.

[0173] Furthermore, based on the sensitivity index of the displacement in three directions of all nodes on the maximum cross section to each parameter to be inverted, the Kriging interpolation method is used to generate a sensitivity index distribution contour map of each parameter to be inverted under displacement in three directions. According to the high sensitivity area of ​​each parameter to be inverted in the sensitivity index distribution contour map, the monitoring points in the panel rockfill dam located in the high sensitivity area or close to the area are selected. Since the parameters to be inverted have a greater influence on the displacement at the above-mentioned monitoring points, the displacement monitoring data of the above-mentioned monitoring points are determined as the monitoring data for parameter inversion. Compared with using the monitoring data of other monitoring points, the inversion analysis of the parameters to be inverted using the above-mentioned monitoring data can effectively improve the accuracy of the parameter inversion analysis of the panel rockfill dam.

[0174] Specifically, according to the sensitivity index distribution contour map of each parameter to be inverted in the three directions, the high sensitivity area of ​​each parameter to be inverted in the three directions can be obtained. According to the actual situation of the project, the monitoring data of the observation points that are close to the three high sensitivity areas and have good data quality can be comprehensively selected as the monitoring data for parameter inversion.

[0175] Figure 6 Shows the parameters K to be inverted for different material partitions on the largest cross section b The vertical displacement sensitivity index distribution contour map; Figure 7 The vertical displacement sensitivity index distribution contour map of the parameter to be inverted n for different material partitions on the maximum cross section is shown. It can be seen that under the vertical displacement of the dam body, for the parameter to be inverted K b , K b1 The high-sensitivity area of ​​K is located near the dam foot in the upstream dam body; b2 The high-sensitivity area of ​​K is located in the middle of the dam body, close to the dam top and dam foundation; b3 The high sensitivity areas of K are mostly concentrated in the downstream dam body near the dam foot. b1Similar but larger in range, mainly located near the dam foot of the upstream dam body; the vertical displacement of most areas in the middle of the dam body shows strong sensitivity to n2; the high sensitivity area of ​​n3 is located near the dam foot of the downstream dam body. The above analysis shows that the high sensitivity areas of the parameters to be inverted in different material partitions are basically located in the material partitions to which they belong, that is, the parameters to be inverted in different material partitions only have a significant impact on the displacement of the dam body in the partitions to which they belong and the adjacent surrounding areas. Therefore, for example, when performing inversion analysis on the parameters to be inverted in the main rockfill area 7, refer to Figure 4 , vertical displacement monitoring data of monitoring points located in and near the main rockfill area 7 should be selected for parameter inversion. Figure 6 (b) When K b2 When performing inversion analysis, it is advisable to select the distance K b2 The vertical displacement monitoring data of monitoring points VS1-02, VS1-03, and VS3-02 close to the high-sensitivity area are analyzed; see Figure 7 (b) When inversion analysis is performed on n2, the vertical displacement monitoring data of monitoring points VS2-02, VS2-03, VS3-01, and VS3-02, which are close to the high-sensitivity area of ​​n2, should be selected for analysis; b2 When performing inversion analysis, it is advisable to select the distance K b2 The vertical displacement monitoring data of the monitoring point VS3-02, which is close to the high sensitivity area of ​​n and the high sensitivity area of ​​n2, are analyzed.

[0176] Figure 8 Shows the parameters K to be inverted for different material partitions on the largest cross section b The distribution contour map of the sensitivity index of horizontal downstream displacement is shown in Figure 2. Fig. 9 The distribution contour map of the sensitivity index of horizontal downstream displacement of the parameter n to be inverted for different material partitions on the maximum cross section is shown. The screening method of monitoring data under vertical displacement is the same as that under vertical displacement. For example, see Figure 8 (b) When K b2 When performing inversion analysis, it is advisable to select the distance K b2 The horizontal displacement monitoring data of monitoring points VS1-02, VS1-03, VS3-01, and VS2-02 close to the high-sensitivity area are analyzed; see Fig. 9 (b) When inversion analysis is performed on n2, the vertical displacement monitoring data of monitoring points VS1-02, VS1-04, VS1-05, VS2-03, and VS3-02, which are close to the high-sensitivity area of ​​n2, should be selected for analysis; b2 When performing inversion analysis, it is advisable to select a distance between K b2The horizontal displacement monitoring data of the monitoring point VS1-02 in the high sensitivity area and which is relatively sensitive to n2 (sensitivity index is 0.3-0.5) are analyzed. The screening method of the monitoring data under horizontal transverse river displacement is similar and will not be repeated here.

[0177] In summary, the specific flow chart of the data screening method of this embodiment can be obtained as follows: Figure 2 shown.

[0178] The present invention uses a finite element model of an earth-rock dam and a parameter group sample set of multiple material parameters to establish a displacement proxy model of all nodes on the finite element model, and determines the parameters to be inverted according to the displacement proxy model, and then determines the monitoring data used for parameter inversion according to the parameters to be inverted, so that the parameters to be inverted that have a greater impact on the dam body displacement and the monitoring data that are more sensitive to the parameters to be inverted can be screened out, which is beneficial to improving the accuracy and efficiency of the inversion analysis.

[0179] The present invention determines the parameters to be inverted of each partition according to the partitions of different materials inside the earth-rock dam body, and then determines the monitoring data used for parameter inversion according to the parameters to be inverted of each partition, so that the screening process of the parameters to be inverted and the monitoring data can be combined with the structural characteristics of the earth-rock dam, and fully consider the mutual influence of stress and strain between multiple partitions and the influence of the deformation of each partition on the overall displacement of the earth-rock dam, thereby improving the accuracy of data screening.

[0180] The above are only a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application is disclosed as above with preferred embodiments, it is not intended to limit the present application. Any technician familiar with the profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A data screening method for earth-rock dam parameter inversion, characterized in that: The method comprises: S1. Establishing a finite element model of an earth-rock dam, and determining a plurality of material parameters to be screened according to a constitutive model adopted by the finite element model; the finite element model has a plurality of nodes; S2, generating a parameter group sample set of the plurality of material parameters, and using the parameter group sample set and the finite element model to establish a displacement proxy model of each node; the displacement proxy model is used to simulate the mapping relationship between the displacement of the node and the plurality of material parameters; S3. Determine the parameters to be inverted from the multiple material parameters according to the displacement proxy models of all nodes, and determine the monitoring data for parameter inversion according to the parameters to be inverted.

2. The method according to claim 1, characterized in that The earth-rock dam is a rockfill dam, and the rockfill dam includes a rockfill body; The constitutive model adopted in the region corresponding to the rockfill body on the finite element model is a nonlinear elastic constitutive model; and the multiple material parameters are all material parameters of the nonlinear elastic constitutive model.

3. The method according to claim 2, characterized in that The rockfill body includes a plurality of rockfill areas with different materials; the parameter group sample set of the plurality of material parameters generated in S2 is specifically: A parameter group sample set of the plurality of material parameters is generated according to the material of any one of the plurality of rockfill areas and according to a preset number of samples.

4. The method according to claim 3, characterized in that In S2, the displacement proxy model of each node is established by using the parameter group sample set and the finite element model, specifically: Using the parameter group sample set and the finite element model, obtaining a displacement sample set of a preset position on the finite element model; Establishing a displacement proxy model of the preset position according to the parameter group sample set and the displacement sample set of the preset position; A displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position.

5. The method according to claim 4, characterized in that The displacement proxy model of each node on the finite element model is established according to the displacement proxy model of the preset position, specifically: Determining a calculation error of the displacement proxy model at the preset position, increasing a preset number of samples when the calculation error does not satisfy a preset error range, and updating the parameter group sample set according to the increased preset number of samples; Re-establishing the displacement proxy model of the preset position according to the updated parameter group sample set until the calculation error of the displacement proxy model of the preset position satisfies the preset error range; According to the parameter group sample set corresponding to the displacement proxy model of the preset position finally established, and using the finite element model, a displacement proxy model of each node on the finite element model is established.

6. The method according to claim 3, characterized in that In S3, the parameters to be inverted are determined from the multiple material parameters according to the displacement proxy models of all nodes, specifically: According to the displacement proxy models of all nodes, the sensitivity index of the displacement of each node to each material parameter is determined; The parameters to be inverted in each rockfill area are determined from the multiple material parameters according to the sensitivity index.

7. The method according to claim 6, characterized in that In S3, the monitoring data used for parameter inversion is determined according to the parameters to be inverted, specifically: The displacement proxy models of all nodes are updated according to the parameters to be inverted in each rockfill area, and the monitoring data for parameter inversion are determined according to the updated displacement proxy models of all nodes.

8. The method according to claim 7, characterized in that The displacement proxy model of all nodes is updated according to the parameters to be inverted in each rockfill area, specifically: Generate a parameter group sample set of all parameters to be inverted according to the materials in each rockfill area; According to the parameter group sample set of all parameters to be inverted, the displacement proxy models of all nodes are updated by using the finite element model.

9. The method according to claim 7, characterized in that: The monitoring data for parameter inversion is determined based on the displacement proxy models of all nodes after the update, specifically: According to the displacement proxy models of all nodes after the update, the sensitivity index of the displacement of each node to each parameter to be inverted is determined; Monitoring data for parameter inversion is determined according to the sensitivity index.

10. The method according to claim 9, characterized in that The monitoring data for parameter inversion is determined according to the sensitivity index, specifically: Determining the sensitivity distribution of each parameter to be inverted on the finite element model according to the sensitivity index; Sensitive monitoring points on the rockfill body are determined according to the sensitivity distribution, and displacement monitoring data of the sensitive monitoring points are determined as monitoring data for parameter inversion.

Images