Grouting simulation and pre-control decision-making method and system based on cross-scale data integration

Abstract

The present invention discloses a grouting simulation and pre-control decision-making method and system based on cross-scale data combination, wherein the method includes: constructing a knowledge graph based on grouting test data, simulation data and corresponding working condition data of indoor grouting tests, using the nodes and edge information of the knowledge graph as input values ​​of a graph convolutional neural network, using known grouting parameters as output values ​​of the graph convolutional neural network, training the graph convolutional neural network to obtain a trained graph convolutional neural network; constructing a target tunnel geological attribute model based on the geological conditions of the target tunnel; obtaining the engineering construction monitoring data of the target tunnel, and aligning the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel in time and space; inputting the aligned data into the trained graph convolutional neural network to obtain grouting prediction data; and adjusting the grouting plan according to the grouting prediction data until the grouting construction is completed.

Description

Technical Field

[0001] The present invention relates to the field of geotechnical engineering technology, and in particular to a grouting simulation and pre-control decision-making method and system based on cross-scale data combination. Background Art

[0002] Tunnel transportation is a vital component of modern urban infrastructure, widely used in various transportation applications, including railways, highways, and subways. With the acceleration of urbanization, tunnel construction is becoming increasingly large-scale, and the safety and stability of tunnel structures have become a core concern in the engineering field. Especially in extreme geological conditions, tunnel excavation faces complex strata and the risk of sudden geological disasters, requiring more precise and efficient grouting reinforcement technology in tunnel construction. Therefore, optimizing the grouting process to improve the accuracy of slurry diffusion and reinforcement effectiveness has become a key technical challenge in tunnel construction.

[0003] Digital twin technology is an innovative technology that connects physical entities with digital models to monitor and simulate physical processes in real time. In tunnel construction, digital twins can help fully understand the actual operating status of the tunnel, accurately predict the impact of the geological environment on the grouting process, and improve the scientific nature and accuracy of the grouting reinforcement process. Although digital twin technology has broad application potential in tunnel engineering, its actual application in the tunnel grouting process still faces many challenges. In particular, in high-risk areas such as tunnel faces, there are many uncertainties, such as stratum heterogeneity, groundwater influence, and slurry flow paths. These factors make the real-time monitoring and dynamic adjustment of digital twin models a huge challenge. Summary of the Invention

[0004] In order to address the deficiencies of the prior art, the present invention provides a grouting simulation and pre-control decision-making method and system based on cross-scale data integration;

[0005] On the one hand, a grouting simulation and pre-control decision-making method based on cross-scale data is provided, including:

[0006] Based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed. The nodes and edge information of the knowledge graph are used as the input values ​​of the graph convolutional neural network, and the known grouting parameters are used as the output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain the trained graph convolutional neural network.

[0007] Based on the geological conditions of the target tunnel, a geological attribute model of the target tunnel is constructed;

[0008] Acquire the target tunnel's engineering construction monitoring data and perform spatiotemporal alignment with the target tunnel's geological attribute model. Input the aligned data into a trained graph convolutional neural network to obtain grouting prediction data. Adjust the grouting plan based on the grouting prediction data until the grouting construction is completed.

[0009] On the other hand, a grouting simulation and pre-control decision-making system based on cross-scale data integration is provided, including:

[0010] A training module is configured to: construct a knowledge graph based on grouting test data, simulation data, and corresponding working condition data of an indoor grouting test; use the nodes and edge information of the knowledge graph as input values ​​of a graph convolutional neural network; use known grouting parameters as output values ​​of the graph convolutional neural network; train the graph convolutional neural network to obtain a trained graph convolutional neural network;

[0011] A construction module is configured to: construct a geological attribute model of a target tunnel based on the geological conditions of the target tunnel;

[0012] The prediction module is configured to: obtain the engineering construction monitoring data of the target tunnel, and align the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel in time and space; input the aligned data into the trained graph convolutional neural network to obtain grouting prediction data; and adjust the grouting plan according to the grouting prediction data until the grouting construction is completed.

[0013] The above technical solution has the following advantages or beneficial effects:

[0014] Based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed. The nodes and edge information of the knowledge graph are used as the input values ​​of the graph convolutional neural network, and the known grouting parameters are used as the output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain the trained graph convolutional neural network. The present invention realizes the training of the graph convolutional neural network using indoor grouting data, providing a basis for cross-scale grouting process modeling.

[0015] The engineering construction monitoring data of the target tunnel is obtained, and the engineering construction monitoring data of the target tunnel is aligned in time and space with the geological attribute model of the target tunnel; the aligned data is input into the trained graph convolutional neural network to obtain grouting prediction data; the grouting plan is adjusted according to the grouting prediction data until the grouting construction is completed. The present invention realizes the alignment of time and space data and the cross-scale grouting pre-control analysis using the trained graph convolutional neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0017] Figure 1 This is a flow chart of the method of embodiment 1. DETAILED DESCRIPTION

[0018] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0019] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments of the present invention. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products, or apparatuses.

[0020] Example 1

[0021] This embodiment provides a grouting simulation and pre-control decision-making method based on cross-scale data integration;

[0022] like Figure 1 As shown in the figure, the grouting simulation and pre-control decision-making method based on cross-scale data combination includes:

[0023] S101: Based on the grouting test data, simulation data, and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed, the nodes and edge information of the knowledge graph are used as input values ​​of the graph convolutional neural network, and the known grouting parameters are used as output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain a trained graph convolutional neural network;

[0024] S102: constructing a geological attribute model of the target tunnel based on the geological conditions of the target tunnel;

[0025] S103: Acquire the engineering construction monitoring data of the target tunnel, and perform spatiotemporal alignment of the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel; input the aligned data into the trained graph convolutional neural network to obtain grouting prediction data; adjust the grouting plan according to the grouting prediction data until the grouting construction is completed.

[0026] Furthermore, the S101: constructing a knowledge graph based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test includes:

[0027] (1-1): Obtain grouting test data, simulation data and corresponding working condition data of indoor grouting test; standardize all the obtained parameters;

[0028] (1-2): Calculate the covariance matrix between each parameter; calculate the eigenvalues ​​and eigenvectors of the covariance matrix;

[0029] (1-3): Sort all eigenvalues ​​in descending order, select the eigenvectors corresponding to the top k eigenvalues ​​as principal components; based on the principal components, calculate the principal component score matrix;

[0030] (1-4): Define the original parameters and principal components as nodes of the knowledge graph, and calculate the correlation coefficient between the nodes. If the correlation coefficient is greater than the set threshold, it means that there is a connection edge between the nodes; otherwise, it means that there is no connection edge between the nodes. The correlation coefficient between the nodes includes: the correlation coefficient between the original parameter nodes and the correlation coefficient between the principal component nodes;

[0031] (1-5): Calculate the factor loading matrix between the original parameter nodes and the principal component nodes. If the factor loading value of an original parameter node in a principal component node is greater than the set threshold, it means that there is a connection edge between the original parameter node and the principal component node. The weight of the connection edge is the absolute value of the factor loading value.

[0032] (1-6): Supplement the attributes of the original parameter nodes and the main component nodes to obtain the knowledge graph.

[0033] Furthermore, the (1-1): obtaining grouting test data, simulation data and corresponding working condition data of the indoor grouting test includes:

[0034] The grouting test data and simulation data (grouting pressure, grouting flow, grouting speed, slurry ratio) of indoor grouting tests of various media (including but not limited to porous media, fractured media, and pipeline media) and the corresponding working condition design (fractured media: fracture size, filling medium type, dynamic water conditions, etc.; pipeline media: pipeline size, filling medium type, dynamic water conditions, etc.; porous media: dynamic water conditions, water richness, porosity, permeability coefficient, ground stress conditions, etc.) are extracted respectively for data and parameters.

[0035] Furthermore, the above (1-1): all acquired parameters are standardized, including:

[0036] Standardize the grouting parameter variable data:

[0037]

[0038] Among them, Z ij represents the value of the i-th sample on the j-th feature after standardization, X ij represents the original data, μ j represents the mean of the jth feature, σ j represents the standard deviation of the j-th feature.

[0039] Furthermore, the step (1-2): calculating the covariance matrix between the parameters includes:

[0040] Calculate the covariance matrix between the parameters:

[0041]

[0042] Where C is the covariance matrix, Z i is the i-th sample, is the mean vector of the samples.

[0043] Furthermore, the step (1-2): calculating the eigenvalues ​​and eigenvectors of the covariance matrix includes:

[0044] Calculate the eigenvalues ​​and eigenvectors of the covariance matrix and perform eigenvalue decomposition on the covariance matrix:

[0045] Cv j =λ j v j

[0046] Among them, λ j represents the jth eigenvalue, v j represents the jth eigenvector.

[0047] Furthermore, (1-3): all eigenvalues ​​are sorted in descending order, and the eigenvectors corresponding to the top k eigenvalues ​​are selected as principal components, including:

[0048] Arrange all eigenvalues ​​in descending order and adjust the order of eigenvectors according to the order. The larger the eigenvalue after sorting, the more data variance the corresponding principal component explains. The eigenvalues ​​and eigenvectors after sorting are:

[0049] λ1≥λ2≥…≥λ p

[0050] Among them, λ1,λ2,...,λ p are the eigenvalues ​​arranged in descending order, and the corresponding eigenvectors are v1,v2,...,vp Select the first k principal components based on the size of the eigenvalues, and select the eigenvectors that meet the following conditions:

[0051] V k =[v1,v2,...,v k ]

[0052] Among them, V k is a matrix consisting of the first k eigenvectors.

[0053] Furthermore, the above (1-3): based on the principal components, calculating the principal component score matrix includes:

[0054] Calculate principal component scores:

[0055] T=Z·V k

[0056] Where T is the principal component score matrix, Z is the standardized data matrix, and V k is a matrix consisting of the first k eigenvectors.

[0057] Define nodes and edges, define each original parameter physical quantity and principal component as a node; and construct edges between original variables and edges between principal components and original variables.

[0058] Furthermore, the correlation coefficients between the original parameter nodes and the original parameter nodes mentioned in (1-4) include:

[0059] Calculate the correlation coefficient between the original variables:

[0060]

[0061] Among them, ρ ij are two variables X i and X j Pearson correlation coefficient between i ,X j ) is their covariance; σ i and σ j is their standard deviation.

[0062] Furthermore, the (1-4): and the correlation coefficients between the principal component nodes include:

[0063] Calculate the correlation coefficient between principal components:

[0064]

[0065] in, are two principal components PC i With PC j The correlation coefficient of .

[0066] Define a variable correlation threshold, if ρ ij If ρ is greater than the threshold, it is considered that the correlation between the two variables is strong. ij If the correlation between the two variables is less than the threshold, it is considered that the correlation between the two variables is weak. If the correlation between the two variables is strong, the correlation coefficient is used as the weight value of the edge. ij If the value is less than a certain minimum value, the edge is not defined. The definition of the edge between the original variables and the pairwise variables between the principal components is realized.

[0067] Furthermore, (1-5): the factor loading matrix between the original parameter nodes and the principal component nodes is calculated. If the factor loading value of an original parameter node on a principal component node is greater than a set threshold, it means that there is a connecting edge between the original parameter node and the principal component node. The weight of the connecting edge is the absolute value of the factor loading value, including:

[0068] Define the edges between the principal components and the original variables and calculate the factor loading matrix:

[0069] L=[v1,v2,...,v k ]

[0070] Where L is the factor loading matrix, v i is the eigenvector of the i-th principal component.

[0071] If the factor loading value of a certain original variable in the principal component is greater than the set threshold, it means that the original variable has a strong contribution to the principal component, and the weight of the edge between the two is the absolute value of the load.

[0072] Furthermore, the above (1-6): supplementing the attributes of the original parameter nodes and the attributes of the principal component nodes to obtain a knowledge graph, including:

[0073] The basic properties (variable name, physical unit, etc.) and statistical properties (variable mean, standard deviation, grouting starting value, maximum and minimum values, variance, skewness, etc.) of each original variable node are improved, and the variables are derived from background information (laboratory monitoring data, simulation, sensor type, environmental conditions) and physical properties (permeability, porosity, ground stress, etc.) are supplemented.

[0074] The basic properties (principal component name), calculation properties (factor loading value of the principal component, main direction of the principal component), and statistical properties (mean value of the principal component, standard deviation of the principal component) of each principal component node are supplemented.

[0075] It should be understood that the node and edge information in the established graph is used as the input features of the graph convolutional neural network. Through graph convolutional reasoning, the grouting effects under different grouting conditions and formation conditions can be inferred based on existing experimental data and simulation data.

[0076] Furthermore, the nodes and edge information of the knowledge graph are used as input values ​​of the graph convolutional neural network, and the known grouting parameters are used as output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain a trained graph convolutional neural network, wherein the loss function in the training process is:

[0077] L total =α1·L prediction +α2·L regularization +α3·L robustness ;

[0078] Among them, α1, α2, and α3 are hyperparameters used to balance the weights of the losses in each part.

[0079] Among them, L prediction The expression is:

[0080]

[0081] Among them, L prediction Represents the prediction error loss, which is used to measure the prediction value and the true value y i Error, dynamic weight ω i , used to increase attention to high deviation samples, β represents a dynamic adjustment parameter, used to control the impact of outliers, Represents the mean of the data set; the purpose of using prediction error loss is to dynamically adjust the focus on different samples and enhance the prediction ability of high deviation areas.

[0082] Furthermore, the L regularization The formula expression is:

[0083]

[0084] Among them, λ1 represents the size of the control weight to prevent overfitting; λ2 represents the balance of the control weight to encourage uniform distribution. The purpose of using the weight regularization loss function is to avoid overfitting and unbalanced distribution caused by excessive weights.

[0085] Among them, L robustness The specific expression is:

[0086]

[0087] Among them, δ represents the outlier threshold, which controls the degree of penalty for excessive errors, and Lrobustness Represents the robustness loss function, which is used to enhance the robustness of the model to outliers. φ(z) represents the kernel function of the loss function, which is used to process the error between the model prediction value and the true value. z represents the difference (error) between the prediction value and the true value, and N represents the total number of data samples.

[0088] Furthermore, the step S102 of constructing a geological attribute model of the target tunnel based on the geological conditions of the target tunnel includes:

[0089] Obtain spatial information and geological attribute information of the target tunnel geological body area, divide the target tunnel geological body area into several grids, and construct a grid model, wherein the address attribute information includes: elastic modulus, water seepage information, porosity, crack location and distribution;

[0090] For each grid area, the acquired geological attribute data is preprocessed. Based on the preprocessed data, geological attributes are assigned to the grid model to obtain the geological attribute model of the target tunnel.

[0091] Among them, the preprocessing includes: refining the grid size; Where Δx is the size of the grid, L is the length of the tunnel region, and N is the number of nodes in the grid.

[0092] Furthermore, the step S103 of acquiring the engineering construction monitoring data of the target tunnel and performing spatiotemporal alignment of the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel may also include: data standardization processing; the data standardization processing specifically includes:

[0093] Standardize the construction monitoring data:

[0094]

[0095] Among them, Z ij represents the jth eigenvalue of the i-th sample after standardization; X ij represents the original data; μ j represents the mean of the jth feature; σ j represents the standard deviation of the j-th feature.

[0096] Furthermore, the S103: obtaining the engineering construction monitoring data of the target tunnel, and performing spatiotemporal alignment of the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel, wherein the spatiotemporal alignment includes: temporal alignment and spatial alignment; the temporal alignment includes:

[0097] First, for the data at different time steps, for two data points (t1, y1) and (t2, y2), the spline interpolation method is used to interpolate the data to the same time interval. The spline interpolation adopts the following formula:

[0098] S(t)=a(t-t1) 3 +b(t-t1) 2 +c(t-t1)+d

[0099] Where a, b, c, and d are polynomial coefficients, which are determined based on the values ​​of adjacent data points, the derivative, and the continuity of the second-order derivative.

[0100] Furthermore, the spatial alignment includes:

[0101] The data points of experimental or simulation data (including spatial position (x i ,y i )) and the corresponding values ​​(z i )like:

[0102] (x1,y1,z1),(x2,y2,z2),...,(x n ,y n ,z n )

[0103] Calculate the distance d between the target point and the known data point i :

[0104]

[0105] Among them, x0, y0 represent the coordinates of the target point; i ,y i Represents the coordinates of the known point i.

[0106] According to the inverse distance weighted method, the weight of each known data point is calculated, and the weight w i With the target point x and the data point (x i ,y i ) between the distance d(x,x) i ) is inversely proportional to:

[0107]

[0108] Among them, w i represents the weight of the known point i to the target point; p represents the weight exponent, which is usually 2.

[0109] Calculate the estimated value z0 of the target point (x0, y0) by weighted averaging:

[0110]

[0111] Among them, z0 represents the estimated value of the target point (x0, y0); w i represents the weight of the known point i; z i Represents the data value of known point i.

[0112] Based on the estimated value z0 of the difference result, the entire spatial data is aligned.

[0113] The aligned data is fed into the graph convolutional neural network, and the grouting equipment parameters and grouting scheme are adjusted in real time based on the real-time predictions of the dynamic water flow field, grouting pressure field, velocity field, and slurry diffusion range. The adjustment process is repeated until the grouting construction is completed, thus achieving pre-control analysis and decision-making for the entire grouting construction process.

[0114] Example 2

[0115] This embodiment provides a grouting simulation and pre-control decision system based on cross-scale data integration;

[0116] Grouting simulation and pre-control decision-making system based on cross-scale data integration, including:

[0117] A training module is configured to: construct a knowledge graph based on grouting test data, simulation data, and corresponding working condition data of an indoor grouting test; use the nodes and edge information of the knowledge graph as input values ​​of a graph convolutional neural network; use known grouting parameters as output values ​​of the graph convolutional neural network; train the graph convolutional neural network to obtain a trained graph convolutional neural network;

[0118] A construction module is configured to: construct a geological attribute model of a target tunnel based on the geological conditions of the target tunnel;

[0119] The prediction module is configured to: obtain the engineering construction monitoring data of the target tunnel, and align the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel in time and space; input the aligned data into the trained graph convolutional neural network to obtain grouting prediction data; and adjust the grouting plan according to the grouting prediction data until the grouting construction is completed.

[0120] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. Grouting simulation and pre-control decision-making method based on cross-scale data combination, characterized by: include: Based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed. The nodes and edge information of the knowledge graph are used as the input values ​​of the graph convolutional neural network, and the known grouting parameters are used as the output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain the trained graph convolutional neural network. Based on the geological conditions of the target tunnel, a geological attribute model of the target tunnel is constructed; Acquire engineering construction monitoring data of the target tunnel, and perform spatiotemporal alignment of the engineering construction monitoring data of the target tunnel with the geological attribute model of the target tunnel; The aligned data is input into the trained graph convolutional neural network to obtain grouting prediction data; the grouting plan is adjusted according to the grouting prediction data until the grouting construction is completed; Based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed, including: Obtain grouting test data, simulation data and corresponding working condition data of indoor grouting test; standardize all the acquired parameters; Calculate the covariance matrix between various parameters; calculate the eigenvalues ​​and eigenvectors of the covariance matrix; Sort all eigenvalues ​​in descending order, select the eigenvectors corresponding to the top k eigenvalues ​​as principal components; based on the principal components, calculate the principal component score matrix; The original parameters and principal components are defined as nodes of the knowledge graph, and the correlation coefficient between the nodes is calculated. If the correlation coefficient is greater than the set threshold, it means that there is a connection edge between the nodes; otherwise, it means that there is no connection edge between the nodes. The correlation coefficient between the nodes includes the correlation coefficient between the original parameter nodes and the correlation coefficient between the principal component nodes. Calculate the factor loading matrix between the original parameter nodes and the principal component nodes. If the factor loading value of an original parameter node in a principal component node is greater than the set threshold, it means that there is a connection edge between the original parameter node and the principal component node. The weight of the connection edge is the absolute value of the factor loading value. The attributes of the original parameter nodes are supplemented, and the attributes of the principal component nodes are supplemented to obtain a knowledge graph.

2. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: All acquired parameters are standardized, including: Standardize the grouting parameter variable data: in, Represents the standardized The sample in The value of the feature, Represents the original data, Indicates the The mean of the features, Indicates the The standard deviation of each feature; Calculate the covariance matrix between various parameters, including: Calculate the covariance matrix between the parameters: in, is the covariance matrix, It is samples, is the mean vector of the samples; Calculate the eigenvalues ​​and eigenvectors of the covariance matrix, including: Calculate the eigenvalues ​​and eigenvectors of the covariance matrix and perform eigenvalue decomposition on the covariance matrix: in, Indicates the eigenvalues, Indicates the feature vectors; Sort all eigenvalues ​​in descending order and select the top ones. The eigenvectors corresponding to the eigenvalues ​​are taken as principal components, including: Arrange all eigenvalues ​​in descending order and adjust the order of eigenvectors according to the order; the larger the eigenvalue after sorting, the more data variance the corresponding principal component explains. The eigenvalues ​​and eigenvectors after sorting are: in, are the eigenvalues ​​in descending order, and the corresponding eigenvectors are ; Select the front according to the size of the eigenvalue principal components, select the eigenvectors that satisfy the following conditions: in, It is from the front A matrix of eigenvectors.

3. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: Based on the principal components, the principal component score matrix is ​​calculated, including: Calculate principal component scores: in, is the principal component score matrix, is the normalized data matrix, It is from the front A matrix of eigenvectors; Define nodes and edges, define each original parameter physical quantity and principal component as a node; and construct edges between original variables and between principal components and original variables; The correlation coefficients between the original parameter nodes include: Calculate the correlation coefficient between the original variables: in, Are two variables and The Pearson correlation coefficient between is their covariance; and is their standard deviation; And the correlation coefficients between principal component nodes, including: Calculate the correlation coefficient between principal components: in, are two principal components and Correlation coefficient of Define a variable correlation threshold, if If the correlation between the two variables is greater than the threshold, If the correlation between the two variables is less than the threshold, it is considered that the correlation between the two variables is weak. If the correlation between the two variables is strong, the correlation coefficient is used as the weight value of the edge. If it is less than a certain minimum value, this edge is not defined.

4. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: Calculate the factor loading matrix between the original parameter nodes and the principal component nodes. If the factor loading value of an original parameter node on a principal component node is greater than the set threshold, it means that there is a connection edge between the original parameter node and the principal component node. The weight of the connection edge is the absolute value of the factor loading value, including: Define the edges between the principal components and the original variables and calculate the factor loading matrix: in, is the factor loading matrix, It is The eigenvectors of the principal components; Supplement the attributes of the original parameter nodes and the main component nodes to obtain a knowledge graph, including: The basic attributes and statistical attributes of each original variable node are improved, and the variables are derived from background information and physical attributes. The basic attributes, calculation attributes, and statistical attributes of each principal component node are supplemented.

5. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: The nodes and edge information of the knowledge graph are used as input values ​​of the graph convolutional neural network, and the known grouting parameters are used as output values ​​of the graph convolutional neural network. The graph convolutional neural network is trained to obtain a trained graph convolutional neural network, wherein the loss function in the training process is: ； in, represents the prediction error loss function, represents the weight regularization loss function, represents the robustness loss function, is a hyperparameter used to balance the weight of each part of the loss.

6. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: Based on the geological conditions of the target tunnel, a geological attribute model of the target tunnel is constructed, including: Obtain spatial information and geological attribute information of the target tunnel geological body area, divide the target tunnel geological body area into several grids, and construct a grid model, wherein the address attribute information includes: elastic modulus, water seepage information, porosity, crack location and distribution; For each grid area, the acquired geological attribute data is preprocessed. Based on the preprocessed data, geological attributes are assigned to the grid model to obtain the geological attribute model of the target tunnel. Among them, the preprocessing includes: refining the grid size; ;in, is the size of the grid, is the length of the tunnel area, is the number of nodes in the grid.

7. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 1 is characterized in that: Acquiring the engineering construction monitoring data of the target tunnel and performing spatiotemporal alignment between the engineering construction monitoring data of the target tunnel and the geological attribute model of the target tunnel, which also includes: data standardization processing; the data standardization processing specifically includes: Standardize the construction monitoring data: in, Represents the standardized The first sample eigenvalues; Represents the original data; Indicates the The mean of the features; Indicates the The standard deviation of each feature; Acquire engineering construction monitoring data of the target tunnel, and perform spatiotemporal alignment between the engineering construction monitoring data of the target tunnel and the geological attribute model of the target tunnel, wherein the spatiotemporal alignment includes: temporal alignment and spatial alignment; the temporal alignment includes: First, the data at different time steps are interpolated by time. For two data points and , use the spline interpolation method to interpolate the data to the same time interval. The spline interpolation adopts the following formula: in, 、 、 、 are the polynomial coefficients, which are determined by the values ​​of adjacent data points, the derivative, and the continuity of the second-order derivative.

8. The grouting simulation and pre-control decision-making method based on cross-scale data integration according to claim 7 is characterized in that: The spatial alignment includes: Data points of experimental or simulation data such as: in, is the spatial location; is the spatial position The corresponding value; Calculate the distance between the target point and the known data points : in, Indicates the coordinates of the target point; Represents a known point coordinates of According to the inverse distance weighted method, the weight of each known data point is calculated. With the target point and data points The distance between Inversely proportional: in, Represents a known point The weight of the target point; represents the weight index; Calculate the target point by weighted average Estimated value : in, Indicates the target point estimated value of; Represents a known point The weight of Represents a known point The data value of ; the estimated value based on the difference result , to achieve alignment of the entire spatial data.

9. Grouting simulation and pre-control decision-making system based on cross-scale data integration, characterized by: include: A training module is configured to: construct a knowledge graph based on grouting test data, simulation data, and corresponding working condition data of an indoor grouting test; use the nodes and edge information of the knowledge graph as input values ​​of a graph convolutional neural network; use known grouting parameters as output values ​​of the graph convolutional neural network; train the graph convolutional neural network to obtain a trained graph convolutional neural network; A construction module is configured to: construct a geological attribute model of a target tunnel based on the geological conditions of the target tunnel; A prediction module is configured to: obtain engineering construction monitoring data of a target tunnel, and perform spatiotemporal alignment of the engineering construction monitoring data of the target tunnel with a geological attribute model of the target tunnel; The aligned data is input into the trained graph convolutional neural network to obtain grouting prediction data; the grouting plan is adjusted according to the grouting prediction data until the grouting construction is completed; Based on the grouting test data, simulation data and corresponding working condition data of the indoor grouting test, a knowledge graph is constructed, including: Obtain grouting test data, simulation data and corresponding working condition data of indoor grouting test; standardize all the acquired parameters; Calculate the covariance matrix between various parameters; calculate the eigenvalues ​​and eigenvectors of the covariance matrix; Sort all eigenvalues ​​in descending order, select the eigenvectors corresponding to the top k eigenvalues ​​as principal components; based on the principal components, calculate the principal component score matrix; The original parameters and principal components are defined as nodes of the knowledge graph, and the correlation coefficient between the nodes is calculated. If the correlation coefficient is greater than the set threshold, it means that there is a connection edge between the nodes; otherwise, it means that there is no connection edge between the nodes. The correlation coefficient between the nodes includes the correlation coefficient between the original parameter nodes and the correlation coefficient between the principal component nodes. Calculate the factor loading matrix between the original parameter nodes and the principal component nodes. If the factor loading value of an original parameter node in a principal component node is greater than the set threshold, it means that there is a connection edge between the original parameter node and the principal component node. The weight of the connection edge is the absolute value of the factor loading value. The attributes of the original parameter nodes are supplemented, and the attributes of the principal component nodes are supplemented to obtain a knowledge graph.

Images