A three-dimensional geological variability characterization method based on multi-source data fusion

Abstract

This invention discloses a three-dimensional geological variability characterization method based on multi-source data fusion, relating to the field of geological exploration technology. It involves using multiple low-fidelity Gaussian process regression models to predict various geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point. These geophysical prediction data are then combined with geological parameter measurement data to construct a first linear regression model, obtaining preliminary geological parameter prediction data. A residual model is then constructed based on the high-fidelity measuring point coordinates and the difference between the geological parameter measurement data and the preliminary geological parameter prediction data. The high-fidelity measuring point coordinates are input into the residual model to obtain residual values. A second linear regression model is then constructed based on the sum of the preliminary geological parameter prediction data and the residual values, along with the geological parameter measurement data. The geological variability characterization results of the target site are calculated using these multiple constructed models. This method can improve the accuracy of three-dimensional geological variability characterization.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, and in particular to a three-dimensional geological variability characterization method based on multi-source data fusion. Background Technology

[0002] Currently, in foundation engineering, accurately characterizing the spatial variability of three-dimensional (3D) underground soil is a prerequisite for ensuring structural safety and economy. Therefore, characterizing three-dimensional geological variability is extremely necessary.

[0003] In existing technologies, high-precision in-situ tests such as Standard Penetration Test (SPT) or Cone Penetration Test (CPT) are commonly used to model geological variability parameters. However, these high-precision in-situ tests are costly, difficult to construct, and difficult to obtain sufficient spatial sampling density. Directly inverting three-dimensional profiles based on sparse high-fidelity data will produce significant statistical uncertainty, resulting in poor accuracy of three-dimensional geological variability characterization. Summary of the Invention

[0004] Therefore, it is necessary to provide a three-dimensional geological variability characterization method based on multi-source data fusion to address the aforementioned technical problems. This method can improve the accuracy of three-dimensional geological variability characterization.

[0005] The present invention adopts the following technical solution:

[0006] This invention provides a three-dimensional geological variability characterization method based on multi-source data fusion, comprising:

[0007] Acquire high-fidelity and low-fidelity data for the target site; the high-fidelity data includes the coordinates of multiple high-fidelity measuring points and the corresponding geological parameter measurement data; the low-fidelity data includes the coordinates of multiple low-fidelity measuring points and various geophysical measurement data corresponding to each low-fidelity measuring point coordinate; the geological parameter measurement data is used to characterize the variability of the corresponding measuring points;

[0008] By modeling with low-fidelity data, we can obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data.

[0009] Multiple low-fidelity Gaussian process regression models are used to predict various geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point. The multiple geophysical prediction data are used as independent variables and geological parameter measurement data are used as dependent variables to construct the first linear regression model.

[0010] Substitute multiple geophysical prediction data into the first linear regression model to obtain the preliminary geological parameter prediction data corresponding to the coordinates of each high-fidelity measuring point. Then, based on the coordinates of multiple high-fidelity measuring points and the difference between the corresponding geological parameter measurement data and the preliminary geological parameter prediction data, a residual model is constructed.

[0011] The residual value corresponding to the coordinates of each high-fidelity measuring point is predicted by the residual model, and the sum of the preliminary geological parameter prediction data and the residual value is used as the independent variable and the geological parameter measurement data is used as the dependent variable to construct a second linear regression model.

[0012] The geological variability of the target site's coordinates, excluding those of the high-fidelity measuring points, was calculated using a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model.

[0013] Optionally, modeling is performed using low-fidelity data to obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data, including:

[0014] For any type of geophysical measurement data, using low-fidelity measurement point coordinates as input and geophysical measurement data as output, Gaussian process regression modeling is performed using multiple low-fidelity measurement point coordinates and corresponding geophysical measurement data to obtain the low-fidelity Gaussian process regression model corresponding to the geophysical measurement data.

[0015] Optionally, the first linear regression model is a multiple linear regression equation; using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables, the first linear regression model is constructed, including:

[0016] A multiple linear regression equation was constructed by using all types of geophysical prediction data as independent variables and geological parameter measurement data as dependent variables.

[0017] Optionally, the residual model is a Gaussian process regression model; using a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model, the geological variability characteristics of the coordinates of the remaining measuring points in the target site, excluding the high-fidelity measuring point coordinates, are calculated, including:

[0018] For any remaining measurement point coordinates at the target site, input the remaining measurement point coordinates into the low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data to obtain multiple geophysical prediction data and the standard deviation of each type of geophysical prediction data.

[0019] Substitute multiple geophysical prediction data into the first linear regression model to obtain preliminary geological parameter prediction values ​​for the coordinates of the remaining measurement points. Then, based on the standard deviation of each geophysical prediction data and the regression coefficient in the first linear regression model, determine the standard deviation of the preliminary geological parameter prediction values.

[0020] Substitute the coordinates of the remaining measuring points into the residual model to obtain the predicted residual values ​​and the standard deviation of the predicted residual values ​​for the remaining measuring point coordinates.

[0021] Substitute the preliminary geological parameter predictions and residual predictions of the remaining measuring point coordinates into the second linear regression model to obtain the target geological parameter prediction data of the remaining measuring point coordinates. Then, based on the regression coefficients, the standard deviation of the preliminary geological parameter predictions, and the standard deviation of the residual predictions in the second linear regression model, determine the standard deviation of the target geological parameter prediction data.

[0022] Based on the predicted data of the target geological parameters and the standard deviation of the predicted data of the target geological parameters, the confidence interval of the predicted data of the target geological parameters under the coordinates of the remaining measuring points is obtained, which is the result of geological variability characterization.

[0023] Optionally, the formula for calculating the standard deviation of the preliminary geological parameter predictions is:

[0024] ;

[0025] in, This represents the standard deviation of the preliminary geological parameter predictions. This indicates the number of types of geophysical measurement data. Indicates the first The regression coefficients corresponding to the geophysical measurement data Indicates the first The standard deviation of geophysical prediction data.

[0026] Optionally, the formula for calculating the standard deviation of the target geological parameter prediction data is:

[0027] ;

[0028] in, This represents the standard deviation of the predicted data for the target geological parameters. This represents the regression coefficients in the second linear regression model. This represents the standard deviation of the preliminary geological parameter predictions. This represents the standard deviation of the predicted residual values.

[0029] Optionally, the method further includes:

[0030] Before training the model using high-fidelity and low-fidelity data, the low-fidelity and high-fidelity data are normalized separately.

[0031] This invention provides a three-dimensional geological variability characterization device based on multi-source data fusion, comprising:

[0032] The acquisition module is used to acquire high-fidelity and low-fidelity data of the target site. The high-fidelity data includes the coordinates of multiple high-fidelity measuring points and the corresponding geological parameter measurement data. The low-fidelity data includes the coordinates of multiple low-fidelity measuring points and various geophysical measurement data corresponding to each low-fidelity measuring point coordinate. The geological parameter measurement data is used to characterize the variability of the corresponding measuring points.

[0033] The first building module is used to model using low-fidelity data to obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data.

[0034] The second construction module is used to predict multiple geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point through multiple low-fidelity Gaussian process regression models, and to construct the first linear regression model by using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables.

[0035] The third construction module is used to substitute multiple geophysical prediction data into the first linear regression model to obtain the preliminary geological parameter prediction data corresponding to the coordinates of each high-fidelity measuring point. It also builds a residual model based on the coordinates of multiple high-fidelity measuring points and the difference between the corresponding geological parameter measurement data and the preliminary geological parameter prediction data.

[0036] The fourth module is used to predict the residual value corresponding to the coordinates of each high-fidelity measuring point through the residual model, and to construct the second linear regression model by using the sum of the preliminary geological parameter prediction data and the residual value as the independent variable and the geological parameter measurement data as the dependent variable.

[0037] The calculation module is used to calculate the geological variability characterization results of the coordinates of the remaining measuring points in the target site, excluding the coordinates of the high-fidelity measuring points, through a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model.

[0038] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for characterizing three-dimensional geological variability based on multi-source data fusion.

[0039] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned three-dimensional geological variability characterization method based on multi-source data fusion.

[0040] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects:

[0041] In this invention, geophysical data predicted by multiple low-fidelity Gaussian process regression models are mapped to high-fidelity geological parameters, thereby converting multiple indirect indicators into direct geological parameters. Then, a residual model is used to capture the prediction bias of the first linear regression model, and a second linear regression model is used to refit the sum of the preliminary geological parameter prediction data and residual values ​​predicted by the first linear regression model with the actual geological parameter measurement data to further correct the geological parameter prediction error. This invention, through the collaborative approach of multiple models, gradually corrects the prediction error, effectively overcoming the limitations of a single model in the characterization of geological variability, thereby improving the accuracy of geological variability characterization. Attached Figure Description

[0042] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0043] Figure 1 A schematic diagram of the process for a three-dimensional geological variability characterization method based on multi-source data fusion provided by the present invention;

[0044] Figure 2 A schematic diagram of another three-dimensional geological variability characterization method based on multi-source data fusion provided by the present invention;

[0045] Figure 3 The present invention provides a three-dimensional domain random field model, wherein (a) figure represents high-fidelity SPT global simulation data, (b) figure represents low-fidelity MASW global simulation data, (c) figure represents low-fidelity ERT global simulation data, and (d) figure represents low-fidelity SPT global simulation data.

[0046] Figure 4 The present invention provides a schematic diagram of extracted measurement data, wherein (a) represents high-fidelity SPT measurement data, (b) represents low-fidelity MASW measurement data, (c) represents low-fidelity ERT measurement data, and (d) represents low-fidelity SPT measurement data.

[0047] Figure 5 The present invention provides a schematic diagram of the distribution of global predicted values ​​and standard deviations, wherein (a) represents high-fidelity SPT global predicted data, and (b) represents high-fidelity SPT global standard deviations;

[0048] Figure 6 This is a schematic diagram of a computer device for implementing a three-dimensional geological variability characterization method based on multi-source data fusion, as provided by the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0050] In existing technologies, low-fidelity data can also be collected using non-invasive instruments such as Multichannel Analysis of Surface Waves (MASW) or Electrical Resistivity Tomography (ERT) in the characterization of geological variability. This includes geophysical measurements or historical in-situ test data from adjacent sites. These methods offer high sampling density and wide coverage.

[0051] Alternatively, a multi-fidelity fusion approach combining low-fidelity and high-fidelity data can be employed. Existing multi-fidelity fusion methods are mostly based on co-kriging, Bayesian compressed sensing, and other similar approaches. While these methods have achieved good results within their existing scopes of application, most require pre-assuming the correlation signs between the sources and are limited to single low-fidelity inputs. Furthermore, when the data sources are heterogeneous and have significantly different dimensions, overfitting or underfitting problems are prone to occur, making it difficult to balance global trends with local biases. Consequently, these methods are difficult to apply rapidly in practical engineering applications.

[0052] Based on this, the present invention provides a three-dimensional geological variability characterization method based on multi-source data fusion. This method eliminates the assumption of prior correlation signs between data sources, effectively captures global trends and local deviations through two-stage linear calibration and residual regression, and is more robust when fusing heterogeneous data. Compared with a single regression model that only uses high-precision data, this method achieves a significant improvement in prediction accuracy and uncertainty quantification capability. Furthermore, it can flexibly incorporate any number and type of low-fidelity data sources, and does not require consistent sample sizes from each source, meeting the needs of complex engineering scenarios.

[0053] The technical solutions provided by the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] Figure 1 This is a schematic diagram of a three-dimensional geological variability characterization method based on multi-source data fusion according to the present invention, which specifically includes the following steps:

[0055] S101, acquire high-fidelity and low-fidelity data of the target site; the high-fidelity data includes the coordinates of multiple high-fidelity measuring points and the corresponding geological parameter measurement data; the low-fidelity data includes the coordinates of multiple low-fidelity measuring points and various geophysical measurement data corresponding to each low-fidelity measuring point coordinate; the geological parameter measurement data is used to characterize the variability of the corresponding measuring points.

[0056] Acquiring high-fidelity and low-fidelity data of the target site includes: collecting high-fidelity data of the target site through high-precision in-situ testing, and collecting low-fidelity data of the target site through non-invasive instruments.

[0057] Among them, the geological parameter measurement data can be obtained by testing the coordinate positions of multiple high-fidelity test points of the target site through SPT or CPT, including parameters such as standard penetration blow count, cone tip resistance or sidewall friction.

[0058] Geophysical measurement data includes, but is not limited to, geophysical measurement data of the target site or historical in-situ test data of adjacent sites.

[0059] Geophysical measurement data for the target site can be obtained by scanning the coordinates of low-fidelity measurement points using MASW or ERT. Geophysical measurement data includes, but is not limited to, MASW shear wave velocity data and ERT resistivity data.

[0060] It should be noted that the specific locations and quantities of high-fidelity and low-fidelity measuring point coordinates can be set according to actual needs, and this embodiment does not limit them here.

[0061] S102 uses low-fidelity data for modeling to obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data.

[0062] The geophysical measurement data includes MASW shear wave velocity data and ERT resistivity data.

[0063] Modeling is performed using low-fidelity data to obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data. This includes: for any type of geophysical measurement data, taking low-fidelity measurement point coordinates as input and geophysical measurement data as output, performing Gaussian process regression modeling using multiple low-fidelity measurement point coordinates and corresponding geophysical measurement data to obtain the low-fidelity Gaussian process regression model corresponding to the geophysical measurement data.

[0064] In one embodiment, the low-fidelity data and high-fidelity data are normalized separately before model training using high-fidelity data and low-fidelity data. It should be noted that the models or regression equations constructed in this invention are all trained using normalized data.

[0065] Specifically, each set of low-fidelity data is... Normalize (mean 0, standard deviation 1), where, Indicates the first The coordinates of low-fidelity measurement points corresponding to various geophysical measurement data. Indicates the first Geophysical measurement data.

[0066] Train a Gaussian Process Regression (GPR) model using normalized low-fidelity data. This yields a low-fidelity Gaussian process regression model for each type of geophysical measurement data, where... Indicates the coordinates of low-fidelity measurement points. Indicates the first A low-fidelity Gaussian process regression model corresponding to geophysical measurement data is used to predict the coordinates of low-fidelity measurement points. The corresponding geophysical data, " indicates that it follows a certain probability distribution, This represents a zero-mean Gaussian process. Let be the autocorrelation function, representing the coordinates of the low-fidelity measurement point. Coordinates of low-fidelity measurement points The correlation between them can be determined by choosing autocorrelation functions such as exponential autocorrelation function (SNX) or radial basis function (RBF).

[0067] It should be noted that this application may also use other machine learning models to train a low-fidelity Gaussian process regression model, and is not limited to a Gaussian process regression model.

[0068] S103 uses multiple low-fidelity Gaussian process regression models to predict various geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point, and constructs the first linear regression model by using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables.

[0069] Specifically, for any high-fidelity measurement point coordinates, the high-fidelity measurement point coordinates are input into the low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data, to obtain the geophysical prediction data at the high-fidelity measurement point coordinates output by each low-fidelity Gaussian process regression model and the standard deviation of each geophysical prediction data.

[0070] In one embodiment, the first linear regression model is a multiple linear regression equation; the first linear regression model is constructed by using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables, including: using all types of geophysical prediction data as independent variables and geological parameter measurement data as dependent variables to construct a multiple linear regression equation.

[0071] The mathematical form of the multiple linear regression equation is:

[0072] (1);

[0073] in, This represents geological parameter measurement data. This indicates the number of types of geophysical measurement data. Indicates the first The regression coefficients corresponding to the geophysical measurement data Indicates the first Geophysical prediction data, This represents the intercept term.

[0074] Specifically, all types of geophysical prediction data are used as independent variables and geological parameter measurement data are used as dependent variables. The least squares method is used to fit the data to determine the regression coefficients and intercept terms corresponding to each type of geophysical measurement data.

[0075] S104. Substitute multiple geophysical prediction data into the first linear regression model to obtain the preliminary geological parameter prediction data corresponding to the coordinates of each high-fidelity measuring point. Then, based on the coordinates of multiple high-fidelity measuring points and the difference between the corresponding geological parameter measurement data and the preliminary geological parameter prediction data, a residual model is constructed.

[0076] For any given high-fidelity measurement point coordinates, all types of geophysical prediction data... Substituting the values ​​into the multiple linear regression equation, we can calculate the preliminary geological parameter prediction data corresponding to the high-fidelity measurement point coordinates. .

[0077] Calculate the difference between the geological parameter measurement data and the preliminary geological parameter prediction data at each high-fidelity measuring point coordinate. .

[0078] After normalizing the coordinates of each high-fidelity measurement point and its corresponding difference combination, a Gaussian process regression model is performed using the normalized high-fidelity measurement point coordinates and their corresponding difference combinations to obtain the residual model. .

[0079] S105, the residual value corresponding to the coordinates of each high-fidelity measuring point is predicted by the residual model, and the sum of the preliminary geological parameter prediction data and the residual value is used as the independent variable and the geological parameter measurement data is used as the dependent variable to construct a second linear regression model.

[0080] For any high-fidelity measurement point coordinates High-fidelity measurement point coordinates Input residual model The residual value is obtained. and the standard deviation of the residuals .

[0081] The second linear regression model is a linear regression equation in one variable, which is:

[0082] (2);

[0083] in, This represents geological parameter measurement data. Represents the regression coefficient. This represents preliminary geological parameter prediction data. Represents the residual value. This represents the intercept term.

[0084] Specifically, the sum of the preliminary geological parameter prediction data and residual values ​​is used as the independent variable, and the geological parameter measurement data is used as the dependent variable. The regression coefficients are determined by fitting multiple high-fidelity measurement points using the least squares method, taking the sum of the preliminary geological parameter prediction data and residual values ​​corresponding to each point, as well as the geological parameter measurement data. and intercept term .

[0085] S106 uses a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model to calculate the geological variability of the coordinates of the remaining measuring points in the target site, excluding the high-fidelity measuring point coordinates.

[0086] In one embodiment, the residual model is a Gaussian process regression model; the geological variability characterization results of the coordinates of the remaining measuring points in the target site, excluding the high-fidelity measuring point coordinates, are calculated using a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model, including the following steps:

[0087] S201: For any remaining measurement point coordinates at the target site, input the remaining measurement point coordinates into the low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data to obtain multiple geophysical prediction data and the standard deviation of each type of geophysical prediction data.

[0088] For any other measuring point coordinates at the target site ,Will The data are input into the low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data constructed in step 102. , get in Geophysical prediction data at the location and the standard deviation of each geophysical prediction data ,in, .

[0089] S202, substitute multiple geophysical prediction data into the first linear regression model to obtain preliminary geological parameter prediction values ​​for the coordinates of the remaining measuring points, and determine the standard deviation of the preliminary geological parameter prediction values ​​based on the standard deviation of each geophysical prediction data and the regression coefficient in the first linear regression model.

[0090] All of Substituting the values ​​into the first linear regression model constructed in step S103, we obtain the coordinates of the remaining measurement points. Preliminary geological parameter predictions at the location .

[0091] Optionally, the formula for calculating the standard deviation of the preliminary geological parameter predictions is:

[0092] (3);

[0093] in, This represents the standard deviation of the preliminary geological parameter predictions. This indicates the number of types of geophysical measurement data. Indicates the first The regression coefficients corresponding to the geophysical measurement data Indicates the first The standard deviation of geophysical prediction data.

[0094] S203, substitute the coordinates of the remaining measuring points into the residual model to obtain the predicted residual values ​​and the standard deviation of the predicted residual values ​​for the remaining measuring point coordinates.

[0095] Will Substitute the residual model constructed in step S104 ,get residual value at and the standard deviation of the residuals .

[0096] S204. Substitute the preliminary geological parameter predictions and residual predictions of the remaining measuring point coordinates into the second linear regression model to obtain the target geological parameter prediction data of the remaining measuring point coordinates. Then, based on the regression coefficients in the second linear regression model, the standard deviation of the preliminary geological parameter predictions, and the standard deviation of the residual predictions, determine the standard deviation of the target geological parameter prediction data.

[0097] Will Preliminary geological parameter predictions at the location and residual prediction values Add them together and substitute them into the second linear regression model to calculate the result. Predicted geological parameters of the target location .

[0098] The formula for calculating the standard deviation of the target geological parameter prediction data is as follows:

[0099] (4);

[0100] in, This represents the standard deviation of the predicted data for the target geological parameters. This represents the regression coefficients in the second linear regression model. This represents the standard deviation of the preliminary geological parameter predictions. This represents the standard deviation of the predicted residual values.

[0101] S205. Based on the predicted data of the target geological parameters and the standard deviation of the predicted data of the target geological parameters, the confidence interval of the predicted data of the target geological parameters under the coordinates of the remaining measuring points is obtained, which is the geological variability characterization result.

[0102] Based on the predicted data of the target geological parameters Standard deviation of the target geological parameter prediction data The coordinates of the remaining measuring points are obtained. The 95% confidence interval for the predicted geological parameters of the target is: .

[0103] Based on the above steps, the 95% confidence interval of the target geological parameter prediction data for the coordinates of each of the remaining measuring points at the target site can be obtained.

[0104] It should be noted that this application is not limited to a 95% confidence interval, and other confidence intervals are also possible. This embodiment is not limited to this and can be set according to requirements.

[0105] In one specific embodiment, this embodiment provides another flowchart of a three-dimensional geological variability characterization method based on multi-source data fusion, such as... Figure 2 As shown, the specific steps include:

[0106] S301 uses a cross-random field algorithm to generate one type of high-fidelity data and three types of low-fidelity data with cross-correlation.

[0107] Specifically, the three-dimensional domain (100m×100m×10m) was divided into two soil models (upper clay layer and lower sand layer). The cross-correlation random field algorithm was used to generate high-fidelity SPT, low-fidelity MASW, low-fidelity ERT, and low-fidelity SPT data in the two soil layers, respectively. The high-fidelity SPT data were in-situ test data of the target site directly obtained through field standard penetration tests; the low-fidelity MASW data were shear wave velocity profile data based on surface wave velocity inversion; the low-fidelity ERT data were resistivity distribution data obtained based on subsurface electrical parameters inversion; and the low-fidelity SPT data were in-situ test data of the adjacent site directly obtained through field standard penetration tests. The mean and coefficient of variation of different types of data in the soil layers are shown in Table 1.

[0108] Table 1. Mean and coefficient of variation of different types of data in soil layers

[0109]

[0110] The spatial autocovariance matrix for both soil layers was constructed using an RBF-type kernel function during the generation process. The expression is as follows: ,in, and Let be the three-dimensional coordinate vector of any two points in space; , which is the variance scaling parameter, used to characterize the overall variability intensity of the target physical quantity in space; , , These represent the autocorrelation lengths in the x, y, and z directions, respectively. , , .

[0111] Construct a 4×4 cross-covariance matrix C with the main diagonal representing the variance of each random field and the off-diagonal elements representing the variance multiplied by 0.6. Perform Cholesky decomposition on C to obtain the lower triangular matrix L. Then multiply the spatial autocovariance matrix of the two soil layers constructed above with L to obtain the multivariate Gaussian random fields of the four types of data in the two soil layers, where the cross-correlation coefficient is 0.6.

[0112] By concatenating the two random fields of soil layers for each data set, a complete three-dimensional domain random field model can be obtained, such as... Figure 3 As shown, (a) represents high-fidelity SPT global simulation data, (b) represents low-fidelity MASW global simulation data, (c) represents low-fidelity ERT global simulation data, and (d) represents low-fidelity SPT global simulation data.

[0113] For the random field model generated above, multi-fidelity measurement data is extracted in the following manner: from the high-fidelity SPT global simulation data. x -y Four SPT drill holes are arranged on the plane, and the four drill holes are located at... x - y The planar coordinates are (25, 25), (75, 25), (75, 75), and (25, 75), with a depth of 10m per hole. Data is sampled at 1m intervals along the depth direction, resulting in 44 high-fidelity SPT data points. Low-fidelity MSASW global simulation data is also included. x - y upper edge of the plane x =50m y =100m y =50m y At 0m, MASW measurement data was sampled, with sampling every 4m along the line and at 1m intervals along the depth direction, resulting in 1144 low-fidelity MASW data points; low-fidelity ERT global simulation data were then used. x - y upper edge of the plane y =50m x =100m x =50m x ERT measurement data were sampled at 0m, with sampling every 4m along the line and at 1m intervals along the depth direction, resulting in 1144 low-fidelity ERT data points; low-fidelity SPT full-domain simulation data were then used. x - y Twenty-five low-fidelity SPT boreholes were arranged in a 25m × 25m grid on the plane, with samples taken at 1m intervals along the depth direction, resulting in a total of 275 low-fidelity SPT data points. The extracted measurement data are as follows: Figure 4 As shown, (a) represents high-fidelity SPT measurement data, (b) represents low-fidelity MASW measurement data, (c) represents low-fidelity ERT measurement data, and (d) represents low-fidelity SPT measurement data.

[0114] S302, construct each GPR model and linear regression model in sequence.

[0115] Based on the above steps, construct a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model. The kernel function for the low-fidelity Gaussian process regression model and the residual model is of type RBF. The relevant parameters of each model and equation are shown in Table 2, where the superscripts 1, 2, and 3 represent low-fidelity MASW, low-fidelity ERT, and low-fidelity SPT data, respectively.

[0116] Table 2. Parameters of each model and equation after training

[0117]

[0118] S303 predicts the parameters of the entire site where the high-fidelity data is located, and obtains the predicted values ​​and standard deviation distribution of the entire site.

[0119] Following step S106, the SPT parameters for the entire site of the high-fidelity SPT are predicted, and the predicted values ​​and standard deviation distributions for the entire site are obtained as follows: Figure 5 As shown, (a) represents the high-fidelity SPT global prediction data, and (b) represents the high-fidelity SPT global standard deviation. Compared with the true value, the R-squared value of the predicted value is... 2 =0.63, mean square error (MSE)=1.70.

[0120] When applying the three-dimensional geological variability characterization method based on multi-source data fusion provided by this invention, it is not necessary to consider... Figure 1 The steps shown are executed in sequence. The specific execution order of each step can be determined as needed, and this invention does not impose any restrictions on it.

[0121] The above describes a method for characterizing three-dimensional geological variability based on multi-source data fusion, provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding device for characterizing three-dimensional geological variability based on multi-source data fusion, the device comprising:

[0122] The acquisition module is used to acquire high-fidelity and low-fidelity data of the target site. The high-fidelity data includes the coordinates of multiple high-fidelity measuring points and the corresponding geological parameter measurement data. The low-fidelity data includes the coordinates of multiple low-fidelity measuring points and various geophysical measurement data corresponding to each low-fidelity measuring point coordinate. The geological parameter measurement data is used to characterize the variability of the corresponding measuring points.

[0123] The first building module is used to model using low-fidelity data to obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data.

[0124] The second construction module is used to predict multiple geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point through multiple low-fidelity Gaussian process regression models, and to construct the first linear regression model by using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables.

[0125] The third construction module is used to substitute multiple geophysical prediction data into the first linear regression model to obtain the preliminary geological parameter prediction data corresponding to the coordinates of each high-fidelity measuring point. It also builds a residual model based on the coordinates of multiple high-fidelity measuring points and the difference between the corresponding geological parameter measurement data and the preliminary geological parameter prediction data.

[0126] The fourth module is used to predict the residual value corresponding to the coordinates of each high-fidelity measuring point through the residual model, and to construct the second linear regression model by using the sum of the preliminary geological parameter prediction data and the residual value as the independent variable and the geological parameter measurement data as the dependent variable.

[0127] The calculation module is used to calculate the geological variability characterization results of the coordinates of the remaining measuring points in the target site, excluding the coordinates of the high-fidelity measuring points, through a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model.

[0128] Specific limitations regarding the 3D geological variability characterization device based on multi-source data fusion can be found in the limitations of the 3D geological variability characterization method based on multi-source data fusion described above, and will not be repeated here. Each module in the aforementioned 3D geological variability characterization device based on multi-source data fusion can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0129] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 A three-dimensional geological variability characterization method based on multi-source data fusion is provided.

[0130] The present invention also provides Figure 6 The schematic diagram of the computer device shown is as follows: Figure 6 As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 1 A three-dimensional geological variability characterization method based on multi-source data fusion is provided.

[0131] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0132] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A three-dimensional geological variability characterization method based on multi-source data fusion, characterized in that, include: Acquire high-fidelity and low-fidelity data of the target site; the high-fidelity data includes the coordinates of multiple high-fidelity measuring points and the corresponding geological parameter measurement data. Low-fidelity data includes the coordinates of multiple low-fidelity measurement points and various geophysical measurement data corresponding to each low-fidelity measurement point coordinate; Geological parameter measurement data are used to characterize the variability of the corresponding measuring points; By modeling with low-fidelity data, we can obtain a low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data. Multiple low-fidelity Gaussian process regression models are used to predict various geophysical prediction data corresponding to the coordinates of each high-fidelity measuring point. The multiple geophysical prediction data are used as independent variables and geological parameter measurement data are used as dependent variables to construct the first linear regression model. Substitute multiple geophysical prediction data into the first linear regression model to obtain the preliminary geological parameter prediction data corresponding to the coordinates of each high-fidelity measuring point. Then, based on the coordinates of multiple high-fidelity measuring points and the difference between the corresponding geological parameter measurement data and the preliminary geological parameter prediction data, a residual model is constructed. The residual value corresponding to the coordinates of each high-fidelity measuring point is predicted by the residual model, and the sum of the preliminary geological parameter prediction data and the residual value is used as the independent variable and the geological parameter measurement data is used as the dependent variable to construct a second linear regression model. The geological variability of the target site's coordinates, excluding those of the high-fidelity measuring points, was calculated using a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model.

2. The method according to claim 1, characterized in that, By modeling with low-fidelity data, we obtain low-fidelity Gaussian process regression models corresponding to each type of geophysical measurement data, including: For any type of geophysical measurement data, using low-fidelity measurement point coordinates as input and geophysical measurement data as output, Gaussian process regression modeling is performed using multiple low-fidelity measurement point coordinates and corresponding geophysical measurement data to obtain the low-fidelity Gaussian process regression model corresponding to the geophysical measurement data.

3. The method according to claim 1, characterized in that, The first linear regression model is a multiple linear regression equation; Using multiple geophysical prediction data as independent variables and geological parameter measurement data as dependent variables, a first linear regression model is constructed, including: A multiple linear regression equation was constructed by using all types of geophysical prediction data as independent variables and geological parameter measurement data as dependent variables.

4. The method according to claim 1, characterized in that, The residual model is a Gaussian process regression model. Using a low-fidelity Gaussian process regression model, a first linear regression model, a residual model, and a second linear regression model, the geological variability of the coordinates of the remaining measurement points at the target site, excluding the high-fidelity measurement points, is calculated, including: For any remaining measurement point coordinates at the target site, input the remaining measurement point coordinates into the low-fidelity Gaussian process regression model corresponding to each type of geophysical measurement data to obtain multiple geophysical prediction data and the standard deviation of each type of geophysical prediction data. Substitute multiple geophysical prediction data into the first linear regression model to obtain preliminary geological parameter prediction values ​​for the coordinates of the remaining measurement points. Then, based on the standard deviation of each geophysical prediction data and the regression coefficient in the first linear regression model, determine the standard deviation of the preliminary geological parameter prediction values. Substitute the coordinates of the remaining measuring points into the residual model to obtain the predicted residual values ​​and the standard deviation of the predicted residual values ​​for the remaining measuring point coordinates. Substitute the preliminary geological parameter predictions and residual predictions of the remaining measuring point coordinates into the second linear regression model to obtain the target geological parameter prediction data of the remaining measuring point coordinates. Then, based on the regression coefficients, the standard deviation of the preliminary geological parameter predictions, and the standard deviation of the residual predictions in the second linear regression model, determine the standard deviation of the target geological parameter prediction data. Based on the predicted data of the target geological parameters and the standard deviation of the predicted data of the target geological parameters, the confidence interval of the predicted data of the target geological parameters under the coordinates of the remaining measuring points is obtained, which is the result of geological variability characterization.

5. The method according to claim 4, characterized in that, The formula for calculating the standard deviation of the preliminary geological parameter predictions is as follows: ； in, This represents the standard deviation of the preliminary geological parameter predictions. This indicates the number of types of geophysical measurement data. Indicates the first The regression coefficients corresponding to the geophysical measurement data Indicates the first The standard deviation of geophysical prediction data.

6. The method according to claim 4, characterized in that, The formula for calculating the standard deviation of the target geological parameter prediction data is: ； in, This represents the standard deviation of the predicted data for the target geological parameters. This represents the regression coefficients in the second linear regression model. This represents the standard deviation of the preliminary geological parameter predictions. This represents the standard deviation of the predicted residual values.

7. The method according to claim 1, characterized in that, The method further includes: Before training the model using high-fidelity and low-fidelity data, the low-fidelity and high-fidelity data are normalized separately.

Images