Geotechnical test result analysis and prediction method based on big data

Abstract

The invention relates to the technical field of civil engineering, in particular to a soil test result analysis and prediction method based on big data, which comprises the following steps: S1, collecting multi-dimensional data in a soil test in real time through a preset sensor array; s2, performing data cleaning and preprocessing on the multi-dimensional data acquired in the S1; s3, calculating distance correlation among different geotechnical parameters to extract characteristic variables; s4, dividing the feature data set into a training set and a test set; s5, constructing a soil engineering performance prediction model; s6, optimizing and adjusting a kernel function and a hyper-parameter of the Gaussian process regression model; and S7, performing real-time prediction by using newly collected soil test data. According to the method, the Gaussian process regression algorithm is combined with the big data technology, so that efficient processing, real-time prediction and dynamic feedback of complex geotechnical test data are realized, the precision and reliability of geotechnical performance prediction are improved, and a scientific basis is provided for construction decision making.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of civil engineering, and particularly relates to a soil test result analysis and prediction method based on big data. BACKGROUND

[0002] Soil test is one of the important means to evaluate the physical and mechanical properties of soil in civil engineering, and is widely used in the determination of key engineering parameters such as foundation bearing capacity, settlement and stability. Through stress, strain and permeability tests on soil, the internal structure and behavior characteristics of soil can be understood, and basic data can be provided for civil engineering design. However, with the increase of large-scale engineering projects, the amount and complexity of soil test data have increased significantly. Traditional data processing methods usually rely on manual analysis or empirical formulas, which cannot effectively process multi-dimensional and large-scale soil test data. In addition, the nonlinear characteristics of soil and the uncertainty of data also bring great challenges to existing analysis methods, resulting in low accuracy and real-time performance of prediction.

[0003] The existing soil test analysis method has the following technical problems: first, the traditional analysis method has limited data processing capacity for large-scale and multi-dimensional data, and it is difficult to effectively extract feature variables under complex soil conditions; second, soil test data often contain a large amount of noise, and existing methods cannot effectively process these noise data, thereby reducing the accuracy of prediction; in addition, the existing model cannot be dynamically optimized, and cannot update and improve the reliability of prediction in real time according to newly collected data. SUMMARY

[0004] Based on the above purpose, the present application provides a soil test result analysis and prediction method based on big data.

[0005] The soil test result analysis and prediction method based on big data comprises the following steps:

[0006] S1: Real-time acquisition of multi-dimensional data in soil test through a pre-set sensor array, wherein the multi-dimensional data includes stress, strain and permeability of soil;

[0007] S2: Data cleaning and preprocessing of the multi-dimensional data collected in S1, wherein the data cleaning includes removing outliers, processing repeated data and filling missing data; the preprocessing includes normalizing the cleaned multi-dimensional data, thereby generating a preprocessed data set;

[0008] S3: Based on the preprocessed data set, the distance correlation between different soil parameters is calculated by using a distance correlation analysis method to extract feature variables and generate a feature data set;

[0009] S4: The feature data set is segmented according to the method of combining time series and spatial features, and is divided into a training set and a test set;

[0010] S5: Based on the training set, a Gaussian process regression algorithm is used for modeling to construct a geotechnical performance prediction model;

[0011] S6: Based on the test set, the geotechnical performance prediction model is verified, the prediction error is calculated, and the kernel function and hyperparameters of the Gaussian process regression model are adjusted through a grid search optimization algorithm to improve the prediction accuracy and stability of the model;

[0012] S7: Based on the optimized geotechnical performance prediction model, real-time prediction is performed using newly collected geotechnical test data to predict the mechanical parameters of the soil, and the prediction results and their confidence intervals are fed back to the construction site.

[0013] Optionally, the S1 specifically includes:

[0014] S11: A predetermined sensor array is arranged in the soil body to be measured, the sensor array including stress sensors, strain sensors, and permeation sensors;

[0015] S12: The stress sensor is a pressure sensor that can record the normal stress and shear stress received by the soil body at different time points at each predetermined collection point;

[0016] S13: The strain sensor is a multi-axis strain gauge that can detect the deformation of the soil body in different directions to capture the tensile, compressive, or shear deformation of the soil body in real time;

[0017] S14: The permeation sensor is a seepage meter that can detect the water flow velocity, pressure change, and moisture content in the soil body.

[0018] Optionally, the S2 specifically includes:

[0019] S21: Outlier rejection, identify and reject outliers in the collected data through a predetermined statistical method, the predetermined statistical method being a three times rule based on standard deviation or a box plot method;

[0020] S22: Repetitive data processing, repetitive detection of multi-dimensional data, comparison using a unique identifier of a hash value, and merging or rejecting of repeatedly appearing identical data points;

[0021] S23: Missing data filling, filling missing data caused by equipment or environmental factors during the collection process based on an interpolation algorithm;

[0022] S24: Normalizing the multi-dimensional data set after cleaning, using maximum and minimum normalization or Z-score standardization to convert data values of different dimensions into the same order of magnitude range, ensuring comparability between different physical quantities and providing consistent input for subsequent analysis;

[0023] S25: Summarizing the multi-dimensional data set after normalization, reorganizing the data structure according to the collection time and spatial dimension, and generating a pre-processing data set.

[0024] Optionally, the S3 specifically includes:

[0025] S31: Based on the pre-processing data set, the multi-dimensional data corresponding to different geotechnical parameters is organized in matrix form to construct a data matrix for calculating distance correlation;

[0026] S32: Calculate the distance matrix. For each pair of geotechnical parameters in the data matrix, use the Euclidean distance to calculate the distance matrix between each pair of parameters;

[0027] S33: Based on the calculated distance matrix, use distance correlation analysis method to calculate the distance correlation coefficient between different geotechnical parameters;

[0028] S34: Feature variable extraction. According to the calculated distance correlation coefficient matrix, set a preset correlation threshold, and select geotechnical parameters with strong correlation as feature variables. Specifically, select variables with correlation coefficients greater than the set threshold as feature variables;

[0029] S35: Finally, construct a feature data set from the extracted feature variables.

[0030] Optionally, the S4 specifically includes:

[0031] S41: Time series segmentation. According to the time dimension in the feature data set, the data is segmented according to the time sequence, and a fixed time interval is set to divide the entire data set into multiple time periods;

[0032] S42: Spatial feature segmentation. Based on the spatial position of the geotechnical test area, the data is segmented according to the spatial features, and by setting a fixed spatial coordinate interval, the data is divided into multiple spatial subsets;

[0033] S43: Dual segmentation combining time series and spatial features. First, based on time series segmentation, the data is divided into multiple time periods, and then according to the spatial features within each time period, multiple data subsets containing both time and spatial features are generated;

[0034] S44: dividing the training set and the test set, and proportionally dividing the training set and the test set from the segmented data subsets, specifically, according to the proportion of 80% as the training set and 20% as the test set.

[0035] Optionally, the S5 specifically includes:

[0036] S51: defining the input and output of the Gaussian process regression model based on the characteristic variables in the training set, the input being the variables in the characteristic data set, and the output being the soil performance parameter to be predicted; the input variables including the stress, strain and permeability of the soil performance parameter, and the output being the soil performance index;

[0037] S52: selecting a radial basis kernel function as the kernel function for the Gaussian process regression model, which reflects the correlation between the input characteristic variables by measuring the Euclidean distance between the input data points, and controls the smoothness and fitting effect of the model;

[0038] S53: calculating the covariance matrix between all input variables in the training set based on the radial basis kernel function selected in S52, the covariance matrix being used to quantify the correlation between each input variable and the target output;

[0039] S54: optimizing the hyperparameters in the Gaussian process regression model, including the length scale of the radial basis kernel function and the noise parameter, by maximizing the log-likelihood function using the training set;

[0040] S55: training the model based on the training data after completing the hyperparameter optimization, and establishing the soil performance prediction model.

[0041] Optionally, the S54 specifically includes:

[0042] S541: defining the log-likelihood function of the input data and the output data in the training set based on the Gaussian process regression model;

[0043] S542: gradient calculation, calculating the gradient of the hyperparameters in order to optimize the log-likelihood function;

[0044] S543: updating the hyperparameters based on the gradient calculation result using the gradient descent method;

[0045] S544: optimizing the length scale and the noise parameter respectively in each iteration;

[0046] S545: setting the convergence condition, specifically, when the change of the log-likelihood function is less than the preset threshold, or when the maximum number of iterations is reached, the iteration is terminated, and the optimal hyperparameters θ are finally obtained * , including the optimal length scale l * and the optimal noise parameter

[0047] Optionally, the S55 specifically comprises:

[0048] S551: determining the final covariance matrix, using the optimal hyperparameter l * and calculating the final covariance matrix K * of the training data;

[0049] S552: calculating the output mean μ(x * ) of the geotechnical performance prediction model based on the training data and the new input data x * ;

[0050] S553: calculating the prediction variance σ 2 (x * ) of the geotechnical performance prediction model for measuring the uncertainty of the prediction;

[0051] S554: after completing the training of the model, the expression of the geotechnical performance prediction model is:

[0052] wherein f(x * ) is the objective function of the geotechnical performance prediction, the prediction result obeys the normal distribution, the mean is μ(x * ), and the variance is σ 2 (x * ).

[0053] Optionally, the S6 specifically comprises:

[0054] S61: verifying the geotechnical performance prediction model based on the test set, inputting the feature variable x test in the test set into the trained Gaussian process regression model to obtain the corresponding prediction value The verification formula is: wherein is the prediction output of the model, and μ(x test ) is the prediction mean calculated according to the training data and the test set data;

[0055] S62: calculating the prediction error for measuring the difference between the prediction value of the model and the actual observation value y test , and specifically using the mean square error as the error measurement standard;

[0056] S63: adjusting the kernel function and the hyperparameter of the Gaussian process regression model using the grid search optimization algorithm, setting the search range of the hyperparameter first, the hyperparameter includes the length scale of the kernel function and the noise parameter, then searching point by point on the preset parameter grid, and selecting the parameter combination that makes the verification error minimum;

[0057] S64: According to the grid search result in S63, select the hyperparameter combination that minimizes the mean square error, and update the geotechnical performance prediction model.

[0058] Optionally, the S7 specifically comprises:

[0059] S71: Collect the latest geotechnical test data in real time through the sensor array, the collected data including the stress, strain and permeability parameters of the soil, and input the latest geotechnical test data as new input data into the optimized geotechnical performance prediction model;

[0060] S72: Real-time prediction of the new input geotechnical test data by using the S6 optimized geotechnical performance prediction model, and calculation of the predicted value of the mechanical parameters of the soil according to the new input data And the corresponding prediction variance σ 2 (x new );

[0061] S73: Based on the predicted mean And the prediction variance σ 2 (x new ), calculate the confidence interval of the soil mechanical parameter prediction result, assuming that the prediction result obeys normal distribution, then the calculation formula of the confidence interval is: Wherein, z is the standard normal distribution critical value corresponding to the confidence level, and σ(x new ) is the square root of the prediction variance;

[0062] S74: Real-time feedback of the prediction result And the corresponding confidence interval CI new to the construction site for reference by the construction personnel.

[0063] The beneficial effects of the present application are:

[0064] The present application, by introducing Gaussian process regression algorithm and combining big data technology, can effectively solve the problems of large amount of geotechnical test data and complex dimension, through feature extraction, data segmentation and real-time prediction of soil stress, strain and permeability parameters, greatly improving the efficiency and prediction accuracy of data processing, through optimization of kernel function and hyperparameters, the model can better adapt to complex geotechnical data and capture its nonlinear characteristics, ensuring the applicability of the model under different geological conditions.

[0065] The present application, through the dynamic updating and feedback function of real-time data acquisition, ensures the timely application of geotechnical test data in the construction site, and the confidence interval of the prediction result provides data support for the construction personnel, which can adjust the construction scheme in time according to the prediction result, improve the safety and construction efficiency of the engineering project, not only improve the reliability of geotechnical performance prediction, but also provide an important basis for scientific decision-making of civil engineering. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Fig. 1 This is a schematic diagram of the geotechnical test result analysis and prediction method according to an embodiment of the present invention;

[0068] Fig. 2 This is a schematic diagram of the method for dividing the training set and the test set according to an embodiment of the present invention. Detailed Implementation

[0069] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0070] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0071] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0072] like Figs. 1-2 As shown, the method for analyzing and predicting geotechnical test results based on big data includes the following steps:

[0073] S1: Real-time acquisition of multi-dimensional data in geotechnical tests through a preset sensor array, including soil stress, strain, and permeability;

[0074] S2: Data cleaning and preprocessing of the multi-dimensional data collected in S1, data cleaning includes removing outliers, handling duplicate data and missing data filling; preprocessing includes normalizing the cleaned multi-dimensional data, and then generating a preprocessed dataset to ensure the comparability and consistency of different dimensional data in subsequent analysis;

[0075] S3: Based on the preprocessed dataset, using the distance correlation analysis (Distance Correlation, dCor) method, the distance correlation between different geotechnical parameters (such as stress, strain, permeability, etc.) is calculated to extract feature variables and generate a feature dataset for subsequent modeling;

[0076] S4: The feature dataset is segmented according to the method of combining time series and spatial features, divided into training set and test set, and the segmented data provides training set and test set suitable for different time and space conditions for Gaussian process regression algorithm;

[0077] S5: Based on the training set, Gaussian process regression algorithm is used for modeling to build a geotechnical performance prediction model. Gaussian process regression algorithm models the nonlinearity and uncertainty in geotechnical performance data, providing predicted values and confidence intervals of soil mechanical parameters (such as foundation bearing capacity, settlement, etc.);

[0078] S6: Based on the test set, the geotechnical performance prediction model is verified, the prediction error is calculated, and the kernel function and hyperparameters of the Gaussian process regression model are adjusted through grid search optimization algorithm to improve the prediction accuracy and stability of the model;

[0079] S7: Based on the optimized geotechnical performance prediction model, real-time prediction is performed using newly collected geotechnical test data to predict soil mechanical parameters, and the prediction results and their confidence intervals are fed back to the construction site.

[0080] S1 specifically includes:

[0081] S11: A predetermined sensor array is arranged in the soil body to be tested, the sensor array includes stress sensors, strain sensors and permeation sensors, the arrangement position of the sensor array is accurately planned according to the depth, regional span and specific geological conditions of the soil body to be tested, to ensure that all types of sensors uniformly cover the entire test area;

[0082] S12: The stress sensor is a high-precision pressure sensor that can record the normal stress and shear stress received by the soil body at different time points at each predetermined collection point;

[0083] S13: The strain sensor is a multi-axis strain gauge that can detect the deformation of the soil body in different directions to capture the tensile, compressive or shear deformation of the soil body in real time;

[0084] S14: The permeation sensor is a seepage meter capable of detecting the water flow velocity, pressure change and moisture content in the soil body; the above steps specify the composition of the sensor array and the functions of various sensors, ensuring accurate collection of soil engineering performance data such as stress, strain and permeability; through the combination of high-precision pressure sensors, multi-axis strain gauges and seepage meters, different mechanical properties of the soil body can be comprehensively monitored, providing reliable data input sources for subsequent data analysis and prediction models.

[0085] S2 specifically includes:

[0086] S21: Outlier rejection, identify and reject outliers in the collected data through a pre-set statistical method, the pre-set statistical method is a three times principle method based on standard deviation or a box plot method, to ensure that each value in the data set is within a reasonable physical range, avoiding data deviation from the actual soil engineering conditions;

[0087] The formula based on the three times principle of standard deviation is: |x i -μ|>3σ, where x i is the data point, μ is the mean of the data set, and σ is the standard deviation of the data set, any x i that exceeds this range is considered an outlier;

[0088] The formula based on the box plot method is: lower limit = Q1-1.5×IQR; upper limit = Q3+1.5×IQR, where IQR = Q3-Q1, any data point that exceeds the upper and lower limits is considered an outlier;

[0089] S22: Repetitive data processing, repetitive detection of multi-dimensional data, comparison using unique identifiers of hash values, merging or rejecting identical data points that appear repeatedly, to ensure that there is no redundant information in the data set, to optimize subsequent processing and storage efficiency; the hash shield comparison formula is:

[0090] Hash value = H(x1,x2,…,x n ), where H represents the hash function, x1,x2,…,x n is the multiple dimensions of the data set, and data points with the same hash value are merged or removed;

[0091] S23: Missing data filling, for missing data caused by equipment or environmental factors during the collection process, fill in the missing data based on interpolation algorithms such as linear interpolation, spline interpolation, etc., to ensure that each data dimension has a complete time or space sequence, improving the integrity of the data; the linear interpolation formula is: where x1,x2 are adjacent known data points, f(x1), f(x2) are the corresponding function values, and f(x) is the missing data point to be interpolated.

[0092] S24: Normalizing the multi-dimensional data set after cleaning, using maximum and minimum normalization or Z-score standardization to convert data values of different dimensions to the same order of magnitude range, ensuring comparability between different physical quantities and providing consistent input for subsequent analysis; the Z-score standardization formula is: where x is the original data, μ is the mean of the data set, σ is the standard deviation of the data set, and z is the standardized data;

[0093] S25: Summarizing the multi-dimensional data set after normalization, reorganizing the data structure according to the collection time and spatial dimension, and generating a pre-processed data set with integrity and consistency; the pre-processed data set generation formula is: D = {(x1, y1, t1), (x2, y2, t2),..., (x n , y n , t n )}, where x and y are spatial positions, t is the time point, and D is the pre-processed data set; through the above steps, the scientificity and operability of each step are ensured, and through outlier rejection, repeated data processing, missing data filling and normalization processing, a complete and consistent pre-processed data set is generated, providing a solid data foundation for subsequent data analysis and model construction.

[0094] S3 specifically includes:

[0095] S31: Based on the pre-processed data set, organize the multi-dimensional data corresponding to different geotechnical parameters (including stress, strain, permeability, etc.) in matrix form, and construct a data matrix for calculating distance correlation; the data matrix form is as follows: where X is the data matrix, m is the sample number, and n is the dimension of the geotechnical parameter;

[0096] S32: Calculate the distance matrix, for each pair of geotechnical parameters in the data matrix, use the Euclidean distance to calculate the distance matrix between each pair of parameters, the distance calculation formula is as follows:

[0097] where d(x i ,x j ) represents the Euclidean distance between the i-th and j-th samples, x ik and x jk represent the values of the i-th and j-th samples in the k-th dimension, and n represents the dimension of the geotechnical parameter;

[0098] S33: Based on the calculated distance matrix, use the distance correlation analysis method to calculate the distance correlation coefficient between different geotechnical parameters, the calculation formula of the distance correlation coefficient dCor(X, Y) is: Wherein, dCor(X, Y) represents the distance correlation between parameters X and Y, dCov(X, Y) represents the distance covariance of parameters X and Y, dCov(X, X) and dCov(Y, Y) respectively represent the distance covariance of parameters X and Y themselves;

[0099] S34: Feature variable extraction, according to the calculated distance correlation coefficient matrix, set a predetermined correlation threshold, and select the soil parameters with strong correlation as the feature variables, specifically select the variables with correlation coefficient greater than the set threshold as the feature variables;

[0100] S35: Finally, the extracted feature variables are constructed into a feature data set, and the format of the feature data set is as follows: D 特征 ={f1, f2, …, f k}, wherein D feature represents the feature data set, f1, f2, …, f k is the selected feature variable, and k is the number of extracted feature variables; the above steps use the distance correlation analysis method to process the soil test data, calculate the distance correlation between different soil parameters, extract the key feature variables, and generate the feature data set, through the reasonable selection of the distance measurement method and the correlation threshold, ensure that the extracted feature variables have high representativeness and accuracy for subsequent modeling and prediction, and provide high-quality data basis for the soil performance prediction model.

[0101] S4 specifically includes:

[0102] S41: Time series segmentation, according to the time dimension in the feature data set, the data is segmented according to the time sequence, a fixed time interval is set, and the entire data set is divided into multiple time periods; specifically, first set the time points t1, t2, …, t m , which are used to correspond to different time stages; then the data set is segmented according to the time points t i , and the data of each time period is allocated to different subsets The formula is as follows: Wherein, is the data subset of the i-th time period, t i and t i+1 are adjacent time points, x1, x2, …, x n represent feature variables of different dimensions;

[0103] S42: Spatial feature segmentation, based on the spatial position of the soil test area, the data is segmented according to the spatial features, and the data is divided into multiple spatial subsets by setting a fixed spatial coordinate interval; specifically, first set the spatial coordinate points (x s , y s), for corresponding to different positions in the geotechnical region; then according to the spatial position (x, y) of each data point, the data is allocated to different spatial region subsets D space , the formula is as follows: wherein, is the data subset of the i th spatial region, x s1 , x s2 and y s1 , y s2 are the upper and lower bounds of the spatial coordinates, x 1, x 2, …, x n is the characteristic variable within the region;

[0104] S43: Dual segmentation combining time series and spatial characteristics, first based on time series segmentation, the data is divided into multiple time periods, and then according to the spatial characteristics within each time period, multiple data subsets containing both time and spatial characteristics are generated; the specific formula is: wherein, represents the data subset that meets both the i th time period and the j th spatial region;

[0105] S44: Divide the training set and the test set, from the segmented data subset, divide the training set and the test set in proportion, specifically according to the proportion of 80% as the training set and 20% as the test set; the division formula is: D train = 0.8 x D combined and D test = 0.2 x D combined , wherein D train is the training set, D test is the test set, and D combined is the data subset after joint segmentation; the above steps ensure that the characteristic data set is fully segmented in time and space dimensions through detailed description of the dual segmentation method of time series and spatial characteristics, and reasonably divide the training set and the test set, providing reliable data sources for subsequent modeling and verification.

[0106] S5 specifically includes:

[0107] S51: Based on the characteristic variables in the training set, define the input and output of the Gaussian process regression model, the input is the variable in the characteristic data set, and the output is the soil performance parameter to be predicted; the input variables include stress, strain and permeability of soil parameters, and the output is soil performance indicators such as foundation bearing capacity or settlement;

[0108] S52: Select radial basis kernel function as the kernel function for the Gaussian process regression model, which reflects the correlation between input characteristic variables by measuring the Euclidean distance between input data points, and controls the smoothness and fitting effect of the model; the calculation formula of the radial basis kernel function is: Where, k(x) p x q ) represents the input data point x p and x q The correlation between them, ||x p -x q || 2 Represents data point x p and x q The squared Euclidean distance between them, l is the length scale hyperparameter of the kernel function, used to control the model's sensitivity to changes in the input data;

[0109] S53: Based on the radial basis function selected in S52, calculate the covariance matrix among all input variables in the training set. The covariance matrix is ​​used to quantify the correlation between each input variable and the target output. Each element of the covariance matrix K is calculated by the kernel function, as shown in the following formula: K pq =k(x p x q ), where K pq Let k(x) represent the covariance between the p-th and q-th data points in the training set. p x q The correlation is obtained by calculating the radial basis kernel function;

[0110] S54: Using the training set, optimize the hyperparameters in the Gaussian process regression model by maximizing the log-likelihood function, including the length scale of the radial basis kernel function and the noise parameter.

[0111] S55: After completing the hyperparameter optimization, the model is trained based on the training data to establish a geotechnical performance prediction model. This model establishes a mapping relationship between input variables and output geotechnical performance indicators through the calculated covariance matrix and optimized hyperparameters.

[0112] S54 specifically includes:

[0113] S541: Based on the Gaussian process regression model, define the log-likelihood function for the input and output data in the training set. Let the log-likelihood function be L(θ), then the expression is: Where L(θ) is the log-likelihood function of the hyperparameter θ, y represents the output data of the training set, K is the covariance matrix, n1 is the number of training data, and |K| represents the determinant of the covariance matrix K. This represents the hyperparameters to be optimized, including the length scale l and noise parameters.

[0114] S542: Gradient calculation, in order to optimize the log-likelihood function, the gradient of the hyperparameter θ is calculated, and the gradient calculation formula is: wherein, represents the gradient of the log-likelihood function with respect to the hyperparameter θ, represents the derivative of the covariance matrix K with respect to the hyperparameter θ;

[0115] S543: Based on the gradient calculation result, the gradient descent method is used to update the hyperparameter; the update formula is: wherein, θ (t) is the hyperparameter value at the tthiteration, η is the learning rate, is the gradient of the log-likelihood function;

[0116] S544: In each iteration, the length scale and noise parameter are optimized respectively, wherein the length scale is used to control the correlation between the input variables, and the noise parameter is used to measure the tolerance of the model to data noise;

[0117] S545: Set the convergence condition, specifically when the change of the log-likelihood function is less than the preset threshold, or when the maximum iteration number is reached, terminate the iteration, and finally obtain the optimal hyperparameter θ * , including the optimal length scale l * and the optimal noise parameter

[0118] S55 specifically includes:

[0119] S551: Determine the final covariance matrix, after completing the hyperparameter optimization, use the optimal hyperparameters l * and to calculate the final covariance matrix K * of the training data, the formula is:

[0120] wherein, K *pq represents the covariance between the training set data points x p and x q , δ pq is the Kronecker delta function, when p=q, δ pq =1, otherwise δ pq =0;

[0121] S552: Based on the training data and the new input data x * , calculate the output mean μ(x * ) of the geotechnical performance prediction model, and the prediction mean μ(x * ) of the Gaussian process regression model, the expression is: wherein, k * is the new input data x *Covariance vector between the training data, K * is the final covariance matrix, y is the target value of the training data

[0122] S553: Calculate the prediction variance σ 2 (x * ) of the geotechnical performance prediction model, which is used to measure the uncertainty of the prediction, and the calculation formula of the prediction variance is: where k(x * , x * ) is the covariance of the new input data itself, represents the correction term caused by the correlation between the input data and the training data

[0123] S554: After the training of the model is completed, the expression of the geotechnical performance prediction model is:

[0124] f(x * ) ~ N(μ(x * ), σ 2 (x * )), where f(x * ) is the target function of the geotechnical performance prediction, the prediction result obeys the normal distribution, the mean value is μ(x * ), and the variance is σ 2 (x * ); Through the above steps, the geotechnical performance prediction model is established, and the final geotechnical performance prediction model can not only give the mean value of the prediction, but also provide the uncertainty measure of the prediction, which provides more reliable prediction basis for practical engineering application.

[0125] S6 specifically includes:

[0126] S61: Based on the test set, the geotechnical performance prediction model is verified, and the feature variable x test in the test set is input into the trained Gaussian process regression model to obtain the corresponding prediction value The verification formula is: where is the prediction output of the model, and μ(x test ) is the prediction mean value calculated according to the training data and the test set data

[0127] S62: Calculate the prediction error to measure the difference between the model prediction value and the actual observation value y test , specifically using the mean square error as the error measurement standard, and the calculation formula is: where y test,i is the actual observation value in the test set, is the model prediction value, and n2 is the sample number in the test set

[0128] S63: Adjust the kernel function and hyperparameters of the Gaussian process regression model using a grid search optimization algorithm. First, set the search range of the hyperparameters, including the length scale of the kernel function and the noise parameter. Then, search point by point on the preset parameter grid, and select the parameter combination that minimizes the validation error;

[0129] The grid search process is as follows:

[0130] 1. Define the hyperparameter grid: where l1, l2, …, l k represent the candidate values of the length scale, represent the candidate values of the noise parameter;

[0131] 2. Calculate the prediction error under each pair of hyperparameter combination: where represents the predicted value using hyperparameters θ;

[0132] S64: According to the grid search results in S63, select the hyperparameter combination that minimizes the mean square error, and update the soil performance prediction model. The updated model can better fit the training data and improve the prediction accuracy of the test data.

[0133] S7 specifically includes:

[0134] S71: Collect the latest soil test data in real time through the sensor array, including the stress, strain, and permeability parameters of the soil. Input the latest soil test data as new input data into the optimized soil performance prediction model;

[0135] S72: Use the S6 optimized soil performance prediction model to predict the new input soil test data in real time, and calculate the predicted value of the soil mechanical parameters 2 (x new ) and the corresponding prediction variance σ 2 (x new ) according to the new input data; The calculation formulas of the prediction mean and variance are as follows: where k new is the covariance vector between the new input data x new and the training data, K * is the covariance matrix of the training data, and y is the target value in the training set;

[0136] S73: Based on the prediction mean and the prediction variance σ 2 (x new ),), the confidence interval of the prediction result of the soil mechanical parameter is calculated, and the prediction result is subject to normal distribution, and the calculation formula of the confidence interval is: wherein z is a standard normal distribution critical value corresponding to a confidence level (for example, for a 95% confidence level, z is approximately 1.96), and sigma (x new ) is the square root of the prediction variance.

[0137] S74: feeding back the prediction result and the corresponding confidence interval CI new to the construction site in real time, specifically, the prediction value and the confidence interval are displayed in a visual form in the construction control system through a data transmission interface (for example, wireless transmission or a cloud computing platform) for reference by construction personnel; the above steps describe how to use newly collected soil test data to predict key soil mechanical parameters in real time based on the optimized soil performance prediction model, and feed back the prediction result and the confidence interval to the construction site, which ensures the reliability and timeliness of the model prediction result, provides accurate decision support for construction personnel, and improves the safety and efficiency of the engineering project.

[0138] The present application encompasses any substitutions, modifications, equivalent methods and solutions made on the essence and scope of the present application. In order for the public to have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits, etc. are not described in detail.

[0139] The above is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be considered as the protection scope of the present application.

Claims

1. A method for analyzing and predicting results of geotechnical tests based on big data, characterized in that, The method comprises the following steps: S1: Real-time acquisition of multi-dimensional data in soil test by a preset sensor array, the multi-dimensional data including stress, strain and permeability of soil; S2: Data cleaning and preprocessing of the multi-dimensional data collected in S1, the data cleaning including removing outliers, processing repeated data and filling missing data; the preprocessing including normalizing the cleaned multi-dimensional data to generate a preprocessed data set; S3: Based on the preprocessed data set, calculating the distance correlation between different soil parameters by a distance correlation analysis method to extract feature variables and generate a feature data set; S4: Dividing the feature data set into a training set and a test set according to a method combining time sequence and spatial features; S5: Based on the training set, modeling by a Gaussian process regression algorithm to construct a soil performance prediction model; S6: Verifying the soil performance prediction model based on the test set, calculating the prediction error, and adjusting the kernel function and hyperparameters of the Gaussian process regression model by a grid search optimization algorithm; S7: Based on the optimized soil performance prediction model, real-time prediction of the mechanical parameters of soil by using newly collected soil test data, and feeding back the prediction results and their confidence intervals to the construction site.

2. The big data-based geotechnical test result analysis and prediction method according to claim 1, characterized in that, The S1 specifically comprises: S11: Arranging the preset sensor array in the soil area to be measured, the sensor array including stress sensors, strain sensors and permeability sensors; S12: The stress sensor is a pressure sensor, which can record the normal stress and shear stress received by the soil at different time points at each preset acquisition point; S13: The strain sensor is a multi-axis strain gauge, which can detect the deformation of the soil in different directions to capture the tensile, compressive or shear deformation of the soil in real time; S14: The permeability sensor is a seepage meter, which can detect the water flow velocity, pressure change and water content in the soil.

3. The big data based geotechnical test result analysis and prediction method according to claim 1, characterized in that, The S2 specifically comprises: S21: Outlier removal, identifying and removing outliers in the collected data by a preset statistical method, the preset statistical method being a three-sigma rule method or a boxplot method based on standard deviation; S22: Repeated data processing, detecting the repeatability of the multi-dimensional data, comparing using a unique identifier of a hash value, and merging or removing the same data points that appear repeatedly; S23: Missing data filling, filling the missing data caused by equipment or environmental factors during the collection process by an interpolation algorithm; S24: Normalizing the cleaned multi-dimensional data set, using maximum and minimum normalization or Z-score standardization to convert the data values of different dimensions to the same order of magnitude range, ensuring comparability between different physical quantities and providing consistent input for subsequent analysis; S25: Summarizing the multi-dimensional data set after normalization, reorganizing the data structure according to the collection time and spatial dimension to generate a preprocessed data set.

4. The big data-based geotechnical test result analysis and prediction method according to claim 1, characterized in that, The S3 specifically comprises: S31: Based on the pre-processed data set, the multi-dimensional data corresponding to different geotechnical parameters is organized in matrix form to construct a data matrix for calculating distance correlation; S32: Calculate the distance matrix. For each pair of geotechnical parameters in the data matrix, calculate the distance matrix between each pair of parameters using the Euclidean distance; S33: Based on the calculated distance matrix, use the distance correlation analysis method to calculate the distance correlation coefficient between different geotechnical parameters; S34: Feature variable extraction. According to the calculated distance correlation coefficient matrix, set a pre-set correlation threshold, and select the geotechnical parameters with strong correlation as the feature variables. Specifically, select the variables with correlation coefficients greater than the set threshold as the feature variables; S35: Finally, construct a feature data set from the extracted feature variables.

5. The big data based geotechnical test result analysis and prediction method according to claim 1, characterized in that, The S4 specifically includes: S41: Time series segmentation. According to the time dimension in the feature data set, the data is segmented in time sequence, a fixed time interval is set, and the entire data set is divided into multiple time periods; S42: Spatial feature segmentation. Based on the spatial position of the geotechnical test area, the data is segmented according to the spatial features. By setting a fixed spatial coordinate interval, the data is divided into multiple spatial subsets; S43: Dual segmentation combining time series and spatial features. First, based on time series segmentation, the data is divided into multiple time periods. Then, according to the spatial features, the data is segmented within each time period to generate multiple data subsets containing both time and spatial features; S44: Divide the training set and the test set. From the segmented data subsets, divide the training set and the test set in proportion. Specifically, allocate 80% as the training set and 20% as the test set.

6. The big data-based geotechnical test result analysis and prediction method according to claim 1, characterized in that, The S5 specifically includes: S51: Based on the feature variables in the training set, define the input and output of the Gaussian process regression model. The input is the variable in the feature data set, and the output is the geotechnical performance parameter to be predicted. The input variables include stress, strain, and permeability geotechnical parameters, and the output is the geotechnical performance index; S52: Select a radial basis kernel function as the kernel function for the Gaussian process regression model. This radial basis kernel function measures the Euclidean distance between input data points to reflect the correlation between input feature variables; S53: Based on the radial basis kernel function selected in S52, calculate the covariance matrix between all input variables in the training set. The covariance matrix is used to quantify the correlation between each input variable and the target output; S54: Use the training set to optimize the hyperparameters in the Gaussian process regression model, including the length scale of the radial basis kernel function and the noise parameter, by maximizing the log-likelihood function; S55: After completing the hyperparameter optimization, train the model based on the training data to establish a geotechnical performance prediction model.

7. The big data-based geotechnical test result analysis and prediction method according to claim 6, characterized in that, The S54 specifically includes: S541: Based on the Gaussian process regression model, define the log-likelihood function of the input data and output data in the training set; S542: Gradient calculation. To optimize the log-likelihood function, calculate the gradient of the hyperparameters; S543: Based on the gradient calculation result, update the hyperparameters using the gradient descent method; S544: In each iteration, the length scale and noise parameters are optimized respectively; S545: Set the convergence condition, specifically when the change of log-likelihood function is less than the preset threshold, or when the maximum iteration number is reached, terminate the iteration, and finally obtain the optimal hyperparameter θ * , including the optimal length scale l * and the optimal noise parameter 8.The big data based geotechnical test result analysis and prediction method according to claim 7, characterized in that, The S55 specifically comprises: S551: determining the final covariance matrix, after completing the hyperparameter optimization, with the optimal hyperparameters l * and computing the final covariance matrix K of the training data * ; S552: Compute the output mean μ(x) of the geotechnical performance prediction model based on the training data and the new input data x * * );​ S553: Calculate the prediction variance σ of the geotechnical performance prediction model 2 (x * ) for measuring the uncertainty of the prediction; S554: After the training of the model is completed, the expression of the geotechnical performance prediction model is: f(x * ) ~ N(μ(x * ), σ 2 (x * )) where f(x * ) is the objective function of geotechnical performance prediction, the prediction result obeys normal distribution, the mean value is μ(x * ), and the variance is σ 2 (x * ). 9.The big data based geotechnical test result analysis and prediction method according to claim 1, wherein, The S6 specifically comprises: S61: verifying the geotechnical performance prediction model based on the test set, inputting the feature variables x in the test set test Enter the trained Gaussian process regression model to obtain the corresponding predicted value The verification formula is: Wherein, The predicted output of the model is μ(x test ) is the predicted mean value calculated according to the training data and test set data; S62 calculate prediction error, measure the difference between model predicted value and actual observation value y test , specifically using mean square error as error measurement standard; S63: The kernel function and hyperparameters of the Gaussian process regression model are adjusted using a grid search optimization algorithm. The search range of the hyperparameters is first set, including the length scale and noise parameters of the kernel function. Then, the parameters are searched point by point on the preset parameter grid, and the parameter combination that minimizes the validation error is selected; S64: According to the grid search results in S63, the hyperparameter combination that minimizes the mean square error is selected, and the geotechnical performance prediction model is updated. 10.The big data based geotechnical test result analysis and prediction method according to claim 1, wherein, The S7 specifically comprises: S71: Real-time acquisition of the latest geotechnical test data through the sensor array, the collected data including the stress, strain and permeability parameters of the soil, and the latest geotechnical test data as new input data is input into the optimized geotechnical performance prediction model; S72: Real-time prediction of the new input geotechnical test data by using the optimized geotechnical performance prediction model of S6, and calculating the predicted value of the mechanical parameters of the soil according to the new input data and the corresponding prediction variance σ 2 (x new ); S73: Based on the predicted mean and the predicted variance 2 (x new ), the confidence interval of the predicted result of the soil mechanical parameter is calculated, and the calculation formula of the confidence interval is: wherein z is the critical value of the standard normal distribution corresponding to the confidence level, and new is the square root of the predicted variance. S74: the prediction result is fed back to the construction site in real time for reference by the construction personnel. and the corresponding confidence interval CI new Real-time feedback to the construction site for reference by the construction personnel.

Images