Variable cross-section composite pile reinforcement design method and system based on machine learning

Abstract

The invention discloses a variable cross-section composite pile reinforcement design method and system based on machine learning. The method comprises the steps that multi-source data of a variable cross-section composite pile is obtained; performing feature extraction on the multi-source data to obtain key features and spatio-temporal features; based on the multi-source data, three-dimensional parametric modeling and nonlinear finite element analysis are carried out, a variable cross-section form and a stress nephogram of the pile body are constructed, and variable cross-section form parameters are obtained; according to the key features, the spatial-temporal features and the variable cross-section morphological parameters, performing multi-target prediction by adopting a preset integrated learning model to obtain a structural safety coefficient, material cost and construction feasibility; and based on the structural safety coefficient, the material cost and the construction feasibility, a mixed integer programming algorithm is adopted, and a steel bar configuration scheme of the variable-cross-section composite pile is generated. According to the method, the efficiency and accuracy of variable-cross-section composite pile reinforcement design can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a variable-section composite pile reinforcement design method and system based on machine learning. Background Art

[0002] Variable-section composite piles are a type of pile foundation that combines different materials or cross-sectional dimensions. Through rational design and optimization, varying pile materials or cross-sectional dimensions at different depths effectively enhance the pile foundation's bearing capacity, deformation resistance, and adaptability to complex geological conditions. This type of pile is suitable for infrastructure construction in soft soils and complex geological environments, effectively addressing issues such as foundation settlement and lateral deformation, and effectively improving the overall stability and safety of the pile foundation.

[0003] Most current variable-section composite pile reinforcement design systems have limitations. Traditional methods often rely on empirical formulas and manual calculations, unable to effectively handle complex multi-source data and the changing engineering environment. This results in a cumbersome design process susceptible to human influence. Regarding optimization algorithms, most traditional systems lack the flexibility to perform multi-objective optimization and typically focus on a single objective, neglecting other key factors such as constructability. Existing variable-section composite pile reinforcement methods suffer from inefficiencies in efficiency and accuracy. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a variable-section composite pile reinforcement design method and system based on machine learning, which can effectively improve the efficiency and accuracy of variable-section composite pile reinforcement design.

[0005] An embodiment of the present invention provides a variable-section composite pile reinforcement design method based on machine learning, comprising: Acquire multi-source data of variable-section composite piles; Extracting features from the multi-source data to obtain key features and spatiotemporal features; Based on the multi-source data, three-dimensional parametric modeling and nonlinear finite element analysis are performed to construct the variable cross-section morphology and stress cloud diagram of the pile body and obtain the variable cross-section morphology parameters; Based on the key features, the spatiotemporal features, and the variable-section morphological parameters, a preset integrated learning model is used to perform multi-objective prediction to obtain the structural safety factor, material cost, and construction feasibility; Based on the structural safety factor, the material cost and the construction feasibility, a mixed integer programming algorithm is used to generate a reinforcement configuration scheme for the variable-section composite pile.

[0006] As an improvement of the above solution, the base models of the integrated learning model include: a random forest model, a gradient boosting decision tree model and a deep residual network model.

[0007] As an improvement of the above solution, the multi-source data includes: engineering geological parameters, load condition parameters, material performance parameters, cross-section parameters, and historical engineering data.

[0008] As an improvement to the above solution, the method of obtaining multi-source data of variable-section composite piles includes: Obtain engineering geological parameters of variable-section composite piles through geological radar detection, in-situ testing, and dynamic monitoring of groundwater levels; Obtain the load condition parameters of the variable-section composite pile through design load spectrum decomposition and construction load simulation; Obtain the material performance parameters of variable-section composite piles through concrete material testing and steel bar performance testing; The cross-sectional parameters of the variable cross-sectional composite pile are obtained through variable cross-sectional geometric parameterization and cross-sectional stiffness parameter calculation. Obtain historical engineering data of variable-section composite piles from the engineering database.

[0009] As an improvement to the above solution, the feature extraction of the multi-source data to obtain key features and spatiotemporal features includes: Normalizing the multi-source data to obtain normalized data; Based on the time correlation and spatial correlation of the normalized data, a spatiotemporal correlation matrix is ​​constructed to obtain spatiotemporal features; Using principal component analysis to reduce the dimension of the normalized data to obtain feature data; According to the feature data, a mutual information entropy algorithm is used to calculate the mutual information between each feature data and the target variable, and several feature data with the largest mutual information are screened out to obtain key features.

[0010] As an improvement to the above solution, the formula for the time correlation is: in, is the temporal correlation between the i-th feature and the j-th feature at time t, is the covariance between the i-th feature and the j-th feature at time t, 、 are the variances of the i-th feature and the j-th feature at time t respectively; The formula for the spatial correlation is: in, is the spatial correlation between the i-th feature and the j-th feature at spatial position s, is the covariance between the i-th feature and the j-th feature at spatial position s, are the variances of the i-th feature and the j-th feature at spatial position s, respectively.

[0011] As an improvement to the above solution, the training method of the ensemble learning model includes: Based on spatiotemporal features, a two-stream neural network is constructed to process temporal features and spatial features respectively, and tensor product operations are used to perform deep interaction between temporal and spatial features to obtain interactive features; The key features of the variable-section composite pile, the variable-section morphological parameters and the interactive features are used as input features of the model, and the base model of the ensemble learning model is trained in parallel; Calculating the feature importance of the input feature and determining the weight of each base model according to the feature importance; Obtaining an initial ensemble learning model according to the base model and the weight of the base model; Based on the initial integrated learning model, multi-objective optimization and Bayesian optimization are performed to obtain a final integrated learning model; wherein the objective function of the multi-objective optimization is constructed based on the structural safety factor, material cost and construction feasibility.

[0012] As an improvement to the above solution, a mixed integer programming algorithm is used to generate a reinforcement configuration scheme for a variable-section composite pile based on the structural safety factor, the material cost, and the construction feasibility, including: Obtain flexural capacity, crack width, and reinforcement ratio; An integer programming model is constructed by using the bending bearing capacity, the crack width, the reinforcement ratio, the structural safety factor, the material cost, and the construction feasibility as constraint conditions, using binary variables to represent the number and distribution of steel bars and constructing an objective function; The integer programming model is input into a mixed integer programming solver for solving, thereby obtaining an optimal reinforcement configuration scheme.

[0013] As an improvement to the above solution, the integer programming model adjusts the total amount of steel bars by minimizing the objective function. The objective function formula is: in, is the cross-sectional area of ​​the j-th steel bar, is a binary variable.

[0014] The embodiment of the present invention further provides a variable-section composite pile reinforcement design system based on machine learning, comprising: A data acquisition module, used to acquire multi-source data of variable-section composite piles; A feature extraction module is used to extract features from the multi-source data to obtain key features and spatiotemporal features; A variable cross-section processing module is used to perform three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, construct the variable cross-section shape and stress cloud map of the pile body, and obtain the variable cross-section shape parameters; A model prediction module is used to perform multi-objective prediction using a preset ensemble learning model based on the key features, the spatiotemporal features, and the variable-section morphological parameters to obtain the structural safety factor, material cost, and construction feasibility; The reinforcement optimization module is used to generate a reinforcement configuration plan for the variable-section composite pile based on the structural safety factor, the material cost and the construction feasibility using a mixed integer programming algorithm.

[0015] Compared with the existing technology, the beneficial effects of the variable-section composite pile reinforcement design method and system based on machine learning provided by the embodiment of the present invention are: by obtaining multi-source data of variable-section composite piles, a variety of influencing factors are fully considered, and the comprehensiveness of the reinforcement scheme is improved; by extracting features from the multi-source data to obtain key features and spatiotemporal features, the most discriminative features can be obtained for subsequent prediction and optimization, thereby improving the reliability of the design; by performing three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, the variable-section morphology and stress cloud map of the pile body are constructed, and the variable-section morphological parameters are obtained, thereby further improving the design reliability. The system can improve the design accuracy and ensure the structural safety of pile foundations under different working conditions. It uses a preset integrated learning model to perform multi-objective prediction based on key features, spatiotemporal features and variable cross-section morphological parameters to obtain the structural safety factor, material cost and construction feasibility. Then, based on the structural safety factor, material cost and construction feasibility, a mixed integer programming algorithm is used to generate the optimal reinforcement configuration scheme. This system can automatically adjust and optimize the reinforcement scheme under complex and changing design requirements to meet the multiple goals of structural safety, cost control and construction feasibility, thereby improving design efficiency, reducing human errors and improving the accuracy of the reinforcement design of variable-section composite piles. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 1 is a flow chart of a variable-section composite pile reinforcement design method based on machine learning provided by an embodiment of the present invention; Figure 2 This is a structural schematic diagram of a variable-section composite pile reinforcement design system based on machine learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] See also Figure 1 , Figure 1 The figure is a flow chart of a variable cross-section composite pile reinforcement design method based on machine learning provided by an embodiment of the present invention. The variable cross-section composite pile reinforcement design method based on machine learning includes: S1: Acquire multi-source data of variable-section composite piles; Specifically, based on geological exploration, load conditions, material properties, cross-sectional parameters and historical data, multi-source data of variable-cross-section composite piles is obtained. Through comprehensive data collection, the embodiment of the present invention can fully consider various influencing factors such as engineering geology, load conditions, material properties, etc., and realize accurate optimization of the composite pile reinforcement scheme.

[0019] As one of the optional embodiments, the multi-source data includes: engineering geological parameters, load condition parameters, material performance parameters, cross-section parameters, and historical engineering data.

[0020] As one of the optional embodiments, the obtaining of multi-source data of the variable-section composite pile includes: Obtain engineering geological parameters of variable-section composite piles through geological radar detection, in-situ testing, and dynamic monitoring of groundwater levels; Obtain the load condition parameters of the variable-section composite pile through design load spectrum decomposition and construction load simulation; Obtain the material performance parameters of variable-section composite piles through concrete material testing and steel bar performance testing; The cross-sectional parameters of the variable cross-sectional composite pile are obtained through variable cross-sectional geometric parameterization and cross-sectional stiffness parameter calculation. Obtain historical engineering data of variable-section composite piles from the engineering database.

[0021] S2: Extract features from the multi-source data to obtain key features and spatiotemporal features; As one of the optional embodiments, the feature extraction of the multi-source data to obtain key features and spatiotemporal features includes: Normalizing the multi-source data to obtain normalized data; Based on the time correlation and spatial correlation of the normalized data, a spatiotemporal correlation matrix is ​​constructed to obtain spatiotemporal features; Using principal component analysis to reduce the dimension of the normalized data to obtain feature data; According to the feature data, a mutual information entropy algorithm is used to calculate the mutual information between each feature data and the target variable, and several feature data with the largest mutual information are screened out to obtain key features.

[0022] Specifically, the multi-source data is normalized by Min-Max normalization, and the normalization formula is: in, represents normalized data, 、 are the minimum and maximum values ​​of the feature in the dataset.

[0023] Furthermore, based on the normalized data and the spatiotemporal correlation of the data, a temporal correlation matrix is ​​constructed. In multi-source data, the values ​​of each feature data at different time points form a time series, and the values ​​at different spatial locations form a spatial series. For time series, the temporal correlation between the feature data is calculated. For spatial series, the spatial correlation between the feature data is calculated. Furthermore, based on the temporal and spatial correlations between the feature data, a spatiotemporal correlation matrix is ​​constructed to capture the temporal and spatial dependencies of each feature data.

[0024] As one of the optional embodiments, the time correlation formula is: in, is the temporal correlation between the i-th feature and the j-th feature at time t, is the covariance between the i-th feature and the j-th feature at time t, 、 are the variances of the i-th feature and the j-th feature at time t respectively; The formula for the spatial correlation is: in, is the spatial correlation between the i-th feature and the j-th feature at spatial position s, is the covariance between the i-th feature and the j-th feature at spatial position s, are the variances of the i-th feature and the j-th feature at spatial position s, respectively.

[0025] Furthermore, for the extraction of key features, the normalized data is first mapped to a new coordinate system through principal component analysis to reduce the dimension of the data; then the amount of information between different feature data is calculated based on the mutual information entropy algorithm, and the most discriminative features are screened as key features.

[0026] In the principal component analysis dimensionality reduction process, the covariance matrix is ​​calculated for the normalized data matrix, and the eigenvalue decomposition of the covariance matrix is ​​performed to obtain the eigenvalues ​​and eigenvectors. Then, the eigenvalues ​​are sorted from large to small, and the principal components corresponding to the first k eigenvectors are selected. These principal components will form a new coordinate system, and the original data will be projected onto these principal components to obtain the feature data after dimensionality reduction. Mutual information entropy is a way to measure the correlation and information between two variables. First, the entropy of each feature is expressed as follows: in, Features The entropy of Features The probability distribution of Features The value of The mutual information between features is calculated based on the entropy of each feature. The formula is: in, for 、 The joint entropy of Then, by calculating the mutual information between each feature data and the target variable, the feature data with the largest mutual information is used as the key feature. The target variable can be obtained from step S3.

[0027] S3: Based on the multi-source data, three-dimensional parametric modeling and nonlinear finite element analysis are performed to construct the variable cross-section morphology and stress cloud diagram of the pile body and obtain the variable cross-section morphology parameters; As one of the optional embodiments, step S3 specifically includes: S31: Set key control sections according to the pile axis, define section control parameters, and construct a non-uniform rational B-spline (NURBS) model; S32: Construct spatial topological relationships by establishing the Frenet frame of the pile axis and calculating the gradient tensor of the cross-section gradient; S33: Based on the concrete and steel parameter data from multiple sources, a nonlinear finite element analysis is performed and a stress invariant cloud diagram is constructed to obtain the variable cross-section morphological parameters. The stress invariant cloud diagram is a graphical representation of the stress distribution of the pile cross section obtained through finite element analysis. The stress magnitude can be viewed through the color gradient of the cloud diagram.

[0028] Specifically, in step S31, the pile axis is fitted through multiple discrete points, or a continuous geometric curve is generated through the collected data. Then, according to the geometric characteristics of the pile body and the construction design requirements, the positions of several key sections are determined, and the control parameters for defining the sections are obtained. Based on the above parameter data, namely the pile axis control points and the section control parameters (such as radius, shape coefficient, stiffness distribution, and weight coefficient set based on the section surface shape), a non-uniform rational B-spline surface model is constructed. The formula is: in, is a point on the surface, are two variables in the parameter space, 、 is the B-spline basis function, is the weight of the control point, is the position of the control point.

[0029] In step S32, the Frenet frame describes the coordinate system of the tangent, normal, and binormal directions of the curve, parameterizes the pile axis, and obtains the tangent, normal, and binormal lines T(s), N(s), and B(s) of the pile axis at position s. Based on the axis position and cross-sectional shape, the deformation and gradient of the cross section on each section are calculated using the formula: Then, get the cross-sectional gradient tensor: Based on the Frenet frame and cross-sectional shape gradient, the local coordinate system is calculated at each key position and the topological relationship of the pile in space is established.

[0030] Furthermore, the embodiment of the present invention also dynamically adjusts the adaptive grid based on the cross-sectional curvature change rate.

[0031] In step S33, when performing finite element analysis, the geometry and material properties (concrete and steel) of the pile must first be modeled, assuming that the nonlinear material behavior is described by a bilinear stress-strain relationship or a more complex constitutive relationship. Taking concrete and steel as an example, assuming that the stress-strain relationship of concrete adopts the Mohr-Coulomb criterion or the Cesar-Castillo constitutive model, the finite element analysis method is used to discretize the geometry and material properties of the pile and transform the problem into a set of algebraic equations. The discretized equations can be expressed as: ,in, is the stiffness matrix, is the displacement vector, is the external force vector. This equation is used to calculate the deformation and displacement of the structure. Stress cloud maps can then be used to display the stress conditions of various parts of the structure based on these displacements and deformations. The variable cross-sectional shape and stress cloud maps of the constructed pile body can be used to obtain the variable cross-sectional shape parameters, providing key input data for subsequent model prediction and reinforcement optimization, improving the accuracy and reliability of reinforcement design.

[0032] S4: Based on the key features, the spatiotemporal features, and the variable-section morphological parameters, a preset integrated learning model is used to perform multi-objective prediction to obtain a structural safety factor, material cost, and construction feasibility; As one of the optional embodiments, the base models of the integrated learning model include: a random forest model, a gradient boosting decision tree model and a deep residual network model.

[0033] Specifically, the ensemble learning model is used to perform multi-objective predictions, including predicting stress distribution under different loads and estimating the feasibility of reinforcement schemes.

[0034] As one of the optional embodiments, the training method of the ensemble learning model includes: Based on spatiotemporal features, a two-stream neural network is constructed to process temporal features and spatial features respectively, and tensor product operations are used to perform deep interaction between temporal and spatial features to obtain interactive features; The key features of the variable-section composite pile, the variable-section morphological parameters and the interactive features are used as input features of the model, and the base model of the ensemble learning model is trained in parallel; Calculating the feature importance of the input feature and determining the weight of each base model according to the feature importance; Obtaining an initial ensemble learning model according to the base model and the weight of the base model; Based on the initial integrated learning model, multi-objective optimization and Bayesian optimization are performed to obtain a final integrated learning model; wherein the objective function of the multi-objective optimization is constructed based on the structural safety factor, material cost and construction feasibility.

[0035] Specifically, based on the constructed spatiotemporal correlation matrix, a two-stream neural network is constructed in the two dimensions of spatial flow and temporal flow, and deep interaction of spatial-temporal features is realized based on tensor product operations. By deeply interacting features in the two dimensions of space and time, the performance of the integrated learning model can be enhanced, so that each base model can better make predictions based on the spatiotemporal characteristics of the data, thereby improving the accuracy of the overall model.

[0036] The two-stream neural network processes the spatial flow and temporal flow in the spatiotemporal correlation matrix respectively, and its output is deeply interacted through tensor product. The formula is: in, Represents the tensor product operation, and deep interaction can be further integrated with features through multi-layer neural networks.

[0037] Furthermore, a parallel computing framework is used to simultaneously train multiple base models: random forest model, gradient boosted decision tree model, and deep residual network model to improve computational efficiency. Among them, the random forest model constructs multiple decision trees and performs voting decisions. The training process includes randomly selecting samples and feature subsets to train each tree; the gradient boosted decision tree model corrects the current model according to the residual of the previous model during each training; the deep residual network model solves the gradient vanishing problem in deep network training by introducing a residual module. The output of the residual module is: ,in, is the output of the residual block, is the input, is the network weight. By combining the advantages of multiple base models, the prediction accuracy is effectively improved.

[0038] Furthermore, the contribution weight of each base model is adjusted according to the importance of each feature. Specifically, the feature importance of each feature can be calculated by information gain, Gini index or gradient value. The machine learning in this application dynamically adjusts the weight of the base model through various mechanisms to adjust its predictive ability on specific features, including: calculating the Gini impurity drop or information gain (entropy reduction) to quantify the contribution of the feature to the prediction; evaluating the performance of each base model on a specific feature subset based on a reserved validation set and dynamically weighting it; using a two-stream neural network to extract cross-sectional morphology and load change features respectively, and adjusting its weight according to the performance of the base model on the fused spatiotemporal features through tensor product fusion.

[0039] Furthermore, assume that the outputs of the three base models are: 、 、 , after determining the weight of each base model, the final prediction output of the integrated learning model is: in, 、 、 are the weights of the three base models respectively.

[0040] When a base model has strong predictive ability on a specific feature, the weight of the base model will be increased, otherwise the weight will be reduced. By dynamically adjusting the contribution weights of each base model, the model can pay more attention to important features, thereby improving the accuracy of the overall prediction.

[0041] Furthermore, a target optimization function was constructed based on the structural safety factor, material cost, and construction feasibility. A crossover and mutation operator was designed to generate new population data and select new individuals to replace the old ones. A Gaussian process function model was constructed based on the Bayesian optimization framework, and iterative optimization was performed by optimizing the acquisition function. Multi-objective optimization was used to balance the safety factor, cost, and feasibility, while hyperparameters and network structure were tuned through Bayesian optimization to obtain the final ensemble learning model. When constructing the crossover and mutation operator, the crossover and mutation operators in the genetic algorithm were used to generate new population data. The crossover operation can adopt single-point crossover, two-point crossover, or uniform crossover, and the mutation operation can randomly change certain decision variables.

[0042] S5: Based on the structural safety factor, the material cost and the construction feasibility, a mixed integer programming algorithm is used to generate a reinforcement configuration plan for the variable-section composite pile.

[0043] As one of the optional embodiments, the generation of a reinforcement configuration scheme for a variable-section composite pile using a mixed integer programming algorithm based on the structural safety factor, the material cost, and the construction feasibility includes: Obtain flexural capacity, crack width, and reinforcement ratio; An integer programming model is constructed by using the bending bearing capacity, the crack width, the reinforcement ratio, the structural safety factor, the material cost, and the construction feasibility as constraint conditions, using binary variables to represent the number and distribution of steel bars and constructing an objective function; The integer programming model is input into a mixed integer programming solver for solving, thereby obtaining an optimal reinforcement configuration scheme.

[0044] Specifically, the structural safety factor, material cost and construction feasibility predicted by the integrated learning model are used as constraints, and the bending bearing capacity, crack width and reinforcement ratio of the national standard are combined to establish a nonlinear constraint equation group. Among them, the bending bearing capacity, crack width and reinforcement ratio are calculated based on the variable cross-section morphological parameters in step S3, and the bending bearing capacity is The calculation formula is: in, is the design compressive strength of concrete, is the cross-sectional area of ​​the steel bar, is the effective height of the component, is the anchorage length of the steel bar; Crack width The calculation formula is usually based on the crack control theory, which is: in, is the stress of the steel bar, is the reinforcement ratio, is a constant; Minimum reinforcement ratio The calculation formula is: in, is the yield strength of steel bars.

[0045] Furthermore, the number and distribution of steel bars in the reinforcement scheme are represented by binary variables, and an integer programming model is constructed. The integer programming model adjusts the total amount of steel bars by minimizing the objective function, and the objective function formula is: in, is the cross-sectional area of ​​the j-th steel bar, is a binary variable.

[0046] Furthermore, the integer programming model is solved using a mixed integer programming (MIP) solver. A new reinforcement configuration scheme is calculated based on the current solution, and the configuration is gradually adjusted to meet the constraints, ultimately obtaining a reinforcement configuration scheme that meets the design requirements. Furthermore, based on the structural parameters of the optimal reinforcement configuration scheme obtained, the parameters are input into CAD to automatically generate vector graphics, obtain corresponding construction drawings, and output construction guidance documents based on the reinforcement configuration scheme and finite element analysis results, realizing the automatic generation of reinforcement design and construction drawings, greatly improving design efficiency and reducing human errors. The embodiment of the present invention achieves efficient and accurate optimization of pile foundation reinforcement, can automatically generate the optimal reinforcement scheme, ensure structural safety, cost control, and construction feasibility, and greatly improve design efficiency and reliability.

[0047] Compared with the prior art, the embodiment of the present invention obtains multi-source data of variable-section composite piles, fully considers various influencing factors, improves the comprehensiveness of the reinforcement scheme, and obtains key features and spatiotemporal features by feature extraction of the multi-source data, so as to obtain the most discriminative features for subsequent prediction and optimization, thereby improving the reliability of the design; by performing three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, constructing the variable-section morphology and stress cloud map of the pile body, and obtaining the variable-section morphological parameters, the design accuracy is further improved, and the structural safety of the pile foundation under different working conditions is ensured; by using a preset integrated learning model for multi-objective prediction based on the key features, spatiotemporal features and variable-section morphological parameters, the structural safety factor, material cost and construction feasibility are obtained, and then based on the structural safety factor, material cost and construction feasibility, a mixed integer programming algorithm is used to generate the optimal steel bar configuration scheme, which can automatically adjust and optimize the reinforcement scheme under complex and changeable design requirements, meet the multiple goals of structural safety, cost control and construction feasibility, improve design efficiency and reduce human errors, and also improve the accuracy of the reinforcement design of variable-section composite piles.

[0048] Correspondingly, the present invention also provides a variable-section composite pile reinforcement design system based on machine learning, which can implement all processes of the variable-section composite pile reinforcement design method based on machine learning in the above embodiment.

[0049] See also Figure 2 , Figure 2 Schematic diagram of a variable cross-section composite pile reinforcement design system based on machine learning provided by an embodiment of the present invention. The variable cross-section composite pile reinforcement design system based on machine learning includes: The data acquisition module 201 is used to acquire multi-source data of variable-section composite piles; A feature extraction module 202 is used to extract features from the multi-source data to obtain key features and spatiotemporal features; The variable cross-section processing module 203 is used to perform three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, construct the variable cross-section shape and stress cloud diagram of the pile body, and obtain the variable cross-section shape parameters; The model prediction module 204 is configured to perform multi-objective prediction using a preset ensemble learning model based on the key features, the spatiotemporal features, and the variable cross-section morphological parameters to obtain a structural safety factor, material cost, and construction feasibility; The reinforcement optimization module 205 is used to generate a reinforcement configuration plan for the variable-section composite pile based on the structural safety factor, the material cost and the construction feasibility using a mixed integer programming algorithm.

[0050] Preferably, the base models of the integrated learning model include: a random forest model, a gradient boosting decision tree model and a deep residual network model.

[0051] Preferably, the multi-source data includes: engineering geological parameters, load condition parameters, material performance parameters, cross-section parameters, and historical engineering data.

[0052] Preferably, the step of obtaining multi-source data of variable-section composite piles includes: Obtain engineering geological parameters of variable-section composite piles through geological radar detection, in-situ testing, and dynamic monitoring of groundwater levels; Obtain the load condition parameters of the variable-section composite pile through design load spectrum decomposition and construction load simulation; Obtain the material performance parameters of variable-section composite piles through concrete material testing and steel bar performance testing; The cross-sectional parameters of the variable cross-sectional composite pile are obtained through variable cross-sectional geometric parameterization and cross-sectional stiffness parameter calculation. Obtain historical engineering data of variable-section composite piles from the engineering database.

[0053] Preferably, the feature extraction module 202 is specifically used to: Normalizing the multi-source data to obtain normalized data; Based on the time correlation and spatial correlation of the normalized data, a spatiotemporal correlation matrix is ​​constructed to obtain spatiotemporal features; Using principal component analysis to reduce the dimension of the normalized data to obtain feature data; According to the feature data, a mutual information entropy algorithm is used to calculate the mutual information between each feature data and the target variable, and several feature data with the largest mutual information are screened out to obtain key features.

[0054] Preferably, the time correlation formula is: in, is the temporal correlation between the i-th feature and the j-th feature at time t, is the covariance between the i-th feature and the j-th feature at time t, 、 are the variances of the i-th feature and the j-th feature at time t respectively; The formula for the spatial correlation is: in, is the spatial correlation between the i-th feature and the j-th feature at spatial position s, is the covariance between the i-th feature and the j-th feature at spatial position s, are the variances of the i-th feature and the j-th feature at spatial position s, respectively.

[0055] Preferably, the variable-section composite pile reinforcement design system based on machine learning is also used for: Based on spatiotemporal features, a two-stream neural network is constructed to process temporal features and spatial features respectively, and tensor product operations are used to perform deep interaction between temporal and spatial features to obtain interactive features; The key features of the variable-section composite pile, the variable-section morphological parameters and the interactive features are used as input features of the model, and the base model of the ensemble learning model is trained in parallel; Calculating the feature importance of the input feature and determining the weight of each base model according to the feature importance; Obtaining an initial ensemble learning model according to the base model and the weight of the base model; Based on the initial integrated learning model, multi-objective optimization and Bayesian optimization are performed to obtain a final integrated learning model; wherein the objective function of the multi-objective optimization is constructed based on the structural safety factor, material cost and construction feasibility.

[0056] Preferably, the reinforcement optimization module 205 is specifically used to: Obtain flexural capacity, crack width, and reinforcement ratio; An integer programming model is constructed by using the bending bearing capacity, the crack width, the reinforcement ratio, the structural safety factor, the material cost, and the construction feasibility as constraint conditions, using binary variables to represent the number and distribution of steel bars and constructing an objective function; The integer programming model is input into a mixed integer programming solver for solving, thereby obtaining an optimal reinforcement configuration scheme.

[0057] Preferably, the integer programming model adjusts the total amount of steel bars by minimizing an objective function, and the objective function formula is: in, is the cross-sectional area of ​​the j-th steel bar, is a binary variable.

[0058] In the specific implementation, the working principle, control process and technical effect achieved by the variable-section composite pile reinforcement design system based on machine learning provided by the embodiment of the present invention are the same as those of the variable-section composite pile reinforcement design method based on machine learning in the above-mentioned embodiment, and will not be repeated here.

[0059] The embodiment of the present invention provides a variable-section composite pile reinforcement design method and system based on machine learning, which has the following beneficial effects: by obtaining multi-source data of the variable-section composite pile, multiple influencing factors are fully considered, the comprehensiveness of the reinforcement scheme is improved, and by extracting the key features and spatiotemporal features of the multi-source data, the most discriminative features can be obtained for subsequent prediction and optimization, thereby improving the reliability of the design; by performing three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, the variable-section morphology and stress cloud map of the pile body are constructed, and the variable-section morphological parameters are obtained, thereby further improving the design accuracy and ensuring the structural safety of the pile foundation under different working conditions; by using a preset integrated learning model to perform multi-objective prediction based on the key features, spatiotemporal features and variable-section morphological parameters, the structural safety factor, material cost and construction feasibility are obtained, and then based on the structural safety factor, material cost and construction feasibility, a mixed integer programming algorithm is used to generate the optimal steel bar configuration scheme, which can automatically adjust and optimize the reinforcement scheme under complex and changing design requirements, meet the multiple goals of structural safety, cost control and construction feasibility, improve design efficiency and reduce human errors, and also improve the accuracy of the variable-section composite pile reinforcement design.

[0060] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A variable cross-section composite pile reinforcement design method based on machine learning, characterized in that: include: Acquire multi-source data of variable-section composite piles; Extracting features from the multi-source data to obtain key features and spatiotemporal features; Based on the multi-source data, three-dimensional parametric modeling and nonlinear finite element analysis are performed to construct the variable cross-section morphology and stress cloud diagram of the pile body and obtain the variable cross-section morphology parameters; Based on the key features, the spatiotemporal features, and the variable-section morphological parameters, a preset integrated learning model is used to perform multi-objective prediction to obtain the structural safety factor, material cost, and construction feasibility; Based on the structural safety factor, the material cost and the construction feasibility, a mixed integer programming algorithm is used to generate a reinforcement configuration scheme for the variable-section composite pile.

2. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 1, characterized in that: The base models of the ensemble learning model include: a random forest model, a gradient boosting decision tree model and a deep residual network model.

3. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 1, characterized in that: The multi-source data includes: engineering geological parameters, load condition parameters, material performance parameters, cross-section parameters, and historical engineering data.

4. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 3, characterized in that: The method of obtaining multi-source data of variable-section composite piles includes: Obtain engineering geological parameters of variable-section composite piles through geological radar detection, in-situ testing, and dynamic monitoring of groundwater levels; Obtain the load condition parameters of the variable-section composite pile through design load spectrum decomposition and construction load simulation; Obtain the material performance parameters of variable-section composite piles through concrete material testing and steel bar performance testing; The cross-sectional parameters of the variable cross-sectional composite pile are obtained through variable cross-sectional geometric parameterization and cross-sectional stiffness parameter calculation. Obtain historical engineering data of variable-section composite piles from the engineering database.

5. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 1, characterized in that: The feature extraction of the multi-source data to obtain key features and spatiotemporal features includes: Normalizing the multi-source data to obtain normalized data; Based on the time correlation and spatial correlation of the normalized data, a spatiotemporal correlation matrix is ​​constructed to obtain spatiotemporal features; Using principal component analysis to reduce the dimension of the normalized data to obtain feature data; According to the feature data, a mutual information entropy algorithm is used to calculate the mutual information between each feature data and the target variable, and several feature data with the largest mutual information are screened out to obtain key features.

6. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 5, characterized in that: The formula for the time correlation is: in, is the temporal correlation between the i-th feature and the j-th feature at time t, is the covariance between the i-th feature and the j-th feature at time t, 、 are the variances of the i-th feature and the j-th feature at time t respectively; The formula for the spatial correlation is: in, is the spatial correlation between the i-th feature and the j-th feature at spatial position s, is the covariance between the i-th feature and the j-th feature at spatial position s, are the variances of the i-th feature and the j-th feature at spatial position s, respectively.

7. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 2, characterized in that: The training method of the ensemble learning model includes: Based on spatiotemporal features, a two-stream neural network is constructed to process temporal features and spatial features respectively, and tensor product operations are used to perform deep interaction between temporal and spatial features to obtain interactive features; The key features of the variable-section composite pile, the variable-section morphological parameters and the interactive features are used as input features of the model, and the base model of the ensemble learning model is trained in parallel; Calculating the feature importance of the input feature and determining the weight of each base model according to the feature importance; Obtaining an initial ensemble learning model according to the base model and the weight of the base model; Based on the initial integrated learning model, multi-objective optimization and Bayesian optimization are performed to obtain a final integrated learning model; wherein the objective function of the multi-objective optimization is constructed based on the structural safety factor, material cost and construction feasibility.

8. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 1, characterized in that: The method of generating a steel bar configuration scheme for a variable-section composite pile using a mixed integer programming algorithm based on the structural safety factor, the material cost, and the construction feasibility includes: Obtain flexural capacity, crack width, and reinforcement ratio; An integer programming model is constructed by using the bending bearing capacity, the crack width, the reinforcement ratio, the structural safety factor, the material cost, and the construction feasibility as constraint conditions, using binary variables to represent the number and distribution of steel bars and constructing an objective function; The integer programming model is input into a mixed integer programming solver for solving, thereby obtaining an optimal reinforcement configuration scheme.

9. The variable cross-section composite pile reinforcement design method based on machine learning according to claim 8, characterized in that: The integer programming model adjusts the total amount of steel bars by minimizing the objective function, and the objective function formula is: in, is the cross-sectional area of ​​the j-th steel bar, is a binary variable.

10. A variable-section composite pile reinforcement design system based on machine learning, characterized in that: include: A data acquisition module, used to acquire multi-source data of variable-section composite piles; A feature extraction module is used to extract features from the multi-source data to obtain key features and spatiotemporal features; A variable cross-section processing module is used to perform three-dimensional parametric modeling and nonlinear finite element analysis based on the multi-source data, construct the variable cross-section shape and stress cloud map of the pile body, and obtain the variable cross-section shape parameters; A model prediction module is used to perform multi-objective prediction using a preset ensemble learning model based on the key features, the spatiotemporal features, and the variable-section morphological parameters to obtain the structural safety factor, material cost, and construction feasibility; The reinforcement optimization module is used to generate a reinforcement configuration plan for the variable-section composite pile based on the structural safety factor, the material cost and the construction feasibility using a mixed integer programming algorithm.

Images