Big data-based uplift pile bearing capacity optimization method

Abstract

This invention discloses a method for optimizing the bearing capacity of tensile piles based on big data. The method includes multi-source data acquisition, dual-control prediction of tensile piles, two-stage progressive parameter optimization of tensile piles, and intelligent optimization of tensile pile bearing capacity. This invention relates to the field of digital data processing technology, specifically to a method for optimizing the bearing capacity of tensile piles based on big data. This scheme innovatively proposes a two-stage progressive approach to optimize structural parameters and construction process parameters, achieving stability and safety coordination of the core bearing performance of tensile piles. It proposes an adaptive residual network dual-control synchronous prediction method with dual physical constraints, ensuring that the bearing performance of the optimized tensile pile scheme meets the standards and deformation is controllable. The design of an adaptive residual network architecture and a multi-constraint dynamic weight total loss function improves the prediction accuracy of the dual-control prediction model for tensile piles. Finally, it proposes an improved optimization algorithm with a dual nonlinear dynamic control mechanism, enhancing the completeness and efficiency of multi-parameter combination optimization for tensile piles.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method for optimizing the bearing capacity of pull-out piles based on big data. Background Technology

[0002] The method for optimizing the bearing capacity of tension piles is a big data-driven intelligent optimization method for the design and optimization of tension piles in civil engineering projects. By collecting multi-source data on tension pile projects and relying on big data processing and artificial intelligence modeling technology, it can accurately optimize and predict the pile design parameters, construction technology and bearing capacity, thereby realizing full-cycle dynamic monitoring, dynamic optimization and scientific quantitative evaluation of tension pile bearing capacity recovery, construction cost control and project safety improvement.

[0003] However, traditional methods for optimizing the bearing capacity of tension piles generally employ a combined optimization mode that integrates structural parameters and construction process parameters. This leads to technical problems such as insufficient tension pile bearing capacity and low construction efficiency in the optimized tension pile bearing capacity scheme. Furthermore, traditional methods for optimizing tension pile bearing capacity typically use a single-objective independent prediction mode without introducing physical constraints that conform to tension pile engineering specifications for rigid anchoring. This results in technical problems such as insufficient accuracy in tension bearing capacity prediction and prediction results easily exceeding the compliance boundaries of the specifications. Existing prediction models applicable to tension piles suffer from technical problems such as gradient vanishing, unstable training convergence, and low prediction accuracy. Existing algorithms applicable to tension pile bearing capacity optimization are prone to premature convergence and getting trapped in local optima in multi-parameter optimization scenarios for tension piles, resulting in insufficient optimization accuracy and stability. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a big data-based optimization method for the bearing capacity of uplift piles. Traditional methods for optimizing uplift pile bearing capacity often employ a hybrid optimization model combining structural parameters and construction process parameters, leading to insufficient uplift pile bearing capacity and low construction efficiency. This invention innovatively proposes a two-stage progressive approach to optimize structural parameters and construction process parameters, improving the stability and safety redundancy of the core bearing capacity of the uplift pile and achieving intelligent optimization of the entire uplift pile bearing capacity process. Furthermore, traditional methods for optimizing uplift pile bearing capacity often use a single-objective independent prediction model without introducing physical constraints conforming to uplift pile engineering specifications for rigid anchoring, resulting in insufficient accuracy and prediction results easily exceeding the compliance boundaries of specifications. This invention innovatively proposes an adaptive residual network dual-control synchronous prediction method with dual physical constraints, constructing a bearing capacity system that conforms to current pile foundation design specifications. The physical constraint equations of load-bearing capacity and the differential equations controlling pile deformation in elastic mechanics improve the synchronicity, accuracy, and stability of the dual-objective prediction of the ultimate bearing capacity and pile top tensile deformation of tensile piles. Addressing the technical problems of gradient vanishing, unstable training convergence, and low prediction accuracy in existing prediction models applicable to tensile piles, this scheme innovatively designs an adaptive residual network architecture and a multi-constraint dynamic weight total loss function. Through an iterative round-dependent dynamic weight strategy, it dynamically balances the loss contribution of data fitting and physical constraints, comprehensively improving the prediction accuracy, convergence stability, and cross-scenario generalization performance of the dual-control prediction model for tensile piles. Furthermore, addressing the technical problems of premature convergence, getting trapped in local optima, and insufficient optimization accuracy and stability in existing algorithms for optimizing the bearing capacity of tensile piles, resulting in the untapped potential for performance improvement, this scheme innovatively proposes an improved optimization algorithm with a dual nonlinear dynamic control mechanism, improving the completeness and comprehensiveness of multi-parameter combination optimization for tensile piles.

[0005] The technical solution adopted by this invention is as follows: The method for optimizing the bearing capacity of pull-out piles based on big data provided by this invention includes the following steps:

[0006] Step S1: Acquisition of multi-source data;

[0007] Step S2: Prediction of dual control of tensile piles;

[0008] Step S3: Two-stage progressive parameter optimization for tension piles;

[0009] Step S4: Intelligent optimization of tensile pile bearing capacity.

[0010] Further, in step S1, the multi-source data acquisition specifically involves collecting data information to obtain raw data for tension pile management, and performing data preprocessing on the raw data for tension pile management to obtain optimized data for tension pile management.

[0011] The raw data for tension pile management includes reference tension pile optimization data and target tension pile optimization data;

[0012] The reference anti-tension pile optimization data and target anti-tension pile optimization data include structural body parameter data, anti-tension pile construction process parameter data, site geological parameter data, site environmental parameter data, and pile-soil interface parameter data;

[0013] The reference tensile pile optimization data also includes the reference tensile pile bearing capacity results and the reference tensile deformation results;

[0014] The data preprocessing specifically involves sequentially performing data cleaning and normalization on the original data for anti-tension pile management to obtain optimized data for anti-tension pile management.

[0015] Furthermore, in step S2, the dual-control prediction of the pull-out pile specifically includes the following steps:

[0016] Step S21: Construct the physical constraints for the dual-control prediction of the tensile pile, specifically by constructing the physical constraint equation for tensile bearing capacity and the differential equation for pile deformation control;

[0017] The physical constraint equation for tensile bearing capacity is used to anchor the compliance boundary of bearing capacity prediction and set a legal compliance threshold for the model's bearing capacity prediction results. The formula used is as follows:

[0018] ;

[0019] In the formula, This represents the standard value of the pull-out bearing capacity of a single pile, where n represents the total number of soil layers traversed by the pile. This represents the reduction factor for the pull-out friction of the i-th soil layer. This represents the standard value of the lateral resistance of the i-th soil layer. Indicates the circumference of the pile. This represents the length of the pile shaft that penetrates the i-th layer of soil. This indicates the standard value of the pile's self-weight;

[0020] The differential equation controlling pile deformation is used to determine the mechanical rationality boundary for anchor deformation prediction, and is used to set the pile end displacement constraint boundary and the pile top force equilibrium boundary; the formulas used are as follows:

[0021] ;

[0022] In the formula, This indicates the elastic modulus of the pile concrete. Indicates the cross-sectional area of ​​the pile. This represents the axial upward displacement at pile depth z, where z represents the pile depth coordinate, and L represents the pile length. The unit frictional resistance of the pile side at depth z is represented by d, where d is the differential symbol.

[0023] Step S22: Establish the dual-control prediction model architecture for tension-resistant piles, specifically by designing an adaptive residual network architecture to establish the dual-control prediction model for tension-resistant piles; the adaptive residual network architecture includes an input layer, an adaptive residual layer, and an output layer;

[0024] The adaptive residual layer is specifically composed of R serially connected residual blocks and an adaptive weighting layer;

[0025] The residual block specifically comprises two fully connected layers and a tanh activation function, which sequentially perform nonlinear mapping operations;

[0026] The adaptive weighting layer specifically configures independent trainable adaptive weights for each residual block and fuses the input features of the residual block with the nonlinear mapping output through a complementary weighting strategy.

[0027] The output layer specifically receives the output features of the adaptive residual layer and outputs the tensile pile bearing capacity results and tensile deformation results synchronously through the linear fully connected layer.

[0028] Step S23: Design the total loss function of multi-constraint fusion. Specifically, firstly, construct a three-dimensional loss system consisting of the loss term for fitting the dual-control target data, the loss term for the physical constraint of bearing capacity, and the loss term for the residual of the deformation differential equation. Then, calculate the dynamic weight of each loss term based on the training iteration rounds. Finally, obtain the total loss function of multi-constraint fusion through weighted fusion.

[0029] The loss term for fitting the dual-control target data is specifically the mean square error between the predicted values ​​of the single pile pull-out bearing capacity and the pull-out deformation at the pile top and the measured values ​​of the field static load test, calculated based on the engineering samples within the training batch.

[0030] The residual loss term of the physical constraint of bearing capacity is specifically calculated based on the physical constraint equation of uplift bearing capacity, which calculates the residual mean square error between the model bearing capacity prediction value and the theoretical calculation value at the physical constraint sampling point.

[0031] The residual loss term of the deformation differential equation is specifically calculated based on the pile deformation control differential equation, the second derivative of the pile displacement predicted by the model is calculated, and the mean square error of the residual of the differential equation is solved.

[0032] The total loss function is specifically calculated by first calculating the training progress normalization factor, and then calculating the dynamic weight of the physical constraint loss term through the exponential scheduling function.

[0033] ;

[0034] ;

[0035] ;

[0036] In the formula, This indicates the current training iteration round, and s represents the weight update frequency. This indicates the total number of iterations in the model training. This indicates the floor function. This represents the normalization factor for the current training progress. and These represent the initial and final values ​​of the physical constraint loss weight for bearing capacity, respectively. and Let represent the initial and final values ​​of the residual loss weights in the deformed differential equation, respectively. This represents the weight of the physical constraint loss for the bearing capacity in the current training iteration. The residual loss weights of the deformed differential equation in the current training iteration are represented.

[0037] Step S24: Iterative model training. Specifically, the reference tensile pile optimization data in the tensile pile management optimization data is used as the training data of the model, and the total loss function is used as the optimization objective of the model training. The tensile pile dual control prediction model is trained, and finally the trained tensile pile dual control prediction model is obtained.

[0038] Furthermore, in step S3, the two-stage progressive parameter optimization of the tension pile specifically includes the following steps:

[0039] Step S31: Construct the parameter optimization algorithm, which specifically includes the following steps:

[0040] Step S311: Initialize the population of search individuals. Specifically, the parameters to be optimized are encoded into search individual position vectors, and M search individual position vectors are generated through random initialization to form the initial search individual population.

[0041] Step S312: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; the objective function value calculated based on the parameter combination corresponding to the position of the search individual, through the corresponding optimized objective function, is used as the fitness value of the search individual;

[0042] Step S313: Calculate the nonlinear control value, specifically by calculating the nonlinear control value for the current iteration number based on the ratio of the current iteration number to the total number of iterations using a logarithmic method; the formula used is as follows:

[0043] ;

[0044] In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the nonlinear control value in the t-th iteration;

[0045] Step S314: Calculate the exploration and development control parameters. Specifically, based on the ratio of the current iteration number to the total number of iterations, calculate the exploration and development control parameters for the current iteration number using exponential decay. The formula used is as follows:

[0046] ;

[0047] In the formula, This represents the exploration and development control parameter in the t-th iteration. This indicates the decay rate of the control parameters being explored and developed. This indicates the decay curvature of the control parameters being explored and developed;

[0048] Step S315: Individual position iterative update, specifically, for each search individual, according to the nonlinear adjustment value and search control parameters, the corresponding position update operation is performed according to the scenario to complete the position iterative update of the search individual;

[0049] Step S316: The search iteration is terminated. Specifically, when the global optimal position of the search individual is higher than the fitness threshold or the number of iterations reaches the maximum number of iterations, the search is terminated and the global optimal position of the search individual is obtained. The global optimal position of the search individual specifically refers to the optimal combination of parameters.

[0050] Step S32: Structural body parameter optimization. Specifically, the structural body parameter data is used as the parameters to be optimized. A parameter optimization algorithm is used for global optimization. During the parameter optimization process, the real-time tensile pile bearing capacity and tensile deformation results corresponding to each set of structural body parameters are obtained in real time through the trained tensile pile dual-control prediction model. Based on the real-time tensile pile bearing capacity and tensile deformation results, the structural body parameter optimization objective function is designed. Finally, the optimal combination of structural parameters is obtained through iterative optimization of the parameter optimization algorithm.

[0051] The objective function for optimizing the structural body parameters is to maximize the real-time tensile pile bearing capacity while ensuring that the real-time tensile deformation is within the allowable range as a rigid constraint.

[0052] Step S33: Construction process parameter optimization. Specifically, the construction process parameters are used as the parameters to be optimized. A parameter optimization algorithm is used for global optimization. During the parameter optimization process, based on the construction process parameters corresponding to the individual search location, the optimal combination of structural parameters and the inherent site geological parameters, site environmental parameters, and pile-soil interface parameters in the target tensile pile optimization data are input into the trained tensile pile dual-control prediction model. The real-time tensile pile bearing capacity result and the real-time pile top tensile deformation prediction result corresponding to each set of construction process parameters are output. A rigid constraint function is constructed using the real-time tensile pile bearing capacity result and the real-time tensile deformation result. At the same time, the construction process optimization objective function is designed. Finally, through iterative optimization of the parameter optimization algorithm, the optimal combination of construction process parameters that meets the constraint conditions and has the optimal objective function is selected, and the optimal combination of construction process parameters is obtained.

[0053] The rigid constraint function is specifically set based on the real-time tensile pile bearing capacity result and the real-time tensile deformation result. First, the real-time tensile pile bearing capacity result must be greater than or equal to the optimal tensile pile bearing capacity result corresponding to the optimal combination of structural parameters in the structural body parameter optimization. At the same time, the real-time tensile deformation result is controlled within the preset allowable range as another rigid constraint condition.

[0054] The objective function for optimizing the construction process is specifically a dual-objective comprehensive optimization function that integrates minimizing construction resource consumption and maximizing construction efficiency. The weighted summation method is used to transform the dual objectives into a single objective for iterative optimization.

[0055] Furthermore, in step S4, the intelligent optimization of the tensile pile bearing capacity is used to precisely and intelligently regulate and optimize the bearing capacity based on the optimal combination of structural parameters and the optimal combination of construction process parameters, combined with the real-time feedback of the tensile pile dual-control prediction model. Specifically, based on the optimal combination of structural parameters and the optimal combination of construction process parameters, a final optimized construction parameter scheme for the tensile pile bearing capacity is formed, ultimately achieving safe and controllable tensile pile bearing capacity and optimal performance.

[0056] The beneficial effects achieved by the present invention using the above solution are as follows:

[0057] (1) In the traditional method of optimizing the bearing capacity of tensile piles, the combined optimization mode of structural body parameters and construction process parameters is generally adopted. During the optimization process, the bearing performance target and the construction efficiency target compete and restrict each other in a disorderly manner, resulting in technical problems such as insufficient tensile pile bearing capacity and low construction efficiency in the optimized tensile pile bearing capacity scheme. This scheme innovatively proposes to carry out the optimization of structural body parameters and construction process parameters in two progressive stages. First, the optimization of structural body parameters is carried out with the maximization of the ultimate bearing capacity of tensile piles as the core optimization target and the compliance of tensile deformation at the pile top as a rigid constraint. Then, the optimal combination of structural parameters is completely fixed and the optimization of construction process parameters is carried out with the maximization of construction process efficiency as the target. This improves the stability and safety redundancy of the core bearing performance of tensile piles, enhances the smoothness of construction process and on-site construction efficiency, and strengthens the performance controllability and on-site adaptability of the optimization scheme. Finally, it achieves the compatibility and synergy of the optimal bearing performance of tensile piles and the maximization of construction efficiency, and completes the intelligent optimization of the entire process of tensile pile bearing capacity.

[0058] (2) In view of the fact that the traditional method for optimizing the bearing capacity of pull-out piles generally adopts a single-objective independent prediction mode, without introducing physical constraints that comply with the engineering specifications for pull-out piles for rigid anchoring, there are technical problems such as insufficient accuracy of the pull-out bearing capacity prediction results and the prediction results easily exceeding the compliance boundary of the specifications. As a result, the optimized pull-out pile scheme has the consequences of insufficient bearing capacity, excessive deformation, and inherent safety hazards. This scheme innovatively proposes an adaptive residual network dual-control synchronous prediction method with dual physical constraints. It constructs a bearing capacity physical constraint equation that conforms to the current pile foundation design specifications and an elastic mechanical pile deformation control. The differential equations anchored the regulatory compliance boundary and the mechanical rationality boundary for the simultaneous prediction of the ultimate bearing capacity and the tensile deformation at the pile top of the tension pile, respectively. This improved the synchronicity, accuracy, and stability of the dual-objective prediction of the ultimate bearing capacity and the tensile deformation at the pile top of the tension pile. It also comprehensively enhanced the cross-scenario generalization and adaptability of the prediction method to different site geological conditions, different pile structures, and combinations of process parameters. Ultimately, it achieved the synchronous, accurate, and compliant prediction of the ultimate bearing capacity and the tensile deformation at the pile top of the tension pile under any combination of structure and process parameters, ensuring that the bearing performance of the optimized tension pile scheme meets the standards and the deformation is controllable.

[0059] (3) In view of the technical problems of gradient vanishing, unstable training convergence and low prediction accuracy in the existing prediction models for tensile piles, this solution innovatively designs an adaptive residual network architecture, configures independent trainable adaptive weights for each serially connected residual block, and designs a multi-constraint dynamic weight total loss function that integrates the data fitting loss of dual control targets, the residual loss of bearing capacity physical constraints and the residual loss of deformed differential equations. Through the dynamic weight strategy that depends on the iteration round, the loss contribution of data fitting and physical constraints is dynamically balanced, which effectively solves the problems of inconsistent prediction accuracy of dual targets, unstable convergence and vanishing gradient of physical constraints in the training process of traditional fixed weight models, and comprehensively improves the prediction accuracy, convergence stability and cross-scenario generalization performance of the tensile pile dual control prediction model.

[0060] (4) In view of the existing algorithms applicable to the optimization of the bearing capacity of tensile piles, in the scenario of multi-parameter optimization of tensile piles, the global exploration and local development capabilities cannot be dynamically adapted with iteration, which easily leads to premature convergence, getting trapped in local optima, and insufficient optimization accuracy and stability. As a result, the potential for improving the performance of tensile piles is not fully explored. This scheme innovatively proposes an improved optimization algorithm with a dual nonlinear dynamic control mechanism. It constructs a logarithmic nonlinear control value and an exponential decay type exploration and development control parameter, which greatly improves the algorithm's global traversal efficiency, optimization convergence speed and optimal solution accuracy in high-dimensional parameter space. It also improves the integrity and comprehensiveness of multi-parameter combination optimization of tensile piles, enhances the algorithm's adaptability to different dimensions of structural body parameters and construction process parameters, and finally obtains the optimal parameter combination of tensile pile structural body parameters and construction process parameters, ensuring that the performance and efficiency indicators of the final output optimization scheme reach the global optimal level. Attached Figure Description

[0061] Figure 1 A flowchart illustrating the big data-based method for optimizing the bearing capacity of pull-out piles provided by this invention;

[0062] Figure 2 This is a flowchart illustrating step S2;

[0063] Figure 3 This is a flowchart illustrating step S3;

[0064] Figure 4 This is a flowchart illustrating step S31;

[0065] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0066] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0067] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0068] Example 1, see Figure 1 The technical solution adopted by this invention is as follows: The method for optimizing the bearing capacity of pull-out piles based on big data provided by this invention includes the following steps:

[0069] Step S1: Multi-source data acquisition, specifically, obtaining optimized management data for tension piles through data collection and preprocessing;

[0070] Step S2: Dual-control prediction of tensile piles. Specifically, firstly, physical constraints for dual-control prediction of tensile piles are constructed. Then, by designing an adaptive residual network architecture, a dual-control prediction model for tensile piles that simultaneously predicts bearing capacity and tensile deformation is established. Next, a multi-constraint dynamic weight total loss function is designed, which integrates the fitting loss of dual-control target data, the residual loss of bearing capacity physical constraints, and the residual loss of deformation differential equations. Finally, the reference tensile pile optimization data in the tensile pile management optimization data is used as the training data of the model. The total loss function is used as the optimization objective for model training. The model is iteratively trained to obtain the trained dual-control prediction model for tensile piles.

[0071] Step S3: Two-stage progressive parameter optimization of tension piles. Specifically, the optimization algorithm is first improved through a dual nonlinear dynamic control mechanism to construct a parameter optimization algorithm. Then, based on the parameter optimization algorithm and the trained tension pile dual-control prediction model, structural body parameter optimization and construction process parameter optimization are carried out in two progressive stages. Finally, the optimal combination of structural parameters and the optimal combination of construction process parameters are obtained.

[0072] The specific dual nonlinear dynamic control mechanism involves constructing a logarithmic nonlinear control value and designing an exponentially decaying exploratory development control parameter.

[0073] Step S4: Intelligent optimization of tensile pile bearing capacity, specifically, based on the optimal combination of structural parameters and the optimal combination of construction process parameters, to form the final optimized construction parameter scheme for tensile pile bearing capacity.

[0074] By performing the above operations, this solution addresses the technical problems of insufficient tensile pile bearing capacity and low construction efficiency in traditional tensile pile bearing capacity optimization methods, which commonly employ a mixed optimization mode combining structural parameters and construction process parameters. This optimization process often results in disordered competition and mutual constraints between bearing capacity and construction efficiency objectives, leading to these issues. The innovative solution proposes a two-stage progressive approach to optimizing structural parameters and construction process parameters. First, structural parameters are optimized with the maximization of the ultimate tensile pile bearing capacity as the core optimization objective and compliance with pile top tensile deformation as a rigid constraint. Then, the optimal combination of structural parameters is completely fixed, and construction process parameters are optimized with the goal of maximizing construction efficiency. This improves the stability and safety redundancy of the core tensile pile bearing capacity, enhances the smoothness of construction procedures and on-site construction efficiency, strengthens the performance controllability and on-site adaptability of the optimization scheme, and ultimately achieves a compatible and synergistic balance between optimal tensile pile bearing capacity and maximized construction efficiency, completing the intelligent optimization of the entire tensile pile bearing capacity process.

[0075] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the multi-source data acquisition specifically involves collecting data information on historical anti-tension pile engineering cases, target project site conditions, and pile-soil interaction system to obtain original anti-tension pile management data, and performing data preprocessing on the original anti-tension pile management data to obtain optimized anti-tension pile management data.

[0076] The raw data for tension pile management includes reference tension pile optimization data and target tension pile optimization data;

[0077] The reference anti-tension pile optimization data and target anti-tension pile optimization data include structural body parameter data, anti-tension pile construction process parameter data, site geological parameter data, site environmental parameter data, and pile-soil interface parameter data;

[0078] The structural body parameter data are used to determine the skeleton parameters of the tensile pile foundation's tensile strength and deformation characteristics, including the tensile pile diameter, pile length, pile body concrete strength grade, pile end form, and pile body cross-sectional form.

[0079] The construction process parameters of the tension piles are used to minimize the consumption of construction resources and maximize the efficiency of construction procedures under safety constraints; these include pile driving method, pile side grouting pressure, pile side grouting volume, and longitudinal reinforcement ratio of the pile body.

[0080] The site geological parameter data include the natural unit weight of the soil layer, the void ratio of the soil layer, the liquid limit index of the soil layer, the plasticity index of the soil layer, the consolidation coefficient of the soil layer, and the horizontal resistance coefficient of the soil around the pile.

[0081] The site environmental parameter data include the site lateral earth pressure coefficient, groundwater buoyancy influence coefficient, site frost heave influence depth, and site soil anisotropy coefficient.

[0082] The pile-soil interface parameter data includes the pile-soil contact surface friction characteristic coefficient, pile-soil bond strength, pile-soil relative stiffness ratio, and pile-soil liquefaction sensitivity index.

[0083] The reference tensile pile optimization data also includes the reference tensile pile bearing capacity results and the reference tensile deformation results;

[0084] The tensile bearing capacity result is specifically the tensile bearing capacity value of a single pile, in kN, which is a deterministic value directly determined through on-site static load test.

[0085] The reference pull-out deformation result is specifically the measured value of the vertical pull-out displacement of the pile top under the ultimate load, in mm, and is a single-point measured value;

[0086] The data preprocessing is used to remove noise and redundant information from the original data, and to eliminate the interference of dimensional differences and category features on subsequent model training and parameter optimization; specifically, data cleaning and normalization processing are performed sequentially on the original data of anti-uplift pile management to obtain optimized data for anti-uplift pile management.

[0087] The data cleaning is used to repair missing data and eliminate the impact of outlier extreme values ​​on the accuracy of analysis; specifically, it involves imputing missing values ​​and removing outliers from the original data; the imputation of missing values ​​is specifically to fill in missing values ​​using the mean imputation method; the removal of outliers is specifically to detect and remove extreme values ​​in the original data using the interquartile range method.

[0088] The normalization process is used to unify the data scale of different features and transform discrete category features into numerical forms that the model can recognize. Specifically, it first uses a label encoding method to map the category fields in the original data to corresponding integer values ​​according to the category values, and then uses the Z-Score standardization method to standardize all continuous variables so that each continuous variable meets a unified numerical scale range.

[0089] Example 3, see Figure 1 and Figure 2This embodiment is based on the above embodiment. In step S2, the dual-control prediction of the pull-out pile is used to simultaneously and accurately predict the ultimate pull-out bearing capacity and the pull-out deformation at the pile top under any combination of parameters, providing a unique and stable compliance verification standard for subsequent two-level parameter optimization; specifically, it includes the following steps:

[0090] Step S21: Construct physical constraints for the dual-control prediction of tensile piles. This is used to set geotechnical mechanical constraint boundaries that conform to the current pile foundation design code for the synchronous prediction of the ultimate tensile bearing capacity and tensile deformation of the pile top. This solves the technical problems of existing pure data-driven prediction models having no physical anchoring, prediction results that violate the laws of engineering mechanics, poor generalization performance, and inability to meet the requirements of dual-control design in engineering. Specifically, this involves constructing the physical constraint equation for tensile bearing capacity and the differential equation for pile deformation control.

[0091] The physical constraint equation for tensile bearing capacity is used to anchor the compliance boundary of bearing capacity prediction, set a legal compliance threshold for the model's bearing capacity prediction results, and ensure that the bearing capacity prediction value output by the model meets the mandatory requirements of the current pile foundation design code; the formula used is as follows:

[0092] ;

[0093] In the formula, This represents the standard value of the pull-out bearing capacity of a single pile, where n represents the total number of soil layers traversed by the pile. This represents the reduction factor for the pull-out friction of the i-th soil layer. This represents the standard value of the lateral resistance of the i-th soil layer, in kPa. This indicates the circumference of the pile, in meters (m). This represents the length of the pile shaft penetrating the i-th layer of soil, in meters. This represents the standard value of the pile's self-weight, in kN, calculated based on the buoyancy of the pile concrete under pull-out conditions.

[0094] The differential equation controlling pile deformation is used to anchor the mechanical rationality boundary of deformation prediction, set continuous static equilibrium constraints for the model's deformation prediction results, ensure that the pile deformation distribution output by the model conforms to the basic principles of elasticity, and set corresponding pile end displacement constraint boundaries. Equilibrium boundary of pile top force The formula used is as follows:

[0095] ;

[0096] In the formula, This represents the elastic modulus of the pile concrete, expressed in kPa. This represents the cross-sectional area of ​​the pile, in m². This represents the axial upward displacement at pile depth z, where z represents the pile depth coordinate, with the centroid of the pile top section as the origin and the vertical downward direction along the pile axis as the positive direction. L represents the pile length in meters. This represents the unit skin friction resistance at depth z, expressed in kPa, derived from the standard value of the side resistance of the i-th soil layer. The value d is calculated based on the characteristics of pile depth distribution, where d is the differential sign.

[0097] Step S22: Establish a dual-control prediction model architecture for tension piles. This architecture is used to construct a neural network architecture that adapts to the requirements of multi-feature input of tension piles, strong nonlinear pile-soil action, and synchronous prediction of dual control targets. It solves the technical problems of gradient vanishing, loss of key features in shallow engineering, and coupling interference in dual-target prediction in traditional fully connected networks. Specifically, a dual-control prediction model for tension piles is established by designing an adaptive residual network architecture. The adaptive residual network architecture includes an input layer, an adaptive residual layer, and an output layer.

[0098] The input layer is used to ensure the effective transmission and preliminary integration of multi-source features of the tension piles; specifically, it receives tension pile management optimization data and converts it into feature dimensions that can be processed by a neural network through a linear mapping layer.

[0099] The adaptive residual layer is specifically composed of R serially connected residual blocks and an adaptive weighting layer;

[0100] The residual block is used to realize the nonlinear deep mapping of the multi-source characteristics of the pull-out pile, and to explore the complex relationship between features such as pile diameter, soil parameters, and pile-soil friction characteristics; it contains two fully connected layers and a tanh activation function, which sequentially perform nonlinear mapping operations;

[0101] The adaptive weighting layer is used to dynamically adjust the feature contribution weights of each residual block, adapting to the differences in the importance of key features under different tensile pile scenarios, and strengthening the cross-layer transmission of core engineering features such as pile side friction and pile stiffness. Specifically, each residual block is configured with independent trainable adaptive weights, and the input features of the residual block are fused with the nonlinear mapping output through a complementary weighting strategy; the formula used is as follows:

[0102] ;

[0103] ;

[0104] ;

[0105] In the formula, Let represent the input feature vector of the r-th residual block. and These are the trainable weight matrices for the first and second fully connected layers of the residual block, respectively. and These are represented as the trainable bias vectors for the first and second fully connected layers, respectively. This represents the intermediate output feature vector between the first fully connected layer and the activation function. This represents the nonlinear mapping characteristic of the r-th residual block. This represents the trainable adaptive weights of the r-th residual block. The output features of the adaptive weighted layer are used as the output of the current residual block and the final output of the adaptive residual layer.

[0106] The output layer is used to directly map the deep features extracted by the adaptive residual layer into engineering physical quantities, avoiding the interference of dual-objective prediction coupling and ensuring the engineering validity of the prediction results; specifically, it receives the output features of the adaptive residual layer and outputs the tensile pile bearing capacity results and tensile deformation results simultaneously through a linear fully connected layer.

[0107] Step S23: Design a multi-constraint fusion total loss function to provide a unified training optimization objective for the constructed adaptive residual network model. This addresses the technical problems of imbalance between data fitting and physical constraints, and the trade-off between bearing capacity and deformation prediction accuracy in dual-control prediction of uplift piles. Through a dynamic weighting strategy dependent on iterative iterations, the contribution of each loss term is dynamically balanced, resolving the shortcomings of traditional fixed-weight models such as unstable training, accuracy imbalance, and prediction results deviating from engineering specifications. Specifically, a three-dimensional loss system is first constructed, comprising the dual-control target data fitting loss term, the bearing capacity physical constraint residual loss term, and the deformation differential equation residual loss term. Then, the dynamic weights of each loss term are calculated based on the training iterations. Finally, the multi-constraint fusion total loss function is obtained through weighted fusion.

[0108] The dual-control target data fitting loss term is used to measure the deviation between the model's predicted values ​​and the actual measured values ​​in the project. Specifically, based on the project samples within the training batch, it calculates the mean square error between the predicted values ​​of single pile pull-out bearing capacity and pile top pull-out deformation and the actual measured values ​​from the field static load test. The formula used is as follows:

[0109] ;

[0110] In the formula, This represents the fitting loss value of the dual-control target data. This represents the total number of samples in a single training batch. This represents the model's predicted value of the single pile pull-out bearing capacity for the j-th sample. This represents the measured true value of the pull-out bearing capacity of a single pile in the j-th sample. This represents the model's predicted value of the pile top pull-out deformation for the j-th sample. This represents the measured true value of the pile top pull-out displacement under the load corresponding to the j-th sample;

[0111] The physical constraint residual loss term for bearing capacity is used to measure the deviation between the model's predicted bearing capacity value and the theoretical value of the legally prescribed formula in the current pile foundation design code, thus anchoring the compliance boundary of the bearing capacity prediction. Specifically, based on the physical constraint equation for tensile bearing capacity, it calculates the mean square error of the residual between the model's predicted bearing capacity value and the theoretically calculated value in the code at the physical constraint sampling points; the formula used is as follows:

[0112] ;

[0113] In the formula, This represents the residual loss value of the physical constraint on bearing capacity. This indicates the number of physical constraint sampling points for bearing capacity in a single training batch, which is consistent with the sample size of the training batch. This represents the model's predicted value of the single pile pull-out bearing capacity at the k-th sampling point;

[0114] The residual loss term of the deformation differential equation is used to measure the deviation between the pile deformation distribution predicted by the model and the axial equilibrium differential equation of elasticity, and to anchor the mechanical rationality boundary of the deformation prediction; specifically, based on the pile deformation control differential equation, the second derivative of the pile displacement predicted by the model is calculated, and the mean square error of the residual of the differential equation is solved; the formula used is as follows:

[0115] ;

[0116] In the formula, This represents the residual loss value of the deformed differential equation. This represents the number of sampling points for the residuals of the deformed differential equations in a single training batch. Sampling is conducted at equal intervals along the pile axis, with no fewer than 20 sampling points per pile. This indicates the pile depth predicted by the model. The axial upward displacement value at the location;

[0117] The total loss function is used to integrate three major constraint objectives: data fitting accuracy, standard compliance, and mechanical rationality. Specifically, it first calculates the training progress normalization factor, and then calculates the dynamic weight of the physical constraint loss term through the exponential scheduling function.

[0118] ;

[0119] ;

[0120] ;

[0121] ;

[0122] In the formula, This indicates the current training iteration round, and s represents the weight update frequency, meaning the dynamic weights are updated once every s iteration rounds. This indicates the total number of iterations in the model training. This indicates the floor function. This represents the normalization factor for the current training progress, and its value ranges from [value range missing]. , and These represent the initial and final values ​​of the physical constraint loss weight for bearing capacity, respectively. and Let represent the initial and final values ​​of the residual loss weights in the deformed differential equation, respectively. This represents the weight of the physical constraint loss for the bearing capacity in the current training iteration, and its value range is... , This represents the residual loss weight of the deformed differential equation in the current training iteration, with a value range of [value range missing]. ;

[0123] Step S24: Iterative model training, used to provide an efficient and stable parameter update strategy for the adaptive residual network model, solves the technical problems of slow convergence speed, easy getting trapped in local optima, vanishing physical constraint gradients, and poor generalization performance of traditional first-order optimizers in the scenario of multi-constraint dual-control prediction of tensile piles. At the same time, it works in deep collaboration with the dynamic weight loss function to ensure that the model training meets the requirements of engineering measurement accuracy and physical constraint compliance. Specifically, the reference tensile pile optimization data in the tensile pile management optimization data is used as the training data of the model, and the total loss function is used as the optimization objective of the model training to train the tensile pile dual-control prediction model, and finally the trained tensile pile dual-control prediction model is obtained.

[0124] The training of the anti-uplift pile dual-control prediction model specifically involves calculating the gradient of the multi-constraint fusion total loss function with respect to all trainable parameters of the model through the backpropagation algorithm, and iteratively updating the weight matrix and bias parameters of the anti-uplift pile dual-control prediction model by combining the gradient descent optimization method based on second-order preconditions. The model parameters are continuously optimized through multiple iterations. When the preset maximum number of training times is reached or the total loss function converges to a set threshold, the iterative training is stopped.

[0125] The gradient descent optimization method based on second-order preconditions is specifically the L-BFGS optimization algorithm.

[0126] By performing the above operations, this solution addresses the technical problems inherent in traditional methods for optimizing the bearing capacity of uplift piles. These methods typically employ a single-objective independent prediction model, failing to introduce physical constraints conforming to the engineering specifications for uplift piles for rigid anchoring. This results in insufficient accuracy in the predicted bearing capacity and a tendency for the predicted results to exceed the compliance boundaries of the specifications, leading to insufficient bearing capacity, excessive deformation, and inherent safety hazards in the optimized uplift pile scheme. This solution innovatively proposes a dual-control synchronous prediction method with adaptive residual networks and dual physical constraints. It constructs a bearing capacity physical constraint equation and an elastic mechanical pile deformation control differential equation that conform to current pile foundation design specifications. This anchors the compliance boundaries of the specifications and the mechanical rationality boundaries for the synchronous prediction of the ultimate bearing capacity and the uplift deformation at the pile top, respectively. This improves the synchronicity, accuracy, and stability of the dual-objective prediction of the ultimate bearing capacity and the uplift deformation at the pile top, comprehensively enhancing the prediction method's cross-scenario generalization and adaptation to different site geological conditions, pile types, and combinations of process parameters. This solution enables simultaneous, accurate, and compliant prediction of the ultimate bearing capacity and top tensile deformation of pull-out piles under arbitrary combinations of structural and process parameters, ensuring that the bearing performance of the optimized pull-out pile scheme meets the standards and that deformation is controllable. Addressing the technical problems of gradient vanishing, unstable training convergence, and low prediction accuracy in existing pull-out pile prediction models, this solution innovatively designs an adaptive residual network architecture. Each serially connected residual block is configured with an independent, trainable adaptive weight. Simultaneously, a multi-constraint dynamic weight total loss function is designed, integrating the data fitting loss of the dual-control objectives, the physical constraint residual loss of bearing capacity, and the residual loss of the deformation differential equation. Through an iterative round-dependent dynamic weight strategy, the loss contributions of data fitting and physical constraints are dynamically balanced, effectively solving the problems of inconsistent prediction accuracy, unstable convergence, and vanishing gradients of physical constraints in the training process of traditional fixed-weight models. This comprehensively improves the prediction accuracy, convergence stability, and cross-scenario generalization performance of the pull-out pile dual-control prediction model.

[0127] Example 4, see Figure 1 , Figure 3 and Figure 4 This embodiment is based on the above embodiment. In step S3, the two-stage progressive parameter optimization of the tension pile is used to achieve the optimal balance between the structural safety and construction technical indicators of the tension pile under the dual control safety constraints of the preset ultimate tensile bearing capacity and the tensile deformation at the pile top, based on the trained multi-output prediction model and strictly following the progressive logic of safety first and efficiency best. This eliminates safety hazards, resource waste, or inefficiency caused by the coupling of structural parameters and process parameters, and ensures that the final optimized solution has compliance, safety, economy, and engineering practicality. Specifically, it includes the following steps:

[0128] Step S31: Construct a parameter optimization algorithm to provide standardized and efficient optimization support for structural parameters and construction process parameters. This algorithm enables comprehensive traversal, safety verification, and optimal selection of both types of parameter combinations, ensuring that the two-level optimization process is logically rigorous and the results are reliable. Specifically, it involves obtaining the optimal parameter combination through an improved optimization algorithm, including the following steps:

[0129] Step S311: Initialize the population search individuals, which are used to encode the structural body parameters and construction process parameters into searchable position vectors of the algorithm, and generate an initial population covering the reasonable search space of the two types of parameters; specifically, the parameters to be optimized are encoded into search individual position vectors, and M search individual position vectors are generated by random initialization to form the initial search individual population;

[0130] The parameters to be optimized are specifically structural parameters and construction process parameters;

[0131] Step S312: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; the objective function value calculated based on the parameter combination corresponding to the position of the search individual, through the corresponding optimized objective function, is used as the fitness value of the search individual;

[0132] The corresponding optimization objective functions include the structural body parameter optimization objective function and the construction process optimization objective function;

[0133] Step S313: Calculate the nonlinear control value, which is used to dynamically adapt the triggering logic of the local development strategy during the algorithm iteration process, reasonably control the execution ratio of the local development strategy in different iteration stages, strengthen the global exploration capability in the early stage of algorithm iteration, and effectively suppress the premature convergence problem of the algorithm. Specifically, based on the ratio of the current iteration number to the total number of iterations, the nonlinear control value of the current iteration number is calculated by logarithmically. The formula used is as follows:

[0134] ;

[0135] In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the nonlinear control value in the t-th iteration;

[0136] Step S314: Calculate the exploration-exploitation control parameter, which is used to dynamically adjust the allocation ratio of global exploration and local exploitation capabilities in the algorithm. Specifically, based on the ratio of the current iteration number to the total number of iterations, the exploration-exploitation control parameter for the current iteration number is calculated using exponential decay. The formula used is as follows:

[0137] ;

[0138] In the formula, This represents the exploration and development control parameter in the t-th iteration. This indicates the decay rate of the control parameter being explored and developed; it is a preset fixed positive value used for control. The decay rate as the iteration process progresses, This indicates the decay curvature of the control parameter being explored and developed; it is a preset fixed positive value used for control. The shape of the decay curve;

[0139] Step S315: Individual position iterative update. This step involves using the coordinated determination of nonlinear control values ​​and search control parameters to execute position update logic according to different scenarios, enabling precise switching between global exploration and local development of the algorithm. Specifically, for each search individual, based on the nonlinear control values ​​and search control parameters, the corresponding position update operation is executed according to the different scenarios to complete the iterative update of the search individual's position. The formula used is as follows:

[0140] ;

[0141] ;

[0142] ;

[0143] ;

[0144] In the formula, Indicates search control parameters. This represents the distance between the currently searched individual and a randomly selected search individual. This represents the distance between the current searched individual and the optimal searched individual. Indicates a random search for individuals. This indicates that the population is searching for the globally optimal position of an individual in the t-th iteration. and Indicates being between Random numbers uniformly distributed within a range This represents the position of the i-th individual in the t-th generation of the population. This represents the position of the i-th individual in the (t+1)-th generation of the population. Indicates being between Random numbers uniformly distributed within a range This represents the parameter that controls the shape of the spiral search, and its value range is... , This represents the random spiral parameter, with a range of values. ;

[0145] Step S316: The search iteration is terminated. Specifically, when the global optimal position of the search individual is higher than the fitness threshold or the number of iterations reaches the maximum number of iterations, the search is terminated and the global optimal position of the search individual is obtained. The global optimal position of the search individual specifically refers to the optimal combination of parameters.

[0146] The globally optimal position of the search individual is specifically the position with the best fitness value selected from the position vectors of all search individuals in the population according to the optimization objective function during all iterations from the initial iteration of the algorithm to the current t-th iteration.

[0147] Step S32: Structural body parameter optimization, which aims to maximize the ultimate tensile bearing capacity by optimizing the structural skeleton design under the premise of eliminating construction process interference, while ensuring that the tensile deformation at the pile top is within the allowable range, and to determine the optimal parameter combination at the structural level. Specifically, the structural body parameter data is used as the parameters to be optimized, and a parameter optimization algorithm is used for global optimization. During the parameter optimization process, the real-time tensile pile bearing capacity result and real-time tensile deformation result corresponding to each set of structural body parameters are obtained in real time through the trained tensile pile dual-control prediction model. Based on the real-time tensile pile bearing capacity result and real-time tensile deformation result, the structural body parameter optimization objective function is designed. Finally, the optimal parameter combination of the structure is obtained through iterative optimization of the parameter optimization algorithm.

[0148] Specifically, to obtain the real-time tensile pile bearing capacity and real-time tensile deformation results, the construction process parameters are fixed to industry standard benchmark values, do not participate in this optimization, and remain constant throughout the process. Based on the structural body parameters corresponding to the search individual location, the fixed construction process parameters are combined with the inherent site geological parameters, site environmental parameters, and pile-soil interface parameters in the target tensile pile optimization data, and input into the trained tensile pile dual-control prediction model to output the real-time tensile pile bearing capacity and real-time tensile deformation results corresponding to the parameter combination.

[0149] The objective function for optimizing the structural body parameters is to maximize the real-time tensile pile bearing capacity while ensuring that the real-time tensile deformation is within the allowable range as a rigid constraint.

[0150] Step S33: Construction process parameter optimization. This step aims to improve construction efficiency, resource consumption, and tensile pile performance by optimizing construction process parameters while keeping the structural body parameters constant. This ensures structural safety while maintaining construction efficiency and operability, and further enhances the tensile pile bearing capacity. Specifically, the construction process parameters are used as the parameters to be optimized. A parameter optimization algorithm is used for global optimization. During the parameter optimization process, based on the construction process parameters corresponding to the search individual location, the optimal combination of structural parameters and the inherent site geological parameters, site environmental parameters, and pile-soil interface parameters in the target tensile pile optimization data are input into the trained tensile pile dual-control prediction model. The model outputs the real-time tensile pile bearing capacity result and the real-time pile top tensile deformation prediction result corresponding to each set of construction process parameters. A rigid constraint function is constructed using the real-time tensile pile bearing capacity result and the real-time tensile deformation result. At the same time, a construction process optimization objective function is designed. Finally, through iterative optimization of the parameter optimization algorithm, the optimal combination of construction process parameters that satisfies the constraint conditions and has the optimal objective function is selected, thus obtaining the optimal combination of construction process parameters.

[0151] The rigid constraint function is specifically set based on the real-time tensile pile bearing capacity result and the real-time tensile deformation result. First, the real-time tensile pile bearing capacity result must be greater than or equal to the optimal tensile pile bearing capacity result corresponding to the optimal combination of structural parameters in the structural body parameter optimization. At the same time, the real-time tensile deformation result is controlled within the preset allowable range as another rigid constraint condition.

[0152] The objective function for optimizing the construction process is specifically a dual-objective comprehensive optimization function that integrates minimizing construction resource consumption and maximizing construction efficiency. A weighted summation method is used to transform the dual objectives into a single-objective iterative optimization; the formula used is as follows:

[0153] ;

[0154] In the formula, This represents the objective function for optimizing the construction process. and These are the weighting coefficients for each item, and their ranges. between, The quantitative indicators representing construction resource consumption are jointly quantified by the pile side grouting volume, pile side grouting pressure, pile longitudinal reinforcement ratio, and the resources corresponding to the pile driving method. This represents a quantitative indicator of construction process efficiency, which is jointly quantified by the time spent driving piles and the rationality of grouting parameters.

[0155] By performing the above operations, this solution addresses the technical problems of existing algorithms applicable to the optimization of the bearing capacity of tensile piles. In the scenario of multi-parameter optimization of tensile piles, the global exploration and local development capabilities cannot be dynamically adapted with iteration, which easily leads to premature convergence, getting trapped in local optima, and insufficient optimization accuracy and stability. As a result, the potential for improving the performance of tensile piles is not fully explored. This solution innovatively proposes an improved optimization algorithm with a dual nonlinear dynamic control mechanism. It constructs a logarithmic nonlinear control value and an exponential decay type exploration and development control parameter, which greatly improves the algorithm's global traversal efficiency, optimization convergence speed, and optimal solution accuracy in high-dimensional parameter space. It also improves the completeness and comprehensiveness of multi-parameter combination optimization of tensile piles and enhances the algorithm's adaptability to different optimization needs of structural body parameters and construction process parameters. Finally, it obtains the optimal parameter combination of structural body parameters and construction process parameters of tensile piles, ensuring that the performance and efficiency indicators of the final output optimization scheme reach the global optimal level.

[0156] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the intelligent optimization of the tensile pile bearing capacity is used to accurately and intelligently regulate and optimize the bearing capacity based on the optimal combination of structural parameters and the optimal combination of construction process parameters, combined with the real-time feedback of the tensile pile dual-control prediction model. Specifically, based on the optimal combination of structural parameters and the optimal combination of construction process parameters, a final optimized construction parameter scheme for the tensile pile bearing capacity is formed, ultimately achieving safe and controllable tensile pile bearing capacity and optimal performance.

[0157] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0158] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0159] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for optimizing the bearing capacity of tension piles based on big data, characterized in that: The method includes the following steps: Step S1: Multi-source data acquisition, specifically, obtaining optimized management data for tension piles through data collection and preprocessing; Step S2: Dual-control prediction of tensile piles. Specifically, firstly, physical constraints for dual-control prediction of tensile piles are constructed. Then, by designing an adaptive residual network architecture, a dual-control prediction model for tensile piles that simultaneously predicts bearing capacity and tensile deformation is established. Next, a multi-constraint dynamic weight total loss function is designed, which integrates the fitting loss of dual-control target data, the residual loss of bearing capacity physical constraints, and the residual loss of deformation differential equations. Finally, the reference tensile pile optimization data in the tensile pile management optimization data is used as the training data of the model. The total loss function is used as the optimization objective for model training. The model is iteratively trained to obtain the trained dual-control prediction model for tensile piles. Step S3: Two-stage progressive parameter optimization of tension piles. Specifically, the optimization algorithm is first improved through a dual nonlinear dynamic control mechanism to construct a parameter optimization algorithm. Then, based on the parameter optimization algorithm and the trained tension pile dual-control prediction model, structural body parameter optimization and construction process parameter optimization are carried out in two progressive stages. Finally, the optimal combination of structural parameters and the optimal combination of construction process parameters are obtained. Step S4: Intelligent optimization of tensile pile bearing capacity, specifically, based on the optimal combination of structural parameters and the optimal combination of construction process parameters, to form the final optimized construction parameter scheme for tensile pile bearing capacity.

2. The method for optimizing the bearing capacity of tension piles based on big data according to claim 1, characterized in that: In step S2, the dual-control prediction of the pull-out pile specifically includes the following steps: Step S21: Construct the physical constraints for the dual-control prediction of the tensile pile, specifically by constructing the physical constraint equation for tensile bearing capacity and the differential equation for pile deformation control; The physical constraint equation for tensile bearing capacity is used to anchor the compliance boundary of bearing capacity prediction and set a legal compliance threshold for the model's bearing capacity prediction results. The formula used is as follows: ； In the formula, This represents the standard value of the pull-out bearing capacity of a single pile, where n represents the total number of soil layers traversed by the pile. This represents the reduction factor for the pull-out friction of the i-th soil layer. This represents the standard value of the lateral resistance of the i-th soil layer. Indicates the circumference of the pile. This represents the length of the pile shaft that penetrates the i-th layer of soil. This indicates the standard value of the pile's self-weight; The differential equation controlling pile deformation is used to determine the mechanical rationality boundary for anchor deformation prediction, and is used to set the pile end displacement constraint boundary and the pile top force equilibrium boundary; the formulas used are as follows: ； In the formula, This indicates the elastic modulus of the pile concrete. Indicates the cross-sectional area of ​​the pile. This represents the axial upward displacement at pile depth z, where z represents the pile depth coordinate, and L represents the pile length. The unit frictional resistance of the pile side at depth z is represented by d, where d is the differential symbol. Step S22: Establish the architecture of the dual-control prediction model for tension piles; Step S23: Design the total loss function for multi-constraint fusion; Step S24: Iterative model training. Specifically, the reference tensile pile optimization data in the tensile pile management optimization data is used as the training data of the model, and the total loss function is used as the optimization objective of the model training. The tensile pile dual control prediction model is trained, and finally the trained tensile pile dual control prediction model is obtained.

3. The method for optimizing the bearing capacity of tension piles based on big data according to claim 2, characterized in that: In step S22, the establishment of the anti-uplift pile dual-control prediction model architecture specifically involves establishing the anti-uplift pile dual-control prediction model by designing an adaptive residual network architecture; the adaptive residual network architecture includes an input layer, an adaptive residual layer, and an output layer; The adaptive residual layer is specifically composed of R serially connected residual blocks and an adaptive weighting layer; The residual block specifically comprises two fully connected layers and a tanh activation function, which sequentially perform nonlinear mapping operations; The adaptive weighting layer specifically configures independent trainable adaptive weights for each residual block and fuses the input features of the residual block with the nonlinear mapping output through a complementary weighting strategy. The output layer specifically receives the output features of the adaptive residual layer and synchronously outputs the tensile pile bearing capacity results and tensile deformation results through the linear fully connected layer.

4. The method for optimizing the bearing capacity of tension piles based on big data according to claim 2, characterized in that: In step S23, the design of the multi-constraint fusion total loss function specifically involves first constructing a three-dimensional loss system consisting of a loss term for fitting the dual-control target data, a loss term for the physical constraint of bearing capacity, and a loss term for the residual of the deformation differential equation; then calculating the dynamic weights of each loss term based on the training iterations; and finally obtaining the multi-constraint fusion total loss function through weighted fusion. The loss term for fitting the dual-control target data is specifically the mean square error between the predicted values ​​of the single pile pull-out bearing capacity and the pull-out deformation at the pile top and the measured values ​​of the field static load test, calculated based on the engineering samples within the training batch. The residual loss term of the physical constraint of bearing capacity is specifically calculated based on the physical constraint equation of uplift bearing capacity, which calculates the residual mean square error between the model bearing capacity prediction value and the theoretical calculation value at the physical constraint sampling point. The residual loss term of the deformation differential equation is specifically calculated based on the pile deformation control differential equation, the second derivative of the pile displacement predicted by the model is calculated, and the mean square error of the residual of the differential equation is solved. The total loss function is specifically calculated by first calculating the training progress normalization factor, and then calculating the dynamic weight of the physical constraint loss term through the exponential scheduling function. ； ； ； In the formula, This indicates the current training iteration round, and s represents the weight update frequency. This indicates the total number of iterations in the model training. This indicates the floor function. This represents the normalization factor for the current training progress. and These represent the initial and final values ​​of the physical constraint loss weight for bearing capacity, respectively. and Let represent the initial and final values ​​of the residual loss weights in the deformed differential equation, respectively. This represents the weight of the physical constraint loss for the bearing capacity in the current training iteration. This represents the residual loss weight of the deformed differential equation in the current training iteration.

5. The method for optimizing the bearing capacity of tension piles based on big data according to claim 1, characterized in that: In step S3, the two-stage progressive parameter optimization of the tension pile specifically includes the following steps: Step S31: Construct a parameter optimization algorithm; Step S32: Structural body parameter optimization. Specifically, the structural body parameter data is used as the parameters to be optimized. A parameter optimization algorithm is used for global optimization. During the parameter optimization process, the real-time tensile pile bearing capacity and tensile deformation results corresponding to each set of structural body parameters are obtained in real time through the trained tensile pile dual-control prediction model. Based on the real-time tensile pile bearing capacity and tensile deformation results, the structural body parameter optimization objective function is designed. Finally, the optimal combination of structural parameters is obtained through iterative optimization of the parameter optimization algorithm. The objective function for optimizing the structural body parameters is to maximize the real-time tensile pile bearing capacity while ensuring that the real-time tensile deformation is within the allowable range as a rigid constraint. Step S33: Construction process parameter optimization. Specifically, the construction process parameters are used as the parameters to be optimized. A parameter optimization algorithm is used for global optimization. During the parameter optimization process, based on the construction process parameters corresponding to the individual search location, the optimal combination of structural parameters and the inherent site geological parameters, site environmental parameters, and pile-soil interface parameters in the target tensile pile optimization data are input into the trained tensile pile dual-control prediction model. The real-time tensile pile bearing capacity result and the real-time pile top tensile deformation prediction result corresponding to each set of construction process parameters are output. A rigid constraint function is constructed using the real-time tensile pile bearing capacity result and the real-time tensile deformation result. At the same time, the construction process optimization objective function is designed. Finally, through iterative optimization of the parameter optimization algorithm, the optimal combination of construction process parameters that meets the constraint conditions and has the optimal objective function is selected, and the optimal combination of construction process parameters is obtained. The rigid constraint function is specifically set based on the real-time tensile pile bearing capacity result and the real-time tensile deformation result. First, the real-time tensile pile bearing capacity result must be greater than or equal to the optimal tensile pile bearing capacity result corresponding to the optimal combination of structural parameters in the structural body parameter optimization. At the same time, the real-time tensile deformation result is controlled within the preset allowable range as another rigid constraint condition. The objective function for optimizing the construction process is specifically a dual-objective comprehensive optimization function that integrates minimizing construction resource consumption and maximizing construction efficiency. The weighted summation method is used to transform the dual objectives into a single objective for iterative optimization.

6. The method for optimizing the bearing capacity of tension piles based on big data according to claim 5, characterized in that: In step S31, the construction of the parameter optimization algorithm specifically includes the following steps: Step S311: Initialize the population of search individuals. Specifically, the parameters to be optimized are encoded into search individual position vectors, and M search individual position vectors are generated through random initialization to form the initial search individual population. Step S312: Calculate the individual fitness value, specifically by calculating the fitness value of the search individual in the population; the objective function value calculated based on the parameter combination corresponding to the position of the search individual, through the corresponding optimized objective function, is used as the fitness value of the search individual; Step S313: Calculate the nonlinear control value, specifically by calculating the nonlinear control value for the current iteration number based on the ratio of the current iteration number to the total number of iterations using a logarithmic method; the formula used is as follows: ； In the formula, t represents the current iteration number. Indicates the maximum number of iterations. This represents the nonlinear control value in the t-th iteration; Step S314: Calculate the exploration and development control parameters. Specifically, based on the ratio of the current iteration number to the total number of iterations, calculate the exploration and development control parameters for the current iteration number using exponential decay. The formula used is as follows: ； In the formula, This represents the exploration and development control parameter in the t-th iteration. This indicates the decay rate of the control parameters being explored and developed. This indicates the decay curvature of the control parameters being explored and developed; Step S315: Individual position iterative update, specifically, for each search individual, according to the nonlinear adjustment value and search control parameters, the corresponding position update operation is performed according to the scenario to complete the position iterative update of the search individual; Step S316: The search iteration is terminated. Specifically, when the global optimal position of the search individual is higher than the fitness threshold or the number of iterations reaches the maximum number of iterations, the search is terminated and the global optimal position of the search individual is obtained. The global optimal position of the search individual specifically refers to the optimal combination of parameters.

7. The method for optimizing the bearing capacity of tension piles based on big data according to claim 1, characterized in that: In step S4, the intelligent optimization of the tensile pile bearing capacity is used to precisely and intelligently regulate and optimize the bearing capacity based on the optimal combination of structural parameters and the optimal combination of construction process parameters, combined with the real-time feedback of the tensile pile dual-control prediction model. Specifically, based on the optimal combination of structural parameters and the optimal combination of construction process parameters, a final optimized construction parameter scheme for the tensile pile bearing capacity is formed, ultimately achieving safe and controllable tensile pile bearing capacity and optimal performance.

Images