Rock-socketed single pile horizontal bearing capacity prediction method and system for digital mechanics twinning

Abstract

The application discloses a rock-socketed single pile horizontal bearing capacity prediction method and system for digital mechanics twinning, relates to the single pile bearing capacity prediction technical field, and constructs a numerical sample library through FDEM numerical simulation, extracts mechanical state variables, and constructs a training set based on the pile top displacement, the pile top load and the mechanical state variables of each loading step; an agent model is constructed with the pile body design parameters, the rock mass design parameters and the target pile top displacement as inputs, the pile top load and the mechanical state variables as outputs, and the agent model is trained by using the training set, and the predicted value output by the agent model is made to approach the FDEM numerical simulation value by optimizing the loss function; according to the pile foundation design parameters to be predicted, the trained agent model is used to obtain the horizontal load-displacement response curve and the mechanical state variable evolution process. Both the prediction accuracy and the calculation efficiency are ensured, and good physical consistency and extrapolation capacity are also possessed.

Description

Technical Field

[0001] This invention relates to the field of single pile bearing capacity prediction technology, and in particular to a method and system for predicting the horizontal bearing capacity of rock-embedded single piles using digital mechanical twins. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Rock-embedded monopiles are widely used foundation types in major engineering projects such as bridges, offshore wind power, and high-rise buildings. Their horizontal bearing capacity directly determines the lateral stability and reliability of the superstructure. Due to the complex interaction mechanism between the pile and the surrounding rock mass, involving highly nonlinear mechanical behaviors such as nonlinear deformation of the rock mass, crack propagation, pile-rock interface slippage, and pile bending moment distribution, predicting the response of rock-embedded monopiles under horizontal loads is extremely challenging. Therefore, accurately predicting their horizontal bearing capacity and load-displacement response has become a crucial aspect of engineering design and safety assessment.

[0004] Currently, the main methods for predicting the horizontal bearing capacity of rock-socketed monopiles include traditional theoretical analytical methods, numerical simulation methods, and data-driven artificial intelligence methods.

[0005] Traditional theoretical methods are mostly based on the Winkler foundation model or the Py curve method. Although they are computationally efficient, they are difficult to accurately reflect the nonlinear damage and progressive failure process of rock masses, and their prediction accuracy is limited, especially under complex geological conditions.

[0006] Numerical simulation methods, such as the finite element method (FEM) and the finite element-discrete element coupled method (FDEM), can simulate the mechanical behavior of pile-rock systems with relatively high precision, but they are computationally expensive and complex to model, making it difficult to meet the needs of rapid, multi-condition iteration in engineering design.

[0007] In recent years, with the development of artificial intelligence technology, pure data-driven AI models (such as neural networks and support vector machines) have been introduced into load-bearing capacity prediction. Although they have strong nonlinear fitting capabilities, their prediction results often lack clear physical mechanism support and have a "black box" problem. They also have insufficient physical consistency in extrapolation prediction and cannot guarantee reliability and generalization ability under unknown working conditions. Summary of the Invention

[0008] To address the aforementioned issues, this invention proposes a method and system for predicting the horizontal bearing capacity of embedded rock piles using digital mechanical twins. This method integrates microscopic physical mechanisms with deep learning technology. By extracting key mechanical state variables from high-fidelity FDEM simulations and constructing a surrogate model architecture that combines a physical kernel with residual learning, physical laws are embedded into the artificial intelligence model in a hard constraint manner. This approach ensures both prediction accuracy and computational efficiency, while also possessing good physical consistency and extrapolation capabilities, thus overcoming the shortcomings of existing methods in terms of accuracy, efficiency, and physical reliability.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] In a first aspect, the present invention provides a method for predicting the horizontal bearing capacity of a rock-socketed monopile using digital mechanical twins, comprising:

[0011] Through FDEM numerical simulation, a numerical sample library of horizontal loading of rock-socketed monopile with various working conditions is constructed. Mechanical state variables characterizing the physical state of the pile-rock system are extracted from the FDEM numerical simulation results. Thus, a training set is constructed based on the pile top displacement, pile top load and mechanical state variables of each loading step.

[0012] A surrogate model is constructed with pile design parameters, rock mass design parameters, and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The surrogate model is trained using a training set, and the predicted values ​​output by the surrogate model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function.

[0013] Based on the pile foundation design parameters to be predicted, a trained surrogate model is used to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables.

[0014] As an alternative implementation, the mechanical state variables include damage factor, plastic strain, and pile bending moment; after aligning the pile top displacement and pile top load of each loading step with the three mechanical state vectors, a structured state evolution sequence is constructed for each FDEM numerical simulation.

[0015] As an alternative implementation method, the process of extracting mechanical state variables includes:

[0016] Along the direction of the pile's embedment depth, the surrounding rock mass is divided into... A continuous control interval;

[0017] For each loading step, traverse all rock mass elements in the FDEM and determine the control interval based on the center coordinates of the rock mass elements;

[0018] Calculate the volume-weighted average of the field variables for all rock mass elements within each control interval. The field variables include damage factor and plastic strain. Also, extract the pile bending moment value at the corresponding depth for each control interval.

[0019] As an alternative implementation, the proxy model includes a physical kernel layer and a residual learning branch;

[0020] The physical kernel layer takes the pile design parameters and rock mass design parameters as inputs and outputs the physical prediction values ​​of the pile top load and the physical prediction values ​​of the pile bending moment distributed along the pile body.

[0021] The residual learning branch takes the pile design parameters, rock mass design parameters, target pile top displacement, and soil reaction force calculated from the physical core layer as inputs, and outputs the pile top load residual, pile body bending moment residual, damage factor prediction value, and plastic strain prediction value. The pile top load residual is the difference between the simulated pile top load value and the physical prediction value of the pile top load, and the pile body bending moment residual is the difference between the simulated pile body bending moment value and the physical prediction value of the pile body bending moment.

[0022] As an alternative implementation method, the loss function Loss is:

[0023] ;

[0024] in, , , and These are the predicted values ​​of the pile top load. Predicted bending moment of pile body Damage factor prediction value and plastic strain prediction values The mean squared error loss; , , , and These are the weighting coefficients; , , and These are the actual values ​​of pile top load, pile bending moment, damage factor, and plastic strain, respectively. This is a physical constraint penalty term.

[0025] As an alternative implementation method, physical constraint penalty item include:

[0026] Calculation of soil reaction force based on predicted pile bending moment ,and Satisfies the static equilibrium equations: ;

[0027] Damage factor predicted value The displacement increases monotonically with loading;

[0028] Damage factor predicted value Predicted bending moment of pile body or soil reaction force Negative correlation.

[0029] Secondly, the present invention provides a system for predicting the horizontal bearing capacity of a rock-embedded single pile for digital mechanical twinning, comprising:

[0030] The numerical simulation module is configured to construct a numerical sample library containing various working conditions of horizontal loading of rock-socketed monopile through FDEM numerical simulation, and extract mechanical state variables that characterize the physical state of the pile-rock system from the FDEM numerical simulation results, thereby constructing a training set based on the pile top displacement, pile top load and mechanical state variables of each loading step.

[0031] The model building and training module is configured to build a proxy model with pile design parameters, rock mass design parameters and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The proxy model is trained using a training set, and the predicted values ​​output by the proxy model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function.

[0032] The prediction module is configured to use a trained surrogate model based on the design parameters of the pile foundation to be predicted to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables.

[0033] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0034] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0035] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect.

[0036] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0037] This invention proposes a method and system for predicting the horizontal bearing capacity of embedded rock monopiles by integrating microscopic physical mechanisms and deep learning technology. Through FDEM numerical simulation, a numerical sample library of horizontal loading conditions for embedded rock monopiles is constructed. Mechanical state variables characterizing the physical state of the pile-rock system are extracted from the FDEM numerical simulation results. A training set is then built based on the pile top displacement, pile top load, and mechanical state variables for each loading step. A surrogate model architecture combining a physical kernel and residual learning is constructed, embedding physical laws into the artificial intelligence model in a hard constraint manner. The model takes pile design parameters, rock mass design parameters, and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. Furthermore, by optimizing the loss function, the predicted values ​​output by the surrogate model are made to approximate the FDEM numerical simulation values, fundamentally solving the problem of insufficient physical consistency in extrapolation predictions by purely data-driven AI models.

[0038] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0040] Figure 1 This is a flowchart of the method for predicting the horizontal bearing capacity of a rock-socketed monopile using digital mechanical twins, as provided in Embodiment 1 of the present invention.

[0041] Figure 2 This is a rendering of the prediction result in the case provided in Embodiment 1 of the present invention;

[0042] Figure 3 The image shows the FDEM simulation results in the case provided in Embodiment 1 of the present invention. Detailed Implementation

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0044] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0045] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms “comprising” and “including”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0046] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0047] Example 1

[0048] This embodiment provides a method for predicting the horizontal bearing capacity of a rock-embedded single pile using digital mechanical twins. Its core lies in constructing a fast, accurate, and physically reliable surrogate model through an innovative path of physical mechanism datafication and intelligent data modeling.

[0049] like Figure 1 As shown, it includes the following steps:

[0050] Through FDEM numerical simulation, a numerical sample library of horizontal loading of rock-socketed monopile with various working conditions is constructed. Mechanical state variables characterizing the physical state of the pile-rock system are extracted from the FDEM numerical simulation results. Thus, a training set is constructed based on the pile top displacement, pile top load and mechanical state variables of each loading step.

[0051] A surrogate model is constructed with pile design parameters, rock mass design parameters, and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The surrogate model is trained using a training set, and the predicted values ​​output by the surrogate model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function.

[0052] Based on the pile foundation design parameters to be predicted, a trained surrogate model is used to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables.

[0053] The method of this embodiment will be described in detail below.

[0054] S1: High-fidelity finite element-discrete element coupled (FDEM) numerical simulation and construction of numerical sample library.

[0055] Specifically:

[0056] A three-dimensional numerical model of a rock-embedded monopile was established using the finite element-discrete element coupling (FDEM) method to accurately simulate the rock cracking and fracturing process.

[0057] By systematically changing key design parameters such as pile diameter, rock embedment depth, rock elastic modulus, cohesion, and internal friction angle, a large number of numerical simulation samples covering different working conditions are generated.

[0058] A horizontal displacement load is applied to each numerical simulation sample to simulate the entire process from initial linearity and nonlinearity to failure. The horizontal load-displacement curve, pile internal forces, and stress, strain, and damage field data of the rock mass are recorded in detail to form a high-fidelity numerical sample library.

[0059] S2: Extract and calculate key mechanical state variables from FDEM simulation results to characterize the physical state of the pile-rock system. The aim is to refine massive, unstructured field data into low-dimensional, structured features with clear physical meaning.

[0060] Specifically:

[0061] 1. Spatial Zoning: Based on the pile-rock interaction mechanism, the surrounding rock mass is divided along the pile's embedment depth direction into zones. A continuous control interval; (for example, a control interval is defined as 0.5 times the pile diameter).

[0062] 2. Integral average of field variables:

[0063] For each loading step, traverse all rock mass elements in the FDEM and determine the control interval to which the rock mass element belongs based on the center coordinates of the rock mass element.

[0064] Then, the volume-weighted average of the field variables for all rock mass elements within each control interval is calculated; the field variables include the damage factor D and the plastic strain. ;

[0065] Thus, the damage factor vector distributed along the depth at each loading step is obtained. and plastic strain vector .

[0066] 3. Pile Response Extraction: At the corresponding depth of each control interval, extract the pile bending moment value M at the corresponding depth of each control interval to form a pile bending moment vector. .

[0067] 4. Construct a structured sequence: Align the pile top displacement U and pile top load H of each loading step with the three key mechanical state vectors mentioned above (the key mechanical state variables are the damage factor vector, plastic strain vector, and pile bending moment vector distributed along the depth direction), and finally construct a structured state evolution sequence for each simulation: The entire numerical sample library is thus transformed into a dataset that can be used for machine learning.

[0068] S3: Based on the paradigm of integrating physical mechanisms and artificial intelligence, a proxy model for embedding physical information is constructed. This proxy model is a deep learning network for embedding physical information. It adopts a multi-task architecture that combines physical kernel and residual learning. Its core idea is that the physical kernel provides the basic solution and AI learns the correction amount. The inputs are pile parameters, rock mass parameters and target displacement, and the outputs are pile top load and key mechanical state variables.

[0069] Specifically, it includes:

[0070] 1. Physical Core Layer: This layer receives pile design parameters and rock mass design parameters, and calculates preliminary physical prediction results based on a simplified physical model, including the predicted physical value of the pile top load. Physical prediction of pile bending moment distributed along the pile body .

[0071] This layer is a differentiable, parameterized, simplified physical model, such as a beam-spring model or a Py curve model based on the Winkler foundation assumption; it receives pile design parameters (such as diameter, rock embedment depth, and pile elastic modulus) and rock mass design parameters (such as rock mass elastic modulus, cohesion, and internal friction angle), and targets the pile top displacement. U A preliminary, physically reasonable prediction result is calculated, including the predicted physical value of the pile top load. Physical prediction of pile bending moment distributed along the pile body This layer ensures that the model can provide predictions that conform to physical laws, even in unknown regions.

[0072] 2. Residual Learning Branch:

[0073] This branch is a deep learning network, such as a multilayer perceptron, unique in that the learning objective is not the complete response, but rather the FDEM simulation values ​​(simulated values ​​of pile top load) used to learn the true response. Simulated bending moment of pile body The residual between the prediction results of the physical kernel layer and the pile top load is the pile top load residual. and pile body bending moment residual And at the same time learn key mechanical state variables;

[0074] Its inputs include pile design parameters, rock mass design parameters, target pile top displacement, and soil reaction force calculated from the physical core layer. This provides rich physical context information; the output is a residual. , And the predicted values ​​of mechanical state variables, i.e., the predicted values ​​of damage factors. and plastic strain prediction values .

[0075] 3. Multi-task output and fusion:

[0076] The final prediction output of the proxy model includes: predicted pile top load: Predicted bending moment of pile body: Simultaneously, it also outputs predicted values ​​of physical and mechanical state variables. and .

[0077] This design allows AI to learn only the deviations in complex physical processes, significantly reducing the learning difficulty and incorporating built-in physical consistency.

[0078] S4: Multi-objective loss function and physical constraint training; the surrogate model is trained using a numerical sample set, and the loss function is optimized to make the surrogate model's predicted values ​​approximate the FDEM simulated values.

[0079] Specifically:

[0080] The loss function, Loss, is a multi-task weighted loss, and its expression is:

[0081] ;

[0082] in, , , and These are the mean square error losses of the predicted values ​​of pile top load, pile bending moment, damage factor, and plastic strain, respectively. , , , and These are the weighting coefficients; , , and These are the actual values ​​of pile top load, pile bending moment, damage factor, and plastic strain, respectively. This is a physical constraint penalty term used to force the model to comply with deeper physical laws.

[0083] Among them, physical constraint penalty item At least including:

[0084] Equilibrium Constraints: Based on Predicted Values ​​of Pile Bending Moment Distribution Calculate soil reaction And check the predictions Does it satisfy the static equilibrium equations? .

[0085] Irreversible constraint: Forced damage factor prediction value The displacement increases monotonically with loading.

[0086] Forced damage factor prediction value Predicted bending moment of pile body or soil reaction force Negative correlation.

[0087] These constraints transform physical laws from soft constraints to hard constraints, ensuring the rationality of extrapolation predictions at the model architecture level.

[0088] S5: Input the design parameters of the rock-socketed single pile to be predicted (pile design parameters and rock mass design parameters) into the trained surrogate model to obtain its complete horizontal load-displacement response curve and the distribution and evolution process of damage factor, plastic strain and pile bending moment along the pile body.

[0089] Through the above steps, this embodiment successfully constructs a digital-mechanical twin prediction method that combines physical reliability and computational efficiency. The following detailed description of these steps, using a basic implementation case, further illustrates the entire process of this method. This case aims to clearly demonstrate the entire process, therefore using laboratory-scale model parameters; however, its principles and methods are applicable to full-scale engineering applications.

[0090] Case objective: To predict the horizontal load-displacement (HU) response curve and internal mechanical state variables of a rock-embedded monopile of a specific size under given rock design parameters.

[0091] S1: FDEM numerical simulation and data generation.

[0092] 1. Model parameter settings:

[0093] Pile design parameters: Pile diameter d = 0.1m, rock embedding depth Elastic modulus of pile .

[0094] Rock mass design parameters: Brittle rock parameters are adopted, including the rock mass elastic modulus. Cohesion c = 1MPa, internal friction angle 40°.

[0095] Loading: Apply horizontal displacement to the top of the pile until... U = 10mm.

[0096] 2. Construct a sample library:

[0097] Around the aforementioned benchmark parameters, 100 sample points are generated within a certain range of fluctuation (e.g., , , wait).

[0098] Each sample was subjected to an FDEM simulation, which took approximately several hours per simulation. In this example, a total of 100 simulations were performed, generating a numerical sample library containing 100 samples.

[0099] S2: Extract key mechanical state variables.

[0100] Take a simulation as an example:

[0101] 1. Spatial partitioning: The embedded segment is divided into There are two intervals, each with a depth of 0.05m.

[0102] 2. Field variable averaging: For the fourth interval (depth 0.15-0.20m), in displacement U When the damage is 5mm, the damage values ​​of all rock mass elements within this interval are statistically analyzed, and their volume-weighted average is calculated. D 4. Similarly, calculate the average plastic strain in this interval. Extract the pile bending moment at the midpoint depth of the interval.

[0103] 3. Constructing the sequence: Repeat this process for all 10 intervals and 100 loading steps to ultimately generate a sequence containing... The state sequence.

[0104] S3: Building and training agent models.

[0105] 1. Physical kernel layer settings: Select the Winkler foundation beam model as the physical kernel. Its foundation reaction coefficient... k The value is estimated from the input rock modulus. Given a displacement, the model quickly calculates the physical prediction of the pile bending moment. Physical prediction of pile top load .

[0106] 2. Residual learning branch setup: Construct a multilayer perceptron (MLP), whose input layer includes: pile diameter d Rock embedding depth L r Rock elastic modulus E r Target pile top displacement U and the ground reaction force calculated by the physical kernel p ( z The MLP outputs four parts: pile top load residuals. Pile body bending moment residual and physical and mechanical state variables and .

[0107] 3. Training: Approximately 100,000 structured data points (100 samples * 1000 loading steps) generated from 100 simulations were divided into training and test sets in an 8:2 ratio. The Adam optimizer was used with a multi-task loss function (where the physical constraint weights are...). (Set to 0.1) Train the surrogate model.

[0108] S4: Implement prediction and verification.

[0109] 1. Input: Predict a new case not seen in the training set: pile diameter d =0.095m, embedment depth rock modulus .

[0110] 2. Prediction: Input this set of parameters into the trained surrogate model, and output the HU curve, the pile bending moment M(z) along the pile depth, and the rock mass damage distribution D(z) along the pile depth. For example... Figure 2 The image shown is a rendering obtained after the generated strain field prediction is transmitted to a 3D visualization platform via a communication protocol.

[0111] 3. Verification: Figure 3 The figure shown is the result of FDEM simulation. The FDEM simulation result is used as a verification benchmark to compare the prediction accuracy.

[0112] Example 2

[0113] This embodiment provides a system for predicting the horizontal bearing capacity of a rock-socketed monopile using digital mechanical twins, including:

[0114] The numerical simulation module is configured to construct a numerical sample library containing various working conditions of horizontal loading of rock-socketed monopile through FDEM numerical simulation, and extract mechanical state variables that characterize the physical state of the pile-rock system from the FDEM numerical simulation results, thereby constructing a training set based on the pile top displacement, pile top load and mechanical state variables of each loading step.

[0115] The model building and training module is configured to build a proxy model with pile design parameters, rock mass design parameters and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The proxy model is trained using a training set, and the predicted values ​​output by the proxy model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function.

[0116] The prediction module is configured to use a trained surrogate model based on the design parameters of the pile foundation to be predicted to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables.

[0117] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0118] In further embodiments, the following is also provided:

[0119] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0120] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0121] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0122] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0123] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0124] A computer program product includes a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0125] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0126] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0127] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0128] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0129] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for predicting the horizontal bearing capacity of a rock-embedded monopile using digital mechanical twins, characterized in that, include: Through FDEM numerical simulation, a numerical sample library of horizontal loading of rock-socketed monopile with various working conditions is constructed. Mechanical state variables characterizing the physical state of the pile-rock system are extracted from the FDEM numerical simulation results. Thus, a training set is constructed based on the pile top displacement, pile top load and mechanical state variables of each loading step. A surrogate model is constructed with pile design parameters, rock mass design parameters, and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The surrogate model is trained using a training set, and the predicted values ​​output by the surrogate model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function. Based on the pile foundation design parameters to be predicted, the trained surrogate model is used to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables. The proxy model includes a physical kernel layer and a residual learning branch; The physical kernel layer takes the pile design parameters and rock mass design parameters as inputs and outputs the physical prediction values ​​of the pile top load and the physical prediction values ​​of the pile bending moment distributed along the pile body. The residual learning branch takes the pile design parameters, rock mass design parameters, target pile top displacement, and soil reaction force calculated from the physical core layer as inputs, and outputs the pile top load residual, pile body bending moment residual, damage factor prediction value, and plastic strain prediction value; the pile top load residual is the difference between the simulated pile top load value and the physical prediction value of the pile top load, and the pile body bending moment residual is the difference between the simulated pile body bending moment value and the physical prediction value of the pile body bending moment; The loss function is: ； in, , , and These are the predicted values ​​of the pile top load. Predicted bending moment of pile body Damage factor prediction value and plastic strain prediction values The mean squared error loss; , , , and These are the weighting coefficients; , , and These are the actual values ​​of pile top load, pile bending moment, damage factor, and plastic strain, respectively. This is a physical constraint penalty term; Physical constraint penalty item include: Calculation of soil reaction force based on predicted pile bending moment ,and Satisfies the static equilibrium equations: ; Damage factor predicted value The displacement increases monotonically with loading; Damage factor predicted value Predicted bending moment of pile body or soil reaction force Negative correlation.

2. The method for predicting the horizontal bearing capacity of a rock-embedded single pile using digital mechanical twins as described in claim 1, characterized in that, The mechanical state variables include damage factor, plastic strain, and pile bending moment. After aligning the pile top displacement and pile top load with the three mechanical state vectors for each loading step, a structured state evolution sequence is constructed for each FDEM numerical simulation.

3. The method for predicting the horizontal bearing capacity of a rock-embedded monopile using digital mechanical twins as described in claim 1, characterized in that, The process of extracting mechanical state variables includes: Along the direction of the pile's embedment depth, the surrounding rock mass is divided into... A continuous control interval; For each loading step, traverse all rock mass elements in the FDEM and determine the control interval based on the center coordinates of the rock mass elements; Calculate the volume-weighted average of the field variables for all rock mass elements within each control interval. The field variables include damage factor and plastic strain. Also, extract the pile bending moment value at the corresponding depth for each control interval.

4. A system for predicting the horizontal bearing capacity of a rock-embedded monopile using digital mechanical twins, characterized in that, The method for predicting the horizontal bearing capacity of a rock-socketed monopile using digital mechanical twins as described in any one of claims 1-3 includes: The numerical simulation module is configured to construct a numerical sample library containing various working conditions of horizontal loading of rock-socketed monopile through FDEM numerical simulation, and extract mechanical state variables that characterize the physical state of the pile-rock system from the FDEM numerical simulation results, thereby constructing a training set based on the pile top displacement, pile top load and mechanical state variables of each loading step. The model building and training module is configured to build a proxy model with pile design parameters, rock mass design parameters and target pile top displacement as inputs, and pile top load and mechanical state variables as outputs. The proxy model is trained using a training set, and the predicted values ​​output by the proxy model are made to approximate the FDEM numerical simulation values ​​by optimizing the loss function. The prediction module is configured to use a trained surrogate model based on the design parameters of the pile foundation to be predicted to obtain the horizontal load-displacement response curve and the evolution process of mechanical state variables.

5. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-3.

6. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-3.

7. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method described in any one of claims 1-3.

Images