Coastal railway bridge pile foundation intelligent design method, device and equipment

Abstract

This application proposes an intelligent design method, device, and equipment for coastal railway bridge pile foundations. First, it collects engineering input information and establishes a design variable parameter space. Then, it mathematically expresses the mechanism formulas and constraints of current railway engineering design specifications, constructing a specification constraint system for coastal railway bridge pile foundations. Soil layers, piles, pile caps, loads, and environmental objects are represented as heterogeneous graphs, and a physical-guided heterogeneous graph expert network model is established to jointly predict key indicators such as bearing capacity. A multi-objective optimization algorithm is used to collaboratively optimize parameters such as pile diameter to obtain the optimal design scheme that meets specification constraints. Finally, the model is corrected based on pile testing and foundation pile inspection results, and the design results are output. This approach achieves a unified coordination of safety, applicability, durability, and economy in the design of coastal railway bridge pile foundations, offering advantages such as high design efficiency, strong engineering applicability, good interpretability, and ease of digital application.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to an intelligent design method, device and equipment for coastal railway bridge pile foundations. Background Technology

[0002] Coastal railway bridges are typically located in coastal mudflats, river-sea confluence areas, bay shallows, and coastal soft soil areas. Their pile foundations are subjected to complex environments characterized by high salinity, high humidity, fluctuating tidal levels, wave erosion, and groundwater corrosion. Furthermore, railway bridges are also subjected to the combined effects of vertical and horizontal loads from trains, braking forces, temperature fluctuations, and occasional extreme conditions. Therefore, higher requirements are placed on the pile foundation's bearing capacity, deformation control, durability, and service reliability. The current railway engineering standard system has established a framework of specifications covering bridge and culvert design, foundation design, engineering geological investigation, concrete durability design, and pile testing.

[0003] The existing design methods for pile foundations of coastal railway bridges generally follow a sequential design model of "obtaining survey parameters - manually selecting pile type and size - verifying according to specifications - adjusting schemes". During the design process, the vertical bearing capacity of a single pile is usually calculated separately, followed by correction for pile group effects, horizontal load analysis, settlement verification, and durability checks, and then repeated calculations are performed based on engineering experience. While this method can meet the requirements of conventional engineering design, it still has significant shortcomings in the context of coastal railway bridges: First, there are many design variables, involving multiple parameters such as pile diameter, pile length, number of piles, pile spacing, pile cap depth, material strength, anti-corrosion measures, and protective layer thickness, resulting in low efficiency of manual iteration; second, under the combined effects of scour, soft soil, chloride erosion, pile group interaction, and train loads, the controlling factors are complex, making it difficult for traditional step-by-step design methods to achieve coordinated optimization of bearing capacity, deformation, durability, and economy; third, specification provisions are usually reflected more in the later verification stage, and many schemes that do not meet the requirements are only eliminated after many calculations, leading to low design efficiency. The aforementioned problems are particularly prominent in basic engineering projects such as coastal railway bridges, which require high reliability and long service life.

[0004] On the other hand, with the development of intelligent design and engineering artificial intelligence, some methods have attempted to use machine learning models to predict foundation bearing capacity, settlement response, or durability performance to improve design efficiency. However, most existing intelligent methods simplify the pile foundation design object to ordinary tabular data, establishing only a black-box mapping relationship of "input parameters - output indicators," which is insufficient to fully express the heterogeneous coupling relationship between "soil layer - single pile - pile cap - load - environment" in the pile foundation of coastal railway bridges. At the same time, these methods often lack explicit embedding of railway engineering code constraints and mechanical mechanism formulas, which can easily lead to situations where the predicted values ​​fit well at the data level but are unusable, unverifiable, or lack interpretability in the sense of engineering codes. For issues particularly critical in coastal environments, such as scour reduction, chloride erosion life, collaborative stress of pile groups, and switching of multi-condition control modes, traditional general machine learning models also struggle to provide a unified solution.

[0005] Furthermore, in the existing design process, survey data, static load test results, low-strain testing, acoustic wave transmission testing, high-strain testing, and subsequent operational monitoring data are often fragmented, and a closed-loop design method applicable to the pile foundations of coastal railway bridges—comprising "standard constraints, mechanism analysis, intelligent prediction, optimized design, and testing feedback correction"—has not yet been formed. Although relevant standards have stipulated the applicable conditions, result verification, and judgment criteria for testing methods such as low-strain reflected wave testing, acoustic wave transmission testing, high-strain testing, single-pile vertical compressive static load testing, single-pile tensile static load testing, and single-pile horizontal static load testing, the relevant testing data are primarily used as a means of construction quality acceptance in the existing design system and have not yet been systematically used for the intelligent correction and iterative updating of the design model.

[0006] Therefore, there is an urgent need to propose a new intelligent design method for the pile foundations of coastal railway bridges. This method should be able to organically combine the constraints of specifications, the pile-soil interaction mechanism, the effects of scour and chloride erosion, multi-condition bearing and deformation analysis, and intelligent optimization algorithms within the framework of China's current railway engineering standards. Furthermore, it should utilize the test piles and pile foundation test results to achieve model feedback correction, thereby improving the efficiency, reliability, interpretability, and engineering applicability of the pile foundation design for coastal railway bridges. Summary of the Invention

[0007] This application proposes an intelligent design method, device, and equipment for the pile foundation of coastal railway bridges, which can solve one of the problems existing in the background art.

[0008] To achieve the above objectives, this application adopts the following technical solution:

[0009] Firstly, a smart design method for pile foundations of coastal railway bridges is provided, including:

[0010] Obtain the engineering input data required for the design of pile foundations for coastal railway bridges;

[0011] The trained neural network is used to process the engineering input data to obtain the prediction results of the design indicators of the coastal railway bridge pile foundation. Combined with a multi-objective optimization algorithm, the design variables of the coastal railway bridge pile foundation are optimized collaboratively to obtain the intelligent design results of the coastal railway bridge pile foundation. The objective function of the multi-objective optimization algorithm is defined by the design variables and the design indicators of the coastal railway bridge pile foundation.

[0012] In one possible design of the first aspect, the method further includes:

[0013] The engineering input data is constructed into a heterogeneous graph. ,in, For a set of nodes, Let be the set of edges. It is a set of relation types.

[0014] In one possible design of the first aspect, the neural network is a physically guided heterogeneous graph expert network, specifically including: a heterogeneous graph encoding module, an expert network module, and a gated fusion module;

[0015] The heterogeneous graph encoding module is used to: extract features from the engineering input data in a layer-by-layer propagation manner to obtain a graph-level embedding vector;

[0016] The expert network module is used to: process the graph-level embedding vector using different experts to obtain the expert prediction values ​​of the corresponding coastal railway bridge pile foundation design indicators;

[0017] The gating fusion module is used to: use a gating mechanism to fuse the expert prediction values ​​into the prediction result.

[0018] In one possible design approach of the first aspect, the loss function of the physically guided heterogeneous graph expert network includes: data loss, physical loss, normalization loss, and regularization loss.

[0019] In one possible design approach of the first aspect, the objective function of the multi-objective optimization algorithm is... for:

[0020] in, For the design variables of pile foundations for coastal railway bridges. This is a set of design variables for the pile foundations of coastal railway bridges. Let a be the sub-objective function, and 'a' be the number of sub-functions. Input data for the project. Here, k represents the constraint number.

[0021] In one possible design of the first aspect, the multi-objective optimization algorithm uses a non-dominated sorting genetic algorithm to obtain the optimal solution set.

[0022] In one possible design approach of the first aspect, the non-dominated sorting genetic algorithm specifically includes:

[0023] Initialize the population;

[0024] For any generation, the performance of each individual in the population is calculated using a surrogate model;

[0025] Based on the performance of individuals, the objective function and constraint values ​​are calculated, and non-dominated sorting, crowding distance evaluation, crossover and mutation are performed to form a new population until convergence.

[0026] In one possible design approach of the first aspect, the engineering input data includes: geological and foundation data, bridge load data, coastal environmental data, and material and construction constraint data; the geological and foundation data includes: thickness, unit weight, cohesion, internal friction angle, deformation modulus, Poisson's ratio, permeability coefficient, pile side resistance characteristic parameters, and pile end resistance characteristic parameters of each soil layer; the bridge load data includes: vertical design load, horizontal design load, and bending moment design value; the coastal environmental data includes: scour depth, surface chloride ion concentration, chloride ion diffusion coefficient, groundwater corrosion environment level, tidal action parameters, and design life; the material and construction constraint data includes: concrete strength index, steel reinforcement strength index, steel reinforcement protective layer thickness, reinforcement ratio, and construction equipment capability parameters;

[0027] The design parameters for the pile foundations of the coastal railway bridges include: design value of vertical bearing capacity of the pile foundation, horizontal displacement of the pile top, settlement of the pile foundation, maximum bending moment of the pile body, remaining service life of the pile foundation, and construction cost.

[0028] The design variables for the pile foundations of the coastal railway bridges include: pile diameter, pile length, number of rows and columns of piles, pile spacing, pile cap depth, concrete strength index, reinforcement ratio, thickness of concrete cover, additional anti-corrosion measures parameters or equivalent durability enhancement parameters.

[0029] In a second aspect, a possible design of the first aspect is provided, in which an apparatus includes one or more functional units or modules for performing the method as described in any possible implementation of the first aspect.

[0030] Thirdly, an electronic device is provided, comprising: a processor, and a memory coupled to the processor, the memory for storing a computer program; the processor for executing the computer program stored in the memory such that the electronic device performs the method as described in any possible implementation of the first aspect.

[0031] Beneficial effects:

[0032] Based on the above technical solution, the following steps are taken: First, engineering input information such as geological stratification, load conditions, scour depth, chloride erosion environment, material parameters, and construction constraints of the coastal railway bridge pile foundation is collected, and a design variable parameter space is established. Then, the mechanism formulas of vertical bearing capacity, scour correction, settlement calculation, horizontal load response, and chloride ion diffusion life are mathematically expressed with the constraints of the current railway engineering design specifications, constructing a specification constraint system for the coastal railway bridge pile foundation. Further, soil layers, piles, pile caps, loads, and environmental objects are represented as heterogeneous graphs, and a physical-guided heterogeneous graph expert network model is established to jointly predict key indicators such as bearing capacity, horizontal displacement, settlement, maximum bending moment, durability, and engineering cost. On this basis, a multi-objective optimization algorithm is used to collaboratively optimize parameters such as pile diameter, pile length, number of piles, pile spacing, pile cap embedment depth, reinforcement ratio, and protective layer thickness to obtain the optimal design scheme that meets the specification constraints. Finally, the model is corrected based on the results of pile testing and pile inspection, and the design results are output. This approach enables a unified coordination of safety, applicability, durability, and economy in the design of coastal railway bridge pile foundations, offering advantages such as high design efficiency, strong engineering applicability, good interpretability, and ease of digital application. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of an intelligent design method for coastal railway bridge pile foundations based on normative constraints and a physical-guided heterogeneous graph expert network, provided in an embodiment of this application. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0036] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification and the above-mentioned figures are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0037] Unless otherwise defined, 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 application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0038] like Figure 1 As shown, this embodiment aims to address the problems in the design of coastal railway bridge pile foundations, such as the post-implementation of regulatory constraints, low efficiency of manual iteration, insufficient consideration of coastal environmental factors such as scouring and chloride erosion, difficulty in coordinating the optimization of load-bearing capacity, deformation, durability, and economy, and the lack of engineering interpretability of existing intelligent methods. It provides an intelligent design method for coastal railway bridge pile foundations based on regulatory constraints and a physical-guided heterogeneous graph expert network. This method organically integrates current railway engineering standards, pile-soil interaction mechanisms, durability evolution laws, and multi-objective optimization algorithms to achieve rapid generation, accurate prediction, collaborative optimization, and feedback correction of pile foundation schemes for coastal railway bridges.

[0039] To achieve the above objectives, the technical solution of this embodiment includes the following steps.

[0040] S1. Acquisition and unified parameterization of engineering input data.

[0041] First, raw data for the target coastal railway bridge project was collected, and a unified set of input parameters was constructed. The input data includes at least four categories: the first category is geological and foundation data. , denoted as:

[0042] in, Let be the thickness of the i-th soil layer. The soil weight, For cohesion, It is the internal friction angle. For compression or deformation modulus, Poisson's ratio, Permeability coefficient, These are characteristic parameters of pile side resistance. is the characteristic parameter of pile end resistance, and m is the total number of soil layers.

[0043] The second category is bridge load data. , denoted as:

[0044] Where, N d For vertical design loads, H d For horizontal design loads, M d T is the design value for bending moment. d Qd represents the equivalent load due to temperature effects, and Qd represents the combined value of other variable actions.

[0045] The third category is coastal environmental data. , denoted as:

[0046] in, For scouring depth, This represents the surface chloride ion concentration or environmental chloride salt characterization value. The chloride ion diffusion coefficient is... These are parameters related to groundwater level and corrosive environment level. It is an indicator of tidal or seawater infiltration frequency. To meet design life requirements.

[0047] The fourth category is material and construction constraint data. , denoted as:

[0048] in, For concrete strength design parameters, For the design index of steel reinforcement strength, For the thickness of the protective layer, For reinforcement ratio, These are construction constraint parameters. These are the equipment capacity constraint parameters.

[0049] The above inputs are uniformly represented as:

[0050] The output of S1 is a standardized engineering input vector u and a layered foundation structure information table, which are used for subsequent construction and performance analysis.

[0051] S2. Parameterization of candidate pile foundation schemes and definition of design variables.

[0052] The candidate schemes for the pile foundation of the coastal railway bridge are parameterized. Each scheme to be designed is represented as a vector of design variables. :

[0053] in, For the pile diameter, For the pile length, and These represent the number of rows and columns of the pile group, respectively. and These represent the pile spacing in two directions. To deepen the foundation, For concrete strength indicators, For reinforcement ratio, For the thickness of the concrete cover for reinforcing bars, Parameters for additional corrosion protection measures or equivalent durability enhancement parameters.

[0054] The total number of piles n is denoted as: ·

[0055] If the pile cap dimensions are automatically generated from the pile group layout, then the pile cap plan dimensions can satisfy:

[0056] in, The width of the foundation. The length of the foundation is... and This is the structural control distance from the center of the side pile to the edge of the pile cap.

[0057] The pile foundation of the coastal railway bridge is expressed as an 11-dimensional design variable vector. By considering the value ranges of various variables, discrete attributes, pile foundation construction specifications, and the coupled constraints of coastal chloride durability, a feasible space of design variables that satisfies engineering feasibility is defined. :

[0058] In the formula: The basic space is a hyperrectangular structure consisting of independent upper and lower bounds for each variable; For all specification, construction, and durability inequality constraints; arbitrary A set of compliant and constructible candidate schemes for coastal railway bridge pile foundations is proposed.

[0059] The output of S2 is the design variable space that the optimizer can search. .

[0060] S3, Mathematicalization of Standard Constraints.

[0061] This embodiment does not leave the specification constraints to be manually verified at the end of the design process; instead, it transforms them into computable inequality constraints and embeds them into the entire process of scheme generation and optimization. (Constraint set) The unified representation is:

[0062] Where K is the total number of constraints.

[0063] 1. Vertical bearing capacity constraint:

[0064] Vertical ultimate bearing capacity of a single pile It can be written as:

[0065] The total resistance of the pile side for:

[0066] Pile end resistance for:

[0067] in, The cross-sectional area of ​​the pile tip. The formula for calculating the perimeter of the pile is as follows:

[0068] Let be the effective penetration length of the pile in the i-th soil layer. These are characteristic parameters of the pile end resistance in the bearing stratum. Design bearing capacity. It can be represented as:

[0069] in This is a correction factor or equivalent safety factor for bearing capacity.

[0070] Considering the pile group effect, the vertical bearing capacity of the pile group Represented as:

[0071] in This represents the efficiency coefficient of the pile group. Therefore, the vertical bearing constraint... Written as:

[0072] 2. Scour correction constraint:

[0073] For coastal railway bridges, the reduction in effective pile length due to scouring must be considered. Define the effective pile length. for:

[0074] The effective length in each soil layer is rewritten as: The ultimate bearing capacity after scouring for:

[0075] Its design value for:

[0076] Design bearing capacity after pile scour for:

[0077] Corresponding scour correction constraints for:

[0078] 3. Settlement constraints:

[0079] Settlement of pile foundation This can be written using the hierarchical summation approach:

[0080] in Let be the average value of the additional stress in the i-th layer. Then the settlement constraint... for:

[0081] in The allowable settlement limit.

[0082] 4. Horizontal displacement and bending moment constraints:

[0083] A single pile or an equivalent pile group satisfies the beam-foundation interaction relationship under horizontal loads:

[0084] in The elastic modulus of the pile material. The moment of inertia of the pile section, Let z be the horizontal displacement of the pile at depth z. This is the horizontal foundation reaction coefficient.

[0085] From this, the horizontal displacement y0 at the pile top and the maximum bending moment Mmax of the pile can be obtained. Corresponding horizontal displacement constraints... Bending moment constraints They are written as:

[0086] in To allow horizontal displacement, This is the design value for the bending load of the pile.

[0087] 5. Combined strength constraint of axial compression and bending moment:

[0088] The pile body satisfies the following under combined stress conditions:

[0089] Where N is the axial pressure on the pile and M is the bending moment of the pile.

[0090] Therefore, the combined strength constraint of axial compression and bending moment It can be written as:

[0091] in and These are the axial compression and bending ultimate bearing values ​​of the pile body, respectively. and This is the resistance reduction factor.

[0092] 6. Durability life constraints:

[0093] In a coastal chloride-salt environment, a chloride ion diffusion model is used to estimate the corrosion initiation time. Chloride ion concentration. It can be represented as:

[0094] in, Let be time, and erf be the error function.

[0095] Corrosion begins when the chloride ion concentration at a depth x = ccov in the concrete cover of the reinforcing steel reaches the critical value Ccr. That is:

[0096] Solve for corrosion initiation time : -2

[0097] Durability constraints Written as:

[0098] 7. Structural and Construction Constraints:

[0099] Depending on equipment capabilities and structural limitations, structural and construction constraints may also be added. - : ≤0 ≤0 ≤0 ≤0

[0100] Boundary constraints such as pile length, pile cap size, and maximum drilling depth of construction equipment.

[0101] in, For the minimum allowable pile diameter, For the maximum allowable pile diameter, This is the minimum allowable center-to-center distance between piles.

[0102] S3 outputs a complete normative constraint system. .

[0103] S4. Construction of heterogeneous pile foundation diagrams for coastal railway bridges.

[0104] This embodiment does not simplify the design object to a regular table, but instead constructs a heterogeneous diagram. :

[0105]

[0106] in For a set of nodes, Let be the set of edges. It is a set of relation types.

[0107] 1. Node type:

[0108] The node set contains the following five categories:

[0109] Soil Node Set :

[0110]

[0111] Pile node set :

[0112]

[0113] Pier Node Set :

[0114]

[0115] Load node set :

[0116]

[0117] Set of environment nodes :

[0118]

[0119] 2. Node characteristics:

[0120] Soil nodal feature vector for:

[0121]

[0122] The initial eigenvector of the j-th pile for:

[0123] [ ]

[0124] in, The elastic modulus of the pile material. Let be the plane coordinates of the j-th pile in the pile cap coordinate system.

[0125] Features of pile cap nodes for:

[0126]

[0127] in, The width of the foundation. The length of the foundation is... This refers to the height of the foundation.

[0128] Load node characteristics for:

[0129]

[0130] Environmental node characteristics for:

[0131]

[0132] 3. Edge type:

[0133] Define soil layer-soil layer edge, pile-soil layer action edge, pile-pile cap connection edge, pile cap-load transfer edge, environment-pile action edge, and environment-soil layer action edge, each corresponding to a different relationship r∈R.

[0134] The output of S4 is a heterogeneous graph representation of engineering objects.

[0135] S5. Physically guided heterogeneous graph expert network model establishment.

[0136] This embodiment employs a physically guided heterogeneous graph expert network for surrogate prediction. The model consists of three sub-modules: a heterogeneous graph encoding module, an expert network module, and a gating fusion module.

[0137] 1. Heterogeneous graph encoding module:

[0138] In the In the layer graph message passing, the embedding of node i is updated as follows:

[0140] in, Let i be the set of neighbors of node i under relation r. Let l be the relation transformation matrix of the l-th level. For relational attention weights, It is a non-linear activation function. The embedding vector of node i in layer (l+1) is obtained by aggregating the neighbor information and its own information in layer l. Let j be the embedding vector of the neighbor node j at layer l. This is the transformation matrix embedded by node i in the l-th layer, used to preserve its own features.

[0141] After multi-layer propagation, a graph-level embedding vector is obtained. :

[0142]

[0143] READOUT is the graph-level pooling function.

[0144] 2. Expert Network Module:

[0145] Set up M expert subnetworks, each modeling a different objective. Let the output of the m-th expert be... for:

[0146]

[0147] in It is a function expression.

[0148] Preferably, the following can be set:

[0149] • Expert 1: Vertical bearing capacity expert;

[0150] • Expert 2: Horizontal displacement and bending moment expert;

[0151] • Expert 3: Settlement expert;

[0152] • Expert 4: Durability and Lifespan Expert;

[0153] • Expert 5: Cost and construction expert.

[0154] Each expert outputs a component vector:

[0155]

[0156] in, Let m be the design value of the vertical bearing capacity of the pile foundation predicted by the expert. Let m be the predicted horizontal displacement of the pile top by the expert. Let m be the predicted settlement of the pile foundation by the expert. Let m be the maximum bending moment of the pile predicted by the expert. Let m be the remaining service life of the pile foundation predicted by the expert. Let m be the construction cost predicted by the m-th expert.

[0157] 3. Gating fusion module:

[0158] The gating network automatically assigns weights to different experts based on the current operating conditions. The weight of the m-th expert... Defined as:

[0159]

[0160] Final output for:

[0161]

[0162] Right now:

[0163]

[0164] in These represent the predicted values ​​for the vertical bearing capacity of the pile foundation, the horizontal displacement of the pile top, the settlement, the maximum bending moment, the durability life, and the project cost, respectively.

[0165] The output of S5 is the PI-HGMoE proxy model before training, which is a highly customized engineering-specific hybrid expert model (MoE) designed primarily for civil engineering (pile foundation design) scenarios.

[0166] S6. Model training methods.

[0167] This embodiment employs a joint training method combining "data loss + physical loss + normalization loss + regularization loss".

[0168] 1. Sample set construction:

[0169] Training sample set Represented as:

[0170]

[0171] in, The true labels for the samples can come from existing engineering databases, finite element / elastoplastic numerical analysis, pile test results, or detection inversion data. Let q be the input value for the q-th sample. This represents the total number of training samples.

[0172] The true label can be written as:

[0173]

[0174] 2. Data loss:

[0175] The data loss is constructed using mean squared error Ldata:

[0176]

[0177] 3. Physical loss:

[0178] To avoid the model only learning black-box mapping, the key mechanism formula of pile foundation is written into the physical loss term.

[0179] Vertical load consistency loss :

[0180]

[0181] in, Let q be the vertical allowable bearing capacity predicted by the model for the q-th sample. The ultimate bearing capacity is calculated based on the pile foundation mechanics formula. For the load-bearing capacity safety partial factor, The total number of samples.

[0182] Flushing to correct consistency loss :

[0183]

[0184] in, The vertical allowable bearing capacity after repainting correction as predicted by the model. The modified ultimate bearing capacity takes into account the effects of scouring.

[0185] Settlement uniformity loss :

[0186]

[0187] in, The settling value of the pile foundation is predicted by the model. This represents the theoretical settlement value calculated using soil mechanics methods.

[0188] Consistent loss of durability life :

[0189]

[0190] in, The model predicts the durability failure time. This represents the theoretical durability life calculated based on Fick's diffusion law.

[0191] Define total physical loss for:

[0192]

[0193] 4. Standardize losses :

[0194] Penalize all predictions that violate normative constraints:

[0195]

[0196] in, Let be the input feature vector of the q-th sample. The model output variable for the q-th sample.

[0197] 5. Regular loss:

[0198] To prevent overfitting, a regularization term Lreg is introduced:

[0199]

[0200] in This is the set of model parameters.

[0201] 6. Total loss function L:

[0202] The ultimate training objective is:

[0203]

[0204] in These are the weighting coefficients for each loss.

[0205] 7. Parameter update:

[0206] Update model parameters using gradient descent or adaptive optimization algorithms. :

[0207]

[0208] in Let t be the learning rate and t be the number of iterations.

[0209] The output of S6 is the trained physics-guided heterogeneous graph expert network agent model. .

[0210] S7. Multi-objective optimization solution.

[0211] By using a trained surrogate model to replace time-consuming and detailed calculations, the optimal pile foundation scheme can be quickly searched in the design variable space.

[0212] 1. Objective function construction:

[0213] This embodiment sets at least the following three objective functions.

[0214] Objective 1: Minimize the total project cost;

[0215] Total cost It can be represented as:

[0216]

[0217] Or it can be broken down by materials and construction:

[0218]

[0219] in These are the unit cost coefficients for concrete, steel reinforcement, and anti-corrosion treatment, respectively. Cost of the foundation platform.

[0220] Objective 2: Minimize deformation response;

[0221] Define the integrated deformation target for:

[0222]

[0223] in and These are the weighting coefficients. The settlement deformation index predicted by the model. This is another type of deformation response predicted by the model.

[0224] Objective 3: Minimize durability risk;

[0225] Define durability risk for:

[0226]

[0227] in, This represents the durability failure time predicted by the model.

[0228] Or equivalently defined as:

[0229]

[0230] If it is necessary to balance low-carbon goals, carbon emission targets can also be added. :

[0231]

[0232] in For concrete volume, For the quality of steel bars, This refers to the amount of anti-corrosion material used.

[0233] 2. Optimize the model:

[0234] Therefore, the optimization problem of this embodiment can be written as:

[0235]

[0236] satisfy

[0237]

[0238] 3. Feasibility Priority and Penalty Function:

[0239] For schemes that violate the constraints, a total violation score is introduced. :

[0240]

[0241] Or equivalent penalty function :

[0242]

[0243] If scalarized targets are adopted It can also be written as:

[0244]

[0245] in The target weight.

[0246] 4. Evolutionary search:

[0247] The non-dominated sorting genetic algorithm NSGA-II is preferred. Population initialization. for:

[0248]

[0249] in, This represents a set of pile foundation design variables.

[0250] For any generation The performance of each individual is calculated through a proxy model. :

[0251]

[0252] Then, based on individual performance, the objective function and constraint values ​​are calculated to complete non-dominated ranking, crowding distance evaluation, crossover, and mutation, forming a new population P:

[0253]

[0254] The termination condition can be set to the maximum algebra. Or Pareto front convergence index satisfy:

[0255]

[0256] in This is an indicator of excessive volume.

[0257] The output of S7 is the Pareto optimal solution set that satisfies the canonical constraints. :

[0258]

[0259] in, This represents the k-th optimal solution in the set, and each solution corresponds to a complete set of pile foundation design variables.

[0260] S8. Recommended scheme selection and detailed review.

[0261] Recommended solutions are selected from the Pareto solution set based on engineering preferences. A comprehensive scoring function is defined. :

[0262]

[0263] in, , , For engineering preference weights, satisfying =1, used to adjust the priority of individual objectives in decision-making.

[0264] in This is the normalized target value. Let:

[0265]

[0266] Recommended scheme Perform another detailed calculation or high-fidelity finite element verification to confirm the consistency between the surrogate model prediction results and the final design values.

[0267] S8's output is the final recommended design scheme. and its integrity performance indicators :

[0268]

[0269] * indicates the updated metric.

[0270] S9. Feedback correction based on pile test and inspection results.

[0271] Before or during construction, the model is corrected by incorporating test piles and pile inspection results. Inspection methods may include low strain, acoustic transmission, high strain, single pile vertical compressive static load, single pile tensile static load, and single pile horizontal static load, etc. The relevant methods and applicable scope are covered by the current "Technical Specification for Pile Inspection in Railway Engineering".

[0272] Set the on-site detection and observation values for:

[0273]

[0274] The subscript obs indicates the observed value.

[0275] Model predictions for:

[0276]

[0277] The subscript pred indicates the predicted value.

[0278] Define the feedback correction loss Lfb as:

[0279]

[0280] The target after incremental update It can be written as:

[0281]

[0282] The updated model parameters are denoted as And based on this, re-output the design. This creates a closed-loop correction process from surveying, design, pile testing, inspection to redesign.

[0283] S10, Output of digital results.

[0284] The final design results will be output as a two-dimensional construction parameter table, a design verification report, a durability assessment report, a recommended scheme comparison table, and an information model data package.

[0285] If integrated into the railway engineering information model system, the output structure should meet the application mode and basic rules specified in the unified standard for railway engineering information models.

[0286] TB / T 10183-2021 was announced and implemented by the National Railway Administration in 2021 and is suitable as the standard basis for digital delivery in this embodiment.

[0287] Output results At least including:

[0288]

[0289] in For the verification report, For information model data.

[0290] Advantages of this embodiment:

[0291] 1. This embodiment moves the specification constraints in the design of coastal railway bridge pile foundations to the entire process of scheme generation and optimization. It transforms the serial process of traditional design mode, which relies on repeated manual calculations and subsequent item-by-item verification, into an integrated process of "real-time discrimination of specification constraints - automatic generation of schemes - rapid performance prediction - multi-objective collaborative optimization". This can significantly improve the design efficiency of coastal railway bridge pile foundations under complex environmental conditions, reduce the repeated calculation of invalid schemes, and improve the automation and intelligence level of design work.

[0292] 2. This embodiment constructs a heterogeneous graph structure of "soil layer - pile - pile cap - load - environment" and uses a physical-guided heterogeneous graph expert network to jointly model indicators such as bearing capacity, settlement, horizontal displacement, pile internal force, durability and engineering cost. Compared with traditional tabular machine learning methods, it can more fully characterize the topological relationships, action paths and coupling mechanisms between multi-source heterogeneous objects in the design of coastal railway bridge pile foundations, thereby improving the model prediction accuracy and enhancing the engineering applicability and interpretability of the method.

[0293] 3. In this embodiment, the physical mechanism formulas such as vertical bearing capacity of pile foundation, scour correction, settlement calculation, horizontal load response and chloride ion diffusion life are embedded into the model training and optimization process. This ensures that the intelligent prediction results are consistent with the laws of engineering mechanics and current railway engineering standards. It avoids the problem that the prediction results of a pure black box model may fit the data but do not meet the specifications or lack engineering rationality, thereby improving the reliability and verifiability of the design results.

[0294] 4. Under the premise of meeting the specifications and constraints of bearing capacity, deformation, durability and structure, this embodiment performs multi-objective collaborative optimization with the goals of engineering cost, deformation response and durability risk. It can achieve a better balance between safety, applicability, durability and economy, and avoid the conservative or uneconomical design caused by the control of a single index in traditional methods. At the same time, this embodiment can also combine the test pile and pile detection results to make feedback corrections to the model parameters and design schemes, forming a closed-loop optimization mechanism covering the entire process of design, construction and testing and verification. Therefore, it has good engineering promotion value and digital application prospects.

[0295] Specific application examples:

[0296] This embodiment selects a single pier pile foundation in the substructure of a coastal railway bridge as the research object. The bridge site is located in a coastal soft soil area, and the site is subject to long-term tidal fluctuations, sea breeze salt spray, and groundwater corrosion. The design service life is 100 years. The design load transferred from the superstructure of the pier to the top surface of the foundation is:

[0297] Vertical design load ;

[0298] Horizontal design load ;

[0299] • Foundation top surface design bending moment ;

[0300] Based on the hydrological and scour analysis of the bridge site, the design scour depth is taken as... The environmental surface chloride ion concentration parameter is taken as follows: The critical concentration for steel corrosion is taken as The equivalent chloride ion diffusion coefficient is taken as The groundwater environment is considered as a highly corrosive environment.

[0301] The site survey revealed that the site can be divided into four layers from top to bottom:

[0302] Layer 1: Silt-clay, thickness Severe Cohesion internal friction angle Compression modulus Characteristic value of pile side resistance ;

[0303] Layer 2: Silty clay, thickness , , , , , ;

[0304] Layer 3: Medium-dense fine sand, layer thickness , , , , , ;

[0305] Layer 4: Moderately weathered silty mudstone (bearing layer), with a top surface burial depth of approximately 24 m. The characteristic value of the pile end resistance is taken as... The corresponding characteristic value of pile side resistance is taken as ;

[0306] In this example, the design variables for pile foundations are set as follows:

[0307]

[0308] Search range for each parameter:

[0309] • Pile diameter : ;

[0310] • Pile length : ;

[0311] ·Number of rows of pile groups Column number The combination range is ;

[0312] • Pile spacing : ;

[0313] • Foundation depth : ;

[0314] • Reinforcement ratio : ;

[0315] • Thickness of concrete cover for reinforcing bars : ;

[0316] I. Mechanical Calculations and Constraint Judgment in Examples

[0317] For any candidate solution, first perform specification constraint discrimination and mechanism calculation according to this embodiment. Taking candidate solution A as an example, take:

[0318]

[0319] That is, the total number of piles Foundation depth reinforcement ratio protective layer thickness .

[0320] 1. Geometric parameters of a single pile:

[0321] The perimeter and tip area of ​​a single pile are respectively:

[0322]

[0323]

[0324] 2. Vertical bearing capacity calculation (excluding scour):

[0325] Assuming the pile passes through the first layer (6 m), the second layer (8 m), the third layer (10 m), and enters the fourth bearing layer (10 m), then the side resistance of the single pile is:

[0326]

[0327] Substitute the values:

[0328]

[0329] The end resistance of a single pile is:

[0330]

[0331] Therefore, the ultimate bearing capacity of a single pile is:

[0332]

[0333] If the load-bearing correction factor is taken The design bearing capacity of a single pile is:

[0334]

[0335] Considering the efficiency coefficient of pile groups The vertical design bearing capacity of the pile group is:

[0336]

[0337] The corresponding vertical bearing constraint value is:

[0338]

[0339] This indicates that the vertical bearing capacity requirement is met before considering scour reduction.

[0340] 3. Calculation of bearing capacity after scour correction:

[0341] Further consideration of scour depth The effective pile length becomes:

[0342]

[0343] After deducting the scour section, the effective length of the pile in each layer is redistributed as follows: Then the side resistance of a single pile after scouring becomes:

[0344]

[0345] Therefore, the ultimate bearing capacity of a single pile after scour is:

[0346]

[0347] The design bearing capacity of a single pile becomes:

[0348]

[0349] The bearing capacity of the pile group after scour is:

[0350]

[0351] The corresponding bearing constraint value after scouring is:

[0352]

[0353] The requirements are still met.

[0354] 4. Settlement calculation:

[0355] Settlement was estimated using the layered summation method. Based on the calculated additional stress, the average additional stress of the four layers was taken as follows: , , , The sedimentation is as follows:

[0356]

[0357]

[0358]

[0359] The estimated settlement is approximately 177.8 mm. If the allowable settlement limit for the bridge pier is taken as 200 mm, then:

[0360]

[0361] The requirements are met.

[0362] 5. Durability life calculation:

[0363] Durability was estimated using a chloride ion diffusion model. Protective layer thickness. The corrosion initiation time is:

[0364]

[0365] Substitute parameters:

[0366]

[0367] Pick ,have to:

[0368]

[0369] Converted to an annual rate, it is approximately:

[0370]

[0371] If lifespan is required In the year:

[0372]

[0373] It meets durability requirements.

[0374] II. Details of Heterogeneous Graph Model Construction

[0375] In this embodiment, a single bridge pier pile foundation design object is represented as a heterogeneous diagram. This includes:

[0376] 4 soil layer nodes;

[0377] 9 pile nodes;

[0378] One foundation node;

[0379] 1 load node;

[0380] One environment node.

[0381] Therefore, the total number of nodes is 16. The edges are constructed as follows:

[0382] • Establish three interlayer edges between adjacent soil layers;

[0383] • Each pile and the four soil layers it passes through establish a pile-soil action edge, totaling... strip;

[0384] • All nine piles are connected to the pile cap via nine separate edges;

[0385] • Corrosion action edges are established between the environmental nodes and the 9 piles, totaling 9 edges;

[0386] • Establish two interaction edges between the environmental node and the upper soft soil layer node.

[0387] Therefore, the total number of edges in this embodiment is 60.

[0388] The initial feature dimension of soil nodes is 9 dimensions, pile nodes are 8 dimensions, cap nodes are 4 dimensions, load nodes are 4 dimensions, and environmental nodes are 6 dimensions. To adapt to the unified graph network encoding, all node features are linearly mapped to a 64-dimensional latent space.

[0389] The graph encoder employs a 3-layer heterogeneous graph attention network, with each layer having a hidden dimension of 64 and 4 attention heads. The layer message passing update formula is:

[0391] The graph-level readout layer employs attention pooling, embedding and aggregating all nodes into a 128-dimensional graph vector. .

[0392] III. Hybrid Expert Network and Output Settings

[0393] This example sets up 5 expert networks:

[0394] • Expert 1: Vertical bearing capacity expert;

[0395] • Expert 2: Horizontal displacement and bending moment expert;

[0396] • Expert 3: Settlement expert;

[0397] • Expert 4: Durability and Lifespan Expert;

[0398] Expert 5: Cost Estimation Expert.

[0399] Each expert network consists of three fully connected layers with 128, 64, and 32 neurons respectively, using GELU as the activation function. The gated network consists of two fully connected layers, outputting five expert weights, which are then normalized using softmax.

[0400]

[0401] The final output is:

[0402]

[0403] In this embodiment, the output quantities correspond to the design bearing capacity of a single pile, the horizontal displacement of the pile top, the settlement of a group of piles, the maximum bending moment, the durability life, and the total cost, respectively.

[0404] IV. Training Data and Training Parameters

[0405] The training samples in this embodiment consist of three parts:

[0406] • 420 sets of historical coastal bridge pile foundation engineering data;

[0407] • 3580 sets of high-fidelity numerical simulation samples generated based on parameterized formation and load combinations;

[0408] • 160 sets of test piles and testing inversion samples.

[0409] Therefore, the total sample size is 4160 groups. The design variables and formation conditions in the sample were all adjusted using Latin hypercube sampling and engineering experience boundaries. The data were divided into training, validation, and test sets in a ratio of 70%:15%:15%.

[0410] The training loss function is:

[0411]

[0412] In this embodiment, the following is taken:

[0413]

[0414] The optimizer uses AdamW with an initial learning rate of The batch size was 32, and the number of training epochs was 300. If the validation set loss did not decrease for 25 consecutive epochs, the training was stopped early. The model converged on the 214th epoch.

[0415] The mean relative errors of each output on the test set are as follows:

[0416] • Load-bearing capacity: 3.4%

[0417] Settlement: 5.1%

[0418] • Horizontal displacement at the pile top: 4.7%

[0419] Maximum bending moment: 5.8%

[0420] Durability lifespan: 4.3%

[0421] • Cost: 2.9%

[0422] This demonstrates that the physical-guided heterogeneous graph expert network in this embodiment can accurately perform joint prediction of key indicators of coastal railway bridge pile foundations.

[0423] V. Multi-objective optimization settings

[0424] This embodiment uses total cost, overall deformation, and durability risk as optimization objectives. The objective functions are as follows:

[0425]

[0426]

[0427]

[0428] Simultaneously satisfy all specification constraints:

[0429]

[0430] The optimization algorithm used is NSGA-II, with a population size of 120, a maximum number of generations of iterations of 180, a crossover probability of 0.9, and a mutation probability of 0.08. The initial population is randomly generated within the design space and enters the evolutionary process after feasibility screening.

[0431] After 180 iterations, 26 Pareto optimal solutions were obtained. Based on the comprehensive scoring function...

[0432]

[0433] Option B is selected as the recommended option, and its main parameters are:

[0434]

[0435]

[0436] VI. Recommended Scheme Results

[0437] For Option B, after prediction by the surrogate model and verification using refined calculations, the following results were obtained:

[0438] • Vertical design bearing capacity of pile groups:

[0439] • Horizontal displacement of pile top:

[0440] • Settlement of pile groups:

[0441] Maximum bending moment:

[0442] • Corrosion initiation life: Year

[0443] Total cost: Yuan

[0444] If the allowable horizontal displacement limit is 12 mm and the settlement limit is 200 mm, then both requirements are met; if the durability life is greater than the design life of 100 years, then the requirements are also met.

[0445] VII. Comparison with Traditional Experience-Based Design Solutions

[0446] To verify the effectiveness of this embodiment, it is compared with solution C obtained by a traditional empirical method. Solution C is:

[0447]

[0448] The results of the review are as follows:

[0449] • Vertical design bearing capacity of pile groups:

[0450] • Horizontal displacement of pile top:

[0451] • Settlement of pile groups:

[0452] • Corrosion initiation life: Year

[0453] Total cost: Yuan

[0454] In comparison, the recommended scheme B in this embodiment reduces the total cost by approximately: [amount missing] while satisfying all load-bearing, deformation, and durability constraints.

[0455]

[0456] Settlement increased slightly but still met the limits, and durability life increased by approximately:

[0457]

[0458] This embodiment demonstrates that it can improve economic efficiency while ensuring safety and durability.

[0459] VIII. Example of Detection Feedback Correction

[0460] During the construction phase, a static load test on the vertical compressive strength of one of the test piles was conducted. The measured ultimate bearing capacity of the single pile was... The model predicted the value as follows: The relative error is Meanwhile, the horizontal displacement at the top of the pile was measured in the single-pile horizontal static load test. The model predicts the value. The relative error is .

[0461] Add the detected feedback samples to the incremental training set to construct the feedback loss:

[0462]

[0463] and with The model was incrementally updated over 20 epochs. After the update, the bearing capacity prediction error was reduced to [missing value]. The horizontal displacement prediction error was reduced to This demonstrates that this embodiment has the ability to perform online corrections using test piles and inspection results.

[0464] IX. Conclusions of the Examples

[0465] This embodiment demonstrates that by constructing a heterogeneous diagram of the pile foundation for coastal railway bridges and utilizing a physics-guided heterogeneous diagram expert network to jointly predict load-bearing capacity, deformation, durability, and cost, and by combining a multi-objective optimization algorithm for scheme search under regulatory constraints, a high-quality pile foundation design scheme that meets the requirements for load-bearing capacity, settlement, horizontal displacement, and durability can be effectively obtained. Compared with traditional empirical design methods, this embodiment has higher design efficiency and better economy while maintaining structural safety and durability. Furthermore, it can utilize pile testing and inspection results for feedback correction, thus making it suitable for intelligent design and digital application of pile foundations for coastal railway bridges.

[0466] This application also provides an apparatus comprising one or more functional units or modules for performing the method described in any of the above embodiments.

[0467] This application also provides an electronic device, including: a processor, and a memory coupled to the processor, the memory being used to store a computer program; the processor being used to execute the computer program stored in the memory, so that the electronic device performs the method as described in any of the above embodiments.

[0468] Electronic devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. These electronic devices may include, but are not limited to, processors and memory.

[0469] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the device via various interfaces and lines.

[0470] The memory can be used to store the computer program, and the processor implements various functions of the electronic device by running or executing the computer program stored in the memory and calling the data stored in the memory.

[0471] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0472] This application also provides a storage medium, which is a computer-readable storage medium. The computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0473] This application also provides a computer program product, including: a computer program or instructions that, when the computer program or instructions are run on a computer, cause the computer to perform any of the above possible implementation methods.

[0474] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A smart design method for pile foundations of coastal railway bridges, characterized in that, include: Obtain the engineering input data required for the design of pile foundations for coastal railway bridges; The trained neural network is used to process the engineering input data to obtain the prediction results of the design indicators of the coastal railway bridge pile foundation. Combined with a multi-objective optimization algorithm, the design variables of the coastal railway bridge pile foundation are optimized collaboratively to obtain the intelligent design results of the coastal railway bridge pile foundation. The objective function of the multi-objective optimization algorithm is defined by the design variables and the design indicators of the coastal railway bridge pile foundation.

2. The method as described in claim 1, characterized in that, The method further includes: The engineering input data is constructed into a heterogeneous graph. ,in, For a set of nodes, Let be the set of edges. It is a set of relation types.

3. The method as described in claim 2, characterized in that, The neural network is a physically guided heterogeneous graph expert network, specifically including: a heterogeneous graph encoding module, an expert network module, and a gated fusion module; The heterogeneous graph encoding module is used to: extract features from the engineering input data in a layer-by-layer propagation manner to obtain a graph-level embedding vector; The expert network module is used to: process the graph-level embedding vector using different experts to obtain the expert prediction values ​​of the corresponding coastal railway bridge pile foundation design indicators; The gating fusion module is used to: use a gating mechanism to fuse the expert prediction values ​​into the prediction result.

4. The method as described in claim 2, characterized in that, The loss function of the physical-guided heterogeneous graph expert network includes: data loss, physical loss, normalization loss, and regularization loss.

5. The method as described in claim 1, characterized in that, The objective function of the multi-objective optimization algorithm for: in, For the design variables of pile foundations for coastal railway bridges. This is a set of design variables for the pile foundations of coastal railway bridges. Let a be the sub-objective function, and 'a' be the number of sub-functions. Input data for the project. Here, k represents the constraint number.

6. The method as described in claim 1, characterized in that, The multi-objective optimization algorithm uses a non-dominated sorting genetic algorithm to obtain the optimal solution set.

7. The method as described in claim 6, characterized in that, Non-dominated sorting genetic algorithms specifically include: Initialize the population; For any generation, the performance of each individual in the population is calculated using a surrogate model; Based on the performance of individuals, the objective function and constraint values ​​are calculated, and non-dominated sorting, crowding distance evaluation, crossover and mutation are performed to form a new population until convergence.

8. The method as described in claim 1, characterized in that, The engineering input data includes: geological and foundation data, bridge load data, coastal environmental data, and material and construction constraint data. The geological and foundation data includes: thickness, unit weight, cohesion, internal friction angle, deformation modulus, Poisson's ratio, permeability coefficient, pile side resistance characteristic parameters, and pile end resistance characteristic parameters for each soil layer. The bridge load data includes: vertical design load, horizontal design load, and design bending moment. The coastal environmental data includes: scour depth, surface chloride ion concentration, chloride ion diffusion coefficient, groundwater corrosion environment level, tidal action parameters, and design life. The material and construction constraint data includes: concrete strength indicators, steel reinforcement strength indicators, steel reinforcement protective layer thickness, reinforcement ratio, and construction equipment capability parameters. The design parameters for the pile foundations of the coastal railway bridges include: design value of vertical bearing capacity of the pile foundation, horizontal displacement of the pile top, settlement of the pile foundation, maximum bending moment of the pile body, remaining service life of the pile foundation, and construction cost. The design variables for the pile foundations of the coastal railway bridges include: pile diameter, pile length, number of rows and columns of piles, pile spacing, pile cap depth, concrete strength index, reinforcement ratio, thickness of concrete cover, additional anti-corrosion measures parameters or equivalent durability enhancement parameters.

9. An intelligent design device for coastal railway bridge pile foundations, characterized in that, It includes one or more functional units or modules for performing the method of any one of claims 1-8.

10. An electronic device, characterized in that, The electronic device includes: a processor, and a memory coupled to the processor. The memory is used to store computer programs; The processor is configured to execute the computer program stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-8.

Images