A basic agent model and a method for lightweight optimization design of a ship with variable thickness steel plate

Abstract

The application discloses a basic agent model and a ship lightweight optimization design method of variable-thickness steel plates, and belongs to the technical field of ship structure optimization related to finite element numerical simulation. The application solves the problem that the traditional ship variable-thickness structure optimization method in the prior art is difficult to obtain an optimal ship variable-thickness structure due to low precision and efficiency; the application establishes a ship longitudinal stiffened plate finite element model based on the geometric configuration of the ship longitudinal stiffened plate; defines core design variables of the ship longitudinal stiffened plate finite element model, sets constraint conditions, and constructs an objective function; adopts a Latin hypercube sampling method to perform test design, integrates output response values, and forms a DOE sample database; adopts a Kriging model to establish an agent model in combination with sample point correlation; and adopts a simulated annealing algorithm to output optimal design variable values. The application effectively improves the ship lightweight optimization efficiency and precision, and can be applied to the lightweight design of variable-thickness steel plate ships.

Description

Technical Field

[0001] This invention belongs to the field of ship structure optimization technology related to finite element numerical simulation, specifically a basic surrogate model and a ship lightweight optimization design method based on variable thickness steel plates. Background Technology

[0002] Lightweight ship structures are a core pathway to achieving green and intelligent ship development. By reducing empty displacement, they can directly increase cargo capacity and reduce fuel consumption and carbon emissions. With the increasing size of ships, traditional uniform thickness steel plate structures, due to low material utilization and limited optimization space, can no longer meet the requirements of lightweighting and high load-bearing capacity. Variable thickness steel plate structures, with their advantage of being able to allocate materials differently according to stress distribution, have become the mainstream direction of structural design.

[0003] As a key component for transmitting loads in the hull structure, the thickness design of longitudinal stiffening plates is closely related to the structural safety, lightweight performance, construction cost and operational economy of ships. For different ship types, nodal stress characteristics and construction process requirements, common traditional steel plate thickness design methods mainly include empirical analogy method, regional equal thickness design method, finite element numerical simulation method and direct calculation method based on standard formulas. The specific problems are as follows: (1) Empirical analogy method uses existing parent ships as references and relies on the designer's experience. It is efficient but lacks innovation and accuracy; (2) Regional equal thickness method simplifies the design based on the structural position, which is convenient for construction but has low material utilization and sacrifices lightweight potential; (3) Finite element analysis method provides detailed mechanical response through high-precision numerical simulation. The results are accurate but the calculation cost is high and it is difficult to use directly for rapid iterative optimization; (4) Standard formula method calculates based on industry standard formulas. It is authoritative and easy to operate, but it is usually conservative and difficult to capture the real performance of complex or non-standard structures.

[0004] Meanwhile, the current optimization of ship variable thickness structures mainly relies on the combination of finite element analysis (FEA) and optimization algorithms, but it faces two major technical bottlenecks: (1) High computational cost: The finite element model of ship structure (such as hull beam, deck, bulkhead) contains a large number of elements and nodes. A single strength, fatigue or buckling analysis takes several hours to several days. If intelligent optimization algorithms such as genetic algorithm and simulated annealing algorithm are used to directly drive the finite element simulation iteration, it often requires thousands or even tens of thousands of analysis calculations. The computational resources and time costs increase exponentially, which is difficult to meet the high-efficiency design requirements of actual engineering. (2) It is difficult to balance optimization accuracy and efficiency: Some studies have tried to use simplified analytical models to replace finite element simulation for optimization, but analytical models often ignore the complex boundary conditions of ship structure, welding residual stress and local stress concentration effects, resulting in a large deviation between the optimization results and the actual engineering scenario. On the other hand, using high-fidelity finite element models for optimization is limited by computational efficiency and cannot achieve large-scale multi-parameter optimization.

[0005] In summary, a basic proxy model and a method for lightweight ship design based on variable thickness steel plates are needed to perform rapid calculations while ensuring accuracy, thereby achieving global optimization, uncertainty quantification, and robust design. Summary of the Invention

[0006] A brief overview of the invention is given below to provide a basic understanding of certain aspects of it. It should be understood that this overview is not an exhaustive summary of the invention. It is not intended to identify key or essential parts of the invention, nor is it intended to limit the scope of the invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to the more detailed description that follows.

[0007] In view of this, in order to solve the problem that traditional ship variable thickness structure optimization methods in the prior art are difficult to obtain the optimal ship variable thickness structure due to low accuracy and efficiency, the present invention provides a basic surrogate model and a ship lightweight optimization design method for variable thickness steel plates.

[0008] The technical solution is as follows: A basic proxy model and a variable thickness steel plate-based lightweight ship optimization design method, comprising the following steps:

[0009] S1. Based on the geometric configuration of the longitudinal stiffening plate of the ship, a parametric finite element model of it is established using finite element software to obtain the finite element model of the longitudinal stiffening plate of the ship;

[0010] S2. Define the core design variables of the finite element model of the longitudinal stiffening plate of the ship. Under the premise of satisfying the set constraints, construct the objective function of the finite element model of the longitudinal stiffening plate of the ship with the goal of minimizing the core design variables.

[0011] S3. The core design variables are used as inputs to the finite element model of the longitudinal stiffening plate of the ship. The Latin hypercube sampling method is used for experimental design. All output response values ​​of the finite element model of the longitudinal stiffening plate of the ship are integrated to form a DOE sample database.

[0012] S4. Based on the sample point data of the DOE sample database, a surrogate model is established by using the Kriging model and combining the random error term obtained by the correlation between sample points;

[0013] S5. Employ simulated annealing algorithm to quickly evaluate a large number of design schemes on the surrogate model, ensuring that the design schemes meet the constraints and the objective function reaches its optimum. Based on the calculated acceptance probability, output the optimal design variable values.

[0014] Furthermore, in S1, the longitudinal stiffening plate of the ship is composed of stiffening strips and panels. According to the arrangement of the stiffening strips and panels, the geometric configuration of the longitudinal stiffening plate of the ship is discretized by selecting the corresponding element type using finite element software, and the overall structure of the longitudinal stiffening plate of the ship is meshed by finite element to complete the construction of the finite element model of the longitudinal stiffening plate of the ship. Subsequently, according to the load differences and functional requirements of each region in the longitudinal stiffening plate of the ship under actual working conditions, the corresponding initial thickness values ​​are assigned to different components.

[0015] Furthermore, in S2, the node thickness of each functional zone of the longitudinal stiffening plate is used as the core design variable. Interval constraints and merging are applied to the node thickness. Under the condition that the constraints are met, the objective function of the finite element model of the longitudinal stiffening plate is constructed with the goal of minimizing the total weight of the longitudinal stiffening plate. The specific steps are as follows:

[0016] Based on the fact that the longitudinal stiffening plate of a ship comprises n variable thickness regions, the plate thickness and stiffening strip dimensions of each region are defined as core design variables. ;

[0017] Core design variables Represented as:

[0018]

[0019] in, Indicates the first Node thickness parameters for each region Indicates matrix transpose;

[0020] The objective function of the finite element model of the longitudinal stiffening plate of a ship is expressed as:

[0021]

[0022] in, The total weight of the longitudinal stiffening plates of the ship. For the first Density of regional materials, For the first Cross-sectional area of ​​each region For the first The length of each region;

[0023] The constraints include strength and stability constraints, specifically:

[0024]

[0025] in, The maximum Von Mises equivalent stress of the structure under design load. For marine steel, For the maximum shear stress, For allowable shear stress, The critical buckling factor of the structure.

[0026] Furthermore, in S3, the core design variables are... The total weight of the longitudinal stiffening plate is used as an input parameter in the finite element model of the ship's longitudinal stiffening plate. and maximum Von Mises equivalent stress As the output response;

[0027] In the sampling process using the Latin hypercube sampling method, based on the core design variables... The dimension determines the number of sample points (following engineering experience, generate 100-200 sets of sample points in total), ensuring that each core design variable... The values ​​are uniformly distributed within their intervals, and the sample points do not overlap. For each set of sample points, the finite element software is automatically driven to update the finite element model of the longitudinal stiffening plate of the ship, apply loads, and perform solution calculations, ultimately outputting the response values. A complete DOE sample database is formed. After sampling, abnormal samples that have not converged in finite element calculation, have abnormal fluctuations in response values, or have deviated significantly from the overall trend are removed to obtain the DOE sample database.

[0028] Furthermore, in S4, the core design variables are constructed using a Kriging model. With response value The mapping relationship between them;

[0029] The mapping relationship is represented as follows:

[0030]

[0031] in, The predicted response value of the surrogate model. For design variables, For global trend functions, For the regression coefficient vector, For random error term, It follows a normal distribution N(0, σ). z ²), The covariance matrix is ​​determined by the correlation between sample points;

[0032] Sample points are defined using the Gaussian correlation function. and sample points Correlation between ;

[0033] Sample points and sample points Correlation between Represented as:

[0034]

[0035] in, For the first Node thickness parameters for each region The dimensions of the core design variables For sample points In the The components of the dimension, For sample points In the The components of the dimension, For the first Dimensional parameters.

[0036] Furthermore, in S5, initializing the simulated annealing algorithm parameters involves generating a new solution through small-amplitude local perturbations, increasing the node thickness in high-stress regions, and thinning the node thickness in low-stress regions, while ensuring the thickness is not less than the minimum wall thickness. This forms the initial thickness scheme;

[0037] During the initial thickness scheme generation process, the constraints of the ship node finite element model are satisfied, ensuring that the thickness change rate of adjacent columns of nodes does not exceed the maximum allowable thickness gradient, and that the node thickness perpendicular to the rolling direction remains consistent. After the strength and stiffness are verified by the finite element solver and the constraints are satisfied, the lighter feasible solution is preferentially accepted according to the objective function of the ship node finite element model. At high temperatures, a heavier feasible solution is accepted to escape local optima. The energy difference obtained is determined using the Metropolis criterion. The reception probability is calculated. According to the reception probability The process is iterated until the temperature drops to the termination value, at which point the iteration stops, and all received feasible solutions are identified and recorded.

[0038] Energy difference Represented as:

[0039]

[0040] in, For the newly generated feasible solution, This is the current feasible solution;

[0041] Receive probability The calculation formula is:

[0042]

[0043] in, Boltzmann's constant, sThis refers to the current annealing temperature parameter.

[0044] The beneficial effects of this invention are as follows: (1) Structural performance and lightweight benefits: Traditional optimization is mostly aimed at plates of uniform thickness or finite discrete thickness, and the material distribution is inflexible. Variable thickness design allows the material to be placed where it is most needed. The node thickness can be mapped to CNC cutting and welding processes, avoiding excessively complex thickness gradients, and taking into account both design accuracy and construction efficiency. The stress, weight, and cost data before and after optimization can form a database, providing a benchmark for the rapid design of similar ships in the future; (2) Improve the computational efficiency of overall optimization: When facing multiple performance constraints such as strength, stiffness, vibration frequency, and fatigue life, traditional ship structural design often adopts a serial and fragmented verification method. For example, it first designs based on strength, then uses simulation to check vibration. If it does not meet the standard, it strengthens locally, which may lead to strength redundancy and weight increase. This cycle of "design-verification-modification" makes the calculation The computational cost increases exponentially with the number of constraints. This invention adopts a surrogate model, which can use advanced machine learning methods such as Kriging model, radial basis function (RBF) neural network or support vector regression (SVR) to train a "meta-model" that can predict multiple responses such as structural stress distribution, natural frequency, and deformation at extremely high speed. This transforms the complex physical field simulation into a data-based, differentiable mathematical function approximation problem, greatly reducing the computational time cost. (3) Ensure global optimum: Traditional empirical methods or local optimization are prone to getting stuck in "local optimum". This invention combines the global approximation capability of the surrogate model with the global search capability of the intelligent optimization algorithm, making it easier to find a better or closer global optimum design scheme in a broad design space, and achieving the "ultimate balance" between weight and performance. Attached Figure Description

[0045] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0046] Figure 1 A flowchart illustrating a basic proxy model and a method for optimizing lightweight ship design using variable-thickness steel plates;

[0047] Figure 2 This is a schematic diagram of the approximate error analysis for the surrogate model;

[0048] Figure 3 To optimize the stress diagram of the front steel plate;

[0049] Figure 4 This is a schematic diagram of the stress in the optimized steel plate. Detailed Implementation

[0050] To make the technical solutions and advantages of the embodiments of the present invention clearer, the exemplary embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0051] refer to Figures 1-4 This embodiment details a basic proxy model and a method for optimizing lightweight ship design using variable-thickness steel plates, specifically including the following steps:

[0052] S1. Based on the geometric configuration of the longitudinal stiffening plate of the ship, a parametric finite element model of it is established using finite element software to obtain the finite element model of the longitudinal stiffening plate of the ship;

[0053] S2. Define the core design variables of the finite element model of the longitudinal stiffening plate of the ship. Under the premise of satisfying the set constraints, construct the objective function of the finite element model of the longitudinal stiffening plate of the ship with the goal of minimizing the core design variables.

[0054] S3. In order to obtain sample data covering the entire design variable space and support the high-precision training of the surrogate model, the core design variables are used as inputs to the finite element model of the longitudinal stiffening plate of the ship. The Latin hypercube sampling (LHS) method is used for design of experiments (DOE) to integrate all output response values ​​of the finite element model of the longitudinal stiffening plate of the ship to form a DOE sample database.

[0055] S4. Based on the sample point data of the DOE sample database, a surrogate model is established by using the Kriging model and combining the random error term obtained by the correlation between sample points;

[0056] S5. The simulated annealing algorithm is used to quickly evaluate a large number of design schemes on the surrogate model, so that the design schemes meet the constraints and the objective function reaches the optimum. Based on the calculated acceptance probability, the optimal design variable values ​​(i.e., the thickness of each part) are output.

[0057] For details, please refer to Figure 2 Actual represents the actual value of the function, and predicted represents the predicted value of the objective function;

[0058] Furthermore, in S1, the longitudinal stiffening plate of the ship is composed of stiffening strips and a panel. According to the arrangement of the stiffening strips and the panel, the geometric configuration of the longitudinal stiffening plate of the ship is discretized by selecting the corresponding element type using finite element software, and the overall structure of the longitudinal stiffening plate of the ship is meshed using finite element methods. This ensures that necessary mesh refinement is performed in areas of stress concentration or critical force transmission paths, thus completing the construction of the finite element model of the longitudinal stiffening plate of the ship. The finite element model of the longitudinal stiffening plate of the ship needs to be able to clearly distinguish different components such as the panel, the longitudinal stiffening web, and the longitudinal stiffening edge strip. Subsequently, based on the differences in loads borne by each region of the longitudinal stiffening plate in actual working conditions and functional requirements, corresponding initial thickness values ​​are assigned to different components as the starting point for subsequent optimization.

[0059] Furthermore, in S2, the node thickness of each functional zone of the longitudinal stiffening plate of the ship is used as the core design variable. Considering the manufacturability of the project, the node thickness is constrained and merged within intervals. Under the condition of satisfying the constraints, the objective function of the finite element model of the longitudinal stiffening plate of the ship is constructed with the goal of minimizing the total weight of the longitudinal stiffening plate. This balances structural lightweighting and economy. Through the core objective function, the core requirement of lightweight design, "to minimize material usage under all constraints," is achieved. The specific steps are as follows:

[0060] Based on the fact that the longitudinal stiffening plate of a ship comprises n variable thickness regions, the plate thickness and stiffening strip dimensions of each region are defined as core design variables. ;

[0061] Core design variables Represented as:

[0062]

[0063] in, Indicates the first Equivalent thickness or node thickness parameters for each region Indicates matrix transpose;

[0064] The objective function of the finite element model of the longitudinal stiffening plate of a ship is expressed as:

[0065]

[0066] in, The total weight of the longitudinal stiffening plates of the ship. For the first Density of regional materials, For the first The cross-sectional area of ​​each region (unit), which serves as The function, For the first The length of each region (unit);

[0067] To ensure the safety of the ship's structure, the constraints include strength and stability constraints, specifically:

[0068]

[0069] in, The maximum Von Mises equivalent stress of the structure under design load is obtained through finite element analysis. For marine steel, For the maximum shear stress, For allowable shear stress, It is the critical buckling factor of the structure, which is used to prevent plate instability.

[0070] Furthermore, in step S3, a data interaction interface is established between the finite element software and the optimization platform to incorporate core design variables. The total weight of the longitudinal stiffening plate is used as an input parameter in the finite element model of the ship's longitudinal stiffening plate. and maximum Von Mises equivalent stress As the output response;

[0071] In the sampling process using the Latin hypercube sampling method, based on the core design variables... The dimension determines the number of sample points (following engineering experience, generate 100-200 sets of sample points in total), ensuring that each core design variable... The values ​​are uniformly distributed within their intervals, and the sample points do not overlap. For each set of sample points, the finite element software is automatically driven to update the finite element model of the longitudinal stiffening plate of the ship, apply loads, and perform solution calculations, ultimately outputting the response values. A complete DOE sample database is formed. After sampling, abnormal samples that have not converged in finite element calculation, have abnormal fluctuations in response values, or have deviated significantly from the overall trend are removed to ensure the effectiveness and reliability of the database, thus obtaining the final DOE sample database.

[0072] Furthermore, in step S4, to replace the high-cost finite element calculation, a Kriging model is used to construct the core design variables. With response value Mapping relationships between (such as maximum stress and weight);

[0073] The mapping relationship is represented as follows:

[0074]

[0075] in, The predicted response value of the surrogate model. For design variables, This is the global trend function, which typically takes the form of a constant μ or a polynomial. For the regression coefficient vector, For random error term, It follows a normal distribution N(0, σ). z ²), The covariance matrix is ​​determined by the correlation between sample points;

[0076] The sample points are defined using the Gaussian Correlation Function. and sample points Correlation between ;

[0077] Sample points and sample points Correlation between Represented as:

[0078]

[0079] in, For the first Equivalent thickness or node thickness parameters for each region The dimensions of the core design variables For sample points In the The components of the dimension, For sample points In the The components of the dimension, For the first The relevant parameters of the dimension determine the rate of change of the surrogate model in different directions. The problem is solved using the Maximum Likelihood Estimation (MLE) method, specifically by solving the following optimization problem:

[0080]

[0081] in, Let R be the number of sample points, and R be the correlation matrix. This is an estimate of the process variance.

[0082] Solve this optimization problem using a genetic algorithm or pattern search method to determine the optimal solution. Value, complete the proxy model establishment.

[0083] Specifically, approximation methods include polynomial response surfaces, Kriging models, radial basis function neural networks, etc. The construction process includes model form selection, fitting, and evaluation of the model's prediction accuracy and generalization ability through methods such as cross-validation. High-precision surrogate models replace computationally expensive finite element analysis, providing a foundation for rapid optimization.

[0084] Furthermore, in S5, the initial solution must be a feasible solution and possess a certain degree of rationality to shorten the optimization iteration cycle. Initializing the simulated annealing algorithm parameters involves generating a new solution through small-amplitude local perturbations, increasing the node thickness in high-stress regions, and thinning the node thickness in low-stress regions, while ensuring that the thickness is not less than the minimum wall thickness. This forms the initial thickness scheme;

[0085] The initial thickness scheme generation process strictly satisfies the constraints of the ship node finite element model, ensuring that the thickness variation rate of adjacent columns of nodes does not exceed the maximum allowable thickness gradient, and that the node thickness perpendicular to the rolling direction remains consistent. After the strength and stiffness are verified by the finite element solver and the constraints are met, the lighter feasible solution is prioritized according to the objective function of the ship node finite element model. At high temperatures, a heavier feasible solution is accepted to escape local optima. The energy difference is determined using the Metropolis criterion. (Including constraint penalty terms), the acceptance probability is calculated. According to the reception probability The process is iterated until the temperature drops to the termination value, at which point the iteration stops, and all received feasible solutions are identified and recorded.

[0086] Energy difference Represented as:

[0087]

[0088] in, For the newly generated feasible solution, This is the current feasible solution;

[0089] Receive probability The calculation formula is:

[0090]

[0091] in, Boltzmann's constant, s This refers to the current annealing temperature parameter.

[0092] Specifically, This indicates the objective function value obtained after substituting the design scheme into the objective function. The simulated annealing algorithm, a heuristic global optimization algorithm, is used to optimize the established surrogate model. The simulated annealing algorithm simulates the solid annealing process. By introducing probabilistic perturbations and acceptance criteria, it can effectively escape local optima and find the global optimum or near-global optimum.

[0093] Although the invention has been described with reference to a limited number of embodiments, those skilled in the art will understand from the foregoing description that other embodiments are conceivable within the scope of the invention described herein. Furthermore, it should be noted that the language used in this specification has been chosen primarily for readability and edibility purposes, and not for the purpose of interpreting or limiting the subject matter of the invention. Therefore, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the appended claims. The disclosure of the invention is illustrative and not restrictive, and the scope of the invention is defined by the appended claims.

Claims

1. A basic proxy model and a method for lightweight ship design based on variable thickness steel plates, characterized in that, Includes the following steps: S1. Based on the geometric configuration of the longitudinal stiffening plate of the ship, a parametric finite element model of it is established using finite element software to obtain the finite element model of the longitudinal stiffening plate of the ship; S2. Define the core design variables of the finite element model of the longitudinal stiffening plate of the ship. Under the premise of satisfying the set constraints, construct the objective function of the finite element model of the longitudinal stiffening plate of the ship with the goal of minimizing the core design variables. S3. The core design variables are used as inputs to the finite element model of the longitudinal stiffening plate of the ship. The Latin hypercube sampling method is used for experimental design. All output response values ​​of the finite element model of the longitudinal stiffening plate of the ship are integrated to form a DOE sample database. S4. Based on the sample point data of the DOE sample database, a surrogate model is established by using the Kriging model and combining the random error term obtained by the correlation between sample points; S5. Employ simulated annealing algorithm to quickly evaluate a large number of design schemes on the surrogate model, ensuring that the design schemes meet the constraints and the objective function reaches its optimum. Based on the calculated acceptance probability, output the optimal design variable values.

2. The basic proxy model and the ship lightweight optimization design method for variable thickness steel plates according to claim 1, characterized in that, In S1, the longitudinal stiffening plate of the ship is composed of stiffening strips and panels. According to the arrangement of stiffening strips and panels, the geometric configuration of the longitudinal stiffening plate of the ship is discretized by selecting the corresponding element type using finite element software, and the overall structure of the longitudinal stiffening plate of the ship is meshed by finite element to complete the construction of the finite element model of the longitudinal stiffening plate of the ship. Subsequently, according to the load differences and functional requirements of each region in the longitudinal stiffening plate of the ship under actual working conditions, the corresponding initial thickness values ​​are assigned to different components.

3. The basic proxy model and the ship lightweighting optimization design method for variable thickness steel plates according to claim 2, characterized in that, In S2, the node thickness of each functional zone of the longitudinal stiffening plate of the ship is used as the core design variable. The node thickness is constrained and merged within intervals. Under the condition of satisfying the constraints, the objective function of the finite element model of the longitudinal stiffening plate of the ship is constructed with the goal of minimizing the total weight of the longitudinal stiffening plate. The specific steps are as follows: Based on the fact that the longitudinal stiffening plate of a ship comprises n variable thickness regions, the plate thickness and stiffening strip dimensions of each region are defined as core design variables. ; Core design variables Represented as: in, Indicates the first Node thickness parameters for each region Indicates matrix transpose; The objective function of the finite element model of the longitudinal stiffening plate of a ship is expressed as: in, The total weight of the longitudinal stiffening plates of the ship. For the first Density of regional materials, For the first Cross-sectional area of ​​each region For the first The length of each region; The constraints include strength and stability constraints, specifically: in, The maximum Von Mises equivalent stress of the structure under design load. For marine steel, For the maximum shear stress, For allowable shear stress, The critical buckling factor of the structure.

4. The basic proxy model and the ship lightweight optimization design method for variable thickness steel plates according to claim 3, characterized in that, In S3, the core design variables The total weight of the longitudinal stiffening plate is used as an input parameter in the finite element model of the ship's longitudinal stiffening plate. and maximum Von Mises equivalent stress As the output response; In the sampling process using the Latin hypercube sampling method, based on the core design variables... The dimension determines the number of sample points, ensuring that each core design variable... The values ​​are uniformly distributed within their intervals, and the sample points do not overlap. For each set of sample points, the finite element software is automatically driven to update the finite element model of the longitudinal stiffening plate of the ship, apply loads, and perform solution calculations, ultimately outputting the response values. A complete DOE sample database is formed. After sampling, abnormal samples that have not converged in finite element calculation, have abnormal fluctuations in response values, or have deviated significantly from the overall trend are removed to obtain the DOE sample database.

5. The basic proxy model and the ship lightweight optimization design method for variable thickness steel plates according to claim 4, characterized in that, In S4, the core design variables are constructed using a Kriging model. With response value The mapping relationship between them; The mapping relationship is represented as follows: in, The predicted response value of the surrogate model. For design variables, For global trend functions, For the regression coefficient vector, For random error term, It follows a normal distribution N(0, σ). z ²), The covariance matrix is ​​determined by the correlation between sample points; Sample points are defined using the Gaussian correlation function. and sample points Correlation between ; Sample points and sample points Correlation between Represented as: in, For the first Node thickness parameters for each region The dimensions of the core design variables For sample points In the The components of the dimension, For sample points In the The components of the dimension, For the first Dimensional parameters.

6. The basic proxy model and the ship lightweight optimization design method for variable thickness steel plates according to claim 5, characterized in that, In step S5, initializing the simulated annealing algorithm parameters involves generating a new solution through small-amplitude local perturbations. This increases the node thickness in high-stress regions and thins the node thickness in low-stress regions, ensuring the thickness is not less than the minimum wall thickness. This forms the initial thickness scheme; During the initial thickness scheme generation process, the constraints of the ship node finite element model are satisfied, ensuring that the thickness change rate of adjacent columns of nodes does not exceed the maximum allowable thickness gradient, and that the node thickness perpendicular to the rolling direction remains consistent. After the strength and stiffness are analyzed and verified by the finite element solver and the constraints are met, according to the objective function of the ship node finite element model, the lighter feasible solution is preferentially accepted. At high temperatures, the heavier feasible solution is accepted to escape local optima. Based on the simulated annealing algorithm, the energy difference is determined using the Metropolis criterion. The reception probability is calculated. According to the reception probability The process is iterated until the temperature drops to the termination value, at which point the iteration stops, and all received feasible solutions are identified and recorded. Energy difference Represented as: in, For the newly generated feasible solution, This is the current feasible solution; Receive probability The calculation formula is: in, Boltzmann's constant, s This refers to the current annealing temperature parameter.

Images