Ship model optimization method and device, electronic equipment and storage medium

Abstract

The invention provides a ship model optimization method and device, electronic equipment and a storage medium, and the method comprises the steps: taking a trained BP neural network as a fitness function of adaptive PSO, and employing the powerful nonlinear mapping capability of the BP neural network to achieve the adaptive PSO; the method can deeply learn and accurately model the relationship between a plurality of complex factors such as fluid dynamic characteristics and navigation conditions and an optimization target, and greatly improves the optimization precision. A multi-objective problem is converted into single-objective optimization through a reasonable weight distribution or objective normalization method, or a Pareto optimal solution set is searched by adopting a multi-objective particle swarm optimization evolutionary strategy. According to the method, the importance of each target can be comprehensively considered in the search process, other key indexes are prevented from being ignored due to the leading of a single target, a more comprehensive and reasonable optimal parameter combination scheme is provided for ship design, and the multi-target processing capability is enhanced.

Description

Technical Field

[0001] This application relates to the field of marine engineering technology, and in particular to a method, apparatus, electronic device and storage medium for optimizing ship models. Background Technology

[0002] Ship model optimization is a core issue in the field of naval engineering, aiming to achieve synergistic optimization of multiple objectives such as propulsion efficiency, hull weight, manufacturing cost, and structural strength by adjusting design parameters.

[0003] To address the problem of traditional BP neural networks easily getting trapped in local optima in ship parameter prediction, a particle swarm optimization algorithm with an adaptive mutation operator is introduced to optimize the initial weights and thresholds of the neural network, thereby improving prediction accuracy and generalization ability.

[0004] However, existing technical solutions have limitations such as single-objective focus, high computational complexity and strong data dependence. They cannot involve multi-objective collaborative optimization, still rely on traditional fitness function design, and are difficult to accurately map the relationship between parameters and performance. Moreover, the computational complexity increases significantly with the number of objectives. Summary of the Invention

[0005] This application provides a ship model optimization method, apparatus, electronic device, and storage medium to address the limitations of existing solutions, such as single-objective focus, high computational complexity, and strong data dependence.

[0006] Firstly, this application provides a ship model optimization method, including: Based on historical ship data, a backpropagation neural network model is constructed and trained. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of the adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters. Initialize the particle swarm, and perform the following steps in each iteration until the set iteration termination condition is met: The current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and the current fitness of each particle is determined based on the prediction performance index. Based on the current fitness of each particle, update the individual historical best position of each particle and the collective historical best position of the particle swarm. Update the position of each particle based on the updated individual historical best position and the group historical best position; Based on the optimization information recorded during the iteration process, the final optimized combination of target ship design parameters is determined.

[0007] In one possible implementation, determining the fitness of each particle based on the predicted performance metric includes: Assign weight values ​​to each of the predicted performance metrics; For each particle, the fitness of the particle is obtained by weighting and summing the values ​​of multiple predicted performance indicators corresponding to the particle according to the weight value.

[0008] In one possible implementation, updating the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle includes: If the current fitness of a particle is better than the fitness corresponding to the particle's individual historical best position, then the particle's individual historical best position is updated using the particle's current position. If the current fitness of a particle is better than the fitness corresponding to the group's historical best position, then the group's historical best position is updated using the particle's current position.

[0009] In one possible implementation, updating the position of each particle based on the updated individual historical best position and the group historical best position includes: Based on the updated individual historical best position and the group historical best position, the velocity change of each particle is determined, and the position of each particle is updated based on the velocity change.

[0010] In one possible implementation, the determination of the velocity change of each particle based on the updated individual historical best position and the group historical best position is achieved through the following formula: ; in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the first The group's historical optimal position in 3D space; The position of each particle is updated based on the velocity change, which is achieved by the following formula: ; in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

[0011] In one possible implementation, determining the final target ship design parameter combination based on the optimization information recorded during the iteration process includes: The candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met is determined as the final target ship design parameter combination.

[0012] In one possible implementation, the optimization information is a Pareto optimal solution set, and the determination of the final target ship design parameter combination based on the optimization information recorded during the iteration process includes: The final combination of target ship design parameters is determined from the Pareto optimal solution set.

[0013] In one possible implementation, determining the final target ship design parameter combination from the Pareto optimal solution set includes: Calculate a comprehensive evaluation value for each scheme in the Pareto optimal solution set, and select the scheme with the best comprehensive evaluation value as the final target ship design parameter combination; or, The system receives a user selection instruction and selects the corresponding scheme from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction.

[0014] Secondly, this application provides a ship model optimization device, comprising: A construction module is used to build and train a backpropagation neural network model based on historical ship data. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. The embedding module is used to embed the trained backpropagation neural network model into the fitness function of the adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters. An initialization module is used to initialize the particle swarm and perform the following steps in each iteration until the set iteration termination condition is met: The first determining module is used to input the current position of each particle into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and determine the current fitness of each particle based on the prediction performance index. The first update module is used to update the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle. The second update module is used to update the position of each particle based on the updated individual historical best position and the group historical best position; The second determining module is used to determine the final optimized combination of target ship design parameters based on the optimization information recorded during the iteration process.

[0015] Thirdly, this application provides an apparatus comprising: a processor and a memory, the processor being configured to execute a ship model optimization program stored in the memory to implement the ship model optimization method described in any one of the first aspects.

[0016] Fourthly, this application provides a storage medium storing one or more programs that can be executed by one or more processors to implement the ship model optimization method described in any one aspect.

[0017] Compared with the prior art, the technical solution provided in this application has the following advantages: The method provided in this application constructs and trains a backpropagation neural network model based on historical ship data. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of an adaptive particle swarm optimization algorithm, wherein the position of each particle in the particle swarm represents a set of candidate ship design parameters. The particle swarm is initialized, and the following steps are executed in each iteration until the set iteration termination condition is met: the current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding predicted performance index, and the current fitness of each particle is determined based on the predicted performance index; the individual historical best position of each particle and the collective historical best position of the particle swarm are updated according to the current fitness of each particle; the position of each particle is updated according to the updated individual historical best position and collective historical best position; and the final optimized target ship design parameter combination is determined based on the optimization information recorded during the iteration process. Using the trained BP neural network as the fitness function of the particle swarm algorithm, the BP neural network, with its powerful nonlinear mapping capability, can deeply learn and accurately model the relationship between these complex factors and the optimization objective. By employing reasonable weight allocation or objective normalization methods, multi-objective problems are transformed into single-objective optimizations, or a multi-objective particle swarm optimization (PSO) algorithm evolutionary strategy is used to find Pareto optimal solutions, thus enhancing multi-objective processing capabilities. Considering the characteristics of ship model parameters, the velocity and position update formulas of the PSO algorithm are improved, and boundary handling mechanisms and constraint handling methods are introduced to ensure the feasibility and practicality of the found optimal parameter combinations. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0021] Figure 1 A flowchart illustrating an embodiment of a ship model optimization method provided in this application; Figure 2 A flowchart illustrating an embodiment of another ship model optimization method provided in this application; Figure 3 A block diagram illustrating an embodiment of a ship model optimization device provided in this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0024] To address the limitations of existing solutions, such as single-objective focus, high computational complexity, and strong data dependence, which prevent multi-objective collaborative optimization and reliance on traditional fitness function design, making it difficult to accurately map the relationship between parameters and performance, and significantly increasing computational complexity with the number of objectives, this application provides a ship model optimization method. This method utilizes a trained BP neural network as the fitness function of a particle swarm optimization (PSO) algorithm. The BP neural network, with its powerful nonlinear mapping capabilities, can deeply learn and accurately model the relationship between these complex factors and the optimization objectives. By employing reasonable weight allocation or objective normalization methods, the multi-objective problem is transformed into single-objective optimization, or a multi-objective PSO evolutionary strategy is used to find the Pareto optimal solution set, enhancing multi-objective processing capabilities. Furthermore, considering the characteristics of ship model parameters, the velocity and position update formulas of the PSO algorithm are improved, and boundary handling mechanisms and constraint handling methods are introduced to ensure the feasibility and practicality of the found optimal parameter combinations.

[0025] Figure 1 This is a flowchart illustrating an embodiment of a ship model optimization method provided in this application. Figure 1 As shown, the method includes the following steps: S101. Based on historical ship data, construct and train a backpropagation neural network model. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index.

[0026] This application is primarily applied to ship model optimization in the field of naval architecture. By constructing a tightly coupled framework of a Back Propagation Neural Network (BP) and Self-Adaptive Particle Swarm Optimization (PSO), the trained BP neural network is directly embedded into the fitness function of the PSO to adjust ship design parameters, achieving coordinated optimization of multiple objectives such as propulsion efficiency, hull weight, manufacturing cost, and structural strength. Specifically, the system can include four main functional units: a data preprocessing unit, a neural network training unit, an adaptive particle swarm optimization unit, and a multi-objective decision-making unit. These units interact through data flow and control flow, forming a closed-loop optimization system.

[0027] In this embodiment, the system first acquires a file containing the ship's original parameters and performance indicators, and checks the file's validity to determine whether it is compliant and meets the required number of parameters and target values. When the file is deemed valid, the system reads the data from the file, acquiring the ship's original parameters and performance indicators as historical data for the ship model to be optimized, and verifies the data range. The system then optimizes the data within the specified range. The data preprocessing module divides the ship model's original parameters and performance indicators, after data detection and range verification, into training and test sets using a preset ratio of 7:3 through simple random partitioning. Then, through normalization, the numerical ranges of different features are adjusted to the same scale, eliminating differences in data dimensions between different features and improving model performance and training speed.

[0028] Furthermore, a three-layer backpropagation (BP) neural network is constructed, consisting of an input layer, a hidden layer, and an output layer. The design parameters of the ship model to be optimized are used as input, and the simulated raw data indicators obtained after optimization are used as output. Using the backpropagation algorithm, the error between the simulated raw data output by the BP neural network and the actual data is calculated. This error is then propagated back to each layer of the BP neural network, and the weights of each layer are adjusted according to the error. This process trains the weights of the BP neural network and optimizes it, establishing a nonlinear mapping relationship between ship parameters and at least one performance indicator.

[0029] S102. Embed the trained backpropagation neural network model into the fitness function of the adaptive particle swarm optimization algorithm.

[0030] In this embodiment, after obtaining the trained BP neural network, the trained BP neural network is embedded into the adaptive PSO fitness function; that is, the BP neural network is used as the fitness function. Specifically, the BP neural network constructs a high-precision prediction model by learning the complex nonlinear mapping relationship between ship parameters and at least one performance index. The particles in the particle swarm are updated through adaptive PSO to make them converge, with the aim of finding the optimal value within a certain range. Here, the position of each particle in the particle swarm represents a set of candidate ship design parameters.

[0031] S103. Initialize the particle swarm and perform the following steps in each iteration until the set iteration termination condition is met.

[0032] In this embodiment, adaptive PSO initializes a particle swarm matrix where each particle element is a random value within a certain range. These elements are then iteratively mapped to ship design parameters. These ship design parameters are input into a backpropagation (BP) neural network for iteration. During each iteration, the particles converge to a single point until a predetermined termination condition is met, thus terminating the iteration and obtaining the particle corresponding to the historical optimal position of the swarm. This yields the candidate ship design parameter combination corresponding to the particle at the historical optimal position of the swarm.

[0033] S104. Input the current position of each particle into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and determine the current fitness of each particle based on the prediction performance index.

[0034] In this embodiment, the current position of each particle in the aforementioned particle swarm is input into a trained BP neural network model. Based on the current position of each particle, the BP neural network model utilizes a complex nonlinear mapping relationship between constructed ship parameters and at least one performance index to obtain the predicted performance index corresponding to each particle. Then, based on the predicted performance index, weights are assigned to the predicted performance index using a weighted method, and the functional mapping relationship between the predicted performance index and fitness is determined according to a preset fitness evaluation rule, thereby determining the current fitness of each particle. Fitness directly reflects the comprehensive performance of this set of ship design parameters in terms of optimization objectives such as propulsion efficiency, hull weight, manufacturing cost, and structural strength. A higher fitness value indicates that the combination of ship design parameters is closer to the optimization objective.

[0035] S105. Based on the current fitness of each particle, update the individual historical best position of each particle and the collective historical best position of the particle swarm.

[0036] In the embodiments of this application, during the iterative process of the particle swarm optimization algorithm, it is necessary to update the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle.

[0037] Specifically, if a particle's current fitness is better than the fitness corresponding to its individual best historical position, then the particle's current position is used to update its individual best historical position; if a particle's current fitness is better than the fitness corresponding to the swarm's collective best historical position, then the particle's current position is used to update the swarm's collective best historical position.

[0038] By continuously updating the individual historical best position of each particle and the collective historical best position of the particle swarm, each particle can use its own experience and the collective experience to adjust its search direction, thereby driving the entire particle swarm to gradually approach the optimal solution.

[0039] S106. Update the position of each particle based on the updated individual historical best position and the group historical best position.

[0040] In this embodiment, during the particle swarm search of the solution space, in order to achieve a balance between global and local searches, the inertia weight is typically dynamically adjusted to balance the global search. The inertia weight controls the degree to which particles inherit their own velocity during the search process. Specifically, based on the updated individual historical best position and the group's historical best position, the velocity change of the particles in the search space is determined, and the positions of each particle in the particle swarm are updated according to this velocity change.

[0041] In an optional embodiment of this application, the velocity change of each particle in the search space is dynamically adjusted using a velocity update formula, thus determining the particle's velocity; the position of the particle in the search space is updated using a position update formula. These velocity and position update formulas enable the particle to progressively search for the optimal solution in the solution space.

[0042] S107. Based on the optimization information recorded during the iteration process, determine the final optimized combination of target ship design parameters.

[0043] In this embodiment, after the adaptive PSO completes its iteration, based on the changes in the historical optimal positions of each particle and the swarm's historical optimal positions recorded during the iteration process, the candidate ship design parameter combination corresponding to the swarm's historical optimal position at the time of iteration termination is obtained and determined as the final target ship design parameter combination. Specifically, when there is only one performance index, the optimization information recorded during the iteration process is the candidate ship design parameter combination corresponding to the swarm's historical optimal position at the time of iteration termination; when there are multiple performance indices, the optimization information recorded during the iteration process is the Pareto optimal solution set, from which the final target ship design parameter combination is determined. This achieves support for dynamic multi-objective trade-offs and the generation of Pareto optimal solution sets, while reducing computational costs and achieving more efficient global optimization.

[0044] The technical solution provided in this application involves constructing and training a backpropagation neural network model based on historical ship data. This model establishes a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of an adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters. The particle swarm is initialized, and the following steps are executed in each iteration until a set iteration termination condition is met: the current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding predicted performance index, and the current fitness of each particle is determined based on the predicted performance index; the individual historical best position of each particle and the collective historical best position of the particle swarm are updated according to the current fitness of each particle; the position of each particle is updated according to the updated individual and collective historical best positions; and the final optimized combination of target ship design parameters is determined based on the optimization information recorded during the iteration process. Using the trained BP neural network as the fitness function of the particle swarm algorithm, the BP neural network, with its powerful nonlinear mapping capability, can deeply learn and accurately model the relationship between these complex factors and the optimization objective. By employing reasonable weight allocation or objective normalization methods, multi-objective problems are transformed into single-objective optimizations, or a multi-objective particle swarm optimization (PSO) algorithm evolutionary strategy is used to find Pareto optimal solutions, thus enhancing multi-objective processing capabilities. Considering the characteristics of ship model parameters, the velocity and position update formulas of the PSO algorithm are improved, and boundary handling mechanisms and constraint handling methods are introduced to ensure the feasibility and practicality of the found optimal parameter combinations.

[0045] Figure 2 A flowchart illustrating an embodiment of another ship model optimization method provided in this application. Figure 2 The process shown includes the following steps: S201. Based on historical ship data, construct and train a backpropagation neural network model. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index.

[0046] In this embodiment, the system first acquires historical data of the ship model to be optimized. This historical data may include the design parameters and performance indicators of the ship model. The system verifies the data range of the design parameters and performance indicators of the ship model and retains the design parameters and performance indicators of the ship model within a preset data range. The retained design parameters and performance indicators of the ship model within the preset data range are then divided into training and testing sets according to a preset ratio. For example, the training and testing sets can be divided into training and testing sets using a simple random partitioning method at a preset ratio of 7:3. The data is then normalized to adjust the numerical ranges of different features to the same scale, eliminating the differences in data dimensions between different features, thus facilitating the subsequent improvement of the stability of the BP neural network training.

[0047] Furthermore, a three-layer backpropagation (BP) neural network is constructed, consisting of an input layer, a hidden layer, and an output layer. The design parameters of the ship model to be optimized are used as input, and the simulated raw data indicators obtained after optimization are used as output. Using the backpropagation algorithm, the error between the simulated raw data output by the BP neural network and the actual data is calculated. This error is then propagated back to each layer of the BP neural network, and the weights of each layer are adjusted according to the error. This process trains the weights of the BP neural network and optimizes it, establishing a nonlinear mapping relationship between ship parameters and at least one performance indicator.

[0048] In this embodiment, the hidden layer in the BP neural network is located between the input and output layers, and its function is to extract and transform features from the design parameter data of the input ship model to be optimized. The tanh function is used as the activation function in the hidden layer, introducing nonlinear factors to enable the neural network to learn and fit complex nonlinear relationships. The training algorithm uses the Adam optimizer, dynamically adjusting the learning rate based on gradient information during training to accelerate the model's convergence speed. Mean squared error (MSE) is used as the loss function during the BP neural network training process to measure the difference between the model's predicted values ​​and the true values. The smaller the value of the loss function, the closer the model's predicted values ​​are to the true values, and the better the model's performance. The mean squared error (MSE) is as follows:

[0049] in, For the true value, For predicted values, Let be the number of samples. For each sample, calculate its true value. Compared with the predicted value The difference between them, and then for this difference The mean square error is calculated by summing the squared errors of all samples and then dividing by the sample size n.

[0050] In this embodiment, the model parameters are adjusted using the Adam optimization algorithm by calculating the gradient of the loss function, thereby gradually reducing the value of the loss function and obtaining a trained BP neural network. The BP neural network, trained using the backpropagation algorithm, can learn the nonlinear mapping relationship between ship design parameters and at least one performance index.

[0051] S202. Embed the trained backpropagation neural network model into the fitness function of the adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters.

[0052] In this embodiment of the application, S202 and Figure 1 The steps S102 in the corresponding embodiments are basically the same, and will not be described again here.

[0053] S203. Initialize the particle swarm and perform the following steps in each iteration until the set iteration termination condition is met.

[0054] In this embodiment of the application, S203 and Figure 1 The steps in the corresponding embodiments are basically the same, and will not be described again here.

[0055] S204. Input the current position of each particle into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and assign a weight value to each prediction performance index.

[0056] In this embodiment, the current position of each particle in the aforementioned particle swarm is input into a trained BP propagation neural network model. This BP neural network model fully learns the complex nonlinear mapping relationship between ship design parameters and multiple key performance indicators, possessing strong generalization ability and prediction accuracy. Therefore, after inputting the current position of each particle, the BP neural network model can output the predicted performance indicator corresponding to each particle.

[0057] Furthermore, after obtaining the predicted performance indicators of each particle, the aforementioned multi-dimensional performance indicators are further converted into fitness values, thereby supporting the iterative optimization of the particle swarm optimization algorithm. Specifically, in this embodiment, a weighted method is used to assign corresponding weight coefficients to each predicted performance indicator based on the importance of actual engineering needs and design goals, so as to facilitate the subsequent comprehensive evaluation of each predicted performance indicator. The weight coefficients can be adjusted according to the priority or importance of the optimization goal.

[0058] S205. For each particle, according to the weight value, the values ​​of multiple prediction performance indicators corresponding to the particle are weighted and summed to obtain the particle's fitness.

[0059] In this embodiment, after assigning corresponding weight coefficients to each predicted performance index using a weighted method, the values ​​of multiple predicted performance indices corresponding to a particle are weighted and summed to obtain the particle's fitness. Based on the current fitness of each particle, a basis is provided for subsequent particle velocity updates, position updates, and global optimal solution searches, thereby driving the optimization process to continuously converge towards the optimal ship design parameter scheme.

[0060] S206. If the current fitness of a particle is better than the fitness corresponding to its historical best position, then update the historical best position of the particle using its current position.

[0061] S207. If the current fitness of a particle is better than the fitness corresponding to the best historical position of the particle swarm, then the best historical position of the particle swarm is updated using the current position of the particle.

[0062] The following is a unified discussion of S206-S207.

[0063] In the embodiments of this application, during the iterative process of the particle swarm optimization algorithm, it is necessary to update the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle.

[0064] Specifically, the current fitness of each particle in the particle swarm is compared with the fitness corresponding to its individual best historical position. Based on a preset fitness criterion, if the current fitness of a particle is better than the fitness corresponding to its individual best historical position, the fitness of that particle is updated. Simultaneously, the fitness of the entire particle swarm is compared with the fitness corresponding to the collective best historical position of the particle swarm. Again, based on a preset fitness criterion, if the fitness of the entire particle swarm is better than the fitness corresponding to the collective best historical position of the particle swarm, the fitness of the collective best historical position is updated. Based on the updated fitness corresponding to both individual and collective best historical positions, the individual best historical position of each particle and the collective best historical position of the particle swarm are updated.

[0065] By comparing the current fitness of each particle with the fitness corresponding to the individual historical best position of each particle and the collective historical best position of the particle swarm, the individual historical best position of each particle and the collective historical best position of the particle swarm are continuously updated, thereby driving the entire particle swarm to gradually approach the optimal solution.

[0066] S208. Based on the updated individual historical best position and the group historical best position, determine the velocity change of each particle, and update the position of each particle based on the velocity change.

[0067] In this embodiment, to achieve a balance between global and local search, the system determines the movement direction and step size of particles in the search space using a velocity update formula. Then, based on the current velocity of each particle in the particle swarm and the updated individual and group historical best positions, the system determines the velocity change of each particle. Finally, based on the current position and the updated velocities of each particle, the position of each particle is updated using a position update formula.

[0068] In an optional embodiment of this application, the velocity change of each particle is determined based on the updated individual historical best position and the group historical best position, using the following formula:

[0069] in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the th The collective historical best position in 3D space. This velocity update formula integrates three parts of information: the particle's current velocity, the difference between the individual's historical best position and the current position, and the difference between the collective historical best position and the current position, making the particle's motion tendency more comprehensive and reasonable.

[0070] The position of each particle is updated based on the change in velocity, using the following formula:

[0071] in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

[0072] The above velocity update formula and position update formula ensure that the particle always meets the actual requirements of ship design during the search process, and ensure that the optimal parameter combination found is feasible and practical.

[0073] S209. When there is only one performance index, the candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met is determined as the final target ship design parameter combination.

[0074] In this embodiment, when the performance index to be optimized is a single objective, the optimization process will iteratively search for that single performance index. During the iteration process, the ship design parameter combination corresponding to the group's historical best position will gradually converge towards the theoretical optimal solution. When the iteration termination condition is met, the candidate ship design parameter combination corresponding to the group's historical best position is the candidate parameter combination that optimizes the single performance index within the entire search space. This candidate parameter combination is then determined as the final target ship design parameter combination.

[0075] S210. When there are multiple performance indicators, the optimization information is the Pareto optimal solution set, and the final combination of target ship design parameters is determined from the Pareto optimal solution set.

[0076] In this embodiment of the application, when there are multiple performance indicators to be optimized, the Pareto optimal solution set is used as the optimization information recorded during the iteration process. Based on the multiple predicted performance indicators corresponding to each particle in the above optimal solution set, thresholds are set for each performance indicator according to the design task and weights are set according to engineering needs, so as to determine the final combination of target ship design parameters from the Pareto optimal solution set.

[0077] In the optional scheme of Embodiment 1 of this application, a comprehensive evaluation value is calculated for each scheme in the Pareto optimal solution set, and the scheme with the best comprehensive evaluation value is selected as the final target ship design parameter combination. By weighing the importance of different performance indicators, the optimal parameter combination is output comprehensively. Alternatively, a user selection instruction is received, and the corresponding scheme is selected from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction. Thus, the importance of each objective can be fully considered during the search process, avoiding the problem of neglecting other key indicators due to the dominance of a single objective, providing a more comprehensive and reasonable optimal parameter combination scheme for ship design, and enhancing multi-objective processing capabilities.

[0078] Figure 2The illustrated process provides a method for ship model optimization. By using a trained BP neural network as the fitness function of an adaptive PSO (Particle Swarm Optimization) algorithm, and leveraging the powerful nonlinear mapping capabilities of the BP neural network, it can deeply learn and accurately model the relationship between numerous complex factors such as hydrodynamic characteristics and navigation conditions and the optimization objective, greatly improving optimization accuracy. Through reasonable weight allocation or objective normalization methods, the multi-objective problem is transformed into single-objective optimization, or a multi-objective particle swarm optimization (PSO) evolutionary strategy is employed to find the Pareto optimal solution set. This ensures that the importance of each objective is comprehensively considered during the search process, avoiding the neglect of other key indicators due to the dominance of a single objective, providing a more comprehensive and reasonable optimal parameter combination scheme for ship design, and enhancing multi-objective processing capabilities.

[0079] Figure 3 This is a block diagram illustrating an embodiment of a ship model optimization device provided in this application. Figure 3 As shown, the device includes: Module 301 is used to construct and train a backpropagation neural network model based on historical ship data. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. Embedding module 302 is used to embed the trained backpropagation neural network model into the fitness function of the adaptive particle swarm optimization algorithm, wherein the position of each particle in the particle swarm represents a set of candidate ship design parameters. Initialization module 303 is used to initialize the particle swarm and perform the following steps in each iteration until the set iteration termination condition is met: The first determining module 304 is used to input the current position of each particle into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and determine the current fitness of each particle based on the prediction performance index. The first update module 305 is used to update the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle. The second update module 306 is used to update the position of each particle based on the updated individual historical best position and the group historical best position; The second determining module 307 is used to determine the final optimized combination of target ship design parameters based on the optimization information recorded during the iteration process.

[0080] In one possible implementation, the first determining module 304 is specifically used to assign a weight value to each of the predicted performance indicators; for each particle, according to the weight value, the values ​​of the multiple predicted performance indicators corresponding to the particle are weighted and summed to obtain the fitness of the particle.

[0081] In one possible implementation, the first update module 305 is specifically configured to update the individual historical best position of a particle using its current position if the current fitness of the particle is better than the fitness corresponding to the individual historical best position of the particle; and to update the group historical best position of the particle swarm using its current position if the current fitness of the particle is better than the fitness corresponding to the group historical best position of the particle swarm.

[0082] In one possible implementation, the second update module 306 is specifically used to determine the velocity change of each particle based on the updated individual historical best position and the group historical best position, and update the position of each particle based on the velocity change.

[0083] In one possible implementation, the determination of the velocity change of each particle based on the updated individual historical best position and the group historical best position is achieved through the following formula: ; in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the first The group's historical optimal position in 3D space; The position of each particle is updated based on the velocity change, which is achieved by the following formula: ; in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

[0084] In one possible implementation, the second determining module 307 is specifically used to determine the candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met as the final target ship design parameter combination.

[0085] In one possible implementation, the optimization information is a Pareto optimal solution set, and the second determining module 307 is specifically used to determine the final target ship design parameter combination from the Pareto optimal solution set.

[0086] In one possible implementation, the second determining module 307 is specifically used to calculate a comprehensive evaluation value for each scheme in the Pareto optimal solution set, and select the scheme with the best comprehensive evaluation value as the final target ship design parameter combination; or, to receive a user selection instruction and select the corresponding scheme from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction.

[0087] like Figure 4 As shown in the figure, this application provides a device including a processor 411, a communication interface 412, a memory 413, and a communication bus 414, wherein the processor 411, the communication interface 412, and the memory 413 communicate with each other through the communication bus 414. Memory 413 is used to store computer programs; In one embodiment of this application, when the processor 411 executes the program stored in the memory 413, it implements the ship model optimization method provided in any of the foregoing method embodiments, including: Based on historical ship data, a backpropagation neural network model is constructed and trained. This model establishes a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of an adaptive particle swarm optimization algorithm, where the position of each particle in the swarm represents a set of candidate ship design parameters. The particle swarm is initialized, and the following steps are performed in each iteration until a set iteration termination condition is met: the current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding predicted performance index, and the current fitness of each particle is determined based on the predicted performance index; the individual historical best position of each particle and the collective historical best position of the particle swarm are updated according to the current fitness of each particle; the position of each particle is updated according to the updated individual historical best position and collective historical best position; and the final optimized combination of target ship design parameters is determined based on the optimization information recorded during the iteration process.

[0088] In one possible implementation, a weight value is assigned to each of the predicted performance metrics; for each particle, the values ​​of the multiple predicted performance metrics corresponding to the particle are weighted and summed according to the weight value to obtain the fitness of the particle.

[0089] In one possible implementation, if the current fitness of a particle is better than the fitness corresponding to the particle's individual historical best position, then the particle's individual historical best position is updated using the particle's current position; if the current fitness of a particle is better than the fitness corresponding to the group historical best position of the particle swarm, then the group historical best position of the particle swarm is updated using the particle's current position.

[0090] In one possible implementation, the velocity change of each particle is determined based on the updated individual historical best position and the group historical best position, and the position of each particle is updated based on the velocity change.

[0091] In one possible implementation, the determination of the velocity change of each particle based on the updated individual historical best position and the group historical best position is achieved through the following formula: ; in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the first The group's historical optimal position in 3D space; The position of each particle is updated based on the velocity change, which is achieved by the following formula: ; in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

[0092] In one possible implementation, the candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met is determined as the final target ship design parameter combination.

[0093] In one possible implementation, the optimization information is a Pareto optimal solution set, from which the final target ship design parameter combination is determined.

[0094] In one possible implementation, a comprehensive evaluation value is calculated for each scheme in the Pareto optimal solution set, and the scheme with the best comprehensive evaluation value is selected as the final target ship design parameter combination; or, a user selection instruction is received, and a corresponding scheme is selected from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction.

[0095] This application embodiment also provides a storage medium (computer-readable storage medium). This storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; it may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; it may also include combinations of the above types of memory. When one or more programs in the storage medium can be executed by one or more processors to implement the ship model optimization method executed on the ship model optimization device side, the processor is used to execute the ship model optimization program stored in the memory to implement the following steps of executing the ship model optimization method on the ship model optimization device side: Based on historical ship data, a backpropagation neural network model is constructed and trained. This model establishes a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of an adaptive particle swarm optimization algorithm, where the position of each particle in the swarm represents a set of candidate ship design parameters. The particle swarm is initialized, and the following steps are performed in each iteration until a set iteration termination condition is met: the current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding predicted performance index, and the current fitness of each particle is determined based on the predicted performance index; the individual historical best position of each particle and the collective historical best position of the particle swarm are updated according to the current fitness of each particle; the position of each particle is updated according to the updated individual historical best position and collective historical best position; and the final optimized combination of target ship design parameters is determined based on the optimization information recorded during the iteration process.

[0096] In one possible implementation, a weight value is assigned to each of the predicted performance metrics; for each particle, the values ​​of the multiple predicted performance metrics corresponding to the particle are weighted and summed according to the weight value to obtain the fitness of the particle.

[0097] In one possible implementation, if the current fitness of a particle is better than the fitness corresponding to the particle's individual historical best position, then the particle's individual historical best position is updated using the particle's current position; if the current fitness of a particle is better than the fitness corresponding to the group historical best position of the particle swarm, then the group historical best position of the particle swarm is updated using the particle's current position.

[0098] In one possible implementation, the velocity change of each particle is determined based on the updated individual historical best position and the group historical best position, and the position of each particle is updated based on the velocity change.

[0099] In one possible implementation, the determination of the velocity change of each particle based on the updated individual historical best position and the group historical best position is achieved through the following formula: ; in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the first The group's historical optimal position in 3D space; The position of each particle is updated based on the velocity change, which is achieved by the following formula: ; in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

[0100] In one possible implementation, the candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met is determined as the final target ship design parameter combination.

[0101] In one possible implementation, the optimization information is a Pareto optimal solution set, from which the final target ship design parameter combination is determined.

[0102] In one possible implementation, a comprehensive evaluation value is calculated for each scheme in the Pareto optimal solution set, and the scheme with the best comprehensive evaluation value is selected as the final target ship design parameter combination; or, a user selection instruction is received, and a corresponding scheme is selected from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction.

[0103] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0105] It should be understood that the terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “described” as used herein may also include the plural forms. The terms “comprising,” “including,” “containing,” and “having” are inclusive and therefore indicate the presence of the stated features, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, elements, components, and / or combinations thereof. The method steps, processes, and operations described herein are not construed as requiring them to be performed in a particular order described or illustrated unless the order of performance is explicitly indicated. It should also be understood that additional or alternative steps may be used.

[0106] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for optimizing a ship model, characterized in that, The method includes: Based on historical ship data, a backpropagation neural network model is constructed and trained. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. The trained backpropagation neural network model is embedded into the fitness function of the adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters. Initialize the particle swarm, and perform the following steps in each iteration until the set iteration termination condition is met: The current position of each particle is input into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and the current fitness of each particle is determined based on the prediction performance index. Based on the current fitness of each particle, update the individual historical best position of each particle and the collective historical best position of the particle swarm. Update the position of each particle based on the updated individual historical best position and the group historical best position; Based on the optimization information recorded during the iteration process, the final optimized combination of target ship design parameters is determined.

2. The method according to claim 1, characterized in that, The determination of the fitness of each particle based on the predicted performance index includes: Assign weight values ​​to each of the predicted performance metrics; For each particle, the fitness of the particle is obtained by weighting and summing the values ​​of multiple predicted performance indicators corresponding to the particle according to the weight value.

3. The method according to claim 1, characterized in that, The step of updating the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle includes: If the current fitness of a particle is better than the fitness corresponding to the particle's individual historical best position, then the particle's individual historical best position is updated using the particle's current position. If the current fitness of a particle is better than the fitness corresponding to the group's historical best position, then the group's historical best position is updated using the particle's current position.

4. The method according to claim 1, characterized in that, The step of updating the position of each particle based on the updated individual historical best position and the group historical best position includes: Based on the updated individual historical best position and the group historical best position, the velocity change of each particle is determined, and the position of each particle is updated based on the velocity change.

5. The method according to claim 4, characterized in that, The velocity change of each particle is determined based on the updated individual historical best position and the group historical best position, using the following formula: ； in, Characterizing the first The particle in the first Dimensional space Speed ​​at the next iteration For inertial weights, , As a learning factor, , It is a random number. Characterizing the first The particle in the first The optimal historical position of an individual in 3D space. Characterizing the particle swarm in the first The group's historical optimal position in 3D space; The position of each particle is updated based on the velocity change, which is achieved by the following formula: ； in, The position of the i-th particle in the d-th dimension during the t-th iteration is represented. Characterizing the first The particle in the first dimensional space ( ) Speed ​​during the next iteration.

6. The method according to claim 1, characterized in that, The determination of the final target ship design parameter combination based on the optimization information recorded during the iteration process includes: The candidate ship design parameter combination corresponding to the group's historical best position when the iteration termination condition is met is determined as the final target ship design parameter combination.

7. The method according to claim 1, characterized in that, The optimization information is a Pareto optimal solution set. The determination of the final target ship design parameter combination based on the optimization information recorded during the iteration process includes: The final combination of target ship design parameters is determined from the Pareto optimal solution set.

8. The method according to claim 6, characterized in that, Determining the final target ship design parameter combination from the Pareto optimal solution set includes: Calculate a comprehensive evaluation value for each scheme in the Pareto optimal solution set, and select the scheme with the best comprehensive evaluation value as the final target ship design parameter combination; or, The system receives a user selection instruction and selects the corresponding scheme from the Pareto optimal solution set as the final target ship design parameter combination according to the instruction.

9. A ship model optimization device, characterized in that, The device includes: A construction module is used to build and train a backpropagation neural network model based on ship historical data. The backpropagation neural network model is used to establish a nonlinear mapping relationship between ship design parameters and at least one performance index. The embedding module is used to embed the trained backpropagation neural network model into the fitness function of the adaptive particle swarm optimization algorithm, where the position of each particle in the particle swarm represents a set of candidate ship design parameters. An initialization module is used to initialize the particle swarm and perform the following steps in each iteration until the set iteration termination condition is met: The first determining module is used to input the current position of each particle into the trained backpropagation neural network model to obtain the corresponding prediction performance index, and determine the current fitness of each particle based on the prediction performance index. The first update module is used to update the individual historical best position of each particle and the collective historical best position of the particle swarm based on the current fitness of each particle. The second update module is used to update the position of each particle based on the updated individual historical best position and the group historical best position; The second determining module is used to determine the final optimized combination of target ship design parameters based on the optimization information recorded during the iteration process.

10. An electronic device, characterized in that, include: A processor and a memory, the processor being configured to execute a ship model optimization program stored in the memory to implement the ship model optimization method according to any one of claims 1-8.

11. A storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the ship model optimization method according to any one of claims 1-8.

Images