Ship multi-objective optimization target relation method and system

Abstract

The invention belongs to the technical field of ship type multi-target optimization design, and discloses a ship type multi-target optimization target relationship classification method and system, a medium, equipment and a terminal, and the method comprises the steps: obtaining a target space through sampling data; calculating the spatial distance between the targets by using a distance calculation formula; clustering the targets by using a clustering algorithm, and evaluating a clustering result; and obtaining a cluster with an optimal clustering effect, and obtaining a target reduction set according to a clustering result, thereby laying a foundation for subsequent ship type multi-target optimization.

Description

Technical Field

[0001] The present invention belongs to the field of ship form optimization design, and relates to a method and system for classifying the relationship of multi-objective optimization objectives of a ship form. Background Art

[0002] At present, the research focus in the direction of objective dimensionality reduction is to reduce the complexity of multi-objective optimization problems. As the number of objectives increases, the high-dimensional objective space poses challenges such as a decline in the convergence ability of optimization algorithms and an increase in computational complexity, and the existence of redundant objectives further increases the problem difficulty. In terms of data processing and analysis, high-dimensional data leads to problems such as an increase in computational costs and the "curse of dimensionality". Objective dimensionality reduction can reduce the data dimension, improve the processing efficiency and effect, and at the same time facilitate data visualization to help discover the internal laws of the data. In addition, to improve the performance of algorithms, objective dimensionality reduction can reduce the number of objectives, lower the computational complexity, and improve the efficiency and robustness of optimization algorithms. In practical applications, multi-objective optimization problems exist in fields such as engineering design, financial investment decision-making, and robot path planning. Objective dimensionality reduction can help find the optimal design solution, formulate reasonable investment strategies, and efficiently plan movement paths, etc., to meet the optimization decision-making needs of various fields.

[0003] In the field of ships, since ship form optimization belongs to a complex engineering system problem with multiple objectives, multiple constraints, and multiple variables, problems such as high dimensionality, high computational cost, and "black box" have led to a deterioration in the quality and efficiency of ship form optimization. At present, the research on design variable dimensionality reduction is relatively mature and has been well applied in the field of ship form optimization. However, there is relatively little research on objective dimensionality reduction.

[0004] At present, certain achievements have been made in the research on the analysis of objective relationships by designers, but there is no research in the field of ships to reduce the difficulty of optimization solution by analyzing objective relationships. For the complex engineering problems of ship form optimization, the increase in the number of objectives makes the objective space more complex, and the Pareto front is difficult to present better, greatly increasing the decision-making difficulty and resulting in a significant decline in the optimization quality.

[0005] Through objective dimensionality reduction, the above problems can be effectively improved. Existing research needs to obtain the Pareto front through preliminary optimization and use Pareto front data analysis to analyze the relationships between objectives. By eliminating redundant relationship objectives and similar relationship objectives, the objective space dimensionality reduction is achieved, and the solution difficulty and decision-making difficulty are reduced.

[0006] At present, the research on achieving objective dimensionality reduction through the analysis and classification of objective relationships is more about obtaining better-quality solutions, but there is not much reduction in computational costs. For practical engineering problems such as ship form optimization, computational costs are also factors that cannot be ignored. Therefore, there is an urgent need to design an optimized objective dimensionality reduction method to analyze using a small amount of simulation data, obtain objective relationships, and achieve objective dimensionality reduction.

[0007] Through the above analysis, the problems and defects existing in the prior art are as follows:

[0008] (1) At present, the target dimensionality reduction method is still in the initial application stage in the ship form optimization design, and there is relatively little in-depth research on the characteristic analysis of the relationship between ship form optimization targets.

[0009] (2) At present, there is relatively little research on using a small number of samples for target relationship analysis in the early stage of optimization, and there is a lack of in-depth analysis of the influence of the sample data volume on the analysis results. Summary of the Invention

[0010] In view of the problems existing in the prior art, the present invention provides a method, system, medium, device and terminal for dimensionality reduction of multi-objective optimization targets of ship form, and particularly relates to a method, system, medium, device and terminal for target relationship analysis based on clustering.

[0011] The present invention is implemented as follows. A method for dimensionality reduction of multi-objective optimization targets of ship form includes calculating the spatial distance between targets through a distance calculation formula according to the performance index information of sample points, and clustering the targets based on the distance criterion using a clustering method to realize the classification of target relationships.

[0012] Further, the method for dimensionality reduction of multi-objective optimization targets of ship form includes the following steps:

[0013] Step 1: Obtain a sample point set through a sampling method, perform performance prediction, and construct a target space;

[0014] Step 2: Calculate the spatial distance between targets according to the distance calculation formula;

[0015] Step 3: Cluster the target set using a clustering algorithm based on the distance between targets, and evaluate the clustering performance using a clustering index algorithm;

[0016] Step 4: According to the clustering index, obtain the clustering result corresponding to the best clustering index, and obtain the corresponding target reduction set.

[0017] Further, the sampling method in Step 1 uses a uniform design method, and the CFD method or an approximate model is used to evaluate the performance of the samples;

[0018] Further, the distance calculation formula between targets in Step 2 uses a cosine distance calculation formula;

[0019] Further, in Step 3, the clustering algorithm uses the K-Means method, and the clustering evaluation index uses the silhouette coefficient method;

[0020] The steps of the K-Means method are as follows:

[0021] (1) Initialization: Determine the value of K and randomly select K initial centroids;

[0022] (2) Assign data points: For each data point, calculate its distance to each centroid. Assign the data point to the cluster corresponding to the nearest centroid;

[0023] (3) Update centroids: For each cluster, calculate the mean of all its data points as the new centroid;

[0024] (4) Iteration: Repeat the assignment and update steps until the centroids are stable or the stopping condition is met;

[0025] (5) Termination: The algorithm ends and outputs the final cluster partition and centroids.

[0026] The calculation steps of the silhouette coefficient are as follows:

[0027] (1) Calculate the within-cluster distance (a(i)): For each data point i, calculate the average distance between it and all other points in the same cluster, denoted as a(i).

[0028]

[0029] where: C i is the cluster to which data point i belongs, |C i | is the size of cluster C i , and distance(i, j) is the distance between data points i and j (usually the Euclidean distance).

[0030] The smaller a(i) is, the closer data point i is to other points in the same cluster;

[0031] (2) Calculate the nearest-cluster distance (b(i)): For each data point i, calculate the average distance between it and all other clusters and find the smallest one, denoted as b(i).

[0032]

[0033] where: C k is the other cluster except C i . |C k | is the size of cluster C k .

[0034] b(i) represents the degree of separation between data point i and the nearest other cluster;

[0035] (3) Calculate the silhouette coefficient of a single data point (s(i)): For each data point i, the formula for calculating its silhouette coefficient s(i) is:

[0036]

[0037] (4) Calculate the overall silhouette coefficient: Take the average of the silhouette coefficients of all data points to obtain the overall silhouette coefficient:

[0038]

[0039] where N is the total number of data points.

[0040] Further, in the fourth step, traverse all possible numbers of clusters, then calculate the silhouette coefficient of each clustering result, extract the clustering result corresponding to the maximum silhouette coefficient, and simultaneously obtain the corresponding target reduction set.

[0041] Another object of the present invention is to provide a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of the ship uncertainty factor classification method.

[0042] Another object of the present invention is to provide a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of the dimensionality reduction of the multi-objective optimization target of the ship type.

[0043] Another object of the present invention is to provide an information data processing terminal for implementing the method for dimensionality reduction of the multi-objective optimization target of the ship type.

[0044] Combined with the above technical solutions and the solved technical problems, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0045] The present invention classifies the target relationships in the multi-objective optimization problem of the ship type. By using the clustering method and combining the cosine distance, a method for dimensionality reduction of the multi-objective optimization target of the ship type is proposed, which can accurately analyze the relationships between the targets in the multi-objective problem of ship type optimization and lay a foundation for the subsequent solution of the multi-objective optimization of the ship type. Therefore, the present invention analyzes the target relationships by using the spatial distance between the targets, thereby eliminating the similar target relationships and realizing the reduction of the target dimension.

[0046] The present invention uses CAESES to perform parametric modeling of the ship type surface, selects design variables, generates samples by using the uniform design method, and evaluates performance indicators by using the CFD method; combines the distance calculation formula and the clustering method to construct a method for dimensionality reduction of the multi-objective optimization target of the ship type. By combining the cosine distance calculation formula and the K-Means algorithm, the present invention analyzes the relationships between the optimization targets according to the performance indicators of the sample points, completes the dimensionality reduction of the targets, and classifies the target relationships into competitive, similar, and constraint relationships, thereby laying a foundation for the subsequent solution of the ship type optimization.

[0047] The present invention provides a method for reducing the dimensions of multi-objective optimization targets of a ship, which can classify the target relationships in the multi-objective optimization problem of ship form, determine competitive, similar, and constraint targets, eliminate similar targets, and achieve the reduction of target dimensions, laying a foundation for the solution of the multi-objective optimization problem of ship form in the next step.

[0048] The expected benefits and commercial value after the transformation of the technical solution of the present invention are as follows: After reducing the dimensions of the targets, the computational cost of solving the multi-objective optimization of ship form can be reduced, the decision-making difficulty can be reduced, which helps designers observe the Pareto front more intuitively and make better decisions.

[0049] The technical solution of the present invention fills the technical gap in the industry at home and abroad: It fills the technical gap in reducing the dimensions of targets in the multi-objective optimization problem of ship form and performs corresponding processing on various different target relationships. Brief Description of the Drawings

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0051] Figure 1 It is a silhouette coefficient curve graph provided by an embodiment of the present invention;

[0052] Figure 2 It is a flow chart of a method for classifying target relationships in multi-objective optimization of ship form provided by an embodiment of the present invention;

[0053] Figure 3 It is a schematic diagram of a method for classifying target relationships in multi-objective optimization of ship form provided by an embodiment of the present invention. Detailed Embodiments

[0054] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0055] Aiming at the problems existing in the prior art, the present invention provides a method, system, medium, device, and terminal for classifying target relationships in multi-objective optimization of ship form, and the present invention will be described in detail below with reference to the drawings.

[0056] This method first performs parametric modeling of the hull surface, selects key design parameters that affect ship resistance and wake non-uniformity, and uses the uniform design method to generate optimized sample points. The purpose of this step is to establish a preliminary objective space with a small number of sample points in the early stage of optimization, so as to conduct objective relationship analysis and dimensionality reduction processing subsequently. Uniform design sampling can effectively cover the design space, ensure the representativeness of different parameter combinations, and improve the efficiency of optimization calculations.

[0057] After obtaining the optimized sample points, calculate the performance indicators for each sample, such as key parameters like hull resistance and wake non-uniformity, and construct the objective space for multi-objective optimization. In this process, obtain objective information through numerical simulation or experimental measurement to form an objective data set. This objective data set is used for subsequent objective relationship analysis to ensure the rationality and effectiveness of the optimization.

[0058] In the objective relationship analysis stage, calculate the similarity between different objectives, and use the cosine distance formula to measure the spatial distance between objectives. The cosine distance can quantify the angular difference between different objective vectors in a high-dimensional space and reflect the similarity between objectives. This measurement method is particularly important for ship form optimization problems because it can effectively reduce data redundancy and improve the accuracy of optimization calculations.

[0059] By calculating the distance between objectives, use the K-Means clustering algorithm to perform clustering analysis on the objective set and divide the objectives into different categories. In the clustering process, the similarity between different objectives is used to guide the classification, enabling the optimization problem to be carried out in a more structured manner. To ensure the rationality of the clustering results, calculate the silhouette coefficient under different numbers of clusters, and select the best clustering result by maximizing the silhouette coefficient, thereby ensuring the accuracy and stability of objective classification.

[0060] After completing the clustering analysis, based on the similarity of objectives in different categories, perform dimensionality reduction processing on the optimization objectives. The core idea of objective dimensionality reduction is to use the clustering results to remove redundant objective information, reduce unnecessary computational complexity in the optimization process, and at the same time maintain the representativeness of the optimization problem. The reduced-dimensional objective subset is easier to handle, making the optimization calculation more efficient.

[0061] After objective dimensionality reduction, the optimization problem is simplified and can execute the multi-objective optimization algorithm in the new objective space. The optimization results can be used to further correct the parametric modeling and objective relationship analysis to achieve iterative improvement of the optimization scheme. This method improves the solution efficiency of the ship form optimization problem, enhances the adaptability of the optimization strategy, and makes the optimization results more in line with the actual engineering requirements through the combination of objective relationship classification, dimensionality reduction, and optimization feedback.

[0062] As Figure 2 shown, the method for reducing the dimensions of multi-objective optimization objectives of the ship form provided by the embodiment of the present invention includes the following steps:

[0063] S101. Obtain performance index data through modeling, sampling, and evaluation;

[0064] S102. Combine the cosine distance formula and the K-Means algorithm according to the collected data to achieve the classification of the multi-objective optimization target relationship of the ship form.

[0065] As a preferred embodiment, as Figure 2 shown, the method for classifying the multi-objective optimization target relationship of the ship form provided by the embodiment of the present invention specifically includes the following steps:

[0066] (1) Implement ship form surface modeling through CAESES modeling software, and select the modeling parameters as design variables;

[0067] (2) Use the uniform design method for sampling and evaluate the performance index of each sample;

[0068] (3) Calculate the cosine distance between samples using the performance index information of the samples;

[0069] (4) Based on the cosine distance between samples, use the K-Means algorithm to cluster the target set and evaluate the clustering result using the silhouette coefficient;

[0070] (5) Traverse all possible numbers of clusters, calculate the silhouette coefficient at the same time, take the clustering result corresponding to the highest silhouette coefficient as the final result, and obtain the corresponding target reduction set.

[0071] The system for classifying the multi-objective optimization target relationship of the ship form provided by the embodiment of the present invention includes:

[0072] A ship form surface modeling module for parametric modeling of the ship form;

[0073] A sample acquisition module for obtaining the corresponding sample set;

[0074] A performance evaluation module for calculating the performance index of the sample;

[0075] A distance calculation module for calculating the distance between targets;

[0076] A clustering analysis module for performing clustering analysis on the target set;

[0077] A clustering evaluation module for evaluating the clustering result.

[0078] In order to prove the creativity and technical value of the technical solution of the present invention, this part is an application embodiment of the technical solution of the claims on a specific product or related technology.

[0079] The present invention takes a 7500t inland river bulk carrier as the research object, and conducts research on the resistance performance and wake non-uniformity under two speeds and two drafts (a total of eight objectives). First, parametric modeling is carried out, and 11 design parameters affecting the ship's resistance performance and wake non-uniformity are selected as design variables. The uniform design method is used to generate 220 sample points, and the performance indexes of each sample are predicted using the Kriging approximation model, forming an initial objective set, which lays a foundation for subsequent objective relationship analysis and objective dimensionality reduction.

[0080] After obtaining the performance indexes of 220 sample points, the present invention uses the cosine distance calculation formula to calculate the cosine distance between every two objectives, and uses this as the basis for the K-Means clustering algorithm. The clustering number k is set to 2-7 respectively, and the corresponding silhouette coefficients are calculated.

[0081] The present invention selects the clustering corresponding to the highest silhouette coefficient as the final result. In the final clustering, similar objectives are grouped together, leaving the objective closest to the clustering center and removing the other objectives in the clustering, thus obtaining the latest objective reduction set.

[0082] The present invention replaces the original objective set with the final objective reduction set, and redefines the multi-objective optimization problem of the 7500t inland river bulk carrier. Due to the reduction of objectives, the calculation cost is reduced in the multi-objective optimization process, the quality of the solutions in the Pareto front is better, and at the same time, due to the reduction of dimensions, it is easier for designers to make decisions.

[0083] Multi-objective dimensionality reduction example for the hull form optimization of a 7500t inland river bulk carrier

[0084] Taking the 7500t inland river bulk carrier of the present invention as the research object, parametric modeling is first carried out, 11 design parameters are selected as design variables, and 220 sample points are generated using uniform design, as shown in the following table:

[0085] Table 1 Design variables

[0086]

[0087] Table 2 Sample point collection

[0088]

[0089] Taking the total resistance R under four working conditions and the minimum WOF at 0.7R as the optimization objectives, that is, directly solving eight optimization objectives, which are defined as follows:

[0090]

[0091] Table 3 Optimization objectives

[0092]

[0093] Meanwhile, it is necessary to ensure that the displacement increases during the optimization process. Therefore, the constraints are as shown in the formula:

[0094]

[0095] Use the Kriging approximation model to calculate and evaluate the performance of the sample set to obtain the target space.

[0096] Use the cosine distance calculation formula to calculate the distance between every two targets, as shown in the following table

[0097] Table 4 Target space distance

[0098]

[0099] Based on the distances between the targets, use the K-Means method to cluster the target set. The number of clusters is 2 - 7, and calculate their corresponding silhouette coefficients respectively, as Figure 1 shown.

[0100] It can be seen that when k = 4, the silhouette coefficient is the largest at 0.6986. Therefore, select the clustering when k = 4 as the final result, as follows:

[0101] Cluster 1: V1T1W, V1T2W, V2T1W, V2T2W;

[0102] Cluster 2: V1T1R;

[0103] Cluster 3: V1T2R;

[0104] Cluster 4: V2T1R, V2T2R.

[0105] Calculate the distance between each target in the cluster and the cluster center, retain the target with the closest distance to the cluster center, and eliminate the remaining targets to obtain the target reduction set: {V1T1W, V1T1R, V1T2R, V2T1R}.

[0106] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated designed hardware. Those of ordinary skill in the art can understand that the above devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code is provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and their modules of the present invention can be implemented by hardware circuits of programmable hardware devices such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above hardware circuits and software such as firmware.

[0107] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A ship type multi-objective optimization target relationship classification method, characterized in that: The method comprises the following steps: Step 1: Obtain the target space through parametric modeling of the hull surface, uniform design sampling, and performance index calculation of the sample points; Step 2: Based on the target information in the target space, calculate the distance between the targets, use the clustering algorithm to perform clustering analysis on the target set, and evaluate the clustering results to obtain the best clustering result; Step 3: Based on the cluster analysis results, perform dimensionality reduction on the target set to remove redundant target information and reduce the target dimension.

2. The ship type multi-objective optimization target relationship classification method according to claim 1 is characterized in that: In the step 1, the hull surface is parametrically modeled, and design parameters that affect the ship resistance and wake unevenness are selected. Sample points are generated using a uniform design method, and the performance index of each sample point is calculated.

3. The ship type multi-objective optimization target relationship classification method according to claim 1 is characterized in that: In the step 2, the cosine distance is used to calculate the spatial distance between the targets, and the K-Means clustering algorithm is used to classify the target set.

4. The ship type multi-objective optimization target relationship classification method according to claim 3 is characterized in that: When the cosine distance is used to calculate the spatial distance between targets, the angle between each target vector and other target vectors is calculated using the components of each target vector, and the target similarity is obtained after normalization.

5. The ship type multi-objective optimization target relationship classification method according to claim 3 is characterized in that: The K-Means clustering algorithm is evaluated based on the silhouette coefficients of different cluster numbers, and the cluster number with the largest silhouette coefficient is selected as the best clustering result.

6. The ship type multi-objective optimization target relationship classification method according to claim 5 is characterized in that: The calculation of the silhouette coefficient is based on the intra-cluster distance and the nearest cluster distance, wherein the intra-cluster distance is calculated based on the Euclidean distance between target vectors in the same cluster, and the nearest cluster distance is calculated based on the Euclidean distance from the target point to the nearest cluster other than the cluster to which it belongs.

7. The ship type multi-objective optimization target relationship classification method according to claim 1 is characterized in that: In the step three, the best clustering result is used to perform target dimension reduction processing and remove redundant target information to reduce the target dimension during the optimization process and improve the effectiveness of target classification.

8. A ship type multi-objective optimization target relationship classification system, characterized in that: The system includes: The target space construction module is used to perform parametric modeling of the hull surface, select the design parameters that affect the ship's resistance and wake unevenness, generate optimized sample points using the uniform design method, calculate the performance index of each sample point, and construct the target space; The target relationship analysis module is used to calculate the spatial distance between targets, classify the target set using a clustering algorithm, evaluate the clustering results, and select the optimal number of clusters; The target dimension reduction processing module is used to perform dimension reduction processing on the target set based on the cluster analysis results, remove redundant target information, and reduce the target dimension.

9. The ship type multi-objective optimization target relationship classification system according to claim 8 is characterized in that: The target relationship analysis module includes: A distance calculation unit, used to calculate the spatial similarity between objects based on cosine distance; A clustering calculation unit, used to classify the target set based on the K-Means clustering algorithm; The clustering evaluation unit is used to calculate the silhouette coefficients under different cluster numbers and select the cluster number with the largest silhouette coefficient as the best clustering result.

10. The ship type multi-objective optimization target relationship classification system according to claim 8, characterized in that: The target dimension reduction processing module includes: A target screening unit, used for selecting a representative target subset based on the clustering results; The dimension optimization unit is used to remove redundant target information and reduce the target dimension during the optimization process.

Images