Longitudinal heterogeneous ship formation-based drag reduction prediction model design method and related device

Abstract

The application relates to a drag reduction prediction model design method based on a longitudinal heterogeneous ship formation and a related device, and belongs to the technical fields of ship engineering and fluid mechanics. The method comprises the following steps: setting longitudinal heterogeneous ship formation numerical simulation parameters in a preset simulation software; adopting a control variable method to design numerical simulation scenes of different formation configurations, determining ship drag coefficients of the longitudinal heterogeneous ship formation under each formation configuration, and analyzing the action law of ship bow flow field velocity and ship wave on shear resistance and differential pressure resistance; and based on the action law, a mixed graph neural network prediction model is used to predict the formation drag of a to-be-formed ship group, wherein the mixed graph neural network prediction model comprises a cooperative architecture encoded by a GATv2 dynamic attention mechanism and a GINE geometric feature. The application can predict and optimize the drag reduction effect of the longitudinal heterogeneous ship formation, and provides a scientific basis and an intelligent decision-making tool for ship formation configuration.

Description

Technical Field

[0001] This invention relates to the fields of marine engineering and fluid mechanics, and in particular to a method and related apparatus for designing a drag reduction prediction model based on longitudinal heterogeneous ship formations. Background Technology

[0002] With the acceleration of globalization, shipping has become a core pillar of international trade, with over 90% of global commercial transport traffic completed via the ocean, making its role in cargo transportation irreplaceable. However, the rapid development of the shipping industry is also accompanied by significant environmental challenges.

[0003] According to data from the International Maritime Organization (IMO), in 2018, the shipping industry accounted for approximately 2.9% of global carbon dioxide emissions, 11% of sulfur oxides, and 15% of nitrogen oxides. Excessive carbon dioxide emissions not only exacerbate global warming but also lead to more frequent extreme weather events and sea-level rise, jeopardizing ecosystem balance. Sulfur oxides and nitrogen oxides, on the other hand, cause acid rain and air pollution, posing serious threats to soil, vegetation, and human health. Therefore, reducing ship drag and thus achieving energy conservation and emission reduction in shipping has become a crucial issue in the development of green transportation. Current technologies typically achieve energy conservation and emission reduction in shipping through fuel substitution and exhaust aftertreatment, but these methods cannot fundamentally reduce shipping energy consumption. Summary of the Invention

[0004] Therefore, it is necessary to provide a drag reduction prediction model design method and related device based on longitudinal heterogeneous ship formations, so as to solve the problem of high energy consumption and emissions of existing shipping vessels with a new approach that does not require modification.

[0005] To address the aforementioned challenges and achieve the goals of drag reduction and energy saving, in a first aspect, this invention provides a drag reduction prediction model design method based on longitudinally heterogeneous ship formations, comprising:

[0006] Set the numerical simulation parameters for longitudinal heterogeneous ship formations within the preset simulation software;

[0007] Numerical simulation scenarios with different formation configurations were designed using the controlled variable method to determine the ship resistance coefficients of longitudinally heterogeneous ship formations under each formation configuration. The formation configurations include the number of ships, speed, spacing, and ship type.

[0008] Analysis of the effects of bow flow velocity and ship wave on shear resistance and differential pressure resistance in longitudinal heterogeneous ship formations with different formation configurations;

[0009] Based on the aforementioned action law, a hybrid graph neural network prediction model is used to predict the formation resistance of the ship group to be formed. The hybrid graph neural network prediction model includes a collaborative architecture that combines the GATv2 dynamic attention mechanism with GINE geometric feature encoding.

[0010] In one possible implementation, numerical simulation parameters for longitudinally heterogeneous ship formations are set within pre-defined simulation software, including:

[0011] In the STAR-CCM+ simulation environment, eight consecutive monitoring points are set in front of the ship, with each monitoring point spaced 0.5m apart, to collect data on the bow flow field; and 40 consecutive monitoring points are set behind the last ship in the formation, with each monitoring point spaced 0.5m apart, to collect data on the changes in the ship's traveling waves.

[0012] In one possible implementation, the numerical simulation scenarios for different formation configurations are designed using the control variable method, including:

[0013] Under the condition that the ship speed and ship spacing are the same, different numbers of ships are configured for the longitudinal heterogeneous ship formation model;

[0014] With the same number of ships and the same spacing between ships, different ship speeds are configured for the longitudinal heterogeneous ship formation model.

[0015] With the same number of ships and ship speeds, different ship spacings are configured for the longitudinal heterogeneous ship formation model.

[0016] In one possible implementation, the analysis of the effects of bow flow velocity and ship wave on shear drag and differential pressure drag based on the ship drag coefficient of longitudinally heterogeneous ship formations with different formation configurations includes:

[0017] During the STAR-CCM+ simulation, the changes in bow flow velocity and wave height over time are statistically analyzed at monitoring points. The average values ​​of these changes are calculated and displayed in a bar chart. Simultaneously, the wave height, wavelength, and period of the wave are statistically analyzed, and the monitoring point data are displayed in a line chart. The line chart is then interpolated.

[0018] In one possible implementation, the effects of bow flow velocity and ship wave on shear drag and differential pressure drag are analyzed, including:

[0019] During the STAR-CCM+ simulation, the changes in bow flow velocity and ship wave height over time are statistically analyzed by monitoring points, and their average values ​​are calculated and displayed in the form of a bar chart. At the same time, the ship wave height, wavelength and period are statistically analyzed, and the monitoring point data are displayed in a line chart, and the curves are smoothed by interpolation.

[0020] In one possible implementation, the formation resistance of the vessel group to be formed is predicted using a hybrid graph neural network prediction model based on the aforementioned interaction law, including:

[0021] The hybrid graph neural network prediction model includes a parallel architecture of GATv2 layer, PNA layer, and GINE layer, which is used for multi-level feature extraction and analysis of model prediction accuracy.

[0022] In one possible implementation, the accuracy of the analysis model prediction includes:

[0023] Three data augmentation measures were adopted to address the scarcity of CFD simulation data, and the prediction accuracy for different formation sizes was quantitatively analyzed using heatmaps.

[0024] In a second aspect, the present invention also provides an electronic device, including a memory and a processor, wherein,

[0025] Memory, used to store programs;

[0026] The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the drag reduction prediction model design method based on longitudinal heterogeneous ship formations in any of the above embodiments.

[0027] Thirdly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instruction, which, when executed by a processor, enables the implementation of the steps in the drag reduction prediction model design method based on longitudinal heterogeneous ship formations of any of the above embodiments.

[0028] The beneficial effects of this invention are: by combining numerical simulation and machine learning methods, it is possible to efficiently predict and optimize the drag reduction effect of longitudinal heterogeneous ship formations, significantly improve the formation navigation efficiency, and provide a scientific basis for ship formation configuration. Attached Figure Description

[0029] To clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments are briefly introduced below. The drawings are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart illustrating a drag reduction prediction model design method based on longitudinal heterogeneous ship formations provided in an embodiment of the present invention;

[0031] Figure 2 A schematic diagram of a longitudinally heterogeneous ship formation model provided in an embodiment of the present invention;

[0032] Figure 3A schematic diagram of a ship hull simulation mesh provided for an embodiment of the present invention;

[0033] Figure 4 This is a diagram showing the setting of ship wave and flow field detection points according to an embodiment of the present invention;

[0034] Figure 5 Waveform diagrams of ships with different formation numbers provided in embodiments of the present invention;

[0035] Figure 6 Total resistance diagrams for different numbers of ship formations provided in embodiments of the present invention;

[0036] Figure 7 The flow field and resistance analysis diagrams of ship formations at different speeds provided in the embodiments of the present invention;

[0037] Figure 8 Six different navigation attitude diagrams for heterogeneous ship formations provided in embodiments of the present invention;

[0038] Figure 9 The hybrid GNN structure provided in the embodiments of the present invention;

[0039] Figure 10 Loss curve of hybrid GNN provided in the embodiments of the present invention;

[0040] Figure 11 The ship formation error distribution provided in the embodiments of the present invention;

[0041] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0042] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0043] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0044] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0045] A specific embodiment of the present invention, such as Figure 1 As shown, a drag reduction prediction model design method based on longitudinal heterogeneous ship formations is disclosed, including:

[0046] S101, set the numerical simulation parameters for longitudinal heterogeneous ship formations within the preset simulation software.

[0047] In this embodiment of the invention, a longitudinally heterogeneous ship formation model was constructed using SolidWorks. Taking three ships as an example, the first two are KCS ships, and the last one is a SERIES60 ship, as shown below. Figure 2 As shown. In the formation, the first vessel is the lead vessel (ld), and the remaining vessels are follower vessels (fw1, fw2, ...). The captain of KCS is denoted as Lk, the captain of SERIES60 is denoted as Ls, and the formation spacing is denoted as D.

[0048] In this embodiment of the invention, the commercial CFD software STAR-CCM+ is used for numerical simulation calculations. Due to the geometric symmetry of the ship's shape, numerical simulations are performed only on half of the ship. In the computational domain of the towing test tank, the ship's center of gravity is set as the origin. Along the X-axis, the distance from the velocity inlet is 18m, and the distance from the pressure outlet is -18m; along the Y-axis, the distance from the ship to the test tank boundary is 18m; along the Z-axis, the distance from the bottom of the test tank is -18m, and the distance from the top is 9m. The hydrostatic pressure corresponds to the calm water surface at the downstream boundary of the test tank. In STAR-CCM+, a mesh is created for the ship and the test tank based on an automatic mesh, as follows: Figure 3 As shown. Surface reconstruction, automatic surface repair, a cut-volume mesh generator, and a prism layer mesh generator were employed. The target surface size was set to 50%, the number of prism layers was 6, and the volume growth rate was slow. To capture Kelvin waves and ensure a rapid transition between the two fluids, surface densification was performed. In STAR-CCM+, drag can be decomposed into shear drag and pressure drag. To further analyze the reasons for the reduction in drag during formation navigation, it is necessary to detect the influence of the bow flow velocity on the ships. Therefore, eight consecutive monitoring points were set up in front of the ships, with each monitoring point spaced 0.5m apart, and the flow field data of the ships during navigation were statistically analyzed, such as... Figure 4 As shown; at the same time, it is also necessary to consider the impact of the ship's sailing waves on the following ships. Therefore, 40 consecutive monitoring points were set up behind the last ship in the formation, with each monitoring point spaced 0.5m apart, to count the changes in the sailing waves of the formation ships.

[0049] To ensure the accuracy of the numerical calculations, a Froude number of 0.2599 was selected, and the resistance of the KCS ship's self-propulsion was numerically simulated. The numerical simulation results were compared with the experimental results presented at the 2010 Gothenburg Symposium, and the expression for the resistance coefficient was studied as follows:

[0050]

[0051] Where cd is the drag coefficient, the average total drag Fd is 42.083 N, the water density is ρ, the ship model speed v is 2.196 m / s, and the wetted surface area A of the ship model is 9.55275 m2.

[0052] The predicted drag coefficient calculated based on the above equation is 0.003654, while the experimental drag coefficient provided at the 2010 Gothenburg Symposium was 0.003711, with a deviation of 1.53%. Increasing the mesh precision will reduce the deviation in predicted drag. Note that the simulation only models half of the ships; to match experimental data, the drag coefficient needs to be multiplied by 2.

[0053] S102, the numerical simulation scenarios of different formation configurations are designed using the controlled variable method, and the ship resistance coefficients of longitudinal heterogeneous ship formations under each formation configuration are determined. The formation configuration includes the number of ships, speed, spacing and ship type.

[0054] In embodiments of the present invention, such as Figure 5 As shown, KCS ship formation models with different numbers of ships were constructed. Under the conditions of a formation spacing of 1 ship length and a sailing speed of 2.12 m / s, the formations of single ships, 2 ships, 3 ships and 4 ships were simulated respectively. The changes in the bow flow field velocity and ship traveling wave were statistically analyzed by monitoring points to analyze their impact on the formation resistance characteristics. Figure 6 The study compares the total resistance of KCS (Kinship Crew) formations with different numbers of vessels under the conditions of a formation spacing of 1 ship length and a speed of 2.12 m / s. It can be seen that the total resistance of the lead ship in the formation is similar to that of a single ship, while the total resistance of the following ships is lower than that of the lead ship. Therefore, the total resistance of the formation is lower than that of a single ship with the same number of vessels.

[0055] In this embodiment of the invention, simulations were conducted on a single ship and a three-ship formation under the same KCS ship type and a formation spacing of 1 ship length, with sailing speeds of 1.275 m / s, 2.12 m / s, and 2.975 m / s, respectively. The single ship is represented by "SG", and the three speeds are labeled as low speed (LS), medium speed (MS), and high speed (HS), respectively. Figure 7 (a) shows the bow flow velocity in single-ship and three-ship formations. Figure 7 (c) shows the corresponding ship shear resistance. As the speed of the formation increases, the velocity difference in the bow flow field between the lead ship and the follower ships gradually increases, and the trend of shear resistance changes is similar. Figure 7 (d) illustrates the pressure drag difference between a single ship and a three-ship formation. At high speeds, the increase in pressure drag is significantly greater than that of shear drag, while the pressure drag difference of the two following ships decreases significantly. Observation Figure 7(b) shows the ship waves of the formation. It was found that the wave height and wavelength of the ship waves were larger at high speeds, and the energy carried was higher. Combined with the significant decrease in pressure drag of the two following ships, it indicates that the high-energy ship waves can reduce pressure drag.

[0056] In this embodiment of the invention, a three-ship formation was constructed under the same KCS ship type and a sailing speed of 2.12 m / s, and simulations were conducted with formation spacing of 0.5 times, 1 times, and 1.5 times the ship length, respectively. The formation spacing was divided into close spacing (ND), medium spacing (MD), and far spacing (FD). Compared with the formation speed condition, the effect of formation spacing on drag reduction is relatively limited, especially from medium spacing to far spacing. As the formation spacing increases, the total resistance of each ship in the formation hardly changes. The proportions of shear drag and differential pressure drag are almost unaffected by the change in formation spacing, similar to the drag distribution pattern at a sailing speed of 2.12 m / s and a formation spacing of 1 times the ship length.

[0057] In this embodiment of the invention, to verify that different types of ships can also reduce resistance, SERIES60 ship formation models with different numbers of vessels were constructed. Simulations were performed for single-ship, 2-ship, 3-ship, and 4-ship formations under conditions of 0.82 times the KCS ship length and a sailing speed of 2.12 m / s. In the SERIES60 formation, the proportion of shear resistance decreased from 62% to approximately 47%, with pressure drag becoming dominant. The resistance reduction rate also increased from 3% to approximately 10%. This change may be related to the smaller size of the SERIES60 vessels (SERIES60 ships weigh 485.6 kg, while KCS ships weigh 823.0451 kg).

[0058] In this embodiment of the invention, a heterogeneous ship formation model is constructed using KCS ships and SERIES60 ships, such as... Figure 8As shown, six heterogeneous ship formation models were constructed, including KCS-SERIES60-KCS (KSK), KCS-SERIES60-SERIES60 (KSS), KCS-KCS-SERIES60 (KKS), SERIES60-SERIES60-KCS (SSK), SERIES60-KCS-SERIES60 (SKS), and SERIES60-KCS-KCS (SKK). Simulations were conducted at a sailing speed of 2.12 m / s and a formation spacing of 0.82 times the length of the KCS ships to analyze the influence of ship type on drag characteristics. The results show that when SERIES60 ships (smaller ships) follow KCS ships (larger ships), their bow flow field velocity is lower, resulting in a corresponding decrease in shear drag. High-energy ship traveling waves help reduce pressure drag, but the reduction is relatively limited. This explains why, in a formation, placing smaller ships behind larger ships can effectively reduce overall drag; when there are many large ships in the formation, placing smaller ships in the middle position can further reduce the total drag of the formation.

[0059] S103, Analysis of the effect of bow flow velocity and ship wave on shear resistance and differential pressure resistance based on the ship resistance coefficient of longitudinal heterogeneous ship formations under different formation configurations.

[0060] In this embodiment of the invention, the post-processing of STAR-CCM+ outputs data on shear resistance, differential pressure resistance, total resistance, bow flow velocity, and ship wave height of the formation during the simulation process. Based on PyCharm, the following data processing is performed: the average value of the monitoring points is calculated and displayed in the form of a bar chart; statistical analysis is performed on the ship wave height, wavelength, and period; the monitoring point data is displayed in a line chart; and the curve is smoothed by interpolation.

[0061] S104, Based on the aforementioned action law, a hybrid graph neural network prediction model is used to predict the formation resistance of the ship group to be formed. The hybrid graph neural network prediction model includes a collaborative architecture using GATv2 dynamic attention mechanism and GINE geometric feature encoding.

[0062] In this embodiment of the invention, a data-driven approach is used to systematically study the coupled effects of ship type, spacing, and speed on resistance characteristics in ship formations, and a hybrid GNN prediction model is constructed. Given the high cost of acquiring STAR-CCM+ simulation data and the dynamic nature of formation size, the graph neural network architecture offers dual advantages: firstly, the graph structure of GNNs naturally adapts to the modeling requirements of a variable number of ship nodes; secondly, its message passing mechanism can effectively capture fluid interactions between ships. Figure 9 As shown, the model employs a three-way parallel architecture to achieve multi-level feature extraction:

[0063] GATv2 layer: Identifies key ship features through a dynamic attention mechanism, providing weighted neighbor information for subsequent aggregation;

[0064] PNA layer: As the core of the architecture, it stabilizes the formation of different sizes through a multi-scale aggregator and integrates the attention weights of GATv2 with the geometric features of GINE.

[0065] GINE layer: A message passing mechanism based on edge features that accurately encodes the fluid interference effects of the relative positions between ships.

[0066] In this embodiment of the invention, to address the challenge of scarce CFD simulation data, three data augmentation measures were implemented: first, the complexity of a single sample was reduced by decoupling flow field features; second, supplementary data was generated using a physical constraint-based interpolation algorithm; and finally, a dataset containing approximately 100,000 samples was constructed (training set:test set = 5:1). Regarding feature engineering, the formation spacing was dimensionless based on the KCS ship length (L / L_KCS∈[0.5,1.5]), while the speed was retained as its original value (v∈[1.275,2.975]m / s) to avoid deviations introduced by inconsistent Froude numbers of heterogeneous ship types.

[0067] In embodiments of the present invention, such as Figure 10 As shown, the model training process exhibits a significant three-stage characteristic: rapid convergence period (0-15 Epochs): the loss function decreases by 86%, and basic feature extraction is completed; fine-tuning period (15-60 Epochs): local optimum search is achieved through parameter oscillation; stable convergence period (>60 Epochs): the final training loss is 0.00148, the test loss is 0.0149, and the fluctuation range is <1.5%.

[0068] In this embodiment of the invention, the prediction accuracy for different formation sizes is quantitatively analyzed using heatmaps. Figure 11 The study found that the prediction error of a 2-ship formation was the lowest (lightest hue in the heatmap), which was due to the completeness of the training data; the error of a 4-ship formation was relatively high (darkest hue), reflecting the challenge of modeling sparse data regions; the accuracy distribution was strongly correlated with the training sample density.

[0069] In the embodiments of this invention, the comparative experiments shown in Table 1 demonstrate that the hybrid GNN significantly outperforms the baseline model in two key metrics: prediction accuracy: MSE reduced by 13.6% (vs. GAT) and 21.3% (vs. random forest); generalization ability: R² improved to a maximum of 0.92, and supports prediction of formations of arbitrary sizes. While random forest has the advantage of computational efficiency, its fixed input dimension makes it unsuitable for application scenarios with varying formation sizes.

[0070] Table 1 Comparison of MSE and R2 for each model

[0071]

[0072] like Figure 12 As shown, the present invention also provides an electronic device 1200. The electronic device 1200 includes a processor 1201, a memory 1202, and a display 1203. Figure 12 Only some components of the electronic device 1200 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0073] In some embodiments, processor 1201 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in memory 1202 or process data, such as the longitudinal heterogeneous ship formation design method of the present invention.

[0074] In some embodiments, processor 1201 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 1201 may be local or remote. In some embodiments, processor 1201 may be implemented on a cloud platform. In some embodiments, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, internal cloud, multi-cloud, etc., or any combination thereof.

[0075] In some embodiments, memory 1202 may be an internal storage unit of electronic device 1200, such as a hard disk or memory of electronic device 1200. In other embodiments, memory 1202 may also be an external storage device of electronic device 1200, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 1200.

[0076] Furthermore, the memory 1202 may include both internal storage units of the electronic device 1200 and external storage devices. The memory 1202 is used to store application software and various types of data installed on the electronic device 1200.

[0077] In some embodiments, display 1203 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 1203 is used to display information from electronic device 1200 and to display a visual user interface. Components 1201-1203 of electronic device 1200 communicate with each other via a system bus.

[0078] In some embodiments, when the processor 1201 executes the longitudinal heterogeneous ship formation design program in the memory 1202, the following steps may be performed:

[0079] Numerical simulations of longitudinally heterogeneous ship formations with different formation configurations were conducted using the controlled variable method to determine the ship resistance coefficients of longitudinally heterogeneous ship formations under each formation configuration.

[0080] A dataset is constructed based on the formation configuration of longitudinal heterogeneous ship formations and the drag coefficients corresponding to each formation configuration. The dataset is then used to train a pre-defined ship formation design model.

[0081] The drag coefficients of different formation configurations of the ships to be formed are predicted using a trained ship formation design model, and the formation configuration with the lowest drag coefficient is determined as the target formation configuration.

[0082] It should be understood that when the processor 1201 executes the longitudinal heterogeneous ship formation design program in the memory 1202, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0083] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions in the longitudinal heterogeneous ship formation design method provided in the above-described method embodiments.

[0084] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0085] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for designing a drag reduction prediction model based on longitudinally heterogeneous ship formations, characterized in that, include: Set the numerical simulation parameters for longitudinal heterogeneous ship formations within the preset simulation software; Numerical simulation scenarios with different formation configurations were designed using the controlled variable method to determine the ship resistance coefficients of longitudinally heterogeneous ship formations under each formation configuration. The formation configurations include the number of ships, speed, spacing, and ship type. Analysis of the effects of bow flow velocity and ship wave on shear resistance and differential pressure resistance in longitudinal heterogeneous ship formations with different formation configurations; Based on the aforementioned operational rules, a hybrid graph neural network prediction model is used to predict the formation resistance of the fleet of ships. The hybrid graph neural network prediction model includes a collaborative architecture of GATv2 dynamic attention mechanism and GINE geometric feature encoding. The step of setting numerical simulation parameters for longitudinally heterogeneous ship formations within the preset simulation software includes: In the STAR-CCM+ simulation environment, eight consecutive monitoring points are set up in front of the ship, with each monitoring point spaced 0.5m apart, to collect data on the bow flow field; and 40 consecutive monitoring points are set up behind the last ship in the formation, with each monitoring point spaced 0.5m apart, to collect data on the changes in the ship's traveling wave. The numerical simulation scenarios for different formation configurations designed using the controlled variable method include: Under the condition that the ship speed and ship spacing are the same, different numbers of ships are configured for the longitudinal heterogeneous ship formation; With the same number of ships and the same spacing between ships, different ship speeds are configured for the longitudinal heterogeneous ship formation. With the same number of ships and ship speeds, different ship spacings are configured for the longitudinal heterogeneous ship formation. The numerical simulation scenarios for different formation configurations designed using the controlled variable method specifically include: A heterogeneous ship formation model was constructed using KCS and SERIES60 ships. Under a given number of ships and the same ship speed and spacing, different combinations of ship types were configured for the longitudinal heterogeneous ship formation. The analysis of the effects of bow flow velocity and ship wave on shear drag and differential pressure drag based on the ship drag coefficient of longitudinally heterogeneous ship formations with different formation configurations includes: During the STAR-CCM+ simulation, the changes in bow flow velocity and wave height over time are statistically analyzed at monitoring points. The average values ​​of these changes are calculated and displayed in a bar chart. Simultaneously, the wave height, wavelength, and period of the wave are statistically analyzed, and the monitoring point data are displayed in a line chart. The line chart is then interpolated. The prediction of the formation resistance of the fleet based on the aforementioned action law using a hybrid graph neural network prediction model includes: The hybrid graph neural network prediction model includes a parallel architecture of GATv2 layer, PNA layer, and GINE layer for multi-level feature extraction and analysis of model prediction accuracy. The prediction accuracy of the analytical model includes: Three data augmentation measures were adopted to address the scarcity of CFD simulation data, and the prediction accuracy for different formation sizes was quantitatively analyzed using heatmaps.

2. An electronic device, characterized in that, Including memory and processor, among which, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the drag reduction prediction model design method based on longitudinal heterogeneous ship formation as described in claim 1.

3. A computer-readable storage medium, characterized in that, Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the drag reduction prediction model design method based on longitudinal heterogeneous ship formations as described in claim 1.

Images