An intelligent prediction method, program, device and storage medium for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network

Abstract

The present application belongs to the technical field of flow field intelligent prediction based on numerical simulation, and particularly relates to a ship surface fluid pressure field intelligent prediction method, program, device and storage medium based on a combination of a reduced-order model and a neural network. The present application maps the unstructured grid data obtained by numerical simulation into structured grid data by using a surrogate model, ensuring the uniformity of the subsequent data processing format. On this basis, the main modal of the surface pressure field and the corresponding grid node normal vector is extracted by the reduced-order model. Then, the full-link neural network is used to realize the rapid prediction of the new ship type basis coefficient, so that for the new ship type, only the design parameters need to be input without inputting the ship grid, the rapid prediction of the ship surface load distribution, the external normal vector distribution and the pressure resistance can be realized, and reliable technical support can be provided for the ship structure strength design and safe navigation at sea.

Description

Technical Field

[0001] This invention belongs to the field of intelligent flow field prediction technology based on numerical simulation, specifically involving an intelligent prediction method, program, device and storage medium for ship surface fluid pressure field based on a combination of reduced-order model and neural network. Background Technology

[0002] Ships are crucial equipment for maritime transportation and resource development, and their structural integrity directly impacts navigation safety and mission execution capabilities. In the complex marine environment, the hull surface is continuously subjected to various dynamic pressure loads, such as the impact of wind and waves, and wave action. These loads are transmitted to the internal structure through the hull surface. Once the load amplitude exceeds the bearing capacity of the materials and structure, it can lead to varying degrees of damage, including localized dents, crack propagation, and component failure. In severe cases, it can even cause the ship to sink, resulting in significant casualties and property losses. Therefore, ensuring structural strength is a fundamental prerequisite for the safe navigation of ships at sea.

[0003] In the early stages of ship design, systematically assessing the structural strength of the hull, especially the pressure distribution on the hull surface under given operating conditions, is a crucial step in mitigating structural safety risks during actual navigation. However, due to the inconsistent topological relationships of computational grids for different ship types in numerical simulations, it is currently difficult to directly compress data from complex ship flow fields (such as the hydrodynamic pressure field experienced by the ship) using reduced-order models. Furthermore, neural networks can typically only rapidly predict overall performance indicators such as drag coefficients and heave / pitch amplitudes, and cannot currently provide rapid predictions of the pressure field distribution on the hull surface. Therefore, developing efficient and accurate rapid prediction technology for the entire ship's pressure field is of paramount importance.

[0004] The rapid whole-ship pressure field prediction technology can dynamically calculate the pressure distribution on the hull surface by combining real-time or predicted sea state information, thereby accurately obtaining key load parameters such as bending moment and torque, providing a reliable data foundation for real-time assessment of structural strength. Under severe sea conditions, this technology can predict loads in a timely manner, determine whether the design limits are approaching, and provide decision support for course and speed optimization. This will help overcome the limitations of traditional safety assessment methods and significantly improve the navigation safety of ships. Summary of the Invention

[0005] The purpose of this invention is to provide a method, program, device and storage medium for intelligent prediction of ship surface fluid pressure field based on a combination of reduced-order model and neural network, which can quickly obtain the surface pressure distribution of target ship type under specific operating conditions.

[0006] A smart prediction method for ship surface fluid pressure field based on a combination of reduced-order model and neural network includes the following steps:

[0007] Determine the predicted operating conditions and obtain a sample dataset of ship types;

[0008] For each ship type sample, the structural design parameter variables of the ship type are obtained, and the hydrodynamic pressure at each grid node of the hull when the ship type is sailing under the predicted working conditions is obtained through numerical simulation method, and then a surrogate model of the ship type sample is constructed.

[0009] Set up structured grid cells, divide each ship type sample according to the structured grid cells, and calculate the cosine vector of each structured grid cell pointing to the outside of the hull.

[0010] Based on the surrogate model of each ship type sample, the predicted value of the hydrodynamic pressure at the center point of each structured grid cell when the ship type is sailing under the predicted operating conditions is obtained.

[0011] Arrange the cosine vectors of each structured grid cell in each ship type sample to construct a structured cosine matrix; arrange the predicted values ​​of the hydrodynamic pressure at the center point of each structured grid cell in each ship type sample to construct a structured hydrodynamic pressure matrix.

[0012] Based on the reduced-order model, the main fundamental modes of the structured cosine matrix and the structured hydrodynamic pressure matrix are obtained, and then the main fundamental coefficients of each ship type sample with respect to the structured cosine matrix and the structured hydrodynamic pressure matrix are calculated.

[0013] Construct the first training set, which includes the main basis coefficients of the ship structural design parameter variables and the structured hydrodynamic pressure matrix corresponding to each ship type sample; use the first training set to train the first neural network model;

[0014] A second training set is constructed, which includes the ship structure design parameter variables and the main basis coefficients of the structured cosine matrix corresponding to each ship type sample; the second training set is used to train the second neural network model.

[0015] The ship's structural design parameters are obtained and input into the trained first and second neural network models respectively. The main basis coefficients of the structured cosine matrix and the structured hydrodynamic pressure matrix are predicted. The cosine vector of each structured grid cell of the ship to be predicted and the predicted value of the hydrodynamic pressure at the center point are obtained. Then, the resultant force of the hydrodynamic pressure on the ship under the predicted conditions is predicted based on numerical integration.

[0016] Furthermore, the proxy model for constructing the ship type sample is specifically as follows:

[0017]

[0018] in, For the first A proxy model for ship type samples; For the first The first of the ship type samples Grid nodes, , For the first The total number of grid nodes in the ship type sample; , , , For parameter vectors; , The first obtained through numerical simulation method When the ship type is sailing under the predicted operating conditions, the first The fluid dynamic pressure at the grid nodes; ; ; For smoothness parameters, ; For length scale parameters, ; It is a gamma function; This is a modified Bessel function of the second kind; To calculate the first The first of the ship type samples Grid nodes and the first The Euclidean distance between grid nodes; , It is random noise.

[0019] Furthermore, the structured mesh unit adopts triangular mesh units, with a total of [number missing]. ;

[0020] The first After the ship type sample is divided into structured grid cells, the first... The coordinates of the three vertices of the structured mesh cell are as follows: , ,

[0021] No. The center point of the structured grid cell is , , , ; ;

[0022] Calculate the normal vector ;

[0023]

[0024]

[0025]

[0026] Calculate the first The center of all structured grid cells in the ship type sample ;

[0027] , ,

[0028] Constructing vectors ,like Then, the inverse of this normal vector is taken as the normal vector pointing outwards from the hull, that is, let... ;

[0029] The normal vector pointing outwards from the hull Normalization yields the first... The first of the ship type samples The structured mesh element corresponds to the cosine vector pointing outwards from the hull. ;

[0030]

[0031] Furthermore, the proxy model based on each ship type sample To obtain the center points of each structured grid cell when the ship type is sailing under the predicted operating conditions. Predicted value of the fluid dynamic pressure at the location ;

[0032] Arrange the cosine vectors of each structured grid cell in the ship type samples to construct a structured cosine matrix. ;

[0033]

[0034] in, ;

[0035] The predicted values ​​of the hydrodynamic pressure at the center point of each structured grid cell for each ship type sample are arranged to construct a structured hydrodynamic pressure matrix. ;

[0036]

[0037] in, .

[0038] Furthermore, the structured hydrodynamic pressure matrix is ​​obtained based on the reduced-order model. The main fundamental modes:

[0039] Constructing a matrix Solve the matrix eigenvalues With feature vectors ;

[0040] If there exists a positive number satisfy Then As a matrix The positive singular values ​​of the eigenvectors As positive singular values The corresponding feature vector;

[0041] Obtain the structured fluid dynamic pressure matrix All positive singular values ​​are sorted in descending order, and the index of the positive singular values ​​is reconstructed as follows: , For matrix The total number of positive singular values;

[0042] Based on the reduction parameters Calculate the structured fluid dynamic pressure matrix Reduced-order cutoff position , , will go A structured fluid dynamic pressure matrix eigenvectors corresponding to positive singular values The primary fundamental mode;

[0043] Then, the structured hydrodynamic pressure matrix of each ship type sample is calculated. Main base coefficients ;

[0044] ,

[0045] Furthermore, the structured cosine matrix is ​​obtained based on the reduced-order model. The main fundamental modes and the acquisition of the structured fluid dynamic pressure matrix The main fundamental modes are the same;

[0046] Based on the structured cosine matrix Main fundamental modes Calculate the structured cosine matrix for each ship type sample. Main base coefficients ;

[0047] , .

[0048] Furthermore, the ship hull design parameters of the vessel to be predicted are obtained. The inputs are fed into the trained first and second neural network models respectively to obtain the predicted main basis coefficients of the structured cosine matrix. Prediction of the main basis coefficients of the structured fluid dynamic pressure matrix The cosine vectors of each structured grid cell of the ship to be predicted are obtained. The predicted value of the hydrodynamic pressure at the center point :

[0049] ,

[0050] Then, based on numerical integration, the resultant force of the fluid pressure experienced by the ship under the predicted operating conditions is predicted. :

[0051]

[0052] in, For the first The first of the ship type samples The area of ​​a structured grid cell.

[0053] A computer device includes a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network.

[0054] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described intelligent prediction method for ship surface fluid pressure fields based on a combination of a reduced-order model and a neural network.

[0055] A computer program product includes computer instructions that, when executed by a processor, implement the steps of the above-described intelligent prediction method for ship surface fluid pressure fields based on a combination of a reduced-order model and a neural network.

[0056] The beneficial effects of this invention are as follows:

[0057] This invention utilizes a surrogate model to map unstructured mesh data obtained from numerical simulations into structured mesh data, ensuring a consistent data processing format. Based on this, a reduced-order model is used to extract the main fundamental modes of the surface pressure field and its corresponding mesh node normal vectors. Subsequently, a fully connected neural network is employed to rapidly predict the fundamental coefficients of the new ship type. Thus, for new ship types, only design parameters need to be input, without requiring input of the hull mesh, to achieve rapid prediction of the hull surface load distribution, external normal vector distribution, and compressive and drag forces. This provides reliable technical support for ship structural strength design and safe navigation at sea. Attached Figure Description

[0058] Figure 1 This is a diagram of the overall architecture of the present invention. Detailed Implementation

[0059] The present invention will now be further described with reference to the accompanying drawings.

[0060] The core innovation of this invention lies in integrating three technical approaches to propose an efficient intelligent prediction method for ship surface fluid pressure fields. First, a Kriging surrogate model is used to map mesh relationships, ensuring consistency in subsequent data processing. Then, a reduced-order model is used to extract the fundamental modes of the surface pressure field and mesh node normal vector data, retaining key fundamental modes and their corresponding basic coefficients by truncation. Finally, a neural network is combined to rapidly predict the basic coefficients for unknown ship types, thereby quickly predicting the distribution of the ship surface pressure field. Based on this method, after inputting the design parameters of the target ship, the basic coefficients corresponding to the hull surface mesh normal vectors and pressure fields can be calculated without inputting the hull mesh, ultimately obtaining the ship surface pressure field distribution and pressure drag values ​​quickly and accurately.

[0061] A smart prediction method for ship surface fluid pressure field based on a combination of reduced-order model and neural network includes the following steps:

[0062] Step 1: Determine the predicted operating conditions and obtain the ship type sample dataset. Each ship type sample includes the structural design parameter variables of the ship type, as well as the hydrodynamic pressure of the ship type at each grid node under the predicted operating conditions obtained by numerical simulation.

[0063] For the Ship type sample, ship type structural design parameter variables are , For the first The first of the ship type samples Ship structure design parameters , , This represents the total number of ship type samples. This represents the total number of ship structural design parameters;

[0064] Establish a Cartesian coordinate system, the first... The total number of grid nodes in the ship type sample is , No. The coordinates of the grid nodes are ; Under the predicted operating conditions, the first The first of the ship type samples The hydrodynamic pressure at the grid node is ;

[0065] If the above data shows excessive fluctuations, it can be standardized to form standardized data with a mean of 0 and a variance of 1.

[0066] Step 2: For each ship type sample, based on the coordinates of each grid node... and the fluid dynamic pressure it experiences Build a proxy model The surrogate model outputs the predicted value of the fluid dynamic pressure at the input node based on the node coordinates.

[0067] While ensuring that the number of grid nodes and topological relationships of the "structured" ship types in each sample are consistent, the mapping relationship between the un"structured" grid data of each sample ship type and its corresponding "structured" grid data is established through the Kriging proxy model.

[0068] The Kriging method primarily assumes that the constructed proxy model follows a Gaussian process:

[0069] Step 2.1: Construct the basis function vector With parameter vector Construct the mean function , and thus construct the first Mean function term matrix of ship type samples ;

[0070] Step 2.2: Take the covariance function as the Matern function and consider random noise. Construct the covariance matrix ;

[0071]

[0072] in, ; The smoothness parameter determines the differentiability of the covariance function. ; The length scale parameter controls the rate at which the correlation decays. ; The gamma function is a generalization of the factorial function to the real number field. This is a modified Bessel function of the second kind; To calculate the first The first of the ship type samples Grid nodes and the first Euclidean distance between grid nodes; random noise It can be taken as ;

[0073] Step 2.3: Solve for the parameter vector Smoothness parameters and length scale parameters , construct the first surrogate model of ship type sample ;

[0074] For the Any coordinate in the ship type sample is Nodes:

[0075]

[0076] in, ; ;

[0077] Step 3: Select triangular mesh elements as structured mesh elements and set the total number of structured mesh elements. For each ship type sample, it is divided into: Each structured grid cell is used to obtain the predicted value of the hydrodynamic pressure at the center point of each structured grid cell under the predicted operating conditions based on the surrogate model of the ship type sample.

[0078] For the Ship type samples are divided into After the first structured mesh cell, the second... The coordinates of the three vertices of the structured mesh cell are as follows: , , Then the first The coordinates of the center point of the structured grid cell are , , , ; and then calculate the first Area of ​​structured grid cells ;

[0079]

[0080] Will Enter the first surrogate model of ship type sample ,get Predicted value of hydrodynamic pressure under the predicted operating conditions ; ;

[0081] Construct the first Structured hydrodynamic pressure vector of ship hull sample ;

[0082] Then, a structured hydrodynamic pressure matrix is ​​constructed for all ship type samples. ;

[0083] Step 4: For each ship type sample, calculate the cosine vector of each structured mesh element pointing outwards from the hull.

[0084] For the The first of the ship type samples Structured mesh cells, calculate normal vectors ;

[0085]

[0086]

[0087]

[0088] Calculate the first The center of all structured grid cells in the ship type sample ;

[0089] , ,

[0090] Constructing vectors ,like Then, the inverse of this normal vector is taken as the normal vector pointing outwards from the hull, that is, let... ;

[0091] The normal vector pointing outwards from the hull Normalization yields the first... The first of the ship type samples The structured mesh element corresponds to the cosine vector pointing outwards from the hull. ;

[0092]

[0093] Construct the first Structured cosine matrix of ship type sample ;

[0094] Then, a structured cosine matrix for all ship type samples is constructed. ;

[0095] Step 5: Structured hydrodynamic pressure matrix for all ship type samples based on the reduced-order model With cosine matrix Perform dimensionality reduction to obtain the matrix Main fundamental modes With matrix Main fundamental modes ;

[0096] Constructing a matrix Solve the matrix eigenvalues With feature vectors ;

[0097] Constructing a matrix Solve the matrix eigenvalues With feature vectors ;

[0098] If there exists a positive number satisfy Then As a matrix The positive singular values ​​of the eigenvectors As positive singular values The corresponding eigenvectors; obtaining the matrix All positive singular values ​​are sorted in descending order, and the index of the positive singular values ​​is reconstructed as follows: , For matrix The total number of positive singular values, ;

[0099] If there exists a positive number satisfy Then As a matrix The positive singular values ​​of the eigenvectors As positive singular values The corresponding eigenvectors; obtaining the matrix All positive singular values ​​are sorted in descending order, and the index of the positive singular values ​​is reconstructed as follows: , For matrix The total number of positive singular values, ;

[0100] Based on the reduction parameters Calculate the matrix Reduced-order cutoff position , , will go matrix eigenvectors corresponding to positive singular values As a matrix The main fundamental modes;

[0101] Based on the reduction parameters Calculate the matrix Reduced-order cutoff position , , will go matrix eigenvectors corresponding to positive singular values As a matrix The main fundamental modes;

[0102] Step 6: Based on the matrix Main fundamental modes Calculate the main base coefficients for each ship type sample. , , Construct samples of each ship type with respect to the matrix. Main basis coefficient vector ;

[0103] According to the matrix Main fundamental modes Calculate the main base coefficients for each ship type sample. , , Construct samples of each ship type with respect to the matrix. Main basis coefficient vector ;

[0104] Step 7: Calculate the ship structure design parameters for each ship type sample. With regard to the matrix Main basis coefficient vector The first training set is constructed, and the first neural network model is trained using the first training set, enabling the first neural network model to output a matrix based on the input ship structure design parameters. Prediction of the main basis coefficient vector;

[0105] Ship structure design parameter variables corresponding to each ship type sample With regard to the matrix Main basis coefficient vector This is used to construct a second training set, and a second neural network model is trained on this set. This enables the second neural network model to output a matrix based on the input ship structure design parameters. Prediction of the main basis coefficient vector;

[0106] Step 8: Obtain the hull structure design parameters of the ship to be predicted. The data are then input into the trained first and second neural network models respectively to obtain the predicted main basis coefficient vectors. and Then, the structured hydrodynamic pressure matrix of the ship to be predicted is calculated. With structured cosine matrix prediction ;

[0107] ,

[0108] This allows for the prediction of the resultant force of fluid pressure experienced by the ship under the predicted operating conditions. ;

[0109]

[0110] The resultant force of fluid pressure The component along the ship's length is the compressive and drag forces experienced by the ship under the predicted operating conditions.

[0111] A smart forecasting system for ship surface fluid pressure field based on a combination of reduced-order models and neural networks consists of the following modules: a surrogate model construction module for grid data mapping, a field data reduction module for acquiring all fundamental modes, a truncation module for filtering the main fundamental modes, a neural network module for predicting the pressure field and normal vector fundamental mode coefficients, and a numerical integration module for achieving rapid forecasting of pressure resistance.

[0112] This invention considers the pressure distribution on the hull surface of different ship types under set operating conditions, enabling the estimation of load distribution and normal vector distribution on the hull surface for any ship type to be predicted. This allows for rapid prediction of the ship's pressure drag within the monitored sea area based on numerical integration. First, the unstructured mesh data obtained from numerical simulation is mapped to a structured mesh using a Kriging surrogate model, ensuring data consistency. Then, the main fundamental modes of the hull surface pressure field and corresponding mesh node normal vectors are extracted using a reduced-order model. Finally, a fully connected neural network is used to quickly predict the fundamental coefficients of the new ship type, thereby predicting its surface load distribution and normal vector distribution, and ultimately its corresponding pressure drag. This method provides efficient and reliable technical support for ship hydrodynamic performance prediction and structural strength assessment.

[0113] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart prediction method for ship surface fluid pressure field based on a combination of reduced-order model and neural network, characterized in that: Determine the predicted operating conditions and obtain a sample dataset of ship types; For each ship type sample, the structural design parameter variables of the ship type are obtained, and the hydrodynamic pressure at each grid node of the hull when the ship type is sailing under the predicted working conditions is obtained through numerical simulation method, and then a surrogate model of the ship type sample is constructed. Set up structured grid cells, divide each ship type sample according to the structured grid cells, and calculate the cosine vector of each structured grid cell pointing to the outside of the hull. Based on the surrogate model of each ship type sample, the predicted value of the hydrodynamic pressure at the center point of each structured grid cell when the ship type is sailing under the predicted operating conditions is obtained. Arrange the cosine vectors of each structured grid cell in each ship type sample to construct a structured cosine matrix; The predicted values ​​of the hydrodynamic pressure at the center point of each structured grid cell of each ship type sample are arranged to construct a structured hydrodynamic pressure matrix. Based on the reduced-order model, the main fundamental modes of the structured cosine matrix and the structured hydrodynamic pressure matrix are obtained, and then the main fundamental coefficients of each ship type sample with respect to the structured cosine matrix and the structured hydrodynamic pressure matrix are calculated. Construct the first training set, which includes the main basis coefficients of the ship structural design parameter variables and the structured hydrodynamic pressure matrix corresponding to each ship type sample; use the first training set to train the first neural network model; A second training set is constructed, which includes the ship structure design parameter variables and the main basis coefficients of the structured cosine matrix corresponding to each ship type sample; the second training set is used to train the second neural network model. The ship's structural design parameters are obtained and input into the trained first and second neural network models respectively. The main basis coefficients of the structured cosine matrix and the structured hydrodynamic pressure matrix are predicted. The cosine vector of each structured grid cell of the ship to be predicted and the predicted value of the hydrodynamic pressure at the center point are obtained. Then, the resultant force of the hydrodynamic pressure on the ship under the predicted conditions is predicted based on numerical integration.

2. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network as described in claim 1, characterized in that: The proxy model for constructing the ship type sample is specifically as follows: in, For the first A proxy model for ship type samples; For the first The first of the ship type samples Grid nodes, , For the first The total number of grid nodes in the ship type sample; , , , For parameter vectors; , The first obtained through numerical simulation method When the ship type is sailing under the predicted operating conditions, the first The fluid dynamic pressure at the grid nodes; ; ; For smoothness parameters, ; For length scale parameters, ; It is a gamma function; This is a modified Bessel function of the second kind; To calculate the first The first of the ship type samples Grid nodes and the first The Euclidean distance between grid nodes; , It is random noise.

3. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network as described in claim 1, characterized in that: The structured mesh unit adopts triangular mesh units, with a total of [number missing]. ; The first After the ship type sample is divided into structured grid cells, the first... The coordinates of the three vertices of the structured mesh cell are as follows: , , No. The center point of the structured grid cell is , , , ; ; Calculate the normal vector ; Calculate the first The center of all structured grid cells in the ship type sample ; ， ， Constructing vectors ,like Then, the inverse of this normal vector is taken as the normal vector pointing outwards from the hull, that is, let... ; The normal vector pointing outwards from the hull Normalization yields the first... The first of the ship type samples The structured mesh element corresponds to the cosine vector pointing outwards from the hull. ; 。 4. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network as described in claim 3, characterized in that: The proxy model based on each ship type sample To obtain the center points of each structured grid cell when the ship type is sailing under the predicted operating conditions. Predicted value of the fluid dynamic pressure at the location ; Arrange the cosine vectors of each structured grid cell in the ship type samples to construct a structured cosine matrix. ; in, ; The predicted values ​​of the hydrodynamic pressure at the center point of each structured grid cell for each ship type sample are arranged to construct a structured hydrodynamic pressure matrix. ; in, .

5. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network according to claim 4, characterized in that: The structured fluid dynamic pressure matrix is ​​obtained based on the reduced-order model. The main fundamental modes: Constructing a matrix Solve the matrix eigenvalues With feature vectors ; If there exists a positive number satisfy Then As a matrix The positive singular values ​​of the eigenvectors As positive singular values The corresponding feature vector; Obtain the structured fluid dynamic pressure matrix All positive singular values ​​are sorted in descending order, and the index of the positive singular values ​​is reconstructed as follows: , For matrix The total number of positive singular values; Based on the reduction parameters Calculate the structured fluid dynamic pressure matrix Reduced-order cutoff position , , will go A structured fluid dynamic pressure matrix eigenvectors corresponding to positive singular values The primary fundamental mode; Then, the structured hydrodynamic pressure matrix of each ship type sample is calculated. Main base coefficients ; ， 。 6. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network as described in claim 5, characterized in that: The structured cosine matrix is ​​obtained based on the order reduction model. The main fundamental modes and the acquisition of the structured fluid dynamic pressure matrix The main fundamental modes are the same; Based on the structured cosine matrix Main fundamental modes Calculate the structured cosine matrix for each ship type sample. Main base coefficients ; ， 。 7. The intelligent prediction method for ship surface fluid pressure field based on a combination of a reduced-order model and a neural network according to claim 6, characterized in that: Obtain the ship structure design parameters of the vessel to be predicted. The inputs are fed into the trained first and second neural network models respectively to obtain the predicted main basis coefficients of the structured cosine matrix. Prediction of the main basis coefficients of the structured fluid dynamic pressure matrix The cosine vectors of each structured grid cell of the ship to be predicted are obtained. The predicted value of the hydrodynamic pressure at the center point : ， Then, based on numerical integration, the resultant force of the fluid pressure experienced by the ship under the predicted operating conditions is predicted. : in, For the first The first of the ship type samples The area of ​​a structured grid cell.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that: When executed by a processor, the computer instructions implement the steps of the method according to any one of claims 1 to 7.

Images