Hull line design method, device and medium based on specified wake field

Abstract

The application discloses a ship hull line design method and device based on a specified wake field, and a medium, and relates to the technical field of ship design. The method comprises the following steps: performing regional feature transformation processing on an initial ship hull surface to obtain a plurality of to-be-predicted ship hull surfaces; inputting the plurality of to-be-predicted ship hull surfaces into a wake field prediction model to obtain a plurality of predicted wake fields; comparing the predicted wake fields with preset specified wake fields to obtain a comparison result; optimizing a previous ship hull surface based on the comparison result, design requirements and optimization targets to obtain an optimized previous ship hull surface; and inputting the optimized previous ship hull surface into the wake field prediction model until the comparison result meets preset requirements, so that it is determined that the ship hull line design is completed. The application is used to solve the problems of long time consumption, low efficiency and poor effect in the prior art when a ship hull line is designed, and the ship hull line meeting the design requirements can be quickly and accurately obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship design, and in particular to a ship hull line design method based on a specified wake field, a device and a medium. BACKGROUND

[0002] The ship wake field is closely related to the propulsion performance of the ship and the energy-saving effect of various energy-saving devices during navigation, and also greatly affects the operating environment of the ship, thereby significantly affecting the cavitation and noise performance of the ship.

[0003] The existing ship design is generally completed in stages. First, the naked ship light body line type is optimized to obtain a ship type with low naked ship hull resistance and a relatively uniform wake field, and then the matching design of the propeller and the design of the hydrodynamic energy-saving device are carried out based on this. The entire design process is based on design experience to continuously attempt design, and is verified in combination with pool tests. The overall time consumption is relatively long, and the efficiency is relatively low.

[0004] However, the energy-saving effect of the naked ship performance excellent ship type matched with the hydrodynamic energy-saving device is not necessarily the best, or the naked ship resistance lowest and the wake field most uniform ship type is not necessarily conducive to the design of the energy-saving device, and is not necessarily the best comprehensive navigation performance. The final effect is not ideal. SUMMARY

[0005] The present application proposes a ship hull line design method based on a specified wake field, a device and a medium to solve the problems of long time consumption, low efficiency and poor effect in the prior art when designing a ship hull line, and to quickly and accurately obtain a ship hull line meeting the design requirements.

[0006] The application embodiment provides a ship hull line design method based on a specified wake field, which comprises the following steps:

[0007] An initial ship hull surface corresponding to a mother ship is obtained, and regional feature transformation processing is performed on the initial ship hull surface to obtain a plurality of to-be-predicted ship hull surfaces;

[0008] The plurality of to-be-predicted ship hull surfaces are input into a pre-trained wake field prediction model to obtain a plurality of predicted wake fields output by the wake field prediction model, wherein the wake field prediction model is trained based on ship hull surface samples and wake field samples;

[0009] The predicted wake field is compared with a preset specified wake field to obtain a comparison result;

[0010] Based on the comparison result and the design requirement, an optimization target for the previous ship body surface is determined, and the previous ship body surface is optimized based on the design requirement and the optimization target to obtain an optimized ship body surface, and the optimized ship body surface is input into the wake field prediction model until the comparison result reaches a preset requirement, so that it is determined that the ship body line design is completed, wherein the previous ship body surface includes the initial ship body surface.

[0011] According to the ship body line design method based on the specified wake field provided in the embodiments of the present application, the optimization target includes an optimization region, an optimization direction and an optimization level;

[0012] The previous ship body surface is optimized based on the design requirement and the optimization target to obtain an optimized ship body surface, including:

[0013] The previous ship body surface is reconstructed to obtain a to-be-designed ship body surface;

[0014] For the to-be-designed ship body surface, the optimization region is designed based on the design requirement and the optimization target by using the optimization direction and the optimization level to obtain an optimized ship body surface.

[0015] According to the ship body line design method based on the specified wake field provided in the embodiments of the present application, before determining the optimization target for the previous ship body surface based on the comparison result and the design requirement, it further includes:

[0016] At least one optimization region conforming to the Gaussian distribution is determined from the to-be-designed ship body surface by using the center point coordinates and the first covariance matrix of the Gaussian function;

[0017] Based on the center point coordinates, the first covariance matrix and the design requirement, a to-be-processed control point is determined from the optimization region, and a design parameter of the to-be-processed control point is determined by using a Gaussian kernel function;

[0018] Based on the design parameter, the optimization direction and the optimization level are determined, wherein the positive and negative of the design parameter are used to determine the optimization direction, and the size of the design parameter is used to determine the optimization level.

[0019] According to the ship body line design method based on the specified wake field provided in the embodiments of the present application, based on the center point coordinates, the first covariance matrix and the design requirement, the to-be-processed control point is determined from the optimization region, including:

[0020] Based on a preset control point selection index formula, the center point coordinates and the first covariance matrix, the to-be-processed control point is determined from the optimization region;

[0021] The control point selection index formula includes:

[0022]

[0023] wherein, ρ P (x) represents a control point selection index, the control point selection index is obtained based on the design requirement, μ n represents a center point coordinate of the nth Gaussian function, σ n represents a first covariance matrix of the nth Gaussian function, x represents a coordinate of a control point to be processed, and N represents a number of Gaussian functions, and the center point coordinate is determined based on the design requirement.

[0024] According to the ship hull line design method based on a specified companion flow field provided in the embodiments of the present application, the design parameters of the control point to be processed are determined by using a Gaussian kernel function, which comprises the following steps.

[0025] The Gaussian kernel function is obtained based on the control point selection index.

[0026] The Gaussian kernel function comprises:

[0027]

[0028] wherein, x' represents a transpose of x, and l represents an Euclidean distance from the coordinate of the control point to be processed to the center point coordinate.

[0029] Based on an eigenvalue decomposition formula, a characteristic basis function corresponding to the Gaussian kernel of the Gaussian kernel function and an eigenvalue corresponding to the characteristic basis function are calculated.

[0030] The eigenvalue decomposition formula comprises:

[0031]

[0032] wherein, represents a characteristic basis function, represents an eigenvalue.

[0033] Based on a modal decomposition formula, the characteristic basis function and the eigenvalue, the control point of the ship hull surface to be designed is projected into an n-dimensional characteristic space to obtain the design parameters.

[0034] The modal decomposition formula comprises:

[0035]

[0036] wherein, represents the control point of the ship hull surface to be designed, A i represents a design parameter, represents an eigenvalue corresponding to the ith-dimensional characteristic space, represents a characteristic basis function corresponding to the ith-dimensional characteristic space.

[0037] The ship hull line design method based on the specified wake field provided in the embodiments of the present application, the design requirements include: design targets and design constraints;

[0038] Before optimizing the previous ship hull surface based on the design requirements and the optimization target, further comprising:

[0039] In the case of determining that the first data dimension corresponding to the predicted wake field is less than the second data dimension corresponding to the mother ship type, a preset optimization function needs to be met;

[0040] The optimization function includes:

[0041]

[0042] x opt represents the optimization result, and J(x) represents a loss function;

[0043] The loss function includes:

[0044]

[0045] wherein x represents the coordinates of the control points of the ship hull surface in the optimization process, u represents the predicted wake field, H() represents an observation operator, represents the transpose of the variable matrix inside the double vertical line multiplied by the subscript matrix multiplied by the variable matrix inside the double vertical line, wherein t=P, R, O and W, P, R, O and W respectively represent the covariance matrix of the variable matrix inside the double vertical line, represents the prior distribution of the input ship type information, representing the influence of prior physical knowledge, x f represents the coordinates of the control points of the initial ship hull surface, H o (u) is the part of the observation operator set as the optimization target, H r (x) is the part of the observation operator set as the constraint, is the specified performance parameter quantity as the constraint, J r (x) is the regularization term of the loss function, which is the part of directly constraining the input parameters, and λ1 and λ2 are the weights of the two constraint terms, representing the strength of the constraint, is the constraint on the ship hull geometry input, is the high-order derivative quantity of the constraint on the ship hull geometry.

[0046] The ship hull line design method based on the specified wake field provided in the embodiments of the present application, after inputting the optimized ship hull surface into the wake field prediction model, further comprising:

[0047] Obtaining the predicted wake field output by the wake field prediction model in the d+1 iteration process;

[0048] calculating a corresponding average wake field and a second covariance matrix of d+1 times;

[0049] calculating a Kalman gain matrix based on the second covariance matrix;

[0050] optimizing a next hull surface based on the Kalman gain matrix, the design requirements and the optimization target.

[0051] According to the ship hull line design method based on a specified wake field provided in the embodiments of the present application, after the Kalman gain matrix is calculated based on the second covariance matrix, the method further comprises:

[0052] determining a correction term corresponding to the dth iteration process, and optimizing a next hull surface based on the correction term;

[0053] wherein the correction term comprises:

[0054]

[0055] wherein λ2 represents a weight, represents a constraint on the hull geometry input at the dth iteration, represents a high-order derivative quantity for constraining the hull geometry, represents a transpose of a variable matrix inside the double vertical lines multiplied by W and then multiplied by the variable matrix inside the double vertical lines, and W represents a covariance matrix of the variable matrix inside the double vertical lines.

[0056] The embodiments of the present application also provide an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the ship hull line design method based on a specified wake field according to any one of the above embodiments when executing the program.

[0057] The embodiments of the present application also provide a non-transitory computer readable storage medium having a computer program stored thereon, and the computer program is executable on a processor to implement the steps of the ship hull line design method based on a specified wake field according to any one of the above embodiments.

[0058] The ship hull line design method, device and medium based on the specified wake field provided by the embodiments of the present application, by obtaining an initial ship hull surface corresponding to a mother ship, and performing regional feature transformation processing on the initial ship hull surface, a plurality of to-be-predicted ship hull surfaces are obtained; the plurality of to-be-predicted ship hull surfaces are input into a wake field prediction model, and a plurality of predicted wake fields output by the wake field prediction model are obtained; the predicted wake fields are compared with the specified wake field, and a comparison result is obtained; based on the comparison result and a design requirement, an optimization target for the last ship hull surface is determined, and the last ship hull surface is optimized based on the design requirement and the optimization target, to obtain an optimized last ship hull surface; the optimized last ship hull surface is input into the wake field prediction model, until the comparison result meets a preset requirement, and the ship hull line design is determined to be completed. The embodiments of the present application adopt the wake field prediction model to perform forward prediction on the wake field, and iteratively design the ship type in a reverse direction based on the prediction result and the comparison result of the specified wake field, so as to predict the ship hull line meeting the design requirement information. It can be seen that the embodiments of the present application can quickly and accurately obtain the ship hull line meeting the design requirement (for example, any one or more of the bare ship performance, the energy-saving effect of the water dynamic energy-saving device, the bare ship resistance, and the navigation performance) and the specified wake field. BRIEF DESCRIPTION OF DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0060] Figure 1 is a flowchart of the ship hull line design method based on the specified wake field provided by the embodiments of the present application;

[0061] Figure 2 is a structural schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0063] The embodiment of the present application provides a ship hull line design method based on a specified wake field. The method can be applied to an intelligent terminal and can also be applied to a server. The present application takes the method applied to the server as an example for illustration, which is for example illustration and does not limit the protection scope of the present application, and some other illustrations in the embodiment are also for example illustration, which will not be illustrated one by one. The specific implementation of the method is as shown in Figure 1

[0064] In step 101, an initial ship hull surface corresponding to a mother ship is obtained, and regional feature transformation processing is performed on the initial ship hull surface to obtain a plurality of to-be-predicted ship hull surfaces.

[0065] Specifically, the initial ship hull surface is parameter sampled by using a set Kalman filtering algorithm to obtain a plurality of to-be-predicted ship hull surfaces. Specifically, the initial ship hull surface is forward integrated, and new data is integrated to obtain a new data set, and the set is updated through a Kalman filtering equation. The data here is point cloud data corresponding to the initial ship hull surface.

[0066] Specifically, which regions are processed can be set in advance according to actual needs of the user, or can be processed randomly.

[0067] In step 102, the plurality of to-be-predicted ship hull surfaces are input into a pre-trained wake field prediction model to obtain a plurality of predicted wake fields output by the wake field prediction model.

[0068] The wake field prediction model is trained based on ship hull surface samples and wake field samples.

[0069] In step 103, the predicted wake field is compared with a preset specified wake field to obtain a comparison result.

[0070] In step 104, based on the comparison result and a design requirement, an optimization target for the previous ship hull surface is determined, and the previous ship hull surface is optimized based on the design requirement and the optimization target to obtain an optimized ship hull surface. The optimized ship hull surface is input into the wake field prediction model until the comparison result meets a preset requirement, and it is determined that the ship hull line design is completed.

[0071] The previous ship hull surface includes the initial ship hull surface.

[0072] The design requirement includes a design target (any one or more of a bare ship performance corresponding to the specified wake field, an energy-saving effect of a water dynamic energy-saving device, a bare ship resistance, and a navigation performance) and a constraint condition (design constraint).

[0073] The optimization target includes an optimization region, an optimization direction, and an optimization level, etc. The optimization direction is used to represent the region change direction, and the optimization level is used to represent the limit threshold of the region change.​

[0074] The ship hull line design method based on the specified wake field provided by the embodiments of the present application obtains an initial ship hull surface corresponding to a mother ship, and performs regional feature transformation processing on the initial ship hull surface to obtain a plurality of to-be-predicted ship hull surfaces; inputs the plurality of to-be-predicted ship hull surfaces into a wake field prediction model to obtain a plurality of predicted wake fields output by the wake field prediction model; compares the predicted wake fields with a preset specified wake field to obtain a comparison result; determines an optimization target for a previous ship hull surface based on the comparison result and a design requirement, and optimizes the previous ship hull surface based on the design requirement and the optimization target to obtain an optimized previous ship hull surface; and inputs the optimized previous ship hull surface into the wake field prediction model until the comparison result meets a preset requirement, at which point it is determined that the ship hull line design is complete. The present application uses the wake field prediction model to perform forward prediction of the wake field, and iteratively designs the ship type in a reverse direction based on the prediction result and the comparison result of the specified wake field to predict a ship hull line that meets the design requirement information. It can be seen that the present application can quickly and accurately obtain a ship hull line that meets the design requirement (for example, any one or more of the bare ship performance, the energy-saving effect of the water dynamic energy-saving device, the bare ship resistance, and the navigation performance) and the specified wake field.

[0075] In one specific embodiment, the wake field prediction model is trained based on ship hull surface samples and wake field samples.

[0076] Specifically, a plurality of ship hull lines are obtained, and corresponding ship hull surfaces are obtained. A nominal wake field corresponding to the ship hull surface is obtained by means of computational fluid dynamics (CFO) and model testing. The ship hull surface is used as a ship hull surface sample, and the nominal wake field is used as a wake field sample to train the wake field prediction model.

[0077] For example, the mapping relationship between the ship hull surface sample and the wake field sample is represented by formula (1):

[0078] u'=f(s')………(1)

[0079] Wherein, u' represents the wake field characteristics corresponding to the wake field sample, for example, the fluid velocity field within 1.2 times the diameter of the propeller disc, s' represents the coordinates corresponding to the ship hull surface sample, expressed in the form of a three-dimensional point cloud, which is a set of spatial coordinate vectors, that is, u'={u' t ,t=1,...,T} and T is the number of point clouds.

[0080] Specifically, the mapping relationship between the ship hull surface sample and the wake field sample is used to train a forward model (wake field prediction model).

[0081] Wherein s' is the FV expression of the low-dimensional ship type parameter obtained by the 3D mFV method combining the point cloud recognition of the ship body surface. u' is the feature extraction method of the wake field based on POD analysis, which realizes the complete expression of the wake field through the low-dimensional decomposition coefficient combined with the feature function.

[0082] Wherein the wake field prediction model is based on a 3D convolutional neural network. The low-dimensional FV three-dimensional matrix and the POD coefficient are input, and a large number of historical calculation data are used to construct a training data set to train the deep neural network. The mapping relationship model obtained after training can realize the input of the specified ship type point cloud and the output of the wake field.

[0083] In one specific embodiment, the initial ship body surface is optimized based on design requirements and optimization objectives, and the specific implementation of obtaining the optimized initial ship body surface includes:

[0084] The previous ship body surface is reconstructed to obtain a to-be-designed ship body surface, and the to-be-designed ship body surface is designed and optimized in the optimization region based on the design requirements and optimization objectives using the optimization direction and optimization level to obtain the optimized ship body surface.

[0085] Specifically, the ship type of the mother ship is analyzed for the first time, the surface is reconstructed based on NURBS, and the ship body surface is designed again based on the previous ship body surface in the subsequent iteration process.

[0086] The present application is based on the regional modification design (regional feature transformation processing) of the feature analysis, which is based on the mother ship, and the optimization region is modified according to the design requirements, which realizes the purpose of generating batch ship type modification based on low-dimensional design parameters.

[0087] In one specific embodiment, based on the comparison result and the design requirement, at least one optimization region conforming to the Gaussian distribution is determined from the to-be-designed ship body surface using the center point coordinates and the first covariance matrix of the Gaussian function before determining the optimization objective for the previous ship body surface; based on the center point coordinates, the first covariance matrix and the design requirement, a to-be-processed control point is determined from the optimization region, and a design parameter of the to-be-processed control point is determined using the Gaussian kernel function; based on the design parameter, an optimization direction and an optimization level are determined.

[0088] Wherein the positive and negative of the design parameter are used to determine the optimization direction, and the size of the design parameter is used to determine the optimization level.

[0089] In one specific embodiment, based on the center point coordinates, the first covariance matrix and the design requirement, the specific implementation of determining the to-be-processed control point from the optimization region includes:

[0090] Based on the preset control point selection index formula, the center point coordinates and the first covariance matrix, the to-be-processed control point is determined from the optimization region.

[0091] wherein the control point selection index formula is shown in formula (2)

[0092]

[0093] wherein ρ P (x) represents the control point selection index, the control point selection index is obtained based on design requirements, μ n represents the center point coordinate of the nth Gaussian function, σ n represents the first covariance matrix of the nth Gaussian function, x represents the coordinate of the control point to be processed, and N represents the number of Gaussian functions. The center point coordinate is determined based on design requirements.

[0094] Specifically, the entire design process includes ship NURBS surface reconstruction, setting of the optimization region, calculation of the Gaussian kernel function, setting of the design parameters (obtained based on the optimization direction and the optimization magnitude), and generation of a new ship point cloud, etc.

[0095] wherein, in order to facilitate the selection of the control points in the optimization region and easy automatic integration optimization, the selection index function of the control points in the optimization region is calculated by using the Gaussian mixture function, and the control point selection index in the natural coordinate system (see formula 2). It only needs to specify the center point coordinate and variance to select a deformation region subject to Gaussian distribution.

[0096] In one specific embodiment, the specific implementation of determining the design parameters of the control point to be processed by using the Gaussian kernel function includes:

[0097] The Gaussian kernel function is obtained based on the control point selection index; the Gaussian kernel corresponding to the Gaussian kernel function and the characteristic value corresponding to the characteristic basis function are calculated based on the eigenvalue decomposition formula; the control points of the ship hull surface to be designed are projected into an n-dimensional characteristic space based on the modal decomposition formula, the characteristic basis function and the characteristic value, to obtain the design parameters.

[0098] wherein the Gaussian kernel function is shown in formula (3):

[0099]

[0100] wherein x' represents the transpose of x, and l represents the Euclidean distance from the coordinate of the control point to be processed to the center point coordinate.

[0101] wherein the eigenvalue decomposition formula is shown in formula (4):

[0102]

[0103] wherein, represents the characteristic basis function, represents the characteristic value.

[0104] Wherein, the modal decomposition formula is shown in formula (5):

[0105]

[0106] Wherein, A i represents the control point of the designed ship surface, represents the eigenvalue corresponding to the i-dimensional feature space, represents the eigenfunction corresponding to the i-dimensional feature space.

[0107] According to the NURBS surface theory, the NURBS control state parameters (coordinates of control points) x = [x p ,y p ,z p ,ω p ] of a certain ship surface are known, and the ship surface value point cloud can be calculated and obtained, and the two are equivalent, so x can be used to represent the ship surface. Wherein, x is obtained based on a three-dimensional rectangular coordinate system and the angle of the corresponding point cloud point. x can be set as f (μ n ,σ n ,A i ), the regional design parameters are (μ n ,σ n ) and the design parameters A i are automatically sampled, and a batch of ship types can be obtained. Wherein, the design region is controlled by μ n and σ n .

[0108] The present application simplifies the high-dimensional complex set design problem to a limited order, realizes the new control point coordinates based on low-dimensional design parameters, and further drives the ship surface deformation to realize the ship type modification.

[0109] In one embodiment, on the basis of the forward evaluation model, the ship type optimization process can be expressed as solving the following optimization problem, that is, given the parent ship, represented by the NURBS control point coordinates, denoted as x f , we want to find a new ship type x * , which meets the input geometric constraint condition (such as the displacement unchanged), and meets the specified target wake field (specified wake field) u * under the premise of the total resistance / received power performance minimum.

[0110] Based on the design requirements and optimization objectives, before optimizing the last ship surface, in the case that the first data dimension corresponding to the predicted wake field is less than the second data dimension corresponding to the parent ship type, the preset optimization function needs to be met.

[0111] where the optimization function is shown in equation (6):

[0112]

[0113] x opt where x represents the coordinates of the control points of the last hull surface, u represents the predicted wake field, H() represents the observation operator,

[0114] where the loss function is shown in equation (7):

[0115]

[0116] where x represents the coordinates of the control points of the last hull surface, u represents the predicted wake field, H() represents the observation operator, represents the transpose of the variable matrix inside the double vertical line multiplied by the subscript matrix multiplied by the variable matrix inside the double vertical line, where t = P, R, O, and W, P, R, O, and W represent the covariance matrix of the variable matrix inside the double vertical line, represents the prior distribution of the input ship type information, representing the influence of prior physical knowledge, H o (u) is the part of the observation operator set as the optimization objective, H r (x) is the part of the observation operator set as the constraint, is the specified performance parameter quantity as a constraint, J r (x) is the regularization term of the loss function, which is the part of directly constraining the input parameters, λ1 and λ2 are the weights of the two constraint terms, representing the strength of the constraint, is the constraint on the hull geometry input, indicates the high-order derivative quantity of the constraint on the hull geometry.

[0117] Specifically, for the specified wake field inverse design problem, the loss function is configured to meet the design requirements:

[0118] (1) Optimization objective part ||H o (u)||: expressed in terms of the minimum mean square error of the predicted wake field and the target wake field, i.e. the L2 norm of the design wake field and the target wake field, see equation (8):

[0119]

[0120] where u = f(x) is the wake field of any ship type, u * is the target wake field, and S is the number of data sampling points in the wake field region.

[0121] (2) Design constraint part represents the deviation of the ship performance constraint where the mapping relationship between the ship geometry and the ship performance parameters is represented by the continuity equation performance parameters such as total resistance or received power.

[0122] (3) Regularization term part J r (x): geometric constraints on the design result, such as displacement, position of the center of buoyancy, etc., represented as:

[0123] where, is the constraint on the hull geometry input, refers to the high-order derivative quantity of the constraint on the hull geometry, such as displacement, position of the center of buoyancy, etc.

[0124] Specifically, when the first data dimension is less than the second data dimension, the inverse problem is a static indeterminate problem, which can be equivalent to solving an optimization problem, i.e. formula (6).

[0125] In one specific embodiment, after inputting the optimized hull surface into the wake field prediction model, the predicted wake field output by the wake field prediction model in the d+1 iteration process is obtained; the average wake field and the second covariance matrix corresponding to d+1 are calculated; the Kalman gain matrix is calculated based on the second covariance matrix; the next hull surface is optimized based on the Kalman gain matrix, the design requirements and the optimization target.

[0126] Specifically, the optimization process replaces the original deterministic variable with the probability distribution of the variable based on the solution of the iterative form set Kalman filtering method for solving the uncertainty Bayesian inference problem, and the optimization problem is converted into finding the input variable x that maximizes p(x|u), and the output variable u = prediction result + ε, where ε represents the prediction error of the wake field prediction model.

[0127] The specific implementation is as follows:

[0128] Sampling:

[0129] On the basis of the parent ship, a Gaussian distribution Latin hypercube sampling method is used to generate a set of multiple hull surfaces to be predicted, for example, the set consists of M sequences where the set is obtained according to the prior distribution of the data.

[0130] Prediction:

[0131] The wake field prediction is performed on each sequence to obtain the prediction result in the d+1 iteration

[0132] Analysis:

[0133] Based on the first calculation formula, the average wake field corresponding to d+1 is calculated, and the second covariance matrix is calculated based on the second calculation formula.

[0134] wherein the first calculation formula is shown in formula (9):

[0135]

[0136] The second calculation formula is shown in formula (10):

[0137]

[0138] The Kalman gain matrix is calculated based on the third calculation formula.

[0139] wherein the third calculation formula is shown in formula (11):

[0140]

[0141] wherein H represents an observation matrix, a matrix corresponding to mapping of a control point from a state space to an observation space, R represents a covariance matrix of a prediction error, and Q represents an uncertainty of the observation matrix.

[0142] The d+1th corresponding set is updated based on the fourth calculation formula and the Kalman gain matrix, to obtain a new set.

[0143] wherein the fourth calculation formula is shown in formula (12):

[0144]

[0145] Specifically, adding constraints in the calculation process of the Kalman filtering method is a design requirement and an important means to avoid excessive illness of the optimization problem. Since the feedback of the Kalman filtering is searching in a Gaussian space, such constraints are generally necessary, otherwise the stability of the inference process will be reduced.

[0146] Specifically, the constraints can be divided into strong constraints and weak constraints. The strong constraints are required to be strictly satisfied and are imposed by a data preprocessing step, and must be accurately satisfied, otherwise the subsequent calculation, model processing or shipbuilding process will be affected. The weak constraints refer to that the final inference result can be as close as possible to the constraint requirement, and strict constraints cannot be guaranteed, which can be imposed in the form of a posteriori estimation.

[0147] In one specific embodiment, after the Kalman gain matrix is calculated based on the second covariance matrix, a correction term corresponding to the dth iteration process is determined; and the set is updated based on the correction term.

[0148] wherein the correction term is shown in formula (13):

[0149]

[0150] The fourth calculation formula after optimization is shown in formula (14):

[0151]

[0152] The present application comprises a forward prediction part based on a wake field prediction model and a reverse design part based on Bayesian inference. The corresponding wake field is obtained through the hull curve input by the user, and then the set Kalman filter algorithm in the form of iteration is applied to realize the hull line design based on Bayesian estimation, to obtain the hull line corresponding to the specified wake field, and realize the rapid design of the hull line.

[0153] Figure 2 An example of a schematic diagram of the physical structure of an electronic device is shown as Figure 2 The electronic device can include a processor 201, a communications interface 202, a memory 203, and a communications bus 204, wherein the processor 201, the communications interface 202, and the memory 203 communicate with each other through the communications bus 204. The processor 201 can call the logic instructions in the memory 203 to execute the hull line design method based on the specified wake field.

[0154] In addition, the logic instructions in the memory 203 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0155] On the other hand, the present application also provides a computer program product, which comprises a computer program stored on a non-transitory computer readable storage medium, and the computer program comprises program instructions, when the program instructions are executed by a computer, the computer can execute the hull line design method based on the specified wake field provided by the above-mentioned methods.

[0156] In yet another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, which is executed by a processor to implement the hull line design method based on the specified wake field provided by the above-mentioned embodiments.

[0157] The apparatus embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0158] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary general hardware platforms, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0159] Finally, it should be noted that: the above only describes the preferred embodiments of the present application, and the present application is not limited to the above embodiments. It can be understood that other improvements and changes directly derived or thought by those skilled in the art without departing from the spirit and concept of the present application should be considered to be included in the protection scope of the present application.

Claims

1. A method for designing ship hull lines based on a specified wake field, characterized in that, The method includes: Obtain the initial hull surface corresponding to the parent ship, and perform regional feature transformation processing on the initial hull surface to obtain multiple hull surfaces to be predicted; Multiple ship hull surfaces to be predicted are input into a pre-trained wake field prediction model to obtain multiple predicted wake fields output by the wake field prediction model, wherein the wake field prediction model is trained based on ship hull surface samples and wake field samples. The predicted wake field is compared with the preset specified wake field to obtain the comparison result; Based on the comparison results and design requirements, the optimization target for the previous hull surface is determined, and the previous hull surface is optimized based on the design requirements and the optimization target to obtain the optimized hull surface. The optimized hull surface is then input into the wake field prediction model until the comparison results meet the preset requirements, at which point the hull profile design is considered complete. The previous hull surface includes the initial hull surface.

2. The hull shape design method based on a specified wake field according to claim 1, characterized in that, The optimization objectives include: optimization area, optimization direction, and optimization magnitude; Based on the aforementioned design requirements and optimization objectives, the previous hull surface is optimized to obtain the optimized hull surface, including: Reconstruct the previous hull surface to obtain the hull surface to be designed; For the hull surface to be designed, the optimization region is designed based on the design requirements and the optimization objectives using the optimization direction and the optimization magnitude to obtain the optimized hull surface.

3. The hull shape design method based on a specified wake field according to claim 2, characterized in that, Before determining the optimization target for the previous hull surface based on the comparison results and design requirements, the following steps are also included: Using the coordinates of the center point of the Gaussian function and the first covariance matrix, at least one optimization region conforming to a Gaussian distribution is determined from the hull surface to be designed; Based on the center point coordinates, the first covariance matrix, and the design requirements, the control points to be processed are determined from the optimization region, and the design parameters of the control points to be processed are determined using the Gaussian kernel function. Based on the design parameters, the optimization direction and optimization magnitude are determined, wherein the sign of the design parameters is used to determine the optimization direction, and the magnitude of the design parameters is used to determine the optimization magnitude.

4. The hull shape design method based on a specified wake field according to claim 3, characterized in that, Based on the center point coordinates, the first covariance matrix, and the design requirements, the control points to be processed are determined from the optimization region, including: Based on the preset control point selection index formula, center point coordinates and first covariance matrix, the control point to be processed is determined from the optimization region; The formula for selecting control points includes: Where, ρ P (x) represents the control point selection index, which is obtained based on the design requirements, μ n σ represents the coordinates of the center point of the nth Gaussian function. n Let x represent the first covariance matrix of the nth Gaussian function, x represent the coordinates of the control point to be processed, and N represent the number of Gaussian functions. The coordinates of the center point are determined based on the design requirements.

5. The hull shape design method based on a specified wake field according to claim 4, characterized in that, The design parameters of the control points to be processed are determined using the Gaussian kernel function, including: The Gaussian kernel function is obtained based on the index selected from the control points. The Gaussian kernel function includes: Where x' represents the transpose of x, and l represents the Euclidean distance from the coordinates of the control point to be processed to the coordinates of the center point; Based on the eigenvalue decomposition formula, the characteristic basis functions corresponding to the Gaussian kernel of the Gaussian kernel function and the eigenvalues ​​corresponding to the characteristic basis functions are calculated. The eigenvalue decomposition formula includes: in, Describe the characteristic basis functions. Represents eigenvalues; Based on the modal decomposition formula, the characteristic basis function, and the eigenvalue, the control points of the hull surface to be designed are projected into an n-dimensional feature space to obtain the design parameters; The modal decomposition formula includes: in, A represents the control point of the hull surface to be designed. i Indicates design parameters, This represents the eigenvalue corresponding to the i-th dimensional feature space. Let represent the characteristic basis function corresponding to the i-th dimension of the characteristic space.

6. The hull shape design method based on a specified wake field according to any one of claims 1-5, characterized in that, The design requirements include: design objectives and design constraints; Before optimizing the previous hull surface based on the aforementioned design requirements and optimization objectives, the following steps are also included: Given that the first data dimension corresponding to the predicted wake field is smaller than the second data dimension corresponding to the mother ship type, the preset optimization function must be satisfied. The optimization function includes: x opt J(x) represents the optimization result, and J(x) represents the loss function. The loss function includes: Where x represents the coordinates of the control point on the previous hull surface, u represents the predicted wake field, and H() represents the observation operator. Let t represent the transpose of the variable matrix inside the double vertical lines, multiplied by the index matrix, and then multiplied by the variable matrix inside the double vertical lines, where t = P, R, O, and W, where P, R, O, and W represent the covariance matrices of the variable matrix inside the double vertical lines, respectively. The prior distribution of the input ship type information represents the influence of prior physical knowledge, x. f H represents the coordinates of the control points of the initial hull surface. o (u) represents the part of the observation operator that is set as the optimization objective, H r (x) represents the constrained part of the observation operator. To specify the performance parameter quantity as a constraint, J r (x) represents the regularization term of the loss function, which directly constrains the input parameters. λ1 and λ2 are the weights of the two constraint terms, indicating the strength of the constraints. To constrain the geometric input of the hull, This refers to a higher-order derived quantity that constrains the geometry of the ship's hull.

7. The hull shape design method based on a specified wake field according to any one of claims 1-5, characterized in that, After inputting the optimized hull surface into the wake field prediction model, the model also includes: Obtain the predicted wake field output by the wake field prediction model during the (d+1)th iteration; Calculate the mean wake field and the second covariance matrix corresponding to the (d+1)th order; Calculate the Kalman gain matrix based on the second covariance matrix; The next hull surface is optimized based on the Kalman gain matrix, the design requirements, and the optimization objective.

8. The hull shape design method based on a specified wake field according to claim 7, characterized in that, After calculating the Kalman gain matrix based on the second covariance matrix, the following steps are also included: Determine the correction term corresponding to the d-th iteration; optimize the next hull surface based on the correction term; The correction terms include: Where λ2 represents the weight, This represents the constraint on the hull geometry input corresponding to the d-th iteration. This refers to higher-order derived quantities that constrain the geometry of a ship's hull. This represents the transpose of the variable matrix inside the double vertical lines multiplied by W and then multiplied by the variable matrix inside the double vertical lines, where W represents the covariance matrix of the variable matrix inside the double vertical lines.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the hull line design method based on a specified wake field as described in any one of claims 1 to 8.

10. A non-transitory 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 hull line design method based on a specified wake field as described in any one of claims 1 to 8.

Images