Parametric design method for ship type based on lightweight network

Abstract

The application discloses a ship type parameterization design method based on a lightweight network, and comprises the following steps: obtaining ship type characteristic parameters for describing an existing sample ship type, and generating original NURBS ship type curved surfaces respectively; dividing the ship type curved surfaces according to a ship type curved surface division method, and exporting all NURBS control points of the divided curved surfaces as three-dimensional coordinate sequences; constructing a ship type data set based on the ship type characteristic parameters and the three-dimensional coordinate sequences of the control points; constructing a lightweight network model PDNN, taking the ship type characteristic parameters as network input and the three-dimensional coordinate sequences of the corresponding ship type curved surfaces as output; evaluating control point data obtained by prediction by using the ship type data set to train the PDNN network model, and generating ship type curved surfaces. The application solves the problems of complicated calculation and redundant parameters of a traditional ship type parameterization design method, can be used for rapid generation and optimization of ship type curved surfaces, and the algorithm of the application performs well in small targets and is superior to other models.

Description

Technical Field

[0001] This invention relates to the fields of ship design and artificial intelligence, and in particular to a parametric design method for ship hull based on lightweight networks. Background Technology

[0002] Parametric hull design is an efficient computer-aided design method for ships. It refers to establishing a special mapping relationship between hull geometric parameters and hull curves or surfaces, assisting designers in quickly generating hull surfaces. This is of great significance for expanding ship design space and optimizing ship design efficiency. Current parametric hull design mainly relies on a large number of geometric parameters, and then uses mathematical interpolation methods to fit hull curves and surfaces. Its calculation process is cumbersome and highly dependent on numerous complex implicit geometric parameters, making it difficult to widely apply in the parametric design of various types of ships.

[0003] With the continuous development of artificial intelligence in recent years, many scholars have begun to study its application in computer-aided design. In ship design, there are already practical applications based on neural network technology for ship resistance prediction and ship performance prediction, which have shown high reliability. Therefore, lightweight neural network methods can be used to address the problems of cumbersome calculation processes and redundant input parameters in traditional parametric ship design methods. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: obtain the ship type feature parameters describing the existing sample ship type, and generate the original NURBS ship type surface respectively; The ship type characteristic parameters are selected from typical parameters that can accurately describe the characteristics of the ship. The surface is divided according to the ship-shaped surface division method, and the NURBS control points of the divided surface are derived as a three-dimensional coordinate sequence. The method for dividing the ship-shaped surface includes dividing the original ship-shaped surface into 11 sub-surfaces along the characteristic curve of the hull, including 5 sub-surfaces at the bow, 4 sub-surfaces amidships, and 2 sub-surfaces at the stern; A total of 334 control points are required for each sub-surface of a single ship type. These are merged to obtain a three-dimensional coordinate sequence with a specified positional order. The sequence length is 3×334. A ship type dataset is constructed based on the ship type feature parameters and the three-dimensional coordinate sequence of control points; A lightweight network model PDNN is constructed, with the ship shape feature parameters as network input and the three-dimensional coordinate sequence of the corresponding ship shape surface as output; In the constructed lightweight network model PDNN, the input layer is the ship-shaped feature parameters, the hidden layer consists of one regularization layer and two fully connected layers with 128 neurons each, and a ReLU activation layer is set after the fully connected layer. The output layer is the three-dimensional coordinate sequence of the control points of the ship-shaped surface. A smoothing layer is added before the network output to perform prior normalization processing on the coordinate data. The PDNN network model is trained using the ship shape dataset to evaluate the predicted control point data and generate the ship shape surface.

[0007] The method for dividing the ship shape surface includes dividing the original ship shape surface into 11 sub-surfaces along the characteristic curve of the hull, including 5 sub-surfaces at the bow, 4 sub-surfaces amidships, and 2 sub-surfaces at the stern; a total of 334 control points are required for each sub-surface of a single ship shape, and the points are merged to obtain a three-dimensional coordinate sequence with a specified positional order, and the sequence length is 3×334.

[0008] The ship type dataset constructed based on the ship type feature parameters and the three-dimensional coordinate sequence of control points includes 173 ship type samples, of which 162 ship type samples are used as the training set and 11 ship type samples are used as the validation set.

[0009] include, In the constructed lightweight network model PDNN, the input layer is the ship-shaped feature parameters, the hidden layer consists of one regularization layer and two fully connected layers with 128 neurons each, and a ReLU activation layer is set after the fully connected layer. The output layer is the three-dimensional coordinate sequence of the control points of the ship-shaped surface. A smoothing layer is added before the network output to perform prior normalization processing on the coordinate data.

[0010] The construction of the lightweight network model PDNN includes, The weighted joint loss of MAE and MSE is used as the loss function. The calculation of the loss function Loss includes: ; Where n represents the number of coordinate data points. Represents the true value of coordinate data. These are the predicted values ​​of the coordinate data. , and represent the weights of MAE and MSE, respectively.

[0011] include, Adam is used as the optimizer, with an initial learning rate of 2×10. -3The iterator is a cosine annealing function, and training stops when the loss converges to a specified threshold.

[0012] The evaluation of the predicted control point data includes, Calculate the mean absolute percentage error between the target value and the predicted value to evaluate model performance; When the error of each parameter is less than 5%, the design requirements are met and the training is complete. Otherwise, adjust and optimize the network structure and retrain; The parameters include waterline length, drainage volume, immersion area, square coefficient, and rhombus coefficient.

[0013] include, The trained PDNN model is deployed in the computer system, including a target ship type feature parameter input module, a control point coordinate prediction module, a ship type surface fitting module, and an IGES file storage module. The target ship type characteristic parameter input module is used by users to input ship type characteristic parameters that meet design requirements; The control point coordinate prediction module is used to predict the coordinates of control points on the hull surface based on the input hull feature parameters; The IGES file storage module is used to store IGES files for ship types that meet design requirements.

[0014] The generation of the ship-shaped surface includes, The deployed PDNN parametric design system was used for actual testing. The target ship shape feature parameters were input, and the system performed coordinate prediction, surface fitting, and surface file output to finally generate the ship shape surface.

[0015] The beneficial effects of this invention are: This invention solves the problems of cumbersome calculations and redundant parameters in the traditional parametric design method for ship shapes, and can be used for the rapid generation and optimization of ship shape surfaces. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below 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. Wherein: Figure 1 A schematic diagram of the basic process of a ship hull parametric design method based on lightweight networks provided in an embodiment of the present invention; Figure 2 A schematic diagram of hull surface partitioning for a hull parametric design method based on lightweight networks provided in an embodiment of the present invention; Figure 3A schematic diagram of surface control points for a ship hull parametric design method based on lightweight networks provided in an embodiment of the present invention; Figure 4 A diagram of the lightweight network model PDNN for a ship hull parametric design method based on lightweight networks provided in an embodiment of the present invention. Figure 5 A comparison diagram of the generated ship form and the target ship form provided by the ship form parameterization design method based on lightweight networks in one embodiment of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0020] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0021] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0022] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Example

[0023] Reference Figures 1-4 As one embodiment of the present invention, a method for parametric design of ship hull based on lightweight networks is provided, comprising: The core of this invention mainly comprises three parts: the first part is the construction of the ship hull dataset, including the collection of ship hull feature parameters and the division of ship hull surfaces; the second part is the construction and optimization of a lightweight network model to better suit parametric ship hull design tasks; the third part is the deployment and application of the model, ultimately generating engineering-usable ship hull surface files; the process is as follows... Figure 1 As shown.

[0024] S1: Obtain the ship type feature parameters describing the existing sample ship types, and generate the original NURBS ship type surfaces respectively. It should be noted that: The ship type characteristic parameters specifically include, Twenty-one typical parameters that can accurately describe ship characteristics were selected as ship type characteristic parameters, including three categories of parameters: main dimensions, ship type coefficients, and other parameters.

[0025] S2: The surface is divided according to the ship-shaped surface partitioning method, and all NURBS control points of the partitioned surface are derived as a three-dimensional coordinate sequence. It should be noted that: The method for dividing the hull shape surface involves dividing the original hull shape surface into 11 sub-surfaces along the characteristic curves of the hull, including 5 sub-surfaces at the bow, 4 sub-surfaces amidships, and 2 sub-surfaces at the stern, as follows: Figure 2 As shown, the NURBS control points of each subsurface of the half-ship are merged and exported to obtain a three-dimensional coordinate sequence with a specified position. A total of 334 control points are required for each subsurface of a single ship type. The merged and exported three-dimensional coordinate sequence with a specified position has a sequence length of 3×334. Figure 3 The NURBS control point coordinates are shown using the upper part of the bow subsurface as an example.

[0026] S3: Construct a ship hull data set based on ship hull feature parameters and the 3D coordinate sequence of control points. It should be noted that: The ship type dataset, constructed based on ship type feature parameters and control point three-dimensional coordinate sequences, includes 173 ship type samples, of which 162 ship type samples are used as the training set and 11 ship type samples are used as the validation set.

[0027] S4: Construct a lightweight network model PDNN, using ship hull feature parameters as network input and the corresponding 3D coordinate sequence of the ship hull surface as output. It should be noted that: include, In the constructed lightweight network model PDNN, the structure of the lightweight network model PDNN is as follows: Figure 4 As shown, the input layer consists of ship-shaped feature parameters, the hidden layer consists of one regularization layer and two fully connected layers with 128 neurons each, and a ReLU activation layer is set after the fully connected layer. The output layer is the three-dimensional coordinate sequence of the control points of the ship-shaped surface. A smoothing layer is added before the network output to perform prior normalization processing on the coordinate data.

[0028] S5: A PDNN network model trained using the ship hull model dataset is used to evaluate the predicted control point data and generate the ship hull surface. It should be noted that: The construction of the lightweight network model PDNN includes, The weighted joint loss of MAE and MSE is used as the loss function. The calculation of the loss function Loss includes: ; Where n represents the number of coordinate data points. Represents the true value of coordinate data. These are the predicted values ​​of the coordinate data. , and represent the weights of MAE and MSE, respectively.

[0029] include, Adam is used as the optimizer, with an initial learning rate of 2×10. -3 The iterator is a cosine annealing function, and training stops when the loss converges to a specified threshold.

[0030] The evaluation of the predicted control point data includes, Calculate the mean absolute percentage error between the target value and the predicted value to evaluate model performance; When the error of each parameter is less than 5%, the design requirements are met and the training is complete. Otherwise, adjust and optimize the network structure and retrain; The parameters include waterline length, drainage volume, immersion area, square coefficient, and rhombus coefficient.

[0031] include, The trained PDNN model is deployed in the computer system, including a target ship type feature parameter input module, a control point coordinate prediction module, a ship type surface fitting module, and an IGES file storage module. The target ship type characteristic parameter input module is used by users to input ship type characteristic parameters that meet design requirements; The control point coordinate prediction module is used to predict the coordinates of control points on the hull surface based on the input hull feature parameters; The IGES file storage module is used to store IGES files for ship types that meet design requirements.

[0032] The generation of the ship-shaped surface includes, The deployed PDNN parametric design system was used for actual testing. The target ship shape feature parameters were input, and the system performed coordinate prediction, surface fitting, and surface file output to finally generate the ship shape surface.

[0033] This invention solves the problems of cumbersome calculations and redundant parameters in traditional parametric ship design methods, and can be used for the rapid generation and optimization of ship surface curves. Example

[0034] Reference Figure 5 This is the second embodiment of the present invention. Unlike the first embodiment, this embodiment provides a verification test of the ship hull parameterization design method based on lightweight networks. This is to verify and explain the technical effects used in this method, and to compare the test results with scientific demonstration methods to verify the real effect of this method.

[0035] In this embodiment, the target values ​​and predicted values ​​of five parameters—waterline length, drainage volume, submerged area, block coefficient, and rhombus coefficient—are compared to determine whether these five parameters meet the design requirements. Table 1 shows the error between the target values ​​and predicted values ​​of the ship type experimental data.

[0036] Table 1: Error table between target and predicted values ​​of ship type test data.

[0037] Waterline length / m 233.534 232.783 0.3226% qualified Drainage volume / m3 51237.215 51976.65 1.4226% qualified Flooded area / m2 10001.96 10066.37 0.6398% qualified Square coefficient / % 56.4 56.7 0.5291% qualified Rhombus coefficient / % 59.8 60.1 0.4992% qualified As shown in Table 1, the errors between the target and predicted values ​​of the five parameters—waterline length, drainage volume, immersion area, square coefficient, and rhombus coefficient—are all less than 5%, which meets the design requirements, and the training is complete.

[0038] Figure 5 The diagram shows a comparison between the predicted and target ship types. The three views are basically the same and meet the engineering requirements of smoothness and seamlessness. This proves that the proposed solution can effectively simplify the parametric design process and generate the required ship surface quickly and accurately.

[0039] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A parametric design method for ship hull forms based on lightweight networks, characterized in that, include: Obtain the ship type feature parameters describing the existing sample ship type, and generate the original NURBS ship type surface respectively; The ship type characteristic parameters are selected from typical parameters that can accurately describe the characteristics of the ship. The surface is divided according to the ship-shaped surface division method, and the NURBS control points of the divided surface are derived as a three-dimensional coordinate sequence. The method for dividing the ship-shaped surface includes dividing the original ship-shaped surface into 11 sub-surfaces along the characteristic curve of the hull, including 5 sub-surfaces at the bow, 4 sub-surfaces amidships, and 2 sub-surfaces at the stern; A total of 334 control points are required for each sub-surface of a single ship type. These are merged to obtain a three-dimensional coordinate sequence with a specified positional order. The sequence length is 3×334. A ship type dataset is constructed based on the ship type feature parameters and the three-dimensional coordinate sequence of control points; A lightweight network model PDNN is constructed, with the ship shape feature parameters as network input and the three-dimensional coordinate sequence of the corresponding ship shape surface as output; In the constructed lightweight network model PDNN, the input layer is the ship-shaped feature parameters, the hidden layer consists of one regularization layer and two fully connected layers with 128 neurons each, and a ReLU activation layer is set after the fully connected layer. The output layer is the three-dimensional coordinate sequence of the control points of the ship-shaped surface. A smoothing layer is added before the network output to perform prior normalization processing on the coordinate data. The PDNN network model is trained using the ship shape dataset to evaluate the predicted control point data and generate the ship shape surface.

2. The ship hull parametric design method based on lightweight networks as described in claim 1, characterized in that: The ship type dataset constructed based on the ship type feature parameters and the three-dimensional coordinate sequence of control points includes 173 ship type samples, of which 162 ship type samples are used as the training set and 11 ship type samples are used as the validation set.

3. The ship hull parametric design method based on lightweight networks as described in any one of claims 1 to 2, characterized in that: The construction of the lightweight network model PDNN includes, The weighted joint loss of MAE and MSE is used as the loss function. The calculation of the loss function Loss includes: ； Where n represents the number of coordinate data points. Represents the true value of coordinate data. These are the predicted values ​​of the coordinate data. and These represent the weights of MAE and MSE, respectively.

4. The ship hull parametric design method based on lightweight networks as described in claim 3, characterized in that: Including Adam as the optimizer, with an initial learning rate of 2×10. -3 The iterator is a cosine annealing function, and training stops when the loss converges to a specified threshold.

5. The ship hull parametric design method based on lightweight networks as described in claim 4, characterized in that: Evaluation of the predicted control point data includes calculating the mean absolute percentage error between the target value and the predicted value to assess model performance; When the error of each parameter is less than 5%, the design requirements are met and the training is complete. Otherwise, adjust and optimize the network structure and retrain; The parameters include waterline length, drainage volume, immersion area, square coefficient, and rhombus coefficient.

6. The hull form parametric design method based on lightweight networks as described in claim 5, characterized in that: include, The trained PDNN model is deployed in the computer system, including a target ship type feature parameter input module, a control point coordinate prediction module, a ship type surface fitting module, and an IGES file storage module. The target ship type characteristic parameter input module is used by users to input ship type characteristic parameters that meet design requirements; The control point coordinate prediction module is used to predict the coordinates of control points on the hull surface based on the input hull feature parameters; The IGES file storage module is used to store IGES files for ship types that meet design requirements.

7. The ship hull parametric design method based on lightweight networks as described in claim 6, characterized in that: The generation of the ship-shaped surface includes applying a pre-deployed PDNN parametric design system for actual testing, inputting the target ship-shaped feature parameters, and then performing coordinate prediction, surface fitting, and surface file output through the system to finally obtain the generated ship-shaped surface.

Images