CN114818128B - Modeling method and optimizing method for ship body local curved surface optimizing neural network - Google Patents
Modeling method and optimizing method for ship body local curved surface optimizing neural network Download PDFInfo
- Publication number
- CN114818128B CN114818128B CN202210421254.5A CN202210421254A CN114818128B CN 114818128 B CN114818128 B CN 114818128B CN 202210421254 A CN202210421254 A CN 202210421254A CN 114818128 B CN114818128 B CN 114818128B
- Authority
- CN
- China
- Prior art keywords
- neural network
- ship
- data
- curved surface
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 45
- 238000003062 neural network model Methods 0.000 claims abstract description 37
- 238000005457 optimization Methods 0.000 claims abstract description 32
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000005516 engineering process Methods 0.000 claims abstract description 8
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 239000002245 particle Substances 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 5
- 238000013461 design Methods 0.000 abstract description 7
- 230000006870 function Effects 0.000 description 13
- 238000004364 calculation method Methods 0.000 description 8
- 238000004088 simulation Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 230000004044 response Effects 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 239000012530 fluid Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004134 energy conservation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000036962 time dependent Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/25—Design optimisation, verification or simulation using particle-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/08—Fluids
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computer Hardware Design (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Fluid Mechanics (AREA)
- Algebra (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Automation & Control Theory (AREA)
- Aviation & Aerospace Engineering (AREA)
- Computational Mathematics (AREA)
- Feedback Control In General (AREA)
Abstract
A ship body local curved surface optimization neural network modeling method and a ship body local curved surface optimization method relate to the field of ship design. Aiming at the problem that a large amount of initial sample data is required to be input in the existing proxy model, the invention provides the technical scheme that: comprising the following steps: selecting a plurality of control points in a ship body to be optimized; changing the coordinates of the control points to obtain new control points corresponding to a plurality of ship types as sample data; acquiring hydrodynamic characteristic values to construct a BP neural network model; acquiring a ship hydrodynamic characteristic value calculated through a BP neural network model, and taking a difference value of the ship hydrodynamic characteristic value corresponding to the set of control point coordinates acquired by a CFD technology as an error value; establishing a Kriging model, wherein an error value is used as output data; taking the sum of the error value and the output data as judgment data; judging, if the judging data meets the preset requirement, combining the BP neural network model and the Kriging model, taking the judging data as output data, and finishing training. The method is suitable for being applied to the work of hull curved surface design.
Description
Technical Field
Relates to the field of ship design, in particular to the field of ship body appearance design.
Background
The neural network is used as a common means in ship hydrodynamic performance prediction, wherein the BP neural network is more used, but the BP neural network needs more samples to reach a certain precision, and the samples cannot be accurately selected when the samples need to be updated.
The optimization of the ship body curved surface is one of important means for realizing ship energy conservation, the ship body curved surface is mainly used for optimizing the hydrodynamic performance of the ship, and in the process of optimizing the hydrodynamic performance of the ship, the prediction of the ship resistance performance is very important. And the ship resistance is an important factor influencing the design work of the ship, and is directly related to the technical and economic indexes of the ship. The current ship resistance performance prediction is mainly based on a hydrodynamic performance result of a model test, and is converted into a real ship scale according to an empirical formula, and the ship hydrodynamic performance prediction is based on CFD simulation.
With the continuous development of the computer field in recent years, the method of computer-based machine learning has also developed very rapidly. The traditional empirical formula has low calculation accuracy, low workload and huge calculation amount in the optimization field of CFD simulation with high accuracy, and the BP neural network can be adopted to build the proxy model to solve the problem, so that the calculation amount is reduced.
At present, a traditional BP neural network model is commonly used in the field of ship type design to accelerate CFD simulation numerical calculation, and a proxy model established by a traditional BP neural network algorithm is established based on a most basic neural network structure, however, a large amount of initial sample data is often required to be input into the existing proxy model, and 500 groups of initial data samples are reached, so that the prediction accuracy of the proxy model can be ensured, and the construction time of a sample set is long. And when the initial sample data are less, the error obtained in the training process is larger, the training model with the standard reaching the precision cannot be obtained, and the sample needs to be updated, the existing BP neural network can only update the sample by randomly increasing the initial sample data, and the sample capable of improving the forecasting precision cannot be accurately selected, so that the problem of low convergence speed is solved, and the superiority of adopting the BP neural network model for ship optimization is weakened.
Disclosure of Invention
Aiming at the problems that the existing BP neural network can only update samples by randomly increasing initial sample data, and can not accurately select samples capable of improving prediction accuracy, the convergence rate is slow, and the superiority of ship optimization by adopting a BP neural network model is weakened, the invention provides the following technical scheme:
the modeling method of the ship body local curved surface optimization neural network comprises the following steps:
step one: selecting a plurality of control points representing the characteristics of the ship body in the ship body to be optimized;
step two: obtaining at least 30 groups of new control points by changing the coordinates of the control points, wherein each group of control points corresponds to a ship shape and is used as sample data;
step three: obtaining the hydrodynamic characteristic value of each ship type ship by adopting a CFD technology;
step four: constructing a BP neural network model, taking the control point coordinates of each group as input layer neurons, corresponding to the ship hydrodynamic characteristic values obtained in the step three for each group of control points, and taking the control point coordinates of each group and the corresponding ship hydrodynamic characteristic values as training data;
step five: the method comprises the steps of collecting a ship hydrodynamic characteristic value calculated through a BP neural network model as first comparison data, taking a ship hydrodynamic characteristic value corresponding to the set of control point coordinates obtained by adopting a CFD technology as second comparison data, and taking a difference value between the first comparison data and the second comparison data as an error value;
step six: establishing a Kriging model, taking the control point coordinates of each group as input data, and taking the corresponding error value obtained in the fifth step as output data;
step seven: taking the sum of the error value obtained in the fifth step and the data output in the sixth step as judgment data;
step eight: judging whether the judging data meets the preset requirement or not, if not, updating the neurons of the input layer through the data with the largest absolute value in the judging data and a group of corresponding control point coordinates, and returning to the step five; if yes, carrying out a step nine;
step nine: and (3) combining the BP neural network model and the Kriging model to serve as a BPNN-Kriging model, taking the judgment data obtained in the step (seven) as output data of the BPNN-Kriging model, and completing training.
Further, the selection manner of the control point in the first step is as follows: selecting coordinate points uniformly distributed in a part of a ship body to be optimized as control points, wherein the control points comprise three pieces of coordinate information: coordinate information in the ship length direction, coordinate information in the ship width direction, and coordinate information in the ship depth direction.
Further, the number of groups of control points obtained in the second step is 70.
Further, the error function of the BP neural network adopts an MSE function.
Further, the activation function of the BP neural network adopts a sigmoid function.
In the sixth step, the solution mode of the Kriging model is as follows: and solving by adopting a particle swarm algorithm.
A computer storage medium for storing a computer program, wherein when the storage medium is run by a computer, the computer executes the modeling method of the ship body local curved surface optimization neural network.
The method for optimizing the local curved surface of the ship body comprises the following steps:
training: establishing a neural network model;
the training steps are specifically as follows: the ship body appearance optimization neural network modeling method;
optimizing: collecting data of a plurality of control points of the ship body to be optimized as input data of the neural network model, and then obtaining local curved surface optimization parameters of the ship body to be optimized through the neural network model.
And the optimizing step is to use a particle swarm algorithm to carry out optimal value solving on the neural network model established in the training step.
A computer storage medium for storing a computer program, wherein the computer executes the hull form optimization method when the storage medium is run by the computer.
The invention has the advantages that:
the modeling method for the ship body local curved surface optimization neural network overcomes the prejudice of the prior art, in order to save time and prepare work, people start in the aspect of a database of the neural network, the calculation speed is improved or the calculation flow is changed, the calculation speed is improved based on a large number of input samples, under the condition that the number of the samples does not reach the standard, a better training result is not obtained by the improvement of the calculation speed, then because the number of the samples is larger, the obtained training result has higher precision, and people are guided not to consider how to obtain the training result with equivalent precision under the condition that the number of the input samples is less.
The technical scheme provided by the invention breaks the prejudice of the prior art, more than 100 groups are needed for modeling in the prior art from the angle of error, and the technical scheme provided by the invention only needs 80 groups, and the operation error of the neural network is optimized by Kriging based on the modeling condition that the number of samples is reduced by more than 20%, so that a training result equivalent to the optimization result of the prior art is obtained.
The modeling method for the ship body local curved surface optimization neural network can establish a training model only by a small amount of sample data, and correct errors through Kriging, so that the cost and time for constructing the training model are reduced on the premise of being equivalent to the precision of the prior art.
According to the technical scheme provided by the invention, the sum of the error value predicted by the Kriging model and the error value predicted by the BP neural network is used as a final prediction result, so that the accuracy of the result is ensured.
According to the technical scheme provided by the invention, the largest error value in the Kriging model prediction is used as a training sample for adaptively updating the BP neural network, so that the accuracy of a result is ensured.
The method is suitable for being applied to the work of hull curved surface design.
Drawings
Fig. 1 is a flowchart of a modeling method for optimizing a neural network for a local curved surface of a hull according to an embodiment.
Fig. 2 is a flowchart of a particle swarm algorithm according to a sixth embodiment;
fig. 3 is a flowchart of a hull local surface optimization method according to an eighth embodiment.
Detailed Description
In a first embodiment, referring to fig. 1, the present embodiment provides a modeling method for optimizing a neural network for a local curved surface of a hull, based on a small amount of input samples, the method including:
step one: selecting a plurality of control points representing the characteristics of the ship body in the ship body to be optimized;
step two: obtaining at least 30 groups of new control points by changing the coordinates of the control points, wherein each group of control points corresponds to a ship shape and is used as sample data;
step three: obtaining the hydrodynamic characteristic value of each ship type ship by adopting a CFD technology;
step four: constructing a BP neural network model, taking the control point coordinates of each group as input layer neurons, corresponding to the ship hydrodynamic characteristic values obtained in the step three for each group of control points, and taking the control point coordinates of each group and the corresponding ship hydrodynamic characteristic values as training data;
step five: the method comprises the steps of collecting a ship hydrodynamic characteristic value calculated through a BP neural network model as first comparison data, taking a ship hydrodynamic characteristic value corresponding to the set of control point coordinates obtained by adopting a CFD technology as second comparison data, and taking a difference value between the first comparison data and the second comparison data as an error value;
step six: establishing a Kriging model, taking the control point coordinates of each group as input data, and taking the corresponding error value obtained in the fifth step as output data;
step seven: taking the sum of the error value obtained in the fifth step and the data output in the sixth step as judgment data;
step eight: judging whether the judging data meets the preset requirement or not, if not, updating the neurons of the input layer through the data with the largest absolute value in the judging data and a group of corresponding control point coordinates, and returning to the step five; if yes, carrying out a step nine;
step nine: and (3) combining the BP neural network model and the Kriging model to serve as a BPNN-Kriging model, taking the judgment data obtained in the step (seven) as output data of the BPNN-Kriging model, and completing training.
Specific:
1. and obtaining a ship type database by a ship surface deformation method.
1.1: and extracting characteristic variables from the mother ship type ship body. The whole ship shape is represented by uniformly distributed coordinate points serving as control points, and the control point set is recorded as M= (M 1 ,m 2 ,m 3 ,...,m n ) Wherein each element in M contains three pieces of position information: m is m i ＝(x i ,y i ,z i ) Wherein x is i Representing the coordinate of the ith point along the ship length direction, y i Representing the coordinate of the ith point in the ship width direction, where z i Representing the coordinates of the ith point in the depth direction.
The ship profile can be represented by the coordinates of the control points, and the curved surface of the ship body is obtained.
1.2: the control point is varied within a certain range. For each control point M in the set M i The coordinates of the three representing directions are changed within a certain range, and the change range is determined according to the ship shape to be optimized, so that the ship shape to be optimized is reasonably optimized.
1.3: after the characteristic variables are changed, a hull curved surface deformation method is adopted to obtain a hull curved surface after the corresponding characteristic variables are changed, so that an initial ship type database is obtained.
2. An initial sample library is extracted from the ship type database, and a hydrodynamic characteristic response value is calculated.
2.1: and sampling a certain number of ship samples from the ship database by adopting a sampling method. And Latin hypercube sampling is adopted for the initial ship type database, so that a corresponding number of ship type samples are obtained.
2.2: the hydrodynamic characteristics corresponding to each sample are calculated by adopting a CFD (CFD, english full name (Computational Fluid Dynamics), namely computational fluid dynamics) numerical simulation method for the extracted ship-shaped samples.
2.3: and taking the characteristic variable of the extracted ship-shaped sample as an initial sample library and taking the corresponding hydrodynamic characteristic value as a response value. In order to construct the proxy model, hydrodynamic characteristic values corresponding to the extracted ship-shaped samples are taken as response values.
3. And establishing a BP neural network model for the initial sample library.
3.1: and taking the characteristic variable of the ship-shaped sample of the initial sample library as input, and taking the corresponding hydrodynamic characteristic as output to construct the BP neural network model. And taking the coordinates of the extracted ship-shaped sample as input and the hydrodynamic characteristic value as output to construct the BP neural network model.
3.2: setting the hidden layer number, the neuron number, the activation function, the loss function, the learning rate and the like of the neural network, and obtaining the trained BP neural network model after a certain number of iterations. The activation function adopts a sigmoid function, the loss function adopts a mean square error MSE function, and the learning rate can be set to be 0.01-0.1. The initial weight and bias are set to random numbers of [0,1], and the following are specific:
the sigmoid function is specifically:
the MSE function is specifically:
wherein,,
sigma represents the input x on the neuron ij Output value of time-dependent sigmoid function, x ij Input representing the jth neuron of the ith hidden layer, m represents the total number of samples, y t Representing the true output of the t-th set of samples,representing the predicted output of the t-th set of samples.
4. And constructing a Kriging model for the prediction error of the BP neural network. And combining the BP neural network model and the Kriging model to form a BPNN-Kriging model.
4.1: and calculating a predicted hydrodynamic characteristic value of the ship-shaped sample through the BP neural network model.
4.2: and taking the difference between the predicted hydrodynamic characteristic value and the hydrodynamic characteristic value calculated by the CFD numerical simulation method as an error value.
4.3: and establishing a Kriging model for the characteristic variable and the corresponding error value of the ship-shaped sample. And taking the coordinates of the extracted ship-shaped sample as input and the error value as output, and establishing a Kriging model.
4.4: combining the two models into a BPNN-Kriging model, and adding the predicted value of the BP neural network model and the predicted error value of the Kriging model as the predicted result.
5. And judging whether the prediction error value of the Kriging model meets the precision requirement, if so, turning to the next step, and if not, updating the initial sample library through the maximum value of the absolute value of the prediction error value, and repeating the steps 3-5.
5.1: and searching the maximum value of the absolute value of the error by adopting a particle swarm optimization algorithm for the Kriging model, judging whether the maximum value of the absolute value of the error meets the precision requirement, and if so, completing construction.
5.2: if the requirements are not met, adding the ship shape corresponding to the maximum value of the absolute value of the error into the initial sample library to update the sample library, and repeating the steps 3-5.
In the second embodiment, the modeling method for optimizing the neural network for the local curved surface of the hull according to the first embodiment is further defined, and the selection mode of the control points in the first step is as follows: selecting coordinate points uniformly distributed in a part of a ship body to be optimized as control points, wherein the control points comprise three pieces of coordinate information: coordinate information in the ship length direction, coordinate information in the ship width direction, and coordinate information in the ship depth direction.
In the third embodiment, the method for modeling the hull local curved surface optimization neural network according to the first embodiment is further limited, and the number of groups of control points obtained in the second step is 70.
In the fourth embodiment, the modeling method for the hull local curved surface optimization neural network provided in the first embodiment is further limited, and the error function of the BP neural network adopts an MSE function.
In a fifth embodiment, the method for modeling a hull local curved surface optimization neural network according to the first embodiment is further defined, and the activation function of the BP neural network is a sigmoid function.
In a sixth embodiment, a description is given of the present embodiment with reference to fig. 2, where the method for modeling a hull local curved surface optimization neural network provided in the first embodiment is further defined, and in the sixth step, a solution mode of the Kriging model is as follows: and solving by adopting a particle swarm algorithm.
An embodiment seven provides a computer storage medium for storing a computer program, where when the storage medium is run by a computer, the computer executes the modeling method for optimizing the local curved surface of the hull provided in any one of the embodiments one to six.
An eighth embodiment is described with reference to fig. 3, and the present embodiment provides a method for optimizing a local curved surface of a hull, where the method includes:
training: establishing a neural network model;
the training steps are specifically as follows: the ship body appearance optimization neural network modeling method;
optimizing: collecting data of a plurality of control points of the ship body to be optimized as input data of the neural network model, and then obtaining local curved surface optimization parameters of the ship body to be optimized through the neural network model.
In the ninth embodiment, the method for optimizing a local curved surface of a hull according to the eighth embodiment is further defined, and the optimizing step is to solve an optimal value of the neural network model established in the training step by using a particle swarm algorithm.
In a tenth aspect, the present embodiment provides a computer storage medium storing a computer program, where when the storage medium is run by a computer, the computer executes the hull form optimizing method provided in the eighth aspect.
An eleventh embodiment is an experimental process of data and records obtained after the optimization of an actual ship shape by using the method for optimizing a local curved surface of a ship body provided in the eighth embodiment, and specifically:
the parameters of the ship shape to be optimized selected in this embodiment are as follows:
ship type parameters to be optimized:
the existing BP neural network is adopted for optimization, 110 groups of samples are adopted through testing, the optimized resistance value is 269.55KN, and compared with the original ship type resistance, the resistance is reduced by 4.34%.
By adopting the scheme, 70 groups of samples are initially input, and after 11 times of sample updating, 81 groups of samples are finally obtained, and the optimized resistance value is 269.92KN. Compared with the original ship type resistance, the resistance is reduced by 4.21 percent.
According to the data, compared with the existing BP neural network technology, the method and the device have the advantages that the similar optimization effect can be obtained under the condition that the input sample number is reduced by 26.36%. The scheme can reduce the input quantity of samples while guaranteeing the precision.
Claims (10)
1. The modeling method of the ship body local curved surface optimization neural network is characterized by comprising the following steps:
step one: selecting a plurality of control points representing the characteristics of the ship body in the ship body to be optimized;
step two: obtaining at least 30 groups of new control points by changing the coordinates of the control points, wherein each group of control points corresponds to a ship shape and is used as sample data;
step three: obtaining the hydrodynamic characteristic value of each ship type ship by adopting a CFD technology;
step four: constructing a BP neural network model, taking the control point coordinates of each group as input layer neurons, corresponding to the ship hydrodynamic characteristic values obtained in the step three for each group of control points, and taking the control point coordinates of each group and the corresponding ship hydrodynamic characteristic values as training data;
step five: the method comprises the steps of collecting a ship hydrodynamic characteristic value calculated through a BP neural network model as first comparison data, taking a ship hydrodynamic characteristic value corresponding to the set of control point coordinates obtained by adopting a CFD technology as second comparison data, and taking a difference value between the first comparison data and the second comparison data as an error value;
step six: establishing a Kriging model, taking the control point coordinates of each group as input data, and taking the corresponding error value obtained in the fifth step as output data;
step seven: taking the error value output by the Kriging model as judging data;
step eight: judging whether the judging data meets the preset requirement or not, if not, updating the neurons of the input layer through the data with the largest absolute value in the judging data and a group of corresponding control point coordinates, returning to the fourth step, and reconstructing the BP neural network model; if yes, carrying out a step nine;
step nine: and combining the BP neural network model and the Kriging model, and taking the BP neural network model and the Kriging model as the BPNN-Kriging model, wherein the output data of the BPNN-Kriging model is obtained by adding the predicted value of the BP neural network model and the predicted error value of the Kriging model, and training is completed.
2. The modeling method of a hull local curved surface optimizing neural network according to claim 1, wherein the selection mode of the control points in the first step is as follows: selecting coordinate points uniformly distributed in a part of a ship body to be optimized as control points, wherein the control points comprise three pieces of coordinate information: coordinate information in the ship length direction, coordinate information in the ship width direction, and coordinate information in the ship depth direction.
3. The modeling method of the hull local curved surface optimization neural network according to claim 1, wherein the number of the groups of the control points obtained in the second step is 70.
4. The modeling method of the hull local curved surface optimization neural network according to claim 1, wherein an error function of the BP neural network adopts an MSE function.
5. The modeling method for the hull local curved surface optimization neural network according to claim 1, wherein the activation function of the BP neural network adopts a sigmoid function.
6. The modeling method of the hull local curved surface optimization neural network according to claim 1, wherein in the sixth step, the solution mode of the Kriging model is as follows: and solving by adopting a particle swarm algorithm.
7. A computer storage medium for storing a computer program, wherein the computer performs the method for modeling a local curved surface of a hull according to any of claims 1-6 when the storage medium is run by a computer.
8. The method for optimizing the local curved surface of the ship body is characterized by comprising the following steps of:
training: establishing a neural network model;
the training steps are specifically as follows: the hull local surface optimization neural network modeling method of any one of claims 1-6;
optimizing: collecting data of a plurality of control points of the ship body to be optimized as input data of the neural network model, and then obtaining local curved surface optimization parameters of the ship body to be optimized through the neural network model.
9. The method of optimizing a local curved surface of a hull according to claim 8, wherein the optimizing step is to solve an optimal value of the neural network model established in the training step by using a particle swarm algorithm.
10. A computer storage medium for storing a computer program, wherein the computer performs the method of optimizing a local curvature of a hull according to any of claims 8-9 when the storage medium is run by a computer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210421254.5A CN114818128B (en) | 2022-04-21 | 2022-04-21 | Modeling method and optimizing method for ship body local curved surface optimizing neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210421254.5A CN114818128B (en) | 2022-04-21 | 2022-04-21 | Modeling method and optimizing method for ship body local curved surface optimizing neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114818128A CN114818128A (en) | 2022-07-29 |
| CN114818128B true CN114818128B (en) | 2023-09-29 |
Family
ID=82505085
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210421254.5A Active CN114818128B (en) | 2022-04-21 | 2022-04-21 | Modeling method and optimizing method for ship body local curved surface optimizing neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114818128B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117150652B (en) * | 2023-08-30 | 2024-04-30 | 武汉理工大学 | Sample point selection method, system and terminal based on ship-shaped constraint space |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100004150A (en) * | 2008-07-03 | 2010-01-13 | 현대중공업 주식회사 | A methodology of hull-form generation by mathematical definition using analytic equations on waterlines of ship |
| CN111506968A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship type optimization method based on BP neural network algorithm |
| CN111506970A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship hydrodynamic performance evaluation method |
| CN111506969A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship type optimization method based on multi-target particle swarm algorithm |
| CN111619755A (en) * | 2020-06-09 | 2020-09-04 | 中国船舶科学研究中心 | Hull profile design method based on convolutional neural network |
| CN111693084A (en) * | 2020-06-23 | 2020-09-22 | 南京航空航天大学 | Measurement error compensation method based on error similarity |
-
2022
- 2022-04-21 CN CN202210421254.5A patent/CN114818128B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20100004150A (en) * | 2008-07-03 | 2010-01-13 | 현대중공업 주식회사 | A methodology of hull-form generation by mathematical definition using analytic equations on waterlines of ship |
| CN111506968A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship type optimization method based on BP neural network algorithm |
| CN111506970A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship hydrodynamic performance evaluation method |
| CN111506969A (en) * | 2020-04-21 | 2020-08-07 | 常熟理工学院 | Ship type optimization method based on multi-target particle swarm algorithm |
| CN111619755A (en) * | 2020-06-09 | 2020-09-04 | 中国船舶科学研究中心 | Hull profile design method based on convolutional neural network |
| CN111693084A (en) * | 2020-06-23 | 2020-09-22 | 南京航空航天大学 | Measurement error compensation method based on error similarity |
Non-Patent Citations (1)
| Title |
|---|
| 基于水动力性能优化的船型设计研究进展;万德成 等;水动力学研究与进展;第34卷(第6期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114818128A (en) | 2022-07-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111027772B (en) | Multi-factor short-term load prediction method based on PCA-DBILSTM | |
| CN110046710A (en) | A kind of the nonlinear function Extremal optimization method and system of neural network | |
| CN109614631B (en) | Aircraft full-automatic pneumatic optimization method based on reinforcement learning and transfer learning | |
| CN112947300A (en) | Virtual measuring method, system, medium and equipment for processing quality | |
| CN110751318A (en) | IPSO-LSTM-based ultra-short-term power load prediction method | |
| CN110197307B (en) | Regional sea surface temperature prediction method combined with attention mechanism | |
| CN114065662B (en) | Method suitable for rapidly predicting airfoil flow field with variable grid topology | |
| CN114912195B (en) | Aerodynamic sequence optimization method for commercial vehicle | |
| CN111832839B (en) | Energy consumption prediction method based on sufficient incremental learning | |
| CN116341097B (en) | Transonic wing optimal design method based on novel high-dimensional proxy model | |
| CN114818128B (en) | Modeling method and optimizing method for ship body local curved surface optimizing neural network | |
| CN111191785A (en) | Structure searching method based on expanded search space | |
| CN114548591A (en) | Time sequence data prediction method and system based on hybrid deep learning model and Stacking | |
| CN104732067A (en) | Industrial process modeling forecasting method oriented at flow object | |
| CN115688276A (en) | Aircraft appearance automatic optimization method, system, equipment and medium based on discrete companion method | |
| CN116757057A (en) | Air quality prediction method based on PSO-GA-LSTM model | |
| CN115099461A (en) | Solar radiation prediction method and system based on double-branch feature extraction | |
| CN114564787A (en) | Bayesian optimization method, device and storage medium for target-related airfoil design | |
| CN112800690B (en) | Underwater folding and unfolding mechanism parameter optimization method based on group intelligent optimization algorithm | |
| CN114330119A (en) | Deep learning-based pumped storage unit adjusting system identification method | |
| CN113868765A (en) | Ship main scale parameter optimization method based on approximate model | |
| CN114936413B (en) | Ship body appearance optimization neural network modeling method and ship body appearance optimization method | |
| CN116628854A (en) | Wing section aerodynamic characteristic prediction method, system, electronic equipment and storage medium | |
| CN108197368B (en) | Method for simply and conveniently calculating geometric constraint and weight function of complex aerodynamic shape of aircraft | |
| CN116894504A (en) | Wind power cluster power ultra-short-term prediction model establishment method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |