Design method of special-shaped underwater pressure-resistant shell

Abstract

The invention discloses a design method of a special-shaped underwater pressure-resistant shell, which comprises the following steps: S1, according to the use environment of a to-be-designed special-shaped underwater pressure-resistant shell, determining various types of structure parameters which are required by the to-be-designed special-shaped underwater pressure-resistant shell and have association relationship and the initial value range of the structure parameters; s2, collecting a plurality of groups of data samples of the structure parameters conforming to the incidence relation in an initial value range; s3, m evaluation indexes used for evaluating the performance of the to-be-designed special-shaped underwater pressure-resistant shell are determined according to the use environment of the to-be-designed special-shaped underwater pressure-resistant shell, original evaluation values corresponding to the m evaluation indexes of each group of data samples are obtained, and the multiple data samples and the corresponding m original evaluation values form a data set; and S4, training an artificial neural network model by using the data set to obtain an initial prediction model for predicting an original evaluation value of the to-be-designed special-shaped underwater pressure-resistant shell. According to the method, multi-dimensional design of the special-shaped underwater pressure-resistant shell can be achieved, the multi-performance requirement is met, the design process is simple, and the design efficiency is high.

Description

Technical Field

[0001] The present invention relates to a design method, and in particular to a design method for a special-shaped underwater pressure hull. Background Art

[0002] An underwater pressure hull is a structure that can withstand external water pressure and plays a key role in protecting internal equipment and personnel, ensuring the realization of the functions of underwater equipment, and improving its performance and efficiency. It is widely used in fields such as submarines, underwater detectors, and underwater production lines. A special-shaped underwater pressure hull is a pressure hull designed based on the principle of bionics and can better meet actual needs.

[0003] However, due to the irregular shape of the special-shaped pressure hull, it is difficult to express its shape structure formulaically and difficult to solve some parameters during design, which increases the difficulty of the optimal design of the special-shaped underwater pressure hull. In the prior art, the equivalent spherical shell theory method for the special-shaped underwater pressure hull stays at the spherical-like structure, and the formula derivation is complex; and only research and design are carried out in terms of mechanical properties, while other aspects are mostly ignored. Therefore, there is an urgent need to study a multi-disciplinary and multi-dimensional optimal design method for it, so that the pressure hull takes into account various performances and simplifies the design process. Summary of the Invention

[0004] Object of the Invention: The object of the present invention is to provide a design method for a special-shaped underwater pressure hull that can achieve multi-dimensional design, take into account various performances, and has a simple design process.

[0005] Technical Solution: A design method for a special-shaped underwater pressure hull disclosed by the present invention includes the following steps

[0006] S1: Determine various types of structure parameters and the initial value ranges of the structure parameters required and having an associated relationship according to the use environment of the to-be-designed special-shaped underwater pressure hull;

[0007] S2: Collect multiple groups of data samples of structure parameters that meet the associated relationship within the initial value range;

[0008] S3: Determine m evaluation indexes for evaluating its performance according to the use environment of the to-be-designed special-shaped underwater pressure hull, and obtain the original evaluation values corresponding to the m evaluation indexes of each group of data samples, and multiple data samples and their corresponding m original evaluation values form a data set;

[0009] S4: Train an artificial neural network model with the data set to obtain an initial prediction model for predicting the original evaluation values of the to-be-designed special-shaped underwater pressure hull;

[0010] S5: Optimize the initial prediction model by using a particle swarm optimization algorithm to obtain a final target prediction model;

[0011] S6: Extract n groups of data samples from the multiple groups of data samples in step S2, and use the genetic algorithm to optimize the obtained n groups of data samples to obtain the optimized n groups of data samples;

[0012] S7: Input the optimized n groups of data samples into the target prediction model to obtain m prediction evaluation values corresponding to the optimized n groups of data samples, and calculate the weight values of m evaluation indicators based on the m prediction evaluation values corresponding to the optimized n groups of data samples;

[0013] S8: Use the ideal distance method and combine the weight values to select a set of optimal data samples from the optimized n groups of data samples for the actual design of the special-shaped underwater pressure hull.

[0014] Further, the multiple types of structural parameters described in step S1 include the major axis length L, the minor axis length B, and the thickness t; the correlation relationship between the multiple types of structural parameters is t = F1(L, B).

[0015] Further, the value range of the initial values in step S1 is as follows: β1 ≤ L ≤ β2; β3 ≤ B ≤ β4; the aspect ratio Φ = B / t, and β5 ≤ Φ ≤ β6.

[0016] Further, the method for collecting data samples of multiple groups of structural parameters that meet the correlation relationship in step S2 is as follows:

[0017] Establish a two-dimensional plane with L and B as variables;

[0018] Under the conditions of β1 ≤ L ≤ β2 and β3 ≤ B ≤ β4, randomly collect multiple groups of corresponding L-B data in the two-dimensional plane;

[0019] Calculate t corresponding to the L-B data one by one based on t = F1(L, B) to obtain L-B-t data;

[0020] Judge whether the Φ of the obtained L-B-t data satisfies β5 ≤ Φ ≤ β6; if it satisfies, the L-B-t constitutes a data sample; if it does not satisfy, randomly extract Φ within the range of β5 ≤ Φ ≤ β6, calculate t' based on Φ = B / t, replace t with t', and the L-B-t constitutes a data sample.

[0021] Further, the m evaluation indicators in step S3 include the maximum stress for evaluating the load-bearing capacity of the special-shaped underwater shell to be designed, the buoyancy coefficient for evaluating the buoyancy performance of the special-shaped underwater shell to be designed The buckling eigenvalue for evaluating the stability of the special-shaped underwater shell to be designed and the intercepted area S = F(B) for evaluating the hydrodynamic performance of the special-shaped underwater shell to be designed; where ρ mat refers to the material density of the special-shaped underwater shell to be designed, ρ refers to the seawater density, V matRefers to the material volume, V dis Refers to the maximum drainage volume; and V mat = S es ·t, S es Refers to the mid-surface area of the special-shaped underwater shell to be designed, S es = F2(L, B).

[0022] Furthermore, the method for obtaining the data set in step S3 is as follows:

[0023] Based on the shape parameter equation of the special-shaped underwater shell to be designed and all the data samples in step S, establish a 3D model of the special-shaped underwater shell to be designed in 3D modeling software;

[0024] Obtain the L-B-t data and V corresponding to each group of data samples in the 3D model dis , and based on Calculate the original evaluation value P1 of the buoyancy coefficient δ corresponding to L-B-t;

[0025] Import the 3D simulation model into the simulation software to obtain the simulation model of the special-shaped underwater shell to be designed;

[0026] Obtain the original evaluation value P2 of the maximum stress and the original evaluation value P3 of the buckling eigenvalue of L-B-t corresponding to the data samples through the simulation model;

[0027] Based on S = F(B), calculate the original evaluation value P4 of the intercepted area S of L-B-t corresponding to the data samples;

[0028] The data set consists of multiple groups of one-to-one corresponding L-B-t and P1-P2-P3-P4.

[0029] Furthermore, in step S6, in addition to satisfying the initial value range and correlation relationship in step S1, the n groups of data samples extracted also need to satisfy the following conditions: β7 ≤ B / L ≤ β8.

[0030] Furthermore, the method for obtaining the optimized n groups of data samples in step S6 is as follows:

[0031] Define the n groups of data samples as the initial population;

[0032] Perform non-dominated sorting on the individuals in the initial population;

[0033] Based on the non-dominated sorting result of the initial population, select the top a individuals in the initial population as excellent individuals. Perform crossover on L and B of any two excellent individuals, and determine whether L, B, and t after crossover simultaneously satisfy the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8; if satisfied, retain the two new offspring individuals generated by this crossover; if not satisfied, do not retain the two new offspring individuals generated by this crossover, and perform crossover again until L, B, and t after crossover simultaneously satisfy the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8;

[0034] After reaching the preset number of crossovers, stop the crossover, and the individuals in the birth population and the retained offspring individuals form a new population;

[0035] Set the mutation probability, perform mutation on L and B of all individuals in the new population, and determine whether L, B, and t generated after individual mutation simultaneously satisfy the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8; if satisfied, retain the new offspring individuals generated by this mutation; if not satisfied, do not retain the new offspring individuals generated by this mutation, and perform mutation on L and B of this individual again until L, B, and t after mutation simultaneously satisfy the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8;

[0036] All the new offspring individuals after the new population is mutated form the target population, and perform crowding degree sorting on the individuals in the target population;

[0037] Select the top n groups of individuals in the crowding degree sorting as the optimized n groups of data samples.

[0038] Furthermore, the method for calculating the weight values of m evaluation indicators in step S7 is as follows:

[0039] Set the first weight of m evaluation indicators as

[0040] Use the C-OWA operator to objectively evaluate the m predicted evaluation values corresponding to n groups of data samples, and obtain the second weight of the set m evaluation indicators as

[0041] Construct a combined weight w based on the first weight and the second weight *g , where k1 is the coefficient of the first weight, k2 is the coefficient of the second weight, and g ∈ [1, m];

[0042] Calculate w based on game theory * and w sub to have the smallest dispersion degree and w * and w obj to have the smallest dispersion degree of the combined coefficients k1 and k2;

[0043] The weight values of m evaluation indicators are w* = [w *1 : w *2 : …: w *m .

[0044] Furthermore, the method for obtaining the optimal data samples in step S8 is as follows:

[0045] Construct a weighted normalized decision matrix based on the weight values of m evaluation indicators

[0046]

[0047] where P n,m represents the predicted evaluation value of the m-th evaluation indicator of the n-th data sample among the optimized n groups of data samples;

[0048] Based on the ideal distance method, let the positive ideal value of the g-th column of be the negative ideal value

[0049]

[0050] Calculate the Euclidean distance between each data sample in the optimized n groups of data samples and the positive ideal value the Euclidean distance between each data sample in the optimized n groups of data samples and the negative ideal value

[0051]

[0052] Calculate the proximity index, and the proximity index of the n Popu -th group of solutions is expressed as:

[0053]

[0054] Select the group of data samples with the largest proximity index among the optimized n groups of data samples for the actual design of the special-shaped underwater pressure hull.

[0055] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: By introducing a variety of structural parameters and a variety of performance indicators, and establishing a mapping relationship between the various structural parameters and the various performance indicators, the multi-dimensional design of the special-shaped underwater pressure hull can be realized, while taking into account the multi-performance requirements of the special-shaped underwater pressure hull. By means of three-dimensional modeling and simulation modeling, the present invention can overcome the problem of insufficient structural parameter data samples, and can assist in the performance of the special-shaped underwater pressure hull designed according to the present invention; and the present invention optimizes the prediction model and data samples successively, which can further improve the performance of the special-shaped underwater pressure hull designed according to the present invention. The present invention simplifies the design process through the artificial neural network model, three-dimensional modeling and simulation modeling, which is beneficial to improving the overall design efficiency. Description of the Drawings

[0056] Figure 1 is a flowchart of the present invention;

[0057] Figure 2 is a schematic structural diagram of the initial prediction model of the embodiment of the present invention;

[0058] Figure 3 is a scatter diagram of the data set of the embodiment of the present invention;

[0059] Figure 4 is an iteration diagram of the particle swarm optimization algorithm of the embodiment of the present invention for optimizing the initial prediction model;

[0060] Figure 5 is a distribution diagram of the weight values of the embodiment of the present invention;

[0061] Figure 6 is a comparison diagram of the predicted evaluation values and the original evaluation values of the usage scenarios 1-4 of the embodiment of the present invention. Detailed Embodiments

[0062] The technical solution of the present invention will be further described below with reference to the drawings.

[0063] Example 1

[0064] A design method for a special-shaped underwater pressure hull disclosed by the present invention, as Figure 1 shown, includes the following steps:

[0065] S1: Determine various types of structural parameters and the initial value ranges of the structural parameters required and having an associated relationship according to the usage environment of the special-shaped underwater pressure hull to be designed.

[0066] Among them, various types of structural parameters include the major axis length L, the minor axis length B, and the thickness t; preferably, the types of structural parameters are not limited to the major axis length, the minor axis length, and the thickness. In actual use, the types of structural parameters required can be increased or decreased according to actual needs. The correlation relationship between various types of structural parameters is t = F1(L, B), that is, there is a constraint relationship between the thickness t, the major axis length L, and the minor axis length B.

[0067] The initial value ranges are as follows: β1 ≤ L ≤ β2; β3 ≤ B ≤ β4; the aspect ratio Φ = B / t, and β5 ≤ Φ ≤ β6. β1, β2, β3, β4, β5, and β6 are all non-negative real numbers, and β1, β2, β3, β4, β5, and β6 can be set according to actual needs.

[0068] S2: Collect multiple groups of data samples of structural parameters that meet the correlation relationship within the initial value range.

[0069] Establish a two-dimensional plane with L and B as variables.

[0070] Under the conditions of β1 ≤ L ≤ β2 and β3 ≤ B ≤ β4, randomly collect multiple groups of corresponding L-B data in the two-dimensional plane; preferably, the random sampling method is preferably Latin hypercube sampling.

[0071] Calculate t corresponding to the L-B data one by one based on t = F1(L, B) to obtain L-B-t data.

[0072] Judge whether the Φ of the obtained L-B-t data satisfies β5 ≤ Φ ≤ β6; if it satisfies, then L-B-t constitutes a data sample; if it does not satisfy, randomly select Φ within the range of β5 ≤ Φ ≤ β6, calculate t' based on Φ = B / t, replace t with t', and L-B-t constitutes a data sample.

[0073] S3: Determine m evaluation indexes for evaluating the performance of the to-be-designed special-shaped underwater pressure hull according to its usage environment, obtain the original evaluation values corresponding to the m evaluation indexes of each group of data samples, and multiple data samples and their corresponding m original evaluation values constitute a data set.

[0074] Among them, the m evaluation indexes include the maximum stress for evaluating the load-bearing capacity of the to-be-designed special-shaped underwater hull, the buoyancy coefficient for evaluating the buoyancy performance of the to-be-designed special-shaped underwater hull The buckling eigenvalue for evaluating the stability of the to-be-designed special-shaped underwater hull and the interception area S = F(B) for evaluating the hydrodynamic performance of the to-be-designed special-shaped underwater hull; where ρ mat refers to the material density of the to-be-designed special-shaped underwater hull, ρ refers to the seawater density, V mat refers to the material volume, V dis refers to the maximum drainage volume; and V mat = Ses ·t, S es Refers to the mid-surface area of the special-shaped underwater shell to be designed, S es = F2(L, B). Preferably, the types of evaluation indicators can be increased or decreased according to requirements during actual use.

[0075] The method for obtaining the data set is as follows:

[0076] Based on the shape parameter equation of the special-shaped underwater shell to be designed and all the data samples in step S, establish a 3D model of the special-shaped underwater shell to be designed in 3D modeling software;

[0077] Obtain the L-B-t data and V corresponding to each group of data samples in the 3D model dis , and based on Calculate the original evaluation value P1 of the buoyancy coefficient δ corresponding to L-B-t;

[0078] Import the 3D simulation model into the simulation software to obtain the simulation model of the special-shaped underwater shell to be designed;

[0079] Obtain the original evaluation value P2 of the maximum stress and the original evaluation value P3 of the buckling eigenvalue of L-B-t corresponding to the data samples through the simulation model;

[0080] Based on S = F(B), calculate the original evaluation value P4 of the intercepted area S2 of L-B-t corresponding to the data samples;

[0081] The data set consists of multiple groups of one-to-one corresponding L-B-t and P1-P2-P3-P4.

[0082] S4: Use the data set to train an artificial neural network model to obtain an initial prediction model for predicting the original evaluation value of the special-shaped underwater pressure-resistant shell to be designed. When training the artificial neural network model, L-B-t is used as the input data, and P1-P2-P3-P4 corresponding to L-B-t is used as the output data.

[0083] Preferably, normalize the data set before using it to train the artificial neural network model.

[0084] Preferably, divide the data set into a training set and a test set according to a preset ratio, use the training set to train the artificial neural network model, and use the test set to test and optimize the initial prediction model obtained after training, which is beneficial to improving the accuracy and stability of the obtained initial prediction model.

[0085] S5: Use the particle swarm optimization algorithm to optimize the initial prediction model to obtain the final target prediction model.

[0086] Such as Figure 2As shown, it is a schematic structural diagram of the initial prediction model. The steps to optimize the initial prediction model using the particle swarm optimization algorithm are as follows:

[0087]

[0088] Among them, is the weight parameter between the i-th node of the input layer and the j-th node of the hidden layer, is the weight parameter between the j-th node of the hidden layer and the k-th node of the output layer, H b j is the constant term parameter of the hidden layer, H b k is the constant term parameter of the output layer. These parameters are used as the positions of the particles in the particle swarm optimization algorithm p x i ; E is H b j , H b k the average error of each dimension of each data sample in the training set by the artificial neural network model. The optimization objective of the particle swarm optimization algorithm is when this average error is minimized. N Sam is the number of samples, and n represents the n-th group of data samples.

[0089] Preferably, for the trained initial prediction model, it is necessary to use the test set data for prediction verification. If the average error of a certain sample in the test set is greater than 5%, retraining is required; if the error is greater than 5% near a certain data sample after multiple trainings, first return to step S2 to supplement L, B, or t with a difference of 10% from the original value through random sampling, and then retrain.

[0090] S6: Extract n groups of data samples from the multiple groups of data samples in step S2, and use the genetic algorithm to optimize the obtained n groups of data samples to obtain the optimized n groups of data samples.

[0091] In addition to meeting the initial value range and correlation relationship in step S1, the extracted n groups of data samples also need to meet the following conditions: β7 ≤ B / L ≤ β8. β7 and β8 are non-negative real numbers.

[0092] The method to obtain the optimized n groups of data samples is as follows:

[0093] Define the n groups of data samples as the initial population; each group of data samples is an individual of the initial population;

[0094] Perform non-dominated sorting on the individuals in the initial population;

[0095] Based on the non-dominated sorting results of the initial population, select the first a individuals in the initial population as excellent individuals. Perform crossover on L and B of any two excellent individuals, and determine whether L, B, and t after crossover simultaneously satisfy the initial value range, the correlation relationship, and β7 ≤ B / L ≤ β8; if satisfied, retain the two new offspring individuals generated by this crossover; if not satisfied, do not retain the two new offspring individuals generated by this crossover, and perform crossover again until L, B, and t after crossover simultaneously satisfy the initial value range, the correlation relationship, and β7 ≤ B / L ≤ β8;

[0096] After reaching the preset number of crossover times, stop crossover, and the individuals in the birth population and the retained offspring individuals form a new population;

[0097] Set the mutation probability, perform mutation on L and B of all individuals in the new population, and determine whether L, B, and t generated after individual mutation simultaneously satisfy the initial value range, the correlation relationship, and β7 ≤ B / L ≤ β8; if satisfied, retain the new offspring individual generated by this mutation; if not satisfied, do not retain the new offspring individual generated by this mutation, and perform mutation on L and B of this individual again until L, B, and t after mutation simultaneously satisfy the initial value range, the correlation relationship, and β7 ≤ B / L ≤ β8;

[0098] All the new offspring individuals after the new population is mutated form the target population, and perform crowding degree sorting on the individuals in the target population;

[0099] Select the first n groups of individuals in the crowding degree sorting as the optimized n groups of data samples.

[0100] S7: Input the optimized n groups of data samples into the target prediction model, obtain m prediction evaluation values corresponding to the optimized n groups of data samples, and calculate the weight values of m evaluation indicators based on the m prediction evaluation values corresponding to the optimized n groups of data samples.

[0101] Set the first weight of m evaluation indicators as

[0102] Use the C-OWA operator to objectively evaluate the m prediction evaluation values corresponding to n groups of data samples, and obtain the second weight of the set m evaluation indicators as

[0103] Construct the combined weight w based on the first weight and the second weight *j , where k1 is the coefficient of the first weight, k2 is the coefficient of the second weight, and j ∈ [1, m];

[0104] Calculate w based on game theory * and w sub such that the dispersion degree between w * and w is the smallest and wobj The combination coefficients k1 and k2 with the smallest degree of dispersion;

[0105] The weight values of m evaluation indicators are w* = [w *1 :w *2 :…:w *m .

[0106] S8: Use the ideal distance method and combine the weight values to select a set of optimal data samples from the optimized n groups of data samples for the actual design of the special-shaped underwater pressure hull.

[0107] Construct a weighted normalized decision matrix based on the weight values of m evaluation indicators

[0108]

[0109] where P n,m represents the predicted evaluation value of the m-th evaluation indicator of the n-th data sample in the optimized n groups of data samples;

[0110] Based on the ideal distance method, let The positive ideal value of the g-th column of The negative ideal value is

[0111]

[0112] Calculate the Euclidean distance between each data sample in the optimized n groups of data samples and the positive ideal value The Euclidean distance between each data sample in the optimized n groups of data samples and the negative ideal value

[0113]

[0114] Calculate the proximity index, and the proximity index of the n Popu th group of solutions is expressed as:

[0115]

[0116] Select a group of data samples with the largest proximity index from the optimized n groups of data samples for the actual design of the special-shaped underwater pressure hull.

[0117] Example 2

[0118] An example of using the design method of a special-shaped underwater pressure hull disclosed in the present invention to design an egg-shaped underwater pressure hull.

[0119] S1: Set the structural parameters as the major axis length L, the minor axis length B, and the thickness t.

[0120] The mid - area of the egg - shaped underwater pressure hull is S es = π·(L·B 2 ) 2 / 3 , and the material volume of the egg - shaped underwater pressure hull is V mat = S es ·t, then the correlation relationship is t = F1(L,B) as t = V mat ·(L·B 2 ) -2 / 3 / π.

[0121] The initial value range is as follows: 2.3 m ≤ L ≤ 2.7 m; 1.7 m ≤ B ≤ 2.3 m; the aspect - thickness ratio Φ = B / t, and 10 ≤ Φ ≤ 60.

[0122] S2: Collect multiple groups of data samples of structural parameters that meet the correlation relationship within the initial value range.

[0123] Establish a two - dimensional plane with L and B as variables.

[0124] Under the conditions of 2.3 m ≤ L ≤ 2.7 m and 1.7 m ≤ B ≤ 2.3 m, use Latin - hypercube sampling to collect multiple groups of corresponding L - B data within the two - dimensional plane.

[0125] Calculate t corresponding to the L - B data one - to - one based on t = F1(L,B) to obtain L - B - t data.

[0126] Judge whether the Φ of the obtained L - B - t data meets 10 ≤ Φ ≤ 606; if it meets, then L - B - t constitutes a data sample; if it does not meet, randomly select Φ within the range of β5 ≤ Φ ≤ β6, calculate t' based on Φ = B / t, replace t with t', and L - B - t constitutes a data sample.

[0127] S3: Determine 4 evaluation indicators for evaluating the performance of the to - be - designed special - shaped underwater pressure hull according to its use environment, obtain the original evaluation values corresponding to the 4 evaluation indicators of each group of data samples, and multiple data samples and their corresponding 4 original evaluation values constitute a data set.

[0128] Among them, the 4 evaluation indicators include the maximum stress for evaluating the load - bearing capacity of the to - be - designed special - shaped underwater hull, the buoyancy coefficient for evaluating the buoyancy performance of the to - be - designed special - shaped underwater hull the buckling eigenvalue for evaluating the stability of the to - be - designed special - shaped underwater hull, and the intercepted area S = F(B) for evaluating the hydrodynamic performance of the to - be - designed special - shaped underwater hull; where ρ mat refers to the material density of the to - be - designed special - shaped underwater hull, ρ refers to the seawater density, V mat refers to the material volume, V dis refers to the maximum drainage volume; and V mat = S es·t, S es Refers to the mid - surface area of the special - shaped underwater shell to be designed, S es = F2(L, B), that is, S es = π·(L·B 2 ) 2 / 3 .

[0129] In terms of hydrodynamic performance, the resistance of the water flow to the egg - shell - shaped underwater shell is selected as the evaluation criterion. The smaller the possible minimum resistance, the better the hydrodynamic performance. When an object in a fluid has a relative velocity with respect to the fluid, the object will be additionally subjected to a force from the fluid. The component parallel to and in the opposite direction of the object's moving velocity is called resistance. The underwater device will inevitably move relative to the water flow, and its resistance value is calculated by the following formula:

[0130]

[0131] In the formula, c res is the resistance coefficient, v is the relative velocity between the device and the water flow, and S is the cross - sectional area. Since for the same pressure - resistant shell under the same working conditions, c res ·ρ·v 2 is a constant value, the hydrodynamic performance is completely expressed by S, and the smaller the value of S, the better. Therefore, S = F(B)=π·(B + t) 2 / 44.

[0132] Based on the shape - parameter equation of the special - shaped underwater shell to be designed and all the data samples in step S2, establish a 3D model of the special - shaped underwater shell to be designed in 3D modeling software; and the shape - parameter equation of the egg - shell - shaped underwater shell is

[0133]

[0134] Obtain the L - B - t data and V dis corresponding to each group of data samples in the 3D model, and calculate the original evaluation value P1 of the buoyancy coefficient δ corresponding to L - B - t based on ;

[0135] Import the 3D simulation model into the simulation software to obtain the simulation model of the special - shaped underwater shell to be designed;

[0136] Obtain the original evaluation value P2 of the maximum stress and the original evaluation value P3 of the buckling eigenvalue of L - B - t corresponding to the data samples through the simulation model;

[0137] Based on S = F(B)=π·(B + t) 2 / 4, calculate the original evaluation value P4 of the intercepted area S of L - B - t corresponding to the data samples.

[0138] The data set consists of multiple sets of corresponding L-B-t and P1-P2-P3-P4. The data set is as shown in Figure 3 the figure.

[0139] S4: Train an artificial neural network model with the data set to obtain an initial prediction model for predicting the original evaluation value of the special-shaped underwater pressure hull to be designed.

[0140] Preferably, divide the data set into a training set and a test set according to a preset ratio, use the training set to train the artificial neural network model, and use the test set to test and optimize the initial prediction model obtained after training, which is beneficial to improving the accuracy and stability of the obtained initial prediction model.

[0141] S5: Optimize the initial prediction model using the particle swarm optimization algorithm to obtain the final target prediction model. As shown in Figure 4 the figure, it is the iteration graph of optimizing the initial prediction model using the particle swarm optimization algorithm.

[0142] is the schematic diagram of the initial prediction model. The steps of optimizing the initial prediction model using the particle swarm optimization algorithm are as follows:

[0143]

[0144] Among them, is the weight parameter between the i-th node of the input layer and the j-th node of the hidden layer, is the weight parameter between the j-th node of the hidden layer and the k-th node of the output layer, H b j is the constant term parameter of the hidden layer, H b k is the constant term parameter of the output layer. These parameters are used as the position of the particle in the particle swarm optimization algorithm p x i ; E is H b j 、 H b k the average error of each dimension of each data sample in the training set by the artificial neural network model. The optimization target of the particle swarm optimization algorithm is when this average error is minimized. N Sam is the number of samples, and n represents the n-th group of data samples.

[0145] Preferably, for the trained initial prediction model, it is necessary to use the test set data for prediction verification. If the average error of a certain sample in the test set is greater than 5%, retrain; if the error is greater than 5% near a certain data sample after multiple trainings, first return to step S2 to supplement L, B, or t that differs from the original value by 10% through random sampling, and then retrain.

[0146] S6: Extract 30 groups of data samples from the multiple groups of data samples in step S2, and use the genetic algorithm to optimize the obtained 30 groups of data samples to obtain 30 optimized groups of data samples.

[0147] When extracting n groups of data samples, in addition to meeting the initial value range and correlation relationship in step S1, the following conditions also need to be met: 0.54 ≤ B / L ≤ 0.84.

[0148] The method for obtaining 30 optimized groups of data samples is as follows:

[0149] Define the 30 groups of data samples as the initial population;

[0150] Perform non-dominated sorting on the individuals in the initial population;

[0151] Based on the non-dominated sorting result of the initial population, select the first a individuals in the initial population as excellent individuals, perform crossover on L and B of any two excellent individuals, and judge whether L, B, and t simultaneously meet the initial value range, correlation relationship, and 0.54 ≤ B / L ≤ 0.84 after crossover; if they meet, retain the two new offspring individuals generated by this crossover; if they do not meet, do not retain the two new offspring individuals generated by this crossover, and perform crossover again until L, B, and t after crossover simultaneously meet the initial value range, correlation relationship, and 0.54 ≤ B / L ≤ 0.84;

[0152] After reaching the preset number of crossover times, stop crossover, and the individuals in the birth population and the retained offspring individuals form a new population;

[0153] Set the mutation probability, perform mutation on L and B of all individuals in the new population, and judge whether L, B, and t generated after individual mutation simultaneously meet the initial value range, correlation relationship, and 0.54 ≤ B / L ≤ 0.84; if they meet, retain the new offspring individual generated by this mutation; if they do not meet, do not retain the new offspring individual generated by this mutation, and re-perform mutation on L and B of this individual until L, B, and t after mutation simultaneously meet the initial value range, correlation relationship, and 0.54 ≤ B / L ≤ 0.84;

[0154] All the new offspring individuals after the new population is mutated form the target population, and perform crowding degree sorting on the individuals in the target population;

[0155] Select the first 30 groups of individuals in the crowding degree sorting as the 30 optimized groups of data samples.

[0156] S7: Input the optimized n groups of data samples into the target prediction model to obtain 4 prediction evaluation values corresponding to the 30 optimized groups of data samples, and calculate the weight values of 4 evaluation indicators based on the 4 prediction evaluation values corresponding to the 30 optimized groups of data samples. The distribution diagram of the weight values is asFigure 5 as shown

[0157] Set the first weight of the 4 evaluation indicators to

[0158] Use the C-OWA operator to objectively evaluate the 4 predicted evaluation values corresponding to 30 groups of data samples, and obtain the second weight of the set 4 evaluation indicators as The first weight and the second weight of this embodiment are shown in Table 1 below, and this embodiment sets four different first weights for four different usage scenarios.

[0159] Table 1

[0160] Category Weight Distribution Principle <![CDATA[Weight ratio (P3:P1:P2:P4)]]> First Weight - Usage Scenario 1 <![CDATA[P3 = P2 > P1 = P4]]> 0.300:0.200:0.300:0.200 First Weight - Usage Scenario 2 <![CDATA[P1>P4>P3=P1]]> 0.100:0.550:0.100:0.250 First Weight - Usage Scenario 3 <![CDATA[P3 = P1 = P2 = P4]]> 0.250:0.250:0.250:0.250 First Weight - Usage Scenario 4 <![CDATA[P2 = P3 > P1]]> 0.400:0.100:0.400:0.000 Second Weight 0.287:0.249:0.227:0.238

[0161] Construct the combined weight w based on the first weight and the second weight *j , where k1 is the coefficient of the first weight, k2 is the coefficient of the second weight, and j ∈ [1, 4];

[0162] Calculate w based on game theory * and w sub with the smallest dispersion, and w * and w obj with the smallest dispersion; that is: min||w * - w||;

[0163] where ||·|| is the two-norm of a vector or matrix, and the matrix form of the equation optimized by the first derivative of the above formula is:

[0164]

[0165] Find the combined coefficients k1 and k2, and perform normalization processing on k1 and k2;

[0166] The weight values of the 4 evaluation indicators are w* = [w *1 : w *2 : …: w *4 . The weight values of this embodiment are shown in Table 2 below:

[0167] Table 2

[0168] Category Game Theory Weight <![CDATA[Combined weights (P3:P1:P2:P4)]]> Usage Scenario 1 1.641:-0.641 0.310:0.175:0.314:0.202 Usage Scenario 2 0.948:0.052 0.204:0.534:0.157:0.105 Usage Scenario 3 0.000:1.000 0.285:0.240:0.278:0.198 Usage Scenario 4 0.935:0.086 0.399:0.114:0.398:0.017

[0169] S8: Use the ideal distance method and combine the weight values to select a set of optimal data samples for the actual design of the special-shaped underwater pressure hull from the optimized n groups of data samples.

[0170] Construct the weighted normalized decision matrix T based on the weight values of the 4 evaluation indicators 30×4 ,

[0171]

[0172] where P n,m represents the predicted evaluation value of the m - th evaluation index of the n - th data sample among the optimized n groups of data samples;

[0173] Based on the ideal distance method, let T 30×4 The positive ideal value of the g - th column of The negative ideal value is

[0174]

[0175] Calculate the Euclidean distance between each data sample in the optimized 30 groups of data samples and the positive ideal value The Euclidean distance between each data sample in the optimized 30 groups of data samples and the negative ideal value

[0176]

[0177] Calculate the proximity index. The proximity index of the n - th Popu group of solutions is expressed as:

[0178]

[0179] Select a group of data samples with the largest proximity index among the optimized 30 groups of data samples for the actual design of the special - shaped underwater pressure hull. The optimized results of this embodiment are shown in Table 3 below, and only the first 3 groups of the optimized 30 groups of data samples are shown in Table 3.

[0180] Table 3

[0181]

[0182]

[0183] As shown in Table 3, the input and output of Use Case 1, Use Case 3, and Use Case 4 are the same. The comparison graph of the predicted evaluation values of Use Case 1, Use Case 3, Use Case 4, and Use Case 2 in this embodiment and the original evaluation values obtained in Step S3. It can be clearly seen from this attached figure that the predicted values and the actual values are almost the same, which can prove that the optimized results obtained by this method have high credibility.

Claims

1. A design method for a special-shaped underwater pressure hull, characterized in that: It includes the following steps: S1: Determine various types of structural parameters and their initial value ranges that are required and related to each other according to the usage environment of the to-be-designed special-shaped underwater pressure hull. S2: Collect multiple groups of data samples of structural parameters that meet the related relationship within the initial value range. S3: Determine m evaluation indicators for evaluating its performance according to the usage environment of the to-be-designed special-shaped underwater pressure hull, and obtain the original evaluation values corresponding to the m evaluation indicators of each group of data samples. Multiple data samples and their corresponding m original evaluation values form a data set. S4: Train an artificial neural network model with the data set to obtain an initial prediction model for predicting the original evaluation values of the to-be-designed special-shaped underwater pressure hull. S5: Optimize the initial prediction model using the particle swarm optimization algorithm to obtain the final target prediction model. S6: Extract n groups of data samples from the multiple groups of data samples in step S2, and perform optimization processing on the obtained n groups of data samples using the genetic algorithm to obtain n groups of optimized data samples. S7: Input the n groups of optimized data samples into the target prediction model to obtain m prediction evaluation values corresponding to the n groups of optimized data samples, and calculate the weight values of the m evaluation indicators based on the m prediction evaluation values corresponding to the n groups of optimized data samples. S8: Select a set of optimal data samples for the actual design of the special-shaped underwater pressure hull from the n groups of optimized data samples using the ideal distance method in combination with the weight values.

2. The design method of the special-shaped underwater pressure hull according to claim 1, characterized in that: The various types of structural parameters described in step S1 include the major axis length L, the minor axis length B, and the thickness t; the related relationship between the various types of structural parameters is t = F1(L, B).

3. The design method of the special-shaped underwater pressure hull according to claim 2, wherein: The initial value ranges described in step S1 are as follows: β1 ≤ L ≤ β2; β3 ≤ B ≤ β4; the aspect ratio Φ = B / t, and β5 ≤ Φ ≤ β6.

4. The design method of the special-shaped underwater pressure hull according to claim 3, characterized in that: The method for collecting multiple groups of data samples of structural parameters that meet the related relationship in step S2 is as follows: Establish a two-dimensional plane with L and B as variables. Under the conditions of β1 ≤ L ≤ β2 and β3 ≤ B ≤ β4, randomly collect multiple groups of corresponding L-B data in the two-dimensional plane. Calculate t corresponding to the L-B data one by one based on t = F1(L, B) to obtain L-B-t data. Judge whether the Φ of the obtained L-B-t data satisfies β5 ≤ Φ ≤ β6; if it satisfies, the L-B-t constitutes a data sample; if it does not satisfy, randomly extract Φ within the range of β5 ≤ Φ ≤ β6, calculate t' based on Φ = B / t, replace t with t', and the L-B-t constitutes a data sample.

5. The design method of the special-shaped underwater pressure hull according to claim 4, characterized in that: The m evaluation indexes in step S3 include the maximum stress for evaluating the load-bearing capacity of the to-be-designed special-shaped underwater shell, and the buoyancy coefficient for evaluating the buoyancy performance of the to-be-designed special-shaped underwater shell The buckling eigenvalue for evaluating the stability of the to-be-designed special-shaped underwater shell and the intercepted area S = F(B) for evaluating the hydrodynamic performance of the to-be-designed special-shaped underwater shell; where ρ mat refers to the material density of the to-be-designed special-shaped underwater shell, ρ refers to the seawater density, V mat refers to the material volume, V dis refers to the maximum drainage volume; and V mat = S es ·t, S es refers to the mid-surface area of the to-be-designed special-shaped underwater shell, S es = F2(L,B).

6. The design method of the special-shaped underwater pressure hull according to claim 5, characterized in that: The method for obtaining the data set in step S3 is as follows: Based on the shape parameter equation of the to-be-designed special-shaped underwater hull and all the data samples in step S, establish a three-dimensional model of the to-be-designed special-shaped underwater hull in 3D modeling software. Obtain the L-B-t data and V corresponding to each set of data samples in the three-dimensional model dis , and based on calculate the original evaluation value P1 of the buoyancy coefficient δ corresponding to L-B-t; Import the three-dimensional simulation model into the simulation software to obtain the simulation model of the to-be-designed special-shaped underwater hull. Obtain the original evaluation value P2 of the maximum stress and the original evaluation value P3 of the buckling eigenvalue of L-B-t corresponding to the data sample through the simulation model. Calculate the original evaluation value P4 of the intercepted area S of L-B-t corresponding to the data sample based on S = F(B). The dataset consists of multiple sets of corresponding L-B-t and P1-P2-P3-P4.

7. The design method of the special-shaped underwater pressure hull according to claim 1, characterized in that: In step S6, when extracting n groups of data samples, in addition to meeting the initial value range and correlation relationship in step S1, the following conditions also need to be satisfied: β7 ≤ B / L ≤ β8.

8. The design method of the special-shaped underwater pressure hull according to claim 1, characterized in that: The method for obtaining the optimized n groups of data samples in step S6 is as follows: Define the n groups of data samples as the initial population; Perform non-dominated sorting on the individuals in the initial population; Based on the non-dominated sorting result of the initial population, select the top a individuals in the initial population as excellent individuals, cross the L and B of any two excellent individuals, and judge whether L, B, and t simultaneously meet the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8 after crossing; if they meet, retain the two new offspring individuals generated by this crossing; if they do not meet, do not retain the two new offspring individuals generated by this crossing, and re-perform the crossing until L, B, and t after crossing simultaneously meet the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8; After reaching the preset number of crossing times, stop crossing, and the individuals in the birth population and the retained offspring individuals form a new population; Set the mutation probability, mutate the L and B of all individuals in the new population, and judge whether L, B, and t generated after individual mutation simultaneously meet the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8; if they meet, retain the new offspring individual generated by this mutation; if they do not meet, do not retain the new offspring individual generated by this mutation, and re-mutate the L and B of this individual until L, B, and t after mutation simultaneously meet the initial value range, correlation relationship, and β7 ≤ B / L ≤ β8; All the new offspring individuals after the new population is mutated form the target population, and perform crowding degree sorting on the individuals in the target population; Select the top n groups of individuals in the crowding degree sorting as the optimized n groups of data samples.

9. The design method of the special-shaped underwater pressure hull according to claim 8, characterized in that: The method for calculating the weight values of m evaluation indicators in step S7 is as follows: Set the first weight of m evaluation indicators as Use the C-OWA operator to objectively evaluate the m predicted evaluation values corresponding to n groups of data samples, and obtain the second weight of the set of m evaluation indicators as Construct the combined weight w based on the first weight and the second weight *g , where k1 is the coefficient of the first weight, k2 is the coefficient of the second weight, and g ∈ [1, m]; Calculating w based on game theory * and w sub has the smallest degree of discreteness and w * and w obj The combination coefficients k1 and k2 with the smallest degree of discreteness; The weight values of m evaluation indicators are w* = [w *1 : w *2 : …: w *m .

10. The design method of the special-shaped underwater pressure hull according to claim 9, characterized in that: The method for obtaining the optimal data sample in step S8 is as follows: Construct a weighted normalized decision matrix based on the weight values of m evaluation indicators Among which P n,m represents the predicted evaluation value of the m-th evaluation index of the n-th data sample among the optimized n groups of data samples; Based on the ideal distance method, assume that the positive ideal value of the g-th column of is Calculate the Euclidean distance between each data sample in the optimized n groups of data samples and the positive ideal value Calculate the Euclidean distance between each data sample in the optimized n groups of data samples and the negative ideal value Calculate the proximity index, the nth Popu proximity index of the group of solutions is expressed as: Select a group of data samples with the largest proximity index from the optimized n groups of data samples for the actual design of the special-shaped underwater pressure hull.

Images