An acoustic metamaterial cabin wall surface acoustic wave regulation interaction gradual design method

Abstract

The application relates to the technical field of acoustic metamaterials, and discloses an acoustic metamaterial cabin wall surface sound wave regulation interaction gradual design method. The acoustic metamaterial is covered on a cabin wall, and the acoustic metamaterial cabin wall surface sound wave regulation interaction gradual design method is characterized in that two sides of the cabin wall surface contacting different media are respectively taken as an A surface and a B surface, and the method comprises the following steps: S1, selecting a periodic structure unit based on acoustic metamaterials to establish a basic three-dimensional structure; S2, collecting characteristic parameters of the basic three-dimensional structure and corresponding output sound pressure levels as a data set; and S3, expanding the data set through a data semantic augmentation method based on a genetic algorithm. The acoustic metamaterial cabin wall surface sound wave regulation interaction gradual design method fully considers the manufacturability of a design structure, feeds back a three-dimensional model based on manufacturing process constraints, considers manufacturing limit sizes, considers application environment limitations, and realizes structure-manufacturing-performance collaborative optimization.

Description

Technical Field

[0001] This invention relates to the field of acoustic metamaterials technology, and in particular to an interactive progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials. Background Technology

[0002] Faced with the urgent need for efficient shipping, larger ships will become a development trend in the shipping industry. However, with the increase in ship size, the power of the main engine and propeller will increase dramatically, and external eddies will easily induce impact noise. During navigation, this will generate more noise, seriously affecting the comfort and safety of shipping. To avoid excessive cabin noise, it is necessary to conduct research on noise suppression of ship walls during the design phase.

[0003] Acoustic metamaterials are structural materials composed of subwavelength resonant or non-resonant units arranged periodically or aperiodically. By designing subwavelength unit structures and introducing abrupt phase changes on the metasurface, sound fields can be manipulated. They offer advantages such as low cost and low thermal viscous loss, significantly increasing the freedom of artificial sound wave manipulation and showing excellent application prospects in noise suppression. Specifically, these subwavelength unit structures, arranged according to certain rules on the surface, can achieve various unique acoustic functions such as diffuse sound reflection and asymmetric sound transmission. Surface sound pressure level is the most direct acoustic indicator for evaluating the noise suppression of acoustic structures. Acoustic structures mainly consist of the sound wave incident surface and the exit surface. The structural parameters of the incident interface and the loss of the exit surface are key factors in the propagation of rigid sound waves and are also the most intuitive and crucial technical indicators for optimizing acoustic structures.

[0004] Current conventional multi-objective optimization methods for acoustic structures struggle to rationally allocate the weights among the coupling interface, solid structure, and acoustic medium in acoustic-structure interaction systems, particularly failing to consider the coordinated control of sound pressure levels from both sides (incident and exit surfaces). Furthermore, most current optimization methods are based on parameter optimization, which, due to small sample sizes, may fall into the "local optimum" trap, exhibiting over-iteration and poor robustness. Additionally, most current optimization methods for acoustic metamaterials are limited to structural optimization, neglecting manufacturing process limitations. Most importantly, in the face of diverse marine environments, the issues of noise radiation and suppression are rarely addressed. Summary of the Invention

[0005] In view of this, this invention proposes an interactive progressive design method for acoustic metamaterial ship cabin wall acoustic wave control. It fully considers the coordinated control of the acoustic wave incident surface and the reflecting surface. Based on the improved data semantic augmentation method, it provides sufficient samples for the optimization of key parameters, reduces the sensitivity of optimization to the sampling rate, and further considers the manufacturability of the designed structure. Based on the manufacturing process constraint feedback optimization of the structural model, it considers the manufacturing limit size and the application environment limitation, and realizes the coordinated optimization of structure, manufacturing and performance. It has broad potential application prospects in the adaptive control of the sound field.

[0006] The technical solution of this invention is implemented as follows: This invention provides an interactive progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials. The acoustic metamaterials are covered on the ship cabin walls. The method is characterized by using the two sides of the ship cabin wall that contact different media as surface A and surface B, respectively. The method includes the following steps:

[0007] S1. Select periodic structural units based on acoustic metamaterials to establish the basic three-dimensional structure;

[0008] S2. Collect the feature parameters of the basic three-dimensional structure and the corresponding output sound pressure level as a dataset;

[0009] S3. Expand the dataset using a data semantic augmentation method based on genetic algorithms;

[0010] S4. Construct a wavelet neural network and train it by expanding the dataset so that the wavelet neural network optimizes the input feature parameters to minimize the sound pressure level of surface A and surface B.

[0011] S5. Set the initial model and extract the feature parameters of the initial model;

[0012] S6. Input the feature parameters of the initial model into the wavelet neural network for optimization;

[0013] S7. Determine whether the feature parameters of the optimized initial model meet the constraints of the additive manufacturing process. If not, return to step S6. If they do meet, determine whether the initial model meets the additive manufacturing process.

[0014] S8. Verify the performance of the manufacturing model under various environmental factors to determine whether it meets the sound pressure level requirements. If it does not meet the requirements, return to step S5. If it does meet the requirements, it is determined to be the final model.

[0015] Based on the above technical solutions, preferably, step S3, the data semantic augmentation method based on genetic algorithms, includes the following steps:

[0016] S31. Extract the feature parameters of the dataset in step S2, encode them into binary, form a number string to simulate chromosomes, and form an initial population.

[0017] S32. Perform single-point binary crossover on the initial population to mutate and form a new population;

[0018] S33. Calculate the fitness of individuals in the population based on the fitness function, and perform evolution with the minimum sound pressure level as the optimization objective;

[0019] S34. Decode the new subset generated by the evolution and construct the parameters of the optimal extractor for the data semantic augmentation method, while obtaining the unclassified augmented dataset.

[0020] S35. Classify the unclassified augmented dataset by comparing the optimized classifier and the trained classifier of the semantic augmentation method with the input data to obtain the classified augmented dataset.

[0021] Based on the above technical solutions, preferably, step S3 further includes Gaussian modeling and distribution estimation of the expanded dataset.

[0022] More preferably, a function to minimize the classification error is introduced into the training classifier, and a sampling rate is set.

[0023] Further preferably, step S3 also includes setting an average recognition rate threshold. When the average recognition rate of the data semantic augmentation method reaches the average recognition rate threshold, the relevant weights of the data semantic augmentation method are corrected by using the feature parameters of the dataset in step S2 as input and the sound pressure levels of the A and B sides of the cabin wall as output, through a linear weighting method.

[0024] More preferably, the linear weighting method includes introducing weighting factors λ1, λ2 and λ3 to establish constraint functions on sound propagation medium 1—ship cabin wall A, sound propagation medium 2—ship cabin wall B and the ship cabin wall.

[0025] Based on the above technical solutions, preferably, step S4, which constructs a wavelet neural network, includes initializing network weights and wavelet parameters, calculating network prediction output and network prediction error, and correcting the relevant weights of the wavelet neural network based on the network prediction error.

[0026] Based on the above technical solutions, preferably, the performance under each environmental factor in step S8 includes at least one of thermal deformation, electromagnetic noise, and vibration noise.

[0027] Based on the above technical solution, preferably, the acoustic wave equation relating the sound pressure, position, and time of the basic three-dimensional structure in step S1 is as follows:

[0028]

[0029] Where v is the acoustic particle vibration velocity, p is the sound pressure, ρ is the fluid mass density, and K is the bulk modulus constant. Let represent the Laplace operator, and t represent time.

[0030] More preferably, in step S2, the output sound pressure level corresponding to the feature parameters of the dataset is obtained by directly performing finite element discretization on the acoustic medium domain according to the Helmholtz equation and establishing the sound field solution. The Helmholtz equation for plane sound waves in an ideal medium is:

[0031]

[0032] Under excitation response, the sound pressure at any point in the design domain of the solid structure is:

[0033] p=∫ Γs iω[g T -(h T -g T Y)(H-GY) -1 G]A -1 (ω)f(ω)ndΓ

[0034] Among them, Γ s Let be a solid, i be the imaginary part, ω be the angular frequency, h and g be column vectors, Y be the corner matrix, H and G be the global system matrices, A(ω) be the frequency response function, f(ω) be the harmonic excitation, and n be the normal vector.

[0035] The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method of the present invention has the following advantages over the prior art:

[0036] (1) By using the improved data semantic augmentation method, sufficient samples are provided for the optimization of key parameters, reducing the sensitivity of optimization to the sampling rate and over-iteration phenomenon of small sample optimization. At the same time, when optimizing feature parameters, the wavelet neural network can reasonably allocate the weights among the coupling interface, solid structure and acoustic medium in the acoustic-solid coupling system. In addition, this method considers the manufacturability of the designed structure, optimizes the structural model based on manufacturing process constraint feedback, considers the manufacturing limit size, considers the application environment limitation, and realizes the structure-manufacturing-performance collaborative optimization.

[0037] (2) By considering the coordinated control of two-sided acoustic waves, the sound pressure levels of surface A and surface B are used as outputs for multi-objective optimization. The manufacturing limit size is considered and fed back to the initial optimization model for further iteration. Temperature cycle thin-wall deformation, electromagnetic noise and mechanical vibration noise during navigation are considered in actual application scenarios. Further performance verification of the optimized structure is carried out to achieve interactive design and incremental optimization from structure to performance. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram illustrating the steps of the acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method of the present invention.

[0040] Figure 2 This is a schematic diagram of the overall structure of the acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method of the present invention.

[0041] Figure 3 This is a schematic diagram of the acoustic wave transmission and three-dimensional model of the acoustic metamaterial ship cabin wall acoustic wave control interactive progressive design method of the present invention. Detailed Implementation

[0042] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0043] like Figure 2 As shown, the present invention relates to an interactive progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials. The acoustic metamaterials are covered on the ship cabin walls. Specifically, it is an integrated interactive progressive collaborative design method of structure-manufacturing-function, which is generally divided into four major modules: preprocessing, structural model, manufacturing model, and functional model.

[0044] The preprocessing module primarily involves establishing a two-dimensional topological model, such as... Figure 3 As shown, in this embodiment, an octagonal truss unit is selected as the two-dimensional unit cell, and the preferred material is AlSi10Mg. Considering the three-dimensional spatial distribution of the unit, a basic three-dimensional conceptual model is constructed. It should be noted that the use of an octagonal truss unit as the two-dimensional unit cell is only an example, and other periodic cubic lattice structures can be used.

[0045] In the structural model module, the feature parameters of the three-dimensional conceptual model are extracted, namely the side length a, wall thickness b, and spatial angle θ. The parameters are optimized based on artificial intelligence methods. An initial model is set. During the optimization process, the asymmetry of the acoustic medium of the cabin wall needs to be considered, that is, the cabin is filled with air and the cabin is filled with water. The artificial intelligence method refers to the parameter optimization model. In this embodiment, wavelet neural network is used to optimize the key dimensions of the feature parameters.

[0046] In the manufacturing model module, the optimized structure is mainly manufactured using additive manufacturing. Considering the process constraints, process parameters are determined, and the limiting dimensions are determined from these parameters. The manufacturing limiting dimension is the minimum wall thickness h. min To meet the minimum dimensions required for additive manufacturing forming accuracy, which are limited by process parameters such as laser power, scanning speed, and scanning spacing, feedback is fed back to the structural model for optimization to determine the final design structure. If the design does not meet the constraints of the process parameters, critical dimensions need to be optimized again.

[0047] In the functional model module, temperature cycling, electromagnetic noise, and mechanical vibration during ship navigation are considered. Finite element analysis is performed on the three operating conditions to verify the rationality of the final design structure. If the environmental performance is not met, the initial model needs to be determined again.

[0048] like Figure 1-3 As shown, this method uses the two sides of the ship's cabin wall that are in contact with different media as surface A and surface B, respectively, and specifically includes steps S1-S8.

[0049] Step S1: Select periodic structural units based on acoustic metamaterials to establish a basic three-dimensional structure.

[0050] In this embodiment, an octagonal truss element is selected as the basic two-dimensional unit cell. Therefore, in this step, based on the three-dimensional initial octagonal truss model, a boundary element model of the acoustic-structure interaction (ASIE) boundary is established. The boundary element model of the acoustic-structure interaction boundary is a numerical calculation method used to simulate ASIE problems. In ASIE problems, the sound field and the solid structure interact, requiring consideration of the propagation and reflection of sound waves on the solid structure. The boundary element model of the acoustic field, based on the boundary element method, represents the sound pressure and particle velocity at the sound field boundary as unknowns on the boundary, and solves them using the basic principles and formulas of the boundary element method. It transforms the sound field boundary conditions on the boundary into boundary element equations, establishes a system of boundary element equations, and solves the sound field distribution using numerical methods. The boundary element model of the acoustic-structure interaction boundary considers the propagation and reflection of sound waves on the solid structure and can be used to simulate the sound field distribution and the response of the solid structure in ASIE problems.

[0051] The equation for the sound wave relationship between sound pressure, location, and time is as follows:

[0052]

[0053] In the formula, v is the vibration velocity of the acoustic particle, p is the sound pressure, ρ is the fluid mass density, and K is the bulk modulus constant. Let represent the Laplace operator, and t represent time.

[0054] Step S2: Collect the feature parameters of the basic three-dimensional structure and the corresponding output sound pressure level as the feature parameters of the dataset.

[0055] Multiple sets of characteristic parameters of different basic three-dimensional structures are collected, and the output sound pressure level corresponding to different characteristic parameters is obtained. The output sound pressure level corresponding to the characteristic parameters can be directly discretized into a finite element method according to the Helmholtz equation to establish a numerical equation for solving the sound field. The Helmholtz equation for plane sound waves in an ideal medium is:

[0056]

[0057] Under excitation response, the sound pressure at any point in the design domain of the solid structure is:

[0058] p=∫ Γs iω[g T -(h T -g T Y)(H-GY) -1 G]A -1 (ω)f(ω)ndΓ

[0059] Among them, Γ s Let be a solid, i be the imaginary part, ω be the angular frequency, h and g be column vectors, Y be the corner matrix, H and G be the global system matrices, A(ω) be the frequency response function, f(ω) be the harmonic excitation, and n be the normal vector.

[0060] After obtaining the dataset, it is used as the basic dataset to prepare for subsequent dataset expansion. In this embodiment, the number of datasets collected is 200, which are divided into training set, validation set and test set, and the ratio of the three is set to 7:2:1.

[0061] Step S3: Expand the dataset using a data semantic augmentation method based on genetic algorithms.

[0062] Data semantic augmentation is a technique used to enhance datasets. By transforming or expanding the original data, the semantic representation of the data is changed, thereby improving the generalization ability and robustness of the model. In this embodiment, the data semantic augmentation method is improved through genetic algorithm and dataset expansion to provide sufficient data for subsequent wavelet neural network training, provide sufficient samples for the optimization of key parameters, and reduce the sensitivity of optimization to sampling rate.

[0063] The data semantic augmentation method based on genetic algorithms specifically includes steps S31-S35.

[0064] Step S31: Extract the feature parameters of the dataset from step S2, encode them into binary, form a number string to simulate chromosomes, and form the initial population.

[0065] Key features, namely side length a, wall thickness b, and spatial angle θ, are extracted, binary encoded, and a string of numbers consisting of 0s and 1s is used to simulate chromosomes to form an initial population.

[0066] Step S32: Perform a single-point binary crossover on the initial population to mutate and form a new population.

[0067] Perform a single-point binary crossover with a mutation rate of 1% to generate a new population from the initial population.

[0068] Step S33: Calculate the fitness of individuals in the population based on the fitness function, and perform evolution with the minimum sound pressure level as the optimization objective.

[0069] The fitness of individuals is calculated based on the fitness function, with the minimum sound pressure level as the optimization objective and the migration number set to 0.15. Evolutionary calculations are performed to generate new subsets.

[0070] Step S34: Decode the new subset generated by the evolution and construct the parameters of the optimal extractor for the data semantic augmentation method, while obtaining the unclassified augmented dataset.

[0071] In this step, the resulting decoded subset is incorporated into the dataset to obtain the expanded, unclassified dataset.

[0072] Step S35: Classify the unclassified augmented dataset by comparing the optimized classifier and the trained classifier of the semantic augmentation method with the input data to obtain the classified augmented dataset.

[0073] In this embodiment, the optimizing classifier and the training classifier of the data semantic augmentation method can be selected as a Bayesian optimizing classifier and a linear training classifier, respectively, to classify the unclassified dataset.

[0074] In addition, this embodiment also sets an average recognition rate threshold. When the average recognition rate of the data semantic augmentation network reaches the average recognition rate threshold, the feature parameters are used as input, and the two sides of the ship's cabin wall that are in contact with different media are respectively regarded as surface A and surface B. The sound pressure level of surface A and surface B of the ship's cabin wall is used as output. The relevant weights of the data semantic augmentation method are corrected by a linear weighting method. The weights are the degree of significance of the parameters on the sound pressure level.

[0075] Specifically, such as Figure 3As shown, the dual-sided acoustic wave coordinated control mainly considers the service environment of the ship's bulkhead. Based on the asymmetry of the acoustic wave propagation medium, during the characteristic parameter optimization process, the objective function that minimizes the sound pressure levels of surface A and surface B is constructed as the output. The weights between the two sides of the ship's bulkhead structure and the acoustic medium are weighted and corrected. Weighting factors λ1, λ2 and λ3 are introduced to establish constraint functions on acoustic propagation medium 1 (air) - ship's bulkhead A, acoustic propagation medium 2 (water) - ship's bulkhead B and the ship's bulkhead, thereby preventing over-iteration and over-optimization of the output and input.

[0076] Meanwhile, a function to minimize the classification error is introduced into the training classifier, and a sampling rate is set. In this embodiment, the sampling rate is set to 0.7 to avoid the selection of the sampling rate hyperparameter affecting the training effect.

[0077] This step expands the sample during the meta-training phase, performs Turkey transformation on the sample features, transfers the features, and performs Gaussian modeling and distribution estimation on similar parameters of the metadata expansion.

[0078] The Turkey transformation method is mainly used for regression of small sample data with unknown population standard deviation. Its basic idea is to compare the absolute value of the difference between the mean values ​​of the same group based on pairwise comparisons of the studentized range. Gaussian modeling and distribution estimation of similar parameters of metadata extension are performed because optimal likelihood estimation cannot be achieved under finite sample conditions. Therefore, Gaussian modeling and distribution estimation of extended data are required to extend the scope of estimation analysis to sparse models in order to meet the reasonable likelihood estimation of the whole.

[0079] Step S4: Construct a wavelet neural network. Train the wavelet neural network by expanding the dataset so that the wavelet neural network optimizes the input feature parameters to the minimum feature parameters of the sound pressure level of surface A and surface B.

[0080] Constructing a wavelet neural network includes initializing network weights and wavelet parameters, calculating the network prediction output and network prediction error, and correcting the relevant weights of the wavelet neural network based on the network prediction error. During the training of the wavelet neural network, the expanded dataset is divided into a training set and a test set. In this embodiment, the division ratio is 7:3, which allows the wavelet neural network to optimize the input feature parameters.

[0081] Step S5: Set up the initial model and extract the feature parameters of the initial model.

[0082] In this step, the initial 3D model is reconstructed. This can be done by manually designing the initial model and then extracting its feature parameters.

[0083] Step S6: Input the feature parameters of the initial model into the wavelet neural network to optimize the feature parameters.

[0084] The initial model feature parameters obtained in step S5 are input into the wavelet neural network for optimization, and the optimized feature parameters are determined to prepare for the final model determination.

[0085] Step S7: Determine whether the optimized feature parameters of the initial model meet the constraints of the additive manufacturing process. If not, return to step S6. If they do meet, obtain the feature parameters of the manufacturing model.

[0086] In this embodiment, the limiting dimension h is determined by using additive manufacturing process parameters as constraints, namely, laser power of 100W, scanning rate of 1m / s, and scanning spacing of 50μm. min The value is 2μm. When all the key optimized parameters are greater than 2μm, the initial model can be determined as a 3D manufacturing model. The manufacturing model obtained in this step can be manufactured through a process. If other processes are used to optimize the dimensions, the parameters of the manufacturing process can be constrained in a similar way. If the constraints of the manufacturing process are not met, it means that the initial model cannot be manufactured. In this case, it is necessary to return to step S6 and re-optimize the feature parameters of the initial model. Of course, it is also possible to directly return to step S5 and rebuild the initial model.

[0087] Step S8: Verify the performance of the manufacturing model under various environmental factors and determine whether it meets the sound pressure level requirements. If it does not meet the requirements, return to step S5. If it does meet the requirements, it is determined to be the final model.

[0088] The performance under various environmental factors includes at least one of heat value deformation, electromagnetic noise and vibration noise. In this embodiment, all three situations are taken into account.

[0089] This embodiment proposes a method for determining a functional model based on calorific value deformation, electromagnetic noise, and vibration noise.

[0090] The parameters of temperature, electromagnetic excitation and mechanical vibration are determined. First, thermo-solid deformation calculation is performed on the manufacturing model under the service temperature cycle to determine the deformation. Then, electromagnetic and mechanical noise calculation is performed to obtain the sound pressure level at this time. It is determined whether the sound pressure level requirement is met. If it is met, the obtained manufacturing model is determined as the final model. If it is not met, return to step S5 to reconstruct the initial model.

[0091] In one specific embodiment, the service temperature cycle is determined and set to 10℃~40℃. Thermo-structural deformation calculations are performed. After meshing, surface B is fixed, a temperature load is applied, and the deformation is determined. When the minimum deformation (20μm) is met, electromagnetic and mechanical noise calculations are performed to obtain the sound pressure level. Due to the complexity of the geometry, tetrahedral mesh elements are used for calculation. For electromagnetic noise calculation, the electromagnetic field and structural field are unidirectionally coupled. A mapping function between the nodes of the electromagnetic field model and the nodes of the structural field model is established based on the finite element method, mapping the electromagnetic force to the structural solver. For mechanical vibration noise calculation, the acoustic finite element mesh layer is filled with idealized air, acoustic excitation is added, and the sound pressure level is solved. When the sound pressure level requirement is met, the final optimized geometry, i.e., the final model, is determined.

[0092] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials, wherein the acoustic metamaterials are applied to the ship cabin walls, characterized in that... Using the two sides of the ship's bulkhead that come into contact with different media as surface A and surface B respectively, the method includes the following steps: S1. Select periodic structural units based on acoustic metamaterials to establish a three-dimensional structure; S2. Collect the feature parameters of the three-dimensional structure and the corresponding output sound pressure level as a dataset; S3. Expand the dataset using a data semantic augmentation method based on genetic algorithms; The data semantic augmentation method based on genetic algorithms in step S3 includes the following steps: S31. Extract the feature parameters of the dataset in step S2, encode them into binary, form a number string to simulate chromosomes, and form an initial population. S32. Perform single-point binary crossover on the initial population to mutate and form a new population; S33. Calculate the fitness of individuals in the population based on the fitness function, and perform evolution with the minimum sound pressure level as the optimization objective; S34. Decode the new subset generated by the evolution and construct the parameters of the optimal extractor for the data semantic augmentation method, while obtaining the unclassified augmented dataset. S35. Classify the unclassified augmented dataset by comparing the optimized classifier and the trained classifier of the semantic augmentation method with the input data to obtain the classified augmented dataset. Step S3 further includes setting an average recognition rate threshold. When the average recognition rate of the data semantic augmentation method reaches the average recognition rate threshold, the relevant weights of the data semantic augmentation method are corrected by using the feature parameters of the dataset in step S2 as input and the sound pressure levels of the A and B sides of the cabin wall as output, through a linear weighting method. S4. Construct a wavelet neural network and train it by expanding the dataset so that the wavelet neural network optimizes the input feature parameters to minimize the sound pressure level of surface A and surface B. S5. Set the initial model and extract the feature parameters of the initial model; S6. Input the feature parameters of the initial model into the wavelet neural network for optimization; S7. Determine whether the feature parameters of the initial model after optimization meet the constraints of the additive manufacturing process. If not, return to step S6. If they do meet, obtain the manufacturing model. S8. Verify the performance of the manufacturing model under various environmental factors to determine whether it meets the sound pressure level requirements. If it does not meet the requirements, return to step S5. If it does meet the requirements, it is determined to be the final model.

2. The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method as described in claim 1, characterized in that: Step S3 also includes performing Gaussian modeling and distribution estimation on the expanded dataset.

3. The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method as described in claim 1, characterized in that: A function to minimize the classification error is introduced into the training classifier, and the sampling rate is set.

4. The interactive progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials as described in claim 1, characterized in that: The linear weighting method includes introducing weighting factors λ1, λ2, and λ3 to establish constraint functions on sound propagation medium 1—ship cabin wall A, sound propagation medium 2—ship cabin wall B, and the ship cabin wall.

5. The interactive progressive design method for acoustic wave modulation of ship cabin walls using acoustic metamaterials as described in claim 1, characterized in that: Step S4, constructing the wavelet neural network, includes initializing network weights and wavelet parameters, calculating the network prediction output and network prediction error, and correcting the relevant weights of the wavelet neural network based on the network prediction error.

6. The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method as described in claim 1, characterized in that: The performance under each environmental factor in step S8 includes at least one of thermal deformation, electromagnetic noise, and vibration noise.

7. The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method as described in claim 1, characterized in that: The acoustic wave equation relating the sound pressure, position, and time of the three-dimensional structure in step S1 is as follows: ； Where v is the vibration velocity of the acoustic particle, and p is the sound pressure. Let K be the mass density of the fluid, and K be the bulk modulus constant. Let represent the Laplace operator, and t represent time.

8. The acoustic metamaterial ship cabin wall acoustic wave modulation interactive progressive design method as described in claim 7, characterized in that: In step S2, the output sound pressure level corresponding to the characteristic parameters of the dataset is obtained by directly discretizing the acoustic medium domain using the Helmholtz equation and establishing the sound field solution. The Helmholtz equation for plane sound waves in an ideal medium is: ； Under excitation response, the sound pressure at any point in the design domain of the solid structure is: ； in, Let be a solid, i be the imaginary part, ω be the angular frequency, h and g be column vectors, Y be the corner matrix, H and G be the global system matrices, A(ω) be the frequency response function, f(ω) be the harmonic excitation, and n be the normal vector.

Images