Forward and reverse fusion design and performance prediction method for mechanical metamaterial structure

Abstract

The invention belongs to the technical field of metamaterial structure design, and aims to solve the problem that metamaterial structure design depends on experience and manual adjustment. The invention provides a forward and reverse fusion design and performance prediction method for a mechanical metamaterial structure, and the method comprises the steps: designing N original metamaterial structures, and extracting N combinations of structure parameters and material parameters; carrying out modeling and simulation to obtain N characteristic parameter groups; constructing a dual-deep learning model comprising a front neural network and a rear neural network, and performing training based on the data; a combination of structure parameters and material parameters of the new metamaterial structure and performance parameters needing to be achieved are given, such as a low-frequency band gap value and a Poisson's ratio; inputting a new combination of structure parameters and material parameters into the double-deep learning model, and judging whether required performance is achieved or not based on characteristic parameters output by the neural network; and if so, outputting the reconstructed optimal original material structure by the post-neural network. According to the invention, the design complexity of the metamaterial can be reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of metamaterial structure design, and in particular relates to a forward and reverse fusion design and performance prediction method for a mechanical metamaterial structure. Background Art

[0002] Mechanical metamaterials are a type of material that exhibits unconventional mechanical properties and structural characteristics through carefully designed geometric structures. They have attracted widespread attention in the field of mechanical research. By carefully designing their microstructures and geometric shapes, metamaterials exhibit unique properties such as negative refractive index, negative Poisson's ratio, negative mass density and negative elastic modulus. The design and preparation of metamaterials have shown great potential in the fields of electromagnetic wave regulation, acoustic wave control, photonics, etc.

[0003] With the development of science and technology, vibration and noise reduction technology has become increasingly important in engineering applications. Traditional methods such as sound-absorbing materials and sound-insulating structures are inefficient or unsuitable in some scenarios. Therefore, metamaterial structure design as an emerging vibration and noise reduction technology has received widespread attention in recent years.

[0004] The existing technical solutions have the following defects in the metamaterial structure design:

[0005] 1. Reliance on experience and manual adjustments: Traditional metamaterial design usually relies on the designer’s experience and manual adjustments, which limits the accuracy and efficiency of the design. Designers need to spend a lot of time and energy to try different structures and parameter combinations, and it is difficult to ensure the optimality of the design.

[0006] 2. High computational cost: The design of metamaterials usually involves a large number of complex calculations and simulations. Changing one of the parameters requires remodeling, especially for multi-scale and multi-physics field coupling problems, the computational cost is huge. This leads to the problem of long design process and high demand for computing resources. There are many dimensional parameters and material parameters in the design process of metamaterials, which affect the final performance of the structure. In recent years, the forward design approach is usually adopted in the design of metamaterial structures. This design method first focuses on the structural design, and then studies its performance characteristics through numerical simulation, simulation and other methods. When designing metamaterial structures, a small number of key parameters are adjusted to study the impact of these parameters on performance. However, when facing specific engineering problems, multiple parameters need to be adjusted to achieve the optimal performance of the metamaterial. To fully understand the impact of each parameter on performance, a large number of calculations are required to determine feasible methods.

[0007] 3. Limited to specific applications: Traditional metamaterial designs are often optimized for specific application scenarios, which leads to insufficient versatility of the design. Once the application scenario or requirements change, the metamaterial structure needs to be redesigned. Summary of the invention

[0008] In order to solve at least one of the above technical problems existing in the prior art, the present invention provides a forward and reverse fusion design and performance prediction method for a mechanical metamaterial structure.

[0009] The present invention is implemented by the following technical solution: a method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure, comprising the following steps:

[0010] S1: Design N original metamaterial structures, extract N combinations of structural parameters and material parameters; perform modeling and simulation to obtain N characteristic parameter groups;

[0011] S2: Build a dual deep learning model including the front neural network and the back neural network;

[0012] S3: Take the combination of N structural parameters and material parameters as input layer data I, and N characteristic parameter groups as output layer data I, and perform training and testing of the front neural network;

[0013] S4: Take the M feature parameter groups that meet the design requirements among the N feature parameter groups as input layer data II, and take the combination of M structural parameters and material parameters that match the M feature parameter groups in S1 as output layer data II, and perform post-neural network training and testing;

[0014] S5: Given a combination of structural parameters and material parameters of a new metamaterial structure, and the design parameter indicators that need to be achieved;

[0015] S6: Input a new combination of structural parameters and material parameters into the dual deep learning model, and determine whether the required design indicators are met based on the characteristic parameter group output by the previous neural network;

[0016] If it is achieved, the post-neural network outputs the reconstructed optimal original material structure;

[0017] If not, the structural parameters and / or material parameters are manually adjusted until the required design indicators are achieved, and then the neural network outputs the reconstructed optimal original material structure.

[0018] Preferably, the characteristic parameter group is a characteristic frequency group related to the band gap width.

[0019] Preferably, the structural parameters of the metamaterial structure include the metamaterial matrix structure size parameters and the local resonance unit size parameters, and the material parameters include material density, Poisson's ratio and Young's modulus;

[0020] A combination of structural parameters and material parameters corresponds to a characteristic frequency group.

[0021] Preferably, a structural band diagram of the metamaterial structure is obtained through modeling and simulation based on COMSOL software;

[0022] Each characteristic frequency group includes K groups of characteristic frequencies, which are obtained from the structural band diagram; each group of characteristic frequencies includes a maximum value of a characteristic frequency and a minimum value of a characteristic frequency; each characteristic frequency group is used to characterize the performance of a metamaterial structure.

[0023] Preferably, the structural low-frequency characteristic index Z(n) is defined to judge the low-frequency characteristics of the metamaterial:

[0024] Define the full bandgap energy band range function F(n) and position function L(n):

[0025] F(n)=min(fr k (n+1))-max(fr k (n))(1.1)

[0026]

[0027] where fr k is the solution of the eigenfrequency calculated along the Brillouin zone of the structure for wave vector k, and n is the number of eigenvalue solutions calculated;

[0028]

[0029] The size of Z(n) is determined by the maximum and minimum values ​​in each set of characteristic frequency solutions.

[0030] Preferably, in step S4, satisfying the low-frequency band gap index means: Z(n)>0; the K groups of characteristic frequencies of the characteristic frequency group of the input layer data II are extracted by Z(n) satisfying the low-frequency band gap index.

[0031] Preferably, the characteristic frequency search reference value is 0 Hz, each characteristic frequency group includes 20 groups of characteristic frequencies, and the number of maximum and minimum values ​​of the corresponding characteristic frequencies is 40.

[0032] Preferably, in step S1, the combination of structural parameters and material parameters and the number N of characteristic frequency groups are greater than 1000, and the number M of characteristic frequency groups after screening in S4 is less than 1000.

[0033] Preferably, the characteristic parameter group includes but is not limited to a characteristic frequency group related to the band gap width, a characteristic sound absorption coefficient group related to the transmission loss, a characteristic resolution group related to the negative refractive index, and a characteristic energy absorption coefficient group related to the negative Poisson's ratio.

[0034] Preferably, the front neural network has a built-in forward optimization module, and the rear neural network has a built-in reverse prediction module; first, the performance of the metamaterial is optimized and designed based on the forward optimization module of the front neural network to explore the structural configuration with the best performance; then, the structural parameters required for the target performance are determined based on the reverse prediction module of the rear neural network.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The present invention can reduce the design complexity of metamaterials. The trained pre-neural network can quickly optimize the metamaterial structure without complicated modeling simulation and performance processing calculations. The inspired metamaterial structural parameters (structural parameters and material parameters) can be directly input into the pre-neural network to directly obtain the performance of the structure. If the structural performance is not ideal, the input parameters can be directly changed and the model parameters can be manually fine-tuned to quickly obtain the performance results of the fine-tuned model. After forward performance optimization and reverse structural optimization, the optimal structure after reconstruction is obtained. This technology is expected to bring significant applications in the fields of precision instruments and rotating machinery and promote the development of the industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 It is a method framework diagram of the present invention;

[0039] Figure 2 is a flow chart of the method of the present invention;

[0040] Figure 3 is a schematic diagram of the front neural network of the dual deep learning model of the present invention;

[0041] Figure 4 Schematic diagram of the post-neural network of the dual deep learning model of the present invention. DETAILED DESCRIPTION

[0042] In conjunction with the drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the conditions under which the present invention can be implemented. Therefore, they have no substantive technical significance. Any structural modification, change in proportional relationship or adjustment of size should fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention. It should be noted that in this specification, relational terms such as first and second are only used to distinguish one entity from other entities, and do not necessarily require or imply any actual relationship or order between these entities.

[0044] In the present application, the characteristic parameter group includes but is not limited to a characteristic frequency group related to the band gap width, a characteristic sound absorption coefficient group related to the transmission loss, a characteristic resolution group related to the negative refractive index, and a characteristic energy absorption coefficient group related to the negative Poisson's ratio.

[0045] like Figures 1 to 4 As shown, this embodiment takes acoustic materials as an example, a method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure includes the following steps:

[0046] S1: Design N original metamaterial structures, extract N combinations of structural parameters and material parameters; perform modeling and simulation to obtain N characteristic frequency groups;

[0047] S2: Build a dual deep learning model including the front neural network and the back neural network;

[0048] S3: using the combination of N structural parameters and material parameters as input layer data I and N characteristic frequency groups as output layer data I to train and test the neural network;

[0049] S4: Take the M characteristic frequency groups that meet the low-frequency band gap index among the N characteristic frequency groups as input layer data II, and take the combination of M structural parameters and material parameters that match the M characteristic frequency groups in S1 as output layer data II, and perform post-neural network training and testing;

[0050] S5: Given a combination of structural parameters and material parameters of a new metamaterial structure, and the low-frequency band gap value to be achieved;

[0051] S6: Input a new combination of structural parameters and material parameters into the dual deep learning model, and determine whether the desired low-frequency band gap value is achieved based on the characteristic frequency group output by the previous neural network;

[0052] If it is achieved, the post-neural network outputs the reconstructed optimal original material structure;

[0053] If not, the structural parameters and / or material parameters are manually adjusted until the desired low-frequency band gap value is reached, and then the neural network outputs the reconstructed optimal original material structure.

[0054] In this embodiment, the structural parameters of the metamaterial structure include the metamaterial matrix structure size parameters and the local resonance unit size parameters, and the material parameters include material density, Poisson's ratio and Young's modulus; in this embodiment, a combination of a structural parameter and a material parameter corresponds to a characteristic frequency group. Each characteristic frequency group includes K groups of characteristic frequencies, and the K groups of characteristic frequencies are obtained from the structural energy band diagram; each group of characteristic frequencies includes a maximum value of a characteristic frequency and a minimum value of a characteristic frequency; each characteristic frequency group is used to characterize the performance of a metamaterial structure. Each characteristic frequency group includes 20 groups of characteristic frequencies, and the number of corresponding maximum and minimum values ​​of characteristic frequencies is 40.

[0055] The specific steps are as follows: first, design the original metamaterial structure with initial inspiration, use software to build a model with specific parameters, and perform numerical analysis to extract characteristic data such as its structural parameters and material parameters, determine its initial performance, and then use the front neural network in the dual deep learning model to perform performance optimization prediction to obtain improved performance. This process is the forward design process. Predicting the structure from known performance is the reverse design process. The process first determines the performance of the designed metamaterial, and then uses the back neural network in the "dual deep learning model" to establish a structural database, reconstruct the model, establish the optimal structure, and complete the design plan.

[0056] In the early stage of the dual deep learning model, the comsol software parametric scanning was used to generate the data required for neural network training. The post-neural network in the dual deep learning model was used to perform structural optimization design, continuously train the deep learning model, extract the structural characteristics of the metamaterial, and finally determine the optimized structure of the metamaterial by combining simulation analysis and experimental verification. By combining forward and reverse design, combined with the dual deep learning model, the metamaterial structural parameters, material parameters and performance indicators are continuously integrated and optimized, thus providing a structural design and performance prediction design scheme for the metamaterial structure from the initial structure to the optimized structure.

[0057] The pre-neural network is used to quickly optimize the performance of metamaterial structures. The rapid optimization process is to model the metamaterial structure with initial inspiration, use the comsol software to perform parametric scanning simulation calculations of material parameters and structural parameters, and calculate the database for training the neural network. The trained pre-neural network can quickly optimize the metamaterial structure. Users do not need to perform complex and tedious modeling simulation and performance processing calculations. They can directly input the inspired metamaterial structure parameters (structural parameters and material parameters) into the pre-neural network to directly obtain the performance of the structure. If the structural performance is not ideal, directly change the input parameters and manually fine-tune the model parameters to quickly obtain the performance results of the fine-tuning of the model.

[0058] This technology has two goals. One is to quickly determine the complete bandgap width (CBW) of the designed dual-unit metamaterial monomer structure. That is, within this width range, any form of structural wave cannot be transmitted. The other is a reverse design. When faced with a specific engineering problem, the desired complete bandgap width is known and the required metamaterial structure is designed.

[0059] First, define the full bandgap energy band range function F(n) and position function L(n):

[0060] F(n)=min(fr k (n+1))-max(fr k (n))(1.1)

[0061]

[0062] where fr k is the solution of the eigenfrequency calculated along the Brillouin zone of the structure for wave vector k; n is the number of eigenvalue solutions calculated;

[0063]

[0064] In the calculation of the low-frequency band gap index Z(n), it is determined by calculating each group of eigenvalues ​​of the wave vector k along the Brillouin zone. The steps of obtaining the structural band diagram include: establishing the Brillouin zone of the metamaterial structure, setting periodic boundary conditions through COMSOL software, introducing the Bloch wave vector k, taking values ​​along the Brillouin zone boundary, calculating the reciprocal lattice, selecting the Wigner-Seitz primitive cell, and converting it into solving the characteristic frequency problem of the structural monomer; setting the characteristic frequency search benchmark value and the required number of characteristic frequencies, and drawing the structural band diagram, wherein the structural band diagram is used to view the band gap range of the structural monomer.

[0065] After multiple calculations in the early stage, when calculating more than 15 groups of characteristic frequencies, the frequency has reached more than 1 kHz. In order to simplify the calculation of different parameters in the initial neural network data training database, we set the number of characteristic frequency solutions to 20 to ensure that most results exceed 500 Hz. If Z(n) is still found to be less than 0 after calculating 20 groups of solutions, it is temporarily assumed that the metamaterial structure does not have low-frequency complete bandgap characteristics within the considered parameter range.

[0066] Analyze the index Z(n). When F(n) is greater than 0, there is a band gap range. When F(n) is less than 0, there is no band gap range. The larger the band gap range, the smaller the band gap position. Therefore, it is meaningless when Z(n) is less than 0. The larger the Z(n), the better the low-frequency band gap performance of this structure. The size of Z(n) is specifically determined by the maximum and minimum values ​​in each group of solutions. In the neural network training process, selecting 40 maximum and minimum values ​​in each group of 20 solutions as the output values ​​of the first neural network can simplify the difficulty of performance prediction. In this embodiment, the K groups of characteristic frequencies of the characteristic frequency group of the input layer data II are extracted by Z(n) that meets the low-frequency band gap index.

[0067] In this application, the algorithm framework consists of four units: data preprocessing unit, front neural network forward optimization unit, data postprocessing unit and back neural network structure reverse prediction unit. This method technology can not only quickly calculate and analyze the performance of the designed dual-unit metamaterial structure, but also predict the corresponding metamaterial structure according to the bandgap range and bandgap position required by the problem.

[0068] The structure of the database is a set of data including a one-to-one set of input layer data (metamaterial structure parameters and material parameters) and a set of output layer data (20 sets of characteristic frequency solutions representing the structure energy band diagram). In the process of neural network training, we selected more than 200,000 data from COMSOL simulation calculations to form more than 1,000 sets of data for model training, and took 10% as the test set. The low-frequency band gap index Z (n) was judged to be greater than 0 for screening. The screened data was less than 1,000 sets, but each set of data included a one-to-one set of input layer data (metamaterial structure parameters and material parameters) and a set of output layer data (20 sets of characteristic frequency solutions representing the structure energy band diagram).

[0069] Data preprocessing unit:

[0070] The data preprocessing unit mainly generates input layer data l and output layer data l. Through the software modeling and analysis of the metamaterial model, it is found that there are 11 necessary structural dimension parameters and 6 material property parameters. If the 17 input layer neuron variables are parametrically scanned at the same time, the required calculation amount is very large. Therefore, the control variables are carried out, and only one set of input layer structural dimension parameters are parametrically scanned each time, and the input layer database is created through the difference.

[0071] Considering the subsequent neural network optimization and adjustment of arbitrary material properties, in theory, the required specific density, Young's modulus and Poisson's ratio can be achieved by adjusting the proportions and structural design of the various components in the composite material. Composite materials are usually composed of two or more materials with different properties, and the material properties can be regulated through appropriate combinations.

[0072] In the comsol software, the output layer data uses OHz as the characteristic frequency search benchmark value to calculate 20 groups of characteristic frequencies. Since the structural energy band diagram needs to be calculated, the 20 groups of calculated characteristic frequencies are processed and the maximum and minimum values ​​of each group of data are taken as the output layer database.

[0073] In order to improve the training speed and generalization ability of the neural network, the input layer data is controlled between 0-100. First, the range of each group of variable parameters is determined, and then each group of parameters is linearly processed, as shown in Formula 1.4:

[0074] x′=k n x0 (1.4)

[0075] x0 is the original data, k n is the linear coefficient of the current parameter, the specific parameter range and k n The values ​​are shown in Table 1 below:

[0076] Table 1 Initial structural parameters in calculation

[0077]

[0078] After linearizing the 11 structural dimension parameters and 6 material property parameters in the table, shortening the dimensional gap, and the 40 (20 groups) characteristic frequencies of the output layer neurons (each group has a maximum and a minimum value), the deviation standardization method is used for normalization, and the data is normalized to [0, 1]. Formula 1.5 is as follows:

[0079]

[0080] Where x′ is the data after linear processing, min is the minimum value in the input layer, max is the maximum value, and x is the normalized data.

[0081] Pre-neural network forward optimization unit:

[0082] Back Propagation Neural Network is used to construct the front neural network. A BP neural network usually consists of three parts: input layer, hidden layer and output layer. Since there are 17 input parameters, the number of nodes in the input layer is 17. These 40 characteristic frequencies are used to calculate the bandgap of the structure in the later stage, and the number of nodes in the output layer is 40.

[0083] The number of hidden nodes in the BP neural network has a certain influence on the performance of the entire network, and is selected using the following empirical formula 1.6:

[0084]

[0085] Where b1 is the number of hidden layers, m1 is the number of input layers, n1 is the number of output layers, a1 is 1-10, in this model a1 is 10, m1 is 17, n1 is 40, b1 is rounded, and the number of hidden layers b1 is 18. Based on the above parameter selection, a 17-18-40 BP neural network is constructed for the design of the metamaterial structure. The input layer data I and the output layer data I are used to train the neural network.

[0086] Data post-processing unit:

[0087] The output layer neural network is the size parameters and material property parameters of the structure. When training the neural network in the early stage, the maximum and minimum frequencies of the output layer data I that meet the low-frequency band gap index are searched, and the output layer data I of the previous neural network is matched to establish a temporary database, including the structural parameters and size parameters of the metamaterial structure.

[0088] The maximum and minimum values ​​of the characteristic frequency of each structure are calculated using a neural network, and the range function F(n) and the position function L(n) are calculated according to formulas 1.1 and 1.2. The low-frequency band gap index Z(n) is calculated by formula 1.3, and the index Z(n) is associated with the corresponding structure to construct a new database: output layer data II, which is used to generate neural network output layer data. In step S1, the combination of structural parameters and material parameters, the number of characteristic frequency groups N is greater than 1000, and the number of characteristic frequency groups M after screening in S4 is less than 1000.

[0089] After the predicted structural parameters and material property parameters are normalized, they need to be denormalized to calculate the original dimensional data x'. According to formula 1.7, in Table 1, the predicted data x' is divided by the linear coefficient k of the current parameter. n to achieve this denormalization process.

[0090]

[0091] Post-neural network structure reverse prediction unit:

[0092] The data sets used in this unit are all from the previous neural network forward optimization unit. First, the output layer data of the previous neural network is screened to retain the data with low-frequency band gap, that is, the indicator Z(n)>0. From formula 1.3, we know that Z(n) is determined by F(n) and L(n). Therefore, when it is necessary to predict a metamaterial structure with a specific band gap performance, F(n) and L(n) are used as the input parameters of the designer, where F(n) and L(n) are determined by 20 sets of characteristic frequencies. Through the algorithm flow chart, the output layer data I of the previous neural network is searched to create 20 sets of characteristic frequencies. In the later structure prediction, 20 sets of characteristic frequency parameters that meet the conditions are calculated through calculation.

[0093] The method was verified by the designed metamaterial structure, and the number of neurons in the output layer was 17. According to formula 1.6, a1 was taken as 10, and the number of neurons in the hidden layer of the neural network was calculated to be 18. A 40-18-17 neural network was established.

[0094] In order to describe the propagation characteristics of the structure, the frequency transmission loss function Tr is defined, as shown in Formula 1.8, where Pout is the output measurement point displacement and Pin is the excitation measurement point displacement.

[0095]

[0096] Advantages of the present invention:

[0097] A forward and reverse fusion design idea is proposed for the optimization of mechanical metamaterial structures. By combining the forward optimization module of the front neural network and the reverse prediction module of the back neural network, the front neural network is first used to optimize the performance of the metamaterial and quickly explore the structural configuration with the best performance; then, the back neural network reverse prediction module is used to accurately determine the structural parameters required under the target performance. This forward and reverse fusion method not only speeds up the design and optimization process of metamaterials, but also improves the accuracy of the prediction of structural properties, achieving efficient transformation from design to application.

[0098] Disadvantages of alternatives:

[0099] 1. Traditional optimization methods: Metamaterial structures are optimized using traditional optimization algorithms, such as genetic algorithms, particle swarm algorithms, etc. However, these methods may require more computing resources and time, and may not accurately predict the performance of the structure.

[0100] 2. Other machine learning models: Use other machine learning models besides neural networks, such as support vector machines, random forests, etc., to predict the performance of metamaterial structures. However, these models may require more data and feature engineering, and their accuracy may not be as good as this neural network.

[0101] 3. Manual design method: Through experience and manual design, try to design a metamaterial structure suitable for a specific application. This method may be limited by the designer's experience and knowledge and is not efficient enough.

[0102] While these alternatives exist, the neural network algorithms and neural network structures of the present technology are a more preferred choice because they are able to optimize structures more quickly and accurately predict performance.

[0103] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for forward and reverse fusion design and performance prediction of mechanical metamaterial structure, characterized in that: The following steps are involved: S1: Design N original metamaterial structures, extract N combinations of structural parameters and material parameters; perform modeling and simulation to obtain N characteristic parameter groups; S2: Build a dual deep learning model including the front neural network and the back neural network; S3: Take the combination of N structural parameters and material parameters as input layer data I, and N characteristic parameter groups as output layer data I, and perform training and testing of the front neural network; S4: Take the M feature parameter groups that meet the design requirements among the N feature parameter groups as input layer data II, and take the combination of M structural parameters and material parameters that match the M feature parameter groups in S1 as output layer data II, and perform post-neural network training and testing; S5: Given a combination of structural parameters and material parameters of a new metamaterial structure, and the design parameter indicators that need to be achieved; S6: Input a new combination of structural parameters and material parameters into the dual deep learning model, and determine whether the required design indicators are met based on the characteristic parameter group output by the previous neural network; If it is achieved, the post-neural network outputs the reconstructed optimal original material structure; If not, the structural parameters and / or material parameters are manually adjusted until the required design indicators are achieved, and then the neural network outputs the reconstructed optimal original material structure.

2. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 1, characterized in that: The characteristic parameter group is a characteristic frequency group related to the band gap width.

3. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 2, characterized in that: The structural parameters of the metamaterial structure include the size parameters of the metamaterial matrix structure and the size parameters of the local resonance unit, and the material parameters include material density, Poisson's ratio and Young's modulus; A combination of structural parameters and material parameters corresponds to a characteristic frequency group.

4. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 2, characterized in that: Through modeling and simulation, the structural band diagram of the metamaterial structure is obtained to characterize its functional parameters; Each characteristic frequency group includes K groups of characteristic frequencies, which are obtained from the structural band diagram; each group of characteristic frequencies includes a maximum value of a characteristic frequency and a minimum value of a characteristic frequency; each characteristic frequency group is used to characterize the performance of a metamaterial structure.

5. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 4, characterized in that: Define the structural low-frequency characteristic index Z(n) to evaluate the low-frequency characteristics of metamaterials: Define the full bandgap energy band range function F(n) and position function L(n): F(n)=min(fr k (n+1))-max(fr k (n) (1.1) where fr k is the solution of the eigenfrequency calculated along the Brillouin zone of the structure for wave vector k, and n is the number of eigenvalue solutions calculated; The size of Z(n) is determined by the maximum and minimum values ​​in each set of characteristic frequency solutions.

6. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 5, characterized in that: In step S4, satisfying the low-frequency band gap index means: Z(n)>0; the K groups of characteristic frequencies of the characteristic frequency group of the input layer data II are extracted by Z(n) satisfying the low-frequency band gap index.

7. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 4, characterized in that: The characteristic frequency search reference value is 0 Hz, each characteristic frequency group includes 20 groups of characteristic frequencies, and the number of maximum and minimum values ​​of the corresponding characteristic frequencies is 40.

8. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 2, characterized in that: In step S1, the combination of structural parameters and material parameters and the number N of characteristic frequency groups are greater than 1000, and the number M of characteristic frequency groups after screening in S4 is less than 1000.

9. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 2, characterized in that: The characteristic parameter groups include but are not limited to a characteristic frequency group related to the band gap width, a characteristic sound absorption coefficient group related to the transmission loss, a characteristic resolution group related to the negative refractive index, and a characteristic energy absorption coefficient group related to the negative Poisson's ratio.

10. The method for forward and reverse fusion design and performance prediction of a mechanical metamaterial structure according to claim 1, characterized in that: The front neural network has a built-in forward optimization module, and the back neural network has a built-in reverse prediction module. First, the performance of the metamaterial is optimized and designed based on the forward optimization module of the front neural network to explore the structural configuration with the best performance. Then, the reverse prediction module of the back neural network is used to determine the structural parameters required for the target performance.

Images