A design method for energy-absorbing metamaterials based on diffusion model and parameter optimization

Abstract

The present invention belongs to the field of energy-absorbing metamaterial design, and specifically relates to a method for designing energy-absorbing metamaterials based on a diffusion model and parameter optimization. The present invention proposes a method for designing energy-absorbing metamaterials based on a diffusion model and parameter optimization, which generates metamaterials given a required stress-strain curve. The diffusion model learns the conditional distribution of metamaterials with given properties, and realizes a one-to-many mapping from properties to geometry. In this method, the diffusion model can generate various metamaterials that meet the given material properties with high accuracy and speed. In order to improve the accuracy and generalization performance of the diffusion model, a new method is proposed to parameterize the generated metamaterial, and at the same time use a genetic algorithm to fine-tune it, so that the optimized mechanical properties of the metamaterial are consistent with the target stress-strain curve. In summary, the present invention has the advantages of fast design speed and high accuracy, and provides a new solution path for realizing more intelligent and customized energy-absorbing metamaterial design.

Description

Technical Field

[0001] The invention belongs to the field of energy absorbing metamaterial design, and in particular relates to an energy absorbing metamaterial design method based on a diffusion model and parameter optimization. Background Art

[0002] Energy-absorbing materials and structures are widely used in aerospace, transportation, human protection and other fields due to their excellent impact resistance. Anti-collision designers consider multiple criteria when designing ideal energy-absorbing metamaterials, including not only high energy absorption performance, but also stability and safety. With the desired energy absorption curve as the design goal and the basic material determined, the remaining design variable is the distribution of the material inside the structure. Material distribution plays a vital role in the formation of the load-displacement response of the structure and the optimization of the energy absorption characteristics of the structure. By carefully manipulating the material distribution, the structural behavior can be customized and the energy absorption performance can be enhanced to achieve specific design goals and meet the desired performance standards. This approach can effectively control the energy absorption characteristics of the structure, allowing designers to optimize the response of the structure to different loading conditions and requirements.

[0003] The forward design method of energy-absorbing structures refers to adjusting geometric parameters and material types through empirical or bionic configuration to achieve different energy absorption effects. Classic energy-absorbing structures include sandwich structures, plate structures, honeycomb structures, foam structures, etc., which have been widely studied in previous studies. The performance of these structures can be adjusted by modifying geometric parameters. In order to improve the energy absorption capacity of the structure, various design schemes have been proposed to improve the classical structure. For example, composite materials are used for reinforcement, and the fold lines of origami structures are used as plastic hinges to improve energy absorption. Bionic design is another method to obtain structures with energy-absorbing properties by imitating the internal structure of organisms. For example, inspired by the Haversian bone structure, tubular nested geometric structures are designed to develop efficient energy-absorbing structures.

[0004] At present, the reverse design method of energy-absorbing structures mainly refers to the design method based on topology optimization

[13] . This method directly meets the design objectives and does not require strict material requirements, which simplifies the design process and improves design efficiency. This is currently a mature field with various methods, including SIMP method, level set method, evolutionary method and moving deformable component method. These methods require rigorous sensitivity derivation and fast convergence rate based on finite element analysis (FEA) to solve the independent variables. The topology optimization of energy-absorbing structures needs to consider the large deformation of the structure and the nonlinear constitutive relationship of the material. Existing research on topology optimization of mechanical metamaterials mainly focuses on objectives such as minimum compliance, negative Poisson's ratio and multi-scale optimization. However, these studies are usually conducted under small deformation and linear elastic assumptions, which makes it challenging to derive sensitivity under large deformation and material nonlinearity. Therefore, gradient-based topology optimization methods currently face an insurmountable bottleneck in the problem of energy absorption under large deformation and elastoplastic behavior. At the same time, design methods based on intelligent algorithms have shown good performance in solving complex multi-objective, non-convex or time-varying optimization problems that are challenging for sensitivity-based methods. The highly flexible model representation of intelligent algorithms improves convergence. However, due to the lack of sensitivity guidance, the optimization efficiency of intelligent algorithms is lower than that of gradient-based algorithms, and this problem becomes more obvious when dealing with significant structural deformations.

[0005] The present invention proposes a design method for energy-absorbing metamaterials based on a diffusion model and parameter optimization. The energy absorption curve is used as a generation condition to guide the conditional diffusion model to generate metamaterials with target mechanical behavior. At the same time, in order to further improve the generalization ability of the model, the parameters of the generated metamaterials are optimized using a genetic algorithm, and finally a metamaterial with target mechanical behavior is obtained. The diffusion model not only makes the design of energy-absorbing metamaterials more flexible and efficient, but also, by introducing parameter optimization, we can fine-tune the generated metamaterials to meet more precise and specific mechanical requirements. Summary of the invention

[0006] The purpose of the present invention is to propose a design method for energy-absorbing metamaterials based on a diffusion model and parameter optimization, aiming to solve the problem of large elastic-plastic deformation that is difficult to solve with traditional gradient algorithms or intelligent algorithms, overcome the efficiency and precision problems in the design of energy-absorbing metamaterials, simplify the design process of energy-absorbing mechanical metamaterials, and provide a new solution path for realizing more intelligent and customized metamaterial design.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A design method for energy-absorbing metamaterials based on diffusion model and parameter optimization, the steps are as follows:

[0009] Step 1.1: Generate as many metamaterials as possible using a topology optimization method based on homogenization solid isotropic material penalty, identify boundaries using a density threshold method, and effectively reduce the detected boundary data points using the Ramer-Douglas-Peucker (RDP) algorithm to obtain a clear and accurate geometric representation for finite element simulation;

[0010] Step 1.2: The metamaterial obtained by topology optimization is simulated using finite element software. The metamaterial is placed between two rigid plates. A vertical displacement is set along the y direction to simulate the compression process, compressing the structure in a quasi-static manner. In order to avoid mutual penetration, considering the interaction between the plate and the structure and the self-interaction within the structure, 316L stainless steel is selected as the reference material, and the Johnson Cook constitutive model is used to simulate the plastic constitutive relationship of the material to obtain the stress-strain curve when the structure is compressed in the y direction.

[0011] Step 1.3: Use principal component analysis to reduce the dimension of the stress-strain curve obtained by simulation, reduce the complexity of the data by finding the principal component, and further normalize the reduced-dimensional data to eliminate the problem of the model being biased towards a certain feature due to differences between features. Finally, the metamaterial and its reduced-dimensional energy absorption curve are used as training data, of which 90% of the generated data is used for training and the remaining 10% is used for verification.

[0012] Step 2.1: Construct a diffusion model. The neural network architecture uses the U-Net network, initialize the network parameters, and use the metamaterial and its energy absorption curve as training data for iterative training in the diffusion model.

[0013] Step 2.2: For a trained diffusion model, given a target stress-strain curve, a family of metamaterials with target mechanical properties can be quickly generated based on the complex nonlinear mapping relationship learned by the diffusion model;

[0014] Step 3.1: Metamaterial parameterization. According to the set partitioning theorem, the generated metamaterial is parameterized into two sets: one represents the void area and the other represents the solid area. The void area is subjected to polygon offset operation to create a polygon with a shape similar to the original polygon.

[0015] Step 3.2: Genetic algorithm is used to solve the inverse design problem, alternating between metamaterial generation and evaluation phases, and using structural encoding as genes affected by evolution, offsetting the voids, and generating the initial population. Then, two metamaterials are randomly selected as parents to derive daughter metamaterials, which have similar feature combinations as the parents. The mutation of the new generation of metamaterials introduces an additional source of structural diversity, where the newly added metamaterials are randomly perturbed by the existing metamaterials. Next, the stress response of each structure is calculated by finite element simulation and the structures are ranked according to the selected fitness function. To accelerate convergence, we adopt the elite retention strategy. Finally, we filter the ranked metamaterials for the next round of iterations. Once the metamaterial fitness reaches the threshold, the cycle is skipped and the best metamaterial is selected from the evaluation population. These metamaterials are identified as possible candidate structures with our desired mechanical properties.

[0016] The beneficial effects of the present invention are:

[0017] 1. This invention proposes a design method for energy-absorbing metamaterials based on diffusion model and parameter optimization, which generates metamaterials given the required stress-strain curve. The diffusion model learns the conditional distribution of metamaterials with given properties, realizing a one-to-many mapping from properties to geometry. In this method, the diffusion model can generate various metamaterials that meet the given material properties with high accuracy and speed.

[0018] 2. In order to improve the accuracy and generalization performance of the diffusion model, a new method is proposed to parameterize the generated metamaterial and fine-tune it using a genetic algorithm. The mechanical properties of the optimized metamaterial are consistent with the target stress-strain curve.

[0019] 3. In summary, the present invention has the advantages of fast design speed and high precision, and provides a new solution path for realizing more intelligent and customized energy-absorbing metamaterial design. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is the algorithm flow chart of the present invention;

[0021] Figure 2 A schematic diagram of the topological structure of the energy-absorbing metamaterial of the present invention;

[0022] Figure 3 It is a schematic diagram of finite element simulation of the present invention;

[0023] FIG4( a ) is a first example of a metamaterial and its stress-strain curve;

[0024] FIG4( b ) is a second example of a metamaterial and its stress-strain curve;

[0025] Figure 4(c) is a third example of a metamaterial and its stress-strain curve;

[0026] FIG4( d ) is a fourth example of a metamaterial and its stress-strain curve;

[0027] Figure 5 It is the principal component cumulative distribution diagram of the present invention;

[0028] Figure 6 is a schematic diagram of the diffusion model of the present invention;

[0029] Figure 7(a) shows the test set curve generation condition 1;

[0030] Figure 7(b) shows the test set curve generation condition 2;

[0031] Figure 7(c) shows the test set curve generation condition 3;

[0032] Figure 7(d) shows the test set curve generation condition 4;

[0033] Figure 7(e) shows the test set curve generation condition 5;

[0034] Figure 8(a) is a schematic diagram of the interpolation curve;

[0035] Figure 8(b) shows the generation of interpolation curve 1;

[0036] Figure 8(c) shows the generation of interpolation curve 2;

[0037] Figure 8(d) shows the generation of interpolation curve three;

[0038] Figure 8(e) is the generation of interpolation curve 4;

[0039] Figure 9(a) is a schematic diagram of the interpolation curve;

[0040] Figure 9(b) shows the generation of interpolation curve 1;

[0041] Figure 9(c) shows the generation of interpolation curve 2;

[0042] Figure 9(d) shows the generation of interpolation curve three;

[0043] Figure 9(e) shows the generation of interpolation curve 4;

[0044] Fig.10 A schematic diagram of the metamaterial offset operation of the present invention;

[0045] Fig.11 This is a schematic diagram of parameter optimization based on genetic algorithm of the present invention;

[0046] Figure 12(a) shows the optimization iteration history;

[0047] Figure 12(b) shows the optimized stress-strain curve;

[0048] Figure 12(c) shows the optimized metamaterial. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the technical solution of the present invention is described in detail below in combination with theoretical formulas, drawings and specific implementation cases. The technical solution adopted by the present invention is described below by examples.

[0050] Figure 1 The overall flow chart of the method of the present invention is divided into three parts: data preparation, diffusion model training and generation, and parameter optimization. In the first step, various metamaterials are obtained by using a topological optimization method based on homogenization, and a CAD model is obtained by using a density threshold method. Then, a finite element simulation is performed to obtain a stress-strain curve under uniaxial compression. The stress-strain curve is reduced in dimension by using the principal component analysis method, and different eigenvalues ​​are mapped to a uniform scale by using a standard normalization method. The metamaterial and stress-strain data processed by dimensionality reduction are used as training data and test data of the diffusion model, respectively. In the second step, the target curve is brought into the trained diffusion model as a condition for generating the corresponding metamaterial. In the third step, the generated metamaterial is parameterized and optimized using a genetic algorithm. Once the fitness of the metamaterial reaches the set threshold, the iteration is terminated, and the best metamaterial is selected from the evaluation population.

[0051] Step 1.1: The mathematical expression of the topology optimization method based on homogenization is as follows:

[0052]

[0053] st:KU=F,k,l=1,…,d

[0054] :

[0055] ∶0≤ρ e ≤1,e=1,…,N

[0056] Where K is the overall stiffness matrix, U and F are displacement vectors and force vectors, d is the design variable dimension, and v is the e is the unit volume, θ is the upper bound of the volume fraction, and the design goal is a function of the homogenized elasticity tensor.

[0057] We perform topology optimization and systematically generate 40,000 metamaterials with different topologies by changing the initial distribution, volume constraints, etc. Both the horizontal and vertical dimensions consist of 128 pixels. To ensure clear boundaries, we use a significant penalty factor such as p ≥ 3, and then binarize the generated metamaterials. Figure 2 The generated metamaterial is intuitively demonstrated.

[0058] Since the metamaterial obtained by topology optimization is represented by grayscale pixels and has jagged edges, it cannot be directly used for finite element simulation. Therefore, the density threshold method is used for boundary identification, and the Ramer-Douglas-Peucker (RDP) algorithm is used to effectively reduce the detected boundary data points to obtain a clear and accurate geometric representation for finite element simulation.

[0059] Step 1.2: Use the commercial finite element software Abaqus to simulate the metamaterial obtained by topology optimization. Figure 3 The finite element simulation boundary conditions are shown. The metamaterial is 10 mm × 100 mm in size and is placed between two rigid plates. A vertical displacement is set along the y direction to simulate the compression process, compressing the structure in a quasi-static manner. In order to avoid mutual penetration, considering the interaction between the plate and the structure and the self-interaction within the structure, 316L stainless steel is selected as the reference material, and the Johnson Cook constitutive model is used to simulate the plastic constitutive relationship of the material, which is expressed as:

[0060]

[0061] The above formula can be regarded as the product of three factor parts, and the three parts represent three mechanical properties respectively. p ) n represents the work hardening effect, represents the strain rate effect, 1-(T * ) m Represents the temperature softening effect. The simulation is only performed at room temperature, so the effects of strain rate, temperature and damage are not considered. The corresponding 316L stainless steel material parameters in this simulation are shown in Table 1.

[0062] Table 1

[0063]

[0064] The solid part of the material was discretized using CPE6M elements under plane strain conditions. R2D2 meshes were used for each rigid plate. The global seed size was set to 0.15 mm to ensure appropriate resolution of the entire simulation domain. Considering that the material exhibits characteristics similar to metal-like elastic-plastic materials, the main mechanism leading to initial structural damage is plastic yielding rather than elastic buckling, so the simulation does not consider the nonlinear buckling effect. Finally, the stress-strain curve of the metamaterial when compressed in the y direction is obtained through post-processing.

[0065] Step 1.3: Perform data analysis on metamaterials and stress-strain through data visualization. In Figure 4, we present four different metamaterials and their stress-strain curves, revealing the significant impact of changes in the microstructure in the metamaterials on the mechanical behavior. By observing these curves, we can see that there are significant differences in the stress-strain behavior between different metamaterials. Further comparing Figure 4(c) and (d), it is worth noting that even similar metamaterials can show large differences in mechanical behavior, indicating that even small changes in metamaterials or subtle adjustments in metamaterial parameters can have a profound impact on mechanical properties. Despite these significant differences, the stress-strain curves in Figure 4(c) and (d) still have a common basic trend. This shows that metamaterials with similar structures have the same trend in their stress-strain curves, providing valuable insights for subsequent metamaterial design.

[0066] Step 1.4: The energy absorption curve is processed by principal component analysis to reduce the dimensionality, and the complexity of the data is reduced by finding the principal components. As shown in Figure 5, it is obvious that the first 15 principal components have the ability to capture 99.97% of the data variance. This shows that even if the dimension is significantly reduced, most of the key information inherent in the data is still retained. Therefore, the original stress data is projected into a space with only 15 dimensions. Further, in order to eliminate the problem of the model being biased towards a certain feature due to differences between features, the data after dimensionality reduction is normalized, which is defined as follows:

[0067]

[0068] Where Y i is the i-th value of the strain-stress curve, Y i,max and Y i,min are the i-th maximum and minimum values ​​in all strain-stress curves, respectively.

[0069] Step 1.5: Since the generated metamaterial has mirror symmetry in both horizontal and vertical directions, in order to make full use of this mirror symmetry and reduce unnecessary redundant data, we selected a quarter of the original images, specifically 64x64 pixel metamaterials, for model training. Finally, we divided 40,000 metamaterials and their stress-strain curves into training and validation sets, of which 90% of the generated data was used for training and the remaining 10% was used for validation.

[0070] Step 2.1: Construct a diffusion model. The neural network architecture uses a U-Net network with attention mechanism and group normalization, as shown in Figure 6As shown in Figure 1, U-Net consists of three main parts: encoder, decoder, and skip connection layer. The encoder compresses the input image into a low-dimensional latent space through a convolutional layer. The decoder accepts the encoded input and maps it to the original resolution. The skip connection layer connects the encoder and decoder, forming a bottleneck in the latent space, emphasizing key features, and promoting bidirectional information flow. Our model mapping maps the metamaterial from 64×64 dimensions to 4×4 dimensions. At the same time, a cross-attention mechanism is used to integrate the energy absorption curve, which is used to control the deep generation model to obtain the corresponding metamaterial according to the target energy absorption curve.

[0071] Initialize the network parameters, use the metamaterial and its energy absorption curve as training data for iterative training in the diffusion model. The selected diffusion model has 1.97 million trainable parameters, the loss function uses MSE loss, Adam is used as the optimizer, and the learning rate is 10 -4 , realize learning rate scheduling, the learning rate decays by 0.98 every 50 epochs, a total of 2000 epochs are performed, the loss function gradually decreases, and finally converges to a lower value;

[0072] Step 2.2: Select 5 stress-strain curves in the test set as generation conditions and put them into the trained diffusion model to generate metamaterials. Then, import the generated metamaterials into Abaqus for quasi-static loading to verify the consistency of the generated metamaterials with the target curves. As a metric to quantify the difference:

[0073]

[0074] Among them, Y g Indicates the generated stress-strain curve, Y t is the target stress-strain curve.

[0075] As shown in Figure 7(a)-(e), each column represents a different working condition. The first row is the target and generated curves, the second row is the metamaterial corresponding to the target curve, and the third row is the generated metamaterial. It can be observed that The result is always below 4%, indicating that the generated metamaterial curve has a high consistency with the target curve. The results show that the proposed diffusion model has high accuracy and reliability in generating metamaterials with ideal mechanical properties. Figure 7(e) shows the difference between the generated curve and the target curve. There are small, yet significant differences between metamaterials, which highlights the ability of the proposed diffusion model to generate metamaterials with significant differences under the target curve. This unique feature increases the flexibility of metamaterial design, enables the generation of metamaterials beyond the training set, and facilitates the discovery of metamaterials that meet the designed mechanical properties.

[0076] Step 2.3: To further evaluate the generative ability of the proposed generative model, we select two stress-strain curves and obtain a family of target curves by linear interpolation between the two curves. Figure 8 shows the first set of interpolation generation results, where the solid line in Figure 8(a) represents the original curve and the dashed line represents the interpolated curve. In Figure 8(b)-(e), the first row shows the target interpolation curve and the generated curve, while the second row shows the generated metamaterial. As can be seen from the figure, the error between the target curve and the generated curve is about 5%, indicating that the generative model exhibits excellent generation ability and can generate metamaterials that match the target curve. It is worth noting that as the target curve changes, the generative model maintains continuity and smoothness when generating metamaterials. There is a gradual transition between adjacent metamaterials, rather than abrupt or incoherent changes, which shows the reliability and stability of the proposed generative model.

[0077] Figure 9 shows the second set of interpolation generation results, which shows that the proposed generative model has good generation ability and can generate metamaterials that match the target curve. However, we also noticed in Figure 9(c) that the generation ability of the generative model has decreased, which indicates that its performance may decrease or produce inaccurate results when faced with new data or data that is significantly different from the training data. It should be noted that although its The error is larger, but the trend of the resulting curve is still similar.

[0078] Step 3.1: In order to ensure the similarity with the initial metamaterial during optimization, a heuristic metamaterial parameterization method is proposed based on the set partition theorem. According to the set partition theorem: a set can be partitioned into multiple non-intersecting subsets, each of which has no intersection, but their union is equal to the original set. The generated metamaterial is parameterized into two sets: one represents the void area and the other represents the solid area. The polygon offset operation is performed on the void area to create a metamaterial with a shape similar to the starting metamaterial, such as Fig.10 shown.

[0079] Step 3.2: Utilize a genetic algorithm to solve the inverse design problem, alternating between metamaterial generation and evaluation phases, and using the structural encoding as the gene subject to evolution. Fig.11As shown, the voids are first offset to generate an initial population. Then, two metamaterials are randomly selected as parent metamaterials from which sub-metamaterials are derived, with similar feature combinations as the parent metamaterials. The mutation of the new generation of metamaterials introduces an additional source of structural diversity, where the newly added metamaterials are randomly perturbed by the existing metamaterials. Next, the stress response of each structure is calculated through finite element simulation and the structures are ranked according to the selected fitness function. To accelerate fusion, an elite retention strategy is adopted. Finally, we filter the ranked metamaterials for the next round of iterations. Once the metamaterial fitness reaches the threshold, the loop is exited and the best metamaterial is selected from the evaluation population. These metamaterials are identified as possible candidate structures with our desired mechanical properties.

[0080] Figure 12 demonstrates the iterative optimization process. As can be seen from Figure 12(a), as the optimization iteration proceeds, the optimal error and the average error decrease rapidly. After only ten generations of iterations, the error of the optimal individual is can be reduced to 6%. Figure 12(b) and (c) show the optimized curves and the corresponding metamaterials. This result verifies the significant effect of the optimization algorithm in reducing errors, highlights the efficiency and effectiveness of the optimization algorithm proposed in this paper, and shows that the algorithm is expected to make up for the problem of generalization performance of the generative model, pointing out a new direction for improving the generalization ability of the generative model.

Claims

1. A design method for energy-absorbing metamaterials based on diffusion model and parameter optimization, characterized in that: Here are the steps: Step 1.1: Generate the metamaterial using the topology optimization method based on the homogenization solid isotropic material penalty, identify the boundary using the density threshold method, effectively reduce the detected boundary data points, and obtain the geometric representation for finite element simulation; Step 1.2: Use finite element software to simulate the metamaterial obtained by topology optimization. The metamaterial is placed between two rigid plates. A vertical displacement is set along the y direction to simulate the compression process, compressing the structure in a quasi-static manner; Step 1.3: Use principal component analysis to reduce the dimension of the stress-strain curve obtained by simulation, reduce the complexity of the data by finding the principal component, and further normalize the reduced-dimensional data to eliminate the problem of the model being biased towards a certain feature due to differences between features; finally, the metamaterial and its reduced-dimensional energy absorption curve are used as training data; Step 2.1: Construct a diffusion model. The neural network architecture uses the U-Net network, initialize the network parameters, and use the metamaterial and its energy absorption curve as training data for iterative training in the diffusion model. Step 2.2: For a trained diffusion model, given a target stress-strain curve, a set of metamaterials with target mechanical properties can be quickly generated based on the complex nonlinear mapping relationship learned by the diffusion model; Step 3.1: Metamaterial parameterization: According to the set partition theorem, the generated metamaterial is parameterized into two sets: one represents the void area and the other represents the solid area. The void area is subjected to a polygon offset operation to create a polygon with a shape similar to the original polygon. Step 3.2: Using a genetic algorithm to solve the inverse design problem, alternating between metamaterial generation and evaluation phases, and using the structural encoding as the gene affected by evolution to offset the voids and generate the initial population; Then, two metamaterials are randomly selected as parent metamaterials for deriving sub-metamaterials, and the derived sub-metamaterials have a similar feature combination to the parent metamaterials; Next, the stress response of each structure is calculated through finite element simulation, and the structures are ranked according to the selected fitness function; Finally, the ranked metamaterials are filtered for the next round of iterations; once the metamaterial fitness reaches a threshold, the loop is skipped and the best metamaterial is selected from the evaluation population.

2. The method for designing an energy-absorbing metamaterial based on a diffusion model and parameter optimization according to claim 1, characterized in that: In the step 1.2, 316L stainless steel is selected as a reference material, and the Johnson Cook constitutive model is used to simulate the plastic constitutive relationship of the material to obtain the stress-strain curve when the structure is compressed in the y direction.

3. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 1 or 2, characterized in that: In step 1.3, 90% of the training data is used for training and 10% is used for validation.

4. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 1 or 2, characterized in that: In step 3.2, in order to accelerate fusion, an elite retention strategy is adopted.

5. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 3, characterized in that: In step 3.2, in order to accelerate fusion, an elite retention strategy is adopted.

6. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 1, 2 or 5, characterized in that: In step 1.1, the Ramer-Douglas-Peucker algorithm is used to effectively reduce the detected boundary data points.

7. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 3, characterized in that: In step 1.1, the Ramer-Douglas-Peucker algorithm is used to effectively reduce the detected boundary data points.

8. The method for designing energy-absorbing metamaterials based on diffusion model and parameter optimization according to claim 4, characterized in that: In step 1.1, the Ramer-Douglas-Peucker algorithm is used to effectively reduce the detected boundary data points.

Images