An artificial intelligence-based electromagnetic performance design method and system for laminated structural materials

Abstract

The present invention discloses an artificial intelligence-based electromagnetic performance design method and system for laminated structural materials, which relates to the field of artificial intelligence technology. The method includes obtaining basic electromagnetic parameters of ply units through experiments, encoding ply units with different electromagnetic absorption properties, and generating random codes for ply configurations based on time dependence. Using electromagnetic simulation software, a simulation database of random codes for ply configurations and corresponding electromagnetic absorption properties is established. The simulation database is used to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model. The prediction model is used to predict the electromagnetic absorption performance of candidate ply configurations, determine whether the electromagnetic absorption performance of the candidate ply configuration is optimal, and if so, determine that the candidate ply configuration is the optimal ply configuration. This method solves the problems faced by electromagnetic absorption materials, such as the difficulty and high cost of searching a large parameter space composed of ply units and material types.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based electromagnetic performance design method and system for laminated structural materials. Background Art

[0002] Laminated composite materials are made by stacking and bonding multiple layers of different materials, or thin sheets (plies) of the same material in different orientations. Their core characteristic is the optimized combination of material properties achieved through layered design. They are widely used in aerospace, marine, and electronic stealth applications.

[0003] The combination of ply units and material types in laminated structures creates an extremely large parameter space, making traditional trial-and-error search methods extremely inefficient. Existing design methods rely on expert knowledge to select ply configurations, which is highly subjective and difficult to systematize. They rely on group-by-group verification through experiments and electromagnetic simulations, resulting in lengthy simulation runs and prohibitively high costs for exploring a wide range of parameters. Summary of the Invention

[0004] The present invention aims to address at least one of the technical problems existing in the prior art. To this end, it proposes an artificial intelligence-based method and system for designing the electromagnetic properties of laminated structural materials. This method addresses the difficulties and high costs associated with searching a large parameter space composed of ply units and material types for electromagnetic absorption materials.

[0005] According to an embodiment of the first aspect of the present invention, an artificial intelligence-based method for designing electromagnetic properties of a laminated structural material includes:

[0006] S100, obtaining basic electromagnetic parameters of ply units through experiments, encoding ply units with different electromagnetic absorption properties, and generating random codes of ply configurations based on time dependency;

[0007] S200, establishing a simulation database of random codes of ply configurations and corresponding electromagnetic absorption performances using electromagnetic simulation software;

[0008] S300, using the simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model;

[0009] S400: Obtain electromagnetic absorption performance of a candidate ply configuration through prediction using the prediction model, determine whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determine that the candidate ply configuration is an optimal ply configuration.

[0010] According to some embodiments of the present invention, in S300, using the simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model includes:

[0011] The encoded ply configuration is input, and the output is the interpolation points of the return loss curve. A stacked residual network architecture is used to perform dimensionality transformation in a deep neural network to obtain a prediction model.

[0012] According to some embodiments of the present invention, a Dropout layer is introduced in the dimensionality transformation process in the deep neural network, and a portion of neuron outputs is randomly discarded with a preset probability in each training batch.

[0013] According to some embodiments of the present invention, the method further includes evaluating the prediction model, dividing part of the data in the simulation database into a test set, evaluating the prediction model using the test set, and determining the prediction model based on the evaluation result.

[0014] According to some embodiments of the present invention, dividing part of the data in the simulation database into a test set, using the test set to evaluate the prediction model, and determining the prediction model based on the evaluation result includes:

[0015] The MSE loss function is used to evaluate the mean square error of the simulation value and the predicted value. The mean square error gradient descent represents a better prediction model. If it no longer decreases after a preset number of verifications or reaches the preset upper limit, the evaluation is stopped and the prediction model is obtained.

[0016] According to some embodiments of the present invention, in S400, obtaining the electromagnetic absorption performance of the candidate ply configuration, determining whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determining that the candidate ply configuration is the optimal ply configuration includes the following steps:

[0017] S401, setting a frequency band target for electromagnetic absorption;

[0018] S402, evaluating the effective absorption bandwidth of the candidate ply configuration using a prediction model at the target frequency band;

[0019] S403, simulating a preset number of candidate ply configurations obtained in each round of Bayesian iteration using the electromagnetic simulation software to obtain simulation results;

[0020] S404: backfill the simulation results into the simulation database and retrain the model until no ply configuration with a better effective absorption bandwidth appears in a new round of candidate ply configurations, which is the ply configuration with the optimal frequency band target.

[0021] According to some embodiments of the present invention, the electromagnetic absorption performance is obtained by analyzing the simulation database to extract indicators characterizing the electromagnetic absorption performance from the data, and the distribution of the indicators characterizing the electromagnetic absorption performance in the parameter space conforms to a Gaussian distribution or a near-Gaussian distribution.

[0022] According to some embodiments of the present invention, the index characterizing the electromagnetic absorption performance includes an interpolation point in a frequency range based on a reflection loss curve representing the electromagnetic absorption performance, and the corresponding index formula is:

[0023]

[0024] Where: N EAB is the number of interpolation points that meet the conditions, EAB is the effective absorption bandwidth, N freq It is a judgment function for whether the return loss meets -10 dB at a certain frequency. freq1 is the start frequency of the target range, and freq2 is the cutoff frequency of the target range.

[0025] According to some embodiments of the present invention, the formula for the electromagnetic absorption performance is:

[0026]

[0027] in:

[0028] EAC is the electromagnetic absorption capacity, RL freq is the reflection loss value at a certain frequency.

[0029] Also disclosed is an artificial intelligence-based electromagnetic performance design system for laminated structural materials, including a memory and a processor, wherein the memory stores a determination program for the artificial intelligence-based electromagnetic performance design method for laminated structural materials, and the processor runs the artificial intelligence-based electromagnetic performance design method for laminated structural materials to enable the artificial intelligence-based electromagnetic performance design system for laminated structural materials to execute the artificial intelligence-based electromagnetic performance design method for laminated structural materials.

[0030] The artificial intelligence-based electromagnetic performance design method and system for laminated structural materials according to embodiments of the present invention have at least the following beneficial effects: by establishing a predictive model that predicts the electromagnetic absorption performance of different ply configurations, the ply configuration with the optimal electromagnetic absorption performance among different candidate ply configurations can be quickly identified. This solves the difficulties and high costs associated with searching a large parameter space composed of ply units and material types for electromagnetic absorption materials, providing an efficient and rapid design method for broadband electromagnetic absorption materials that does not rely on expert experience.

[0031] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0033] Figure 1 This is a flowchart of an artificial intelligence-based electromagnetic performance design method for laminated structural materials according to an embodiment of the present invention;

[0034] Figure 2 A flowchart of a progressive Bayesian optimization algorithm according to an embodiment of the present invention;

[0035] Figure 3 This is a probability distribution diagram of EAB in 10,000 sets of simulation data according to an embodiment of the present invention;

[0036] Figure 4 A diagram showing the corresponding relationship between EAB and EAC in 10,000 sets of simulation data according to an embodiment of the present invention;

[0037] Figure 5 This is a probability distribution diagram of EAC in 10,000 sets of simulation data according to an embodiment of the present invention;

[0038] Figure 6 In one embodiment of the present invention, after Bayesian iteration, the linear fitting curve of the data predicted by the prediction model and the simulation data is obtained, and the linear correlation is represented by the coefficient of determination R and the relative error;

[0039] Figure 7 In the Bayesian iterative process of an embodiment of the present invention, the first 100 candidates are the prediction situation of one round of iteration;

[0040] Figure 8 A schematic diagram of a laminated structural material composed of three materials and a total of 34 layers according to an embodiment of the present invention;

[0041] Figure 9 for Figure 8 Prediction and simulation results of representative laminated structural materials, including reflection loss curves across the 1-18 GHz frequency range;

[0042] Figure 10 The predicted and simulated reflection loss curves of the laminated structural material after optimization design for the 1-4 GHz sub-band;

[0043] Figure 11 The predicted and simulated reflection loss curves of the laminated structural material after optimization for the 4-8 GHz sub-band;

[0044] Figure 12Predicted and simulated reflection loss curves of laminated structural materials optimized for the 8-18 GHz sub-band;

[0045] Figure 13 This is a structural block diagram of a system for designing electromagnetic properties of laminated structural materials according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0047] In the description of the present invention, it should be understood that descriptions involving orientation, such as the orientation or positional relationship indicated by up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0048] In the description of the present invention, "a plurality" refers to more than two. The use of "first" or "second" is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of the indicated technical features, or implicitly indicating the order of the indicated technical features.

[0049] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, and connecting should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0050] Explanation of terms:

[0051] Lamination configuration: the sequence of laminated structural materials from top to bottom, how they are laid; a layup unit is a layer of material, one in the layup sequence; layup is stacking layers one by one.

[0052] Reference Figure 1 As shown, the present invention discloses an artificial intelligence-based electromagnetic performance design method for laminated structural materials, comprising:

[0053] S100, obtaining basic electromagnetic parameters of ply units through experiments, encoding ply units with different electromagnetic absorption properties, and generating random codes of ply configurations based on time dependency;

[0054] S200, establishing a simulation database of random codes of ply configurations and corresponding electromagnetic absorption performances using electromagnetic simulation software;

[0055] S300, using a simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model;

[0056] S400 , obtaining electromagnetic absorption performance of a candidate ply configuration through prediction using a prediction model, determining whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determining the candidate ply configuration as an optimal ply configuration.

[0057] In some embodiments of the invention, basic electromagnetic parameters of the ply unit are obtained through experiments, including the real part and imaginary part of the dielectric constant, and the real part and imaginary part of the magnetic permeability.

[0058] A self-coded script was used to implement batch processing of finite element simulations in conjunction with CST Studio. This process established a simulation database of approximately 10,000 data sets. This was done to reduce the time and uncertainty associated with manual processing, as performing electromagnetic simulations for each randomly generated ply configuration is tedious and repetitive. A script was used to import each ply configuration into the electromagnetic simulation software, assign material properties, and set a series of electromagnetic simulation boundaries, modes, and frequency ranges.

[0059] The integrated simulation database was randomly divided into 80% as a training set, 15% as a validation set, and 5% as a test set. The training set was used to train the prediction model, representing the implicit mapping relationship between ply configuration and electromagnetic absorption performance. The validation set was used to verify the trained prediction model, identify better architectures, and improve the model's generalization capabilities. The test set was used to evaluate performance, using the mean squared error (MSE) loss function to assess the mean squared error (MSE) between the simulated and predicted values. Gradient descent of the mean squared error indicates a better prediction model. Evaluation ceased if the mean squared error stopped decreasing after 50 validation runs or reached a limit of 300 runs, resulting in a prediction model.

[0060] The trained prediction model can predict the electromagnetic absorption performance of a given ply code, enabling cost-effective prediction and subsequent iterative optimization of unknown ply configurations. By inputting the codes corresponding to different candidate ply configurations and outputting the corresponding electromagnetic absorption performance, the ply configuration with the best electromagnetic absorption performance is selected. This ply configuration then becomes the structure of the laminated material.

[0061] The present invention solves the problems faced by electromagnetic absorption materials, such as the difficulty and high cost of searching a large parameter space composed of ply units and material types, and provides an efficient and fast design method for broadband electromagnetic absorption materials that does not rely on expert experience.

[0062] In some embodiments of the present invention, in S300, using a simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model includes:

[0063] The encoded ply configuration is input, and the output is the interpolation points of the return loss curve. A stacked residual network architecture is used to perform dimensionality transformation in a deep neural network to obtain a prediction model.

[0064] In this embodiment, a simulation database trains a proxy model for predicting electromagnetic absorption performance under different ply configurations. The input is the encoded ply configuration, and the output is the interpolated points of the return loss curve. A stacked residual network architecture is used to transform dimensions in deep neural networks. This architecture comprises an input layer, multiple hidden layers, and an output layer. The number of neurons in each layer and the number of layers are transformed by a certain multiple of the gradient, solving the problems of vanishing and exploding gradients in deep networks. In backpropagation, the gradient is transmitted layer by layer using the chain rule. The absolute value of the derivative of the activation function is usually less than 1. After multiplication of multiple layers, the gradient decays exponentially, making it difficult to update shallow network parameters. Alternatively, if the weights are initialized too large or the activation function derivative is greater than 1, the gradient may grow exponentially, resulting in unstable parameter updates. Traditional networks have linear inter-layer transformations, while residual connections introduce nonlinear superposition, diversifying the gradient propagation paths, enhancing stability, and increasing the depth and expressiveness of the network. In this process, the internal architecture of the neural network is continuously trained by inputting different data. Each time a dimension transformation is completed, a prediction model is obtained, and the mean square error of the simulation value and the predicted value is evaluated by using the MSE loss function. The mean square error gradient descent represents a better prediction model. If it no longer decreases after 50 verifications or reaches the upper limit of 300 times, the evaluation is stopped and the prediction model is obtained.

[0065] Furthermore, the Dropout layer is introduced in the dimensionality transformation process in the deep neural network, and a part of the neuron output is randomly discarded with a preset probability in each training batch.

[0066] The Dropout layer is introduced during the dimensionality transformation process, which randomly discards a portion of neuron outputs with a probability of 25% in each training batch, reducing the model complexity and computational cost while increasing the robustness and generalization ability of the model.

[0067] In some embodiments of the present invention, the prediction model is further evaluated, part of the data in the simulation database is divided into a test set, the prediction model is evaluated using the test set, and the prediction model is determined according to the evaluation result.

[0068] Furthermore, part of the data in the simulation database is divided into a test set, the prediction model is evaluated using the test set, and the prediction model is determined based on the evaluation results to include:

[0069] The MSE loss function is used to evaluate the mean square error of the simulation value and the predicted value. The mean square error gradient descent represents a better prediction model. If it no longer decreases after a preset number of verifications or reaches the preset upper limit, the evaluation is stopped and the prediction model is obtained.

[0070] In this example, the integrated simulation database was randomly divided into 80% as a training set, 15% as a validation set, and 5% as a test set. The training set was used to train the prediction model, representing the implicit mapping relationship between ply configuration and electromagnetic absorption performance. The validation set was used to validate the trained prediction model, identify better architectures, and improve the model's generalization capabilities. The test set was used to evaluate performance using the mean squared error (MSE) loss function to assess the mean squared error (MSE) between the simulated and predicted values. Gradient descent of the mean squared error indicates a better prediction model. Evaluation ceases after 50 validation runs or reaches a limit of 300 runs, resulting in a predicted model.

[0071] In some embodiments of the present invention, the electromagnetic absorption performance is analyzed by analyzing a simulation database to extract indicators characterizing the electromagnetic absorption performance from the data, and the distribution of the indicators characterizing the electromagnetic absorption performance in the parameter space conforms to a Gaussian distribution or a near-Gaussian distribution.

[0072] refer to Figures 3 to 5 As shown, Figure 3 The analysis shows the probability distribution of the effective absorption bandwidth (EAB) for 10,000 data sets. The 4-5 GHz range accounts for the largest proportion, at 32.95%, while the probability of distribution within the 7-8 GHz range is only 0.52%. In this paper, to provide guidance for ply design with excellent electromagnetic absorption capabilities, we focus on the distribution percentage of the widest EAB in the histogram and the corresponding ply configuration. However, based on the EAB data distribution, exploring this focus through extensive simulation calculations is inefficient, and the resulting EAB is still relatively narrow.

[0073] In general, the electromagnetic absorption capacity (EAC) can be used to characterize the electromagnetic absorption performance. The formula is:

[0074]

[0075] Where: EAC is the electromagnetic absorption capacity, which is expressed as the integral area to show the overall absorption capacity of the laminated material, RL freq is the return loss value at a certain frequency, freq1 is the start frequency of the target range, and freq2 is the cutoff frequency of the target range.

[0076] By studying the scattered distribution of electromagnetic absorption capacity (EAC) with respect to EAB, such as Figure 4As shown in Figure 2, it can be found that the EAB corresponding to the EAC value varies from a larger value range. For example, when the EAC value is 133 dB·GHz, the EAB can vary from 0.76 to 7.08 GHz, and as shown in Figure 2 Figure 5 The probability distribution of EAC shown in the figure is mostly concentrated in the range of 121-141 dB·GHz, accounting for as much as 82.42%. This indicates that in 10,000 sets of data, as many as 82.42% of the ply configurations have similar EACs. Highly concentrated data will mask the characteristics and regularities of the ply configuration, making it difficult to predict the mapping relationship between electromagnetic performance and ply configuration.

[0077] On the contrary, EAB presents a nearly Gaussian distribution in the data structure from 0 to 8 GHz, indicating that the use of EAB values ​​can better characterize the electromagnetic absorption performance related to the ply configuration, thereby facilitating the discovery of the implicit mapping relationship between the two. Therefore, in the subsequent EAB The EAB represented by the value is used to perform the ply reverse design based on Bayesian optimization.

[0078] In some embodiments of the present invention, the index characterizing the electromagnetic absorption performance includes an interpolation point in the frequency range based on a reflection loss curve representing the electromagnetic absorption performance. The corresponding index formula is:

[0079]

[0080] Where: N EAB is the number of interpolation points that meet the conditions, EAB is the effective absorption bandwidth, N freq is a judgment function for whether the reflection loss meets -10 dB at a certain frequency, freq1 is the starting frequency of the target range, and freq2 is the cutoff frequency of the target range. Using the above formula, the effective frequency bandwidth within the target frequency range at which the material reflection loss meets the electromagnetic stealth requirement is calculated.

[0081] refer to Figure 2 As shown, in some embodiments of the present invention, in S400, obtaining the electromagnetic absorption performance of a candidate ply configuration through prediction by the prediction model, and determining whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determining that the candidate ply configuration is an optimal ply configuration includes the following steps:

[0082] S401, setting a frequency band target for electromagnetic absorption;

[0083] S402, evaluating the effective absorption bandwidth of the candidate ply configuration using a prediction model at the target frequency band;

[0084] S403, simulating a preset number of candidate ply configurations obtained in each round of Bayesian iteration using the electromagnetic simulation software to obtain simulation results;

[0085] S404: backfill the simulation results into the simulation database and retrain the model until no ply configuration with a better effective absorption bandwidth appears in a new round of candidate ply configurations, which is the ply configuration with the optimal frequency band target.

[0086] refer to Figure 8 , which is a schematic diagram of the ply configuration of the laminated structural material provided in a specific embodiment of the present invention. As can be seen from the figure, the structure is composed of three materials, showing a certain gradient distribution.

[0087] refer to Figure 9 In this embodiment, the frequency band target for electromagnetic absorption is 1-18 GHz, that is, freq1 = 1 GHz, freq2 = 18 GHz. Figure 7 As shown in the figure, during the progressive Bayesian optimization iteration process, the electromagnetic simulation results of the top 100 candidate ply configurations in a new round are continuously added to retrain the prediction model until better electromagnetic absorption performance no longer appears in the new round of candidate ply configurations.

[0088] Using trained prediction models to approximate the electromagnetic absorption performance of different ply configurations, the simulation speed is increased by about 10% compared to CST Studio finite element simulation while maintaining high prediction accuracy. 6 times (0.02ms). However, considering the wide parameter space (3 34 ), the challenge lies in how to find the mapping relationship between ply configuration and electromagnetic performance, and further accelerate the reverse design of electromagnetic performance.

[0089] After every 100 Bayesian optimization runs, the optimization results are simulated through CST Studio and added to the database to retrain the prediction model, continuously improving its accuracy and generalization capabilities.

[0090] The electromagnetic target is changed and set according to the iterative effect. After each change, the optimal solution under the previous electromagnetic target is provided as a priori condition to the subsequent process.

[0091] By building a machine learning framework that combines progressive Bayesian optimization algorithms, Figure 8 The total design space consisting of three electromagnetic absorption materials is 3 34 Prediction and optimization of unknown ply configurations, such as Figure 6 As shown in the figure, after progressive optimization, the coefficient of determination of the prediction model is 0.964 and the mean absolute error is 0.27 GHz, which proves the reliability of the prediction model.

[0092] like Figure 9 As shown in the electromagnetic performance prediction diagram of the laminated structural material provided in this embodiment, it can be seen from the figure that, in the frequency band target of electromagnetic absorption with 1-18 GHz, the broadband electromagnetic absorption laminated structural material after optimized design achieves an effective absorption bandwidth of 9.6 GHz, and the simulation results and the prediction results have a good degree of matching.

[0093] Expand the electromagnetic target to other ranges and repeat the above process to finally obtain the ply configuration with effective absorption bandwidth of the target in this frequency band. Figures 10 to 12 , respectively taking 1-4 GHz, 4-8 GHz and 8-18 GHz as the frequency band targets, and similarly going through steps S402 to S404 , the optimal ply configuration for the frequency band targets is obtained.

[0094] like Figures 10 to 12 As shown in the figure, there are prediction and simulation diagrams of electromagnetic performance of laminated structural materials optimized for the frequency bands of 1-4GHz, 4-8GHz and 8-18GHz. The width of the frequency band below -10dB shown in the figure indicates that good electromagnetic absorption is achieved for the sub-band.

[0095] refer to Figure 13 As shown, the present invention also discloses an artificial intelligence-based electromagnetic performance design system for laminated structural materials, including a memory and a processor, wherein the memory stores a determination program for the artificial intelligence-based electromagnetic performance design method for laminated structural materials, and the processor runs the determination program for the artificial intelligence-based electromagnetic performance design method for laminated structural materials, so that the artificial intelligence-based electromagnetic performance design system for laminated structural materials executes the artificial intelligence-based electromagnetic performance design method for laminated structural materials.

[0096] Since the electromagnetic performance design system for laminated structural materials based on artificial intelligence adopts all the technical solutions of the electromagnetic performance design method for laminated structural materials based on artificial intelligence in the above-mentioned embodiment, it has at least all the beneficial effects brought by the technical solutions of the above-mentioned embodiment, which will not be repeated here.

[0097] The embodiments of the present invention are described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the scope of the present invention.

Claims

1. A method for designing electromagnetic properties of laminated structural materials based on artificial intelligence, characterized in that: include: S100, obtaining basic electromagnetic parameters of ply units through experiments, encoding ply units with different electromagnetic absorption properties, and generating random codes of ply configurations based on time dependency; S200, establishing a simulation database of random codes of ply configurations and corresponding electromagnetic absorption performances using electromagnetic simulation software; S300, using the simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model; S400, obtaining electromagnetic absorption performance of a candidate ply configuration through prediction using the prediction model, determining whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determining the candidate ply configuration as an optimal ply configuration; The electromagnetic absorption performance is obtained by analyzing the simulation database to extract an index representing the electromagnetic absorption performance from the data, and the distribution of the index representing the electromagnetic absorption performance in the parameter space conforms to a Gaussian distribution or a near-Gaussian distribution; The index characterizing the electromagnetic absorption performance includes an interpolation point in a frequency range based on a reflection loss curve representing the electromagnetic absorption performance. The corresponding index formula is: Where: N EAB is the number of interpolation points that meet the conditions, EAB is the effective absorption bandwidth, N freq It is a judgment function for whether the return loss meets -10 dB at a certain frequency. freq1 is the start frequency of the target range, and freq2 is the cutoff frequency of the target range.

2. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 1, characterized in that: In S300, the use of the simulation database to train a proxy model for predicting electromagnetic absorption performance under different ply configurations to obtain a prediction model includes: The encoded ply configuration is input, and the output is the interpolation points of the return loss curve. A stacked residual network architecture is used to perform dimensionality transformation in a deep neural network to obtain a prediction model.

3. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 2, characterized in that: The Dropout layer is introduced during the dimensionality transformation process in the deep neural network, and a portion of neuron outputs is randomly discarded with a preset probability in each training batch.

4. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 1, characterized in that: The method also includes evaluating the prediction model, dividing part of the data in the simulation database into a test set, evaluating the prediction model using the test set, and determining the prediction model according to the evaluation result.

5. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 4, characterized in that: Dividing part of the data in the simulation database into a test set, using the test set to evaluate the prediction model, and determining the prediction model according to the evaluation result includes: The MSE loss function is used to evaluate the mean square error of the simulation value and the predicted value. The mean square error gradient descent represents a better prediction model. If it no longer decreases after a preset number of verifications or reaches the preset upper limit, the evaluation is stopped and the prediction model is obtained.

6. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 1, characterized in that: In S400, obtaining the electromagnetic absorption performance of the candidate ply configuration and determining whether the electromagnetic absorption performance of the candidate ply configuration is an optimal value, and if so, determining that the candidate ply configuration is the optimal ply configuration includes the following steps: S401, setting a frequency band target for electromagnetic absorption; S402, evaluating the effective absorption bandwidth of the candidate ply configuration using a prediction model at the target frequency band; S403, simulating a preset number of candidate ply configurations obtained in each round of Bayesian iteration using the electromagnetic simulation software to obtain simulation results; S404: backfill the simulation results into the simulation database and retrain the model until no ply configuration with a better effective absorption bandwidth appears in a new round of candidate ply configurations, which is the ply configuration with the optimal frequency band target.

7. The electromagnetic performance design method of laminated structural materials based on artificial intelligence according to claim 1, characterized in that: The formula for the electromagnetic absorption performance is: in: EAC is the electromagnetic absorption capacity, RL freq is the return loss value at a certain frequency, freq1 is the start frequency of the target range, and freq2 is the cutoff frequency of the target range.

8. An artificial intelligence-based electromagnetic performance design system for laminated structural materials, characterized by: It includes a memory and a processor, wherein the memory stores a determination program of the electromagnetic performance design method of laminated structural materials based on artificial intelligence, and the processor runs the determination program of the electromagnetic performance design method of laminated structural materials based on artificial intelligence, so that the electromagnetic performance design system of laminated structural materials based on artificial intelligence executes the electromagnetic performance design method of laminated structural materials based on artificial intelligence as described in any one of claims 1 to 7.

Images