A Wave Absorption Optimization Method and System Based on Spectrum Recognition

Abstract

This invention relates to the field of microwave absorbing material optimization technology, and discloses a microwave absorbing optimization method and system based on graph recognition. The method acquires configuration data information of multiple types of microwave absorbing materials, constructs a material performance graph containing material configuration characteristics and composite layout relationships, and performs graph structure analysis on the graph to extract topological dependencies between nodes and complete multi-hop information transmission. Based on the topological dependencies and multi-hop information transmission results, it identifies key structural combinations that play a dominant role in microwave absorbing performance, and uses these as constraint references to set an optimization variable constraint space containing material type, thickness, morphology, and arrangement sequence. Using the material performance graph and the optimization variable constraint space as screening conditions, an optimization-guided algorithm selects configuration schemes with high microwave absorbing performance prediction scores from the graph. This method achieves intelligent optimization design of microwave absorbing material configurations, effectively improving the accuracy and efficiency of microwave absorbing performance optimization.

Description

Technical Field

[0001] This invention relates to the field of microwave absorbing material optimization technology, specifically to a microwave absorbing optimization method and system based on spectrum recognition. Background Technology

[0002] Microwave absorbing materials have crucial application value in modern technology, and their performance directly affects the development of key areas such as the anti-interference capability of electronic devices and the stealth effect of military equipment. With technological advancements, various application scenarios are placing higher demands on the absorption frequency band, intensity, and other performance indicators of microwave absorbing materials. Therefore, how to efficiently optimize the configuration of microwave absorbing materials has become an urgent problem to be solved. Traditional optimization methods for microwave absorbing materials have significant drawbacks, mostly relying on manual experience for material type selection, thickness determination, morphological structure design, and arrangement sequence planning. This subjective judgment-based approach not only consumes a lot of time and energy but also makes it difficult to fully grasp the complex relationship between material configuration characteristics and composite layout.

[0003] Traditional optimization methods, when dealing with multivariate constrained problems, lack scientific and systematic analytical tools and cannot deeply explore the topological dependencies between material properties. This leads to a high degree of blindness in the optimization process and makes it difficult to accurately identify the key structural combinations that dominate the microwave absorption performance. Furthermore, some existing optimization algorithms suffer from low computational efficiency and slow convergence speed when facing complex situations with multiple parameters and constraints in microwave absorbing material optimization, making it difficult to quickly obtain reliable optimization results. This severely restricts the research and development progress and practical application effects of microwave absorbing materials. Therefore, there is an urgent need for a new method and system that can overcome the shortcomings of traditional methods and achieve efficient and accurate optimization of microwave absorbing material configurations. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing absorption based on spectrum recognition, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a method and system for absorption optimization based on spectrum recognition, the method comprising:

[0006] Obtain configuration data information for various types of microwave absorbing materials;

[0007] Using the configuration data as the foundation, a material performance map containing material configuration characteristics and composite layout relationships is constructed. Graph structure analysis is performed on the material performance map to extract topological dependencies between nodes and complete multi-hop information transmission. Based on the topological dependencies and multi-hop information transmission results, key structural combinations that play a dominant role in microwave absorption performance are identified. These key structural combinations are used as constraint references to set an optimization variable constraint space containing material type, thickness, morphology, and arrangement sequence. Using the material performance map and the optimization variable constraint space as screening conditions, configuration schemes with high microwave absorption performance prediction scores are selected from the material performance map based on an optimization-guided algorithm.

[0008] Preferably, constructing a material property map containing the relationship between material configuration characteristics and composite layout based on the configuration data information includes:

[0009] Collect configuration feature data of different microwave absorbing materials, the configuration feature data including material microstructure parameters and macroscopic composite layout information; record the composite layout relationship between each material in the configuration feature data; label the corresponding microwave absorption intensity parameters for the configuration feature data; use the labeled configuration feature data as graph nodes and the composite layout relationship as graph edge weights to establish a material performance graph that associates nodes with edge weights.

[0010] Preferably, graph structure analysis is performed on the material property map to extract topological dependencies between nodes and complete multi-hop information transfer, including:

[0011] The node features and edge weights of the material property map are convolved using graph convolution operations to extract the topological dependencies between nodes. The topological dependencies are weighted using a graph attention mechanism to generate attention weights that include multi-hop neighborhood information. Based on the attention weights, multi-hop information is transferred between nodes, and node feature representations that fuse multi-hop information are output.

[0012] Preferably, the key structural combinations that play a dominant role in absorption performance, identified based on the topological dependencies and multi-hop information transmission results, include:

[0013] Analyze the influence of each node's connection edge weight in the topological dependency relationship, and filter the node combinations corresponding to edge weights whose influence exceeds a preset threshold; combine the contribution values ​​of each node's feature representation in the multi-hop information transmission results, and extract node combinations whose contribution values ​​are higher than a preset contribution value; perform intersection matching between the filtered node combinations and the extracted node combinations, and output the matching result as the key structure combination.

[0014] Preferably, an optimization variable constraint space is set, including material type, thickness, morphology, and arrangement sequence, including:

[0015] The range of selectable types of absorbing materials is determined, including carbon-based materials, magnetic materials, and composite nanomaterials; the thickness range of each type of material is set, and the range is determined according to the requirements of the material's absorption frequency band; the selectable morphology of the material's microstructure is limited, including granular, sheet-like, and fibrous forms; and the arrangement sequence rules of the materials in the composite system are specified, including alternating arrangement, layered stacking, and gradient distribution.

[0016] Preferably, the configuration scheme with the highest predicted microwave absorption performance score is selected from the material property spectrum based on an optimization-guided algorithm, including:

[0017] An initial score value is assigned to each node combination in the material performance spectrum; the initial score value is adjusted to generate a predicted score based on the matching degree between the absorption intensity parameter corresponding to the node combination and the optimization variable constraint space; the node combinations are sorted from high to low according to the predicted scores, and the top preset number of node combinations in the sorting results are selected as candidate configuration schemes; the scheme with the highest predicted score among the candidate configuration schemes is selected as the final recommended configuration scheme.

[0018] Preferably, the initial score is adjusted to generate a predicted score based on the degree of matching between the absorption intensity parameters corresponding to the node combination and the optimization variable constraint space, including:

[0019] Calculate the matching degree between the material type in the node combination and the range of selectable types in the optimization variable constraint space; calculate the conformity between the material thickness in the node combination and the thickness value range in the optimization variable constraint space; calculate the consistency between the morphology and structure in the node combination and the selectable forms in the optimization variable constraint space; calculate the fit between the permutation sequence in the node combination and the permutation sequence rule in the optimization variable constraint space; sum the matching degree, conformity, consistency, and fit by weight, and output the adjusted prediction score.

[0020] Preferably, the present invention further includes a spectrum recognition-based absorption optimization system, the system being used to implement the above-described spectrum recognition-based absorption optimization method, the system comprising:

[0021] The spectrum construction module is used to acquire configuration data information of multiple types of microwave absorbing materials and construct a material performance spectrum that includes the material configuration characteristics and the relationship between the composite layout.

[0022] The graph recognition module is used to perform graph structure analysis on the material property spectrum, extract the topological dependencies between nodes and complete multi-hop information transmission, and identify key structural combinations based on the transmission results.

[0023] An optimization guidance module is used to set an optimization variable constraint space including material type, thickness, morphology, and arrangement sequence, and select a configuration scheme with a high microwave absorption performance prediction score from the material performance spectrum based on the optimization guidance algorithm.

[0024] A configuration recommendation module is configured to output the high-scoring configuration scheme selected by the optimization guidance module.

[0025] Preferably, the spectrum construction module includes a data collection unit, a relationship recording unit, a parameter annotation unit, and a spectrum establishment unit; the data collection unit is used to collect configuration feature data of different absorbing materials; the relationship recording unit is used to record the composite layout relationship between each material in the configuration feature data; the parameter annotation unit is used to annotate the corresponding absorbing intensity parameters for the configuration feature data; the spectrum establishment unit is used to use the annotated configuration feature data as spectrum nodes and the composite layout relationship as spectrum edge weights to establish a material performance spectrum that associates nodes with edge weights.

[0026] Preferably, the graph recognition module includes a topology extraction unit, an information transmission unit, and a structure screening unit; the topology extraction unit is used to perform graph convolution operations on the node features and edge weight information of the material performance graph to extract the topological dependencies between nodes; the information transmission unit is used to assign weights to the topological dependencies through a graph attention mechanism to complete multi-hop information transmission between nodes; the structure screening unit is used to screen key structural combinations that play a dominant role in the absorption performance based on the topological dependencies and the multi-hop information transmission results.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] By constructing a material performance map, the configurational characteristics and composite layout relationships of the absorbing materials are organically integrated in a graph structure, making the material performance information more intuitive and clear, and facilitating in-depth analysis of the interactions between various factors. Using graph structure analysis to extract node topological dependencies and complete multi-hop information transmission, the correlation effects of each part in the material configuration can be comprehensively captured, accurately identifying key structural combinations that play a decisive role in absorbing performance, and providing more targeted directions for optimization work.

[0029] By establishing a constraint space for optimization variables encompassing multiple dimensions such as material type, thickness, morphology, and arrangement sequence, and combining this with an optimization-guided algorithm to screen configuration schemes with high prediction scores from material performance maps, the entire process from data collection to scheme generation is automated, significantly reducing reliance on human experience. This method comprehensively considers multiple optimization variables and their constraints, and through scientific calculation and analysis, quickly and efficiently screens configuration schemes with excellent microwave absorption performance, significantly improving the efficiency and accuracy of microwave absorbing material optimization. The system's map construction module ensures data integrity and accuracy, the map recognition module delves into the intrinsic relationships of material properties, and the optimization guidance and configuration recommendation modules achieve intelligent generation and accurate recommendation of optimization schemes. The collaborative work of these modules makes the entire system highly operable and practical, providing effective technical support for microwave absorbing material researchers, helping to obtain higher-performance microwave absorbing materials, and promoting the progress and development of microwave absorbing material technology in various application fields. Attached Figure Description

[0030] Figure 1 This is a schematic diagram illustrating the working principle of the absorption optimization method based on spectrum recognition described in this invention.

[0031] Figure 2 Flowchart for constructing material property maps;

[0032] Figure 3 A flowchart for graph structure analysis and multi-hop information transmission;

[0033] Figure 4 A flowchart for identifying key structural combinations. Detailed Implementation

[0034] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Please see Figures 1-4 This invention provides a wave absorption optimization method based on spectrum recognition, the specific implementation of which is as follows:

[0036] Obtain configuration data for various types of microwave absorbing materials. This configuration data covers detailed structural information for different microwave absorbing materials and forms the basis for all subsequent operations.

[0037] Using configuration data as the foundation, a material performance map is constructed that includes material configuration characteristics and composite layout relationships. Specifically, configuration characteristic data of different microwave absorbing materials are first collected, including material microstructure parameters and macroscopic composite layout information; then, the composite layout relationships between the materials in these configuration characteristic data are recorded; next, the corresponding microwave absorption intensity parameters are labeled for the configuration characteristic data; finally, the labeled configuration characteristic data are used as map nodes, and the composite layout relationships are used as map edge weights to establish a material performance map that associates nodes with edge weights.

[0038] Graph structure analysis is performed on the material property map to extract topological dependencies between nodes and complete multi-hop information transfer. Graph convolution is used to convolve the node features and edge weights of the material property map to extract the topological dependencies between nodes. A graph attention mechanism is used to assign weights to the topological dependencies, generating attention weights that include multi-hop neighborhood information. Based on these attention weights, multi-hop information transfer between nodes is completed, and node feature representations incorporating multi-hop information are output.

[0039] The key structural combinations that play a dominant role in absorption performance are identified based on topological dependencies and multi-hop information transmission results. The influence of the edge weights connecting each node in the topological dependencies is analyzed, and node combinations corresponding to edge weights with influence exceeding a preset threshold are selected. Combining the contribution values ​​of each node's feature representation in the multi-hop information transmission results, node combinations with contribution values ​​higher than a preset contribution value are extracted. The selected node combinations are then matched with the extracted node combinations, and the matching result is used as the key structural combination.

[0040] Using key structural combinations as constraint references, an optimization variable constraint space is defined, encompassing material type, thickness, morphology, and arrangement sequence. The range of selectable microwave absorbing materials is determined, including carbon-based materials, magnetic materials, and composite nanomaterials. Thickness ranges for each material type are defined, determined based on the required microwave absorption frequency bands. The selectable morphologies of the material's microstructure are limited, including granular, sheet-like, and fibrous forms. The arrangement sequence rules for materials in the composite system are specified, including alternating arrangement, layered stacking, and gradient distribution.

[0041] Using material performance maps and optimization variable constraint spaces as screening criteria, configuration schemes with high microwave absorption performance prediction scores are selected from the material performance maps based on an optimization-guided algorithm. An initial score is assigned to each node combination in the material performance map. Based on the matching degree between the microwave absorption intensity parameters corresponding to the node combination and the optimization variable constraint space, the initial score is adjusted to generate a predicted score. Specifically, this involves calculating the matching degree between the material type in the node combination and the range of selectable types in the optimization variable constraint space, the conformity of the material thickness with the thickness range, the consistency of the morphology with the selectable forms, and the fit of the permutation sequence with the permutation sequence rule. These degrees are then weighted and summed to obtain the predicted score. The node combinations are sorted from highest to lowest predicted score, and the top preset number of node combinations are selected as candidate configuration schemes. The scheme with the highest predicted score among the candidate configuration schemes is selected as the final recommended configuration scheme.

[0042] Example 1: During the construction of material performance maps, comprehensive data on the configuration characteristics of different microwave absorbing materials were collected. These materials encompass various types. For carbon-based materials, microstructural parameters such as graphite interlayer spacing, porosity, and specific surface area need to be obtained, as these parameters directly affect the electromagnetic loss characteristics of the material. Simultaneously, macroscopic composite layout information is acquired, such as whether it is a single layer or mixed with other materials in the composite material, and its volume percentage in the overall structure. For magnetic materials, microstructural parameters include grain size, magnetic domain structure, and saturation magnetization, while macroscopic composite layout information involves the arrangement when combined with other materials—whether it is layered or dispersed in the matrix. Data collection on the configuration characteristics of composite nanomaterials is more complex, requiring the acquisition of microscopic parameters such as nanoparticle size, morphology, and crystal form, as well as macroscopic composite layout information such as the dispersion state of nanoparticles in the matrix and interfacial bonding.

[0043] After collecting the configuration feature data, it is necessary to record the detailed composite layout relationships between the materials. This process requires clarifying the spatial relationships of different materials within the composite system. For example, are two materials distributed in parallel layers, arranged in an interlaced pattern, or is one material uniformly dispersed as particles within the matrix of another material? Simultaneously, the thickness ratio of each layer should be recorded; for instance, in a layered structure, the specific thickness of each layer and the thickness ratio between layers should be specified. Furthermore, the interfacial bonding methods of the materials in the composite system also need to be recorded—whether it is mechanical bonding, chemical bonding, or physical adsorption, etc. These composite layout relationships will significantly impact the microwave absorption performance of the materials.

[0044] The collected configuration feature data is labeled with corresponding absorption intensity parameters. Obtaining these parameters requires specific testing methods, typically using equipment such as a vector network analyzer, to test the material within a defined frequency range. During testing, the material is prepared as a sample of a specific shape and size, placed in the testing apparatus, and electromagnetic waves are emitted and received by the instrument. The reflected electromagnetic waves are then calculated, and this reflection loss value is a key indicator of the absorption intensity parameter. During the labeling process, it is crucial to ensure that each configuration feature data corresponds to an accurate absorption intensity parameter so that the subsequent construction of the spectrum accurately reflects the relationship between configuration and performance.

[0045] After completing data collection, relationship recording, and parameter annotation, the material performance map was constructed. The annotated configurational features were used as map nodes, each containing complete microstructural parameters, macroscopic composite layout information, and corresponding absorption intensity parameters. For example, a node might contain information about a carbon-based material with a specific graphite interlayer spacing, a certain percentage of porosity, a layered distribution in the composite system with a fixed thickness, and a corresponding absorption intensity parameter with a specific reflection loss at a certain frequency band. The composite layout relationships were used as map edge weights, determined by considering both the density of the layout and its impact on absorption performance. Higher edge weights were assigned to denser layouts with a greater impact on absorption performance, and vice versa. This method established a material performance map linking nodes and edge weights, allowing the map to visually demonstrate the connections between different configurational features and composite layout relationships, as well as their impact on absorption performance.

[0046] In constructing the spectral map, it is crucial to ensure the accuracy and completeness of the data. The collected configurational characteristic data must undergo rigorous screening and verification to eliminate outliers and errors, ensuring the reliability of the map. Simultaneously, the determination of edge weights must be based on extensive experimental data and theoretical analysis, avoiding subjective assumptions to guarantee that the map accurately reflects the intrinsic relationship between material properties and configuration. Furthermore, spectral construction is a dynamic process; as new configurational data and absorption intensity parameters are acquired, the map can be continuously updated and improved, making it more comprehensive and accurate in reflecting the material's performance characteristics.

[0047] The material performance map constructed through the above steps provides an important foundation for subsequent graph structure analysis, key structural combination identification, and configuration optimization. It visually presents complex material configuration information and performance parameters in graphical form, facilitating in-depth analysis and processing using graph theory and machine learning methods, thereby enabling optimized design of microwave absorbing materials.

[0048] Example 2: When performing graph structure analysis on a material property map, graph convolution is first performed to extract the topological dependencies between nodes. The graph convolution operation targets the node features and edge weights in the material property map. Node features include the material's microstructural parameters, macroscopic composite layout information, and corresponding absorption intensity parameters, while edge weights represent the composite layout relationships between different materials. In practice, these node features and edge weights are input into a graph convolutional neural network, and specific convolution kernels are designed to perform operations on them. For example, for a given node, the features and edge weights of its neighboring nodes are processed by the convolution kernel, thereby capturing the impact of the structural association between that node and its neighboring nodes on the absorption performance. During this process, node features need to be normalized to ensure comparability across different dimensions. Simultaneously, edge weights also need to be appropriately encoded so that they can be effectively processed by the graph convolutional neural network. Through multiple convolution operations, the topological dependencies between nodes are gradually extracted. These dependencies reflect the impact of the interaction between different configuration features and composite layout relationships on the absorption performance.

[0049] After extracting topological dependencies, the next step is to assign weights to them using a graph attention mechanism, generating attention weights that incorporate multi-hop neighborhood information. The core of the graph attention mechanism is calculating the attention coefficient between nodes, which represents the degree of attention one node pays to another. When calculating the attention coefficient, the feature vectors of the nodes and the edge weights between them are considered. Specifically, for each node, its feature vector is multiplied by the feature vectors of its neighboring nodes, then weighted according to the edge weights, and finally the attention coefficient is obtained through an activation function. To enhance the model's expressive power, a multi-head attention mechanism is typically used, that is, multiple attention heads are used simultaneously for calculation, and the results are concatenated or averaged. In this way, different attention weights can be assigned to different topological dependencies. Topological dependencies that have a greater impact on absorption performance are assigned higher attention weights, while those with a smaller impact are assigned lower weights. The generated attention weights include multi-hop neighborhood information, that is, they consider not only the information of directly adjacent nodes but also the information of neighboring nodes after multiple steps of propagation, thus capturing the structural features in the graph more comprehensively.

[0050] After generating attention weights, multi-hop information transfer between nodes is completed based on these weights, outputting node feature representations that fuse multi-hop information. The multi-hop information transfer process is iterative; each node collects information from neighboring nodes according to its attention weights and fuses this information with its own features. Specifically, for each node, its new feature representation is obtained by weighted summation of the features of neighboring nodes according to their attention weights, plus its own features. In this process, information is not only transferred between directly adjacent nodes but also through intermediate nodes to more distant nodes, achieving multi-hop propagation. For example, a node's information can be transferred through intermediate nodes to nodes two, three, or even more hops away, allowing each node to acquire a wider range of information from the graph. Through multiple information transfer iterations, the node's feature representation is continuously updated, gradually fusing more multi-hop neighboring information. The final output node feature representation no longer only contains the node's initial features but also incorporates relevant features from its neighboring nodes, more comprehensively reflecting the node's structural position in the entire graph and its impact on absorption performance.

[0051] In performing graph convolution operations, graph attention mechanisms, and multi-hop information transfer, it is necessary to appropriately set the model's hyperparameters, such as the size of the convolution kernel, the number of attention heads, and the number of iterations. These hyperparameter settings directly affect the extraction of topological dependencies between nodes, the generation of attention weights, and the quality of multi-hop information transfer. For example, the size of the convolution kernel determines the neighborhood range that each convolution operation can capture, the number of attention heads affects the model's ability to focus on different feature dimensions, and the number of iterations determines the depth of information transfer. Therefore, during implementation, it is necessary to determine appropriate hyperparameters through experiments and debugging based on the scale and complexity of the material property map to ensure effective extraction of topological dependencies, accurate generation of attention weights, and high-quality multi-hop information transfer.

[0052] Furthermore, during the entire graph structure analysis process, attention must be paid to data processing and model training. For the input node features and edge weights, necessary preprocessing is required, such as handling missing values ​​and filtering outliers, to ensure data quality. During model training, an appropriate loss function, such as the mean squared error loss function or the cross-entropy loss function, needs to be selected, and the model parameters optimized according to the task requirements of absorption performance prediction. Simultaneously, to avoid model overfitting, regularization methods, such as L1 or L2 regularization, and dropout techniques can be employed.

[0053] By performing graph structure analysis on the material property map through the above steps, the topological dependencies between nodes can be effectively extracted, attention weights containing multi-hop neighborhood information can be generated, and multi-hop information transfer between nodes can be completed, ultimately yielding node feature representations that fuse multi-hop information. These processed node feature representations provide an important basis for subsequent identification of key structural combinations that play a dominant role in microwave absorption performance. This enables the accurate screening of structural combinations that have a significant impact on microwave absorption performance from complex material property maps, providing support for the optimized design of microwave absorbing materials.

[0054] Example 3: When identifying key structural combinations based on topological dependencies and multi-hop information transmission results, it is necessary to analyze the influence of the connection weights of each node in the topological dependencies. In the material property map, the weights represent the composite layout relationship between nodes, and their values ​​reflect the strength of the influence of the corresponding structural association on the wave absorption performance. In specific implementation, it is necessary to perform a comprehensive analysis of all node connection weights, and determine the relative influence of each weight by statistically analyzing the distribution of weights in the map, such as the average value and variance of the weights. For example, if the value of a certain weight is significantly greater than that of most weights in the map, it can be preliminarily judged that the structural association represented by the node combination corresponding to that weight has a greater impact on the wave absorption performance. Subsequently, a preset threshold is set, which can be determined based on the statistical characteristics of the weights, such as using the average value of the weights plus a certain multiple of the standard deviation as the threshold. The node combinations corresponding to the weights whose influence exceeds the preset threshold are screened out to form the first type of candidate node combinations.

[0055] Simultaneously, it is necessary to extract node combinations by combining the contribution values ​​of the feature representations of each node in the multi-hop information transmission results. After multi-hop information transmission, the feature representation of each node integrates information from itself and neighboring nodes, and the contribution value is used to measure the importance of the feature representation in reflecting the absorption performance. There are various methods to calculate the contribution value; for example, the gradient attribution method can be used, which determines the contribution value by calculating the gradient of the node feature representation with respect to the output of the absorption performance prediction model. For each node feature representation, the higher its contribution value, the greater the impact of the structural information contained in the feature representation on the absorption performance. A preset contribution value is set, and node combinations with contribution values ​​higher than the preset contribution value are extracted to form the second type of candidate node combinations.

[0056] The intersection matching of the first type of candidate node combinations and the second type of candidate node combinations is performed. The purpose of intersection matching is to find node combinations that simultaneously meet the requirements of edge weight influence and node characteristic contribution value. Such node combinations are more likely to be key structural combinations that play a dominant role in the absorption performance. Specifically, for each node combination in the first type of candidate node combinations, it is checked whether it exists in the second type of candidate node combinations. If it exists, the node combination is included in the matching result.

[0057] In the above process, a key formula is involved in calculating the contribution value of the node feature representation, as follows: in, Represents a node The contribution value represented by the feature; The number of dimensions in the feature representation; Represents the prediction function for absorption performance For nodes Feature representation of the first dimension The partial derivative of , whose absolute value reflects the degree of influence of the feature of that dimension on the prediction result; For the first The weights of the features can be pre-set based on the importance of the features, for example, by domain knowledge or feature selection methods.

[0058] When analyzing the impact of edge weights, it is crucial to consider the reasonableness of the preset threshold. Setting the threshold too high may result in too few selected node combinations, overlooking some important structural combinations; setting it too low may include too many node combinations, increasing the complexity of subsequent processing. Therefore, in practical applications, the preset threshold needs to be adjusted through multiple experiments based on the specific characteristics of the graph and the optimization objectives to ensure that the selected node combinations reflect the main structural influences without being overly broad.

[0059] When calculating the contribution value of nodal features, it is necessary to ensure the accuracy of the microwave absorption performance prediction model. The prediction model can employ various machine learning methods, such as neural networks and random forests. The training data for the model should come from the nodal features and corresponding absorption intensity parameters in the material property spectrum. During model training, it is crucial to set the model parameters appropriately to avoid overfitting or underfitting, ensuring that the prediction model accurately reflects the relationship between nodal features and microwave absorption performance, thereby making the calculated contribution value reliable.

[0060] When performing intersection matching, it's crucial to pay attention to how node combinations are defined. A node combination can be a pair of two nodes or a combination of multiple nodes, depending on the graph structure and analysis requirements. For each node combination, it's necessary to clearly define the nodes it contains and the connections between them to ensure the accuracy of the intersection matching.

[0061] The entire identification process needs to consider the dynamic updating of the spectrum. When new nodes and edge weight information are added to the material property spectrum, it is necessary to re-analyze the degree of influence of edge weights, calculate contribution values, and perform intersection matching to ensure that the identified key structural combinations can reflect the latest spectrum information, thereby providing accurate constraint references for the optimized design of microwave absorbing materials.

[0062] By following the steps above, key structural combinations that play a dominant role in microwave absorption performance can be identified from the material property spectrum. These key structural combinations reflect the parts of the material configuration characteristics and composite layout that have the greatest impact on microwave absorption performance. Using them as constraint references can guide the optimization direction in the subsequent optimization process, making the setting of optimization variable constraint space and the selection of configuration schemes more targeted, thereby improving the efficiency and effectiveness of microwave absorption material optimization design.

[0063] Example 4: When setting the optimization variable constraint space, which includes material type, thickness, morphology, and arrangement sequence, specific constraints need to be gradually clarified from multiple dimensions. Taking a specific absorption frequency band requirement as an example, the range of selectable types of absorbing materials is determined. This range covers three major categories: carbon-based materials, magnetic materials, and composite nanomaterials. Carbon-based materials can specifically include graphene, carbon nanotubes, and carbon fibers. For example, in high-frequency absorption scenarios, graphene is included as a selectable type because it has good conductivity and a large specific surface area, and can effectively absorb electromagnetic waves through dielectric loss. Magnetic materials can include ferrites, metallic magnetic powders, etc. Nickel-zinc ferrites have suitable permeability at specific frequency bands and can consume electromagnetic wave energy through magnetic loss, so they are also included in the selectable range. Composite nanomaterials, such as ferrite-graphene composite nanomaterials, combine the advantages of magnetic loss of magnetic materials and dielectric loss of carbon-based materials, and may have good application potential in broadband absorption, so they are also considered as one of the selectable types.

[0064] After identifying the material type, it is necessary to set the thickness range for each type of material. The determination of the thickness range is closely based on the required absorption frequency band of the material. Taking microwave absorbing materials operating in the 2-4 GHz frequency band as an example, for graphene, theoretical analysis of the propagation characteristics and loss mechanism of electromagnetic waves in the material shows that when the thickness of the graphene layer is 5-15 μm, its dielectric loss peak matches the electromagnetic waves in this frequency band well, achieving a good absorption effect. Therefore, the thickness range of graphene is set to 5-15 μm. For nickel-zinc ferrite, considering the relationship between its magnetic loss characteristics and material thickness, in the 2-4 GHz frequency band, when the thickness is 10-25 μm, the imaginary part of the permeability is large, and the magnetic loss is significant. Therefore, the thickness range of nickel-zinc ferrite is set to 10-25 μm. For ferrite-graphene composite nanomaterials, due to the synergistic effect of their composite structure, the thickness range may differ from that of single materials. Analysis shows that in this frequency band, a thickness range of 15-30 μm can achieve a good balance between the dielectric loss and magnetic loss of the composite system. Therefore, its thickness range is set to 15-30 μm.

[0065] The available morphologies of materials are limited to granular, sheet-like, and fibrous forms. Different morphologies have varying effects on the microwave absorption performance of materials. For example, granular structures like ferrite particles, with their spherical or near-spherical shapes, can be uniformly dispersed in composite systems, facilitating multiple scattering of electromagnetic waves within the material and increasing energy loss paths. Sheet-like graphene, with its two-dimensional planar structure, can form a conductive network in composite materials, enhancing dielectric loss. Simultaneously, the interlayer interfaces of the sheet-like structure can also reflect and scatter electromagnetic waves. Fibrous carbon fibers, with their high aspect ratio, can be arranged along certain directions in composite materials, forming continuous conductive channels. This not only improves the material's conductivity but also increases electromagnetic wave loss through the interaction between fibers. In practical implementation, the appropriate morphology must be selected based on the absorption frequency band and the characteristics of the material type. For instance, for carbon-based materials requiring enhanced dielectric loss, sheet-like graphene or fibrous carbon fibers are preferred; for magnetic materials requiring high magnetic loss, granular ferrite can be chosen.

[0066] The rules governing the arrangement of materials in a composite system include alternating arrangements, layered stacking, and gradient distribution. Taking layered stacking as an example, when designing a composite microwave absorbing material containing graphene and nickel-zinc ferrite layers, the graphene and nickel-zinc ferrite layers can be stacked alternately. The thickness of each layer is selected according to a pre-defined range, such as 10 μm for the graphene layer and 15 μm for the nickel-zinc ferrite layer. This alternating arrangement causes multiple reflections and refractions due to differences in the electromagnetic parameters of the materials as electromagnetic waves propagate between different material layers, increasing energy loss. Gradient distribution is applied, for example, when designing a composite microwave absorbing material where the electromagnetic parameters gradually change from the surface to the bottom layers. The surface layer uses a thicker... Thin graphene materials (e.g., 8 μm thick), a middle layer of ferrite-graphene composite nanomaterials (20 μm thick), and a bottom layer of thicker nickel-zinc ferrite material (25 μm thick) are used to gradually increase the dielectric constant and magnetic permeability of the materials from the surface to the bottom, forming a gradient distribution. This arrangement can effectively reduce the reflection of electromagnetic waves on the material surface, allowing more electromagnetic waves to enter the material and be absorbed. The alternating arrangement rule can also be applied to composite systems of three materials, such as graphene layers, nickel-zinc ferrite layers, and carbon fiber layers arranged alternately, with the thickness of each layer selected within its respective range, to achieve the synergistic effect of multiple loss mechanisms.

[0067] When setting the constraint space for optimization variables, attention must be paid to the correlation between the constraints in each dimension. For example, the choice of material type will affect the setting of the thickness range; different types of materials have different suitable thickness ranges due to their different electromagnetic properties. The morphology of the material will affect its arrangement sequence in the composite system; sheet-like materials are more suitable for layered stacking or gradient distribution, while particulate materials are more suitable for uniform dispersion in the matrix. At the same time, the feasibility of the actual preparation process must also be considered. The set thickness range and arrangement sequence should be within the capabilities of existing preparation technologies. For example, the thickness of each layer in a layered stacked structure must meet the accuracy requirements of the preparation equipment, and the parameter variation gradient of a gradient distribution structure must conform to the controllable range of the material synthesis process.

[0068] The setting of the optimization variable constraint space is not static and can be adjusted according to actual needs and new experimental data. When the absorption frequency band changes, the selectable range of material types and the thickness range may need to be redefined. When a new morphological structure or arrangement sequence rule is found to improve the absorption performance, it can be included in the constraint space to expand the optimization possibilities. Through this dynamic adjustment, the optimization variable constraint space can always adapt to different absorption requirements and technological developments, providing accurate and reasonable constraints for selecting configuration schemes with high absorption performance prediction scores in the material performance spectrum. This ensures that the optimized configuration scheme not only meets the absorption performance requirements but also has practical fabrication feasibility.

[0069] Example 5: When selecting configurations with high predicted absorption performance scores based on the optimization-guided algorithm in the material property spectrum, an initial score value should be assigned to each node combination in the spectrum. The initial score value should be set based on the absorption intensity parameter corresponding to the node combination. For example, the reflection loss value marked on the node combination can be directly used as the initial score. If the reflection loss of a node combination corresponding to the absorption intensity parameter in a specific frequency band is -30dB, then its initial score value is set to 30. When assigning initial scores, it is necessary to ensure that the score value of each node combination can intuitively reflect the relative superiority or inferiority of its current absorption performance. For node combinations with smaller reflection loss values ​​(i.e., better absorption performance), a higher initial score value should be assigned.

[0070] The initial score is adjusted based on the matching degree between the absorption intensity parameters corresponding to the node combination and the optimization variable constraint space, thus generating the predicted score. This process involves matching degree calculations across multiple dimensions. First, the matching degree between the material type in the node combination and the range of selectable types in the optimization variable constraint space is calculated. Assuming the selectable types in the optimization variable constraint space are carbon-based materials, magnetic materials, and composite nanomaterials, if the material type of a node combination is a composite of carbon-based and magnetic materials, then its matching degree is 100%. If the material type includes other materials not within the selectable range, the matching degree is calculated based on the proportion of materials within the selectable range. For example, if a node combination consists of carbon-based materials and unknown materials, with carbon-based materials accounting for 50%, then the matching degree is 50%.

[0071] Calculate the degree of conformity between the material thickness and the thickness range in the optimization variable constraint space. Taking the thickness of carbon-based materials in a certain node combination as an example, if the thickness range of carbon-based materials in the optimization variable constraint space is 5-15 μm, and the thickness of carbon-based materials in this node combination is 8 μm, which is within the range, then the conformity is 100%. If the thickness is 20 μm, exceeding the upper limit of the range, the conformity is calculated based on the degree of deviation. For example, if it exceeds the upper limit by 5 μm, the conformity can be reduced by 10% for every 1 μm exceeding the upper limit, resulting in a conformity of 50%.

[0072] Calculate the consistency degree between the morphological structure and the optional forms in the optimization variable constraint space. Optional forms include granular, sheet-like, and fibrous forms. If the morphological structure of the material in the node combination is a composite of sheet-like graphene and granular ferrite, then each morphological structure corresponds to an optional form, and the overall consistency degree is the average of the consistency degrees of each part. For example, if the consistency degree of the sheet-like structure is 100% and the consistency degree of the granular structure is 100%, then the overall consistency degree is 100%. If there are non-optional forms, such as rod-like structures, the consistency degree is determined based on the degree of similarity. For example, if the rod-like structure and the fibrous structure have a certain degree of similarity, the consistency degree can be set to 70%.

[0073] Calculate the fit between the permutation sequence and the permutation sequence rules in the optimization variable constraint space. The permutation sequence rules include alternating arrangement, layered stacking, and gradient distribution. If the permutation sequence of the node combination is an alternating arrangement of graphene layers and ferrite layers, which conforms to the alternating arrangement rule, the fit is 100%. If the permutation sequence is a random mixture, the fit is determined according to the degree of matching with each rule. For example, if there is some similarity with the layered stacking rule, the fit can be set to 60%.

[0074] The weighted sum of the matching degree, conformity degree, consistency degree, and fit degree is used to obtain the adjusted prediction score. The weights should be set according to the degree of influence of each dimension on the absorption performance. For example, the material type has a more critical influence on the absorption performance, and its weight can be set to 0.4. The weights of thickness, morphology, and arrangement sequence are 0.3, 0.2, and 0.1, respectively. Assuming that the matching degree of a certain node combination is 100%, the conformity degree is 80%, the consistency degree is 90%, and the fit degree is 70%, then its prediction score is: 100%×0.4+80%×0.3+90%×0.2+70%×0.1=0.4+0.24+0.18+0.07=0.89, that is, the prediction score is 89 points.

[0075] After calculating the predicted scores for all node combinations, the combinations are sorted from highest to lowest score. The sorting process must ensure the accuracy of the score calculations and the consistency of the sorting logic. For node combinations with the same score, a secondary sort can be performed based on the specific values ​​of their absorption intensity parameters or other secondary dimensions. The top preset number of node combinations in the sorted results are selected as candidate configuration schemes. This preset number can be determined based on actual needs, such as selecting the top 10 node combinations to ensure sufficient candidate schemes for subsequent screening.

[0076] The scheme with the highest prediction score among the candidate configuration schemes is selected as the final recommended configuration scheme. For example, the highest-scoring node combination among the candidate configuration schemes has a prediction score of 95 points. Its material type is a combination of carbon-based materials and composite nanomaterials, with thicknesses of 10 μm and 20 μm, respectively. Its morphological structure is plate-like and granular, and its arrangement sequence is a gradient distribution. This scheme has the highest comprehensive matching degree under the premise of satisfying the optimization variable constraint space, so it is selected as the final recommended configuration scheme.

[0077] During implementation, it is crucial to maintain consistency in the initial score allocation criteria to avoid score deviations due to inconsistent standards. The calculation methods for each dimension's matching degree, conformity, consistency, and fit must be clear and repeatable to ensure the objectivity of the predicted scores. Weight settings should be based on domain knowledge and practical experience; if necessary, multiple weight settings can be compared and analyzed to select the weight combination that best reflects the importance of each dimension. The selection of the preset number of options must balance optimization efficiency and solution diversity. Too few preset options may miss potential configuration solutions; too many preset options will increase the workload of subsequent screening.

[0078] The implementation of the optimized guidance algorithm needs to be combined with the dynamic updating of material property spectra. When new node combinations are added to the spectra or the absorption intensity parameters of existing node combinations are updated, the prediction scores need to be recalculated and reordered to ensure that the recommended configuration schemes are always based on the latest spectra information. At the same time, the recommended configuration schemes need to consider the feasibility of actual fabrication processes. This can be achieved by introducing a fabrication feasibility dimension into the prediction score calculation, or by conducting a fabrication feasibility assessment after candidate scheme screening, to avoid recommending schemes with high theoretical scores but which cannot be actually fabricated.

[0079] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0080] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A microwave absorption optimization method based on spectrum recognition, characterized in that, The method includes: Obtain configuration data information for various types of microwave absorbing materials; Using the configuration data as the foundation, a material performance map containing material configuration characteristics and composite layout relationships is constructed. Graph structure analysis is performed on the material performance map to extract topological dependencies between nodes and complete multi-hop information transmission. Based on the topological dependencies and multi-hop information transmission results, key structural combinations that dominate the microwave absorption performance are identified. These key structural combinations are used as constraint references to set an optimization variable constraint space containing material type, thickness, morphology, and arrangement sequence. Using the material performance map and the optimization variable constraint space as screening conditions, configuration schemes with high microwave absorption performance prediction scores are selected from the material performance map based on an optimization-guided algorithm. Based on the configuration data information, a material property map containing the material configuration characteristics and composite layout relationship is constructed, including: Collect configuration feature data of different microwave absorbing materials, the configuration feature data including material microstructure parameters and macroscopic composite layout information; record the composite layout relationship between each material in the configuration feature data; label the corresponding microwave absorption intensity parameters for the configuration feature data; use the labeled configuration feature data as graph nodes and the composite layout relationship as graph edge weights to establish a material performance graph that associates nodes with edge weights; Define an optimization variable constraint space that includes material type, thickness, morphology, and arrangement sequence, including: The range of possible types of absorbing materials is determined, including carbon-based materials, magnetic materials, and composite nanomaterials; the thickness range of each type of material is set, and the range is determined according to the requirements of the material's absorbing frequency band; the optional morphology of the material's microstructure is limited, including granular, sheet-like, and fibrous forms; and the arrangement sequence rules of the materials in the composite system are specified, including alternating arrangement, layered stacking, and gradient distribution. Based on the material property spectrum, an optimization-guided algorithm is used to select configuration schemes with high microwave absorption performance prediction scores, including: An initial score is assigned to each node combination in the material performance spectrum; the initial score is adjusted to generate a predicted score based on the matching degree between the absorption intensity parameter corresponding to the node combination and the optimization variable constraint space; the node combinations are sorted from high to low according to the predicted scores, and the top preset number of node combinations in the sorting results are selected as candidate configuration schemes; the scheme with the highest predicted score among the candidate configuration schemes is selected as the final recommended configuration scheme.

2. The absorption optimization method based on spectrum recognition as described in claim 1, characterized in that, Perform graph structure analysis on the material property map to extract topological dependencies between nodes and complete multi-hop information transfer, including: The node features and edge weights of the material property map are convolved using graph convolution operations to extract the topological dependencies between nodes. The topological dependencies are weighted using a graph attention mechanism to generate attention weights that include multi-hop neighborhood information. Based on the attention weights, multi-hop information is transferred between nodes, and node feature representations that fuse multi-hop information are output.

3. The absorption optimization method based on spectrum recognition as described in claim 2, characterized in that, Based on the aforementioned topological dependencies and multi-hop information transmission results, key structural combinations that play a dominant role in absorption performance are identified, including: Analyze the influence of each node's connection edge weight in the topological dependency relationship, and filter the node combinations corresponding to edge weights whose influence exceeds a preset threshold; combine the contribution values ​​of each node's feature representation in the multi-hop information transmission results, and extract node combinations whose contribution values ​​are higher than a preset contribution value; perform intersection matching between the filtered node combinations and the extracted node combinations, and output the matching result as the key structure combination.

4. The absorption optimization method based on spectrum recognition as described in claim 1, characterized in that, Based on the degree of matching between the absorption intensity parameters corresponding to the node combination and the optimization variable constraint space, the initial score value is adjusted to generate a predicted score, including: Calculate the matching degree between the material type in the node combination and the range of selectable types in the optimization variable constraint space; calculate the conformity between the material thickness in the node combination and the thickness value range in the optimization variable constraint space; calculate the consistency between the morphology and structure in the node combination and the selectable forms in the optimization variable constraint space; calculate the fit between the permutation sequence in the node combination and the permutation sequence rule in the optimization variable constraint space; sum the matching degree, conformity, consistency, and fit by weight, and output the adjusted prediction score.

5. A microwave absorption optimization system based on spectrum recognition, used to implement the microwave absorption optimization method based on spectrum recognition as described in any one of claims 1 to 4, characterized in that, The system includes: The spectrum construction module is used to acquire configuration data information of multiple types of microwave absorbing materials and construct a material performance spectrum that includes the material configuration characteristics and the relationship between the composite layout. The graph recognition module is used to perform graph structure analysis on the material property spectrum, extract the topological dependencies between nodes and complete multi-hop information transmission, and identify key structural combinations based on the transmission results. An optimization guidance module is used to set an optimization variable constraint space including material type, thickness, morphology, and arrangement sequence, and select a configuration scheme with a high microwave absorption performance prediction score from the material performance spectrum based on the optimization guidance algorithm. A configuration recommendation module is configured to output the high-scoring configuration scheme selected by the optimization guidance module.

6. The absorption optimization system based on spectrum recognition as described in claim 5, characterized in that, The spectrum construction module includes a data collection unit, a relationship recording unit, a parameter annotation unit, and a spectrum building unit. The data collection unit is used to collect configuration feature data of different absorbing materials. The relationship recording unit is used to record the composite layout relationship between each material in the configuration feature data. The parameter annotation unit is used to annotate the corresponding absorbing intensity parameters for the configuration feature data. The spectrum building unit is used to use the annotated configuration feature data as spectrum nodes and the composite layout relationship as spectrum edge weights to build a material performance spectrum that associates nodes with edge weights.

7. The absorption optimization system based on spectrum recognition as described in claim 5, characterized in that, The graph recognition module includes a topology extraction unit, an information transmission unit, and a structure filtering unit; the topology extraction unit is used to perform graph convolution operations on the node features and edge weight information of the material property graph to extract the topological dependencies between nodes. The information transmission unit is used to assign weights to the topological dependencies through a graph attention mechanism to complete multi-hop information transmission between nodes; The structure screening unit is used to screen key structural combinations that play a dominant role in the absorption performance based on the topological dependency relationship and the multi-hop information transmission results.

Images