Preparation method of functional polyimide composite material with machine learning-assisted Janus structure

Abstract

The invention belongs to the technical field of high polymer material manufacturing, and discloses a preparation method of a functional polyimide composite material of a machine learning assisted Janus structure, and the method comprises the following steps: S1, establishing a database; s2, data preprocessing; s3, designing a basic structure and a preparation process; s4, constructing and training a machine learning model; s5, testing and evaluating the model; s6, feature importance analysis; s7, optimization design of the composite material; and S8, experimental verification. The composite material and the preparation process thereof are obtained by applying a machine learning algorithm model, analyzing the influence of various factors on the performance of the functional polyimide composite material and evaluating the degree of influence of key characteristics on the performance, the material has a functional layered structure, the surface layer serves as an impedance matching layer, and the middle layer serves as an electromagnetic wave absorption layer. The substrate layer serves as a reflecting layer, the structure is packaged completely after hot pressing, the overall wave absorbing efficiency and thermal management performance of the material are improved, and the manufacturing requirements of various high-end electronic devices can be met.

Description

Technical Field

[0001] The present invention relates to the technical field of polymer material manufacturing, and in particular to a method for preparing a functional polyimide composite material with a machine learning-assisted Janus structure. Background Art

[0002] Absorbing materials have recently become the focus of key technology areas, especially in improving the clarity of wireless signal transmission, reducing electromagnetic noise interference, and optimizing the electromagnetic compatibility of electronic devices. With the continuous advancement of technology, these materials are rapidly developing in the direction of being lighter, thinner, and more efficient. Research and development of advanced absorbing materials plays a vital role in ensuring the security and reliability of information communications and in future technological innovation and applications.

[0003] The current research focus of absorbing materials is often on building efficient conductive networks. However, relying solely on highly conductive networks to manufacture electromagnetic shielding materials is not a perfect solution. For example, although most metals (such as copper, silver, and gold) and their composite materials (including metal fabrics, metal foams, and metal laminates) effectively reflect electromagnetic radiation due to their ultra-high conductivity, they may also cause secondary environmental pollution. In contrast, carbon-based materials (such as graphene, carbon nanotubes, and carbon black) have become the dominant materials for electromagnetic wave absorption due to their good dielectric constants. However, these materials alone often fail to meet the needs for broadband and high-intensity absorption performance. In this context, the importance of studying thermal conductivity cannot be ignored. Effective thermal management is key to ensuring that electronic devices maintain stability and extend their life during high-performance operation. Especially in dense electronic components and microchips, excellent thermal conductive materials can prevent heat accumulation, thereby avoiding performance degradation and device damage caused by overheating. Therefore, materials that have both absorbing and good thermal conductivity are particularly important in modern electronic and communication equipment. Such materials can not only absorb and shield unnecessary electromagnetic interference, but also effectively manage heat flow and improve the overall performance and safety of the equipment. However, due to the high complexity and uncertainty in the design of components, microstructures and preparation processes of materials with both wave absorption and good thermal conductivity, and the wide range of factors involving materials, ratios, structures, temperature, pressure, time and many other factors that influence each other and change in a complex manner, traditional design and experimental methods, as well as conventional AI technologies, are unable to achieve the prediction and optimization of functional polyimide composite materials with a specific optimal combination of wave absorption and good thermal conductivity. Summary of the invention

[0004] The present invention is to solve the problems of the prior art and overcome the above-mentioned shortcomings of the prior art. The present invention provides a method for preparing a functional polyimide composite material with a machine learning-assisted Janus structure. The machine learning technology is used to construct a model that can predict the functional polyimide composite material. The model optimizes the design of the composite material and the preparation process based on the functional layered Janus structure by evaluating the degree to which key features affect the performance. The material properties and interactions of each layer are precisely controlled to significantly improve the absorption efficiency of electromagnetic waves, and at the same time enhance the thermal conductivity of the material to improve development efficiency and reduce test costs. A new technical concept is provided for manufacturing more efficient wave-absorbing and heat-conducting integrated materials with specific composite performance combination requirements to meet the needs of manufacturing high-end and precision electronic equipment with dual requirements of electromagnetic compatibility and thermal management.

[0005] To achieve the above object, the technical solution adopted by the present invention is: A method for preparing a functional polyimide composite material with a machine learning-assisted Janus structure comprises the following steps: S1. Establish database Collect known functional polyimide composite material information, including classification characteristics, numerical characteristics and preparation process, and establish a database; S2. Data Preprocessing Convert categorical features into binary sub-category features ("0" and "1"), preprocess the data, and divide it into training and test sets according to the ratio; S3: Design of basic Janus structure and preparation process The basic structure of the polyimide composite material to be prepared, including a Janus surface layer, an intermediate layer and a base layer, is designed, and then the basic components, proportions and preparation process of the composite material are designed. The process is as follows: each layer of the composite material is prepared, and then hot-pressed; S4. Build and train machine learning models Select a machine learning algorithm, use the training set to build a machine learning prediction model, use the characteristic information of the composite material as input parameters, evaluate the accuracy of the machine learning prediction model with evaluation indicators, and use cross-validation to screen the best machine learning prediction model; S5. Model testing and evaluation Use the test set to test the accuracy of the machine learning model and use relevant evaluation indicators to evaluate the performance. Use the Pearson correlation coefficient to analyze and verify the importance of features to ensure the accuracy and reliability of the model; S6. Feature Importance Analysis The recursive feature elimination method is combined with the machine learning prediction model to effectively identify key feature variables. Based on the analysis results of the Pearson correlation coefficient, these key features are further adjusted and the model parameters and feature selection are optimized to accurately adjust the model and ensure the highest prediction accuracy and improve the value of practical applications, and obtain model prediction results and key features; S7. Optimal design of composite materials Use machine learning to assist in optimizing composite material selection and structural design. Based on the model prediction results and key features obtained in the above steps, design a Janus structure polyimide composite material with microwave absorption and thermal conductivity and its preparation process. According to the key features identified by the machine learning model, determine the Janus structure, components and ratios of each layer, and preparation process of the optimized composite material. S8. Experimental verification Janus structure composite material samples were prepared, and control samples with reverse characteristics were prepared simultaneously. Then, comprehensive performance tests were performed on these samples to verify the wave absorption and thermal conductivity of the composite material samples. Those that met the design requirements were selected as the optimized functional polyimide composite materials.

[0006] A functional polyimide composite material is obtained by using the aforementioned machine learning-assisted design method, and comprises: a surface layer, an intermediate layer and a base layer stacked in sequence, wherein the surface layer uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves enter the interior of the material without being reflected; the intermediate layer uses a boron nitride coating modified with ferroferric oxide as an electromagnetic wave absorption layer, the ferroferric oxide causes eddy current loss, converts electromagnetic energy into thermal energy, and the introduction of high thermal conductivity boron nitride promotes rapid dissipation of thermal energy, thereby improving the overall wave absorption efficiency and thermal management performance; the base layer introduces a carbonized graphene / carbon fiber felt layer as a reflection layer to fully reflect the electromagnetic waves that are not absorbed into the absorption layer for multiple absorption; the wave absorption layer and the conductive layer are filled with polyimide resin to improve the structural stability of the material; finally, by hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure, thereby obtaining a functional polyimide composite material with excellent electromagnetic wave management and efficient thermal management.

[0007] The method for preparing the functional polyimide composite material with machine learning-assisted Janus structure provided by the present invention has the following obvious outstanding substantial characteristics and significant advantages compared with the prior art: 1. The functional polyimide composite material and design method provided by the present invention can cope with the challenges of difficult effective prediction and optimized design brought about by the complex changes in various aspects of the material. Based on machine learning technology, especially using AI models such as the XGBoost algorithm and preprocessed data, an accurate prediction model is constructed and trained. Machine learning understands and analyzes how the parameter characteristics of various materials affect the final material performance, thereby optimizing the composition and microstructure of the material, and ultimately efficiently and intelligently designing an integrated wave-absorbing and heat-conducting material with specific performance combination requirements; this intelligent design process not only improves development efficiency, but also reduces the cost of experiments, providing new technical concepts for manufacturing better performance parameter combinations and more efficient integrated wave-absorbing and heat-conducting materials.

[0008] 2. The composite material and method provided by the present invention adopt machine learning technology to construct a model capable of predicting functional polyimide composite materials. The precise machine learning model obtained through pre-training analyzes the importance of material characteristics. The model optimizes the design of composite materials and preparation processes based on functional layered Janus structures by evaluating the degree to which key characteristics affect performance. The material properties and interactions of each layer are precisely controlled to significantly improve the absorption efficiency of electromagnetic waves while enhancing the thermal conductivity of the material. This intelligent material design process significantly shortens the R&D cycle, can effectively improve the efficiency of material R&D, shorten the R&D cycle, and significantly save scientific research costs such as design and testing. It provides a new technical concept for the development of a series of composite materials with specific performance combinations for efficient electromagnetic wave management and thermal management.

[0009] 3. The composite material and method provided by the present invention, the multilayer microstructure of the composite material obtained by the optimized design of the model can achieve efficient electromagnetic wave absorption and shielding through the special design of different layers; the porous structure design of the surface layer reduces the reflection of electromagnetic waves and allows electromagnetic waves to enter the interior of the material, the middle absorption layer converts electromagnetic energy into thermal energy through eddy current loss, and the conductive layer of the substrate layer enhances the electromagnetic shielding capability, while reflecting the unabsorbed electromagnetic waves back to the middle layer for reabsorption. This synergistic effect significantly improves the overall absorption and shielding efficiency of the composite material, and provides a new technical concept for the industrial preparation of more efficient wave-absorbing and heat-conducting integrated materials with specific composite performance combination requirements. The composite material can be widely used in the manufacture of high-end, precision electronic equipment (especially communication equipment) with dual requirements of electromagnetic compatibility and thermal management, and can ensure the security and reliability of information communication.

[0010] 4 The composite material and method provided by the present invention, the composite material is infused with polyimide resin in the absorbing coating and the conductive layer, and the different layers are firmly connected by hot pressing technology, thereby enhancing the overall structural stability and durability of the material. This structural design not only ensures the effective blocking of electromagnetic waves, but also ensures the mechanical strength and environmental adaptability of the material in long-term use.

[0011] 5. The composite material and method provided by the present invention can significantly reduce the time and resources required for traditional trial and error methods by using machine learning models to predict the best combination of material components, microstructures, hot pressing process parameters, etc. Through the optimization design predicted by the model, the technical solution verified by the test can directly enter the production process, avoiding unnecessary waste of materials and time delays, thereby improving production efficiency and reducing costs. The processing technology of the polyimide composite obtained by the optimal process parameters is more perfect, the mechanical properties of the polyimide composite are better, and the service life is longer.

[0012] 6. The composite material and method provided by the present invention have highly customized model optimization. By using multiple machine learning models to fit and predict the relationship between the components, microstructures, hot pressing process parameters and tensile strength and its corresponding thermal stability of each layer of the polyimide composite, the model with the best prediction performance can be identified. This not only improves the accuracy of the model, but also ensures the precise adjustment of process parameters and maximizes the performance of the material by selecting the most suitable model.

[0013] 7. Compared with traditional physical experimental methods, the use of machine learning models in the present invention reduces the demand for chemicals for experiments and tests and potential environmental pollution; the application of this technology is in line with the development concept of green environmental protection and supports sustainable development goals. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a graph of prediction results of the machine learning prediction model in the test set according to an embodiment of the present invention; Figure 2 It is a heat map of the Pearson correlation coefficients between the characteristic variables of the embodiment of the present invention; Figure 3 A schematic diagram of a process for preparing a functional polyimide composite material provided by the present invention; Figure 4 is the reflection coefficient R of Example 2 of the present invention and Comparative Example 1; Figure 5 SER values ​​of different impedance matching layers in the embodiments of the present invention. DETAILED DESCRIPTION

[0015] In order to accurately describe the purpose, technical solutions and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0016] The above scheme is further described below in conjunction with the accompanying drawings and specific embodiments. The equipment and models in each embodiment adopt existing technologies, and all materials are commercially available products.

[0017] Basic Example The method for preparing a functional polyimide composite material with a Janus structure assisted by machine learning provided in this embodiment comprises the following steps: S1. Establish database Collecting known functional polyimide composite material information, including classification characteristics, numerical characteristics and preparation process, and establishing a database; specifically, by collecting information related to a variety of known functional polyimide composite materials to establish a database, the information and database include: the structure of the impedance matching layer (including porous structure and film structure), the filler of the impedance matching layer (carbon material and magnetic material), the absorber (including Fe3O4 and other absorbing materials), the stacking order of layers (such as impedance matching layer-absorbing layer-conductive layer, etc.), the molding process (including freeze drying, hot pressing and coating), the material structure, the conductive layer filler (such as carbon material and magnetic material), the thickness and the reflection coefficient R; S2. Data Preprocessing The classification features are converted into binary sub-category features ("0" and "1"), the data are preprocessed, and divided into a training set and a test set according to a ratio; specifically, the classification features in step S1 are converted into binary sub-category features ("0" and "1"), where "0" indicates that the sample does not belong to the sub-category and "1" indicates that the sample belongs to the sub-category; the conversion results are as follows: structure of the impedance matching layer: porous structure (1), thin film structure (0); filler of the impedance matching layer: carbon material (1) and magnetic material (0); absorber: Fe3O4 (1), other absorbing materials (0); stacking order of layers: impedance matching layer-absorbing layer-conductive layer (1), other stacking order (0); molding process: freeze drying, hot pressing and coating; material structure: asymmetric (1), symmetric (0); conductive layer filler: carbon material (1), magnetic material (0), etc.; S3: Design of basic Janus structure and preparation process The basic structure of the polyimide composite material to be prepared, including the surface layer, the middle layer and the base layer forming the Janus structure, is designed, and then the basic components, proportions and preparation process of the composite material are designed. The process is as follows: each layer of the composite material is prepared, and then hot-pressed; S4. Build and train machine learning models Select a machine learning algorithm, use the training set to build a machine learning prediction model, use the characteristic information of the composite material as an input parameter, evaluate the accuracy of the machine learning prediction model with an evaluation index, and use cross-validation to screen the best machine learning prediction model; the machine learning algorithm includes: one or more of the extreme gradient boosting (XGBoost) method, the ridge regression (RidgeRegresso) method, the gradient boosted decision tree (GBDT) method, and the random forest regression method; S5. Model testing and evaluation Use the test set to test the accuracy of the machine learning model and use relevant evaluation indicators to evaluate the performance. Use the Pearson correlation coefficient to analyze and verify the importance of features to ensure the accuracy and reliability of the model; In machine learning, the Ridge Regressor linear regression algorithm is first used, and the correlation coefficient R and the root mean square error (RMSE) are used as indicators to evaluate the accuracy of the model. On the training set, the regression coefficient between the reflection coefficient R predicted by the optimal model obtained by this algorithm and the actual value is 0.74, and the root mean square error is 0.19. When this model is applied to the test set, the regression coefficient between the predicted value and the actual value is 0.67, and the root mean square error is 0.18, indicating that the model has certain predictive ability on the test set, but there is room for improvement.

[0018] Subsequently, the Extreme Gradient Boosting (XGBoost) algorithm was introduced. The results of applying this method to the training set showed that the regression coefficient between the reflection coefficient R predicted by the model and the actual value was increased to 0.84, and the root mean square error was reduced to 0.11. The regression coefficient applied to the test set reached 0.89, and the root mean square error was 0.12. These results show that the XGBoost model is superior to the ridge regression model in terms of accuracy and stability, showing higher prediction accuracy, and is more suitable for predicting the reflection coefficient R of composite materials, assisting the design and development of composite materials.

[0019] S6. Feature Importance Analysis The recursive feature elimination method is combined with the machine learning prediction model to effectively identify key feature variables. Based on the analysis results of the Pearson correlation coefficient, these key features are further adjusted and the model parameters and feature selection are optimized to accurately adjust the model and ensure the highest prediction accuracy and improve the value of practical applications, and obtain model prediction results and key features; S7. Optimal design of composite materials Use machine learning to assist in optimizing composite material selection and structural design. Based on the model prediction results and key features obtained in the above steps, design a Janus structure polyimide composite material with microwave absorption and thermal conductivity and its preparation process. According to the key features identified by the machine learning model, determine the Janus structure, components and ratios of each layer, and preparation process of the optimized composite material. Among them, the optimization design of the functional layered Janus structure composite material is to regulate the dielectric properties of the surface layer, the middle layer and the base layer. The difference in conductivity makes each surface show different electromagnetic wave reflection characteristics, including the following steps: The surface of S7-1 uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves can enter the interior of the material without being reflected; The surface polyimide felt is not filled with resin, maintaining its porous structure, thereby allowing electromagnetic waves to enter; the middle layer and the base layer are filled with polyimide resin to prevent electromagnetic waves from penetrating; The S7-2 intermediate layer uses a boron nitride coating modified with ferroferric oxide as the electromagnetic wave absorption layer. The ferroferric oxide causes eddy current loss, converting electromagnetic energy into thermal energy, while the introduction of high thermal conductivity boron nitride promotes rapid dissipation of thermal energy, thereby improving the overall wave absorption efficiency and thermal management performance. The S7-3 substrate layer introduces a carbonized graphene / carbon fiber felt layer as a reflective layer, which fully reflects the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; S7-4 fills the absorbing layer and the conductive layer with polyimide resin to improve the structural stability of the material; and through hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure, thereby obtaining an optimized structure of the polyimide composite material with excellent electromagnetic wave management and efficient thermal management.

[0020] The optimized design preparation method comprises the following steps: A1. Prepare the base layer as the reflective layer: evenly coat the graphene solution on the surface of the polyimide (PI) felt, stack and prepare a multi-layer polyimide felt / graphene structure, put the multi-layer graphene / PI felt structure into a special mold, apply vertical downward pressure to ensure that the layers are tightly bonded, and perform high-temperature carbonization treatment; A2. Preparation of an intermediate layer as an absorption layer: Boron nitride (BN) powder is exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare functionalized boron nitride nanosheets (f-BNNSs), and f-BNNSs and ferroferric oxide (Fe3O4) are evenly mixed and coated on the surface of the carbonized PI felt.

[0021] A3. Prepare the surface layer as the impedance matching layer: immerse the PI felt in a low concentration of graphene solution and perform ultrasonic treatment for half an hour to ensure that the graphene particles are evenly attached to the surface of the felt; A4. Hot pressing of each layer: PI solution is prepared by dissolving PI powder in an organic solvent, and the conductive layer coated with the absorber is immersed in the PI solution. The solvent is evaporated by programmed temperature increase; the impedance matching layer and the cured absorption layer-conductive layer are connected by hot pressing using polyimide resin as an adhesive. During the hot pressing process, the PI resin is further cured to form a seamless connection, and finally an optimized preparation method for functional polyimide composite materials is obtained.

[0022] S8. Experimental verification Janus structure composite material samples were prepared, and control samples with reverse characteristics were prepared simultaneously. Then, comprehensive performance tests were performed on these samples to verify the wave absorption and thermal conductivity of the composite material samples. Those that met the design requirements were selected as the optimized functional polyimide composite materials.

[0023] A functional polyimide composite material is obtained by using the aforementioned machine learning-assisted design method, and comprises: a surface layer, an intermediate layer and a base layer stacked in sequence, wherein the surface layer uses porous graphene / polyimide felt as an impedance matching layer to regulate the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves enter the interior of the material without being reflected; the intermediate layer uses a boron nitride coating modified with ferroferric oxide as an electromagnetic wave absorption layer, and the ferroferric oxide causes eddy current loss, converting electromagnetic energy into heat energy, and the introduction of high thermal conductivity boron nitride promotes rapid dissipation of heat energy, thereby improving the overall wave absorption efficiency and thermal management performance; the carbonized graphene / carbon fiber felt layer is introduced into the base layer as a reflective layer to fully reflect the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; the wave absorbing layer and the conductive layer are filled with polyimide resin to improve the structural stability of the material; finally, by hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure to obtain a functional polyimide composite material with excellent electromagnetic wave management and efficient thermal management; the number of layers of the multilayer polyimide felt / graphene structure is 3-x layers of graphene / PI felt arranged alternately (x≥5). In the electromagnetic shielding composite material, the thickness of the surface layer and the base layer are both 500μm-3000μm.

[0024] The functional polyimide composite material based on machine learning-assisted Janus structure design and the preparation method thereof provided in the embodiment of the present invention mainly analyzes the influence of various factors on the performance of the functional polyimide composite material by using the extreme gradient boosting (XGBoost) algorithm, evaluates the degree to which key features affect the performance, and designs and optimizes the Janus structure composite material based on functional stratification. The surface layer uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves enter the interior of the material without being reflected; the middle layer uses a boron nitride coating modified with ferroferric oxide as an electrical In the magnetic wave absorption layer, ferroferric oxide causes eddy current loss, converting electromagnetic energy into heat energy, while the introduction of high thermal conductivity boron nitride promotes rapid dissipation of heat energy, thereby improving the overall absorption efficiency and thermal management performance; the base layer introduces a carbonized graphene / carbon fiber felt layer as a reflective layer to fully reflect the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; the absorbing layer and the conductive layer are filled with polyimide resin to improve the structural stability of the material; and through hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure, and finally a functional polyimide composite material with excellent electromagnetic wave management and efficient thermal management is obtained.

[0025] Example 1 The method for preparing a functional polyimide composite material with a Janus structure assisted by machine learning provided in an embodiment of the present invention is a specific application based on the above-mentioned basic embodiment. It is to design a functional polyimide composite material based on a Janus structure and a preparation method thereof, so that the composite material finally has a specific performance combination, and uses composite materials with different structures and electromagnetic response characteristics as functional modules to achieve assembly according to the requirements of the target performance combination. Among them, a graphene / polyimide felt with a porous structure is used to adjust the surface impedance of the material and allow electromagnetic waves to be incident; a ferroferric oxide / functional boron nitride nanosheet coating is used as a microwave absorption layer to convert electromagnetic energy into thermal energy through eddy current loss, and the incorporation of thermally conductive boron nitride helps to dissipate heat energy while improving the overall thermal conductivity; carbonized graphene / carbon fiber felt is used as a conductive layer to ensure electromagnetic shielding performance and reflect electromagnetic waves that are not absorbed by the top and middle layers for reabsorption; the absorbing coating and the conductive layer are filled with polyimide resin to improve the stability of the material and prevent electromagnetic waves from penetrating; finally, through hot pressing technology, polyimide resin is used as an adhesive to connect different layers, and finally a functional polyimide composite material with a specific combination of properties is obtained.

[0026] A method for preparing a functional polyimide composite material having a machine learning-assisted Janus structure with a specific performance combination further adopts the following technical solutions on the basis of the basic embodiment: S1. Design of Janus structure of functional materials with specific performance combinations: Based on theory and preliminary data, a basic scheme of a three-layer polyimide composite material with integrated microwave absorption and thermal conductivity and Janus structure is proposed, including material composition, ratio, structure of each layer and basic preparation process, and then machine learning is used to assist in optimizing material selection, structural design and preparation process; S2. Establishing a database: According to the basic scheme of the three-layer polyimide composite material, known composite material related information is collected to establish a database, which includes classification features and numerical features; S3. Data preprocessing: Convert categorical features into binary sub-category features (“0” and “1”), preprocess the data, and divide the data into training and test sets in proportion; S4. Construct and verify the machine learning model: select a machine learning algorithm, use the training set to construct a machine learning prediction model, use the characteristic information of the composite material as the input parameter, evaluate the accuracy of the machine learning prediction model with the evaluation index, and use cross-validation to screen the best machine learning prediction model; the machine learning algorithm selects the XGBoost algorithm, such as Figure 1 As shown in the figure, the regression coefficient between the reflection coefficient R predicted by the best model obtained on the training set and the actual value is 0.84, and the root mean square error is about 0.11. The reflection coefficient R in the test set is predicted using the trained prediction model, and the regression coefficient between the model prediction value and the actual value reaches 0.89, and the root mean square error is about 0.12. It can be seen that the XGBoost model has a high accuracy and can be used to predict the reflection coefficient R and assist in the design and development of composite materials.

[0027] S5. Model testing and evaluation: Use the test set to test the accuracy of the machine learning model and use relevant evaluation indicators to evaluate the performance. Use the Pearson correlation coefficient to analyze and verify the importance of features to ensure the accuracy and reliability of the model; S6. Feature importance analysis: Recursive feature elimination combined with machine learning prediction model is used to effectively identify key feature variables. Based on the analysis results of Pearson correlation coefficient, these key features are further adjusted and model parameters and feature selection are optimized to accurately adjust the model and ensure the highest prediction accuracy and improve the value of practical applications.

[0028] S7. Design of functional materials with Janus structure: Based on the model prediction results and key features obtained in the above steps, design polyimide composite materials with optimized Janus structure and good microwave absorption and thermal conductivity; The process includes: regulating the dielectric properties of the top layer, the middle layer and the bottom layer. The difference in electrical conductivity makes each surface exhibit different electromagnetic wave reflection characteristics. The surface layer uses polyimide felt without resin filling to maintain its inherent porous structure, thereby allowing electromagnetic waves to enter. The middle layer and the base layer are filled with polyimide resin to prevent electromagnetic waves from penetrating.

[0029] S8. Experimental verification: Based on the key features identified by the machine learning model, prepare Janus structure composite material samples and simultaneously prepare control samples with reverse features. Perform comprehensive performance tests on these samples to verify their microwave absorption and thermal conductivity.

[0030] The preparation method of the optimized functional polyimide composite material with Janus structure comprises the following steps: A1. Preparation of reflective layer: The graphene solution is evenly coated on the surface of polyimide (PI) felt, and a multi-layer polyimide felt / graphene structure is prepared by stacking. The multi-layer graphene / PI felt structure is placed in a special mold, and vertical downward pressure is applied to it to ensure that the layers are tightly bonded, and then a high-temperature carbonization treatment is performed; The number of layers of the multilayer polyimide felt / graphene structure is 3-x layers of graphene / PI felt arranged alternately (x≥5), preferably 5-7 layers. As the number of alternating layers increases, the internal defects of the material also increase, reducing the effective absorption of energy.

[0031] A2. Preparation of absorption layer: Boron nitride (BN) powder is exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare functionalized boron nitride nanosheets (f-BNNSs). f-BNNSs and ferroferric oxide (Fe3O4) are evenly mixed and coated on the surface of carbonized PI felt.

[0032] A3. Preparation of impedance matching layer: immerse the PI felt in a low concentration of graphene solution and perform ultrasonic treatment for half an hour to ensure that the graphene particles are evenly attached to the surface of the felt. The thickness of the surface layer and the base layer are both between 500μm and 3000μm. Thinner absorbing materials cannot provide sufficient dielectric loss and magnetic loss, and have low mechanical strength; thicker absorbing materials will increase the weight of the overall structure.

[0033] A4. Composite materials are obtained by hot pressing: PI solution is prepared by dissolving PI powder in an organic solvent, the conductive layer coated with the absorber is immersed in the PI solution, and the solvent is evaporated by programmed temperature increase; the impedance matching layer and the cured absorption layer-conductive layer are connected by hot pressing with polyimide resin as an adhesive, and the PI resin is further cured during the hot pressing process to form a seamless connection.

[0034] The pressure applied in the hot pressing process is 0-20 MPa, the hot pressing temperature is 150° C.-200° C., and the hot pressing time is 20 min-60 min.

[0035] The functional polyimide composite material with machine learning-assisted optimization of Janus structure and the preparation method thereof provided in the embodiment of the present invention analyzes the influence of various factors on the performance of the functional polyimide composite material by using the extreme gradient boosting (XGBoost) algorithm, and evaluates the degree to which the key features affect the performance. A Janus structure based on functional layering is designed, in which the surface layer uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves enter the interior of the material without being reflected; the middle layer uses a boron nitride coating modified with ferroferric oxide as an electromagnetic wave absorber. The ferroferric oxide causes eddy current loss, converting electromagnetic energy into thermal energy, while the introduction of high thermal conductivity boron nitride promotes the rapid dissipation of thermal energy, thereby improving the overall wave absorption efficiency and thermal management performance; the carbonized graphene / carbon fiber felt layer is introduced into the base layer as a reflective layer to fully reflect the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; the wave absorption layer and the conductive layer are filled with polyimide resin to improve the structural stability of the material; and through hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure to obtain a polyimide composite material with excellent electromagnetic wave management and efficient thermal management.

[0036] The composite material sample prepared in this embodiment has a vertical thermal conductivity of 0.47 Wm - 1 K -1 ; At the same time, the reflection coefficient R obtained through testing is 0.49.

[0037] Example 2 See also Figure 3 , a schematic diagram of the process for preparing a functional polyimide-based composite material provided by an embodiment of the present invention.

[0038] The method for preparing a functional polyimide composite material with a Janus structure assisted by machine learning provided in an embodiment of the present invention is a specific application based on the above basic embodiment, which is basically the same as Example 1, except that: Firstly, based on theory and preliminary data, the basic structure and preparation method of a Janus structure polyimide composite material with integrated microwave absorption and thermal conductivity are conceived and proposed, and then machine learning is used to assist in optimizing material selection, structural design and preparation process. By collecting known information related to composite materials, a database is established, which contains the following information: the structure of the impedance matching layer (including porous structure and film structure), the filler of the impedance matching layer (carbon material and magnetic material), the absorber (including Fe3O4 and other absorbing materials), the stacking order of layers (such as impedance matching layer-absorbing layer-conductive layer, etc.), the molding process (including freeze drying, hot pressing and coating), the material structure, the conductive layer filler (such as carbon material and magnetic material), the thickness and the reflection coefficient R; some of these features are classification features, and some are numerical features. The classification features are converted into binary subcategory features ("0" and "1"), where "0" means that the sample does not belong to the subcategory and "1" means that the sample belongs to the subcategory. The conversion results are as follows: impedance matching layer structure: porous structure (1), thin film structure (0); impedance matching layer filler: carbon material (1) and magnetic material (0); absorber: Fe3O4 (1), other absorbers (0); layer stacking order: impedance matching layer-absorber layer-conductive layer (1), other stacking order (0); molding process: freeze drying, hot pressing and coating; material structure: asymmetric (1), symmetric (0); conductive layer filler: carbon material (1), magnetic material (0). Preprocessing The dataset is randomly shuffled and then divided into a training set and a test set in a ratio of 8:2. The training set is used to train the machine learning model, and the test set is used to evaluate the generalization ability of the training model.

[0039] This embodiment selects the XGBoost algorithm, and uses the correlation coefficient R and the root mean square error RMSE as evaluation indicators for judging the accuracy of the model. The regression coefficient between the reflection coefficient R predicted by the best model obtained on the training set and the actual value is 0.84, and the root mean square error is about 0.11. The reflection coefficient R in the test set is predicted using the trained prediction model, and the regression coefficient between the model prediction value and the actual value reaches 0.89, and the root mean square error is about 0.12. Therefore, the XGBoost model has a high accuracy and can be used for the prediction of the reflection coefficient R to assist in the design and development of composite materials.

[0040] The calculation results of the Pearson correlation coefficient between the 11 characteristic variables are as follows: Figure 2 As shown. The results show that the absolute value of the Pearson correlation coefficient between any two feature variables is less than 0.7, so there is no collinearity between the feature variables. Since a lower reflection coefficient R indicates better absorbing performance, the higher negative correlation between the feature and the reflection coefficient R indicates that the feature is more critical to the absorbing performance.

[0041] The six key characteristic variables were effectively identified by recursively eliminating the nested XGBoost model: the structure of the impedance matching layer, the filler of the impedance matching layer, the absorber, the stacking order of the layers, the molding process and the thickness. According to the Pearson correlation coefficient heat map and the six key characteristic variables, the following functional materials were designed: impedance matching layer structure: porous structure; impedance matching layer filler: carbon material; absorber selection: Fe3O4; layer stacking order: impedance matching layer-absorber layer-conductive layer; molding process: coating and hot pressing; thickness: 3-6mm. The designed materials were experimentally verified, and the optimized preparation method is as follows: (1) Soak the polyimide (PI) felt in the prepared acetone-ethanol mixed solution, ultrasonically treat for 2 hours to remove the attachments on the fiber surface, and rinse with deionized water until no foam is generated. Cut two pieces of PI felt into a size of 20 mm × 60 mm. According to the mass ratio of graphene to carboxylated cellulose nanofibers of 10:1, the graphene concentration is 100 mg / ml and the carboxylated cellulose nanofiber concentration is 10 mg / ml, and ultrasonic treatment is performed for 0.5 hours. The mixed solution is evenly coated on the surface of the PI felt. In order to ensure the uniformity and good adhesion of the coating, the coated PI felt sample is placed in an oven at 75°C and dried for 12 hours to obtain a single-layer graphene / PI felt preform. A layer of PI felt is stacked on the surface of the single-layer graphene / PI felt preform and fixed in a specific compression mold. The structure of the multi-layer PI felt is compressed together by applying vertical pressure, and a carbonization treatment is carried out at 1400°C in an argon environment. The temperature is increased to 800°C at a heating rate of 10°C / min; the temperature is increased to 1100°C at a heating rate of 8°C / min; the temperature is increased to 1400°C at a heating rate of 5°C / min and kept warm for 2h; the temperature is reduced to 1000°C at a cooling rate of 5°C / min; the temperature is reduced to 800°C at a cooling rate of 8°C / min. The sample is cooled with the furnace to obtain a conductive layer with a three-layer structure.

[0042] (2) h-BN powder was exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare f-BNNSs. In short, 2 g of h-BN powder was first added to a mixed solvent of deionized water and isopropanol, ultrasonically treated for 6 h, and then centrifuged to obtain f-BNNSs. Fe3O4 was added in a ratio of 1:1 and then evenly coated on the surface of the conductive layer to obtain an absorbing coating.

[0043] (3) Take the dried PI powder, add the polar organic solvent N,N-dimethylacetamide (DMAC) to prepare a PI solution with a mass fraction of 8wt%, and ultrasonically treat for 0.5h to ensure that the PI powder is completely and evenly dissolved. Soak the conductive layer / absorbent layer in the solution and move it into an oven under vacuum for 12h to remove bubbles, then program the vacuum oven to evaporate the solvent.

[0044] (4) The PI felt was soaked in a graphene solution with a concentration of 20 mg / ml and ultrasonically treated for 0.5 h to ensure that the graphene was evenly dispersed in the felt, and then dried to obtain an impedance matching layer.

[0045] (5) Using polyimide resin as an adhesive, the impedance matching layer and the conductive layer / absorbing layer are compounded by a hot pressing process to form a structurally complete multifunctional composite material. The hot pressing process is carried out at a pressure of 5 MPa, lasts for 30 minutes, and the hot pressing temperature is set to 165°C.

[0046] The thermal conductivity of the composite material sample prepared in this embodiment was tested on a thermal conductivity tester, and the vertical thermal conductivity of the composite material was calculated to be 0.55 Wm -1 K -1 ; At the same time, the reflection coefficient R obtained through testing is 0.47.

[0047] Comparative Example 1 According to the same preparation steps as in Example 2, only the filler of the impedance matching layer in the key characteristic variable is changed to graphene-free and a composite material is prepared, as follows: The steps for selecting key feature variables in machine learning remain unchanged, and the experimental verification steps are as follows: (1) Soak the polyimide (PI) felt in the prepared acetone-ethanol mixed solution, ultrasonically treat for 2 hours to remove the attachments on the fiber surface, and rinse with deionized water until no foam is generated. Cut two pieces of PI felt into a size of 20 mm × 60 mm. According to the mass ratio of graphene to carboxylated cellulose nanofibers of 10:1, the graphene concentration is 100 mg / ml and the carboxylated cellulose nanofiber concentration is 10 mg / ml, and ultrasonic treatment is performed for 0.5 hours. The mixed solution is evenly coated on the surface of the PI felt. In order to ensure the uniformity and good adhesion of the coating, the coated PI felt sample is placed in an oven at 75°C and dried for 12 hours to obtain a single-layer graphene / PI felt preform. A layer of PI felt is stacked on the surface of the single-layer graphene / PI felt preform and fixed in a specific compression mold. The structure of the multi-layer PI felt is compressed together by applying vertical pressure, and a carbonization treatment is carried out at 1400°C in an argon environment. The temperature is increased to 800°C at a heating rate of 10°C / min; the temperature is increased to 1100°C at a heating rate of 8°C / min; the temperature is increased to 1400°C at a heating rate of 5°C / min and kept warm for 2h; the temperature is reduced to 1000°C at a cooling rate of 5°C / min; the temperature is reduced to 800°C at a cooling rate of 8°C / min. The sample is cooled with the furnace to obtain a conductive layer with a three-layer structure.

[0048] (2) h-BN powder was exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare f-BNNSs. In short, 2 g of h-BN powder was first added to a mixed solvent of deionized water and isopropanol, ultrasonically treated for 6 h, and then centrifuged to obtain f-BNNSs. Fe3O4 was added in a ratio of 1:1 and then evenly coated on the surface of the conductive layer to obtain an absorbing layer.

[0049] (3) Take the dried PI powder, add the polar organic solvent N,N-dimethylacetamide (DMAC) to prepare a PI solution with a mass fraction of 8wt%, and ultrasonically treat for 0.5h to ensure that the PI powder is completely and evenly dissolved. Soak the conductive layer / absorbent layer in the solution and move it into an oven under vacuum for 12h to remove bubbles, then program the vacuum oven to evaporate the solvent.

[0050] (4) Use PI felt as an impedance matching layer.

[0051] (5) Using polyimide resin as an adhesive, the impedance matching layer and the conductive layer / absorbing layer are compounded by a hot pressing process to form a structurally complete multifunctional composite material. The hot pressing process is carried out at a pressure of 5 MPa, lasts for 30 minutes, and the hot pressing temperature is set to 165°C.

[0052] Figure 4 The improvement of the reflection coefficient R by adding carbon material into the filler of the impedance matching layer is demonstrated.

[0053] Comparative Example 2 According to the same preparation steps as Example 2, only the impedance matching layer structure in the key characteristic variable is changed to a non-porous structure and a composite material is prepared, as follows: The steps for selecting key feature variables in machine learning remain unchanged, and the experimental verification steps are as follows: (1) Soak the PI felt in the prepared acetone-ethanol mixed solution, ultrasonically treat for 2 hours to remove the attachments on the fiber surface, and rinse with deionized water until no foam is generated. Cut two pieces of PI felt into a size of 20 mm × 60 mm. According to the mass ratio of graphene to carboxylated cellulose nanofibers of 10:1, the graphene concentration is 100 mg / ml and the carboxylated cellulose nanofiber concentration is 10 mg / ml, and ultrasonic treatment is performed for 0.5 hours. The mixed solution is evenly coated on the surface of the PI felt. In order to ensure the uniformity and good adhesion of the coating, the coated PI felt sample is placed in an oven at 75°C and dried for 12 hours to obtain a single-layer graphene / PI felt preform. Two single-layer graphene / PI felt preforms and a layer of PI felt are stacked and fixed in a specific compression mold. The structure of the multi-layer PI felt is compressed together by applying vertical pressure, and carbonized at 1400°C in an argon environment. The temperature is increased to 800°C at a heating rate of 10°C / min; the temperature is increased to 1100°C at a heating rate of 8°C / min; the temperature is increased to 1400°C at a heating rate of 5°C / min and kept warm for 2h; the temperature is decreased to 1000°C at a cooling rate of 5°C / min; the temperature is decreased to 800°C at a cooling rate of 8°C / min. The sample is cooled with the furnace to obtain a conductive layer with a five-layer structure.

[0054] (2) h-BN powder was exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare f-BNNSs. In short, 2 g of h-BN powder was first added to a mixed solvent of deionized water and isopropanol, ultrasonically treated for 6 h, and then centrifuged to obtain f-BNNSs. Fe3O4 was added in a ratio of 1:1 and then evenly coated on the surface of the conductive layer to obtain an absorbing coating.

[0055] (3) The PI felt was soaked in a graphene solution with a concentration of 20 mg / ml and ultrasonically treated for 0.5 h to ensure that the graphene was evenly dispersed in the felt, and then dried to obtain an impedance matching layer.

[0056] (4) Take dry polyimide (PI) powder, add polar organic solvent N,N-dimethylacetamide (DMAC) to prepare a PI solution with a mass fraction of 8wt%, and ultrasonically treat for 0.5h to ensure that the PI powder is completely and evenly dissolved. Soak the conductive layer / absorbing layer / impedance matching layer in the solution and move it into an oven under vacuum for 12h to remove bubbles, then program the vacuum oven to evaporate the solvent.

[0057] pass Figure 5 It can be seen that the SER value obtained by this method is as high as 9.8dB. Compared with the impedance matching layer using the PI felt without resin filling, the porous structure is retained due to the lack of resin filling, which promotes the penetration of electromagnetic waves and multiple internal reflections, and improves the absorption capacity of electromagnetic waves. Therefore, the SER value is reduced from 8dB to 2.3dB.

[0058] The preparation method of the functional polyimide composite material with machine learning-assisted Janus structure provided in the above-mentioned embodiments of the present invention focuses on using machine learning technology to construct a model capable of predicting functional polyimide composite materials. The model optimizes the design of composite materials and preparation processes based on functional layered Janus structures by evaluating the degree to which key features affect performance. The material properties and interactions of each layer are precisely controlled to significantly improve the absorption efficiency of electromagnetic waves, while enhancing the thermal conductivity of the material, thereby improving development efficiency and reducing test costs, providing new technical concepts for manufacturing more efficient wave-absorbing and thermally conductive integrated materials with specific composite performance combination requirements.

[0059] The above embodiments of the present invention combine components, formulas and preparation processes, use multiple machine learning models for the optimal design of functional polyimide composite materials and preparation processes, and use the trained models to predict the changing relationship between the combined effect of the wave absorption performance and thermal conductivity of the functional polyimide composite materials, and obtain the optimal structure, components, ratios and preparation process parameters of the composite materials, thereby solving the technical problems that cannot be solved by the prior art, and solving the problem that the design of the components, microstructures and preparation processes of materials with wave absorption and good thermal conductivity is highly complex and uncertain. The linkage relationship between many mutually influencing and complex changing factors such as the components, ratios, structures, temperature, pressure, time and other factors of the composite materials can be predicted, thereby realizing the prediction and optimization of functional polyimide composite materials with a specific optimal combination of wave absorption and good thermal conductivity.

[0060] The above describes the embodiments of the present invention in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made according to the purpose of the invention. Any changes, modifications, substitutions, combinations or simplifications made according to the spirit and principle of the technical solution of the present invention should be equivalent replacement methods. As long as they meet the purpose of the invention and do not deviate from the technical principles and inventive concepts of the present invention, they belong to the protection scope of the present invention.

Claims

1. A method for preparing a functional polyimide composite material with machine learning-assisted Janus structure, characterized in that: The following steps are involved: S1. Establish database Collect known functional polyimide composite material information, including classification characteristics, numerical characteristics and preparation process, and establish a database; S2. Data Preprocessing Convert categorical features into binary sub-category features ("0" and "1"), preprocess the data, and divide it into training and test sets according to the ratio; S3: Design of basic Janus structure and preparation process The basic structure of the polyimide composite material to be prepared, including a Janus surface layer, an intermediate layer and a base layer, is designed, and then the basic components, proportions and preparation process of the composite material are designed. The process is as follows: each layer of the composite material is prepared, and then hot-pressed; S4. Build and train machine learning models Select a machine learning algorithm, use the training set to build a machine learning prediction model, use the characteristic information of the composite material as input parameters, evaluate the accuracy of the machine learning prediction model with evaluation indicators, and use cross-validation to screen the best machine learning prediction model; S5. Model testing and evaluation Use the test set to test the accuracy of the machine learning model and use relevant evaluation indicators to evaluate the performance. Use the Pearson correlation coefficient to analyze and verify the importance of features to ensure the accuracy and reliability of the model; S6. Feature Importance Analysis The recursive feature elimination method is combined with the machine learning prediction model to effectively identify key feature variables. Based on the analysis results of the Pearson correlation coefficient, these key features are further adjusted and the model parameters and feature selection are optimized to accurately adjust the model and ensure the highest prediction accuracy and improve the value of practical applications, and obtain model prediction results and key features; S7. Optimal design of composite materials Use machine learning to assist in optimizing composite material selection and structural design. Based on the model prediction results and key features obtained in the above steps, design a Janus structure polyimide composite material with microwave absorption and thermal conductivity and its preparation process. According to the key features identified by the machine learning model, determine the Janus structure, components and ratios of each layer, and preparation process of the optimized composite material. S8. Experimental verification Janus structure composite material samples were prepared, and control samples with reverse characteristics were prepared simultaneously. Then, comprehensive performance tests were performed on these samples to verify the wave absorption and thermal conductivity of the composite material samples. Those that met the design requirements were selected as the optimized functional polyimide composite materials.

2. The method for preparing a functional polyimide composite material with machine learning-assisted Janus structure according to claim 1, characterized in that: The step S1 specifically collects information related to a variety of known functional polyimide composite materials to establish a database, and the information and database include: the structure of the impedance matching layer, the filler of the impedance matching layer, the absorber, the stacking order of the layers, the molding process, the material structure, the conductive layer filler, the thickness and the reflection coefficient R.

3. The method for preparing a functional polyimide composite material with machine learning-assisted Janus structure according to claim 1, characterized in that: The data preprocessing in step S2 converts the classification features in step S1 into binary sub-category features "0" and "1", where "0" indicates that the sample does not belong to the sub-category, and "1" indicates that the sample belongs to the sub-category; the conversion results are as follows: structure of the impedance matching layer: porous structure (1), thin film structure (0); filler of the impedance matching layer: carbon material (1) and magnetic material (0); absorber: Fe3O4 (1), other absorbing materials (0); stacking order of layers: impedance matching layer-absorbing layer-conductive layer (1), other stacking order (0); molding process: freeze drying, hot pressing and coating; material structure: asymmetric (1), symmetric (0); conductive layer filler: carbon material (1), magnetic material (0).

4. The method for preparing a functional polyimide composite material with machine learning-assisted Janus structure according to claim 1, characterized in that: The machine learning algorithm in step S4 includes: one or more of the extreme gradient boosting (XGBoost) method, the ridge regression method, the gradient boosted decision tree (GBDT) method and the random forest regression method.

5. The method for preparing a functional polyimide composite material with machine learning-assisted Janus structure according to claim 1, characterized in that: The optimization design of the functional layered Janus structure composite material in step S7 is to adjust the dielectric properties of the surface layer, the middle layer and the base layer. The difference in conductivity makes each surface show different electromagnetic wave reflection characteristics, including the following steps: The surface of S7-1 uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves can enter the interior of the material without being reflected; The surface polyimide felt is not filled with resin, maintaining its porous structure, thereby allowing electromagnetic waves to enter; the middle layer and the base layer are filled with polyimide resin to prevent electromagnetic waves from penetrating; The S7-2 intermediate layer uses a boron nitride coating modified with ferroferric oxide as the electromagnetic wave absorption layer. The ferroferric oxide causes eddy current loss, converting electromagnetic energy into thermal energy, while the introduction of high thermal conductivity boron nitride promotes rapid dissipation of thermal energy, thereby improving the overall wave absorption efficiency and thermal management performance. The S7-3 substrate layer introduces a carbonized graphene / carbon fiber felt layer as a reflective layer, which fully reflects the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; S7-4 fills the absorbing layer and the conductive layer with polyimide resin to improve the structural stability of the material; and through hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure, thereby obtaining an optimized structure of the polyimide composite material with excellent electromagnetic wave management and efficient thermal management.

6. The method for preparing a functional polyimide composite material with machine learning-assisted Janus structure according to claim 1, characterized in that: The preparation method optimized in step S7 comprises the following steps: A1. Prepare the base layer as the reflective layer: evenly coat the graphene solution on the surface of the polyimide (PI) felt, stack and prepare a multi-layer polyimide felt / graphene structure, put the multi-layer graphene / PI felt structure into a special mold, apply vertical downward pressure to ensure that the layers are tightly bonded, and perform high-temperature carbonization treatment; A2. Preparation of an intermediate layer as an absorption layer: Boron nitride (BN) powder is exfoliated and modified by ultrasonic-assisted liquid phase exfoliation to prepare functionalized boron nitride nanosheets (f-BNNSs), and f-BNNSs and ferroferric oxide (Fe3O4) are evenly mixed and coated on the surface of the carbonized PI felt. 7.A3. Prepare the surface layer as the impedance matching layer: immerse the PI felt in a low concentration graphene solution and ultrasonicate for half an hour to ensure that the graphene particles are evenly attached to the surface of the felt; A4. Hot pressing of each layer: PI solution is prepared by dissolving PI powder in an organic solvent, and the conductive layer coated with the absorber is immersed in the PI solution. The solvent is evaporated by programmed temperature increase; the impedance matching layer and the cured absorption layer-conductive layer are connected by hot pressing using polyimide resin as an adhesive. During the hot pressing process, the PI resin is further cured to form a seamless connection, and finally an optimized preparation method for functional polyimide composite materials is obtained.

8. A functional polyimide composite material, characterized in that: The method is obtained by using the machine learning-assisted design method described in any one of claims 1 to 7, and comprises: a surface layer, an intermediate layer and a base layer stacked in sequence, wherein the surface layer uses porous graphene / polyimide felt as an impedance matching layer to adjust the surface impedance characteristics of the material so that a large amount of incident electromagnetic waves enter the interior of the material without being reflected; the intermediate layer uses a boron nitride coating modified with ferroferric oxide as an electromagnetic wave absorption layer, the ferroferric oxide causes eddy current loss, converts electromagnetic energy into thermal energy, and the introduction of high thermal conductivity boron nitride promotes rapid dissipation of thermal energy, thereby improving the overall wave absorption efficiency and thermal management performance; the base layer introduces a carbonized graphene / carbon fiber felt layer as a reflection layer to fully reflect the unabsorbed electromagnetic waves into the absorption layer for multiple absorption; the wave absorption layer and the conductive layer are filled with polyimide resin to improve the structural stability of the material; finally, by hot pressing, the polyimide resin is used as an adhesive to completely encapsulate the designed Janus structure, thereby obtaining a functional polyimide composite material with excellent electromagnetic wave management and efficient thermal management.

9. The functional polyimide composite material according to claim 8, characterized in that: The number of layers of the multi-layer polyimide felt / graphene structure is 3-x layers of graphene / PI felt arranged alternately (x≥5).

10. The functional polyimide composite material according to claim 8, characterized in that: In the electromagnetic shielding composite material, the thickness of the surface layer and the base layer are both between 500 μm and 3000 μm.

Images