Carbon fiber reinforced thermoplastic composite material performance database, construction method and application thereof

Abstract

The invention belongs to the technical field of high-performance composite materials, and discloses a carbon fiber reinforced thermoplastic composite material performance database, a construction method and application thereof, and the method comprises the following steps: S1, database structure construction; s2, experimental sample collection and data standardization; s3, feature engineering and variable reduction; s4, training a machine learning model; s5, constructing and verifying an adaptive model; and S6, data expansion and feedback optimization. According to the method, material performance prediction and formula parameter reverse design under target performance are realized through systematic acquisition and normalization processing of three types of data of material components, preparation process and performance characterization and building of a nonlinear mapping model among a material structure, a process and performance through a machine learning method. The database can be used for intelligently recommending a high-performance composite material combination scheme, is suitable for rapid screening and customized development of various thermoplastic composite materials, effectively reduces the research and development cost and development cycle, and improves the material design efficiency.

Description

Technical Field

[0001] The present invention relates to interdisciplinary fields such as high-performance composite materials engineering, data-driven material design, and machine learning predictive modeling, and specifically to a carbon fiber reinforced thermoplastic composite material performance database, construction method, and application thereof. Background Art

[0002] With the urgent demand for lightweight and high-performance structural materials in fields such as aerospace, rail transportation, and high-end manufacturing, carbon fiber reinforced thermoplastic composites (CFRTP) have become a key candidate for structural and functional integration due to their excellent specific strength, thermal stability, and hot processability. The performance of CFRTP is not only controlled by the properties of the carbon fibers and matrix resin, but also by the combined effects of multiple factors such as fiber volume fraction, arrangement, interfacial compatibility, interface reinforcement measures, and processing paths.

[0003] In practical applications, CFRTP systems often involve multi-scale structural design (such as the fiber-matrix-interface three-phase structure) and multi-parameter processing (such as temperature, pressure, and dwell time), with significant nonlinear coupling between these variables. This makes the traditional material development model, which relies on expert experience and trial-and-error testing, inefficient. This not only results in long development cycles, opaque parameter optimization space, low R&D efficiency, and high costs, but also lacks the ability to accurately predict performance mechanisms.

[0004] To address these challenges, the Materials Genome Initiative (MGI) provides a theoretical path to break through these bottlenecks. Its core concept is to achieve predictable, controllable, and reversible design of material properties through a three-pronged research paradigm: high-throughput preparation, high-throughput characterization, and big data mining. This approach combines high-throughput experiments, database systems, and machine learning modeling to achieve predictable, controllable, and reversible design of material properties. This approach has been initially validated in fields such as metals and ceramics. However, for CFRTP systems with complex interfacial behaviors and high-dimensional parameter spaces, a systematic database platform and structure-activity mapping modeling method have yet to be established.

[0005] Existing composite material design paradigms are often based on a linear process of "empirical hypothesis - material preparation - characterization and testing - data induction - correction and optimization." These processes rely heavily on manual experimentation and empirical parameter adjustment, lacking data-driven knowledge expression and intelligent decision-making mechanisms. In CFRTP systems in particular, issues such as interface regulation mechanisms, thermal-mechanical coupling responses, and nonlinear failure paths pose significant challenges to structure-activity modeling, making it difficult to meet the demands for rapid matching and coordinated optimization of multi-objective performance. This is particularly true in carbon fiber reinforced thermoplastic composite (CFRTP) systems, where multi-scale and multi-physics coupling behaviors are involved (e.g., interface transfer mechanisms and the regulation of mechanical responses by thermal flux perturbations), making accurate modeling of material structure-activity relationships extremely challenging.

[0006] In summary, in the field of complex properties such as mechanical properties and heat resistance, there are currently no mature database tools and intelligent reasoning systems, which make it difficult to support the rapid screening, performance regulation and customized development of material combinations in high-end application scenarios, and cannot meet the material screening and customization needs of high-performance fields such as aviation, rail transportation, and electronic packaging. Summary of the Invention

[0007] The present invention aims to provide a carbon fiber reinforced thermoplastic composite material performance database, construction method, and application thereof. Combining material components, processing technology, and performance characterization data, the present invention uses machine learning methods to establish a multivariate mapping relationship between composition, structure, process, and performance to achieve material performance prediction, reverse design, and virtual screening. This serves as an important supporting platform for intelligent material design and performance prediction, and enables the database to be used to intelligently recommend high-performance composite material combination solutions, effectively reducing R&D costs and development cycles, improving material design efficiency, and being applied to the rapid screening and customized development needs of thermoplastic composites in multiple technical fields.

[0008] The technical solution adopted by the present invention to achieve the above-mentioned purpose is: A method for constructing a carbon fiber reinforced thermoplastic composite material performance database, characterized by comprising the following steps: S1. Database structure construction The following three types of sub-databases are established in the database system: Component database, recording the original material information used to construct carbon fiber reinforced thermoplastic composites, including but not limited to: the type and physicochemical properties of the thermoplastic resin matrix; the type, linear density, surface treatment method, fiber arrangement and volume fraction of the carbon fiber; the type, polarity matching parameters and addition ratio of the interfacial modifier; the auxiliary phase and its addition type and ratio; Prepare a process parameter database to record the processing path information of composite materials, including but not limited to: molding method; temperature, pressure, holding time, cooling curve during heat treatment; processing equipment type and process window setting; material pretreatment parameters; Performance database, containing performance response data obtained from standardized experimental tests, including: mechanical properties, thermal properties, and microscopic characterization parameters; S2. Experimental sample collection and data normalization Based on preset combinations, systematically prepare composite material samples with different matrix-carbon fiber-interface agent ratios, and test material properties using unified testing standards. All raw data are uniformly converted to units, normalized, missing value filled, and coded to form a standardized vectorized data format. S3, Feature Engineering and Variable Dimensionality Reduction Perform categorical variable embedding encoding, numerical variable normalization, variable selection or dimensionality reduction; S4. Machine Learning Model Training Supervised learning methods are used to model and train the data, constructing a nonlinear mapping model between component parameters + process parameters → multiple output performance indicators. Model training uses cross-validation and holdout validation methods. Output accuracy indicators include: determination coefficient R², root mean square error RMSE, and mean absolute error MAE.

[0009] S5. Adaptation model construction and verification An adaptation model is constructed based on the trained model to describe the mapping relationship between components, processes, and performance. The model output is constrained by setting boundary conditions for input variables to form an adaptation structure suitable for reverse design. S6. Data expansion and feedback optimization By adding sample design and prediction-experiment closed-loop verification mechanisms, the database boundaries are continuously expanded, and the model parameters are dynamically optimized and updated to adapt to the diverse CFRTP architectures and performance distribution characteristics.

[0010] A carbon fiber reinforced thermoplastic composite material performance database is characterized in that it is constructed based on the method of the carbon fiber reinforced thermoplastic composite material performance database.

[0011] The application of the carbon fiber reinforced thermoplastic composite material performance database in the intelligent design of composite materials comprises the following steps: (1) Setting target performance index parameters, including at least one mechanical property and at least one thermal property; (2) Importing the target performance parameters as input into the trained adaptation model; (3) The model performs prediction and inference under set constraints based on the known component-process-performance mapping relationship in the database; (4) Output the material formula combination, carbon fiber volume fraction and molding process parameters that meet the target performance as a recommended preparation plan to guide the actual preparation or virtual screening of materials; (5) Import the recommended solution into the simulation verification system or the actual preparation process for performance verification, and further optimize the model based on the feedback results; (6) The recommended design scheme is further integrated into a virtual screening platform, simulation optimization system, or intelligent manufacturing process for rapid material screening, structural optimization, or personalized customization.

[0012] The beneficial effects of the present invention are:

[0013] 1. This invention utilizes materials genome engineering methods to construct a structured, high-dimensional, scalable carbon fiber-reinforced thermoplastic composite performance database. Combined with machine learning modeling strategies, this approach enables systematic modeling and reverse engineering from material composition, structure, and process parameters to performance responses. Compared to traditional material development models that rely on experience and experimentation, this invention constructs a high-dimensional, structured performance database for carbon fiber-reinforced thermoplastic composites. It also incorporates machine learning methods to establish a component-structure-process-performance mapping model, making it a key support platform for intelligent material design and performance prediction.

[0014] 2. The present invention provides a method for constructing a carbon fiber reinforced thermoplastic composite material performance database and its application. This method systematically collects and normalizes three types of data: material composition, preparation process, and performance characterization, and constructs a nonlinear mapping model between material structure, process, and performance through machine learning methods. Furthermore, by establishing an adaptation model, material performance prediction and reverse design of formulation parameters under target performance can be achieved, enabling the database to be used for intelligently recommending high-performance composite material combination solutions, effectively reducing R&D costs and development cycles, and improving material design efficiency. This method and database are applicable to the rapid screening and customized development of thermoplastic composite materials in fields such as aerospace, rail transportation, and lightweight automobiles.

[0015] 3. The method and application provided by the present invention for constructing a carbon fiber reinforced thermoplastic composite (CFRTP) performance database based on multidimensional experimental data, closely combined with a machine learning model, can realize the intelligent prediction and reverse design of mechanical properties and heat resistance properties, and can therefore be widely used in the rapid development and virtual screening systems of high-performance structural materials.

[0016] 4. The construction method, database, and application provided by the present invention can significantly improve the efficiency of material design: by constructing a data-driven structure-process-performance mapping system, material performance prediction and solution pre-screening can be quickly completed without the need for extensive physical testing, effectively reducing the R&D cycle and testing costs.

[0017] 5. The construction method, database, and application provided by the present invention support reverse deduction and customized design of multiple performance targets: the adaptation model constructed by the present invention can reversely recommend material components and process combinations based on the set target performance (such as tensile strength ≥350 MPa, Tg ≥240 ℃), significantly improving the target adaptation accuracy and path planning capabilities.

[0018] 6. The construction method, database, and application provided by the present invention can enhance the interpretability of the material structure-activity mechanism: Based on feature engineering and algorithm visualization, the present invention can analyze the influence weight of each input parameter (such as fiber volume fraction, arrangement mode, molding temperature, etc.) on the output performance, provide data support for the material property evolution mechanism, and enhance the physical relevance and industrial applicability of the model.

[0019] 7. The construction method, database and application provided by the present invention support multi-source heterogeneous data integration and high-throughput experimental feedback mechanism: the database structure is compatible with experimental test data, simulation prediction data and industrial field data, and has the ability to continuously expand and dynamically update, forming a closed-loop optimization system of experiment-simulation-data fusion.

[0020] 8. The construction method, database and application provided by the present invention can adapt to the structural and functional integrated material screening needs in high-end engineering fields: the system is particularly suitable for batch recommendation and performance optimization of lightweight, high-strength, and heat-resistant composite materials in aerospace, rail transportation, electronic packaging and other fields, which can greatly shorten the material selection time and improve the adaptation rate.

[0021] 9. The construction method, database and application provided by the present invention support the integrated application of intelligent manufacturing: the database and prediction model can serve as the "process-performance linkage module" of the digital composite material production line, realizing the automated control logic of raw material input - performance prediction - process optimization - online feedback, enabling high-reliability manufacturing systems.

[0022] 10. The construction method, database and application provided by the present invention are highly scalable: the database architecture and model training process of the present invention can be generalized and applied to a variety of thermoplastic resin matrices and other reinforced material systems, and can also be extended to the collaborative optimization design of multi-objective performance such as flame retardancy, conductivity, and electromagnetic shielding.

[0023] The above is an overview of the technical solution of the invention. The present invention will be further described below in conjunction with specific implementation methods. DETAILED DESCRIPTION

[0024] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose, the following is a detailed description of the applicable process and actual effects of the method of the present invention in combination with representative embodiments under different situations.

[0025] Basic Example

[0026] The method for constructing a carbon fiber reinforced thermoplastic composite material performance database provided in this embodiment includes the following steps: S1. Database structure construction The following three types of sub-databases are established in the database system to form the overall database: Component database, recording the original material information used to construct carbon fiber reinforced thermoplastic composites, including but not limited to: the type and physicochemical properties of the thermoplastic resin matrix; the type, linear density, surface treatment method, fiber arrangement and volume fraction of the carbon fiber; the type, polarity matching parameters and addition ratio of the interfacial modifier; the auxiliary phase and its addition type and ratio; In the component database, the types of carbon fibers include chopped fibers, continuous fibers, and fabrics; auxiliary phases include toughening agents and lubricants; Prepare a process parameter database to record the processing path information of composite materials, including but not limited to: molding method; temperature, pressure, holding time, cooling curve during heat treatment; processing equipment type and process window setting; material pretreatment parameters; In the preparation process parameter database, molding methods include hot pressing, injection molding, automatic wire placement / thermosetting molding, etc.; material pretreatment parameters include drying temperature, moisture content control, etc. Performance database, including performance response data obtained from standardized experimental tests, including but not limited to: mechanical properties, thermal properties, microscopic characterization parameters, etc.; In the performance database, the mechanical properties of performance response data include tensile strength, Young's modulus, flexural strength, flexural modulus, impact toughness, interlaminar shear strength, and fatigue limit; thermal properties include glass transition temperature Tg, heat deformation temperature HDT, thermal conductivity, and thermal stability; microscopic characterization parameters include XRD crystal structure, SEM interface morphology, and DMA storage modulus change spectrum; The three types of sub-databases mentioned above all support structured format management, use relational databases (SQL) or attribute-oriented object databases (such as HDF5 and MongoDB) for storage, and support heterogeneous experimental data integration and multi-table query interfaces; The data sources of the database constructed in step S1 include: laboratory sample test data, industrial production process monitoring data, open source literature database and finite element simulation output data; data from different data sources are all processed through consistency mapping and dimensional unification before being used for model training and verification in steps S4 and S5.

[0027] The thermoplastic resin matrix is ​​one of polyetheretherketone (PEEK), polyamide (PA6), polyphenylene sulfide (PPS), polypropylene (PP), and polyetherimide (PEI).

[0028] S2. Experimental sample collection and data normalization Based on preset combinations, systematically prepare composite material samples with different matrix-carbon fiber-interfacial agent ratios and test material properties using unified testing standards (such as ISO 527, ISO 11357, and ASTM D648). All raw data are uniformly converted to units, normalized, missing value filled, and coded to form a standardized vectorized data format. S3, Feature Engineering and Variable Dimensionality Reduction Perform categorical variable embedding encoding (one-hot embedding or label encoding); normalize numerical variables (z-score or min-max); and perform variable selection or dimensionality reduction (based on Pearson correlation coefficient analysis, principal component analysis (PCA), and recursive elimination (RFE) screening). S4. Machine Learning Model Training Supervised learning methods are used to model and train data, constructing a nonlinear mapping model between component parameters + process parameters → multiple output performance indicators. Model training uses cross-validation (k-fold) and group-wise cross-validation. Output accuracy indicators include: coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). The machine learning models include support vector machine (SVM), random forest (RF), gradient boosting tree (GBDT); one or more combinations of multi-layer perceptron (MLP) neural network and convolutional neural network (CNN); and graph neural network (GNN) is used to embed the graph structure of composite material components into model.

[0029] S5. Adaptation model construction and verification An adaptation model is constructed based on the trained model to describe the mapping relationship between components, processes, and performance. The model output is constrained by setting boundary conditions for input variables (such as upper and lower limits on component ratios and process window ranges), forming an adaptation structure suitable for reverse design. S6. Data expansion and feedback optimization By adding sample design and prediction-experiment closed-loop verification mechanisms, the database boundaries are continuously expanded, and model parameters are dynamically optimized and updated to adapt to diverse CFRTP architectures and performance distribution characteristics. Data expansion is the expansion of the database to other performance dimensions based on functional requirements, including collaborative modeling and trade-off optimization of multiple physical properties such as conductivity, flame retardancy, aging resistance, and electromagnetic shielding.

[0030] A carbon fiber reinforced thermoplastic composite material performance database is constructed based on the method of the above-mentioned carbon fiber reinforced thermoplastic composite material performance database.

[0031] The application of the carbon fiber reinforced thermoplastic composite material performance database in the intelligent design of composite materials comprises the following steps: (1) Setting target performance index parameters, including at least one mechanical property and at least one thermal property; (2) Importing the target performance parameters as input into the trained adaptation model; The target performance index parameters include: tensile strength, heat deformation temperature, Tg, thermal conductivity, impact toughness, and any combination of these performance index parameters; (3) The model performs prediction and inference under set constraints based on the known component-process-performance mapping relationship in the database; (4) Output the material formula combination, carbon fiber volume fraction and molding process parameters that meet the target performance as a recommended preparation plan to guide the actual preparation or virtual screening of materials; (5) Import the recommended solution into the simulation verification system or the actual preparation process for performance verification, and further optimize the model based on the feedback results; (6) The recommended design scheme is further integrated into a virtual screening platform, simulation optimization system, or intelligent manufacturing process for rapid material screening, structural optimization, or personalized customization.

[0032] The database can also be expanded to other performance dimensions according to functional requirements, including but not limited to the collaborative modeling and trade-off optimization of multiple physical properties such as conductivity, flame retardancy, aging resistance, and electromagnetic shielding.

[0033] Example 1 This embodiment provides a carbon fiber reinforced thermoplastic composite material performance database, construction method, and application provided in this embodiment. This embodiment is a specific implementation of the basic embodiment, covering the entire process from data acquisition, database structure design, machine learning modeling to intelligent reverse design, and realizing systematic correlation modeling and optimization design of material components, process parameters, and performance responses. The database construction method specifically includes the following steps: 1. Database structure construction Construct three types of sub-databases in the MySQL relational database: (1) Component database: used to enter the type and parameters of thermoplastic resin matrix, including PEEK (density, Tg, melting point, molecular weight and distribution, etc.), PA6, PPS, etc.; carbon fiber types are T700, T800, T1000 grade continuous fibers, recording linear density, surface treatment method, volume fraction; interfacial modifier and addition ratio; auxiliary phase and addition ratio and other parameters.

[0034] (2) Preparation process parameter database: used to record the hot pressing process path, including pre-drying temperature, moisture content; molding temperature, pressure, holding time, cooling curve rate; molding method.

[0035] (3) Performance database: various data obtained through tests in accordance with ISO 527, ISO 11357 and other standards. Mechanical properties include tensile strength, flexural strength, and interlaminar shear strength; thermal properties include Tg, HDT, and thermal conductivity.

[0036] 2. Experimental Sample Collection and Data Normalization A total of 24 samples were prepared using four different matrix combinations (PEEK / PA6 / PPS / PP) and two carbon fiber types (chopped / continuous), along with various surface treatments. Tensile, flexural, impact, and thermal performance tests were performed. All experimental data were standardized (e.g., stress was uniformly expressed in MPa), normalized (min-max scaling), and interpolated (KNN interpolation). Variable encoding was then converted to vectorized data in standard CSV and HDF5 formats.

[0037] 3. Feature Engineering and Dimensionality Reduction One-hot encoding was used for categorical variables, and z-score normalization was performed on numerical variables. Pearson correlation coefficient and PCA were used for dimensionality reduction, and principal components with a cumulative explained variance of more than 90% were retained for subsequent modeling.

[0038] 4. Machine Learning Model Training Random forest (RF), gradient boosted tree (GBDT), and multi-layer perceptron (MLP) models were used for model comparison. Component and process parameters were used as input, and performance indicators were trained as multi-output targets. A combination of 10-fold cross-validation and leave-one-out cross-validation was employed. The RF model achieved the best prediction performance, with a tensile strength R² of 0.91, an RMSE of 8.3 MPa, and a MAE of 5.7 MPa.

[0039] 5. Adaptation model construction and verification Based on the trained GBDT+RF fusion model, an adaptation module is constructed. The input boundary conditions include: the carbon fiber volume fraction does not exceed 60%, the processing temperature is not higher than 400 ℃, etc. The model outputs recommended material components and process parameter combinations based on the input target performance to support reverse design.

[0040] 6. Data expansion and feedback optimization DoE design is introduced, and the Latin hypercube design method is used to generate new sample combinations. Combined with the finite element thermal conductivity prediction results and experimental test data, a closed-loop update mechanism is adopted to expand the database and retrain the model to improve the generalization ability.

[0041] 7. Intelligent Design Application The trained model was integrated into the design platform, and the target performance was set as follows: tensile strength ≥ 150 MPa, thermal conductivity ≥ 0.5 W·m⁻¹·K⁻¹, and Tg ≥ 140 ℃. The results were input into the adaptation model, and the recommended formula that met the target was output: PA6 matrix + chopped carbon fiber (volume fraction 38%) + surface treatment agent A (1.5 wt%) + hot pressing (340 ℃ / 2 MPa / 10 min). This formula was used as the virtual screening result for material preparation.

[0042] 8. Multiple performance index combination and customized design The target performance combination is: flexural strength > 200 MPa, HDT > 130 °C, and impact toughness > 30 kJ / m². The platform can adjust the recommendation strategy based on different target weights (user-defined) to achieve coordinated optimization of mechanical and thermal properties.

[0043] 9. Integrate applications into manufacturing processes The model and database structure have been integrated into the virtual platform for the material development process and connected to the Simulia / Abaqus simulation system interface to achieve a closed-loop linkage from structural simulation to formulation design, and support online screening, intelligent recommendation and rapid preparation of customized composite materials.

[0044] Example 2 This example provides a carbon fiber-reinforced thermoplastic composite performance database, construction method, and application. This is a refinement of Basic Example 1, providing a prediction and reverse design solution for PEEK-based continuous carbon fiber composites in high-performance structural applications. This example aims to verify the prediction accuracy and reverse recommendation capabilities of the adaptation model constructed in this invention for high-performance structural material design. First, target performance parameters are set as tensile strength ≥ 350 MPa and heat deflection temperature (HDT) ≥ 260°C, with the application scenario being aviation structural components.

[0045] The candidate matrix screened from the database is PEEK, and the reinforcement is continuous carbon fiber fabric (single-layer density of 180 g / m²). The model boundary conditions are set as follows: carbon fiber volume fraction range is 50–60%, molding temperature is between 350–380 °C, and holding time is 6–12 min.

[0046] The recommended results calculated by the adaptation model are: carbon fiber volume fraction is 55%, hot pressing temperature is 365 ℃, pressure is 8 MPa, holding time is 10 min, and cooling rate is controlled within 5 ℃ / min.

[0047] Three groups of composite materials were prepared according to the recommended scheme. Their tensile strength and HDT were tested according to ISO 527 and ISO 178 standards. The results showed an average tensile strength of 372 MPa and an HDT of 273 °C, which were within ±5% of the model prediction error. This verifies the practicality of the proposed method in the intelligent recommendation and precision control of high-strength structural parts.

[0048] Example 3 This embodiment provides a carbon fiber reinforced thermoplastic composite material performance database, construction method, and application provided in this embodiment. It is a specific embodiment based on Basic Example 1, providing an application solution for rapid screening and low-cost optimization of chopped fiber / PP-based lightweight CFRTP materials.

[0049] The goal of this example is to achieve design optimization of lightweight, low-cost thermoplastic composite materials with the following target properties: tensile strength ≥ 120 MPa, HDT ≥ 110 °C, and density ≤ 1.2 g / cm³, for use in automotive instrument panel support structures.

[0050] The system initialization database contains 20 groups of original samples of PP / chopped carbon fiber system, the reinforcement is chopped carbon fiber (length 6 mm), the matrix is ​​homopolymer PP, and the interface agent is maleic anhydride grafted polypropylene (addition amount 0.5 wt%).

[0051] By setting the performance target range, the adaptation model returns the recommended parameters: fiber volume fraction of 25%, hot pressing temperature of 190 °C, holding time of 5 minutes, pressure of 6 MPa, and free cooling rate.

[0052] Samples prepared according to the recommended protocol were tested, and the tensile strength was 126 MPa, the HDT was 117°C, and the density was 1.15 g / cm³, meeting expectations. This validated the system's ability to quickly adapt to mid-range and low-end performance requirements and its low resource consumption.

[0053] Example 4 This embodiment provides a carbon fiber reinforced thermoplastic composite material performance database, construction method and application provided in this embodiment, which is a concretization based on Basic Example 1, providing a new material system integration-PEI-based composite material subsystem expansion and model migration verification solution.

[0054] To verify the scalability of the database and model system, this example selected PEI (polyetherimide), which did not appear in the initial sample set, as the matrix material, constructed a new data subsystem, and carried out transfer learning training.

[0055] Fifteen groups of PEI-based composite samples were prepared, including chopped, continuous, and woven fiber morphologies. These samples varied in fiber volume fraction (20–50%), molding temperature (280–320°C), and cooling rate (5–20°C / min). Multiscale performance parameters, including tensile modulus, Tg (derived by DSC), and DMA storage modulus spectra, were obtained using standardized testing methods and entered into the database extension node.

[0056] An incremental learning algorithm was used to fine-tune the weights based on the original model architecture. The migrated model achieved R² predictions of 0.91 (tensile strength) and 0.94 (Tg) for the PEI system properties, demonstrating the model's adaptability and generalizability across systems.

[0057] Example 5 This embodiment provides a carbon fiber reinforced thermoplastic composite material performance database, construction method, and application provided in this embodiment, which is a concretization based on Basic Example 1, and provides a recommended optimization scheme for structure-function integrated composite materials under multi-objective collaborative design.

[0058] This embodiment simulates an electronic packaging application scenario. The target performance indicators include tensile strength ≥ 300 MPa, thermal conductivity ≥ 1.5 W / (m·K), and impact strength ≥ 25 kJ / m², which are typical structural-thermal functional synergy requirements.

[0059] The model inputs include four components: a PPS matrix, continuous carbon fiber fabric, a thermal conductivity enhancer (carbon nanotubes (CNTs), and an interfacial coupling agent. The CNT loading ranged from 0–2 wt%, and the fiber volume fraction ranged from 40–50%. A multi-objective Bayesian optimization strategy was used to construct a performance trade-off function.

[0060] The recommended parameter combination is: PPS + continuous CF (Vf = 45%) + 1.2% CNT + 0.3% interface agent, with a molding temperature of 295°C, a holding pressure of 10 minutes, and a pressure of 10 MPa. Test results showed a tensile strength of 328 MPa, a thermal conductivity of 1.64 W / (m·K), and an impact strength of 26.1 kJ / m², all meeting the design requirements. This demonstrates the high-dimensional modeling capabilities and collaborative recommendation advantages of the proposed system under multi-objective optimization.

[0061] The above-mentioned embodiments of the present invention have demonstrated that the present invention is applicable to various performance target scenarios such as high strength, light weight, thermal conductivity, and heat resistance. It is systematic, generalizable, and practical, and can provide effective support for the digital and intelligent design of composite materials. The method includes the systematic collection and normalization of three types of data: material components, preparation processes, and performance characterizations, and constructs a nonlinear mapping model between material structure-process-performance through machine learning methods; by establishing an adaptation model, the prediction of material properties and the reverse design of formulation parameters under target performance can be achieved; the database can be used to intelligently recommend high-performance composite material combination solutions, effectively reducing R&D costs and development cycles, and improving material design efficiency. The method and database can be applied to the rapid screening and customized development of thermoplastic composite materials in the fields of aerospace, rail transportation, and lightweight automobiles.

[0062] Furthermore, within the scope of the present invention, other embodiments obtained by selecting different database types, preparation processes, components, and their proportions based on different needs through the carbon fiber reinforced thermoplastic composite material performance database, construction method, and application can achieve the intended technical effects of the present invention. Therefore, these embodiments are not listed one by one.

[0063] The above description is merely a preferred embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any technical features that are the same or similar to those of the above embodiments of the present invention are within the scope of protection of the present invention.

Claims

1. A method for constructing a carbon fiber reinforced thermoplastic composite material performance database, characterized in that: The following steps are involved: S1. Database structure construction The following three types of sub-databases are established in the database system: Component database, recording the original material information used to construct carbon fiber reinforced thermoplastic composites, including but not limited to: the type and physicochemical properties of the thermoplastic resin matrix; the type, linear density, surface treatment method, fiber arrangement and volume fraction of the carbon fiber; the type, polarity matching parameters and addition ratio of the interfacial modifier; the auxiliary phase and its addition type and ratio; Preparation process parameter database, recording composite material processing path information, including but not limited to: molding method; Temperature, pressure, holding time, and cooling curve during heat treatment; Processing equipment type and process window setting; Material pretreatment parameters; Performance database, containing performance response data obtained from standardized experimental tests, including: mechanical properties, thermal properties, and microscopic characterization parameters; S2. Experimental sample collection and data normalization Based on preset combinations, systematically prepare composite material samples with different matrix-carbon fiber-interface agent ratios, and test material properties using unified testing standards. All raw data are uniformly converted to units, normalized, missing value filled, and coded to form a standardized vectorized data format. S3, Feature Engineering and Variable Dimensionality Reduction Perform categorical variable embedding encoding, numerical variable normalization, variable selection or dimensionality reduction; S4. Machine Learning Model Training Supervised learning methods are used to model and train data, constructing a nonlinear mapping model between component parameters + process parameters → multiple output performance indicators. Model training uses cross-validation and holdout validation methods. Output accuracy indicators include: determination coefficient R², root mean square error RMSE, and mean absolute error MAE. S5. Adaptation model construction and verification An adaptation model is constructed based on the trained model to describe the mapping relationship between components, processes, and performance. The model output is constrained by setting boundary conditions for input variables to form an adaptation structure suitable for reverse design. S6. Data expansion and feedback optimization By adding sample design and prediction-experiment closed-loop verification mechanisms, the database boundaries are continuously expanded, and the model parameters are dynamically optimized and updated to adapt to the diverse CFRTP architectures and performance distribution characteristics.

2. The method for constructing a carbon fiber reinforced thermoplastic composite material performance database according to claim 1, characterized in that: The database constructed in step S1 supports structured format management, adopts relational database or attribute-oriented object database for storage, and supports heterogeneous experimental data integration and multi-table query interface; In the component database, the types of carbon fibers include chopped fibers, continuous fibers, and fabrics; auxiliary phases include toughening agents and lubricants; In the preparation process parameter database, molding methods include hot pressing, injection molding, automatic wire placement / thermosetting molding, etc.; material pretreatment parameters include drying temperature and moisture content control; In the performance database, the mechanical properties of performance response data include tensile strength, Young's modulus, flexural strength, flexural modulus, impact toughness, interlaminar shear strength, and fatigue limit; thermal properties include glass transition temperature Tg, heat deformation temperature HDT, thermal conductivity, and thermal stability; microscopic characterization parameters include XRD crystal structure, SEM interface morphology, and DMA storage modulus change spectrum.

3. The method for constructing a carbon fiber reinforced thermoplastic composite material performance database according to claim 1, characterized in that: The data sources of the database constructed in step S1 include: laboratory sample test data, industrial production process monitoring data, open source literature database and finite element simulation output data; data from different data sources are all processed through consistency mapping and dimensional unification before being used for model training and verification in steps S4 and S5.

4. The method for constructing a carbon fiber reinforced thermoplastic composite material performance database according to claim 1, characterized in that: The thermoplastic resin matrix in step S1 is one of polyetheretherketone (PEEK), polyamide (PA6), polyphenylene sulfide (PPS), polypropylene (PP), and polyetherimide (PEI).

5. The method for constructing a carbon fiber reinforced thermoplastic composite material performance database according to claim 1, characterized in that: The machine learning model in step S4 is a combination of one or more of support vector machine (SVM), random forest (RF), gradient boosting tree (GBDT), multi-layer perceptron (MLP) neural network, and convolutional neural network (CNN); and graph neural network (GNN) is used to embed the composite material component graph structure into a model.

6. The method for constructing a carbon fiber reinforced thermoplastic composite material performance database according to claim 1, characterized in that: The data expansion in step S6 is to expand the database to other performance dimensions according to functional requirements, including: collaborative modeling and trade-off optimization of multiple physical properties such as conductivity, flame retardancy, aging resistance, and electromagnetic shielding.

7. A carbon fiber reinforced thermoplastic composite material performance database, characterized in that: It is constructed based on the method of the carbon fiber reinforced thermoplastic composite material performance database described in any one of claims 1 to 6.

8. An application of the carbon fiber reinforced thermoplastic composite material performance database according to claim 7 in the intelligent design of composite materials.

9. The use according to claim 8, characterized in that The following steps are involved: (1) Setting target performance index parameters, including at least one mechanical property and at least one thermal property; (2) Importing the target performance parameters as input into the trained adaptation model; (3) The model performs prediction and inference under set constraints based on the known component-process-performance mapping relationship in the database; (4) Output the material formula combination, carbon fiber volume fraction and molding process parameters that meet the target performance as a recommended preparation plan to guide the actual preparation or virtual screening of materials; (5) Import the recommended solution into the simulation verification system or the actual preparation process for performance verification, and further optimize the model based on the feedback results; (6) The recommended design scheme is further integrated into a virtual screening platform, simulation optimization system, or intelligent manufacturing process for rapid material screening, structural optimization, or personalized customization.

10. The application according to claim 9, wherein the target performance indicator parameters include: Tensile strength, heat distortion temperature, Tg, thermal conductivity, impact toughness, and any combination of these performance index parameters.