Composite material analysis method and system based on machine learning
By constructing a multi-source literature database and optimizing process parameters using integrated learning models, the problem of low efficiency in composite material analysis and optimization is solved, and efficient and accurate composite material performance analysis and process design are achieved to ensure the acquisition of high-performance materials.
Patent Information
- Application Number
- CN202510595574.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2025-08-15
AI Technical Summary
The prior art has low efficiency and high resource consumption in composite material analysis and optimization, making it difficult to efficiently and accurately analyze the impact of factors such as different proportions and particle sizes on the performance of composite materials.
Using a machine learning-based method, a multi-source literature database is constructed, data processing and optimization is carried out through an integrated learning model, and parameter optimization is selected in combination with the process flow of composite materials to determine high-performance process parameters.
It realizes efficient analysis of composite materials, breaks through the limitations of traditional trial and error methods, provides convenient process design, ensures the acquisition of high-performance composite materials, and ensures its application.
Smart Images
Figure CN120496709A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the interdisciplinary field of materials computational science and intelligent manufacturing, and in particular to a composite material analysis method and system based on machine learning. Background Art
[0002] With the development and application of industry, the performance requirements for materials in high-temperature, wear-resistant and high-intensity working environments are becoming increasingly stringent. To overcome these limitations, ceramic matrix composites (CMCs) have emerged. These materials are gradually becoming an alternative to traditional metal materials due to their excellent high-temperature stability, oxidation resistance, wear resistance and high strength. For example, Al2O3 / SiC composites. Al2O3 ceramics, with their high hardness, strong high-temperature oxidation resistance, and low cost, have become the most widely used engineering ceramics. However, the low fracture toughness of Al2O3 ceramics limits their application in high-temperature, high-speed working conditions such as high-speed cutting. To improve the toughness and wear resistance of Al2O3 ceramics, various secondary phases are often added to the Al2O3 matrix to achieve a toughening effect. SiC, due to its excellent mechanical properties, is often used as a reinforcing particle. The addition of SiC particles to the Al2O3 matrix can significantly improve the bending strength, hardness, and fracture toughness of Al2O3 ceramics. As a result, Al2O3 / SiC composites have demonstrated excellent performance in various fields. For example, in the tool field, Al2O3 ceramic tools incorporating SiC exhibit excellent wear resistance and a long service life during high-speed cutting. In coating applications, coatings incorporating Al2O3 / SiC composites can significantly improve the coating's density and ablation resistance, providing effective protection for the substrate in high-temperature environments. In the field of bulletproof armor, Al2O3 / SiC nanocomposite ceramics were prepared by hot pressing and sintering, so that the SiC located at the grain boundaries can deflect cracks and promote transgranular fracture; the SiC located in the Al2O3 grains can pin the grain boundaries and hinder crack propagation, thereby reducing the maximum impact force of the bullet and enhancing the material's impact resistance and crushing energy dissipation capabilities.
[0003] While composite materials hold enormous potential, their performance is influenced by a multitude of factors during testing. These factors are complex and intertwined, making optimization of material properties extremely challenging. Currently, composite material analysis and optimization typically rely on extensive experimental data and empirical formulas, but these methods are inefficient when faced with complex, multidimensional problems and require significant resources and time. Therefore, efficiently and accurately analyzing the impact of factors such as different ratios and particle sizes on composite material performance has become a pressing issue in composite material analysis and optimization. Summary of the Invention
[0004] The purpose of the present invention is to provide a composite material analysis method and system based on machine learning, which can realize efficient analysis of composite materials based on machine learning and determine the high performance parameters of composite materials, thereby providing convenience for the process design of composite materials.
[0005] To achieve the above objectives, the present invention provides the following technical solution: a composite material analysis method based on machine learning, comprising:
[0006] Identify target composite materials;
[0007] Construct a multi-source literature database based on target composite materials;
[0008] Obtain implementation data from a multi-source literature database;
[0009] Performing standardization on the implementation data to obtain standardized implementation data, and dividing the standardized implementation data into a training set and a test set;
[0010] Build an integrated learning model;
[0011] The training set and the test set are used to optimize the ensemble learning model to obtain the ensemble learning optimization model;
[0012] The process parameters of the target composite material are obtained by combining the integrated learning optimization model with the process flow of the target composite material to optimize the parameters.
[0013] Furthermore, standardization is performed on the implementation data, including:
[0014] Performing data standardization on the implementation data information of the literature file to obtain first processed implementation data;
[0015] The first processing implementation data information is cleaned to obtain standardized implementation data.
[0016] Furthermore, when constructing the ensemble learning model, the XGBoost regression algorithm is used to establish a mapping relationship between the preparation parameters and the performance indicators, and the model parameters are set for the ensemble learning model.
[0017] Furthermore, when optimizing the ensemble learning model using the training set and the test set, the optimization objective function is used to determine the optimal parameters of the model, where the optimization objective function is:
[0018]
[0019] In the above formula, y i is the actual result value of the i-th implementation data, is the model output value of the i-th implementation data, γ is the complexity penalty coefficient of the decision tree, T is the number of leaf nodes of the decision tree, λ is the regularization parameter, and wj is the weight of the j-th leaf node.
[0020] Furthermore, after obtaining the process parameters of the target composite material, the process parameters of the target composite material are verified, including:
[0021] Conducting tests according to the process parameters of the target composite material and the process flow of the target composite material to obtain the target test composite material;
[0022] The performance analysis is carried out on the target test composite material to obtain the test analysis results.
[0023] Furthermore, a multi-source literature database is constructed based on the target composite material, including:
[0024] Analyze the composition of the target composite material to determine the composition of the composite material;
[0025] Collecting literature information according to the components of the composite material to obtain a first collected literature file;
[0026] Combining the components of the composite material to collect literature information, to obtain a second collected literature file;
[0027] A document file database is constructed based on the first collected document files and the second collected document files to obtain a multi-source document database.
[0028] Furthermore, the standardized implementation data is divided into a training set and a test set, including:
[0029] Determine the ratio of training set and test set;
[0030] The standardized implementation data were randomly arranged to obtain the allocation sequence;
[0031] According to the allocation rules, the standardized implementation data in the allocation sequence are allocated to the training set and the test set in turn to obtain the training set and the test set.
[0032] Furthermore, the training set and test set are used to optimize the ensemble learning model, including:
[0033] Train the ensemble learning model through the training set to determine the ensemble learning training model;
[0034] Using the test set to test the ensemble learning training model, and determining whether the ensemble learning training model is optimal based on the test data analysis, to obtain a first analysis result;
[0035] When the first analysis result shows that the ensemble learning training model is optimal, the ensemble learning training model at this time is the ensemble learning optimization model;
[0036] When the first analysis result is that the integrated learning training model has not reached the optimal level, the model training data of the training set when training the integrated learning model and the model test data of the test set when testing the integrated learning training model are obtained, and the training set and test are updated according to the model training data and model test data. Then, the updated training set and test set are used to continue to optimize the integrated learning model until the integrated learning training model reaches the optimal level, thereby obtaining an integrated learning optimization model.
[0037] A composite material analysis system based on machine learning, comprising: a target confirmation unit, a literature collection unit, a data processing unit, a model construction unit, a model optimization unit, and a parameter optimization unit;
[0038] The target confirmation unit is used to determine the target composite material;
[0039] The document collection unit is used to construct a multi-source document database based on the target composite material;
[0040] The data processing unit is used to obtain implementation data from a multi-source document database, perform standardization processing on the implementation data to obtain standardized implementation data, and then divide the standardized implementation data into a training set and a test set;
[0041] The model building unit is used to build an integrated learning model;
[0042] The model optimization unit is used to optimize the ensemble learning model using the training set and the test set to obtain the ensemble learning optimization model;
[0043] The parameter optimization unit is used to optimize parameters by combining an integrated learning optimization model with the process flow of the target composite material to obtain the process parameters of the target composite material.
[0044] Furthermore, the composite material analysis system further comprises: a parameter verification unit;
[0045] After obtaining the process parameters of the target composite material, the parameter verification unit verifies the process parameters of the target composite material, including:
[0046] Conducting tests according to the process parameters of the target composite material and the process flow of the target composite material to obtain the target test composite material;
[0047] The performance analysis is carried out on the target test composite material to obtain the test analysis results.
[0048] The present invention realizes efficient analysis of composite materials based on machine learning and determines the process parameters of high-performance composite materials, breaking through the limitations of traditional trial-and-error methods, providing convenience for the process design of composite materials, enabling better acquisition of high-performance composite materials, and thus providing guarantees for the use of composite materials.
[0049] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the structures specifically pointed out in the application documents.
[0050] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0052] Figure 1 Schematic diagram of the steps of the composite material analysis method according to the present invention;
[0053] Figure 2 Schematic diagram of part of step 4 in the composite material analysis method of the present invention;
[0054] Figure 3 Schematic diagram of the process parameter verification steps in the composite material analysis method of the present invention;
[0055] Figure 4 is an overall schematic diagram of the composite material analysis method according to the present invention;
[0056] Figure 5 This is a schematic diagram of step 2 in the composite material analysis method of the present invention;
[0057] Figure 6 This is a graphical representation of data for analyzing the prediction accuracy of the model for the performance indicator flexural strength in the composite material analysis method of the present invention;
[0058] Figure 7 This is a graphical representation of data for analyzing the prediction accuracy of the model using the performance indicator Vickers hardness in the composite material analysis method of the present invention;
[0059] Figure 8 This is a graphical representation of data for analyzing the prediction accuracy of the model for the performance indicator fracture toughness in the composite material analysis method of the present invention;
[0060] Figure 9 A schematic diagram of the composite material analysis system according to the present invention;
[0061] Figure 10 This is another schematic diagram of the composite material analysis system according to the present invention. DETAILED DESCRIPTION
[0062] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0063] like Figure 1 As shown, an embodiment of the present invention provides a composite material analysis method based on machine learning, comprising:
[0064] Step 1: Determine the target composite material;
[0065] Step 2: Construct a multi-source literature database based on the target composite material;
[0066] Step 3: Obtain implementation data from a multi-source literature database;
[0067] Step 4: Standardize the implementation data to obtain standardized implementation data, and divide the standardized implementation data into a training set and a test set;
[0068] Step 5: Build an integrated learning model;
[0069] Step 6: Use the training set and the test set to optimize the ensemble learning model to obtain the ensemble learning optimization model;
[0070] Step 7: Optimize the parameters of the target composite material through the integrated learning optimization model to obtain the process parameters of the target composite material.
[0071] In the above technical solution, when the standardized implementation data is divided into a training set and a test set, the training set and the test set are divided according to a ratio, for example, the training set and the test set are divided according to a ratio of 4:1.
[0072] In the above technical solution, when constructing the integrated learning model, the XGBoost regression algorithm is used to establish the mapping relationship between the preparation parameters and the performance indicators.
[0073] In the above technical solution, the process parameters include: the composition ratio of the composite material, the particle size of the composite material, the temperature and pressure of the process, etc.
[0074] In the above technical solution, obtaining implementation data in a multi-source document database includes: identifying document files, reading the contents of the document files, and obtaining document content information;
[0075] The implementation data is determined in the document content information, and data extraction is performed on the implementation data to obtain the implementation data of the document file.
[0076] In the above technical solution, when optimizing parameters through the integrated learning optimization model combined with the process flow of the target composite material, the target performance of the target composite material is determined based on the application of the target composite material, and the preparation parameters are determined through the integrated learning optimization model combined with the target performance. Then, a genetic algorithm is used to optimize parameters in combination with the process flow and preparation parameters of the target composite material, thereby obtaining the optimal process parameters of the target composite material.
[0077] The above technical solution realizes efficient analysis of composite materials based on machine learning, so that a large amount of experimental data and empirical formula analysis can be completed in a relatively short time, and the process parameters of high-performance composite materials can be determined, so that high-performance composite materials can be obtained according to the process parameters of composite materials, which provides convenience for the process design of composite materials, thereby better obtaining high-performance composite materials and providing guarantees for the use of composite materials. By constructing a multi-source literature database based on the target composite material, data collection is carried out according to the target composite material, ensuring that the multi-source literature database is information about the target composite material, thereby avoiding the ensemble learning model from performing invalid analysis on irrelevant implementation data, reducing time waste, and by standardizing the implementation data, ensuring the consistency of the implementation data, avoiding the implementation data dimensional inconsistency that cannot be processed by the ensemble learning model, and also avoiding the learning model error illusion caused by the inconsistent implementation data dimensionality, providing guarantees for the ensemble learning optimization model, and by combining the ensemble learning optimization model with the process flow of the target composite material for parameter optimization, the process parameters of the target composite material can be used to obtain high-performance composite materials in the process of composite material production, providing guarantees for the application of the target composite material.
[0078] In one embodiment provided by the present invention, Figure 2 As shown, the implementation data is standardized, including:
[0079] D1. Perform data cleaning on the implementation data information of the literature file to obtain the first processed implementation data;
[0080] D2. Perform data standardization on the first processing implementation data information to obtain standardized implementation data.
[0081] In the above technical solution, the implementation data includes: test implementation data and test result data, and the implementation method is only the method for testing the test implementation data in the literature.
[0082] In the above technical solution, when performing data cleaning on the implementation data information of the literature file, abnormal indicators are eliminated and missing data are supplemented for the first processed implementation data.
[0083] In the above technical solution, when performing data standardization on the first processing implementation data information, the unit system data is converted into international standard units and equivalent conversion is performed on the data of different implementation methods.
[0084] The above technical solution provides a guarantee for the data analysis and processing of the integrated learning model by making the data in different documents have the same dimension when the implementation data is standardized, avoiding the situation where the model analysis cannot be performed due to different data dimensions or the model analysis error is large, thereby ensuring the optimization of the integrated learning optimization model. Moreover, by identifying and obtaining the implementation data and implementation methods in the document files, the information extraction of the document files is realized, the redundancy of irrelevant information is reduced, and the subsequent analysis and processing is only carried out on the implementation data of the identified document files, which provides convenience for data standardization and data cleaning, improves the efficiency of data standardization and data cleaning, and avoids errors in data standardization and data cleaning. In addition, the consistency of the implementation data is guaranteed by data standardization, avoiding the large model analysis error caused by inconsistent dimensions that affects the optimization of the integrated learning model. At the same time, the integrity of the implementation data is guaranteed by data cleaning, avoiding data anomalies or data missing that affect the optimization efficiency of the integrated learning model, thereby providing a guarantee for the optimization of the integrated learning model.
[0085] In one embodiment provided by the present invention, when constructing an integrated learning model, an XGBoost regression algorithm is used to establish a mapping relationship between preparation parameters and performance indicators, and model parameters are set for the integrated learning model.
[0086] In the above technical solution, the model parameters include: the depth of the decision tree and the learning rate. For example, the maximum depth of the decision tree is 3 and the learning rate is 0.05.
[0087] In the above technical solution, the preparation parameters refer to the composition ratio, addition time, etc. of the composite material during the preparation process.
[0088] The above technical solution fully utilizes multi-core CPU and cluster resources through the XGBoost regression algorithm, effectively improves the model training speed and reduces time consumption. It also uses a tree-based model as a base learner to automatically process feature interactions, capture complex patterns in the data, and efficiently realize the fitting relationship between preparation parameters and performance indicators, facilitating the optimization of the integrated learning model, so that the integrated learning optimization model can be obtained in a shorter time. In addition, the XGBoost regression algorithm can process various types of data, thereby improving the applicability of the integrated learning model.
[0089] In one embodiment provided by the present invention, when optimizing an ensemble learning model using a training set and a test set, an optimization objective function is used to determine the optimal parameters of the model, wherein the optimization objective function is:
[0090]
[0091] In the above formula, y i is the actual result value of the i-th implementation data, is the model output value of the i-th implementation data, γ is the complexity penalty coefficient of the decision tree, T is the number of leaf nodes of the decision tree, λ is the regularization parameter, and w j is the weight of the j-th leaf node.
[0092] In the above technical solution, in the optimization objective function, Measure the overall prediction accuracy of the model, and use γT and Implementing double compound regularization can not only prevent the decision tree from over-splitting and overfitting, but also prevent the model from being sensitive to noise due to excessive weights.
[0093] The above technical solution uses a high-complexity joint control model to perform loss function analysis, and uses objective data to feedback the optimization effect of the integrated learning model, making it more intuitive to clarify whether the model parameters have reached the optimal situation, avoiding insufficient optimization of the integrated learning model resulting in large errors in model parameters or continuing to optimize the model when the model parameters have reached the optimal state, resulting in a waste of time. In addition, a regularization term is added to the optimization objective function to achieve double compound regularization, which can not only prevent the decision tree from over-splitting and avoid excessive weights that cause the model to be sensitive to noise, but also can simultaneously constrain the model structure and parameters through the complexity penalty coefficient and regularization parameter of the decision tree. By adjusting γ and λ, the structural complexity and weight smoothness of the tree can be controlled respectively. It is suitable for different data distributions (such as high variance or high deviation scenarios) to effectively prevent model overfitting, improve flexibility, and make the generalization ability stronger.
[0094] In one embodiment provided by the present invention, Figure 3 and Figure 4 As shown, after obtaining the process parameters of the target composite material, the process parameters of the target composite material are further verified, including:
[0095] S1. Conducting a test according to the process parameters of the target composite material and the process flow of the target composite material to obtain a target test composite material;
[0096] S2. Perform performance analysis on the target test composite material and obtain test analysis results;
[0097] In the above technical solution, when conducting experiments according to the process parameters of the target composite material combined with the process flow of the target composite material, in the process flow of the target composite material, the components of the composite material are added and the experimental preparation is regulated according to the process parameters of the target composite material to obtain the target composite material.
[0098] In the above technical solution, the performance analysis of the target test composite material includes:
[0099] Determine the purpose of the target composite material, and perform attribute characteristic analysis based on the purpose of the target composite material to determine the target attribute characteristics;
[0100] Conduct current performance analysis on the target composite material to determine the current performance of the target composite material;
[0101] Combined with the target attribute characteristics, the current performance of the target composite material is analyzed to see whether it meets the requirements and the test analysis results are obtained.
[0102] In the above technical solution, if the test analysis result shows that the current performance of the target composite material meets the requirements, the target composite material is prepared according to the process parameters of the target composite material at this time; if the test analysis result shows that the current performance of the target composite material does not meet the requirements, the process parameters of the target composite material are further optimized until the test analysis result shows that the current performance of the target composite material meets the requirements, the process parameters of the target composite material are determined, and then the target composite material is prepared according to the process parameters of the target composite material at this time.
[0103] The above technical solution breaks through the limitations of the traditional trial and error method by verifying the process parameters of the target composite material. By experimentally preparing the process parameters of the target composite material, the performance of the target composite material prepared based on the process parameters of the target composite material can be clarified, avoiding the waste of resources caused by blind preparation of composite materials and ensuring the application needs of the target composite material.
[0104] In one embodiment provided by the present invention, Figure 5 As shown, a multi-source literature database is constructed based on the target composite material, including:
[0105] B1. Analyze the composition of the target composite material to determine the composition of the composite material;
[0106] B2. Collecting literature information according to the components of the composite material to obtain a first collected literature file;
[0107] B3. Collecting literature information on the components of the composite material to obtain a second collected literature file;
[0108] B4. Construct a document database based on the first collected document files and the second collected document files to obtain a multi-source document database.
[0109] In the above technical solution, the components of the composite material refer to the main components of the composite material. For example, the target composite material is an Al2O3 / SiC composite material, and its components are SiC and Al2O3.
[0110] In the above technical solution, literature information is collected based on the composite material components. The composite material components are then specified according to different standards, such as SiC content (0-25wt%), SiC particle size (0.03-1μm), Al2O3 particle size (0.06-1μm), etc. Literature searches are then conducted based on performance indicators, resulting in multiple literature documents on the performance analysis of the composite material components, thereby determining the literature information collection results. Among these performance indicators, flexural strength, Vickers hardness, fracture toughness, etc.
[0111] In the above technical solution, the multi-source document database corresponds to the composite material. When constructing the document database based on the first and second collected document files, identification information is obtained for the target composite material, and the document database is created based on the identification information. The first and second collected document files are then stored in the document database to obtain the multi-source document database. Furthermore, when the first and second collected document files are stored in the document database, the first and second collected document files, obtained in real time, are analyzed in conjunction with the multi-source document database to determine whether they are already existing files. If so, there is no need to store the information in the document database again. If not, the information is stored in the document database.
[0112] The above technical solution analyzes the composition of the target composite material to clarify the composition of the composite material, so that literature information can be collected according to the composition of the composite material, improving the comprehensiveness of literature information collection, and ensuring that the multi-source literature database can contain literature files about composite materials in various situations, so that the multi-source literature database can better optimize the integrated learning model, improve the accuracy of the process parameters of the target composite material, and then better generate and manufacture the target composite material based on the process parameters of the target composite material, ensuring the application of the target composite material. Moreover, when the first collected literature file and the second collected literature file are stored in the literature file database, by combining the multi-source literature database for analysis, the same literature file is avoided from being stored multiple times in the literature file database, the effectiveness of information storage is improved, and the same literature file is avoided from appearing in the literature file database.
[0113] In one embodiment provided by the present invention, the standardized implementation data is divided into a training set and a test set, including:
[0114] Determine the ratio of training set and test set;
[0115] The standardized implementation data were randomly arranged to obtain the allocation sequence;
[0116] According to the allocation rules, the standardized implementation data in the allocation sequence are allocated to the training set and the test set in turn to obtain the training set and the test set.
[0117] In the above technical solution, the ratio of the training set to the test set can be set and adjusted according to needs. Usually, the number of implementation data in the training set is greater than the number of implementation data in the test set. For example, the ratio of the training set to the test set is 4:1.
[0118] In the above technical solution, the allocation rule is related to the total number of standardized implementation data. When the standardized implementation data in the allocation sequence are allocated to the training set and the test set in sequence according to the allocation rule, the total number of standardized implementation data is determined, and the ratio of the training set to the test set is analyzed to see if there is any excess in the total number of standardized implementation data. If there is any excess, the allocation rule is to first allocate the training set, then the test set, then the training set, and then the test set according to the ratio of the training set to the test set, until the excess standardized implementation data can no longer be fully allocated to the training set and the test set in one round, and then allocate the excess standardized implementation data to the training set. If there is no excess, cross-allocation is performed according to the ratio of the training set to the test set until all standardized implementation data are allocated.
[0119] The above technical solution determines the ratio of the training set and the test set so that when the standardized implementation data are divided into the training set and the test set, the training set and the test set are determined according to the ratio of the training set and the test set, and the number of standardized implementation data in the training set is greater than the number of standardized implementation data in the test set. The training set provides sufficient samples so that the integrated learning model can learn the subtle differences and complex structures of the data and optimize the parameters of the integrated learning model itself. At the same time, the test set is used to verify the accuracy and reliability of the integrated learning training model when facing new data, thereby improving the generalization ability of the integrated learning model, avoiding overfitting of the integrated learning model, and ensuring the optimization effect of the integrated learning model.
[0120] In one embodiment provided by the present invention, optimizing an ensemble learning model using a training set and a test set includes:
[0121] Train the ensemble learning model through the training set to determine the ensemble learning training model;
[0122] Using the test set to test the ensemble learning training model, and determining whether the ensemble learning training model is optimal based on the test data analysis, to obtain a first analysis result;
[0123] When the first analysis result shows that the ensemble learning training model is optimal, the ensemble learning training model at this time is the ensemble learning optimization model;
[0124] When the first analysis result is that the integrated learning training model has not reached the optimal level, the model training data of the training set when training the integrated learning model and the model test data of the test set when testing the integrated learning training model are obtained, and the training set and test are updated according to the model training data and model test data. Then, the updated training set and test set are used to continue to optimize the integrated learning model until the integrated learning training model reaches the optimal level, thereby obtaining an integrated learning optimization model.
[0125] In the above technical solution, the training set and the test set are updated according to the model training data and the model test data, including: performing model error analysis on the model training data and the model test data respectively to determine the model error of the implementation data; dividing the implementation data according to the model error, and dividing the implementation data into a first partition set and a second partition set in combination with the error preset value; wherein the model error of the implementation data of the first partition set is less than the model error of the implementation data of the second partition set; performing implementation data feature analysis on the first partition set and the second partition set to obtain the implementation data features of the first partition set and the implementation data features of the second set, wherein the implementation data features are a combination of the composition of the composite material and the performance indicators, such as: SiC content and flexural strength, SiC particle size and Vickers hardness, Al2O3 particle size and flexural strength, etc. According to the implementation data types of the first partition set and the implementation data types of the second partition set, the implementation data type characteristics with larger model errors are determined, and data is collected according to the implementation data type characteristics with larger model errors to obtain secondary collected data; implementation data acquisition and standardization processing are performed on the secondary collected data to obtain newly added standardized implementation data; model SHAP eigenvalues are calculated for the standardized implementation data in the first partition set and the second partition set, and implementation data are eliminated in proportion according to the SHAP eigenvalues to obtain remaining standardized implementation data; the remaining standardized implementation data are analyzed in combination with the model training data or the model test data to determine whether the remaining standardized implementation data need to be labeled and corrected. If standard correction is required, the remaining standardized implementation data that need standard correction are corrected to obtain the processed remaining standardized implementation data; the processed remaining standardized implementation data are integrated with the newly added standardized implementation data, and the integrated data are allocated according to the proportion of the training set and the test set according to the allocation rule to obtain an updated training set and an updated test set.
[0126] In the above technical solution, when the updated training set and test set are used to continue to optimize the integrated learning model, the updated training set is used to continue model training for the integrated learning training model to obtain a re-trained integrated learning training model, and the updated test set is used to test the re-trained integrated learning training model. Then, based on the test data analysis, it is determined whether the integrated learning training model has reached the optimal level. If it has reached the optimal level, the re-trained integrated learning training model is the integrated learning optimization model. Otherwise, the training set and test are continued to be updated.
[0127] In the above technical solution, the random forest model is trained with the training set data. The model will be able to identify the relationship between each input parameter and performance indicator, and improve the prediction accuracy through ensemble learning. Subsequently, the trained model is verified using the test set data to evaluate its performance on unknown data. The prediction accuracy of the model will be measured by the root mean square error (RMSE), the coefficient of determination (R 2 ) and other indicators to verify the effectiveness of the model. Performance indicators include: bending strength, Vickers hardness, fracture toughness, etc. Figure 6 、 Figure 7 、 Figure 8 As shown, the analytical data of different performance indicators on the prediction accuracy of the model.
[0128] The above technical solution optimizes the integrated learning model through the training set, and detects the integrated learning optimization model through the test set, thereby ensuring the accuracy of the integrated learning optimization model. When the first analysis result shows that the integrated learning training model has not reached the optimal level, the training set and test are updated according to the model training data and model test data, so that the training set and test set are adjusted according to the model training data and model test data of the integrated learning model, thereby optimizing the shortcomings of the integrated learning model, enabling the machine learning model to better adapt to changes in actual applications, and improving the generalization ability and performance of the model.
[0129] like Figure 9 As shown, an embodiment of the present invention provides a composite material analysis system based on machine learning, comprising: a target confirmation unit, a literature collection unit, a data processing unit, a model construction unit, a model optimization unit and a parameter optimization unit;
[0130] The target confirmation unit is used to determine the target composite material;
[0131] The document collection unit is used to construct a multi-source document database based on the target composite material;
[0132] The data processing unit is used to obtain implementation data from a multi-source document database, perform standardization processing on the implementation data to obtain standardized implementation data, and then divide the standardized implementation data into a training set and a test set;
[0133] The model building unit is used to build an integrated learning model;
[0134] The model optimization unit is used to optimize the ensemble learning model using the training set and the test set to obtain the ensemble learning optimization model;
[0135] The parameter optimization unit is used to optimize parameters by combining an integrated learning optimization model with the process flow of the target composite material to obtain the process parameters of the target composite material.
[0136] In one embodiment provided by the present invention, Figure 10 As shown, the composite material analysis system further includes: a parameter verification unit;
[0137] After obtaining the process parameters of the target composite material, the parameter verification unit verifies the process parameters of the target composite material, including:
[0138] Conducting tests according to the process parameters of the target composite material and the process flow of the target composite material to obtain the target test composite material;
[0139] The performance analysis is carried out on the target test composite material to obtain the test analysis results.
[0140] A composite material analysis system based on machine learning corresponds to a composite material analysis method based on machine learning. The working principle and beneficial effects of the composite material analysis system have been described in the corresponding composite material analysis method embodiment and will not be repeated here.
[0141] Those skilled in the art should understand that the first and second in the present invention merely refer to different application stages.
[0142] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the disclosure herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow from the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.
[0143] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims
1. A composite material analysis method based on machine learning, characterized in that: include: Identify target composite materials; Construct a multi-source literature database based on target composite materials; Obtain implementation data from a multi-source literature database; Performing standardization on the implementation data to obtain standardized implementation data, and dividing the standardized implementation data into a training set and a test set; Build an integrated learning model; The training set and the test set are used to optimize the ensemble learning model to obtain the ensemble learning optimization model; The process parameters of the target composite material are obtained by combining the integrated learning optimization model with the process flow of the target composite material to optimize the parameters.
2. The composite material analysis method according to claim 1, characterized in that: Standardize implementation data, including: Performing data standardization on the implementation data information of the literature file to obtain first processed implementation data; The first processing implementation data information is cleaned to obtain standardized implementation data.
3. The composite material analysis method according to claim 1, characterized in that: When constructing the ensemble learning model, the XGBoost regression algorithm is used to establish a mapping relationship between the preparation parameters and the performance indicators, and the model parameters are set for the ensemble learning model.
4. The composite material analysis method according to claim 1, characterized in that: When optimizing the ensemble learning model using the training set and the test set, the optimization objective function is used to determine the optimal model parameters. The optimization objective function is: In the above formula, y i is the actual result value of the i-th implementation data, is the model output value of the i-th implementation data, γ is the complexity penalty coefficient of the decision tree, T is the number of leaf nodes of the decision tree, λ is the regularization parameter, and w j is the weight of the j-th leaf node.
5. The composite material analysis method according to claim 1, characterized in that: After obtaining the process parameters of the target composite material, the process parameters of the target composite material are also verified, including: Conducting tests according to the process parameters of the target composite material and the process flow of the target composite material to obtain the target test composite material; The performance analysis is carried out on the target test composite material to obtain the test analysis results.
6. The method according to claim 1, characterized in that Construct a multi-source literature database based on target composite materials, including: Analyze the composition of the target composite material to determine the composition of the composite material; Collecting literature information according to the components of the composite material to obtain a first collected literature file; Combining the components of the composite material to collect literature information, to obtain a second collected literature file; A document file database is constructed based on the first collected document files and the second collected document files to obtain a multi-source document database.
7. The method according to claim 1, characterized in that The standardized implementation data is divided into training and test sets, including: Determine the ratio of training set and test set; The standardized implementation data were randomly arranged to obtain the allocation sequence; According to the allocation rules, the standardized implementation data in the allocation sequence are allocated to the training set and the test set in turn to obtain the training set and the test set.
8. The method according to claim 2, characterized in that Optimize the ensemble learning model using training and test sets, including: Train the ensemble learning model through the training set to determine the ensemble learning training model; Using the test set to test the ensemble learning training model, and determining whether the ensemble learning training model is optimal based on the test data analysis, to obtain a first analysis result; When the first analysis result shows that the ensemble learning training model is optimal, the ensemble learning training model at this time is the ensemble learning optimization model; When the first analysis result is that the integrated learning training model has not reached the optimal level, the model training data of the training set when training the integrated learning model and the model test data of the test set when testing the integrated learning training model are obtained, and the training set and test are updated according to the model training data and model test data. Then, the updated training set and test set are used to continue to optimize the integrated learning model until the integrated learning training model reaches the optimal level, thereby obtaining an integrated learning optimization model.
9. A composite material analysis system based on machine learning, characterized in that: The composite material analysis system includes: a target confirmation unit, a literature collection unit, a data processing unit, a model construction unit, a model optimization unit and a parameter optimization unit; The target confirmation unit is used to determine the target composite material; The document collection unit is used to construct a multi-source document database based on the target composite material; The data processing unit is used to obtain implementation data from a multi-source document database, perform standardization processing on the implementation data to obtain standardized implementation data, and then divide the standardized implementation data into a training set and a test set; The model building unit is used to build an integrated learning model; The model optimization unit is used to optimize the ensemble learning model using the training set and the test set to obtain the ensemble learning optimization model; The parameter optimization unit is used to optimize parameters by combining an integrated learning optimization model with the process flow of the target composite material to obtain the process parameters of the target composite material.
10. The composite material analysis system according to claim 9, characterized in that: The composite material analysis system further includes: a parameter verification unit; After obtaining the process parameters of the target composite material, the parameter verification unit verifies the process parameters of the target composite material, including: Conducting tests according to the process parameters of the target composite material and the process flow of the target composite material to obtain the target test composite material; The performance analysis is carried out on the target test composite material to obtain the test analysis results.