Cyanate ester / quartz fiber mechanical property optimization method based on machine learning and related device

Abstract

The invention belongs to the technical field of material science, and discloses a cyanate ester / quartz fiber mechanical property optimization method based on machine learning and a related device. Recommending a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material; utilizing a trained machine learning model to recommend components and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester; and screening the recommended components and process parameters of the cyanate ester to obtain the components and process parameters of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material. According to the method, limited production line data can be utilized, the experiment frequency and cost are reduced, the cyanate performance is rapidly optimized, and then the high-performance composite material is obtained.

Description

Technical Field

[0001] The present invention belongs to the field of material science and technology, and specifically relates to a method for optimizing the mechanical properties of cyanate ester / quartz fibers based on machine learning and a related device. Background Art

[0002] Fiber reinforced composite materials are composite materials made of reinforcing fibers and matrix materials through molding, pultrusion, winding and other processes. Cyanate ester / quartz fiber composite materials are widely used in aircraft, ground radar, satellite, 5G communication antenna covers due to their high strength, light weight, excellent wave transmission performance and strong design. With the development of aerospace and national defense equipment construction, higher requirements are placed on the tensile strength of cyanate ester / quartz fiber composite materials. Only by mastering the laws that affect mechanical properties can we try to control the experimental parameters and ensure that the material has excellent performance during service.

[0003] At present, in actual production, composite materials need to be made into composite rolls by composite coating machines and composite impregnation machines. After the composite materials are rolled, they need to be layered and cured before delivery to sample and test their tensile strength to ensure that their performance is qualified. This product production and performance testing process is time-consuming and labor-intensive. Shortening the delivery cycle of composite materials and reducing the cost of sampling and testing are expected to become important ways for companies to reduce costs and increase efficiency. The performance of composite materials is closely related to the performance of resins. The mechanical properties of cyanate esters are determined by the ingredients and curing process, but there are many combinations of catalyst toughening agent content and curing process, which are difficult to verify one by one through experiments. How to use limited production line data, reduce the number and cost of experiments, and quickly optimize the performance of cyanate esters to obtain high-performance composite materials has become an urgent problem to be solved. Summary of the invention

[0004] In order to solve the defects existing in the above-mentioned prior art, the purpose of the present invention is to provide a method and related device for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning. The present invention can utilize limited production line data, reduce the number and cost of experiments, quickly optimize the performance of cyanate ester, and thus obtain high-performance composite materials.

[0005] To achieve the above purpose, the technical solution adopted by the present invention is as follows: A method for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning includes the following steps: Using the trained neural network prediction model, a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material are recommended; the input of the neural network prediction model is the tensile modulus and tensile strength of the cyanate ester, and the output is the tensile strength of the cyanate ester / quartz fiber composite material; Using the trained machine learning model, the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester are recommended; the input of the machine learning model is the composition and process parameters of the cyanate ester, and the output is the tensile modulus and tensile strength of the cyanate ester; The recommended cyanate ester components and process parameters are screened to obtain the cyanate ester components and process parameters corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

[0006] Preferably, the process of using the trained neural network prediction model to recommend the target tensile modulus and tensile strength of the cyanate ester corresponding to the target tensile strength of the cyanate ester / quartz fiber composite material includes: Using the gradient descent algorithm, the tensile strength target of the cyanate ester / quartz fiber composite material, the initial value of the cyanate ester tensile modulus and the initial value of the tensile strength are first set, and then the mean square error is used as the loss function, and the gradient of the loss function and the input of the neural network prediction model is calculated. The cyanate ester tensile modulus and tensile strength are continuously iterated and optimized. When the loss function is stable, the neural network model corresponding to the tensile strength of the cyanate ester / quartz fiber composite material output by the neural network prediction model is input into the tensile modulus and tensile strength of the cyanate ester as the tensile modulus target and tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

[0007] Preferably, the process of using the trained machine learning model to recommend the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester includes: By calculating the tensile modulus and tensile strength of the cyanate ester corresponding to the same component and process combination output by the machine learning model, and the Euclidean distance from the tensile modulus and tensile strength of the target cyanate ester, several groups of cyanate ester component and process combinations corresponding to the smaller Euclidean distance are selected as the components and process parameters of the cyanate ester corresponding to the tensile modulus target and tensile strength target of the cyanate ester; wherein the components and process parameters of the cyanate ester corresponding to the tensile modulus target and tensile strength target of the cyanate ester are the components and process parameters of the cyanate ester that can make the tensile strength of the cyanate ester / quartz fiber composite material not less than the tensile strength target of the cyanate ester / quartz fiber composite material.

[0008] Preferably, the training process of the neural network prediction model includes: The neural network prediction model is pre-trained using the original virtual data set to obtain a pre-trained neural network prediction model; wherein the construction process of the original virtual data set includes: using the Standard module in the Abaqus software to perform finite element calculations to obtain the tensile strength of the cyanate ester / quartz fiber composite material corresponding to different cyanate ester tensile moduli and tensile strengths, and the tensile strength of the cyanate ester / quartz fiber composite material corresponding to the different cyanate ester tensile moduli and tensile strengths constitutes the original virtual data set; The pre-trained neural network prediction model is fine-tuned using the original real data set to obtain the trained neural network prediction model; wherein the construction process of the original real data set includes: conducting a tensile mechanical property test of cyanate ester / quartz fiber composite materials under different cyanate ester ingredients and curing process parameters to obtain the tensile strength of the cyanate ester / quartz fiber composite materials corresponding to different cyanate ester tensile modulus and tensile strength, and the tensile strength of the cyanate ester / quartz fiber composite materials corresponding to the different cyanate ester tensile modulus and tensile strength constitute the original real data set.

[0009] Preferably, the training process of the machine learning model includes: The tensile modulus and tensile strength test data corresponding to different process combinations of cyanate ester components are constructed into a data set for training a machine learning model; the data set for training a machine learning model includes: phenolphthalein polyaryletherketone content, polysulfone content, polyimide content, acetylacetonate cobalt content, organic tin content, nonylphenol content, curing temperature, curing time, curing pressure, tensile modulus and tensile strength; The machine learning model is trained using the data set for training the machine learning model to obtain a trained machine learning model; during training, the input parameters of the machine learning model are first normalized, and then the input is processed by the machine learning model to obtain the tensile modulus and tensile strength of the cyanate ester.

[0010] Preferably, the machine learning model adopts a linear regression model, a polynomial regression model, a support vector machine regression model, a random forest regression model, an extreme gradient boosting regression model or a Gaussian process regression model; The neural network prediction model adopts an artificial neural network regression model.

[0011] Preferably, the tensile modulus and tensile strength of the cyanate ester are normalized, and the normalized result is processed using a neural network prediction model to obtain the tensile strength of the cyanate ester / quartz fiber composite material.

[0012] The present invention also provides a cyanate ester / quartz fiber mechanical property optimization system based on machine learning, comprising: A first screening unit is used to recommend a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material by using a trained neural network prediction model; the input of the neural network prediction model is the tensile modulus and tensile strength of the cyanate ester, and the output is the tensile strength of the cyanate ester / quartz fiber composite material; A second screening unit is used to recommend the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester by using the trained machine learning model; the input of the machine learning model is the composition and process parameters of the cyanate ester, and the output is the tensile modulus and tensile strength of the cyanate ester; The third screening unit is used to screen the recommended cyanate ester composition and process parameters to obtain the cyanate ester composition and process parameters corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

[0013] The present invention also provides an electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the cyanate ester / quartz fiber mechanical property optimization method based on machine learning as described above in the present invention.

[0014] The present invention also provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning as described above is implemented in the present invention.

[0015] The present invention has the following beneficial effects: The present invention uses artificial intelligence to mine the relationship between the performance of component materials and the performance of composite materials for the composite material production process based on the performance data of component materials, establishes a digital analysis of performance evolution, and can realize the rapid prediction of composite material performance. After the composite material is rolled, by using the predicted value instead of the mechanical property test value of the composite material, the frequency of composite material performance sampling and testing can be greatly reduced, the delivery cycle can be shortened, and the economic cost can be reduced. On this basis, combined with big data analysis, it can also realize product quality abnormality tracing and auxiliary product process optimization design, and accelerate the product research and development process. For the optimized cyanate ester mechanical property target, active learning can select samples for testing in a targeted manner according to the prediction results of the current model, and quickly narrow the scope of the optimal solution space. The model obtained by active learning can provide targeted guidance, help to quickly find the optimal ingredients in the material design and optimization process, thereby improving the mechanical properties of cyanate ester. Active learning can improve sample utilization efficiency, reduce experimental costs, quickly explore the optimal solution space, and provide guidance for material design and optimization in the process of realizing the optimal ingredients of cyanate ester. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute an improper limitation of the present invention. In the drawings: Figure 1 This is a schematic diagram of the mechanical property prediction and optimization process of cyanate ester / quartz fiber composite materials based on machine learning in the present invention.

[0017] Figure 2 It is a schematic diagram of the optical microscope structure of the cyanate ester / quartz fiber composite material in the present invention.

[0018] Figure 3 It is a schematic diagram of the finite element periodic mesh division in the present invention.

[0019] Figure 4 It is a schematic diagram of the tensile force-displacement curve of the cyanate ester / quartz fiber composite material simulated by finite element in the present invention.

[0020] Figure 5 The figure is a schematic diagram comparing the predicted data and actual data of the tensile strength fitting of the neural network prediction model of the present invention to the cyanate ester / quartz fiber composite material.

[0021] Figure 6 Schematic diagram of the tensile modulus model performance comparison of the machine learning model used in the present invention.

[0022] Figure 7 Schematic diagram of the tensile strength model performance comparison of the machine learning model used in the present invention.

[0023] Figure 8The figure is a schematic diagram comparing the predicted data and the actual data of the Gaussian process regression model for fitting the tensile modulus of cyanate ester in the present invention.

[0024] Fig. 9 The figure is a schematic diagram comparing the predicted data and the actual data of the Gaussian process regression model for fitting the tensile strength of cyanate ester in the present invention. DETAILED DESCRIPTION

[0025] The present invention will be described in detail below in conjunction with the accompanying drawings and specific embodiments. The exemplary embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0026] See also Figure 1 The method for optimizing the mechanical properties of the cyanate ester / quartz fiber composite material of the present invention comprises the following steps: Obtaining original data of composite materials obtained through tensile mechanical property test of cyanate ester / quartz fiber composite materials and finite element simulation calculation; the original data is the tensile modulus and tensile strength of cyanate ester / quartz fiber composite materials corresponding to the tensile strength under different composition processes of cyanate ester; the original virtual data set obtained through finite element simulation calculation is the original real data set, and the data set obtained through tensile mechanical test is the original real data set; using the original virtual data set obtained through finite element simulation calculation as a pre-training data set for training a neural network prediction model, dividing the data set into a training set and a test set; training a neural network prediction model according to the training set, and evaluating the accuracy of the neural network prediction model through the test set; using the original real data set obtained through the tensile mechanical property test of cyanate ester / quartz fiber composite materials to fine-tune the neural network prediction model and obtain a trained neural network prediction model; The trained neural network prediction model is used to recommend the cyanate ester tensile modulus target (referring to the target value of the tensile modulus of the cyanate ester) and the tensile strength target (referring to the target value of the tensile strength of the cyanate ester / quartz fiber composite material) corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material; the recommendation process uses a gradient descent algorithm, firstly gives the tensile strength target of the cyanate ester / quartz fiber composite material and the initial value of the tensile modulus of the cyanate ester and the initial value of the tensile strength, uses the mean square error (Mean Square Error) as the loss function, calculates the gradient of the loss function and the input (i.e., the tensile modulus of the cyanate ester and the tensile strength), and continuously iterates and optimizes the tensile modulus of the cyanate ester and the tensile strength, so that the loss function is stable, at which time the predicted value of the neural network prediction model is close to the tensile strength target of the cyanate ester / quartz fiber composite material initially set, and the tensile modulus and tensile strength of the cyanate ester corresponding to the predicted value of the tensile strength of the cyanate ester / quartz fiber composite material at this time are used as the recommended target (i.e., the tensile modulus target and the tensile strength target of the cyanate ester); the loss function means square error calculates the variance of the predicted value of the neural network prediction model and the tensile strength target of the composite material; The original cyanate data set of the machine learning model is obtained according to the cyanate test data, and the original cyanate data set is divided into a training set and a test set; the original cyanate data set is the tensile modulus and tensile strength corresponding to different cyanate component processes; the machine learning model is trained according to the training set, and the accuracy of the machine learning model is evaluated through the test set to obtain the trained machine learning model; the components and process parameters corresponding to the tensile modulus and tensile strength targets of the cyanate are recommended according to the trained machine learning model; the recommendation process is carried out by calculating the Euclidean distance between the tensile modulus and tensile strength of the cyanate corresponding to the same component process combination predicted by the machine learning model and the target cyanate tensile modulus and tensile strength, and sorting them from small to large The invention discloses a method for preparing the cyanate ester composite material of the present invention, wherein the composition and process combination of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester is selected; wherein the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester are the composition and process parameters of the cyanate ester that can make the tensile strength of the cyanate ester / quartz fiber composite material not less than the tensile strength target of the cyanate ester / quartz fiber composite material; wherein the number of the composition and process combination of the cyanate ester can be determined according to actual conditions, and can also be flexibly adjusted. The present invention does not make any specific limitation, as long as the tensile strength of the cyanate ester / quartz fiber composite material can be not less than the tensile strength target of the cyanate ester / quartz fiber composite material. In the above scheme, the typical component and process combination of cyanate ester includes, in terms of mass percentage: the toughening agent component of cyanate ester is phenolphthalein polyaryletherketone (PEK-C) 0-20%, polysulfone (PSF) 0-20%, polyimide (PI) 0-20%, and the total toughening agent components add up to no more than 20%; the catalyst component is cobalt acetylacetonate 0-0.08%, organic tin 0-0.08%, and the total catalyst components add up to no more than 0.08%; the co-catalyst component is nonyl 0-8%; the curing temperature is 170-220°C, the curing time is 120-480min, and the curing pressure is 0.2-0.8MPa. When recommending, the machine learning model traverses all the components and process combinations of cyanate ester, and uses the machine learning model to predict the tensile modulus and tensile strength of cyanate ester under different component and process combinations.

[0027] Finally, the recommended cyanate ester can be synthesized and made into cyanate ester / quartz fiber composite materials, and the mechanical properties can be tested to verify the recommended results and obtain the optimized composite materials.

[0028] In the above scheme of the present invention, the Standard module in the Abaqus software can be used to perform finite element simulation calculations to obtain relevant mechanical parameters of the cyanate ester / quartz fiber composite material. When performing finite element simulation calculations, the tensile modulus and tensile strength of different cyanate esters are changed to obtain the corresponding tensile strength of the cyanate ester / quartz fiber composite material.

[0029] In the above scheme of the present invention, the mechanical property test includes: a cyanate ester tensile test to obtain the tensile strength and tensile modulus of the cyanate ester; a cyanate ester / quartz fiber composite material tensile test to obtain the tensile strength of the cyanate ester / quartz fiber composite material.

[0030] In the above solution of the present invention, the neural network data set division method is a 10-fold cross validation method, comprising: Divide the training set into 10 parts; 9 of them are used as training sets in turn to train the neural network prediction model, and the remaining 1 is used as a test set to find the optimal parameters and calculate the prediction accuracy of the test set; The prediction accuracy is averaged and used as an evaluation of the prediction accuracy of the model.

[0031] In the above solution of the present invention, the machine learning model data set partitioning method is a leave-one-out cross-validation method, comprising: Divide the total number of training sets K into K parts; Take turns using K-1 copies as training sets to train the machine learning model, and use the remaining 1 copy as a test set to find the optimal parameters and calculate the prediction accuracy of the test set; The prediction accuracy is averaged and used as an evaluation of the prediction accuracy of the model.

[0032] In the above scheme of the present invention, the machine learning model can adopt any one of a linear regression model (Linear), a polynomial regression model (Poly), a support vector machine regression model (SVR), a random forest regression model (RF), an extreme gradient boosting regression model (XGBR) and a Gaussian process regression model (GPR).

[0033] In the above scheme of the present invention, the regression coefficient and the root mean square error are used to measure the prediction accuracy of the machine learning model on the test set.

[0034] In the above scheme of the present invention, the predicted result is detected by a tensile testing machine, including: measuring the force and displacement curve during the experiment by a sensor to obtain a stress-strain curve; and calculating the elastic modulus and tensile strength according to the stress-strain curve.

[0035] Example Please refer to Figure 1 The actual example of the machine learning-based optimization method for the mechanical properties of cyanate ester / quartz fiber includes the following steps: Step 1, obtaining original data obtained through mechanical property test and finite element simulation calculation of cyanate ester / quartz fiber composite material; Specifically, mechanical property tests and finite element simulations were conducted on cyanate ester / quartz fiber under different cyanate ester ingredients and curing process parameters, and the tensile strength of the cyanate ester / quartz fiber composite material was simulated based on the finite element method.

[0036] In this embodiment, the tensile modulus and tensile strength of cyanate ester are obtained by changing the cyanate ester ingredients and the curing process parameters, and the corresponding tensile strength mechanical performance index is obtained by synthesizing the cyanate ester / quartz fiber composite material.

[0037] Specifically, finite element calculations were performed in the Standard module of the Abaqus software to obtain the tensile strength of the cyanate ester / quartz fiber composite material corresponding to different cyanate ester tensile modulus and tensile strength.

[0038] The fiber size and void size were determined under an optical microscope, and a unit cell model of cyanate / quartz fiber composite material was established in UG software. Figure 2 Schematic diagram of the optical microscope structure of cyanate / quartz fiber composite material. The unit cell model is structured using Hypermesh software to generate a periodic grid. Figure 3 The figure is a schematic diagram of the periodic grid of the unit cell model. In the Abaqus software, the Easypbc plug-in is used to apply periodic boundary conditions to the unit cell to obtain the elastic modulus parameters of the composite material.

[0039] The tensile strength of the composite material is studied by using the Abaqus custom field subroutine (USDFLD), defining field variables and establishing the relationship between the properties of the component materials cyanate ester and quartz fiber and the changes in field variables. The failure criterion of the component material adopts the maximum stress criterion, and USDFLD is used as the interface for secondary development. The constitutive model of the component material is defined in the USDFLD subroutine, and the Abaqus main program is used to simulate the tensile condition. Finally, the tensile strength of the cyanate ester / quartz fiber composite material is obtained. Figure 4 Schematic diagram of the tensile force-displacement curve of cyanate ester / quartz fiber composite material simulated by finite element method.

[0040] The ingredients and process parameters of the cyanate ester component material of the cyanate ester / quartz fiber composite material are adjusted to obtain the corresponding tensile strength of the final composite material.

[0041] In this embodiment, the toughening agent component of the cyanate ester is selected as phenolphthalein polyaryletherketone (PEK-C) 0-20%, polysulfone (PSF) 0-20%, polyimide (PI) 0-20%, and the total toughening agent components are not more than 20%; the catalyst component is cobalt acetylacetonate 0-0.08%, organic tin 0-0.08%, and the total catalyst components are not more than 0.08%; the co-catalyst component is nonyl 0-8%; the curing temperature is 170-220°C, the curing time is 120-480min, and the curing pressure is 0.2-0.8MPa.

[0042] Step 2, dividing the data set according to the original data set and training the neural network prediction model; Specifically, the original data set for training the neural network prediction model is obtained based on the data generated by the finite element simulation. The finite element method generates a data set for the tensile strength of cyanate ester / quartz fiber, and the data set is divided into a training set and a test set. The data set division method in this step is the 10-fold cross-validation method: the training set is divided into 10 parts, 9 of which are used as training sets to train the neural network prediction model in turn, and the remaining 1 part is used as the test set. The prediction accuracy of the test set is calculated, and the prediction accuracy is averaged as an evaluation of the model prediction accuracy to determine the optimal hyperparameters of the neural network prediction model.

[0043] The data set includes three features, namely, the tensile modulus and tensile strength of cyanate ester, and the tensile strength of cyanate ester / quartz fiber composite material.

[0044] The data set is normalized. Since the units of the input data of the neural network prediction model are different, the range of some data may be particularly large, resulting in slow convergence of the neural network and long training time. Since the range of the activation function of the output layer of the neural network prediction model is limited, the target data of the network training needs to be mapped to the range of the activation function.

[0045] Training a neural network prediction model based on the training data set and adjusting and optimizing the parameters of the model, and evaluating the accuracy of the prediction of the neural network prediction model on the test set; For the tensile strength dataset, the best neural network structure after debugging is: the number of neural network layers is 6, of which there are 4 hidden layers. The number of neurons in the input layer is 2, the number of neurons in the hidden layer is 256, 128, 64, 32, and the number of neurons in the output layer is 1. The hidden layer uses the relu activation function, the number of training iterations is 100, and the learning rate is 0.0001; the optimal network structure parameters are saved as the pre-training model for transfer learning.

[0046] Importing the optimal model network structure parameters fitted by finite element data, using the cyanate ester / quartz fiber composite material test data to fine-tune the neural network prediction model and obtain a neural network prediction model trained after transfer learning; Observe the performance of the neural network prediction model in predicting the tensile strength of cyanate ester / quartz fiber composites. Figure 5 This is a schematic diagram comparing the predicted data and actual data of the tensile strength fitting of the cyanate ester / quartz fiber composite material by the neural network prediction model after transfer learning in the present invention, and the optimal network structure parameters are retained as the base model of the gradient descent optimization algorithm.

[0047] The maximum tensile strength of cyanate / quartz fiber composite materials in the current data set is 597MPa. The tensile strength target value of cyanate / quartz fiber composite materials is set to 600MPa, the cyanate tensile modulus is initialized to 3000MPa, and the tensile strength is initialized to 60MPa. There are 16 sets of cyanate tensile modulus and tensile strength combinations. The initialized cyanate tensile modulus and tensile strength are added with a random value to ensure that the 16 sets of initial values ​​are different. The gradient descent method is used to recommend the target modulus and tensile strength of cyanate. The mean square error is used as the loss function. The gradient of the loss function and the input (i.e., cyanate tensile modulus and tensile strength) is calculated. The cyanate tensile modulus and tensile strength are continuously iterated and optimized to make the loss function stable. At this time, the predicted value of the neural network prediction model is close to the tensile strength target of the composite material initially set. The cyanate tensile modulus and tensile strength corresponding to the predicted value at this time are used as the recommended target. The final recommended cyanate ester mechanical property combination 1 is a tensile modulus of 4202.93MPa, a tensile strength of 66.46MPa, and a corresponding composite tensile strength of 599.89MPa; the cyanate ester mechanical property combination 2 is a tensile modulus of 4125.73MPa, a tensile strength of 73.26MPa, and a corresponding composite tensile strength of 598.73MPa.

[0048] The neural network prediction model is an artificial neural network regression model, which uses the root mean square error (RMSE) and regression coefficient (R 2 ) to observe and measure the prediction accuracy of the model on the test set.

[0049] Step 3, please refer to Figure 1 , establish a machine learning model and recommend the ingredients and process parameters corresponding to the mechanical properties of cyanate esters; Specifically, the original data set of the machine learning model is obtained according to the tensile modulus and tensile strength test data corresponding to different component process combinations of cyanate esters, and the original data set is divided into a training set and a test set.

[0050] The data set division method in this step is the leave-one-out cross-validation method: divide the total number of training sets into K parts, use K-1 of them as training sets in turn to train the machine learning model, and use the remaining 1 part as the test set to find the optimal parameters and calculate the prediction accuracy of the machine learning model in the test set. The prediction accuracy is averaged as an evaluation of the prediction accuracy of the machine learning model.

[0051] The data set contains 11 features, namely, phenolphthalein polyaryletherketone (PEK-C) content, polysulfone (PSF) content, polyimide (PI) content, cobalt acetylacetonate content, organic tin content, nonylphenol content, curing temperature, curing time, curing pressure, tensile modulus of cyanate ester and tensile strength of cyanate ester.

[0052] Normalize the data set. Since the units of the input data are different, some data have a particularly large range. When the numerical ranges of the features vary greatly, the model may give higher weights to features with larger numerical ranges, thereby ignoring features with smaller numerical ranges, even though these features may be more important for the prediction results. Therefore, normalization is needed to ensure that all features are treated fairly during the model training process to improve the accuracy of the model.

[0053] The machine learning model is trained according to the training set, and the prediction accuracy of the machine learning model is evaluated by the test set to obtain a trained machine learning model. A total of six models are used to fit the tensile modulus of cyanate esters, and the six models are linear regression model (Linear), polynomial regression model (Poly), support vector machine regression model (SVR), random forest regression model (RF), extreme gradient boosting regression model (XGBR), and Gaussian process regression model (GPR). Figure 6 The figure is a schematic diagram of the performance comparison of the tensile modulus model of the machine learning model used in this embodiment. It can be seen that the regression coefficient and root mean square error of the Gaussian process regression model are better than those of other models, and the Gaussian process regression model has the highest prediction accuracy for the tensile modulus of cyanate esters.

[0054] The machine learning model is trained according to the training set, and the prediction accuracy of the machine learning model is evaluated through the test set to obtain a trained machine learning model. A total of six models are used to fit the tensile strength of cyanate esters, and the six models are linear regression model (Linear), polynomial regression model (Poly), support vector machine regression model (SVR), random forest regression model (RF), extreme gradient boosting regression model (XGBR), and Gaussian process regression model (GPR). Figure 7Schematic diagram of the performance comparison of the tensile strength model of the machine learning model used in this embodiment. It can be seen that the regression coefficient and root mean square error of the Gaussian process regression model are better than those of other models, and the tensile strength prediction accuracy of cyanate ester using the Gaussian process regression model is the highest.

[0055] Observe the performance of cyanate ester tensile modulus and tensile strength in Gaussian process regression model, Figure 8 This is a comparison chart of the predicted data and actual data of the tensile modulus of cyanate esters by the Gaussian process regression model in the present invention; Fig. 9 This is a comparison chart of the predicted data and actual data of the tensile strength of cyanate esters by the Gaussian process regression model in the present invention. The Gaussian process regression model has excellent performance in the tensile modulus and tensile strength of cyanate esters, so the Gaussian process model is recommended as the base model for cyanate ester experiments.

[0056] The cyanate component process space is defined, and the component process combination includes: the toughening agent component of cyanate is phenolphthalein polyaryletherketone (PEK-C) 0-20%, step length 1%; polysulfone (PSF) 0-20%, step length 1%; polyimide (PI) 0-20%, step length 1%; the total toughening agent components add up to no more than 20%; the catalyst component cobalt acetylacetonate 0-0.08%, step length 0.01%; organic tin 0-0.08%, step length 0.01%; the total catalyst components add up to no more than 0.08%; the co-catalyst component nonyl 0-8%, step length 1%; the curing temperature is 170-220℃, step length 10℃; the curing time is 120-480min, step length 60min; the curing pressure is 0.2-0.8MPa, step length 0.1MPa. All component and process combinations are traversed, and the machine learning model is used to predict the tensile modulus and tensile strength of cyanate under different component process combinations.

[0057] By calculating the Euclidean distance between the tensile modulus and tensile strength of the cyanate ester corresponding to the same component process combination predicted by the Gaussian process regression model and the target cyanate ester tensile modulus and tensile strength, and sorting the Euclidean distances from small to large, the components and process parameters corresponding to the mechanical performance target of the cyanate ester are recommended. The Euclidean distance calculation formula is as follows:

[0058] in , are the means of the predicted values ​​of tensile modulus and tensile strength of cyanate ester by Gaussian process regression model; is the Euclidean distance currently calculated; and The target cyanate ester tensile modulus and tensile strength are respectively selected. The composition and process parameters of the cyanate ester corresponding to the smaller Euclidean distance are selected.

[0059] The recommended cyanate esters were synthesized and tensile tests were performed to verify the model optimization effect.

[0060] For the target combination of cyanate ester tensile modulus and tensile strength, the recommended cyanate ester 1 composition and process parameters are shown in Table 1: Table 1

[0061] Example 1 of the present invention uses the recommended cyanate ester for verification, and conducts a tensile test according to GB / T2567-2021 "Test Method for Properties of Resin Castings". A universal testing machine is used at room temperature. The force and displacement curve during the experiment is measured by a sensor, and then the stress-strain curve is obtained, and the tensile strength is obtained by curve analysis.

[0062] The total length of the tensile specimen of the test material of Example 1 of the present invention is 200 mm, the length of the working section is 60 mm, the width is 10 mm, the thickness is 4 mm, and the tensile rate is 2 mm / min.

[0063] The stress-strain curve of the cyanate ester in the tensile state is drawn based on the force and displacement curve measured by the sensor. According to the theoretical calculation formula, the experimentally measured tensile modulus is 3.8GPa and the tensile strength is 68MPa, which is close to the mechanical property combination of cyanate ester recommended by the neural network, which is 4202.93MPa modulus and 66.46MPa strength.

[0064] For the cyanate ester tensile modulus and tensile strength combination 2, the recommended cyanate ester 2 composition and process parameters are shown in Table 2: Table 2

[0065] Example 2 of the present invention uses the recommended cyanate ester for verification, and conducts a tensile test according to GB / T2567-2021 "Test Method for Properties of Resin Castings". A universal testing machine is used at room temperature. The force and displacement curve during the experiment is measured by a sensor, and then the stress-strain curve is obtained, and the tensile strength is obtained by curve analysis.

[0066] The total length of the tensile specimen of the test material of Example 2 of the present invention is 200 mm, the length of the working section is 60 mm, the width is 10 mm, the thickness is 4 mm, and the tensile rate is 2 mm / min.

[0067] The stress-strain curve of the cyanate ester in the tensile state is drawn based on the force and displacement curve measured by the sensor. According to the theoretical calculation formula, the experimentally measured tensile modulus is 3.6 GPa and the tensile strength is 70 MPa, which is close to the modulus 4125.73 MPa and the strength 73.26 MPa of the cyanate ester mechanical property combination 2 recommended by the neural network.

[0068] Step 4, impregnating the cyanate synthesized in step 3 with quartz fiber to obtain a composite material roll; Specifically, tensile tests were performed on the synthesized cyanate ester / quartz fiber composites to verify the model prediction effect.

[0069] Example 3 of the present invention uses a composite material impregnated with resin and fiber for verification. The tensile test is carried out in accordance with ASTM D3039 "Standard Test Method for Tensile Properties of Polymer-Based Composite Materials" and is carried out at room temperature using a universal testing machine. The force and displacement curve during the experiment is measured by a sensor, and then the stress-strain curve is obtained to guide the calculation of the tensile strength.

[0070] The total length of the tensile test specimen of the test material of Example 3 of the present invention is 250 mm, the length of the working section is 138 mm, the width is 25 mm, the thickness is 2.25 mm, and the tensile rate is 2 mm / min.

[0071] The stress-strain curve of the cyanate ester / quartz fiber composite material in the tensile state was drawn based on the force and displacement curve measured by the sensor. According to the theoretical calculation formula, the experimentally measured tensile strength is 607MPa, which is greater than 600MPa and meets the optimization requirements of the tensile strength of the composite material.

[0072] Example 4 of the present invention uses a composite material impregnated with resin and fiber for verification. The tensile test is carried out in accordance with ASTM D3039 "Standard Test Method for Tensile Properties of Polymer-Based Composite Materials" and is carried out at room temperature using a universal testing machine. The force and displacement curve during the experiment is measured by a sensor, and then the stress-strain curve is obtained to guide the calculation of the tensile strength.

[0073] The total length of the tensile test specimen of the test material of Example 4 of the present invention is 250 mm, the length of the working section is 138 mm, the width is 25 mm, the thickness is 2.25 mm, and the tensile rate is 2 mm / min.

[0074] The stress-strain curve of the cyanate ester / quartz fiber composite material in the tensile state was drawn based on the force and displacement curve measured by the sensor. According to the theoretical calculation formula, the experimentally measured tensile strength is 615MPa, which meets the optimization requirements of the composite tensile strength.

[0075] From the above examples, it can be seen that the present invention provides a method for optimizing the mechanical properties of cyanate ester / quartz fiber composite materials based on machine learning. By combining finite element and experimental data, the machine learning method can quickly optimize the ingredients and process parameters of cyanate ester corresponding to the required tensile strength of the composite material. It can effectively solve the problem of time-consuming and labor-intensive traditional "trial and error method" for testing the performance of composite materials, provide greater guidance and help for process production, and provide targeted guidance to help quickly find the optimal ingredients in the process of material design and optimization, greatly reducing the R&D and experimental costs required for the development of new materials.

[0076] The present invention also provides a cyanate ester / quartz fiber mechanical property optimization system based on machine learning, which is used to implement the optimization method provided in any of the above embodiments, and the system includes: A first screening unit is used to recommend a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material by using a trained neural network prediction model; the input of the neural network prediction model is the tensile modulus and tensile strength of the cyanate ester, and the output is the tensile strength of the cyanate ester / quartz fiber composite material; A second screening unit is used to recommend the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester by using the trained machine learning model; the input of the machine learning model is the composition and process parameters of the cyanate ester, and the output is the tensile modulus and tensile strength of the cyanate ester; The third screening unit is used to screen the recommended cyanate ester composition and process parameters to obtain the cyanate ester composition and process parameters corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

[0077] The embodiments of the present invention also provide corresponding electronic devices and computer-readable storage media for implementing the solutions provided by the embodiments of the present invention.

[0078] Among them, the device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes, so that the device executes the machine learning-based cyanate / quartz fiber mechanical property optimization method described in any embodiment of the present application.

[0079] The storage medium stores a computer program, wherein the computer program, when executed by the processor, implements the method for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning as described in any embodiment of the present application.

[0080] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning, characterized in that: The process includes the following: Using the trained neural network prediction model, a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material are recommended; the input of the neural network prediction model is the tensile modulus and tensile strength of the cyanate ester, and the output is the tensile strength of the cyanate ester / quartz fiber composite material; Using the trained machine learning model, the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester are recommended; the input of the machine learning model is the composition and process parameters of the cyanate ester, and the output is the tensile modulus and tensile strength of the cyanate ester; The recommended cyanate ester components and process parameters are screened to obtain the cyanate ester components and process parameters corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

2. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The process of using the trained neural network prediction model to recommend the target tensile modulus and tensile strength of the cyanate ester corresponding to the target tensile strength of the cyanate ester / quartz fiber composite material includes: Using the gradient descent algorithm, the tensile strength target of the cyanate ester / quartz fiber composite material, the initial value of the cyanate ester tensile modulus and the initial value of the tensile strength are first set, and then the mean square error is used as the loss function, and the gradient of the loss function and the input of the neural network prediction model is calculated. The cyanate ester tensile modulus and tensile strength are continuously iterated and optimized. When the loss function is stable, the neural network model corresponding to the tensile strength of the cyanate ester / quartz fiber composite material output by the neural network prediction model is input into the tensile modulus and tensile strength of the cyanate ester as the tensile modulus target and tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

3. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The process of using the trained machine learning model to recommend the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester includes: By calculating the tensile modulus and tensile strength of the cyanate ester corresponding to the same component and process combination output by the machine learning model, and the Euclidean distance from the tensile modulus and tensile strength of the target cyanate ester, several groups of cyanate ester component and process combinations corresponding to the smaller Euclidean distance are selected as the components and process parameters of the cyanate ester corresponding to the tensile modulus target and tensile strength target of the cyanate ester; wherein the components and process parameters of the cyanate ester corresponding to the tensile modulus target and tensile strength target of the cyanate ester are the components and process parameters of the cyanate ester that can make the tensile strength of the cyanate ester / quartz fiber composite material not less than the tensile strength target of the cyanate ester / quartz fiber composite material.

4. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The training process of the neural network prediction model includes: The neural network prediction model is pre-trained using the original virtual data set to obtain a pre-trained neural network prediction model; wherein the construction process of the original virtual data set includes: using the Standard module in the Abaqus software to perform finite element calculations to obtain the tensile strength of the cyanate ester / quartz fiber composite material corresponding to different cyanate ester tensile moduli and tensile strengths, and the tensile strength of the cyanate ester / quartz fiber composite material corresponding to the different cyanate ester tensile moduli and tensile strengths constitutes the original virtual data set; The pre-trained neural network prediction model is fine-tuned using the original real data set to obtain the trained neural network prediction model; wherein the construction process of the original real data set includes: conducting a tensile mechanical property test of cyanate ester / quartz fiber composite materials under different cyanate ester ingredients and curing process parameters to obtain the tensile strength of the cyanate ester / quartz fiber composite materials corresponding to different cyanate ester tensile modulus and tensile strength, and the tensile strength of the cyanate ester / quartz fiber composite materials corresponding to the different cyanate ester tensile modulus and tensile strength constitute the original real data set.

5. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The training process of the machine learning model includes: The tensile modulus and tensile strength test data corresponding to different process combinations of cyanate ester components are constructed into a data set for training a machine learning model; the data set for training a machine learning model includes: phenolphthalein polyaryletherketone content, polysulfone content, polyimide content, acetylacetonate cobalt content, organic tin content, nonylphenol content, curing temperature, curing time, curing pressure, tensile modulus and tensile strength; The machine learning model is trained using the data set for training the machine learning model to obtain a trained machine learning model; during training, the input parameters of the machine learning model are first normalized, and then the input is processed by the machine learning model to obtain the tensile modulus and tensile strength of the cyanate ester.

6. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The machine learning model adopts a linear regression model, a polynomial regression model, a support vector machine regression model, a random forest regression model, an extreme gradient boosting regression model or a Gaussian process regression model; The neural network prediction model adopts an artificial neural network regression model.

7. The method for optimizing mechanical properties of cyanate ester / quartz fiber based on machine learning according to claim 1, characterized in that: The tensile modulus and tensile strength of cyanate ester were normalized, and the normalized results were processed using a neural network prediction model to obtain the tensile strength of cyanate ester / quartz fiber composite materials.

8. A cyanate ester / quartz fiber mechanical properties optimization system based on machine learning, characterized in that: include: A first screening unit is used to recommend a tensile modulus target and a tensile strength target of the cyanate ester corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material by using a trained neural network prediction model; the input of the neural network prediction model is the tensile modulus and tensile strength of the cyanate ester, and the output is the tensile strength of the cyanate ester / quartz fiber composite material; A second screening unit is used to recommend the composition and process parameters of the cyanate ester corresponding to the tensile modulus target and the tensile strength target of the cyanate ester by using the trained machine learning model; the input of the machine learning model is the composition and process parameters of the cyanate ester, and the output is the tensile modulus and tensile strength of the cyanate ester; The third screening unit is used to screen the recommended cyanate ester composition and process parameters to obtain the cyanate ester composition and process parameters corresponding to the tensile strength target of the cyanate ester / quartz fiber composite material.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the machine learning-based cyanate / quartz fiber mechanical property optimization method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the method for optimizing the mechanical properties of cyanate ester / quartz fiber based on machine learning as described in any one of claims 1 to 7 is implemented.

Images