A method and device for predicting polymer material properties

Abstract

A polymer material performance prediction method and device, belonging to the technical field of polymer material performance prediction, the method polymerizes and connects several monomer molecules in sequence to form a molecular chain; performs relaxation processing through a GAFF force field to obtain a polymer structure model; obtains set physical and chemical properties through computational simulation; performs fusion and cleaning processing on the set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints, and selects characteristic data related to target performance; constructs several preliminary performance prediction models according to a regression model; performs parameter optimization through a grid search method to select a performance prediction model; obtains a trained performance prediction model through training; inputs the set characteristic data into the trained performance prediction model to obtain the target performance corresponding to the set characteristic data. The present invention can realize the combination of computational simulation and artificial intelligence methods, improve the efficiency of performance prediction, reduce the amount of experimental data required, and improve the accuracy of polymer material performance prediction.

Description

Technical Field

[0001] The present invention relates to the technical field of polymer material performance prediction, and in particular to a polymer material performance prediction method and device. Background Art

[0002] Polymer materials are materials composed of high molecular weight compounds, which have the characteristics of large molecular weight and complex structure. They are usually connected by a large number of repeating units to form a chain or network structure. Their wide applications include packaging materials, automotive parts, aerospace components, medical equipment, electronic housings, and daily consumer goods such as clothing and kitchen utensils. Accurately predicting the properties of polymer materials is crucial to the development and application of polymer materials. By predicting properties such as mechanical properties, flame retardancy, and chemical resistance, researchers can design polymer materials with specific functions according to their intended use. Predicting polymer materials can improve the efficiency of material development, reduce the number of experimental trial and error, and reduce R&D costs and resource consumption.

[0003] At present, the performance of polymer materials is mainly predicted by two methods: experiments and molecular dynamics (MD) simulation. Among them, the experimental method has the problem of high cost. Traditional experimental methods require a lot of time and resources, including raw material synthesis, sample preparation and characterization testing. At the same time, there is also a problem of limited scope. Due to cost and time constraints, the experimental method can only explore a relatively limited number of material types. By combining different monomers and additives, a rich variety of potential polymer materials can be obtained, while the experimental method is difficult to traverse the vast material space, limiting the discovery of better materials. The molecular dynamics (MD) simulation method also has the problem of high computational cost. Although MD simulation provides details and features at the molecular level, it has improved efficiency compared to the experimental method, but the computational cost consumed for high-throughput screening of a wide range of chemical space is also relatively high; the molecular dynamics (MD) simulation method also has the problem of scale limitation. Due to the limitations of time and space scales, MD simulation is mainly used at the nanometer and nanosecond scales. It is usually difficult to accurately simulate the behavior of materials at a larger scale, resulting in errors between the calculated value and the true value.

[0004] Therefore, how to invent a method for predicting the properties of polymer materials to improve prediction efficiency while reducing the amount of experimental data and lowering costs has become an urgent problem to be solved. Summary of the invention

[0005] To this end, the present invention provides a method and device for predicting the properties of polymer materials, which can realize the combination of computational simulation and artificial intelligence methods, improve the efficiency of performance prediction, reduce the amount of experimental data required, and improve the accuracy of polymer material performance prediction.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a method for predicting polymer material properties, comprising:

[0007] Polymerize and connect a number of monomer molecules in sequence to form a molecular chain; place the molecular chains in a set container, and perform relaxation processing through a GAFF force field to obtain a polymer structure model;

[0008] Performing computational simulation on the monomer molecules, the molecular chains and the polymer structure model to obtain set physical and chemical properties;

[0009] The set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints are fused and cleaned to select feature data related to target performance; the feature data are randomly divided into a training set and a test set;

[0010] Constructing several preliminary performance prediction models according to the regression model; optimizing the parameters of the preliminary performance prediction model by a grid search method, and selecting a performance prediction model; training the performance prediction model by the training set and the test set, and obtaining the trained performance prediction model;

[0011] The set feature data is input into the trained performance prediction model, and the target performance corresponding to the set feature data is obtained through prediction processing by the performance prediction model.

[0012] As a preferred embodiment of a method for predicting the properties of polymer materials, in the process of relaxing the molecular chain through the GAFF force field to obtain the polymer structure model, the expression of the GAFF force field is:

[0013] ;

[0014] In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

[0015] As a preferred scheme for a method for predicting the properties of polymer materials, the set physical and chemical properties obtained by computational simulation include: molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain forces, gyration radius, cohesive energy and solubility parameter.

[0016] As a preferred solution of a polymer material performance prediction method, in the process of selecting the characteristic data related to the target performance, the correlation between the characteristic data and the target performance is evaluated by the Pearson correlation coefficient; the calculation formula of the Pearson correlation coefficient is:

[0017] ;

[0018] In the formula, X i is the characteristic data; i is the target performance; X ave is the average value of the feature; Y ave is the target performance average.

[0019] As a preferred solution of a polymer material performance prediction method, in the process of optimizing the parameters of the preliminary performance prediction model by a grid search method and selecting a performance prediction model, several model parameters are verified by the grid search method, and the preliminary performance prediction model with the highest test set determination coefficient is selected as the performance prediction model; the calculation formula of the test set determination coefficient is:

[0020] ;

[0021] In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

[0022] The present invention also provides a polymer material performance prediction device, based on the above polymer material performance prediction method, comprising:

[0023] A polymer structure model acquisition module is used to sequentially polymerize and connect a number of monomer molecules to form a molecular chain; a number of the molecular chains are placed in a set container, and relaxation is performed through a GAFF force field to obtain a polymer structure model;

[0024] A physical and chemical property calculation simulation module is used to perform calculation simulation on the monomer molecule, the molecular chain and the polymer structure model to obtain the set physical and chemical properties;

[0025] A feature data selection module is used to fuse and clean the set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints, and select feature data related to target performance; and randomly divide the feature data into a training set and a test set;

[0026] The performance prediction model construction and training module is used to construct several preliminary performance prediction models based on the regression model; optimize the parameters of the preliminary performance prediction model by the grid search method to select the performance prediction model; train the performance prediction model by the training set and the test set to obtain the trained performance prediction model;

[0027] The performance prediction module is used to input the set feature data into the trained performance prediction model, and obtain the target performance corresponding to the set feature data through prediction processing by the performance prediction model.

[0028] As a preferred solution of a polymer material performance prediction device, in the polymer structure model acquisition module, in the process of relaxing the molecular chain through the GAFF force field to obtain the polymer structure model, the expression of the GAFF force field is:

[0029] ;

[0030] In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

[0031] As a preferred embodiment of a polymer material performance prediction device, in the physical and chemical property calculation simulation module, the set physical and chemical properties obtained by calculation and simulation include: molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain force, gyration radius, cohesive energy and solubility parameter.

[0032] As a preferred solution of a polymer material performance prediction device, in the feature data selection module, in the process of selecting the feature data related to the target performance, the correlation between the feature data and the target performance is evaluated by the Pearson correlation coefficient; the calculation formula of the Pearson correlation coefficient is:

[0033] ;

[0034] Where, X i is the characteristic data; i is the target performance; X ave is the characteristic mean value; Y ave is the target performance average.

[0035] As a preferred solution of a polymer material performance prediction device, in the performance prediction model construction and training module, in the process of optimizing the parameters of the preliminary performance prediction model by a grid search method and selecting a performance prediction model, several model parameters are verified by the grid search method, and the preliminary performance prediction model with the highest test set determination coefficient is selected as the performance prediction model; the calculation formula of the test set determination coefficient is:

[0036] ;

[0037] In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

[0038] The present invention has the following advantages: several monomer molecules are polymerized and connected in sequence to form a molecular chain; several of the molecular chains are placed in a set container, and relaxation treatment is performed through the GAFF force field to obtain a polymer structure model; the monomer molecules, the molecular chains and the polymer structure model are computationally simulated to obtain set physical and chemical properties; the set physical and chemical properties, molecular SMILES, molecular graphs and molecular fingerprints are fused and cleaned, and characteristic data related to target performance are selected; the characteristic data are randomly divided into a training set and a test set; several preliminary performance prediction models are constructed according to a regression model; the preliminary performance prediction model is parameter optimized through a grid search method to select a performance prediction model; the performance prediction model is trained through the training set and the test set to obtain the trained performance prediction model; the set characteristic data is input into the trained performance prediction model, and the target performance corresponding to the set characteristic data is obtained through prediction processing by the performance prediction model. The present invention realizes the combination of computational simulation and artificial intelligence methods, improves the efficiency of performance prediction, and reduces the amount of experimental data required. For example, molecular dynamics (MD) and quantum mechanics (QM) calculations provide a wealth of physics-based information on molecular interactions and properties. By building a high-throughput computational process to generate a large amount of data for artificial intelligence, the reliance on experimental data can be reduced, making the performance prediction process simple and fast. The present invention also combines the information of atoms, molecular chains, and aggregate states obtained by computational simulation with the formulation components and process parameters, and significantly improves the accuracy of polymer material performance prediction by fusing information of different scales as the input of the artificial intelligence model. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0040] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0041] Figure 1 A schematic flow chart of a method for predicting polymer material properties provided in Example 1 of the present invention;

[0042] Figure 2 A schematic diagram of a monomer molecule repeating unit in a possible embodiment provided in Example 1 of the present invention;

[0043] Figure 3 A schematic diagram of a polymer all-atom structure model in a possible embodiment provided in Embodiment 1 of the present invention;

[0044] Figure 4 A schematic diagram of a model prediction result in a possible embodiment provided in Embodiment 1 of the present invention;

[0045] Figure 5 This is a schematic diagram of the architecture of a polymer material performance prediction device provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0046] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] Example 1

[0048] See also Figure 1 Embodiment 1 of the present invention provides a method for predicting polymer material properties, comprising the following steps:

[0049] S1, polymerizing and connecting a number of monomer molecules in sequence to form a molecular chain; placing the molecular chains in a set container, performing relaxation treatment through a GAFF force field, and obtaining a polymer structure model;

[0050] S2, performing computational simulation on the monomer molecules, the molecular chains and the polymer structure model to obtain set physical and chemical properties;

[0051] S3, fusing and cleaning the set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints, selecting feature data related to target performance; and randomly dividing the feature data into a training set and a test set;

[0052] S4. Construct several preliminary performance prediction models according to the regression model; optimize the parameters of the preliminary performance prediction model by grid search method and select the performance prediction model; train the performance prediction model by the training set and the test set to obtain the trained performance prediction model;

[0053] S5. Input the set feature data into the trained performance prediction model, and obtain the target performance corresponding to the set feature data through prediction processing by the performance prediction model.

[0054] In this embodiment, in step S1, a plurality of monomer molecules are polymerized and connected in sequence to form a molecular chain; the plurality of molecular chains are placed in a set container and subjected to relaxation treatment by a GAFF force field to obtain a polymer structure model;

[0055] Specifically, SMILES (Simplified Molecular Linear Input Specification, which uses a string of characters to represent the molecular structure of a compound, which encodes the atoms, bonds, and connectivity information within the molecule) is used to describe the chemical information of the monomer, such as tetrahydrofuran, and to specify the number and position of the connection points when the monomer is polymerized (such as Figure 2 As shown, the molecular connection sites are marked with C1-C4). Multiple monomer molecules are polymerized and connected in sequence to form a molecular chain, and then multiple chains are randomly stacked and placed in a cubic box with a length of 500 angstroms. After sufficient relaxation using the GAFF force field, a stable polymer structure model is obtained.

[0056] The expression of the GAFF force field is:

[0057] ;

[0058] In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

[0059] In this embodiment, in step S2, computational simulation is performed on the monomer molecule, the molecular chain and the polymer structure model to obtain set physical and chemical properties;

[0060] Specifically, various physical and chemical properties are counted and calculated based on monomer molecules, molecular chains and polymer structure models, including but not limited to molecular weight, charge number, molecular surface area, lowest / highest molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain forces, gyration radius, cohesive energy and solubility parameters. For example, molecular weight, charge number, molecular surface area, lowest / highest molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds and molecular chain forces can be automatically calculated based on the SMILES corresponding to the monomer molecules by using python to write scripts and import them into the rdkit library. For the gyration radius, molecular dynamics simulation is performed on the polymer all-atom structure model, using lammps software and GAFF force field, setting simulation temperature, pressure, step size and time, and then analyzing the dynamic simulation trajectory, and the average value of the distance from each part of each molecular chain to the center of the molecular chain is the gyration radius. The cohesive energy is calculated by the group contribution method. The literature is consulted to obtain the contribution of each group to the cohesive energy and the sum is used to obtain the cohesive energy of the molecule. Similarly, the sampling group contribution method can be used to calculate the molecular molar volume. The ratio of cohesive energy to molar volume is the cohesive energy density, and the solubility parameter is obtained by taking the square root of the cohesive energy density.

[0061] In this embodiment, in step S3, the set physical and chemical properties, molecular SMILES, molecular graph, and molecular fingerprint are fused and cleaned, and feature data related to the target performance are selected; the feature data are randomly divided into a training set and a test set;

[0062] Specifically, the data obtained by the above computational simulation and process parameters (such as temperature, pressure, time, etc.) as well as the encoded molecular SMILES, molecular graphs, molecular fingerprints, etc. are fused, cleaned and selected to screen out the features related to the target performance as the input of the model. The Pearson correlation coefficient is used to evaluate the correlation between the feature (X) and the target performance (Y). When the correlation coefficient is greater than 0.6, the feature is strongly correlated with the target property. The calculation formula of the Pearson correlation coefficient is:

[0063] ;

[0064] In the formula, X i is the characteristic data; i is the target performance; X ave is the characteristic mean value; Y ave is the target performance average.

[0065] The molecular fingerprint is calculated using rdkit based on molecular SMILES, and the present invention includes but is not limited to Morgan fingerprint, MACCS fingerprint, AtomPair fingerprint and TopologicalTorsion fingerprint.

[0066] The feature data is randomly divided into a training set and a test set in a ratio of 7:3 for training the performance prediction model.

[0067] In this embodiment, in step S4, several preliminary performance prediction models are constructed according to the regression model; the parameters of the preliminary performance prediction model are optimized by the grid search method to select the performance prediction model; the performance prediction model is trained by the training set and the test set to obtain the trained performance prediction model;

[0068] Specifically, regression models are selected including but not limited to linear regression, support vector machine regression, gradient descent regression, decision tree regression, random forest regression, neural network regression, etc. to build performance prediction models, and the model parameters are adjusted and optimized.

[0069] During the model training process, the grid search method was used to try different model parameters and select the test set determination coefficient R 2 The highest model is used as the performance prediction model. The calculation formula of the test set determination coefficient is:

[0070] ;

[0071] In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

[0072] In this embodiment, in step S5, the set feature data is input into the trained performance prediction model, and the target performance corresponding to the set feature data is obtained through prediction processing by the performance prediction model.

[0073] Specifically, after the model training is completed, the corresponding target performance can be output by inputting the corresponding features. Taking the Young's modulus of polyurethane as an example, after feature engineering, each polyurethane material sample can be represented by a vector of length 10. This vector is a feature with a high correlation with the Young's modulus that is retained after screening. Then, the vector is input into the trained model to obtain the corresponding Young's modulus.

[0074] In a possible embodiment, an example of prediction of polyurethane mechanical properties is provided as follows:

[0075] T1. Polymer structure model acquisition: Polyurethane elastomer is composed of three raw materials: polyol, isocyanate and chain extender. According to the chemical reaction rules, the corresponding reaction sites of the molecules are specified (such as Figure 2 C1-C4), the reaction sites are sequentially connected to form molecular chains and aggregated structures, such as Figure 3 As shown;

[0076] T2. Calculation and simulation of physical and chemical properties: Use rdkit, gaussian, lammps and other software to calculate molecules, molecular chains and all-atom structures to obtain information such as molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain forces, gyration radius, cohesive energy, solubility parameters, etc.

[0077] T3. Feature data selection: The component information (monomer types, this data set includes 5 polyols, 6 isocyanates and 2 chain extenders and their dosages), synthesis information (prepolymerization temperature and time, chain extension temperature and time) and computational simulation data were screened for correlation analysis, and the features with a high correlation with Young's modulus were retained;

[0078] T4. Train multiple models using the selected feature data, including but not limited to linear regression, support vector machine regression, gradient descent regression, decision tree regression, random forest regression, neural network regression, etc. By adjusting the model parameters, the test set is modeled and the model parameters are adjusted and optimized. The prediction results of the best model are as follows: Figure 4 As shown;

[0079] T5. Performance prediction. After determining the raw materials and temperature for polyurethane synthesis, the characteristic data obtained by calculation are combined with the characteristics with a high correlation with the breaking strength as the model input to obtain the predicted breaking strength value.

[0080] In summary, the present invention polymerizes and connects several monomer molecules in sequence to form a molecular chain; places several of the molecular chains in a set container, performs relaxation processing through the GAFF force field, and obtains a polymer structure model; performs computational simulation on the monomer molecules, the molecular chains, and the polymer structure model to obtain set physical and chemical properties; performs fusion and cleaning processing on the set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints, and selects characteristic data related to target performance; randomly divides the characteristic data into training sets and test sets; constructs several preliminary performance prediction models according to the regression model; optimizes the parameters of the preliminary performance prediction model through the grid search method, and selects a performance prediction model; trains the performance prediction model through the training set and the test set to obtain the trained performance prediction model; inputs the set characteristic data into the trained performance prediction model, and obtains the target performance corresponding to the set characteristic data through the prediction processing of the performance prediction model. The present invention realizes the combination of computational simulation and artificial intelligence methods, improves the efficiency of performance prediction, and reduces the amount of experimental data required. For example, molecular dynamics (MD) and quantum mechanics (QM) calculations provide a wealth of physics-based information on molecular interactions and properties. By building a high-throughput computational process to generate a large amount of data for artificial intelligence, the reliance on experimental data can be reduced, making the performance prediction process simple and fast. The present invention also combines the information of atoms, molecular chains, and aggregate states obtained by computational simulation with the formulation components and process parameters, and significantly improves the accuracy of polymer material performance prediction by fusing information of different scales as the input of the artificial intelligence model.

[0081] It should be noted that the method of the embodiment of the present disclosure can be performed by a single device, such as a computer or a server. The method of the present embodiment can also be applied in a distributed scenario and completed by multiple devices cooperating with each other. In the case of such a distributed scenario, one of the multiple devices can only perform one or more steps in the method of the embodiment of the present disclosure, and the multiple devices will interact with each other to complete the described method.

[0082] It should be noted that the above describes some embodiments of the present disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the above embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0083] Example 2

[0084] See also Figure 5Embodiment 2 of the present invention further provides a polymer material performance prediction device, comprising:

[0085] The polymer structure model acquisition module 001 is used to sequentially polymerize and connect a number of monomer molecules to form a molecular chain; the molecular chains are placed in a set container, and relaxation is performed through a GAFF force field to obtain a polymer structure model;

[0086] The physical and chemical property calculation simulation module 002 is used to perform calculation simulation on the monomer molecule, the molecular chain and the polymer structure model to obtain the set physical and chemical properties;

[0087] The feature data selection module 003 is used to perform fusion and cleaning processing on the set physical and chemical properties, molecular SMILES, molecular graph, and molecular fingerprint, and select feature data related to the target performance; and randomly divide the feature data into a training set and a test set;

[0088] The performance prediction model construction and training module 004 is used to construct several preliminary performance prediction models according to the regression model; optimize the parameters of the preliminary performance prediction model by the grid search method to select the performance prediction model; train the performance prediction model by the training set and the test set to obtain the trained performance prediction model;

[0089] The performance prediction module 005 is used to input the set feature data into the trained performance prediction model, and obtain the target performance corresponding to the set feature data through prediction processing by the performance prediction model.

[0090] In this embodiment, in the polymer structure model acquisition module 001, in the process of relaxing the molecular chain through the GAFF force field to obtain the polymer structure model, the expression of the GAFF force field is:

[0091] ;

[0092] In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

[0093] In this embodiment, in the physicochemical property calculation simulation module 002, the set physicochemical properties obtained by calculation simulation include: molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, intermolecular chain force, gyration radius, cohesive energy and solubility parameter.

[0094] In this embodiment, in the feature data selection module 003, in the process of selecting the feature data related to the target performance, the correlation between the feature data and the target performance is evaluated by the Pearson correlation coefficient; the calculation formula of the Pearson correlation coefficient is:

[0095] ;

[0096] In the formula, X i is the characteristic data; i is the target performance; X ave is the characteristic mean value; Y ave is the target performance average.

[0097] In this embodiment, in the performance prediction model construction and training module 004, in the process of optimizing the parameters of the preliminary performance prediction model by the grid search method and selecting the performance prediction model, several model parameters are verified by the grid search method, and the preliminary performance prediction model with the highest test set determination coefficient is selected as the performance prediction model; the calculation formula of the test set determination coefficient is:

[0098] ;

[0099] In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

[0100] It should be noted that the information interaction, execution process and other contents between the modules of the above-mentioned system are based on the same concept as the method embodiment in Example 1 of the present application, and the technical effects they bring are the same as those of the method embodiment of the present application. For specific contents, please refer to the description in the method embodiment shown above in the present application, and will not be repeated here.

[0101] Example 3

[0102] Embodiment 3 of the present invention provides a non-transitory computer-readable storage medium, in which a program code for a polymer material property prediction method is stored. The program code includes instructions for executing embodiment 1 or any possible implementation method of a polymer material property prediction method.

[0103] Computer-readable storage media can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0104] Example 4

[0105] Embodiment 4 of the present invention provides an electronic device, including: a memory and a processor;

[0106] The processor and the memory communicate with each other via a bus; the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute a polymer material performance prediction method of embodiment 1 or any possible implementation thereof.

[0107] Specifically, the processor can be implemented by hardware or by software. When implemented by hardware, the processor can be a logic circuit, an integrated circuit, etc.; when implemented by software, the processor can be a general-purpose processor implemented by reading software codes stored in a memory. The memory can be integrated in the processor or can be located outside the processor and exist independently.

[0108] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable systems. The computer instructions can be stored in a computer-readable storage medium, or transmitted from a computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center.

[0109] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing system, they can be concentrated on a single computing system, or distributed on a network composed of multiple computing systems, and optionally, they can be implemented by a program code executable by a computing system, so that they can be stored in a storage system and executed by the computing system, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0110] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. A method for predicting polymer material properties, characterized in that: include: Polymerize and connect a number of monomer molecules in sequence to form a molecular chain; place the molecular chains in a set container, and perform relaxation processing through a GAFF force field to obtain a polymer structure model; Performing computational simulation on the monomer molecules, the molecular chains and the polymer structure model to obtain set physical and chemical properties; The set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints are fused and cleaned to select feature data related to target performance; the feature data are randomly divided into a training set and a test set; Constructing several preliminary performance prediction models according to the regression model; optimizing the parameters of the preliminary performance prediction model by a grid search method, and selecting a performance prediction model; training the performance prediction model by the training set and the test set, and obtaining the trained performance prediction model; The set feature data is input into the trained performance prediction model, and the target performance corresponding to the set feature data is obtained through prediction processing by the performance prediction model.

2. A polymer material performance prediction method according to claim 1, characterized in that: In the process of relaxing the molecular chain through the GAFF force field to obtain the polymer structure model, the expression of the GAFF force field is: ； In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

3. A polymer material performance prediction method according to claim 2, characterized in that: The set physicochemical properties obtained by computational simulation include: molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain forces, gyration radius, cohesive energy and solubility parameter.

4. A polymer material performance prediction method according to claim 3, characterized in that: In the process of selecting the characteristic data related to the target performance, the correlation between the characteristic data and the target performance is evaluated by the Pearson correlation coefficient; the calculation formula of the Pearson correlation coefficient is: ； Where, X i is the characteristic data; i is the target performance; X ave is the average value of the feature; Y ave is the target performance average.

5. A polymer material performance prediction method according to claim 4, characterized in that: In the process of optimizing the parameters of the preliminary performance prediction model and selecting the performance prediction model by the grid search method, several model parameters are verified by the grid search method, and the preliminary performance prediction model with the highest test set determination coefficient is selected as the performance prediction model; the calculation formula of the test set determination coefficient is: ； In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

6. A polymer material performance prediction device, using a polymer material performance prediction method according to any one of claims 1 to 5, characterized in that: include: A polymer structure model acquisition module is used to sequentially polymerize and connect a number of monomer molecules to form a molecular chain; a number of the molecular chains are placed in a set container, and relaxation is performed through a GAFF force field to obtain a polymer structure model; A physical and chemical property calculation simulation module is used to perform calculation simulation on the monomer molecule, the molecular chain and the polymer structure model to obtain the set physical and chemical properties; A feature data selection module is used to fuse and clean the set physical and chemical properties, molecular SMILES, molecular graphs, and molecular fingerprints, and select feature data related to target performance; and randomly divide the feature data into a training set and a test set; The performance prediction model construction and training module is used to construct several preliminary performance prediction models based on the regression model; optimize the parameters of the preliminary performance prediction model by the grid search method to select the performance prediction model; train the performance prediction model by the training set and the test set to obtain the trained performance prediction model; The performance prediction module is used to input the set feature data into the trained performance prediction model, and obtain the target performance corresponding to the set feature data through prediction processing by the performance prediction model.

7. A polymer material performance prediction device according to claim 6, characterized in that: In the polymer structure model acquisition module, in the process of relaxing the molecular chain through the GAFF force field to obtain the polymer structure model, the expression of the GAFF force field is: ； In the formula, k r , k θ and V n are force constants; r eq and θ eq All are equilibrium structural parameters; q i and q j is the partial atomic charge; ε is the Lennard-Jones well depth; A ij and B ij is the Lennard-Jones radius; λ is the phase angle of the torsion angle parameter; r is the bond length; θ is the bond angle; R ij is the distance between atoms i and j.

8. A polymer material performance prediction device according to claim 7, characterized in that: In the physicochemical property calculation simulation module, the set physicochemical properties obtained by calculation simulation include: molecular weight, charge number, molecular surface area, lowest / highest energy molecular orbital, band gap, bond dissociation energy, dipole moment, number of intramolecular hydrogen bonds, number of intermolecular hydrogen bonds, molecular chain force, gyration radius, cohesive energy and solubility parameter.

9. A polymer material performance prediction device according to claim 8, characterized in that: In the feature data selection module, in the process of selecting the feature data related to the target performance, the correlation between the feature data and the target performance is evaluated by the Pearson correlation coefficient; the calculation formula of the Pearson correlation coefficient is: ； Where, X i is the characteristic data; Y i is the target performance; X ave is the characteristic mean value; Y ave is the target performance average.

10. A polymer material performance prediction device according to claim 9, characterized in that: In the performance prediction model construction and training module, in the process of optimizing the parameters of the preliminary performance prediction model by the grid search method and selecting the performance prediction model, several model parameters are verified by the grid search method, and the preliminary performance prediction model with the highest test set determination coefficient is selected as the performance prediction model; the calculation formula of the test set determination coefficient is: ； In the formula, is the true value of sample i; is the predicted value of sample i; is the average of the true values.

Images