A method and system for encoding based on the topology of a polymer material

Abstract

The application relates to the technical field of material design, in particular to a coding method and system based on a high polymer material topological structure, which comprises the following steps: firstly, a high polymer topological structure is constructed based on group characteristics and group adjacency relations; then, high polymer structure information is acquired based on the high polymer topological structure, and the high polymer structure information is digitally coded; finally, the high polymer structure information is digitally decoded to obtain the adjacency relations and group attributes of the high polymer material, and the high polymer material is subjected to visual processing. The application defines a new language system based on the group scale in combination with the advantages of the SMILES coding compiling method, that is, a new coding mode, and applies the new coding mode to structure extraction of the high polymer material and topological graph conversion, thereby serving as the input parameter of the high polymer material, reducing the number of parameters, reducing the requirement for the data volume, and reflecting the periodicity and complex spatial structure of the high polymer material.

Description

Technical Field

[0001] This application relates to the field of materials design technology, and in particular to a coding method and system based on the topological structure of polymer materials. Background Technology

[0002] The traditional "trial and error" approach to designing polymer materials with targeted properties is an empirical method that relies on researchers' experience and expertise. This involves conducting repeated experiments to study the material's physical properties, structure, and chemical reactions, ultimately leading to material design. However, this method is complex, time-consuming, and labor-intensive, requiring significant investment of resources and time in experiments and testing. Therefore, researchers have demanded more efficient material design methods. Thanks to advancements in artificial intelligence, surrogate models based on graph neural networks (GNNs) can accurately predict the properties of compounds based on their topological structure, avoiding the trial-and-error process and greatly improving the efficiency of materials research and development.

[0003] The commonly used SMILES encoding transforms the molecular structure into a series of characters to represent its molecular structure, thereby describing the molecular topology. SMILES encoding describes atoms and their bonding patterns, and can also concisely and clearly describe ring structures, branched structures, etc., accurately and uniquely representing chemical structures. Therefore, SMILES encoding currently has wide applications in molecular topology representation and in GNN prediction of molecular properties. From a theoretical perspective, SMILES encoding can also describe the topology of some polymer materials, thus enabling the use of GNNs to design high-performance polymer materials.

[0004] However, in the field of polymer materials, the current approach of using the SMILES encoding language system to convert polymer materials into topological graphs and then using GNNs to predict their properties based on the topological graphs of polymers still faces the following difficulties: (1) Some chemical properties, such as chemical reactions and intermolecular interactions, involve complex physicochemical mechanisms. More features based on physicochemical expertise need to be added to the topological graphs of polymer materials to assist in model training. However, in polymer materials, these chemical features are often not simply described by atomic-scale SMILES encoding. (2) Polymer materials have a certain periodicity and their composition is extremely complex, often having a network structure. SMILES encoding can only describe the monomers of polymer materials in the form of strings, making it difficult to represent the network cross-linking structure and periodic structure in polymers. It is difficult to extract the complete structural information of complex polymers, and further, it is difficult to extract the topological graph from the structural formula of polymers. (3) Based on the topology graph transformed by SMILES encoding, GNN can be used to predict the performance of polymer materials based on the topology graph. However, in reverse design, the final product is a string describing the atomic-scale structure of polymer materials. The atomic-scale structure design has certain difficulties in the synthesis and characterization in experiments. Summary of the Invention

[0005] This application provides an encoding method and system based on the topological structure of polymer materials, which encodes each group and its bonding relationship in the polymer material, thereby converting data from a professional database in the field of target polymer materials into input parameters of a graph neural network. The converted input parameters can effectively complete the training of the neural network.

[0006] To address the aforementioned technical problems, in a first aspect, embodiments of this application provide an encoding method based on the topological structure of polymer materials, comprising the following steps: First, constructing a polymer topological structure based on group characteristics and group adjacency relationships; then, acquiring polymer structural information based on the polymer topological structure and digitally encoding the polymer structural information; finally, digitally decoding the polymer structural information to obtain the adjacency relationships and group properties of the polymer material, and visualizing the polymer material.

[0007] In some exemplary embodiments, a polymer topology is constructed based on group characteristics and group adjacency relationships, including: defining and describing group characteristics, and extracting specific groups in the polymer material; describing group adjacency relationships, and extracting the internal connections and structural features of the polymer material; using specific groups as nodes in the polymer structure topology diagram, and connecting each node based on the internal connections and structural features of the polymer material to form a polymer topology, thereby obtaining the spatial arrangement of each group within the molecule.

[0008] In some exemplary embodiments, defining and describing group characteristics includes: segmenting the structure of a polymer based on the structural formula of the polymer in the database and in conjunction with chemical knowledge, and representing the segmented groups with different codes.

[0009] In some exemplary embodiments, the adjacency relationship of groups is described, including: defining the adjacency relationship between polymer groups, and describing whether different groups in the polymer material are adjacent to each other, and the connection sites between adjacent groups.

[0010] In some exemplary embodiments, the polymer structure information is digitally encoded, including: digitally processing the adjacency relationships of functional groups using data conversion codes to obtain digital adjacency relationships; converting the digital adjacency relationships into adjacency matrices, node feature matrices, and edge feature matrices based on functional group features; and describing the topological graph structure of the polymer and the physicochemical information of each edge and each node in the polymer topological graph based on the adjacency matrix, node feature matrix, and edge feature matrix.

[0011] In some exemplary embodiments, the adjacency matrix represents the pairwise connection relationship between groups and is used to describe the topological graph structure of the polymer; the node feature matrix is ​​used to describe the feature information on the nodes, including the node weight and functionality; the edge feature matrix is ​​used to describe the features on the connecting edges of the topological graph structure of the polymer, including the chemical bond bond energy.

[0012] In some exemplary embodiments, the polymer structure information is digitally decoded to obtain the adjacency relationships and group properties of the polymer material, and the polymer material is visualized. This includes: by interpreting the encoding rules and algorithms, processing the adjacency matrix of the polymer using code, and converting it into the original adjacency relationships of polymer groups; extracting the group information of the polymer based on the feature matrix of polymer nodes and the feature matrix of edges, converting the encoded polymer material information into polymer structure information that can be manually identified before encoding, and obtaining the decoded information; and visualizing the structure of the polymer material based on the decoded information.

[0013] In some exemplary embodiments, after visualizing the polymer material, the encoding method further includes: obtaining a visualized structure based on the visualization results; and performing experimental synthesis and characterization based on the visualized structure to verify the results of the reverse design.

[0014] Secondly, embodiments of this application also provide an encoding system based on the topological structure of polymer materials, comprising: a topological structure construction module, an encoding module, and a decoding module connected in sequence; wherein, the topological structure construction module is used to construct a polymer topological structure based on group characteristics and group adjacency relationships; the encoding module is used to obtain polymer structure information based on the polymer topological structure and digitally encode the polymer structure information; the decoding module is used to digitally decode the polymer structure information to obtain the adjacency relationships and group properties of the polymer material, and to visualize the polymer material.

[0015] In some exemplary embodiments, the topology construction module includes a group feature unit, a group adjacency relationship unit, and a node connection unit. The group feature unit is used to define and describe group features and extract specific groups in the polymer material. The group adjacency relationship unit is used to define and describe group adjacency relationships and extract the internal connections and structural features of the polymer material. The node connection unit is used to use the specific groups as nodes in the polymer structure topology diagram and connect the nodes based on the internal connections and structural features of the polymer material to form a polymer topology structure, thereby obtaining the spatial arrangement of the various groups within the molecule.

[0016] The technical solution provided in this application has at least the following advantages:

[0017] This application provides a coding method and system based on the topological structure of polymer materials. The coding method includes the following steps: First, constructing a polymer topological structure based on group characteristics and group adjacency relationships; then, acquiring polymer structural information based on the polymer topological structure and digitally encoding the polymer structural information; finally, digitally decoding the polymer structural information to obtain the adjacency relationships and group properties of the polymer material, and visualizing the polymer material. This application provides a coding method and system based on the topological structure of polymer materials. Combining the advantages of the SMILES coding compilation method, this application defines a new language system, i.e., a new coding method, based on the group scale, and applies it to the structural extraction of polymer materials, performing topological graph transformation as input parameters for the polymer material. This reduces the number of parameters, alleviates the data volume requirements, and can reflect the periodicity and complex spatial structure of polymer materials. Attached Figure Description

[0018] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations do not constitute a limitation on the embodiments, and unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0019] Figure 1A flowchart illustrating a coding method based on the topological structure of polymer materials provided in this application embodiment;

[0020] Figure 2 A schematic diagram of a coding system based on the topology of polymer materials provided in this application embodiment;

[0021] Figure 3 Example diagram of functional group definition information in polymer structures provided in embodiments of this application;

[0022] Figure 4 Example diagram of polymer adjacency relationships and group information extraction provided in the embodiments of this application;

[0023] Figure 5 This is an example diagram illustrating the extraction of adjacency relationships and functional group information of copolymers provided in embodiments of this application.

[0024] Figure 6 Example diagram of modified polymer adjacency relationships and group information extraction provided in the embodiments of this application;

[0025] Figure 7 A flowchart illustrating the translation group information and adjacency relationships provided in embodiments of this application;

[0026] Figure 8 Example diagrams showing the conversion of polymer structure diagrams into topological diagrams provided in the embodiments of this application;

[0027] Figure 9 The training set R provided for the embodiments of this application 2 A schematic diagram;

[0028] Figure 10 The test set R provided for the embodiments of this application 2 A schematic diagram. Detailed Implementation

[0029] As the background technology shows, the current approach of using a language system based on SMILES encoding to convert polymer materials into topological graphs and then using GNNs to predict the properties of polymers based on these topological graphs still has many problems.

[0030] To address the aforementioned technical problems, this application provides an encoding method and system based on the topological structure of polymer materials. The encoding method includes the following steps: First, constructing a polymer topological structure based on group characteristics and group adjacency relationships; then, acquiring polymer structural information based on the polymer topological structure and digitally encoding the polymer structural information; finally, digitally decoding the polymer structural information to obtain the adjacency relationships and group properties of the polymer material, and visualizing the polymer material. This application provides an encoding method based on the topological structure of polymer materials, which can encode each group and its bonding relationships in the polymer material, thereby representing the polymer structural information in the form of a topological graph, thus providing input parameters for a GNN surrogate model. Relying on the input parameters transformed by this application, the GNN model can accurately predict the performance of polymer materials, thereby assisting in the design of polymer materials.

[0031] The embodiments of this application will now be described in detail with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details have been provided in the embodiments of this application to facilitate a better understanding of the application. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments.

[0032] See Figure 1 This application provides an encoding method based on the topological structure of polymer materials, comprising the following steps:

[0033] Step S1: Construct the polymer topology based on group characteristics and group adjacency relationships.

[0034] Step S2: Based on the polymer topology, obtain polymer structure information and digitally encode the polymer structure information.

[0035] Step S3: Digitally decode the polymer structure information to obtain the adjacency relationship and group properties of the polymer material, and then visualize the polymer material.

[0036] The encoding method based on the topology of polymer materials provided in this application can transform data from specialized databases in the field of target polymer materials into input parameters for a graph neural network (GNN). The transformed input parameters can effectively train the neural network. In the absence of specialized database support, users can collect relevant literature through various channels, such as major literature retrieval platforms (CNKI, Web of Science), academic journals, etc., and then extract relevant data from the literature. First, step S1 defines the polymer topology using language; then, step S2 digitizes the polymer structure information; and finally, step S3 digitizes the polymer structure information. Furthermore, the decoded information can be used to visualize the polymer structure, and experimental synthesis and characterization can be performed based on the visualized structure to verify the results of the reverse design.

[0037] In some embodiments, the construction of the polymer topology in step S1 based on group characteristics and group adjacency relationships includes the following steps:

[0038] Step S101: Define and describe the characteristics of functional groups and extract specific functional groups from the polymer material.

[0039] Step S102: Describe the adjacency relationship of functional groups and extract the internal connections and structural features of the polymer material.

[0040] Step S103: Using specific functional groups as nodes in the polymer structure topology diagram, and based on the internal connections and structural characteristics of the polymer material, connect each node to form a polymer topology structure, thereby obtaining the spatial arrangement of each functional group within the molecule.

[0041] In some embodiments, the definition and description of group characteristics in step S101 includes: dividing the structure of the polymer according to the structural formula of the polymer in the database and combining chemical knowledge, and representing the divided groups with different codes.

[0042] Specifically, in defining and describing the characteristics of functional groups, this application designs a language system based on the functional group scale for describing the topological structure of polymer materials. Specifically, this language system uses easily recognizable symbols to represent each functional group appearing in the polymer material. In practical applications, based on the structural formulas of polymers in the database and combined with chemical knowledge, the structure is segmented, and the segmented functional groups are represented by different symbols. Through this step, this application can extract and quantitatively analyze specific functional groups contained in polymers and use them as nodes in the polymer structure topology diagram.

[0043] In some embodiments, step S102 describes the adjacency relationship of groups, including: defining the adjacency relationship between polymer groups, and describing whether different groups in the polymer material are adjacent, and the connection sites between adjacent groups.

[0044] Specifically, in the description of group adjacency relationships, this application further defines the adjacency relationships between polymer groups, which can clearly reflect whether different groups in the polymer material are adjacent, and the connection sites between adjacent groups, thereby revealing the internal connections and structural features of the molecule. This application uses the groups extracted in step S101 as nodes in the polymer structure topology map. Through the information extracted from the adjacency relationships, the nodes of each group can be connected to form a complete topology map structure, that is, the spatial arrangement of each group within the molecule. Topology plays a crucial role in the stability, reactivity, and function of molecules. Understanding the molecular topology helps predict and explain its behavior in chemical reactions or biological activities. Furthermore, the relevant information of the polymer topology map structure can be used as input parameters for a GNN, thereby establishing a structure-property mapping relationship for polymer materials.

[0045] In some embodiments, step S2 involves digitally encoding the polymer structure information, including:

[0046] Step S201: Use data conversion code to digitize the adjacency relationship of groups to obtain digitized adjacency relationships.

[0047] Step S202: Based on group characteristics, the digitized adjacency relationships are transformed into an adjacency matrix, a node feature matrix, and an edge feature matrix.

[0048] Step S203: Based on the adjacency matrix, node feature matrix, and edge feature matrix, describe the topological graph structure of the polymer and the physicochemical information of each edge and each node in the polymer topological graph.

[0049] In some embodiments, in steps S202 and S203, the adjacency matrix represents the pairwise connection relationship between groups and is used to describe the topological graph structure of the polymer; the node feature matrix is ​​used to describe the feature information on the nodes, including the node weight and functionality; the edge feature matrix is ​​used to describe the features on the connecting edges of the topological graph structure of the polymer, including the chemical bond bond energy.

[0050] Step S2 primarily involves the digital encoding of polymer structure information. Data encoding is the process of converting information into a specific format or rule, representing and storing this information in a digital form, transforming abstract concepts into a computer-recognizable form, making them easier to process, transmit, and analyze. The advantage of digitization lies in its ability to achieve efficient data management and rapid information exchange. In this application, the specific method for implementing the digitization of polymer structure data is as follows:

[0051] The adjacency relationships extracted in step S1 are digitized using data conversion code. Further, based on the extracted polymer group characteristics, the digitized adjacency relationships are transformed into an adjacency matrix, a node feature matrix, and an edge feature matrix. The adjacency matrix represents the pairwise connections between groups. Based on the adjacency matrix, the topological graph structure of the polymer can be described. Features on the connecting edges (such as chemical bond energies) are described by the edge feature matrix; feature information on the nodes (such as node weights and functionality) is described by the feature matrix. Based on these three matrices, not only can the topological graph structure of the polymer be described, but also the physicochemical information of each edge and node in the polymer topological graph can be described. This information can then be used as input parameters to help the neural network better understand the characteristics of each node and edge, facilitating neural network learning and inference.

[0052] In some embodiments, step S3 involves digitally decoding the polymer structure information to obtain the adjacency relationships and group properties of the polymer material, and then visualizing the polymer material, including the following steps:

[0053] Step S301: By interpreting the encoding rules and algorithms, the adjacency matrix of the polymer is processed using the code and transformed into the original adjacency relationship of the polymer groups.

[0054] Step S302: Based on the feature matrix of polymer nodes and the feature matrix of edges, extract the group information of polymers, convert the encoded polymer material information into polymer structure information that can be manually identified before encoding, and obtain the decoded information.

[0055] Step S303: Visualize the structure of the polymer material based on the decoded information.

[0056] Step S3 primarily involves the digital decoding of polymer structure information. The polymer material, designed through reverse engineering using computer algorithms, is still described using adjacency matrices, node feature matrices, and edge feature matrices. To more vividly understand the designed polymer structure and properties, the designed polymer material needs to be visualized, i.e., decoded according to the above encoding method, to obtain the adjacency relationships and group properties of the polymer material, thereby facilitating the experimental synthesis of this polymer material. During the data decoding process, by interpreting the encoding rules and algorithms, the adjacency matrix of the polymer is processed using the code, transforming it into the original adjacency relationships of polymer groups; further, based on the feature matrices of polymer nodes and edges, the group information of the polymer is extracted, thus converting the encoded polymer material information into polymer structure information that can be manually identified before encoding.

[0057] In some embodiments, after visualizing the polymer material in step S3, the above encoding method further includes: obtaining the visualized structure based on the visualization results; and performing experimental synthesis and characterization based on the visualized structure to verify the results of the reverse design.

[0058] See Figure 2 This application also provides an encoding system based on the topology of polymer materials, including: a topology construction module 101, an encoding module 102, and a decoding module 103 connected in sequence; wherein, the topology construction module 101 is used to construct a polymer topology based on group characteristics and group adjacency relationships; the encoding module 102 is used to obtain polymer structure information based on the polymer topology and to digitally encode the polymer structure information; the decoding module 103 is used to digitally decode the polymer structure information to obtain the adjacency relationships and group properties of the polymer material and to visualize the polymer material.

[0059] In some embodiments, the topology construction module 101 includes a group feature unit 1011, a group adjacency relationship unit 1012, and a node connection unit 1013. The group feature unit 1011 is used to define and describe group features and extract specific groups in the polymer material. The group adjacency relationship unit 1012 is used to define and describe group adjacency relationships and extract the internal connections and structural features of the polymer material. The node connection unit 1013 is used to use the specific groups as nodes in the polymer structure topology diagram and connect each node based on the internal connections and structural features of the polymer material to form a polymer topology structure and obtain the spatial arrangement of each group in the molecule.

[0060] The coding method and system based on the topology of polymer materials provided in this application will be described in detail below through specific embodiments.

[0061] Step 1: In chemistry, when breaking down a complex polymer structure into individual functional groups, it is necessary to follow relevant chemical principles. Therefore, we break it down according to the following principles and number the resulting functional groups:

[0062] (1) This application takes into account the strength of the bonds and the convenience of connection, and divides the polymer structure at the covalent single bonds inside the polymer material. Covalent bonds are one of the most basic connection methods in polymer materials. Unlike the ionic bonds, hydrogen bonds and other interaction forces existing inside the polymer, the group separation of the polymer material at the covalent bonds does not significantly affect the independence of the internal structure. At the same time, this separation method also facilitates the conversion of the polymer structure topology.

[0063] (2) This application identifies and separately numbers groups with specific chemical properties and functions in polymers. Groups are important parts of polymers that determine their chemical properties and reaction behavior. Using them as building blocks to divide the polymer structure can preserve the internal chemical structure information of the polymer to the greatest extent while constructing a simplified topological structure of the polymer material. When defining and classifying groups, a group with unique functions should not be split into multiple parts, but should be defined as a whole and assigned a specific code.

[0064] (3) This application takes into account the interactions between multiple groups. The connection methods between different groups constitute polymer three-dimensional structures with different morphologies, including cage-like structures, comb-like structures, and star-like structures. The groups in these special structures have certain interactions and may exhibit reactivity completely different from that of linear polymers. Therefore, these special polymer structures with different morphologies must be reflected in the topology diagram. For this reason, when dividing the groups, it is necessary to ensure that the division method does not destroy the special three-dimensional structure of the polymer in order to correctly describe the structural information of the polymer.

[0065] In summary, this application defines as follows: Figure 3 The groups and their symbols shown describe the group information in the polymer structure.

[0066] Step 2: Based on the polymer group codes defined in Step 1, extract the adjacency and characteristic relationships of the polymer structure formula. The specific principles are as follows:

[0067] (1) For data in databases or literature, it is necessary to extract the characteristics of adjacency relationships, functional groups, and chemical bonds based on their molecular structure diagrams. The format for the designation of functional groups in adjacency relationships is as follows: number of times the same functional group appears + designation (e.g., ...). Figure 3(As shown) + Connection Position. In the definition of the connection position, one-hot encoding is used to reflect the site where two groups connect. The numbering rules for the connection positions are as follows: the site connecting to the polymer main chain is position 1, and the remaining positions are numbered clockwise. Additionally, if all sites of a group are chemically equivalent (-CH2-), then a connection position does not need to be defined. When extracting adjacency relationships, two adjacent groups are considered as a pair, and their group codes are written together, separated by a comma, with parentheses at both ends, such as (1A1, 1D). The polymer chain is traversed sequentially, and considering the periodicity of the polymer, the adjacency relationship needs to connect from the tail to the head of the polymer to obtain the overall adjacency relationship of the polymer. For the extraction of group features, the format is as follows: Total Group Weight (Weight of Individual Group; Functionality; Source Raw Material Code). If the same group appears multiple times in a polymer, it is represented by multiple parentheses, and the total weight of the group is the sum of the weights of each group within the parentheses. Figure 4 The example shown is a polymer adjacency relationship and group feature extraction.

[0068] (2) For AB binary copolymer polymers, this application considers four connection modes of the two repeating units when defining their adjacency relationships, namely AA, AB, BA, and BB, and reflects them in the adjacency relationships. The adjacency relationship definition method for multi-component copolymers is the same as that for binary copolymers. When extracting group information, the weight of each group is assigned based on its molar ratio and is included in the total group weight. Figure 5 The image shows an example of a typical polymer copolymer.

[0069] (3) For modified polymers with modifications on the polymer chain, this application considers both the modified and unmodified parts when defining their adjacency relationships, and writes out their adjacency relationships separately. When extracting group information, the weight of each group is also assigned based on its molar ratio. Figure 6 The image shows a typical example of a polymer modified on a polymer chain.

[0070] Step 3: Convert the collected adjacency relationships into a topological graph structure and input parameters for the GNN. This application uses code to process and extract the adjacency relationships and group features of polymer materials, and its specific implementation method is as follows: Figure 7As shown in the diagram, firstly, the self-developed code in this application is used to read the functional group information of the polymer material, and each functional group is individually numbered according to its code and frequency of occurrence. Simultaneously, its features are extracted to construct a node feature matrix. Secondly, the extracted adjacency relationships are transformed according to the aforementioned individual functional group numbers, completing the digitization process of the adjacency relationships. Next, the code is used to statistically analyze the adjacent functional groups of each functional group based on each translated pair of adjacency relationships, constructing a polymer structure topology graph, and inputting the relevant information into the adjacency matrix. Finally, the features corresponding to each edge in the polymer topology graph are extracted and correlated with the adjacency matrix to construct an edge feature matrix. In this way, we successfully established the relationship between the polymer structure and its topology graph. Figure 8 The example shown is a polymer structural formula and its transformed topological diagram.

[0071] Step 4: Convert the GNN input parameters into adjacency relations and functional group features. After designing polymers based on GNN, the final calculation result is an adjacency matrix based on UTF-8 encoding. However, it is difficult for humans to intuitively understand the polymer structure through the abstract adjacency matrix. Therefore, this application also completed the conversion of the GNN input parameters, such as the adjacency matrix, into the adjacency relations and functional group features in Step 2 based on the code, thereby better visualizing the polymer structure. Combined with the polymer functional group encoding table established in Step 1, this facilitates experimental synthesis and characterization by researchers.

[0072] This application presents a polymer information encoding method suitable for converting the structure of polymer materials into custom information, and the encoding results can provide input parameters for subsequent machine learning algorithms. Its key points are:

[0073] (1) Describing the structural information of polymers at the scale of functional groups or molecular fragments, and in the field of chemistry, decomposing massive polymer structural formulas into independent fragments to represent their topological structures. Specifically, it has the following characteristics:

[0074] (1.1) Regarding the sites for dividing polymer fragments: the chemical structure of polymers is divided at the chemical bonds within the polymer material, thus dividing it into several groups or molecular fragments;

[0075] (1.2) Regarding polymer group nodes: Groups or molecular fragments within the polymer are treated as independent nodes, and the nodes are described using specific codes. The adjacency relationships of the polymer are extracted based on the above codes, and its topological structure is extracted based on computer language.

[0076] (1.3) In terms of complex polymer structures: It can represent and design complex polymer structures such as polymer copolymers, network structures and modified polymer chains to reflect the adjacency relationship of each group inside them, and perform statistical analysis of their characteristics in the overall polymer.

[0077] (2) This application can be used to realize the mutual conversion between polymer structures, topological diagrams, and GNN model input parameters, and then use machine learning to assist in the design and optimization of polymer materials. Specifically, it has the following features:

[0078] (2.1) Regarding polymer topology diagrams: At the group or molecular fragment scale, establish the transformation rules between the structure of polymers and their topology diagrams, effectively transform the adjacency relationships between molecular fragments or groups, and use topology diagrams to effectively represent polymer structures;

[0079] (2.2) Regarding the conversion of GNN model input parameters: The code is used to complete the accurate conversion from polymer adjacency relations and features to GNN model input parameters. The input parameters of the GNN model include the adjacency matrix, the feature matrix of the node, and the feature matrix of the edge.

[0080] (2.3) Regarding the translation of GNN model input parameters: The input parameters of the GNN model—adjacency matrix, node feature matrix and edge feature matrix—are translated using code, and converted into data such as the adjacency relationship and group features of polymers that are easy for humans to identify, thereby creating a polymer structure diagram.

[0081] Compared with existing technologies, the coding method and system based on the topology of polymer materials provided in this application have the following advantages: This application solves the problem that SMILES coding is difficult to represent the physical knowledge of polymers in the field of polymers. Many physicochemical information of polymers can be represented by features at the group level, thus this application can solve the problem of topological representation of polymer materials; furthermore, based on the groups and their adjacency relationships defined in this application, the structure of polymers can be described directly from the polymer material itself, without relying on monomers, and the periodicity and network structure of polymers can be fully represented; simultaneously, the adjacency relationships extracted in this application can be converted into GNN input parameters and polymer topology structures, and the conversion of GNN network input parameters into adjacency relationships can also be realized. This allows for reverse design to complete the design of novel polymer materials and transforms them into polymer structural formulas that can be visualized in experiments.

[0082] To verify the feasibility of the encoding method and system based on polymer material topology provided in this application, the molecular encoding method defined in this application can be based on data in a database and converted into input parameters usable by a GNN. Specifically, this includes the adjacency matrix of the polymer material, the feature matrices of nodes and edges, and the performance data of the polymer material. This application uses data from an ablation-resistant polymer material structure-performance database as an example, and utilizes a GNN model to determine the 5% thermal decomposition temperature (T) of the materials in the database. 5％ Based on the aforementioned polymer coding method, and utilizing 1490 polymer structure data entries from the database, as well as T... 5％Performance data samples were collected to obtain the corresponding adjacency matrix, node and edge feature matrices. A crystal graph convolutional neural network (CGCNN) regression model was then used to analyze the T-values ​​of the materials in the database. 5％ Performance is predicted.

[0083] The CGCNN model network architecture consists of three convolutional layers, one hidden layer, and one fully connected layer. Each convolutional layer has 64 convolutional kernels to extract and process feature information from the polymer topology map. The hidden layer has 128 hidden features, which are passed to subsequent neural network layers for further processing and learning of task-related features. The fully connected layer uses the Softplus activation function to map the output features to the final prediction result space.

[0084] When training and testing the model, T 5％ As training labels for the CGCNN model, the adjacency matrix and feature matrix were used as input parameters. 10% of the original data was used as the validation set for training the network, and 10% was used as the test set to validate the model and evaluate its prediction performance. The results are as follows: Figure 9 , Figure 10 As shown.

[0085] The training and testing results above show that, using the adjacency matrix transformed by this invention and the feature matrices of nodes and edges as input parameters, combined with the CGCNN model, the T-value of ablation-resistant polymer materials can be accurately predicted. 5％ Performance. Based on the above examples, the molecular coding method of this application can also perform regression prediction of the properties of other polymer materials (such as tensile strength, glass transition temperature, etc.) based on the structural information of polymers, thereby realizing accurate reverse design of polymer materials.

[0086] Based on the above technical solutions, embodiments of this application provide an encoding method and system based on the topological structure of polymer materials. The encoding method includes the following steps: First, constructing a polymer topological structure based on group characteristics and group adjacency relationships; then, acquiring polymer structural information based on the polymer topological structure and digitally encoding the polymer structural information; finally, digitally decoding the polymer structural information to obtain the adjacency relationships and group properties of the polymer material, and visualizing the polymer material. This application provides an encoding method and system based on the topological structure of polymer materials. Combining the advantages of the SMILES encoding compilation method, this application defines a new language system based on the group scale, i.e., a new encoding method, and applies it to the structural extraction of polymer materials, performing topological graph transformation as input parameters for the polymer material. This reduces the number of parameters, alleviates the data volume requirements, and can reflect the periodicity and complex spatial structure of polymer materials.

[0087] Those skilled in the art will understand that the above-described embodiments are specific examples of implementing this application, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of this application. Any person skilled in the art can make their own modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A coding method based on the topological structure of polymer materials, characterized in that, Includes the following steps: Based on group characteristics and group adjacency relationships, polymer topologies are constructed; Based on the polymer topology, polymer structure information is obtained and the polymer structure information is digitally encoded. The polymer structure information is digitally decoded to obtain the adjacency relationship and group properties of the polymer material, and the polymer material is then visualized. The polymer structure information is digitally encoded, including: The adjacency relationships of functional groups are digitized using data conversion codes to obtain digitized adjacency relationships; Based on group characteristics, the digitized adjacency relationships are transformed into adjacency matrices, node feature matrices, and edge feature matrices; Based on the adjacency matrix, node feature matrix, and edge feature matrix, the topological graph structure of polymers and the physicochemical information of each edge and each node in the polymer topological graph are described. The adjacency matrix represents the pairwise connection relationships between groups, and the adjacency matrix is ​​used to describe the topological structure of the polymer. The node feature matrix is ​​used to describe the feature information on the nodes, including the node's weight and functionality. The feature matrix of the edges is used to describe the features on the connecting edges of the topological graph structure of the polymer, and the features on the connecting edges include chemical bond bond energies.

2. The encoding method based on the topological structure of polymer materials according to claim 1, characterized in that, The construction of polymer topologies based on group characteristics and group adjacency relationships includes: Define and describe the characteristics of functional groups, and extract specific functional groups from polymer materials; The adjacency relationships of functional groups are described to extract the internal connections and structural features of polymer materials; Using the specific functional groups as nodes in the polymer structure topology diagram, and based on the internal connections and structural characteristics of the polymer material, the nodes are connected to form a polymer topology, thereby obtaining the spatial arrangement of the various functional groups within the molecule.

3. The encoding method based on the topological structure of polymer materials according to claim 2, characterized in that, The definition and description of the characteristics of the functional groups include: Based on the structural formulas of polymers in the database, and combined with chemical knowledge, the polymer structure is segmented, and the segmented groups are represented by different codes.

4. The encoding method based on the topological structure of polymer materials according to claim 2, characterized in that, The description of the adjacency relationship of the groups includes: Define the adjacency relationship between polymer groups and describe whether different groups in polymer materials are adjacent, as well as the connection sites between adjacent groups.

5. The encoding method based on the topological structure of polymer materials according to claim 1, characterized in that, The polymer structure information is digitally decoded to obtain the adjacency relationships and group properties of the polymer material, and the polymer material is then visualized, including: By interpreting the rules and algorithms of the encoding, the adjacency matrix of polymers is processed using the code and transformed into the original adjacency relationships of polymer groups. Based on the feature matrix of polymer nodes and the feature matrix of edges, the group information of polymer is extracted, and the encoded polymer material information is converted into polymer structure information that can be manually identified before encoding, thus obtaining the decoded information. The structure of polymer materials is visualized based on the decoded information.

6. The encoding method based on the topological structure of polymer materials according to claim 1, characterized in that, After visualizing the polymer material, the encoding method further includes: Based on the visualization results, obtain the visualization structure; Experimental synthesis and characterization were performed based on the visualized structure to verify the results of the reverse design.

7. A coding system based on the topological structure of polymer materials, the system being used to implement the coding method based on the topological structure of polymer materials as described in any one of claims 1 to 6, characterized in that, The system includes: a topology construction module, an encoding module, and a decoding module connected in sequence; wherein, The topology construction module is used to construct polymer topologies based on group characteristics and group adjacency relationships; The encoding module is used to obtain polymer structure information based on the polymer topology and to digitally encode the polymer structure information. The decoding module is used to digitally decode the polymer structure information to obtain the adjacency relationship and group properties of the polymer material, and to visualize the polymer material.

8. The coding system based on the topological structure of polymer materials according to claim 7, characterized in that, The topology construction module includes group feature units, group adjacency relationship units, and node connection units; The group feature unit is used to define and describe group features and extract specific groups from polymer materials; The group adjacency relationship unit is used to define and describe the group adjacency relationship and extract the internal connections and structural features of polymer materials; The node connection unit is used to connect the nodes in the polymer structure topology diagram with the specific group as the node, and based on the internal connection and structural characteristics of the polymer material, to form a polymer topology structure and obtain the spatial arrangement of the various groups in the molecule.

Images