A method and device for generating a polyimide molecular structure based on monomer combination

Abstract

The application relates to the technical field of computer-aided design of high polymer materials, and discloses a polyimide molecular structure generation method and device based on monomer combination, which comprises the following steps: constructing a synthesizable monomer database; constructing a molecular performance prediction model; generating a target polyimide structure based on Monte Carlo tree search; and outputting a candidate scheme. The device corresponds to the method. According to the application, the chemical search space is controlled within a reasonable and operable range by limiting the basic construction monomer; the actual synthesizability of the generated structure is guaranteed from the source by combining the commercially available or known synthesizable monomers; and the intelligent algorithm is used to systematically explore the monomer combination space, so that the discovery efficiency of novel high-performance polyimide materials is significantly improved.

Description

Technical Field

[0001] This application relates to the field of computer-aided design technology for polymer materials, specifically a method and apparatus for generating polyimide molecular structures based on monomer combinations. Background Technology

[0002] Polyimide (PI) is a class of high-performance polymers containing imide rings in its main chain. Due to its excellent thermal stability, high mechanical strength, good chemical stability, and superior electrical insulation properties, it occupies an important position in cutting-edge fields such as aerospace, microelectronics processing, and optoelectronic displays. In recent years, with the booming development of emerging industries such as flexible electronic devices and 5G high-frequency communications, more stringent and diverse requirements have been placed on the performance of polyimide materials. How to quickly and accurately design and screen novel polyimide molecular structures that meet specific performance indicators, such as specific glass transition temperatures, has become a key technical challenge that urgently needs to be solved in this field. Summary of the Invention

[0003] The purpose of this application is to provide a method and apparatus for generating polyimide molecular structures based on monomer combinations, so as to solve the technical problem in the prior art that it is difficult to quickly and accurately design and screen novel polyimide molecular structures that meet specific performance indicators.

[0004] To achieve the above objectives, this application provides a method for generating polyimide molecular structures based on monomer combinations, the method comprising:

[0005] S10: Construct a synthetic monomer database, wherein the synthetic monomer database stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers;

[0006] S20: Construct a molecular property prediction model, which is used to predict specific properties of polyimide molecules, including at least the glass transition temperature of polyimide molecules;

[0007] S30: Generate the target polyimide structure based on Monte Carlo tree search. This search generation process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as the target polyimide structure.

[0008] S40: Output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties.

[0009] As a preferred embodiment, S10 specifically includes:

[0010] Systematically collect known dianhydride and diamine monomer structural information from publicly available chemical substance databases or literature; clean and standardize the collected data to remove erroneous or incomplete records; convert the processed monomer molecular structures into a computer-recognizable standardized coding format to establish a synthetic monomer database.

[0011] As a preferred embodiment, S20 specifically includes:

[0012] The molecular performance prediction model is a machine learning model using a graph neural network architecture. The molecular performance prediction model takes the standardized code or its corresponding graph structure representation as input, performs feature extraction and nonlinear mapping, and outputs the corresponding performance prediction value. The performance prediction value includes at least the predicted glass transition temperature. The molecular performance prediction model is trained using polyimide with known structures and its performance data.

[0013] As a preferred option, S30 specifically includes:

[0014] Using dianhydride and diamine monomers from the synthetic monomer database as basic nodes in the search space, an iterative search is performed using the Monte Carlo tree search algorithm. During the search process, candidate polyimide structures are generated by simulating the polymerization reaction of dianhydride and diamine, and the generated candidate polyimide structures are evaluated using the molecular performance prediction model. The evaluation results are converted into a reward function to guide the search direction of the Monte Carlo tree search algorithm. The evaluation results include at least the predicted glass transition temperature.

[0015] Preferably, the construction of the molecular performance prediction model includes:

[0016] A training dataset is constructed based on the known molecular structure of polyimide and its corresponding experimental measurements, wherein the experimental measurements include at least the glass transition temperature.

[0017] A graph-based representation method is used to extract atomic information as node features and chemical bond information as edge features for the molecular graph structure of each polyimide molecule.

[0018] Construct a graph convolutional neural network model to predict specific properties of polyimide molecules;

[0019] The graph convolutional neural network model is trained using the training dataset under supervised learning conditions to optimize the model parameters and obtain the molecular performance prediction model.

[0020] Preferably, the hidden layer of the molecular performance prediction model is a multi-layer graph convolutional layer, which is used to aggregate neighborhood node information layer by layer to capture the local and global structural features of the molecule.

[0021] Preferably, the iterative search using the Monte Carlo tree search algorithm includes:

[0022] A search tree is constructed, with the root node representing the initial state, and the dianhydride monomer and diamine monomer constituting the action space to be selected in the search tree.

[0023] The Monte Carlo tree search algorithm explores the optimal combination through multiple iterative cycles. Each iteration includes: starting from the root node, selecting child nodes downwards in the tree according to a preset selection strategy until a leaf node with unexpanded child nodes is reached; if there are unexplored actions at the current leaf node, one or more new child nodes are created and added to the tree; based on the selected monomer pair composed of dianhydride monomer and diamine monomer, using a preset chemical reaction template, the molecular structure of the corresponding polyimide repeating unit is generated, and the generated molecular structure is input into the analytical performance prediction model, outputting the predicted glass transition temperature of the structure, and using the predicted glass transition temperature as the reward value for this simulation; the obtained reward value is propagated upwards along the path to update the statistical information of all nodes on the path from the evaluated node to the root node, in order to guide the node selection in subsequent iterations.

[0024] To achieve the above objectives, this application also provides an apparatus for generating polyimide molecular structures based on monomer combinations, applying the polyimide molecular structure generation method based on monomer combinations as described above. The apparatus includes:

[0025] A database construction module is used to construct a synthetic monomer database, which stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers.

[0026] A prediction model building module is used to build a molecular property prediction model, which is used to predict specific properties of polyimide molecules, including at least the glass transition temperature of polyimide molecules.

[0027] The target search module is used to generate target polyimide structures based on Monte Carlo tree search. The search generation process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as the target polyimide structures.

[0028] The target output module is used to output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties.

[0029] To achieve the above objectives, this application also provides a computer device for generating polyimide molecular structures based on monomer combinations, including at least one processor, at least one memory, and a data bus;

[0030] The processor and the memory communicate with each other via the data bus;

[0031] The memory stores program instructions that can be executed by the processor, which invokes the program instructions to execute the method for generating polyimide molecular structures based on monomer combinations as described above.

[0032] To achieve the above objectives, this application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the method for generating polyimide molecular structures based on monomer combinations as described above.

[0033] Beneficial effects: The method and apparatus for generating polyimide molecular structures based on monomer combination in this application control the chemical search space within a reasonable and operable range by limiting the basic building monomers; it ensures the practical synthesability of the generated structure from the source by combining commercially available or known synthetic monomers; and it significantly improves the discovery efficiency of novel high-performance polyimide materials by systematically exploring the monomer combination space using intelligent algorithms. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart illustrating the method for generating polyimide molecular structures based on monomer combinations, provided in an embodiment of this application.

[0036] Figure 2 A schematic diagram of the overall process for generating polyimide molecular structures based on monomer combinations, provided in an embodiment of this application;

[0037] Figure 3 This is a schematic diagram of the basic structure of the molecular performance prediction model provided in the embodiments of this application;

[0038] Figure 4 A flowchart illustrating the specific implementation of polyimide structure generation based on Monte Carlo tree search provided in this application embodiment;

[0039] Figure 5 The diagram shows the structure of the apparatus for generating polyimide molecular structures based on monomer combinations, as provided in the embodiments of this application. In the diagram: 10, database construction module; 20, prediction model construction module; 30, target search module; 40, target output module.

[0040] The implementation, functional features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0041] The technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0042] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0043] Analysis of existing design methods for novel polyimide structures reveals at least the following limitations:

[0044] I. Existing technologies typically employ traditional trial-and-error methods based on experience. Researchers select dianhydrides and diamine monomers for synthesis and testing based on past experience and literature reports. This method lacks systematic theoretical guidance, exhibits strong randomness, and suffers from long synthesis and characterization cycles and high costs for single samples, making it difficult to effectively explore a broad chemical space, resulting in a high failure rate and low efficiency in the development of new materials.

[0045] Second, existing technologies employ structure generation methods based on combinations of functional groups or fragments. These methods generate candidate molecules by exhaustively listing combinations of functional groups or structural fragments. However, this approach is highly susceptible to the "chemical space explosion" problem, leading to an exponential increase in computational and screening costs. More importantly, the molecular structures generated by this method often ignore the rules of chemical synthesis and stability limitations, resulting in a large number of candidate structures that cannot be synthesized in reality and lack practical application value.

[0046] Third, existing technologies attempt to utilize deep learning-based molecular generation models (such as variational autoencoders (VAEs) and generative adversarial networks (GANs). These methods typically require massive amounts of training data, and the models often sample molecules in a continuous latent space, making it difficult to guarantee that the generated topological structures conform to basic chemical valence bond theory, resulting in extremely poor syntheticability of the generated molecules. Furthermore, the "black box" nature of deep learning models makes the generation process lack interpretability, making it difficult to provide experimenters with specific synthetic path guidance.

[0047] In summary, existing technologies generally suffer from prominent problems such as uncontrollable chemical space search scale, low computational screening efficiency, and inability to guarantee the syntheticity of generated structures. Therefore, there is an urgent need to develop a technical solution that can effectively control the search space, reliably guarantee the syntheticity of candidate structures, and enable systematic and efficient design of new materials.

[0048] Reference Figure 1 , Figure 1 This is a flowchart illustrating a method for generating polyimide molecular structures based on monomer combinations, provided in an embodiment of this application.

[0049] To address the aforementioned technical limitations, such as Figure 1 As shown, this embodiment discloses a method for generating polyimide molecular structures based on monomer combinations. The method includes:

[0050] Reference Figure 2 , Figure 2 This is a schematic diagram of the overall process of the method for generating polyimide molecular structures based on monomer combinations, provided in the embodiments of this application.

[0051] The present embodiment will now briefly describe a method for generating polyimide molecular structures based on monomer combinations. For example... Figure 2 As shown, in the specific application of this embodiment, we performed the following processing:

[0052] First, data collection and preprocessing are performed, specifically by collecting molecular data from public databases and encoding them in SMILES format.

[0053] Secondly, a property prediction model is constructed, specifically using GCN (Graph Convolutional Network) to construct a molecular glass transition temperature prediction model.

[0054] Next, the PI structure is generated based on the single-unit combination method, specifically based on Monte Carlo tree search.

[0055] Finally, candidate molecular structures are output, specifically those with higher Tg values.

[0056] S10: Construct a synthetic monomer database, which stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers.

[0057] Specifically, S10 includes:

[0058] Systematically collect known dianhydride and diamine monomer structural information from publicly available chemical substance databases or literature; clean and standardize the collected data to remove erroneous or incomplete records; convert the processed monomer molecular structures into a computer-recognizable standardized coding format to establish a synthetic monomer database.

[0059] In the specific application of this embodiment, the data collection and preprocessing performed in S10 includes:

[0060] For monomer data collection, this includes extensively gathering molecular structure data of known dianhydride and diamine monomers from authoritative public chemical databases (such as PubChem, ChemicalBook, etc.) and relevant scientific literature and patents. These data sources mainly include compounds that have been experimentally proven to be synthesizable or commercially available, thus ensuring that the constructed monomer database has a high degree of synthetic availability.

[0061] Data cleaning involves rigorously screening and correcting the collected raw data. Entries with incorrect chemical structure connections, abnormal atomic valence states, duplicate records, or missing key information are removed to ensure the accuracy and reliability of the data.

[0062] For standardized coding, this includes using cheminformatics tools (such as RDKit) to uniformly convert the cleaned monomer molecular structures into simplified molecular linear input specifications (SMILES) strings, providing a standardized data format for subsequent computer processing and model input.

[0063] S20: Construct a molecular property prediction model to predict specific properties of polyimide molecules, including at least the glass transition temperature of the polyimide molecules.

[0064] Specifically, S20 includes:

[0065] The molecular performance prediction model is a machine learning model using a graph neural network (GNN) architecture. The molecular performance prediction model takes standardized encoding or its corresponding graph structure representation as input, and outputs corresponding performance prediction values ​​after feature extraction and nonlinear mapping. The performance prediction values ​​include at least the predicted glass transition temperature. The molecular performance prediction model is trained using polyimide with known structures and its performance data.

[0066] As a preferred embodiment of this invention, the construction of the molecular performance prediction model includes:

[0067] A training dataset is constructed based on the known molecular structure of polyimide and its corresponding experimental measurements, including at least the glass transition temperature.

[0068] A graph-based representation method is used to extract atomic information as node features and chemical bond information as edge features for the molecular graph structure of each polyimide molecule.

[0069] Construct a graph convolutional neural network model to predict specific properties of polyimide molecules;

[0070] The graph convolutional neural network model is trained using a training dataset under supervised learning conditions to optimize the model parameters and obtain a molecular performance prediction model.

[0071] As a preferred embodiment of this example, the hidden layer of the molecular performance prediction model is a multi-layer graph convolutional layer, which is used to aggregate neighborhood node information layer by layer to capture the local and global structural features of the molecule.

[0072] In this specific application, the aim is to establish a deep learning model capable of rapidly predicting the glass transition temperature (Tg) of polyimide based on its chemical structure. The specific process is as follows:

[0073] Training dataset construction: Collect existing datasets containing known polyimide molecular structures and their corresponding experimentally measured Tg values. The data is also cleaned, removing records with structural errors or missing / abnormal Tg values ​​to ensure the quality of the training data.

[0074] Feature Extraction: In this embodiment, a graph-based representation method is used. For the SMILES string of each polyimide molecule, tools such as RDKit are used to convert it into a molecular graph structure. Atomic information (such as atom type, hybridization state, etc.) is extracted as node features of the graph, and chemical bond information (such as bond type, conjugation status, etc.) is extracted as edge features of the graph.

[0075] Reference Figure 3 , Figure 3 This is a schematic diagram of the basic structure of the molecular performance prediction model provided in the embodiments of this application.

[0076] like Figure 3 As shown, the construction of the above-mentioned Graph Convolutional Neural Network (GCN) architecture includes:

[0077] Input layer: Receives the node features and edge features of the molecular graph extracted above.

[0078] Hidden Layers: Contain at least one, preferably multiple graph convolutional layers. In this preferred embodiment, two cascaded graph convolutional layers (e.g., conv1 and conv2) are defined to aggregate neighborhood node information layer by layer to capture the local and global structural features of the molecule. The preferred type of graph convolutional layer in this embodiment is GINEConv (graph isomorphic network convolution with edge features). After each graph convolutional operation, a non-linear activation function (e.g., LeakyReLU or ReLU) is preferably connected to enhance the model's expressive power, and a batch normalization (BatchNorm) layer can be selectively added to accelerate convergence and improve stability.

[0079] Following the graph convolutional layers, a fully connected (FC) module is connected to map the high-dimensional graph-level features to the final prediction output space. This module may contain one or more fully connected layers. For example, the first fully connected layer maps the feature dimensions to a smaller space (e.g., from 256 dimensions to 128 dimensions) and processes them using the ReLU activation function. To prevent overfitting, a Dropout mechanism is preferably introduced between the fully connected layers, with the dropout probability set between 0.1 and 0.5, such as 0.2 in this embodiment.

[0080] Output layer: The last fully connected layer maps the features to a scalar value, namely the predicted glass transition temperature Tg.

[0081] Model training: The GCN model described above is trained using the constructed training dataset to optimize the model parameters so that it can accurately predict the Tg value of unknown polyimide structures.

[0082] S30: Generate target polyimide structures based on Monte Carlo tree search. This search process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as target polyimide structures.

[0083] Specifically, S30 includes:

[0084] Using dianhydride and diamine monomers from the synthetic monomer database as basic nodes in the search space, an iterative search is performed using the Monte Carlo tree search algorithm. During the search process, candidate polyimide structures are generated by simulating the polymerization reaction of dianhydride and diamine, and the generated candidate polyimide structures are evaluated using a molecular performance prediction model. The evaluation results are converted into a reward function to guide the search direction of the Monte Carlo tree search algorithm. The evaluation results include at least the predicted glass transition temperature.

[0085] As a preferred embodiment of this invention, an iterative search is performed using the Monte Carlo tree search algorithm, including:

[0086] A search tree is constructed, with the root node representing the initial state, and the dianhydride monomer and diamine monomer constituting the action space to be selected in the search tree.

[0087] The Monte Carlo tree search algorithm explores the optimal combination through multiple iterative cycles. Each iteration includes: starting from the root node, selecting child nodes downwards in the tree according to a preset selection strategy until a leaf node with unexpanded child nodes is reached; if there are unexplored actions at the current leaf node, one or more new child nodes are created and added to the tree; based on the selected monomer pair composed of dianhydride monomer and diamine monomer, using a preset chemical reaction template, the molecular structure of the corresponding polyimide repeating unit is generated, and the generated molecular structure is input into the performance prediction model, outputting the predicted glass transition temperature of the structure, which is used as the reward value for this simulation; the obtained reward value is propagated upwards along the path to update the statistical information of all nodes on the path from the evaluated node to the root node, in order to guide the node selection in subsequent iterations.

[0088] Reference Figure 4 , Figure 4 This is a flowchart illustrating the specific implementation of polyimide structure generation based on Monte Carlo tree search, as provided in this application embodiment.

[0089] like Figure 4 As shown, in the specific application of this embodiment, we utilize the Monte Carlo Tree Search (MCTS) algorithm to perform an efficient search within the combinatorial space defined by the dianhydride and diamine monomer libraries constructed in step S10. The specific process is as follows:

[0090] Initialization: Establish a search tree, with the root node representing the initial state (no monomer selected). The dianhydride monomer and diamine monomer in step S10 constitute the action space (nodes) to be selected in the search tree.

[0091] Iterative Search: The MCTS algorithm explores the optimal combination through multiple iterative loops (e.g., setting the number of loops to 1000, 2000, or more, depending on the computational resources and search depth requirements; this embodiment uses 2000 loops as an example). Each iteration typically includes the following four stages:

[0092] (a) Selection: Starting from the root node, child nodes are selected downwards in the tree according to a specific selection strategy (such as the Upper Confidence Bounds algorithm, UCT) until a leaf node with unexpanded child nodes is reached. In this embodiment, this corresponds to selecting a dianhydride monomer and a diamine monomer in sequence.

[0093] (b) Extension: If there are unexplored actions (i.e., singleton pairs that have not yet been combined) in the current leaf node, create one or more new child nodes and add them to the tree.

[0094] (c) Evaluation: Based on the selected dianhydride and diamine monomer pair, using a pre-defined chemical reaction template (simulating the condensation reaction of dianhydride and diamine), the molecular structure (represented by SMILES) of the corresponding polyimide repeating unit is generated using tools such as RDKit. Subsequently, the generated polyimide structure is input into the molecular performance prediction model trained in step S20, and the model outputs the predicted Tg value of the structure. This predicted Tg value (or the score calculated based on the Tg value) is used as the reward value for this simulation. The reward function is designed to guide the search towards higher Tg values.

[0095] (d) Backtracking: The obtained reward value is propagated backward along the path to update the statistics of all nodes on the path from the evaluated node to the root node (such as the number of visits and the average reward value), which is used to guide the selection of nodes in subsequent iterations.

[0096] S40: Output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties.

[0097] In the specific application of this embodiment, after completing a preset number of MCTS iterations, the search results are sorted according to the predicted Tg values, and suitable combinations of dianhydride and diamine monomers are selected based on expert knowledge. The polyimide molecular structures, monomer information, and predicted Tg values ​​corresponding to these combinations are output, forming a final candidate list for researchers to conduct subsequent synthesis verification.

[0098] In summary, the monomer combination-based polyimide molecular structure generation method of this embodiment is an automated polyimide structure design method based on a limited combination of dianhydride and diamine monomers, combined with machine learning performance prediction and Monte Carlo tree search algorithm. It overcomes the problems of chemical space explosion and poor syntheticability in the prior art, and achieves the following: by limiting the basic building monomers, the chemical search space is controlled within a reasonable and operable range; based on the combination of commercially available or known synthetic monomers, the practical syntheticability of the generated structure is guaranteed from the source; and the monomer combination space is systematically explored using intelligent algorithms, which significantly improves the discovery efficiency of novel high-performance polyimide materials.

[0099] Based on the above, the method for generating polyimide molecular structures based on monomer combinations in this embodiment has at least the following advantages compared with the prior art:

[0100] This embodiment creatively adopts a "monomer-based combination" strategy. On the one hand, by limiting the number of basic building blocks (dianhydrides and diamines), the huge chemical space is effectively confined within a calculable and controllable reasonable range, overcoming the problem of "chemical space explosion". On the other hand, since the selected monomers are all derived from reported or commercially available mature compounds, the feasibility of the final polyimide candidate structure being synthesized in the laboratory is greatly improved compared to freely generated molecular fragments.

[0101] This embodiment combines the efficient optimization capabilities of the Monte Carlo Tree Search (MCTS) algorithm with the fast and accurate evaluation capabilities of a deep learning model. The MCTS algorithm can intelligently plan search paths within a vast combinatorial space, avoiding blind exhaustive searches; while the graph neural network-based prediction model provides a fast and reliable performance feedback mechanism for MCTS. The synergistic effect of these two algorithms significantly improves the efficiency of screening high-performance materials, greatly shortens the R&D cycle, and reduces trial-and-error costs.

[0102] Reference Figure 5 , Figure 5 This is a structural block diagram of a polyimide molecular structure generation device based on monomer combination provided in an embodiment of this application.

[0103] like Figure 5 As shown, this embodiment also discloses a device for generating polyimide molecular structures based on monomer combinations. The device utilizes the method for generating polyimide molecular structures based on monomer combinations as described above. The device includes:

[0104] The database construction module is used to build a synthetic monomer database, which stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers.

[0105] The prediction model building module is used to build a molecular property prediction model, which is used to predict specific properties of polyimide molecules, including at least the glass transition temperature of polyimide molecules.

[0106] The target search module is used to generate target polyimide structures based on Monte Carlo tree search. This search generation process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as target polyimide structures.

[0107] The target output module is used to output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties.

[0108] This embodiment also discloses a computer device for generating polyimide molecular structures based on monomer combinations, including at least one processor, at least one memory, and a data bus;

[0109] The processor and memory communicate with each other via a data bus;

[0110] The memory stores program instructions that can be executed by the processor, which calls the program instructions to execute the method for generating polyimide molecular structures based on monomer combinations as described above.

[0111] This embodiment also discloses a storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the method for generating polyimide molecular structures based on monomer combinations as described above.

[0112] It should be noted that the monomer-based polyimide molecular structure generation device, computer equipment, and storage medium of this embodiment correspond to the aforementioned monomer-based polyimide molecular structure generation method. Therefore, any content not specifically described in the monomer-based polyimide molecular structure generation device, computer equipment, and storage medium of this embodiment, including but not limited to functional definitions, working principles, and technical effects, can be referred to the description in the aforementioned monomer-based polyimide molecular structure generation method, and will not be repeated here.

[0113] In the embodiments provided in this application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor may be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments may be performed by a computer program instructing the associated hardware. During implementation, the program may be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media may be any available medium accessible to a computer. Computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0114] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for generating polyimide molecular structures based on monomer combinations, characterized in that, The method includes: S10: Construct a synthetic monomer database, wherein the synthetic monomer database stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers; S20: Construct a molecular property prediction model, which is used to predict specific properties of polyimide molecules, including at least the glass transition temperature of polyimide molecules; S30: Generate the target polyimide structure based on Monte Carlo tree search. This search generation process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as the target polyimide structure. S40: Output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties; S30, specifically includes: Using dianhydride and diamine monomers from the synthetic monomer database as basic nodes in the search space, an iterative search is performed using the Monte Carlo tree search algorithm. During the search process, candidate polyimide structures are generated by simulating the polymerization reaction of dianhydride and diamine, and the generated candidate polyimide structures are evaluated using the molecular performance prediction model. The evaluation results are converted into a reward function to guide the search direction of the Monte Carlo tree search algorithm. The evaluation results include at least the predicted glass transition temperature.

2. The method for generating polyimide molecular structures based on monomer combinations according to claim 1, characterized in that, S10 specifically includes: Systematically collect known dianhydride and diamine monomer structural information from publicly available chemical substance databases or literature; clean and standardize the collected data to remove erroneous or incomplete records; convert the processed monomer molecular structures into a computer-recognizable standardized coding format to establish a synthetic monomer database.

3. The method for generating polyimide molecular structures based on monomer combinations according to claim 1, characterized in that, S20 specifically includes: The molecular performance prediction model is a machine learning model using a graph neural network architecture. The molecular performance prediction model takes the standardized code or its corresponding graph structure representation as input, performs feature extraction and nonlinear mapping, and outputs the corresponding performance prediction value. The performance prediction value includes at least the predicted glass transition temperature. The molecular performance prediction model is trained using polyimide with known structures and its performance data.

4. The method for generating polyimide molecular structures based on monomer combinations according to claim 3, characterized in that, The construction of the molecular performance prediction model includes: A training dataset is constructed based on the known molecular structure of polyimide and its corresponding experimental measurements, wherein the experimental measurements include at least the glass transition temperature. A graph-based representation method is used to extract atomic information as node features and chemical bond information as edge features for the molecular graph structure of each polyimide molecule. Construct a graph convolutional neural network model to predict specific properties of polyimide molecules; The graph convolutional neural network model is trained using the training dataset under supervised learning conditions to optimize the model parameters and obtain the molecular performance prediction model.

5. The method for generating polyimide molecular structures based on monomer combinations according to claim 4, characterized in that, The hidden layer of the molecular performance prediction model is a multi-layer graph convolutional layer, which is used to aggregate neighborhood node information layer by layer to capture the local and global structural features of the molecule.

6. The method for generating polyimide molecular structures based on monomer combinations according to claim 1, characterized in that, The iterative search using the Monte Carlo tree search algorithm includes: Construct a search tree, with the root node representing the initial state, and the dianhydride monomer and diamine monomer constituting the action space to be selected in the search tree. The Monte Carlo tree search algorithm explores the optimal combination through multiple iterative cycles. Each iteration includes: starting from the root node, selecting child nodes downwards in the tree according to a preset selection strategy until a leaf node with unexpanded child nodes is reached; if there are unexplored actions at the current leaf node, one or more new child nodes are created and added to the tree; based on the selected monomer pair composed of dianhydride monomer and diamine monomer, using a preset chemical reaction template, the molecular structure of the corresponding polyimide repeating unit is generated, and the generated molecular structure is input into the performance prediction model, outputting the predicted glass transition temperature of the structure, which is used as the reward value for this simulation; the obtained reward value is propagated upwards along the path to update the statistical information of all nodes on the path from the evaluated node to the root node, in order to guide the node selection in subsequent iterations.

7. A device for generating polyimide molecular structures based on monomer combinations, employing the method for generating polyimide molecular structures based on monomer combinations as described in any one of claims 1 to 6, characterized in that, The device includes: A database construction module is used to construct a synthetic monomer database, which stores standardized codes corresponding to the structural information of dianhydride monomers and diamine monomers. A prediction model building module is used to build a molecular property prediction model, which is used to predict specific properties of polyimide molecules, including at least the glass transition temperature of polyimide molecules. The target search module is used to generate target polyimide structures based on Monte Carlo tree search. This search generation process is used to find high-performance polyimide candidate structures with target properties in the combinatorial space and define them as the target polyimide structures. The target output module is used to output candidate solutions, which include the target polyimide structure and its molecular formula, the corresponding dianhydride and diamine monomer composition information, and the predicted specific properties.

8. A computer device for generating polyimide molecular structures based on monomer combinations, characterized in that, Includes at least one processor, at least one memory, and a data bus; The processor and the memory communicate with each other via the data bus; The memory stores program instructions that can be executed by the processor, which invokes the program instructions to execute the method for generating polyimide molecular structures based on monomer combinations as described in any one of claims 1 to 6.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method for generating polyimide molecular structures based on monomer combinations as described in any one of claims 1 to 6.

Images