A method, device and medium for processing structured data

Abstract

The present application discloses a method, device and medium for processing structural data, which is applied to multiple structural data processing devices, and each structural data processing device executes the method in parallel, and the method includes: standardizing the original data of the non-spatial structure of the cell to obtain a structural knowledge graph of the three-dimensional structure composed of cells; constructing a training data set for training the training model according to a pre-constructed question-answering strategy and structural knowledge graph. Thus, under the parallel processing of multiple structural data processing devices, the original data is standardized quickly and efficiently to obtain a structural knowledge graph, and a training data set is generated through a pre-constructed question-answering strategy, so that the model training data is efficiently processed to obtain high-quality training data. In addition, after training with the high-quality training data set provided by the present application, the overall performance of the model can be improved, thereby improving the performance of the cells and materials.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device and medium for processing structured data. Background Art

[0002] Cells refer to the basic units that make up materials. Factors such as the shape, size, and arrangement of cells will affect the overall performance of the material. Since there is a close relationship between the microstructure of a material (i.e., cells) and its performance, a reasonable design of cells can achieve the regulation and optimization of the macroscopic performance of the material, thereby meeting the needs of different application scenarios such as aerospace, medical equipment, energy engineering, and electronic technology.

[0003] With the rapid development of large language models, in a feasible implementation method of regulating material properties through cell design, the large language model and cell design can be combined to explore the performance boundaries of the cell structure through the large language model, thereby improving the performance of the material.

[0004] When exploring better cell structures through large language models to improve material performance, the performance of large language models is crucial. The performance of large language models is closely related to the quality of training data. How to obtain high-quality structural training data, train large language models, improve the performance of large language models, and thus improve the performance of cells and materials is an urgent problem to be solved by technicians in this field. Summary of the invention

[0005] In view of this, one aspect of the present application provides a method for processing structured data, which is applied to a structured data processing device in a data processing system, wherein the structured data processing device includes a plurality of structured data processing devices, each of which executes the method for processing structured data in parallel, and the method includes:

[0006] Obtain the original data of the non-spatial structure of the cell;

[0007] The raw data is standardized to obtain a structural knowledge graph of a three-dimensional structure composed of the cells; wherein the structural knowledge graph includes structural performance and basic structural data in a spatial structure;

[0008] According to the pre-constructed question-answering strategy and the structural knowledge graph, a training data set for training the model to be trained is constructed.

[0009] Optionally, the raw data is subjected to standardization processing, including:

[0010] Converting the original data into point-bar data formed by connecting the coordinates of every two nodes in the spatial structure;

[0011] The basic structure data is obtained by merging the point rods in the point rod data that are located on the same straight line and have intersections.

[0012] Optionally, merging the point bars in the point bar data that are on the same straight line and have intersections includes:

[0013] According to the node coordinates, the point rod data is converted into vector data in the same direction;

[0014] Determine, based on the vector data, whether there is a point-rod data group in which the point-rods are parallel to each other and have intersections in the point-rod data;

[0015] If so, the node coordinates in the point-rod data group are sorted according to the same direction; and based on the first two node coordinates in the sorting result, all the point-rod data in the point-rod data group are merged into a target vector in the same direction.

[0016] Optionally, constructing a training data set for training the model to be trained according to the pre-constructed question-answering strategy and the structural knowledge graph includes:

[0017] Based on the structural knowledge graph, generate a set of answers to the three-dimensional structure under different question-answering strategies;

[0018] Counting the number of three-dimensional structures under different answer sets;

[0019] Constructing a correspondence between the three-dimensional structure, the question-answering strategy, and the answer set;

[0020] The training data set is obtained according to the number of the three-dimensional structures and the corresponding relationship.

[0021] Optionally, the method for processing structure data further includes:

[0022] Mixing and acquiring data of different structural properties in a preset ratio from the training data set to obtain a target training data set;

[0023] Training the model to be trained using the target training data set;

[0024] If a fine-tuning instruction is received, the preset ratio is adjusted according to the fine-tuning instruction to optimize the model to be trained.

[0025] Optionally, the step of obtaining data of different structural properties in a preset ratio from the training data set by mixing to obtain a target training data set includes:

[0026] Determining whether the data of each of the structural performances in the training data set meets the data requirement of the preset ratio;

[0027] If satisfied, obtain the preset proportion of data to obtain the target training data set;

[0028] If not, the data of the structural performance that does not meet the preset ratio requirement is repeatedly acquired until the preset ratio is met to obtain the target training data set.

[0029] Optionally, the structural performance includes at least one of volume fraction, structural topological performance and structural mechanical performance; when the structural performance includes structural mechanical performance, the raw data is standardized, including:

[0030] Physical simulation is performed on the basic structure data to determine the structural mechanical properties; wherein the structural mechanical properties include ultimate strength properties and Young's modulus properties.

[0031] Another aspect of the present application provides a structure data processing device, comprising:

[0032] A raw data acquisition module, used to acquire raw data of the non-spatial structure of the cell;

[0033] A standardization processing module, used for performing standardization processing on the original data to obtain a structural knowledge graph of the three-dimensional structure composed of the cells; wherein the structural knowledge graph includes structural performance and basic structural data in the spatial structure;

[0034] A construction module is used to construct a training data set for training the model to be trained based on a pre-built question-answering strategy and the structural knowledge graph.

[0035] Another aspect of the present application provides a structure data processing device, including a memory and a processor, wherein the memory stores a computer program that can be executed on the processor, and the processor implements the steps of the structure data processing method when executing the program.

[0036] Another aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the program implements the steps of the method for processing structured data when executed by a processor.

[0037] The present application provides a method, device and medium for processing structural data, which have the following beneficial effects: under the architecture of parallel processing of multiple structural data processing devices, the original data of cells of non-spatial structures can be quickly and efficiently standardized to obtain a structural knowledge graph including basic structural data in spatial structures and cell structural performance. Furthermore, a training data set is generated through a pre-built question-and-answer strategy to achieve efficient processing of model training data, and high-quality training data can be obtained while ensuring data processing efficiency. Therefore, after the model is continuously trained with the high-quality training data set provided by the present application, the overall performance of the model can be improved. By combining the model with cells, the boundaries of cell structural performance can be explored, cell performance can be improved, and material performance can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flowchart of a method for processing structured data provided in an embodiment of the present application;

[0039] Figure 2 A schematic diagram of the structure of a data processing system provided in an embodiment of the present application;

[0040] Figure 3 A flowchart of a method for processing structured data provided by another embodiment of the present application;

[0041] Figure 4 A schematic diagram of a structural data processing device provided in an embodiment of the present application;

[0042] Figure 5 A schematic diagram of the structure of a structural data processing device provided in another embodiment of the present application.

[0043] The accompanying drawings are marked as follows: 1 is a data management platform, 2 is a structural data processing device, 50 is a memory, 51 is a processor, 52 is a display screen, 53 is an input and output interface, 54 is a communication interface, 55 is a power supply, 56 is a communication bus, 501 is a computer program, 502 is an operating system, and 503 is data. DETAILED DESCRIPTION

[0044] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0045] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0046] Figure 1 A flowchart of a method for processing structured data provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method includes:

[0047] S10: obtaining the original data of the non-spatial structure of the cell;

[0048] Figure 2 This is a structural diagram of a data processing system provided in an embodiment of the present application. First, it should be noted that the structural data processing method provided in the embodiment of the present application is applied to Figure 2 The structural data processing device 2 in the data processing system shown, that is, in an optional embodiment, the execution subject of the present application can be the structural data processing device 2.

[0049] like Figure 2 As shown, the data processing system includes a data management platform 1 and a plurality of structure data processing devices 2, wherein the data management platform 1 is used to distribute the original data of the cell to the plurality of structure data processing devices 2, that is, to distribute the data processing tasks. In an optional embodiment, the data management platform 1 evenly distributes the original data of the cell to each structure data processing device 2 for data processing according to the number of structure data processing devices 2. For example, if the data currently to be processed includes 10,000 pieces, and the number of structure data processing devices 2 is 10, 1,000 pieces of data are distributed to each structure data processing device 2.

[0050] In another optional embodiment, the data can be distributed according to the amount of remaining unprocessed data of each structural data processing device 2. The structural data processing device 2 with a large amount of remaining unprocessed data sends less data, and the structural data processing device 2 with a small amount of remaining unprocessed data sends more data, thereby ensuring maximum resource utilization.

[0051] In a specific embodiment, the method for processing structural data provided by the present application obtains the original data of the cell through step S10, and the original data is the data of the non-spatial structure generated by the rule. In an optional embodiment, the method for generating cells by rules refers to establishing a three-dimensional space coordinate system with the vertices of a cube as the origin, labeling the specified points on the cube, that is, marking the specified points with serial numbers, and generating cells based on multiple specified points on the cube. Therefore, the original data of the cell refers to an array composed of the serial numbers of the specified points that generate the cell.

[0052] For example, the original data of the non-spatial structure generated by the rule is [(4, 19), (16, 20), (10, 20)], where 4, 10, 16, 19, 20 represent the specified points on the cube with serial numbers 4, 10, 16, 19, 20 respectively. Obviously, the original data cannot be directly parsed into data in the spatial structure, that is, the spatial structure data of the cell cannot be directly determined, and thus the spatial structure data of the three-dimensional structure composed of the cell cannot be determined. In other words, the original data of the cell can be understood as a type of data that cannot be directly parsed into a spatial structure.

[0053] S11: Standardize the original data to obtain a structural knowledge graph of a three-dimensional structure composed of cells; wherein the structural knowledge graph includes structural performance and basic structural data in the spatial structure;

[0054] Since the original data of the cell cannot be directly parsed into data in the spatial structure, in an optional embodiment, the basic structural data of the cell in the spatial structure can be obtained by standardizing the original data in step S11. In addition, the structural performance of the cell can be determined, and then the performance of the three-dimensional structure can be determined. Thus, the structural performance of the three-dimensional structure and the basic structural data are combined to form a structural knowledge graph.

[0055] In an optional embodiment, the basic structure data includes point rod data formed by connecting the coordinates of every two nodes in the spatial structure. In order to improve the quality of the training data set, the standardization processing can be the conversion of the original data into point rod data in the spatial structure that can be directly parsed. It can also include eliminating redundant point rods in the point rod data, or merging point rods that meet the conditions. Specifically, the condition can be that parallel and intersecting point rods are merged.

[0056] In another optional embodiment, the standardization process may further include unifying the basic structure data into data under the same standard coordinate system, that is, ensuring that the scales of the node coordinates in the point bar data are the same. In addition, the standardization process also includes processing the basic structure data into data with uniform precision, for example, retaining the node coordinates to 3 decimal places.

[0057] It is worth noting that the structural properties of the three-dimensional structure may include but are not limited to volume fraction, structural topological properties and structural mechanical properties, wherein the structural topological properties include connectivity, symmetry, etc., and the structural mechanical properties include ultimate strength properties and Young's modulus properties, etc. In a specific embodiment, the structural topological properties are the intrinsic properties of the three-dimensional structure and can be directly obtained. The volume fraction and structural mechanical properties need to be determined through standardization.

[0058] S12: Based on the pre-built question-answering strategy and structural knowledge graph, a training dataset is constructed for training the model to be trained.

[0059] It should be noted that the training data set obtained by the method provided in this application can be used to train the model to be trained, thereby improving the performance of the model to be trained. In an optional embodiment, the model to be trained can be a large language model, and this application does not limit the model to be trained.

[0060] In a specific embodiment, a training data set can be constructed based on a pre-constructed question-answering strategy and the structural knowledge graph obtained in step S11. In an optional embodiment, the question-answering strategy can be a generative strategy, for example, describing a given three-dimensional structure. The question-answering strategy can also be a judgment strategy, for example, judging whether a given three-dimensional structure is connected.

[0061] It is understandable that based on the pre-built question-answering strategy and the structural knowledge graph of the three-dimensional structure, a training data set of questions about the three-dimensional structure and the corresponding answers can be generated. The training data set can be used to train the training model to obtain a model that can analyze the three-dimensional structure. Therefore, by combining the trained model with the cell design, the boundaries of the cell performance can be explored, that is, how to improve the performance of the cell, and then improve the performance of the material composed of the cell.

[0062] The method for processing structural data provided in the embodiment of the present application can quickly and efficiently perform standardized processing on the original data of cells of non-spatial structures under the architecture of parallel processing of multiple structural data processing devices 2, and obtain a structural knowledge graph of a three-dimensional structure composed of cells. Furthermore, a training data set is generated through a pre-built question-and-answer strategy to achieve efficient processing of model training data, and high-quality training data can be obtained while ensuring data processing efficiency. Thus, after the model is continuously trained with the high-quality training data set provided by the present application, the overall performance of the model can be improved. By combining the model with cells, the boundaries of cell structure performance can be explored, cell performance can be improved, and material performance can be improved.

[0063] As an optional embodiment, the raw data is standardized, including:

[0064] Convert the original data into point-bar data formed by connecting the coordinates of every two nodes in the spatial structure;

[0065] After merging the point rods in the point rod data that are located on the same straight line and have intersections, the basic structure data is obtained.

[0066] It is understandable that the original data of the cell cannot be directly parsed into the data of the spatial structure. Therefore, the original data needs to be standardized in order to obtain the basic data in the spatial structure. Specifically, the original data is first converted into a point-rod structure formed by connecting the coordinates of every two nodes in the spatial structure.

[0067] For example, the original data is [(4, 19), (16, 20), (10, 20)]. Each serial number can be converted into a node coordinate point, and the node coordinates corresponding to two serial numbers can be connected to obtain a point rod. For example, (4, 19) can be converted into a point rod.

[0068] In the original data of the cell, there are 6 node coordinates and 3 point rods. In an optional embodiment, the three-dimensional structure is composed of 8 cells, that is, the cell is one eighth of the three-dimensional structure. Therefore, when converting to a three-dimensional structure to obtain the basic structural data of the three-dimensional structure, the original data needs to be flipped to restore the entire three-dimensional structure. Thus, 48 ​​node coordinates and 24 point rods can be obtained.

[0069] Furthermore, in an optional embodiment, considering that the large language model waiting for training has a limit on the size of the input data to be trained, it is necessary to clean the data. Specifically, after the original data is converted into point-bar data, the point-bar data that are located on the same straight line and have intersections are merged. That is, the point-bars that are on the same straight line and have intersections can be merged into one point-bar data, thereby reducing the data volume of the basic structure data.

[0070] Figure 3 A flowchart of a method for processing structured data provided by another embodiment of the present application is provided. Based on the above embodiment, as an optional embodiment, Figure 3 As shown in the figure, the point bars in the point bar data that are located on the same straight line and have intersections are merged, including:

[0071] S30: converting the point rod data into vector data in the same direction according to the node coordinates;

[0072] In a specific embodiment, all point rod data are converted into vector data in the same direction according to the node coordinates, that is, two coordinates on the point rod are converted into vectors, and the directions of the vectors of different point rods are the same.

[0073] S31: Determine, based on the vector data, whether there is a point-rod data set in which the point-rods are parallel to each other and have intersections in the point-rod data; if so, execute steps S32 and S33.

[0074] Further, it is determined whether any of the point rods are parallel to each other or intersecting according to the vector data of the point rods. Specifically, in an optional embodiment, when determining whether two point rods are parallel, the vector data corresponding to the two point rods can be calculated. If the calculation result is zero, it indicates that the two point rods are parallel. Otherwise, the two point rods are not parallel.

[0075] Under the condition of parallelism, to determine whether two point rods have an intersection, the vector difference of the two vector data can be calculated, and the projection of the vector difference on any of the original vectors can be calculated. If this projection is within the length range of the vector, it indicates that the two point rods have an intersection.

[0076] When it is determined based on the vector data calculation that there are point rods that are both parallel and intersecting, such point rod data are classified into a point rod data group.

[0077] Of course, if there is no point-rod data set in which the point rods are parallel to each other and have intersections in the point-rod data, the point-rod data can be directly used as the basic structure data.

[0078] S32: sorting the node coordinates in the point-bar data group in the same direction;

[0079] S33: According to the coordinates of the first two nodes in the sorting result, all point-bar data in the point-bar data group are merged into a target vector in the same direction.

[0080] Furthermore, after obtaining the point rod data group that satisfies the parallel and intersecting conditions, the point rods in each point rod data group need to be merged. Specifically, the node coordinates in the point rod data group are first sorted in the same direction through step S32. It should be noted that the sorting direction is the same as the direction of the vector. After sorting, the first two node coordinates in the sorting result are used as the merged coordinates through step S33, that is, the first two node coordinates are converted into a vector to obtain the vector after the point rod data group is merged, thereby completing the merging of parallel and intersecting point rods.

[0081] It is worth noting that the training data set may not only include point rods, but may also include expressions such as tubes and plates. Therefore, in an optional embodiment, after merging the parallel and intersecting point rod data to obtain the basic structure data, the basic structure data is converted into a description method of a standard JSON schema, and a specified algorithm is called to calculate the volume fraction of each point rod structure, that is, the structural performance of the volume fraction of the three-dimensional structure can be obtained. This application does not limit the specified algorithm.

[0082] Therefore, the method for processing structural data provided in the embodiment of the present application merges parallel and intersecting point rods, that is, cleans the point rod data, reduces the amount of input data for the model to be trained, and ensures both training quality and training efficiency.

[0083] In an optional embodiment, a training data set for training a model to be trained is constructed according to a pre-constructed question-answering strategy and a structural knowledge graph, including:

[0084] Based on the structural knowledge graph, a set of answers with three-dimensional structures under different question-answering strategies is generated;

[0085] Count the number of three-dimensional structures under different answer sets;

[0086] Construct the correspondence between the three-dimensional structure, question-answering strategy and answer set;

[0087] According to the number and corresponding relationship of the three-dimensional structures, a training data set is obtained.

[0088] In a specific embodiment, the knowledge graph composed of structural performance and basic structural data needs to form a data set that can be used to train the model to be trained based on the question-answering strategy. Specifically, based on the structural knowledge graph, a set of answers under different question-answering strategies is generated.

[0089] It is understandable that different questions (i.e., corresponding question-answering strategies) may be asked about the three-dimensional structure to obtain corresponding answer sets, wherein the question-answering strategy may be a generative strategy, for example, the generative strategy describes a given three-dimensional structure. In this case, different three-dimensional structures may obtain different answer sets under the generative strategy.

[0090] The question-answering strategy can also be a judgment strategy, for example, the judgment strategy is to judge whether a given three-dimensional structure is connected, so that under this judgment strategy, a connectivity answer set of different three-dimensional results can be obtained. In an optional embodiment, the question-answering strategy can also be a structure construction strategy, for example, the structure construction strategy is to construct a connected three-dimensional structure with 10 point rods, and correspondingly, the connected three-dimensional structure with 10 point rods is used as the answer set.

[0091] In addition, in another optional embodiment, the question-answering strategy may include a question-answering strategy about structural knowledge in addition to a strategy about a given structure, for example, what is a cell. It should be noted that in a specific embodiment, the preset question-answering strategy may be updated and expanded to meet the needs of different scenarios.

[0092] In addition, it should be noted that the pre-established question-answering strategy can be based on cells or on a three-dimensional structure composed of cells. This application does not limit this, as long as a structural knowledge graph about the three-dimensional structure can be obtained.

[0093] In an optional embodiment, in order to facilitate the subsequent training of the training model, the number of stereoscopic structures under different difficult sets can be counted. At the same time, a correspondence between the stereoscopic structure, the question-answering strategy and the answer set is established, thereby obtaining a training data set based on the statistical number of stereoscopic structures and the corresponding relationship. Table 1 is a correspondence table of a stereoscopic structure, question-answering strategy and answer set provided in an embodiment of the present application. For ease of understanding, it will be explained in conjunction with Table 1 below.

[0094] Table 1 A corresponding relationship table provided in the embodiment of the present application

[0095]

[0096] As described in Table 1, the stereostructure includes stereostructure A, stereostructure B, and stereostructure C, and the question-answering strategies include strategy 1 and strategy 2. Based on the structural knowledge graphs corresponding to stereostructure A, stereostructure B, and stereostructure C, the answer sets corresponding to each stereostructure under different question-answering strategies can be obtained. For example, the answer corresponding to stereostructure A under strategy 1 is A1.

[0097] It should be noted that different correspondence tables may be set for different three-dimensional structures, or all three-dimensional structures may be set in the same correspondence table, which is not limited in this application.

[0098] In an optional embodiment, in order to further improve the performance of the model to be trained, the method for processing structure data provided in the embodiment of the present application further includes:

[0099] From the training data set, data of different structural performances in a preset ratio are mixed and obtained to obtain a target training data set;

[0100] Train the model to be trained using the target training data set;

[0101] If a fine-tuning instruction is received, the preset ratio is adjusted according to the fine-tuning instruction to optimize the model to be trained.

[0102] It is understandable that different proportions of data, that is, the proportions of data with different structural properties, can improve the performance, generalization ability, and performance of the model to be trained on a specific task. Therefore, in an optional embodiment, after obtaining the training data set, data with different structural properties at a preset proportion can be mixed from all training data sets to obtain a target training data set.

[0103] That is to say, data of different structural performances are mixed and acquired in a certain proportion to obtain a target training data set, and the model to be trained is trained with the target training data set, thereby improving the overall performance of the model to be trained.

[0104] In order to further improve the performance of the model to be trained and ensure that the model to be trained can be applied to a specific application scenario, in an optional embodiment, the user can adjust the preset ratio according to the training results. Figure 2 The structural data processing device 2 shown sends a fine-tuning instruction so that the structural data processing instruction adjusts the preset ratio according to the fine-tuning instruction, thereby optimizing the model to be trained.

[0105] Based on the above embodiment, as an optional embodiment, data of different structural performances with a preset ratio are mixed and obtained from the training data set to obtain a target training data set, including:

[0106] Determine whether the data of each structural performance in the training data set meets the data requirements of the preset ratio;

[0107] If satisfied, obtain the preset proportion of data to obtain the target training data set;

[0108] If it is not satisfied, the data of the structural performance corresponding to the preset ratio requirement that does not meet the preset ratio requirement is repeatedly obtained until the preset ratio is met to obtain the target training data set.

[0109] When mixing and obtaining data of different structural properties as the target training data set, first determine whether the data of each structural property in the training data set meets the requirements of the preset ratio. If so, directly mix and obtain the data of the preset ratio.

[0110] If it is not satisfied, repeatedly obtain data that does not meet the preset ratio until the preset ratio is met. For example, the structural performance includes structural performance A, structural performance B, and structural performance C, and the basic training data corresponding to structural performance A is 100,000, the basic training data corresponding to structural performance B is 100,000, and the basic training data corresponding to structural performance C is 500,000, and the preset ratio is 1:2:3.

[0111] When data is obtained according to the preset ratio, 100,000 pieces of data of structural performance A, 200,000 pieces of data of structural performance B, and 300,000 pieces of data of structural performance C should be obtained. At this time, the basic training data corresponding to structural performance B is obtained twice repeatedly, and the basic training data that meets the preset ratio is randomly obtained from structural performance C, that is, 300,000 pieces of data are randomly obtained from 500,000 pieces of data.

[0112] Therefore, the method for processing structural data provided in the embodiment of the present application improves the generalization ability of the model to be trained by mixing data of structural performance in different proportions to train the model to be trained, thereby improving the overall performance of the model.

[0113] As an optional embodiment, the structural performance includes at least one of volume fraction, structural topological performance and structural mechanical performance. Wherein, when the structural performance includes structural mechanical performance, the raw data is standardized, including:

[0114] Physical simulation is performed on the basic structural data to determine the structural mechanical properties, wherein the structural mechanical properties include ultimate strength properties and Young's modulus properties.

[0115] Specifically, the basic structure data is first converted into a standard STL format, and the basic structure data in the STL format is subjected to physical simulation to obtain structural mechanical properties including ultimate strength performance and Young's modulus performance. It can be understood that through physical simulation, not only can the structural mechanical properties be determined, but physical simulation can also save costs compared to 3D printing.

[0116] It should be noted that the structural topological performance may include but is not limited to validity, connectivity and symmetry, where validity refers to whether the basic structural data are all within the cube that constructs the three-dimensional structure, connectivity refers to whether the point-rod structures are connected, and symmetry refers to whether the three-dimensional structure or the point-rod is symmetrical about an axis, plane, etc.

[0117] The structural topological performance is the attribute characteristic of the three-dimensional structure itself, and will not change due to the merging of point-rod structures, that is, it will not change due to standardization processing, and can be directly obtained based on the obtained three-dimensional structure.

[0118] The volume fraction is an important parameter in materials science and engineering. In the embodiment of the present application, the volume fraction can be the proportion of each point-rod structure to the total volume fraction. The present application does not limit the specific method for determining the volume fraction.

[0119] In fact, structural properties include at least one of volume fraction, structural topological properties and structural mechanical properties, and of course may also include other properties, as long as the properties can be used to construct a structural knowledge graph of a three-dimensional structure, and this application does not limit this.

[0120] In the above embodiments, the method for processing structure data is described in detail. The present application also provides a corresponding embodiment of a structure data processing device.

[0121] Figure 4 A structural diagram of a structural data processing device provided in an embodiment of the present application is shown in FIG. Figure 4As shown, the data processing unit includes:

[0122] The original data acquisition module 40 is used to acquire the original data of the non-spatial structure of the cell;

[0123] A standardization processing module 41 is used to perform standardization processing on the original data to obtain a structural knowledge graph of a three-dimensional structure composed of cells; wherein the structural knowledge graph includes structural performance and basic structural data in the spatial structure;

[0124] The construction module 42 is used to construct a training data set for training the model to be trained based on the pre-constructed question-answering strategy and structural knowledge graph.

[0125] In addition, the structure data processing device provided in the embodiment of the present application also includes:

[0126] A raw data conversion module, used to convert raw data into point-bar data formed by connecting the coordinates of every two nodes in the spatial structure;

[0127] The point-rod data merging module is used to merge the point-rods that are located on the same straight line and have intersections in the point-rod data to obtain the basic structure data.

[0128] A point-rod data conversion module is used to convert the point-rod data into vector data in the same direction according to the node coordinates;

[0129] The first processing module is used to determine whether there is a point-rod data group in which the point rods are parallel to each other and have intersections in the point-rod data based on the vector data; if so, sort the node coordinates in the point-rod data group in the same direction; and merge all the point-rod data in the point-rod data group into a target vector in the same direction based on the first two node coordinates in the sorting result.

[0130] The answer set generation module is used to generate a set of answers with three-dimensional structures under different question-answering strategies based on the structural knowledge graph;

[0131] The three-dimensional structure quantity statistics module is used to count the number of three-dimensional structures under different answer sets;

[0132] The correspondence building module is used to build the correspondence between the three-dimensional structure, the question-answering strategy, and the answer set;

[0133] The training data set acquisition module is used to obtain the training data set according to the number and corresponding relationship of the three-dimensional structures.

[0134] A target training data set acquisition module is used to obtain data of different structural properties in a preset proportion from the training data set to obtain a target training data set;

[0135] The training module is used to train the training model using the target training data set;

[0136] The fine-tuning module is used to adjust the preset ratio according to the fine-tuning instruction if a fine-tuning instruction is received to optimize the model to be trained.

[0137] The second processing module is used to determine whether the data of each structural performance in the training data set meets the data requirement of a preset ratio; if so, obtain the data of the preset ratio to obtain the target training data set; if not, repeatedly obtain the data of the structural performance corresponding to the preset ratio requirement that does not meet the preset ratio requirement until the preset ratio is met to obtain the target training data set.

[0138] The simulation module is used to perform physical simulation on the basic structure data to determine the mechanical properties of the structure; wherein the mechanical properties of the structure include ultimate strength properties and Young's modulus properties.

[0139] Figure 5 A structural diagram of a structural data processing device provided in another embodiment of the present application is shown in FIG. Figure 5 As shown, the structure data processing device includes: a memory 50 for storing computer programs;

[0140] The processor 51 is used to implement the steps of the method for processing structured data mentioned in the above embodiment when executing a computer program.

[0141] The structural data processing device provided in this embodiment may include but is not limited to a smart phone, a tablet computer, a laptop computer, or a desktop computer.

[0142] Among them, the processor 51 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 51 can be implemented in at least one hardware form of a digital signal processor (Digital Signal Processor, referred to as DSP), a field programmable gate array (Field-Programmable Gate Array, referred to as FPGA), and a programmable logic array (Programmable Logic Array, referred to as PLA). The processor 51 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a central processing unit (Central Processing Unit, referred to as CPU); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 51 may be integrated with a graphics processing unit (Graphics Processing Unit, referred to as GPU), and the GPU is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 51 may also include an artificial intelligence (Artificial Intelligence, referred to as AI) processor, which is used to process computing operations related to machine learning.

[0143] The memory 50 may include one or more computer-readable storage media, which may be non-transitory. The memory 50 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In this embodiment, the memory 50 is at least used to store the following computer program 501, wherein, after the computer program is loaded and executed by the processor 51, it can implement the relevant steps of the method for processing structural data disclosed in any of the aforementioned embodiments. In addition, the resources stored in the memory 50 may also include an operating system 502 and data 503, etc., and the storage method may be temporary storage or permanent storage. Among them, the operating system 502 may include Windows, Unix, Linux, etc. Data 503 may include but is not limited to relevant data involved in the method for processing structural data, etc.

[0144] In some embodiments, the structure data processing device may further include a display screen 52 , an input and output interface 53 , a communication interface 54 , a power supply 55 , and a communication bus 56 .

[0145] Those skilled in the art will understand that Figure 5 The structure shown in the figure does not constitute a limitation on the structure of the data processing device, and may include more or less components than those shown in the figure.

[0146] The structural data processing device provided in the embodiment of the present application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the structural data processing method in the above embodiment.

[0147] It should be noted that, although the operations are depicted in a specific order in the accompanying drawings, this should not be understood as requiring these operations to be performed in the specific order shown or to be performed sequentially, or requiring all illustrated operations to be performed to achieve the desired results. In some cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system modules and components in the above-described embodiments should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product, or packaged into multiple software products.

Claims

1. A method for processing structured data, characterized in that: A structure data processing device applied to a data processing system, the structure data processing device comprising a plurality of structure data processing devices, each of the structure data processing devices executing the structure data processing method in parallel, the method comprising: Obtain the original data of the non-spatial structure of the cell; The raw data is standardized to obtain a structural knowledge graph of a three-dimensional structure composed of the cells; wherein the structural knowledge graph includes structural performance and basic structural data in a spatial structure; Constructing a training data set for training the model to be trained according to the pre-constructed question-answering strategy and the structural knowledge graph; The raw data is subjected to standardization processing, including: Converting the original data into point-bar data formed by connecting the coordinates of every two nodes in the spatial structure; After merging the point rods in the point rod data that are located on the same straight line and have intersections, the basic structure data is obtained; The step of constructing a training data set for training the model to be trained according to the pre-constructed question-answering strategy and the structural knowledge graph includes: Based on the structural knowledge graph, generate a set of answers to the three-dimensional structure under different question-answering strategies; Counting the number of three-dimensional structures under different answer sets; Constructing a correspondence between the three-dimensional structure, the question-answering strategy, and the answer set; The training data set is obtained according to the number of the three-dimensional structures and the corresponding relationship.

2. The method for processing structured data according to claim 1, characterized in that: Merging the point bars in the point bar data that are located on the same straight line and have intersections, including: According to the node coordinates, the point rod data is converted into vector data in the same direction; Determine, based on the vector data, whether there is a point-rod data group in which the point-rods are parallel to each other and have intersections in the point-rod data; If so, the node coordinates in the point-rod data group are sorted according to the same direction; and based on the first two node coordinates in the sorting result, all the point-rod data in the point-rod data group are merged into a target vector in the same direction.

3. The method for processing structured data according to claim 1, characterized in that: The method further comprises: Mixing and acquiring data of different structural properties in a preset ratio from the training data set to obtain a target training data set; Training the model to be trained using the target training data set; If a fine-tuning instruction is received, the preset ratio is adjusted according to the fine-tuning instruction to optimize the model to be trained.

4. The method for processing structured data according to claim 3, characterized in that: The step of mixing and obtaining data of different structural properties in a preset ratio from the training data set to obtain a target training data set includes: Determining whether the data of each of the structural performances in the training data set meets the data requirement of the preset ratio; If satisfied, obtain the preset proportion of data to obtain the target training data set; If not, the data of the structural performance that does not meet the preset ratio requirement is repeatedly acquired until the preset ratio is met to obtain the target training data set.

5. The method for processing structured data according to claim 1, characterized in that: The structural performance includes at least one of volume fraction, structural topological performance and structural mechanical performance; when the structural performance includes structural mechanical performance, the raw data is standardized, including: Physical simulation is performed on the basic structure data to determine the structural mechanical properties; wherein the structural mechanical properties include ultimate strength properties and Young's modulus properties.

6. A structural data processing device, characterized in that: include: A raw data acquisition module, used to acquire raw data of the non-spatial structure of the cell; A standardization processing module, used for performing standardization processing on the original data to obtain a structural knowledge graph of the three-dimensional structure composed of the cells; wherein the structural knowledge graph includes structural performance and basic structural data in the spatial structure; A construction module, used to construct a training data set for training the model to be trained according to the pre-constructed question-answering strategy and the structural knowledge graph; A raw data conversion module, used to convert raw data into point-bar data formed by connecting the coordinates of every two nodes in the spatial structure; The point-rod data merging module is used to merge the point-rods that are located on the same straight line and have intersections in the point-rod data to obtain basic structure data; The answer set generation module is used to generate a set of answers with three-dimensional structures under different question-answering strategies based on the structural knowledge graph; The three-dimensional structure quantity statistics module is used to count the number of three-dimensional structures under different answer sets; The correspondence building module is used to build the correspondence between the three-dimensional structure, the question-answering strategy, and the answer set; The training data set acquisition module is used to obtain the training data set according to the number and corresponding relationship of the three-dimensional structures.

7. A structural data processing device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the program, the steps of the method for processing structured data according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for processing structured data described in any one of claims 1 to 5 are implemented.

Images