Knowledge graph customization method and application method oriented to industrial drawing knowledge

Abstract

The invention provides a knowledge graph customization method for industrial drawing knowledge and an application method. The knowledge graph customization method comprises the following steps: S1, collecting a plurality of drawing sample groups; each drawing sample group comprises two single drawing samples, namely an initial-format drawing and a corresponding standardized-format drawing; s2, marking entities, attribute values and relationships in the single drawing sample to form a single drawing knowledge graph; s3, identifying and aligning the same entity in all the single-drawing knowledge maps, and connecting all the single-drawing knowledge maps through the aligned same entity to form a special knowledge map; and S4, the special knowledge graph obtains the capability of converting the drawings in any format into the drawings in the standardized format by identifying and learning the processing process of converting the drawings in the initial format into the drawings in the standardized format in each drawing sample group. According to the special knowledge graph provided by the invention, data from different sources can be accurately aligned, and meanwhile, cross-scene expansion application can be realized through scene driving.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of knowledge graph, in particular to a knowledge graph customization method and application method for industrial drawing knowledge. BACKGROUND

[0002] As a structured knowledge representation form, knowledge graph has been widely used in many fields, including search engines, recommendation systems, intelligent question answering, etc. Conventional knowledge graphs are mostly extracted from a large amount of text data, and the extracted entity, relationship and event information are organized into structured knowledge representation. Although knowledge graph technology has made significant progress, the constructed knowledge graph.

[0003] In the process of industrial drawing knowledge, the format standards of drawings in different fields may have certain differences, different enterprises have different common names for the same object, and the adopted drawing formats are not uniform. Therefore, users often face problems such as format differences of drawings from different sources and semantic conflicts, which affect the alignment of key information in drawings and the application of cross-field drawing information. However, traditional knowledge graphs are less targeted at non-text data such as industrial drawings, and it is difficult to cover the complex standards of industrial drawings, which requires a large amount of manual rules or annotated data fine-tuning.

[0004] Therefore, in the process of industrial drawing knowledge, how to convert unstructured drawing information into a unified knowledge system through structured data to support component recognition and classification, design defect detection, multi-modal data fusion, intelligent review and deep information mining is one of the technical problems to be solved in the field.

[0005] In summary, the existing knowledge graph for industrial drawing knowledge at least has the following problems: 1. Poor alignment effect of data from different sources; 2. Difficult cross-scene application. SUMMARY

[0006] The present application aims to overcome the defects of the prior art and provides a knowledge graph customization method and application method for industrial drawing knowledge, which can solve the problems of poor alignment of knowledge graph for data from different sources and difficult cross-scene application in the prior art.

[0007] In order to achieve the above and other purposes, the present application is realized by including the following technical solutions: as a first aspect, the present application proposes a knowledge graph customization method for industrial drawing knowledge, including the steps of:

[0008] S1, collect a plurality of groups of drawing sample groups; each group of drawing sample groups includes two single drawing samples, which are initial format drawings and corresponding standardized format drawings, respectively;

[0009] S2, label the entities, attribute values and relationships in a single drawing sample to form a single drawing knowledge graph;

[0010] S3, identify the same entity in all single drawing knowledge graphs, connect all single drawing knowledge graphs through the same aligned entity to form a special knowledge graph;

[0011] S4, the special knowledge graph obtains the ability to convert any format drawing into a standardized format drawing by identifying and learning the processing process of converting the initial format drawing in each group of drawing sample groups into a standardized format drawing.

[0012] In an embodiment, in the step S1, the collected drawing sample groups are data cleaned to remove repeated, incorrect or incomplete data.

[0013] In an embodiment, in the step S2, the accuracy of the labeled data is verified by a combination of manual sampling and automatic verification tools.

[0014] In an embodiment, in the step S3, the same entity is an entity that expresses different but actually refers to the same object.

[0015] In an embodiment, whether the entities expressed differently are the same entity is determined by comparing the similarity of the attribute values and relationships of the entities expressed differently.

[0016] In an embodiment, the knowledge graph customization method further comprises step S5, periodically updating the drawing sample groups and industry standard changes, and expanding the special knowledge graph using knowledge reasoning technology.

[0017] As a second aspect, the present application proposes a knowledge graph application method for industrial drawing knowledge, characterized in that it comprises the steps of: adopting the knowledge graph customization method of the first aspect to construct the special knowledge graph; inputting industrial drawings, and the special knowledge graph converts the industrial drawings into a standardized drawing format output.

[0018] In an embodiment, the special knowledge graph labels the text, symbols and graphics in the industrial drawings by identifying the mapping relationship between the key information and structured data in the industrial drawings; and converts the labeling of the industrial drawings according to the character features and layout rules specific to the standardized drawing format.

[0019] In an embodiment, the knowledge graph application method further comprises the step of: inputting industrial drawings and searching for alternative processing technologies in the special knowledge graph according to cross-scene needs.

[0020] The one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0021] The present application constructs a professional knowledge graph by connecting all single drawing knowledge graphs through aligning the same entities in different single drawing knowledge graphs, can accurately align data from different sources, and uniformly convert drawing in any format into standardized format drawing. At the same time, the professional knowledge graph realizes the fusion and expansion between cross-scene knowledge graphs, and through scene driving, the professional knowledge graph can be flexibly applied in different scenes, and a balance point between individualization and safety control is found to ensure the safety and individualization of the professional knowledge graph for industrial drawing knowledge. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 A schematic diagram of a knowledge graph for industrial drawing knowledge is shown. DETAILED DESCRIPTION

[0023] The embodiments of the present application provide a knowledge graph customization method and application method for industrial drawing knowledge, which can solve the problems of poor alignment effect of knowledge graph on data from different sources and difficult cross-scene application in the prior art.

[0024] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application. In addition, it should be noted that only parts related to the present application are shown in the drawings for convenience of description.

[0025] The general idea of the technical solutions provided by the present application is as follows: a knowledge graph customization method for industrial drawing knowledge, comprising the steps of: S1, collecting a plurality of drawing sample groups; each of the drawing sample groups comprises two single drawing samples, namely an initial format drawing and a corresponding standardized format drawing; S2, labeling entities, attribute values and relationships in the single drawing sample to form a single drawing knowledge graph; S3, identifying and aligning the same entity in all single drawing knowledge graphs, connecting all single drawing knowledge graphs through the aligned same entity to form a special knowledge graph; S4, the special knowledge graph obtains the ability to convert any format drawing into a standardized format drawing by identifying and learning the processing process of converting the initial format drawing in each of the drawing sample groups into a standardized format drawing.

[0026] The main idea is to connect all single drawing knowledge graphs by aligning the same entities in different single drawing knowledge graphs to construct a professional knowledge graph, which can accurately align data from different sources and convert any format drawing into a standardized format drawing. At the same time, the professional knowledge graph realizes the fusion and expansion of cross-scene knowledge graphs, and through scene driving, the professional knowledge graph can be flexibly applied in different scenes, and a balance point between individualization and safety control is found to ensure the safety and individualization of the professional knowledge graph for industrial drawing knowledge.

[0027] After introducing the basic principles of the application, various non-limiting embodiments of the application will be specifically introduced below in conjunction with the drawings.

[0028] Example one:

[0029] The embodiment provides a knowledge graph customization method for industrial drawing knowledge, which can use an AI agent to customize a knowledge graph, and specifically includes the following steps:

[0030] S1, collect multiple groups of drawing sample sets for AI agent training and learning; each group of drawing sample sets includes two single drawing samples, which are initial format drawing and corresponding standardized format drawing;

[0031] Further, the collected drawing sample sets can be data cleaned to remove duplicate, incorrect or incomplete data to ensure the data quality of the drawing sample sets.

[0032] S2, the AI agent labels the entities, attribute values and relationships in the single drawing sample to form a single drawing knowledge graph;

[0033] Specifically, since the information in the drawing includes multi-modal data such as text, symbols and graphics, and the drawing may have non-structured data such as hand-drawing or hand-labeling, the AI agent can combine 2 or more of the following technical means: knowledge fusion, contrast learning, large model, joint reasoning and multi-language processing, to realize efficient construction and application of the knowledge graph. Please refer to Figure 1 In the single drawing knowledge graph, a rounded box can be used to represent an entity, which is a specific component or object, such as a nut, a bolt, a gear, an electric motor, etc.; a box is used to represent an attribute value, such as a size value, a material, a power value, etc.; an arrow is used to represent a relationship, such as an assembly relationship, a position relationship, a function relationship, etc.; when extracting entities, attribute values and relationships, the importance of entity and relationship connection should also be extracted.

[0034] Further, a data verification mechanism can be established, i.e. the accuracy of the labeled data is verified by a combination of manual sampling and automatic verification tools to ensure that the labeled data is accurate and correct.

[0035] S3, the AI agent identifies and aligns the same entity in all single drawing knowledge graphs, connects all single drawing knowledge graphs through the aligned same entity, and forms a special knowledge graph;

[0036] The same entity is an entity that is expressed differently but actually refers to the same object. Specifically, whether the entities expressed differently are the same entity can be determined by comparing the similarity of the attribute values and relationships of the entities expressed differently. The specific determination process is as follows: first, set the similarity or consistency requirement according to the use scenario; then compare the similarity of the attribute values and relationships of the entities expressed differently in different drawings; when the similarity meets the set requirement, it can be determined that the entities expressed differently in different drawings are the same entity, otherwise they are different entities.

[0037] For example, the entity screw used in a drawing is labeled as a bolt in some drawings, both of which have relationships such as diameter, thread number, length, shape, and the specific values of diameter, thread number, length, and shape are attribute values. In a scenario with low precision requirement, as long as the values of diameter and length are the same, it can be determined that the screw in a drawing and the bolt in some drawings are the same entity. However, in a scenario with high precision requirement, all relationships and attribute values need to be highly consistent before it can be determined that the screw in a drawing and the bolt in some drawings are the same entity.

[0038] Specifically, the precision requirement of the use scenario can be confirmed through the question and answer interaction content between the user and the AI agent. The information provided by the user can include the specific use scenario, the specific attribute values and relationships that need to be consistent, etc.

[0039] S4, the special knowledge graph obtains the ability to convert any format drawing into a standardized format drawing by identifying and learning the processing process of converting the initial format drawing in each group of drawing sample groups into a standardized format drawing;

[0040] Specifically, in the initial stage of model training of the special knowledge graph, the samples are the existing initial format drawings and the processed standardized format drawings: the model of the special knowledge graph first extracts key information such as component entities (e.g. part name), material attributes (e.g. steel type), labeling information (e.g. dimensional tolerance), and component relationship (e.g. assembly relationship) from the initial format drawings; then identifies modification content (e.g. size adjustment, material replacement) and new relationship (e.g. optimized assembly logic) from the processed standardized format drawings; finally, by learning the processing process between the two, such as how to extract key information and how to transform key information, the method of converting other format drawings into standardized format drawings is learned, which has specific character features and layout rules.

[0041] S5, periodically updating the drawing sample set and industry standard changes, extending the special knowledge graph using knowledge reasoning technology. By incorporating new design drawing data and industry standard changes, the timeliness and completeness of the dataset are maintained.

[0042] Embodiment two:

[0043] The embodiment provides an application method of a knowledge graph for industrial drawing knowledge, comprising the steps of: adopting the knowledge graph customization method of embodiment one to build the special knowledge graph; inputting an industrial drawing, and converting the industrial drawing into a standardized drawing format output by the special knowledge graph.

[0044] Specifically, the special knowledge graph labels the text, symbols and graphics in the industrial drawing by identifying the mapping relationship between the key information and structured data in the industrial drawing; and converts the labeling of the industrial drawing according to the character features and typesetting rules specific to the standardized drawing format.

[0045] Further, the special knowledge graph for industrial drawing knowledge covers professional knowledge such as materials, processes and equipment, and provides comprehensive technical support for industrial production. In order to improve the process design efficiency, the special knowledge graph can also be used for cross-scene expansion application to realize process design optimization. For example, in a certain scene, a user inputs an industrial drawing, which labels the material and the effect achieved. In the case of a specific material, the hardness and processing conditions and other key elements of the material are aligned through the professional knowledge graph, alternative processing processes are searched in the professional knowledge graph, and the processing cost and yield conditions are compared to select the optimal process for the user to refer to.

[0046] The construction process of the professional knowledge graph with cross-scene expansion application function is as follows:

[0047] The mapping rules of geometric features and functional attributes are established, the professional knowledge and data in different fields are sorted out, and a unified semantic representation is constructed; using natural language processing technology, the key elements (elements vary according to the scene) such as materials, processes and effects in the knowledge graph are extracted to obtain standardized entities (standardized entities can be flexibly adjusted for different scenes), the relationship and attribute value between the standardized entities and other entities are clarified; the differences between field terms and standards are learned, the same entities expressed in different ways are aligned, the knowledge fusion algorithm is adopted to integrate and correlate the knowledge in different fields; using knowledge reasoning technology, the knowledge graph is extended to generate new knowledge and relationships for user reference; using the effect evaluation method, the application effect of the knowledge graph is verified and optimized, the application scheme of the knowledge graph is iterated and optimized, and the application effect is improved.

[0048] In conclusion, the special knowledge graph provided by the application can accurately align data from different sources, and unify any format of drawing into a standardized format drawing. At the same time, the fusion and expansion between cross-scene knowledge graphs are realized, the flexible application of professional knowledge graph in different scenes can be realized through scene driving, and a balance point between individualization and safety control is found, so that the safety and individualization demand of the professional knowledge graph for industrial drawing knowledge are ensured.

[0049] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the same technology of the present application, the present application also intends to include these modifications and variations.

Claims

1. A knowledge graph customization method for industrial drawing knowledge, characterized by: Including steps: S1. Collect multiple sets of drawing sample groups; each set of drawing sample groups includes two single drawing samples, namely, an initial format drawing and a corresponding standardized format drawing; S2. Mark out the entities, attribute values, and relationships in a single drawing sample to form a single drawing knowledge graph; S3. Identify and align the same entity in all single drawing knowledge graphs, and connect all single drawing knowledge graphs through the aligned same entity to form a dedicated knowledge graph; S4. The dedicated knowledge graph obtains the ability to convert drawings in any format into standardized format drawings by identifying and learning the processing process of converting initial format drawings into standardized format drawings in each group of drawing sample groups.

2. The knowledge graph customization method according to claim 1, characterized in that: In step S1, data cleaning is performed on the collected drawing sample group to remove duplicate, erroneous or incomplete data.

3. The knowledge graph customization method according to claim 1, characterized in that: In step S2, the accuracy of the marked data is verified by combining manual sampling and automated verification tools.

4. The knowledge graph customization method according to claim 1, characterized in that: In step S3, the same entity is an entity that is expressed differently but actually refers to the same object.

5. The knowledge graph customization method according to claim 4, characterized in that: By comparing the attribute values ​​and relationship similarities of entities expressing different expressions, it is determined whether the entities expressing different expressions are the same entity.

6. The knowledge graph customization method according to claim 1, characterized in that: It also includes step S5, regularly updating the drawing sample group and industry standard changes, and using knowledge reasoning technology to expand the dedicated knowledge graph.

7. A knowledge graph application method for industrial drawing knowledge, characterized by: Including steps: Adopting the knowledge graph customization method according to any one of claims 1 to 6 to construct the dedicated knowledge graph; Industrial drawings are input, and the dedicated knowledge graph converts the industrial drawings into a standardized drawing format for output.

8. The knowledge graph application method according to claim 7, characterized in that: The dedicated knowledge graph identifies the mapping relationship between the key information and structured data in the industrial drawings, and marks the text, symbols and graphics in the industrial drawings; and converts the annotations of the industrial drawings according to the character features and typesetting rules unique to the standardized drawing format.

9. The knowledge graph application method according to claim 7, characterized in that: It also includes the steps of: inputting industrial drawings and searching for alternative processing techniques in the dedicated knowledge graph based on cross-scenario requirements.

Images